Revisiting the Petrogenesis of Pyroclastic Glass Bead Deposits at the Apollo 15 and 17 Sites

Abstract

The Apollo 15 low-titanium and Apollo 17 high-titanium pyroclastic glass beads are among the most primitive magmatically derived samples obtained from the Moon. Two key samples, the low-Ti Apollo 15426 green glass clod and the high-Ti Apollo 74220 orange glass are morphologically distinct, where the Apollo 15 beads are larger (~107 μm along maximum axis) and more fractured, and the Apollo 17 are smaller (~42 μm) and less fractured. In this study, holohyaline beads as well as crystallized beads were examined from both samples using petrography, electron microprobe analysis, and laser-ablation inductively coupled plasma mass spectrometry. Crystallized beads show compositional variability in major, minor, and trace elements and enable examination of magmatic mineral fractionation processes during cooling of both deposits. The Apollo 15426 beads experienced variable olivine crystallization, whereas the Apollo 74220 beads experienced both olivine and ilmenite crystallization. Holohyaline beads from both deposits show more limited major, minor, and trace element variability than their crystallized counterparts. Trace element abundance data for individual holohyaline beads show that in Apollo 74220, they are tightly clustered at ~30× Carbonaceous Ivuna chondrite [CI] with negative Eu anomalies and subchondritic Nb/Ta, interpreted to reflect the presence of late-stage magma ocean cumulates overturned into an otherwise primitive mantle source. Incompatible trace element abundances for holohyaline beads in 15426 are supra-chondritic from ~8× CI, to >80× CI, with pronounced relative depletions in Sr and Eu for the most incompatible element enriched beads, which represent a distinct bead group within the deposit. Apollo 15426 beads have elevated Ni and Co abundances at the edges of the beads compared to their centers. These data are interpreted to reflect a more complex magmatic evolution of the 15426 deposit, beginning with (i) initial magma generation, storage, and assimilation within shallower low- and high-Ca pyroxene bearing magma ocean cumulates (15B,C); (ii) mobilization of the earlier magmas by more recently generated primitive magmas (15A); (iii) eruption and crystallization of some beads (15D,E); and (iv) later jumbling of the deposit, possible impact contamination and addition of exotic basaltic bead components (J Group). In contrast, the 74220 data show no discernable difference between Ni and Co abundances at the edges and centers supporting prior observations for limited melt fractionation and an absence of meteoritic components. Both deposits are likely to have been formed in the presence of a transient atmosphere. Using 74220 melt compositions from this study, post-entrapment crystallization abundances range from 266 to 1130 μg/g for H₂O, 36 to 68 μg/g for F, 441 to 832 μg/g for S, and 0 to 2.31 μg/g for Cl, consistent with prior studies and suggesting up to ~0.1 wt % H₂O in the melt, with considerably less in the source. The role that late-stage magma ocean cumulates rich in ilmenite and high-Ca pyroxene might play in modifying this volatile element estimate, however, casts remaining doubt on the volatile element abundance and evolution of the primitive Moon.

INTRODUCTION

The Apollo 15425/6 low-Ti green and Apollo 74220 high-Ti orange pyroclastic glass beads are amongst the most primitive volcanically derived samples from the Moon (Delano, 1979, 1986) (Fig. 1). Pyroclastic bead deposits at both the Apollo 15 and 17 landing sites are thought to have formed through eruptive fire fountaining at the lunar surface (Heiken & McKay, 1978), resulting in broadly coherent bead deposits and predictable volatile distributions (e.g. Fogel & Rutherford, 1995; Renggli et al., 2017). Based on U-Pb isotope data, the ages of the deposits have been determined to be similar, at ~3.48 Ga for the 74220 beads, and ~3.41 Ga for the 15426 beads (Tatsumoto et al., 1973; Tera & Wasserburg, 1974). Compared to mare basalts, the Apollo 15 and 17 pyroclastic glasses are both considered to have formed from partial melts that originated at greater depths (Krawczynski & Grove, 2012; Barr & Grove, 2013), experienced significantly less fractional crystallization (Delano, 1986; Shearer & Papike, 1993; Papike et al., 1998), and originated from a generally more volatile-rich source (e.g. Saal et al., 2008; Hauri et al., 2011; McCubbin et al., 2023). Since the Apollo missions, over 100 pyroclastic glass deposits have been detected across the lunar surface from remote sensing (Hawke & Head III, 1980; Gaddis et al., 1985, 1998, 2003, 2011; Hawke et al., 1989; Gustafson et al., 2012), and while their total volumes are likely to be limited (Head, 1976; Head & Coffin, 1997), their distribution and pervasiveness indicate that they are important for understanding lunar interior processes.

Limitations remain in understanding the petrogenesis of both the low-Ti (e.g. 15426) and high-Ti (e.g. 74220) pyroclastic glasses. While their mafic compositions are important for determining possible models of lunar differentiation, ranging from whole-Moon melting and complete magma ocean differentiation, to only a shallow magma ocean (Shearer et al., 1989; Longhi, 1992; Snyder et al., 1992), the details of their petrogenesis remains incomplete. Much of the pioneering work on the composition of the 15426 and 74220 glasses was done using major element compositions by Delano (1979, 1986) who showed that both pyroclastic glass bead deposits underwent limited post-melt modification and were volcanic in origin. Furthermore, Delano (1986) recognized five compositional groups (15A, 15B, 15C, 15D, 15E) within the 15426 glass bead deposit. Subsequently, workers have produced both major, minor, and trace element abundance data on Apollo 15 and 17 glass beads to confirm and extend these groupings (e.g. Steele et al., 1992; Shearer & Papike, 1993).

The results of petrological experiments using Apollo 15426 glass compositions have indicated that the most primitive 15A glasses likely originated from depths up to ~460 km in the Moon (Elkins-Tanton et al., 2003). The 15A glasses are considered parental to the 15B and 15C glass groups, which experienced assimilation of shallow cumulates during ascent of deeper-derived melts through the lunar interior (Elkins-Tanton et al., 2003). Barr & Grove (2013) used a refined composition of the 15A glass and showed saturation of olivine + orthopyroxene on the liquidus at 1520 °C and 2.1 GPa (~420 km). From this, Barr & Grove (2013) proposed that the Apollo 15A green glasses are the result of melts of the deep lunar mantle unaffected by magma ocean processes that adiabatically rose, penetrated, infiltrated, and variably included materials from late-stage cumulates. Variable trace element abundances in the A15 glasses have been suggested to result from polybaric partial melting and the heterogeneity of overturned cumulates at the base of the magma ocean (Barr & Grove, 2013), although equilibrium melting of a generally homogeneous source region has also been suggested (Steele et al., 1992).

Experiments have also been conducted for high-Ti Apollo 74220 glass compositions. In those experiments, olivine and orthopyroxene are saturated at 2.5 GPa (~500 km) and 1530 °C with the multiple saturation point showing sensitivity to the intensive variable fO₂, which causes up to a 300-km shift in the estimated minimum depth of origin (Longhi, 1992; Hess, 1993; Krawczynski & Grove, 2012). Olivine-dominated cumulate sources that have been hybridized with late cumulate minerals, including ilmenite, as well as trapped residual lunar magma ocean liquids have been invoked to explain the major and trace element patterns of the 74220 glass (Hughes et al., 1989). The experimental constraints for both the Apollo 15 and 17 glasses, therefore, locate their initial partial melting as occurring at significant depth in the lunar interior (>400 km), often with addition of partial melts from late-stage magma ocean cumulates rich in ilmenite, high Ca-pyroxene and apatite (Apollo 17), or both low- and high-Ca pyroxene (Apollo 15).

An alternative viewpoint is that the pyroclastic glass beads are not representative of the lunar mantle (e.g. Stolper et al., 1974; Wood & Ryder, 1977; Albarede et al., 2013; Albarède et al., 2015). In such a scenario, the glass compositions would not be faithful recorders of partial melting due to interaction with materials during transit to the surface, and the depths of origin derived from experimentally determined multiple saturation pressures would be erroneous. Evidence in support of contamination of both deposits comes from the higher abundances, in chondritic-relative proportions, of highly siderophile elements (HSE) and ¹⁸⁷Os/¹⁸⁸Os on the exteriors (etchates) of several glass bead samples, including 15426 and 74220, relative to their interiors (Walker et al., 2004). This possibly suggests some assimilation of chondritic impactor-contaminated material on the outer portions of the beads, although whether this was added during or after eruption is unconstrained. Another outstanding issue is that the relationship of mineral phases associated with the 74220 beads are not well understood. In particular, olivine grains (>0.2 mm) coated with or within orange glass and containing volatile-rich melt inclusions have been reported (Hauri et al., 2011; Chen et al., 2015). Additionally, metal grains (Weitz et al., 1997) have been discovered within the deposit, but their relationship to fractional crystallization of the beads is not well understood.

To more fully understand the petrogenesis of the Apollo 15426 and 74220 glass beads, we report new major element data obtained by electron probe microanalysis, in addition to trace element data obtained by laser-ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) on beads from the 15426 and 74220 pyroclastic deposits. With these data, we not only examine prior constraints on 15426 and 74220 from major, minor, and trace element abundance data but also provide new observations, including recognition of a distinct bead group in 15426, and illustrate the importance of fractional crystallization acting within the different bead deposits for understanding their petrogenesis.

SAMPLES AND METHODS

Samples

The green glass clod sample, identified by sample numbers 15425, 15426, 15427, and 15365–15377, was discovered around Spur Crater on the Apennine Front of the Apollo 15 mission (e.g. Butler, 1971; Reid et al., 1972; Tatsumoto et al., 1987). Sample 15426 contains surface-correlated volatiles and is enriched in Zn, Cd, Br, Se, Te, Ge, In, Tl, Bi, Ag, and Sb (Ganapathy et al., 1973; Morgan & Wandless, 1984). Prior work has shown that major element compositions in 15426 are relatively restricted (SiO₂ = 43–48 wt %; TiO₂ = 0.26–0.43 wt %; MgO = 16–18 wt %) with distinct glass groupings or sub-types proposed (Delano, 1986). Reported chondrite-normalized rare earth element (REE) patterns for 15426 range from relatively flat to being slightly enriched in the heavy REE (Ma et al., 1981; Galbreath et al., 1990; Steele, 1992). We examined a suite of major, minor, and trace element data of 15426 using electron probe microanalysis (EPMA, n = 179 distinct analyses) and LA-ICP-MS (n = 113 distinct analyses) on a commissioned and specially prepared polished 1″ glass round 50-μm-thick section (15426 180) from NASA JSC.

The orange pyroclastic glass clod 74220 was collected at Shorty Crater during the Apollo 17 mission and was subsequently identified as forming part of a more widely distributed pyroclastic glass deposit (Heiken & McKay, 1974, 1977; Meyer et al., 1975). It has been suggested that the impactor responsible for the creation of Shorty Crater penetrated the basalt flow that overlay the pyroclastic deposit (Schmidt, 1990). The Apollo 17 soil forming 74220 is nearly pure glass, but only about one-third of the sample is unbroken spheres and ovoid beads, with the remainder made up of broken spheres and angular glass fragments (Heiken et al., 1974; Heiken & McKay, 1974). Sample 74220 is enriched in volatile elements compared to mare basalts, with the beads having coatings of volatile elements (Zn, S, Pb, and Cl) (Nunes et al., 1974; Meyer et al., 1975; Butler Jr. & Meyer, 1976). Reported bulk major element compositions of the 74220 glasses vary (SiO₂ = 38–55 wt %; TiO₂ = 3.6–13 wt %; MgO = 2–17 wt %; Delano, 1986; Chen et al., 2015), and the trace element signatures for the 74220 glass beads exhibit a flat to concave down chondrite-normalized incompatible element pattern with a negative Eu anomaly and a positive Nb anomaly (Chen et al., 2015). We collected a suite of major, minor, and trace element data for 74220 using the EPMA (n = 419) and using LA-ICP-MS (n = 119) on a commissioned and specially prepared polished 1″ glass round 50-μm-thick section (74220 701) from NASA JSC.

Microscopy and particle size distribution

The ~50-μm polished sections of 15426 180 and 74220 701 were examined using a transmitted/reflected light microscope (Figs 2 and 3). Most of the glass beads are exposed in the same plane, embedded within epoxy. Plane-polarized, cross-polarized, and reflected light images of these sections were made prior to performing in situ analyses. To determine the particle size distribution (PSD), reflected or transmitted light images were scaled and contrasted and colors were optimized to separate the beads from the surrounding epoxy. The scaled images were then analyzed using imageJ software. To assess the veracity of the PSD method, images of 74220, 701 at 1.5×, 5×, 10×, and 20× magnification were all measured to obtain a range of PSD that were found to agree within 5% uncertainties (e.g. Day et al., 2006). Modes were determined on olivine, ilmenite, and glass using imageJ software. Fragmented versus intact bead textures were determined using imageJ software, which allowed determination of the deviation from a sphere for each bead.

Electron probe microanalysis

Sample 15426 180 was analyzed for major and minor elements using a JEOL 8530F EPMA at NASA-JSC using a 5-μm spot size, an accelerating voltage of 15 keV, and a beam current of 15 nA. The standards used were both natural and synthetic minerals and glasses. Sample 74220 701 was analyzed for major and minor elements using a JEOL JXA 8200 EPMA at the Institute of Meteoritics, in the Department of Earth and Planetary Sciences at the University of New Mexico (UNM). Backscatter electron (BSE) images were obtained during quantitative analyses. Quantitative analyses were conducted using a 5-μm spot size, an accelerating voltage of 15 keV, and a beam current of 15 nA. The following standards were used for EPMA at UNM: quartz (Si), rutile (Ti), almandine (Al, Fe), chromite (Cr), spessartine (Mn), olivine (Mg), pyrope (Ca), albite (Na), orthoclase (K), Durango apatite (P), and pyrite (S). The following standards were used for EPMA at JSC: diopside (Ca and Si), rutile (Ti), apatite-wilberforce (P), rhodonite (Mn), orthoclase (K), albite (Na), Sitkin-Anorthite (Al), chromite (Cr), troilite (S), A99 glass (Fe), and MPI glass (Mg). The quality of the analyses was assessed using electron microprobe totals. Peak count times of 60 seconds and background count times of 30 seconds were used for Na, P, S, K, Ti, Cr, and Mn, while peak count times of 30 seconds and background count times of 15 seconds were used for Si, Mg, Al, Ca, and Fe. The new data are reported in full in Tables S1 and S2.

LA-ICP-MS

Thirty-five trace element abundances were determined using a New Wave UP-213 nm laser ablation system coupled to a ThermoScientific iCAP Q inductively coupled plasma mass spectrometer at the Scripps Isotope Geochemistry Laboratory using methods similar to those described in Kumler & Day (2021). Monitored masses were ⁷Li, ¹¹B, ⁴⁴Ca, ⁴⁵Sc, ⁴⁸Ti, ⁵¹V, ⁵²Cr, ⁵⁷Fe, ⁵⁹Co, ⁶⁰Ni, ⁸⁸Sr, ⁸⁹Y, ⁹⁰Zr, ⁹³Nb, ⁹⁵Mo, ¹³⁷Ba, ¹³⁹La, ¹⁴⁰Ce, ¹⁴¹Pr, ¹⁴⁶Nd, ¹⁴⁷Sm, ¹⁵³Eu, ¹⁵⁷Gd, ¹⁵⁹Tb, ¹⁶³Dy, ¹⁶⁵Ho, ¹⁶⁶Er, ¹⁶⁹Tm, ¹⁷²Yb, ¹⁷⁵Lu, ¹⁷⁸Hf, ¹⁸¹Ta, ¹⁸²W, ²³²Th, and ²³⁸U, with the focus being on cosmochemically ‘refractory’ elements rather than moderately volatile or volatile elements. Rasters and spot sizes of 50, 75, 100, and 150 μm were used depending on the size of the bead targeted with a laser repetition rate of 5 Hz and a fluence of ~4 Jcm⁻². Ablation took place in a 3-cm³ cell. The cell was flushed with He-gas and was mixed with Ar carrier-gas flow of ~1 L/min. Data collection lasted for 60 seconds for each analysis with 20 seconds of background and 40 seconds of laser ablation. The washout time was set for 120 seconds.

Each analysis was normalized to calcium measured by EPMA. Iron and Ti were both measured by EPMA and LA-ICP-MS and act as an independent check of data quality. Only larger beads within 15426 180 and 74220 701 were analyzed to accurately determine elemental abundances. Crystallized beads were measured, and edges and centers of the beads were also analyzed. These data are compiled in Tables S3 and S4, along with new ilmenite data for mare basalt samples 12063 and 15555 used for comparative purposes (Table S5; see below). Due to the ablation areas (50–150 μm) and associated volumes, it is likely that crystallized components in both glass samples reflect mixtures of materials (i.e. glass + olivine ± ilmenite). Although we report these data, we acknowledge significant variability in the reported abundances may be possible, although ratios of elements should be unaffected by corrections to an assumed Ca content.

Standard reference material (SRM) glass values of NIST 610, BHVO 2-g, BIR 1-g, and BCR 2-g were from a compiled database of preferred values from GEOREM (Jochum et al., 2005). Calibration versus only NIST 610 or by producing calibration curves versus all four glass reference materials were performed and were found to give nearly identical results. Consequently, reported data are corrected relative to NIST 610. The SRM show good accuracy and reproducibility, generally better than 10% (Tables S6–S9), over >2 years of analytical campaign. Reported data are above the detection limit, which were determined from the 3-sigma SD of the gas blank (Table S10).

To examine the role of ilmenite in the source of the lunar pyroclastic glass beads, we report ilmenite trace element abundance data from the Apollo 12063 51 and 15555 955 mare basalts for comparative purposes. For this work, it is assumed that ilmenite in these samples reproduce compositions of ilmenite formed at the end of magma ocean crystallization. A spot size of 100 μm was used and monitored masses were ²⁴Mg, ²⁹Si, ⁴⁸Ti, ⁵¹V, ⁵²Cr, ⁵⁵Mn, ⁵⁷Fe, ⁵⁹Co, ⁶⁰Ni, ⁶³Cu, ⁶⁶Zn, ⁸⁹Y, ⁹⁰Zr, ⁹³Nb, ¹⁷⁸Hf, ¹⁸¹Ta, and ²⁰⁸Pb, with the focus being on the high-field- strength elements (HSFE). Raw data were corrected relative to an assumed Fe content in ilmenite (351 400 μg/g or ~ 45 wt % FeO) and to BHVO-2g and BCR-2g. The reproducibility of the SRM NIST 610 was generally better than 10% for these analyses. These data are reported in Table S5.

RESULTS

Petrographic observations and PSD

The dominantly green-colored beads in 15426 (Fig. 2) are noticeably larger and fragmented when compared to the smaller, rounded, and orange-to-black-colored beads that compose 74220 (Fig. 3). The 74220 701 section is dominantly composed of glass beads. Section 15426 180 is made of beads, bead fragments, and unusually shaped bead fragments, most notably characterized by Bead J (Fig. 2d). Bead types recognized in both samples are those with no obvious crystalline features, even at the microscopic scale, which are termed holohyaline. Holohyaline beads are particularly notable in 74220 as they are always orange in color. Beads that contain crystals of olivine (15426) and olivine ± ilmenite (74220) have glass between the crystals, which is termed vitrophyric material. In the case of the 74220 sample, these crystallized beads are typically dark orange to black in color. Within the Apollo 74220 beads, several different types of olivine textures, including acicular, dendritic, and subequant, were identified. Within 15426, only acicular textures of olivine were observed. Under high magnification (10× or greater), crystalline materials appear as dark opaque grains in plane polarized light. Both polished sections studied are almost exclusively composed of holohyaline or crystallized beads, with the notable exception of minute (<10-μm) metal grains interstitial to the beads. For 15426 180, there is a ~1:3 ratio of holohyaline to crystallized beads, similar to that previously observed in 15427 (Arndt et al., 1984). For 74220 701, there is a ~1:1 ratio of holohyaline to crystallized beads, approximately consistent with other sections of 74220 (Arndt & von Engelhardt, 1987).

The PSD were determined for both 15426 180 and 74220 701 (Fig. 4). For 15426, the average bead size is 65 μm along the maximum axis, with an unfractured average bead size of 107 ± 5 μm. Sample 15426 contains some large beads, with sizes up to nearly 2 mm (Fig. 2). Approximately 55% of the beads in 15426 are spheroidal, and the beads and fragments generally have a higher aspect ratio (~1.7) compared with 74220. Determination of 74220 bead size at four different magnifications (1.5, 5, 10, and 20 ×), counting over 10 000 individual beads, gave an average intact bead size of 42 μm along the maximum axis, and this value is within the median bead sizes of 35–51 μm previously reported in Taylor et al. (2018). It should be noted that the beads in 74220, 701 are not always completely spherical with a maximum (long) to minimum (short) axis aspect ratio of ~1.5, with approximately 65% of the beads being unfractured and spheroidal. This fraction of intact beads is much higher than previously reported on larger fractions of the same bead deposit (Heiken et al., 1974; Heiken & McKay, 1974).

Major element abundances of the Apollo 15426 and 74220 glass beads

The major element chemistry of beads in 15426 and 74220 (Figs 5–7) are presented in Table 1 and Tables S1 and S2. For holohyaline bead data, edge and center compositions are reported for both sections. The term ‘edge’ is used to define the rims of beads as observed in thin-section, which depend on the size of the bead. The term ‘center’ refers to direct measurement of the observed center of the 2-D bead in a thin-section without obvious inclusion of bead edges. As much as 30% of the beads in both 15426 and 74220 have undergone some degree of crystallization of either olivine and/or ilmenite within vitrophyric material, and these crystalline materials are also reported.

Apollo 15426 glass

Apollo 15426 beads were classified as low-Ti compositions by Delano (1986). Analyses of holohyaline beads from this study have an average center bead composition of 0.38 ± 0.04 wt % TiO₂ (n = 65), consistent with a low-Ti description (Fig. 1). The average MgO content of the 15 426 holohyaline center beads is 17.41 ± 0.35 wt %, with Al₂O₃ values from 7.5–8.1 wt % and an average FeO content of 19.72 ± 1.22 wt % (Fig. 6). Analyses of the beads in the polished sections enabled identification of four of the five chemical groupings (no Group C was observed in this study) determined by Delano (1986), with most of the studied beads from this work belonging to group A or D (Fig. 5). No major differences in composition exist between the edge or center data for the holohyaline beads.

For crystallized beads within 15426, we find consistent modal percentages of 33–36.5% olivine to 63.5–67% vitrophyric material for individual beads (e.g. Fig. 3c, d, and f). The dataset for major and minor element abundances in the crystallized beads includes individual analyses of olivine and vitrophyric material. Crystallized beads exhibit a larger range in values of ~4.2–41.4 wt % MgO and up to 0.7 wt % TiO₂. Data for SiO₂ versus TiO₂ or CaO in the crystallized beads are positively correlated and negatively correlated for MgO (Fig. 6). Crystallized beads have Al₂O₃ values ranging from 0.1 to 22.5 wt %, and FeO contents range from 5.0 to 23.3 wt %; these highly variable major abundances relative to holohyaline are due to the presence or absence of olivine in the analyses, with variations reflecting beam overlap of phases (Fig. 6). A notable feature of the new data are the offsets of group B samples from the A group beads that make up most analyses. Another important observation is that the crystallized beads appear to originate from an A group initial composition, rather than from the C or B groups (Figs 5 and 6).

A distinct set of beads, most notably represented by Bead ‘J’ (Fig. 2d), stand out from the other 15426 bead analyses. These beads are referred to as the J Group to distinguish them clearly from previously reported 15426 bead types. Bead J is large (>1 × 1 mm) and has an unusual morphology (Fig. 2d), which was found to be a commonality to other beads in the J Group. These beads can contain partly crystallized materials, including very fine (<10 μm) plagioclase, within them. J Group glasses have 9–10 wt % MgO, which are lower than 15426 holohyaline beads. They also have Al₂O₃ values ranging from 10.3 to 10.7 wt % and unusually high TiO₂ values ranging from 4.1 to 4.7 wt % (Fig. 6). These J Group glasses have similarities to ‘yellow glasses’ noted in Wood & Ryder (1977) and to some high Ti glasses from 15426 measured previously (Fig. 1).

Apollo 74220 glass

The Apollo 74220 sample contains crystallized and holohyaline beads. The samples have been referred to as a high-Ti glass by Delano (1986), and the new analyses of holohyaline beads support this, with average TiO₂ contents of 9.05 ± 0.22 wt % (n = 186). The 74 220 holohyaline beads have an average of ~14.4 ± 0.7 wt % MgO, and an average FeO content of 22.2 ± 0.2 wt %. Both centers and edges of holohyaline beads give essentially identical average compositions and span a relatively limited range of major element compositions; no distinctive groupings are evident for the 74220 beads (Figs 5 and 7).

Crystallized bead materials in 74220 span a much larger range in major element abundances than the holohyaline centers and edges (Fig. 7). For example, the TiO₂ contents within the crystallized beads range from 0.8 to 58.1 wt %, with FeO from 15.5 to 41.3 wt %, and MgO from 3.3 to 37.3 wt %. These large variations reflect analysis of mineralogical (olivine, ilmenite [high Ti, high Fe]) and vitrophyric (low Ti, low Fe) constituents that are remnants from olivine and ilmenite crystallization within the crystallized beads. Beam overlap with these phases during analysis explains the wide variation in compositions observed. Mineralogical variability is supported by the variable modal abundances of phases within 74220 crystallized beads. For example, modal percentages of phases in three different beads examined (e.g. Fig. 3d and f) gave (1) 71.1% olivine: 1.2%, ilmenite: 27.7% vitrophyric bead material; (2) 55% olivine, 5% ilmenite, 40% vitrophyric bead material; and (3) 43% olivine, 2% ilmenite, 55% vitrophyric bead material.

Minor and trace element abundances

We report minor and trace element abundance data that were determined on polished sections of 74220 and 15426 using LA-ICP-MS. These data are summarized in Table 1 and can be found in their entirety in Tables S3 and S4. We also report some major element data determined by LA-ICP-MS. The ablation pits for LA-ICP-MS are significantly larger than the beam size used during EPMA such that heterogeneities in glasses (for example, Bead J) will be accentuated.

Apollo 15426 minor and trace element abundances

The 15426 holohyaline center and edge compositions span a relatively wide range of compatible (e.g. Ni = 154 ± 4 μg/g [Center] to 205 ± 10 μg/g [Edge]), large ion lithophile (e.g. Sr = 27 ± 0.5 μg/g [Center] to 32 ± 2 μg/g [Edge]) and REE contents (e.g. La = 1.3 ± 0.1 μg/g [Center] to 1.8 ± 0.2 μg/g [Edge]) (Table 1), with crystallized beads spanning an even larger range of compositions (Fig. 8). The Apollo 15426 holohyaline bead centers and edges generally exhibit a relatively flat CI-chondrite-normalized REE pattern with slight negative Eu anomalies (Fig. 8). The incompatible trace element patterns of 15426 holohyaline beads are suprachondritic with more absolute variation in abundances than 74220 (Fig. 9). Incompatible trace element patterns of 15426 from this study are similar to those of others (Ma et al., 1981; Korotev, 1987; Steele et al., 1992); however, previous studies have suggested incompatible trace elements extend to lower abundances for 15426. Holohyaline center and edge data give ratios of La/Yb (1.3 ± 0.5 [Center] to 1.3 ± 0.2 [Edge]), Nb/Ta (21 ± 3 [Center] to 21 ± 4 [Edge]), and Eu* (0.8 ± 0.15 [Center] to 0.7 ± 0.1 [Edge]) that are reasonably restricted (Fig. 10). In a Zr/Hf versus Nb/Ta plot, 15 426 holohyaline centers and edges plot around the chondritic values and at the uppermost end of the lunar trend identified by Münker et al. (2003) (Fig. 11).

The J Group beads are highly distinct in the incompatible trace element compositions to the other glass groups in 15426. J Group beads have relatively lower Ni (53 ± 16 μg/g), higher Sr (124 ± 22 μg/g) and La (18 ± 3 μg/g) and higher La/Yb (2.6 ± 0.2), and lower Nb/Ta (18.6 ± 3) and Eu* (0.4 ± 0.02 μg/g) than the other groups (Fig. 10). J Group beads also have high distinct REE and incompatible element abundance profiles (Figs 8 and 9), with similarities to potassium (K), REE, phosphorous (P)-rich components (KREEP)-rich impactor materials such as, for example, impact melt breccias from the Apollo 15 (Ryder & Spudis, 1987; Korotev, 2000) or Apollo 17 landing sites (e.g. Norman et al., 2002).

Crystallized beads in 15426 have a greater range in trace element variability compared to the holohyaline beads from the same deposit. Due to these large variations and uncertainties in trace element abundances, we do not plot them apart from in Figs 6 or 7. In Fig. 10, the 15426 crystallized beads generally extend to lower Ni for a given FeO content, but otherwise have similar albeit more variable Nb/Ta, La/Yb and Eu* to the holohyaline beads.

Apollo 74220 minor and trace element abundances

The 74220 holohyaline center and edge compositions span a restricted range of compatible (e.g. Ni = 55 ± 6 μg/g; all uncertainties at 1 SD), large ion lithophile (e.g. Sr = 205 ± 2 μg/g [Center] to 212 ± 5 μg/g [Edge]) and REE contents (e.g. La = 5.8 ± 0.1 μg/g [Center] to 6.2 ± 0.2 μg/g [Edge]) contents (Table 1). Crystallized beads span a larger range of compositions, although these are not plotted on CI-chondrite-normalized diagrams due to the possible variability in absolute concentrations due to under- or over-correction to Ca for the ablation volumes (see Samples and methods).

The Apollo 74220 CI-chondrite-normalized REE pattern displays a concave down ‘gull wing’ pattern expressed by lower light REE and heavy REE relative to middle REE abundances, and a negative Eu anomaly at ~55 × CI chondrite (Fig. 8). The average holohyaline heavy REE abundance is slightly lower than the average light REE abundance. These data agree with previous data presented by (Philpotts et al., 1974). The CI-chondrite-normalized REE and incompatible trace element patterns for 74220 are suprachondritic at between ~10 and 100× CI (Figs 8 and 9). Trace element abundance data from this study are broadly similar to previous studies (Shearer & Papike, 1993; Chen et al., 2015). Most incompatible element abundances for individual beads, bead edges, and bead centers show that they are tightly clustered, with the notable exceptions of Th and U. The CI-chondrite-normalized incompatible trace element pattern shows distinctive positive Ba, Nb, Ta, Hf, and Ti anomalies (Fig. 9). Ratios of La/Yb (1.5 ± 0.1 [Center] to 1.7 ± 0.3 [Edge]), Nb/Ta (15 ± 1.4 [Center] to 15.1 ± 2 [Edge]), and Eu* (0.8 ± 0.04 [Center] to 0.8 ± 0.06 [Edge]) are all restricted (Fig. 10). In a Zr/Hf versus Nb/Ta plot, 74 220 holohyaline centers and edges plot below the chondritic values at the lower end of the lunar trend identified by Münkeret al. (2003) (Fig. 11).

Holohyaline bead edges and centers were analyzed to examine if geochemical differences exist between them. There is no appreciable difference between edge and center Co analyses (Edge versus Center for Co = 61 ± 4 μg/g versus 60 ± 1 μg/g, respectively), nor between edge and center Ni analyses (Fig. 10). Overall, the REE show no major differences between the edges and the centers in the 74220 holohyaline beads.

Crystallized beads in 74220 have a greater range in trace element variability compared to the holohyaline beads. Cobalt abundances in crystallized beads range from 0.6 to 180 μg/g, and Ni abundances range from 1.1 to 259 μg/g. Crystallized beads span a greater range in heavy and light REE than holohyaline beads. For example, abundances of La range from 1.4 to 7.6 μg/g, and Yb abundances range from 0.3 to 5.5 μg/g. As noted above, due to the large variations and uncertainties in trace element abundances for the 74220 crystallized beads, we do not plot them apart from in Fig. 6 and 7.

DISCUSSION

Limitations of the in situ measurement approach

At least three limitations are recognized with the approach of this study. First, the analyzed beads are within polished sections, and due to this method of preparation, it is not known where each bead was exactly sectioned. Samples were requested without prejudice, so sampling of the beads from the sample clods to make the section is also unknown. Since we do not know where each bead was sectioned, it is possible that some of our analyses over-represent the compositions of bead margins and underrepresent bead centers. If the beads were sliced close to their margins rather than directly through the center, analyses would preferentially sample the edges of the bead, as noted previously (Ganapathy et al., 1973; Nunes et al., 1974; Meyer et al., 1975; Butler Jr. & Meyer, 1976; Morgan & Wandless, 1984). This issue was mostly obviated during LA-ICP-MS by the observation of penetration of the laser beam through particularly thin ovoid slices and the selection of the largest beads, decreasing the chance of measuring slices away from bead centers.

Second, addition of epoxy during the making of the thin section and grain sorting and/or modification of the thin section has the potential to affect the PSD. In this case, there might be a depletion in the smallest size fraction within the thin sections (e.g. left-hand most regions of Fig. 4).

A third limitation of the method is that of the beam size of the in situ methods compared to the size of the bead. Both 15426 and 74220 have beads larger than 50 microns in grain size, but in the case of 74220, such beads are limited in number, and this was particularly exacerbated by the smaller average bead size due to fragmented and broken beads in the polished sections. For EPMA, beam sizes were consistently smaller than the bead size, but for LA-ICP-MS, the beam size used ranged between 50 and 150 microns. In the case of EPMA analyses, there was clearly overlap between phases within crystallized beads. For LA-ICP-MS sessions, larger beads were chosen for analysis in both samples meaning that the smaller bead fractions were not sampled as comprehensively. This size-filtering may explain the absence of group 15C in the 15426 glasses from this study, which may be in the smaller grain-size range for the sample that we studied. For 74220, the remarkable consistency in compositions for holohyaline edges and centers might suggest no significant variations, even within the finer fractions. In some cases, the beam size also introduced challenges since it was larger or overlapped with the beads analyzed. As stated above, many of the beads, particularly in 15426, are fragmented. In the beads that are fragmented, it cannot be stated with certainty where the center or the edges of the beads were measured, since the bead was not fully intact. Throughout the remainder of this work, these issues are outlined, where relevant.

Physical characteristics of pyroclastic beads and implications for volcanic processes

The bead size and morphology vary between the 15426 and 74220 beads, with the PSD calculated in this study following bulk sieved sample measurements reasonably well (Fig. 4). Some differences are notable, however, likely from creating the sections and the techniques used to determine the PSD (Limitations of the in situ measurement approach). Our PSD data for the Apollo 74220 beads suggests a slightly smaller average grain size compared to that of the PSD data by Taylor et al. (2018). Those authors used laser particle size analysis and wet sieving, while our study used imageJ analysis of back-scatter electron images on 2-D surfaces. Both techniques have drawbacks. For example, sieving <20-μm particles is not reproducibly accurate due to the tendency of smaller sieve openings to clog. Laser diffraction methods also depend on particle shape and could distort the PSD of irregular particles (Taylor et al., 2018). Our samples are not markedly different from previous measurements and can be deemed representative. The pristine Apollo samples are stored and processed in the Apollo curation lab at NASA JSC under an inert atmosphere of dry (<10 μL/L H₂O) high-purity gaseous nitrogen, conditions designed to limit sample degradation over multi-decade or longer timescales (McCubbin et al., 2019). From our analysis, our results support that of Taylor et al. (2018) that lunar samples 74220 and 15426 have not degraded significantly over time, in contrast to earlier reports (Cooper et al., 2015).

Both samples 74001 and 74220 were collected in close proximity at Station 4 of the Apollo 17 landing site. It was determined previously that at the bottom (~80-cm depth) of the 74001/2 core (which contains orange and crystallized beads) the ratio of crystallized beads to holohyaline components was higher than near the top of the core, where holohyaline beads were more dominant (Heiken & McKay, 1978). The Apollo 74220 beads exhibit a range of compositions, degree of crystallization, textures, and other physical characteristics, consistent with previous work (Arndt & von Engelhardt, 1987). Within the Apollo 74220 beads, several different types of olivine textures, including acicular, dendritic, and subequant, can be identified. These olivine textures were identified within the 74001/2 drill core where Weitz et al. (1999) considered the following five types of beads, which exist along a continuum: (1) orange holohyaline beads; (2) partially crystallized beads; (3) strongly crystallized beads; (4) completely crystallized beads; and (5) brown beads. Our study did not identify brown beads, but we observed a spectrum within section 74220 701 from orange holohyaline beads to strongly or completely crystallized beads. The crystallized beads likely underwent prolonged cooling compared with holohyaline beads, most evident in almost completely crystallized fragments (e.g. Fig. 3c).

It has been suggested that once the cooling rate of the beads falls below 100 °C/s, ilmenite and olivine crystallization can occur (Arndt & von Engelhardt, 1987), meaning that the crystallized brown and black beads cooled more slowly than the orange beads. Weitz et al. (1999) attributed this difference in cooling rate to the location within the plume, where beads closer to the vent cooled more slowly where there was a higher optical density. Alternatively, it has been proposed that the beads did not form in a vacuum (Hui et al., 2018), consistent with independent evidence that the Moon may have had transient atmospheres during its formation (Needham & Kring, 2017; van Kooten et al., 2020). It is also possible that thermal blanketing by overlying deposits further played a role in modifying cooling rates resulting in crystallized and non-crystallized beads that were jumbled by later impacts. In this sense, 74220 seems to be a mixture of deposit materials, akin to 74001/2.

The Apollo 15426 beads are generally larger than those of 74220. Furthermore, the crystallized beads in section 74220 701 have variable ratios of olivine, ilmenite, and vitrophyric components, whereas section 15426 180 has a consistent olivine to vitrophyric component ratio in crystalline beads and, overall, is dominantly holohyaline. The olivine within 15426 are acicular, resembling a quench texture or hopper-type morphology for some olivine. Arndt et al. (1984) reported that the free flight cooling rates of crystal-free and crystalline beads would be 1500 °C/s (for beads with a diameter of 0.22 mm) to 4200 °C/s (for beads with a diameter of 0.094 mm), which are much faster than their estimated critical cooling rate of 1 °C/s for the actual formation of 15426 green glass beads. This critical cooling rate of 1 °C/s is consistent with eruption into a hot gaseous medium (Arndt et al., 1984). While these authors considered this to be gas ejected with the molten droplets, we suggest that a transient lunar atmosphere was present during the eruption of both the Apollo 15426 and 74220 glass bead deposits, in line with other observations (Needham & Kring, 2017; Hui et al., 2018; van Kooten et al., 2021).

As noted previously, the Group D and E glasses come from crystal fractionation of a Group A composition, suggesting that the Group A composition is the dominant composition of the deposit. A particular puzzle for the 15426 glass deposit is how the Group B and C glasses were preserved without mixing if the molten beads were all ejected together. This might suggest either (1) that mixing between the magmas responsible for the glasses occurred at a very late stage, just prior to eruption to the lunar surface or (2) the glass types were mechanically mixed after the formation of the deposit, perhaps by impact disruption of Spur Crater. This would also be consistent with the higher amount of fragmentation in 15426 compared with 74220, due to regolith or bead deposit reworking. In the latter case, this would suggest either multiple deposits, or variability in eruptive composition with time, akin to recent time-series observations of terrestrial basaltic volcanic eruptions (e.g. Bindeman et al., 2022; Day et al., 2022).

J Group bead characteristics in 15426

A notable feature of the presented data is identification of a distinct bead grouping within the 15426 green glass clods, distinct from any other holohyaline and crystallized beads in the sample. The J Group beads (named after Bead J in 15426 180) are characterized by non-spheroidal shapes (e.g. Fig. 2d), anomalously high TiO₂ (>4 wt %; Fig. 6a), REE (Fig. 8a) and incompatible trace element abundances (Fig. 9a), but relatively low Ni contents relative to the other 15426 glass beads. Similar compositions of material have been previously noted for 15426, and were identified as ‘yellow glasses’ (Wood & Ryder, 1977). The yellow beads, which likely equate to the Bead J group of this study, were estimated to make ~1% of the 15426 deposit (Wood & Ryder, 1977), consistent with the modal percentages of J Group beads from the polished section examined in this study.

Four possibilities are considered to explain these anomalous materials. The first of which is that they represent agglutinate material picked up by magmas responsible for 15426 and that became entrained within the deposit. This explanation is not favored due to the absence of obvious regolith components (e.g. gas bubbles, metal) or, indeed, the absence of such high TiO₂ material at the Apollo 15 site. An alternative explanation is that the J Group represents a distinct batch of magma. This scenario would occur if the 15426 parental melt stalled in the lunar crust, underwent fractional crystallization, and was subsequently reinvigorated by a new batch of more primitive magma. An issue with this scenario is the extensive degree of fractional crystallization that would be required to produce the REE and incompatible element enrichment in the J Group, relative to the other 15426 beads, and the lower SiO₂ contents for a given MgO or CaO content (Fig. 6). This means the J Group beads are not extreme fractional crystallization products. While it is possible to potentially explain their composition by >90% olivine and plagioclase (95:5 distribution) fractionation, this would lead to a melt enriched in SiO₂ and depleted in MgO, which is not the case.

Of the final two possibilities, the first is that the J Group beads are from intermediate-Ti basaltic volcanic fire fountaining, perhaps akin to compositions observed at the Apollo 12 site. In this circumstance, they would have a significant KREEP component to explain the elevated REE and incompatible trace element components of these beads. Such an origin would be consistent with the relatively low siderophile contents of the J Group beads, at ~37 μg/g Co and ~53 μg/g Ni (Table 1), broadly within the range of endogenous lunar mare basalts (Day, 2020).

The other plausible alternative is that the Bead J group represent relatively low-Ni basaltic impact melts. In particular, the J Group glasses resemble some impact melt breccias with KREEP (e.g. Norman et al., 2002; McIntosh et al., 2020; Fig. 9a) and are similar to basaltic impact glasses from both soil of Station 9a at the Apollo 15 site (Korotev, 1987) and the Apollo 14 soil, 14 163 (Nemchin et al., 2022) (Fig. 12). However, it is also notable that of the >15 types of KREEP impact melt breccias, at the Apollo 12, 14, 15, 16, and 17 sites, few have basaltic compositions akin to the J Group glasses (e.g. Korotev, 2000; Korotev et al., 2011). The compositions of the J Group glasses could be explained by admixing of a Station 9a-type soil component (Korotev, 1987) and KREEP basalt or KREEP-rich impact melt, similar to mixing scenarios for Apollo 12 soils (Korotev et al., 2011). In particular, the J Group glasses have relatively low Ni and Co, elevated incompatible trace element abundances, and similarly intermediate Ti and compositions to 14163 basaltic glasses. Notwithstanding, if the J Group glasses represented exogenous components within the 15426 deposit, then they would have implications for measurement of Os isotopes and HSE abundances within it. Walker et al. (2004) found that most of the HSE complement was in the etchate (>90%) for the 80–200-μm size fraction with a nearly flat chondrite-relative HSE pattern. If the J Group beads were formed in impacts and have similar abundances of these elements to known lunar impact melts, then the presence of 1% of the J Group could lead to them contributing >10% of the HSE fraction (e.g. Norman et al., 2002; McIntosh et al., 2020). In this circumstance, the presence of a small impact glass contribution within the 15426 deposit would explain its higher general HSE contents when compared with the 74220 deposit.

In what we consider the most likely scenario of a basaltic volcanic origin for the J Group, assuming that the 15426 deposit was erupted and emplaced at ~3.41 Ga, this would suggest incorporation of this KREEP-rich material into the bead deposit either during or after this time. Reworking of the deposit could have occurred during disturbance of the green glass clods from various impact events that occurred at the site. We therefore suggest that the Bead J (yellow beads) are from an intermediate Ti, KREEP source from within the Procellarum KREEP Terrane. They were either delivered from locations distal from the Apollo 15 site or represent an unrecognized intermediate-Ti basalt component from the Hadley–Appenine front that was then effectively mixed into the deposit during or after its emplacement.

Origin of Ni and Co variations between the Apollo 15 and 17 deposits

In their analysis of trace elements within Apollo 15 green glass particles, Steele et al. (1992) noted a positive correlation between incompatible trace elements, like Sm and Co and Ni. They explained this behavior due to the possibility of a low oxidation state resulting in mineral–melt partition coefficients less than unity for Co and Ni. The new correlated LA-ICP-MS dataset for Sm and Co are presented in Fig. 13a for both 74220 and 15426 and solely for 15426 in Fig. 13b, along with fields of data from Steele et al. (1992). Systematic correlations like those presented by Steele et al. (1992) are not as clearly observed in the new 15426 dataset. Center and raster data for 15426 show a wide range of Co for a reasonably restricted range of Sm, and the edge data trend to higher Sm and Co concentrations (Fig. 13b). The J Group beads have low Co and high Sm contents, higher even than the 74220 beads (Fig. 13a). These data do not support the concept that Co or Ni behave as incompatible elements in either of the bead deposits. Instead, Co and Ni behave as compatible elements.

The Apollo 15426 and Apollo 74220 holohyaline beads exhibit distinct Ni and Co abundances from one another, with the highest Ni and Co contents generally conforming to the bead centers of 74220, with the lowest contents in crystalline beads (Fig. 14). For 15426, Ni and Co contents appear to be highest at the edges of some beads (Fig. 14). The Apollo 15426 data also have a steeper slope (Ni/Co = 1:2 to 1:3) compared to that of the Apollo 74220 data (Ni/Co = 1:1). The defined slopes of both deposits intersect at ~55 μg/g Co and ~55 μg/g Ni, which is similar to the estimated Ni and Co contents of mare basalt source regions (Day, 2020). We suggest that the different trends in Ni and Co for 15426 and 74220 are likely to be accounted for not only by mineral fractionation processes both within their mantle sources but also during the observed fractionations of some beads during crystallization at the surface.

To examine Ni/Co trends in the Apollo 15426 and 74220 glasses, modeling of Ni and Co behavior during partial melting and fractional crystallization was undertaken (Fig. 14). A similar method was used in Day (2020), although no sulfide or metal was assumed in the source or during fractional crystallization in this study. For the 15426 glasses, a source mineralogy of 70% olivine, 20% low-Ca pyroxene (orthopyroxene), and 10% high-Ca pyroxene (clinopyroxene) was assumed, while for the 74220 glasses, we assumed a 50:50 mixture of olivine and ilmenite. Although these endmembers are likely to be extreme, they reflect (1) possible assimilation of shallow low- and high-Ca bearing magma ocean cumulates, for the 15426 glasses (e.g. 15B and 15C) and (2) the presence of residual ilmenite in the source of the 74220 glasses. For partition coefficients, we used a preferred D_NiOlivine of 15.5 (after Mysen, 1978), a preferred D_CoOlivine of 3 (after Klemme et al., 2006) and for low-Ca pyroxene (preferred D_NiLowCa = 5; D_CoLowCa = 2; Irving & Frey, 1978]) and clinopyroxene (preferred D_NiHighCa = 2.6 [Mysen, 1978]; D_CoHighCa = 1.3; [Beattie, 1994]). Cobalt and Ni are also likely to be relatively compatible in ilmenite and spinel (D_NiOpaques = 3.8–6.2; D_CoOpaques = 1.4–3.3; Higuchi & Nagasawa, 1969; Ewart & Griffin, 1994), and we used the highest reported D values for both. Modeling results, shown in Fig. 14, illustrate that partial melt generation and interaction of melts with magma ocean cumulates either rich in low- and high-Ca pyroxene (for 15426), or in olivine and ilmenite (for 74220) can explain the different Ni/Co ratios of the two deposits. Later fractionation and crystallization within the deposits can explain the low Ni and Co contents of some later glasses (e.g. the 15 D and E group glasses). This suggests that a major control on the siderophile element geochemistry of both deposits are the sources and later magma ocean cumulates that primary melts came into contact with. The role of sulfide or metal fractionation is less clear. Metal blebs with the 74220 beads examined by Weitz et al. (1997) have a high Ni/Co ratio compared to the edge and center ratios of the pure holohyaline analyses for 74220, possibly suggesting late-stage fractionation of metals during crystal–liquid fractionation.

Another important observation from the Ni and Co data is that the edges of the Apollo 15426 beads have elevated Ni and Co abundances compared to their centers, while the Apollo 17 holohyaline data do not exhibit such differences (Fig. 14). Considered were several possibilities to explain the elevated Ni and Co abundances on the edges of the 15426 beads compared to the centers. Analytical issues can be ruled out because analytical sessions over a prolonged period showed identical results despite modifying run conditions to examine possible artefacts (Tables S3 and S4). The process that resulted in elevated Ni and Co abundances on the exterior of the 15426 beads must then have had to occur either syn- or post- eruption to explain the center–edge differences. A possibility is that high Ni and Co were incorporated during eruption onto the exterior of the beads by regolith that has experienced meteoritic contamination. Elevated Ni and Co abundances on the edges of the 15426 beads, along with data from Walker et al. (2004) that showed that the etchate of 15426 sample contains high HSE concentrations in chondritic relative abundances, might support the possibility of assimilation of crustal materials that have experienced meteoritic contamination. However, modeling regolith assimilation based on Ni and Co abundances is not possible for the Apollo 15426 sample because Apollo 15 regolith and impact melt breccias do not extend to the elevated Co abundances of the Apollo 15426 beads (e.g. Chapter 8 of Heiken et al., 1991). Further work is therefore required to explain the distinction between bead edges and centers for Co and Ni in sample 15426.

Cause and timing of magma ocean cumulate contributions to 15426 melts

As noted above, investigation of the petrogenesis of the Apollo 15426 green glass clod has shown distinct groupings, arguing that simple fractional crystallization or fractional melting models were inconsistent with olivine fractionation, particularly based on NiO abundances in the glasses (Delano, 1986; Elkins-Tanton et al., 2003). Furthermore, on a fayalite–forsterite–silica ternary diagram, some bead subgroups cannot be explained based on melting trends since they indicate a source with olivine magnesian numbers too low for the lunar mantle (Stolper et al., 1974; Grove, 1981; Elkins-Tanton et al., 2003). The results from this study show that the crystallized beads appear to form dominantly from a Group 15A glass composition (e.g. Fig. 5). In contrast, the 15C and B glass compositions have higher SiO₂ for a given MgO, CaO, or FeO content than the 15A glass compositions (Fig. 6), with the 15 426 holohyaline glasses exhibiting an order of magnitude range in their incompatible element patterns (e.g. Delano, 1979, 1986; Steele et al., 1992, see also Table S11, for a summary of group 15A to 15D compositions). These lines of evidence all suggest significant variability and fractionation in trace elements.

Experimental studies have proposed that the petrogenesis of 15426 beads depends on their subgroup. In the model by Elkins-Tanton et al. (2003) the parental melts for Group A may be explained by congruent melting over a small pressure range at ~440-km average depth, and the Group B and C glasses are best explained by Group A melts that were hybridized with magmas derived from melting of shallower (possible ~120-km depth) magma ocean cumulate sources. Subsequently Barr & Grove (2013) proposed a model where partial melting of undifferentiated lunar mantle produced melts that infiltrated and selectively melted late-stage magma ocean cumulates. Our results are broadly consistent with these concepts, but largely indicate a predominance of 15A type compositions in the sample that was studied.

Although the range of incompatible trace element abundances can be potentially explained by a deep primitive magma (15A group) interacting with shallower magma ocean cumulates, the physical process by which this would occur remains enigmatic because these putative mantle-derived partial melts would need to remain separate, without mixing, until both were close to or at the surface during eruption. One possibility is stalling of a contaminated magma (e.g. 15B or 15C) at the lunar crust–mantle boundary and invigoration by a deeper melt (e.g. 15A composition) driving it to the surface. This scenario would be similar to what physically happens to basaltic magmas on Earth, where early magmas stalled, followed by reinvigoration by deeper, hotter magma pulses that results in volcanic eruption (e.g. Day et al., 2022). We suggest that several pulses of magma generation may have occurred to form the 15426 glass bead deposit. In the first instance, a deeply derived magma was produced and transited into overlying magma ocean cumulates, where it stalled and assimilated shallow magma ocean cumulate materials (Fig. 15, i), producing the 15B and 15C glass compositions. Subsequently, another pulse of deeply derived uncontaminated magma reinvigorated the stalled melts and resulted in eruption at the lunar surface (Fig. 15, ii) where the beads were erupted into transient atmospheric conditions. Some 15A magmas were crystallized or partially crystallized, resulting in the 15D and 15E compositions. Reworking of the deposit likely led to jumbling of the various glasses and no preservation of eruptive stratigraphy, while simultaneously incorporating the J Group (yellow) glasses.

Evidence for magma ocean cumulate components in the 74220 magma

Trace element abundance patterns for crystallized beads from 74220 are consistent with removal and addition of olivine and ilmenite (Table S4). When crystalline beads are excluded, the 74220 holohyaline beads exhibit major, minor, and trace element abundances that cluster tightly suggesting a homogenous melt composition, with a negative Eu anomaly of 0.78. There are two likely scenarios to explain the Eu anomaly exhibited by the high-Ti glasses. First, a substantial portion of the source of the 74220 beads could have formed after plagioclase removal from the lunar magma ocean, as suggested by Shearer & Papike (1993). Second, there was significant interaction of 74220 melts with magmas that had seen earlier plagioclase removal. This alternative seems less likely given the lack of evidence for relatively shallow magma ocean cumulate assimilation, as observed for example in 15426. Regardless of which scenario is correct, the source of the 74220 beads is unlikely to be pristine undifferentiated mantle, consistent with the high-Ti contents of the beads and HFSE fractionation suggesting contribution from a hybridized and deep cumulate mantle source, possibly with a substantial ilmenite-rich and high-Ca pyroxene cumulate component. There is no evidence from the glasses for apatite in the source, either from P contents, which are low, or the high P/Ce ratio. Zirconium, Hf, Nb, and Ta are refractory trace elements with broadly similar geochemical behavior. Apollo 74220 holohyaline data have subchondritic Nb/Ta and Zr/Hf (Fig. 10). Münker (2010) reported that in general, subchondritic Nb/Ta and Zr/Hf ratios indicate an ilmenite–clinopyroxene-rich source. Experimental evidence indicates that ilmenite was left in the 74220 source after melting because ilmenite does not appear on the liquidus of the 74220 melt composition at any of the multiple saturation points determined for 74220 (Krawczynski & Grove, 2012). Apollo 74220 Zr/Hf and Nb/Ta ratios from this study may be consistent with the concept that ilmenite was completely removed during the partial melt formation of the 74220 parental magma. Apollo 15555 and 12063 ilmenite data are similar to that of 74220, with Nb/Ta ratios below the chondritic ratio (Fig. 10). The source region of these ilmenites was modelled using partition coefficients from (Klemme et al., 2006). The calculated compositions with ilmenite in the source are primarily in the lower lefthand quadrant of Fig. 10, exhibiting Zr/Hf and Nb/Ta below that of the chondritic ratios (Zr/Hf =34.3 ± 0.3; Nb/Ta =19.9 ± 0.5). The similarity between HFSE in 74220 holohyaline beads and 15555 and 12063 ilmenites is consistent with all samples being derived from an ilmenite–clinopyroxene-rich source and for the presence of residual ilmenite in the source.

Unlike for 15426, it appears that the 74220 bead deposit is more homogeneous, coming from deeply derived melts with an ilmenite overturn component (Fig. 15). These beads did not experience shallower assimilation or stalling, but instead experienced significant fractional crystallization during eruption into a transient atmosphere at the lunar surface. Their compositions, with negative Eu* anomalies, high Ti and REE contents, and sub-chondritic Nb/Ta and Zr/Hf indicate a mixture of primitive lunar mantle and cumulate overturn components with ilmenite and high-Ca pyroxene that formed after plagioclase saturation in the magma ocean. This would imply that the 74220 glass bead deposit, while containing a magma ocean overturn component, can be considered as generally more primitive than some of the glass groups (15B, C, D, E) from the 15426 deposit.

Implications for fractional crystallization and melt inclusion compositions in olivine

The crystalline materials within 74220 suggest saturation of both olivine and ilmenite, with crystallization occurring during quenching or subsolidus during slow cooling. Hauri et al. (2011) measured olivine crystals and their hosted melt inclusions within the pyroclastic glass. The olivine hosts are in equilibrium with the surrounding holohyaline components, assuming an Fe–Mg exchange coefficient (⁠|${K_d}_{Ol- Melt}^{Fe- Mg}=\frac{X_{melt}^{Mg}\times{X}_{Ol ivine}^{Fe}}{X_{melt}^{Fe}\times{X}_{Ol ivine}^{Mg}}$|⁠) K_d = 0.27 (Fig. 16). However, the melt in the olivine-hosted melt inclusions are not in equilibrium with the olivine. Olivine–melt inclusion pairs in the Hauri et al. (2011) study exhibit Fe–Mg exchange K_d values ranging from 0.060 to 0.245, which are attributed to post-entrapment crystallization of olivine on the inside walls of the olivine host, a common feature of olivine-hosted melt inclusions (Sobolev, 1996).

Melt inclusions are, in principle, a direct sample of the magma at the time of host-phase crystallization, and a melt inclusion within an early crystallizing phase, such as olivine, has the potential to provide a close approximation of the parent magma (Sobolev, 1996; Danyushevsky et al., 2004). Interpreting melt inclusions with respect to the parent magma, however, can be challenging because melts fractionate during cooling through crystallization of the host phase onto the inclusion walls and potentially through crystallization of daughter minerals (Sobolev, 1996). Melt inclusions can represent snap shots of melt captured throughout the entire crystallization sequence and do not represent solely the initial stages of crystallization (Basu Sarbadhikari et al., 2011). These factors can lead to difficulty in accurately interpreting the volatile element composition of the interior of the Moon based on 74220 samples alone (Hauri et al., 2011; Chen et al., 2015; Ni et al., 2017, 2019; McCubbin et al., 2023). Hauri et al. (2011) indicated that the amount of post-entrapment crystallization has a minor effect on the volatile contents of the melt inclusions. Hauri et al. (2011) calculated and corrected melt inclusion compositions by adding back equilibrium olivine in 0.1% increments, recalculating the melt composition at every step until the melt inclusions were in Fe-Mg equilibrium with olivine crystals in which they were hosted, as evidenced by olivine-melt K_d values of 0.250–0.280. Post-entrapment crystallization can lead to the depletion of elements that are compatible in the host mineral (Roedder, 1979; Danyushevsky et al., 2002), which in 74220 would be the olivine crystallized beads as evidenced by Fig. 3. The corrected volatile element contents were determined by multiplying the percentage of post-entrapment crystallization with the measured volatile element abundance and then subtracting that from the original measured composition (Hauri et al., 2011), leading to a lower volatile element abundance than measured. Post-entrapment crystallization–corrected abundances reported by Hauri et al. (2011) range from 270 to 1202 μg/g for H₂O, 36.6 to 72.1 μg/g for F, 447 to 884 μg/g for S, and 0 to 2.17 μg/g for Cl. We recalculated post entrapment crystallization and volatile element abundances within melt inclusions using melt compositions from this study and found a similar amount of post-entrapment crystallization to Hauri et al. (2011). The amount of post-entrapment crystallization was calculated using a similar methodology but with major element data from this study and olivine major element data from Hauri et al. (2011). Calculations suggest relatively minimal changes to the H₂O, F, Cl, and S contents, with ~15% differences in calculated post-entrapment crystallization. Corrected post-entrapment crystallization abundances re-calculated from this study range from 266 to 1130 μg/g for H₂O, 36 to 68 μg/g for F, 441 to 832 μg/g for S, and 0 to 2.31 μg/g for Cl. These abundances of volatile elements suggest up to ~0.1 wt % H₂O in the melt and less in the source of the 74220 pyroclastic deposit (Ni et al., 2019). The average of 25–30% post-entrapment crystallization approximately matches the olivine crystallization of the beads themselves as noted from our modal analyses. Volatile element abundance estimates by Hauri et al. (2011) are therefore robust and systematic of the petrogenesis of the Apollo 17 pyroclastic glass bead deposit. A remaining uncertainty, however, lies in the contribution of late-stage ilmenite + clinopyroxene magma ocean cumulates that overturned into the source of the high-Ti orange glasses. Modeling indicates that these late-stage products could be enriched in volatile elements (e.g. Dhaliwal et al., 2018) and so lead to an overestimate of water in the primitive lunar interior. Alternatively, late-stage magma ocean products may also be depleted in volatile elements such as water (e.g. Day & Moynier, 2014) and so lead to a reduction in the estimated original water content. Determining between these possibilities is important for understanding the initial volatile contents and later volatile element evolution of the Moon.

CONCLUSIONS

The Apollo 15 and Apollo 17 lunar pyroclastic glasses are important samples for constraining the nature of the lunar interior and the differentiation processes that acted on the Moon. New petrography shows the presence of a range of variably crystallized bead types in both the Apollo 15 (represented by 15426) and Apollo 17 (represented by 74220) bead deposits. For both deposits, major and minor element abundance data reveal significant compositional differences between crystallized and holohyaline beads. Accounting for this late-stage crystallization in both deposits, responsible for the 15D and 15E type beads in 15426 and ilmenite–olivine-bearing beads in 74220 enables elucidation of glass bead deposit formation at both the Apollo 15 and Apollo 17 sites. In particular, the Apollo 15 beads are larger and more fractured, and their melt composition was more heterogenous, with beads with REE abundances ~6 × CI chondrite representing the 15A, B (and C, not observed in this study) types and another, the J Group (yellow) bead population, at ~60 × CI chondrite and with negative Sr and Eu anomalies. Apollo 17 beads are smaller, less fractured, and more homogenous, suggesting more limited fractional crystallization and a more homogenous melt source. The Apollo 74220 beads display subchondritic Nb/Ta and Zr/Hf values and are similar to ilmenites from some Apollo 12 and 15 mare basalts that are assumed to have compositions of late magma ocean cumulate ilmenite. The HFSE signature of 74220 coupled with the results of phase equilibrium experiments on 74220 melt compositions are consistent with ilmenite being residual in the source of 74220 parental melts. The negative Eu anomalies and HFSE characteristics in the Apollo 74220 analyses indicates partial derivation from, or interaction with, magma ocean cumulates likely from overturn of late-stage ilmenite + high-Ca pyroxene. The negative Eu and Sr anomalies within the lower REE abundance beads of 15426 groups 15C (and C) indicate that they are derived from material that experienced plagioclase removal, consistent with derivation from or interaction with magma ocean cumulates or their melts. These results indicate that the 15426 glass deposit represents a jumble of primitive (15A), exotic (J Group), and hybridized beads from assimilation of magma ocean cumulates (15B,C) as well as beads that fractionated during eruption to the lunar surface (15D,E), that were erupted into a transient lunar atmosphere and mixed together, possibly by impact. The high Ni and Co contents at the edge of the 15426 beads may also suggest some impact contamination affected the deposit. 74220 glass deposit was likely erupted from a magma from a more homogeneous source, also erupted into a transient atmosphere, and saw some extent of fractionation of ilmenite and olivine in some beads, with limited evidence for impact contamination from siderophile elements (Ni, Co). The 74220 glass deposit is important due to the high estimated water contents from melt inclusions of olivine associated with it (Hauri et al., 2011). Our calculations suggest minimal changes to the H₂O, F, Cl, and S contents, with ~15% differences in calculated post-entrapment crystallization. The average of 25–30% post-entrapment crystallization approximately matches the olivine crystallization of the beads themselves, but the presence of a transient lunar atmosphere during eruption, and the presence of overturned late-stage ilmenite + high-Ca pyroxene magma ocean cumulates in the sources means these volatile contents may not solely reflect the composition of the primitive silicate Moon.

AUTHOR CONTRIBUTIONS

J.M.D.D. designed the research; E.C.M, F.M., K.V.K, R.H., M.P. and J.M.D.D. performed the research; J.M.D.D. and F.M. contributed new reagents/analytical tools; E.C.M and J.M.D.D. analyzed the data and wrote the draft manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

DATA AVAILABILITY

The data underlying this article are available in the article and in its online supplementary material.

SUPPLEMENTARY DATA

Supplementary data are available at Journal of Petrology online.

ACKNOWLEDGEMENTS

We thank NASA CAPTEM for the provision of samples. Support for this work came from the NASA Emerging Worlds (NNX15AL74G; 80NSSC19K0932) program. F.M.M. acknowledges support from NASA’s Planetary Science Research Program during this work. Reviews and comments by T. Fagan, M. Norman, R. Korotev, and an anonymous reviewer are greatly appreciated.

REFERENCES