A tale of five enclaves: Mineral perspectives on origins of mafic enclaves in the Tuolumne Intrusive Complex

Mafic magmatic enclaves are nearly ubiquitous in arc plutons. Their origins were the subject of significant debate over the past century (summarized by Barbarin and Didier, 1991). Proposed origins include reworked (i.e., chemically modified) xenoliths (e.g., Bowen, 1922; Bateman et al., 1963), disrupted, preexisting mafic dikes (Roddick and Armstrong, 1959; Cobbing and Pitcher, 1972), and residual material (restite) from the magma source region (White and Chappell, 1977). The similarities in mineral assemblages of enclaves and their host granitic rocks led other workers to propose origins as “autoliths”: essentially cumulates formed in the host magma (e.g., Pabst, 1928) or fine-grained marginal facies disrupted into the host magma (e.g., Bateman et al., 1963; Didier, 1973).

Another explanation for the mineralogical similarities between enclaves and host is that the enclaves are products of fragmentation of magmas that were undercooled when injected into the host magma (Eichelberger, 1978; Gamble, 1979; Reid et al., 1983; Vernon, 1983, 1984; Bacon, 1986; Barnes et al., 1986). This interpretation recognizes mafic enclaves as potential recorders of the nature of recharge magmas and of magma mixing processes. The literature now contains hundreds of studies on enclave textures, mineral assemblages, and geochemical compositions, and it is generally accepted that most such enclaves result from magma mingling and/or mixing in crustal magma reservoirs.

In the volcanology community, enclaves as evidence for magma influx are particularly important because recharge of mafic magmas is a potential trigger for eruptions (e.g., Feeley et al., 2008; Shane at al., 2008; Humphreys et al., 2009; Ruprecht and Bachmann, 2010). Moreover, in volcanic systems, rapid quenching of eruptive products, including mafic enclaves, permits assessment of original (high-temperature) mineral assemblages and compositions of minerals and melt (glass), providing clear information about mixing end members (e.g., Tepley et al., 1999; Salisbury et al., 2008; Schmidt and Grunder, 2011; Ruprecht et al., 2012; Chadwick et al., 2013; Allan et al., 2017; Humphreys et al., 2019). In some instances, the compositions of quenched mafic enclaves may represent the composition of mafic end-member magmas (but see Bacon, 1986).

Unlike mafic enclaves from volcanic rocks, mafic enclaves in plutons commonly display mineral assemblages and major-element mineral compositions identical to those in the adjacent host (e.g., Vernon, 1983; Barbarin, 1990, 2005; Barnes et al., 1990; Dorais et al., 1990; Allen, 1991), and in some instances, the enclaves contain large crystals (e.g., alkali feldspar megacrysts and quartz ocelli) that would not be expected to crystallize from a melt of the enclave’s bulk composition (e.g., Hibbard, 1991; Baxter and Feely, 2002). The uniformity in mineral assemblage is interpreted to result in part from physical mixing (e.g., Vernon, 1983, 1990), but also from diffusive exchange between enclave and host magmas (e.g., Baker, 1991; Tepper and Kuehner, 2004; Humphreys et al., 2010). Textural features, including fine-scale crystal habits and inclusion relationships within such enclaves preserve evidence of undercooling and mixing (e.g., Hibbard, 1981, 1991; Vernon, 1983, 1990), so it is interesting to ask whether mineral compositions and zoning patterns may, in fact, also preserve a magmatic evolutionary history distinct from that of the host.

As with mafic enclaves in volcanic rocks, the bulk compositions of mafic enclaves in plutons have been interpreted by some to represent the composition of potential parental magmas or of potential mafic end members of magma mixing (e.g., Reid et al., 1983). However, if it can be shown that physical and/or chemical exchange occurred between an enclave and its host magma, then the enclave composition will not be an accurate representation of a mafic end-member magma, whether parental or not (e.g., Vernon, 1990; Barbarin and Didier, 1992; Browne et al., 2006).

In this study, we analyzed major- and trace-element compositions of calcic amphibole and plagioclase in five enclaves and their adjacent host rocks: the equigranular Half Dome unit of the Tuolumne Intrusive Complex (TIC; Bateman and Chappell, 1979; Huber et al., 1989). Our goal was to determine whether trace-element abundances in amphibole and plagioclase in enclaves are distinct from those in the host, and if so, to interpret the reasons for these distinctions. We found that while major-element compositions of enclave minerals are similar to those in the host, trace-element abundances and zoning patterns are generally dissimilar to the host phases. Moreover, although some large crystals in enclaves were inherited from the adjacent host magma, it is more common that these large crystals are dissimilar to equivalent crystals in typical equigranular Half Dome (eHD) rocks, with the implication that some enclaves record multiple magma mixing and/or mingling events. Our results indicate that although the mafic enclaves studied are related to magma influx, (1) the new magma was not necessarily mafic, and (2) enclave bulk compositions are likely to reflect early mixing of crystals and melt from the host, and in some cases followed by loss of melt to the host.

The TIC (Fig. 1) is one of four Late Cretaceous complexes that make up the Sierra Crest suite of the Sierra Nevada batholith (Coleman and Glazner, 1997). It consists of five mapped units, the outermost of which is Kuna Crest Granodiorite and equivalent rocks exposed along the margin of the TIC. Successively interior units are the Half Dome Granodiorite, which is subdivided into outer equigranular, and inner porphyritic units, Cathedral Peak Granodiorite, and Johnson Granite Porphyry (Bateman and Chappell, 1979; Huber et al., 1989). Zircon U-Pb ages range from 94.86 ± 0.29 Ma to 83.86 ± 0.31 Ma (Paterson et al., 2016). The enclaves described here were collected from the eHD unit, in which U-Pb (zircon) ages range from 91.7 ± 0.2–89.6 ± 0.2 Ma (Coleman et al., 2004; Memeti et al., 2010). The youngest of these ages overlap with those of the porphyritic Half Dome, which is consistent with typical gradational contacts between the two units (e.g., Žák and Paterson, 2005; Paterson et al., 2016).

The eHD is primarily composed of granodiorite but ranges from tonalite to leucogranite (e.g., Bateman and Chappell, 1979). It displays variably sharp to gradational contacts with adjacent units (Bateman and Chappell, 1979; Žák and Paterson, 2005; Memeti et al., 2010; Paterson et al., 2016). The eHD is characterized by large (cm-scale), euhedral hornblende phenocrysts and prominent titanite crystals as long as 1.5 cm (Supplemental File 1¹). It is also heterogeneous, with variations in the abundance and proportion of hornblende and biotite, schlieren banding, mafic enclaves (Figs. 2 and 3), and a variety of magmatic structures (Žák and Paterson, 2005; Paterson et al., 2008; Paterson, 2009; Ardill et al., 2020). In addition, Huber et al. (1989) and Coleman et al. (2012) mapped elongate felsic bodies within the eHD, which the latter authors interpreted to represent differentiates of distinct magma batches.

Samples were hammered from outcrops or spalled slabs. Where possible, both the enclave and immediately adjacent host were sampled. In the case of sample TLM-53, the host is the matrix of an enclave dike (Fig. 3D). Therefore, the host of the dike itself was also sampled (TLM-54). Two or more polished thin sections were prepared, including ones that crossed the enclave-host contact where possible. Representative pieces of the enclaves were crushed, powdered in alumina, and fused into glass disks using a 1:2 ratio of sample to flux (Claisse M4 Fluxer: 49.75% lithium metaborate, 49.75% lithium tetraborate, and 0.5% lithium bromide). Major- and minor-element compositions (SiO₂, TiO₂, Al₂O₃, Fe₂O₃, MnO, MgO, CaO, Na₂O, K₂O, and P₂O₅) were determined by X-ray fluorescence (XRF) at the Texas Tech University Geoanalytical Laboratory using a Thermo Scientific ARL Perform’X. Element abundances were determined with operating conditions that varied from 30 to 60 kV, 60–120 mA, and count times ranging from 8 to 40 seconds. Analyses were calibrated using a suite of international standards. Data were collected during four separate analytical sessions. Precision and accuracy were determined using the standards AGV-1, BHVO-2, and GSP-1. For individual analytical sessions, the relative standard deviation (RSD) for all elements was <3%. However, the %RSD value determined for the combined analytical sessions is higher for Mg, up to 8%. Additional enclave data were obtained by X-ray fluorescence at Pomona College. Rock powders were fused in a 1:2 ratio with Spectromelt A10 (di-lithium tetraborate) flux, following procedures described in Johnson et al. (1999). The beads were analyzed using a 3.0 kW Panalytical Axios wavelength-dispersive XRF spectrometer at Pomona College. Each analysis represents an average of three measurements. Bulk-rock compositions are presented in Supplemental File 2 (^{footnote 1}).

Major-element concentrations of hornblende (Hbl) and plagioclase were analyzed by electron microprobe at the University of Oklahoma. Operating conditions were 20 kV accelerating voltage, 20 nA beam current, and 1–2 µm spot size, using natural and synthetic standards. Hornblende site occupancies were estimated on the basis of 23 oxygens and classified according to Leake et al. (1997; see Esawi, 2004). Trace-element abundances were measured using in situ laser ablation–inductively coupled plasma mass spectrometry (LA-ICP MS) at Texas Tech University on polished sections. Ablation was done using an NWR 213 nm solid-state laser with a dual-volume cell. The ablated aerosol was analyzed using an Agilent 7500CS quadrupole ICP-MS. Nominal operating conditions for Hbl were spot diameter 40 µm, laser pulse rate of 5 Hz, and fluence of 7.1–8.8 J/cm². Nominal operating conditions for plagioclase were 60 µm spot size, 5 Hz pulse rate, and fluence of 2.5–3.5 J/cm². The spot size used for plagioclase analysis was chosen to permit spatially resolved analysis of thin, calcic mantles. This relatively small size meant that some elements, particularly the heavy rare-earth elements (REE), were below detection limits.

For each analysis, 25 s of background (laser off) and ∼60 s of signal were recorded. The primary analytical standard, U.S. Geological Survey (USGS) glass GSD, was analyzed after every 5–10 unknowns. Precision was determined by repeated analysis of basaltic glass BHVO-2g. Long-term precision (RSD) ranges from 2.1%–9.2%, and <6% for most trace elements. Trace-element abundances were normalized to that of CaO for amphibole and SiO₂ for plagioclase. The ablation spectra and reduced data were inspected for anomalously high counts or spikes of P, Ti, and Zr. Such analyses were omitted from the data set. For Hbl, limits of detection (LOD) are generally less than 1% of the abundance reported, but reach 5% for Rb and Ta and 13% for Th and U. For plagioclase, LOD are <2% of reported abundances for Ca, Fe, Ga, Sr, Ba, La, and Ce, 4%–5% for Ti and Zn, 8% for Pr, and ∼18% for Nd and Eu. The high LOD for Pr, Nd, and Eu is due to the small spot size. Results of Hbl and plagioclase analyses are presented in Supplemental Files 3 and 4 (^{footnote 1}), respectively. The locations of analytical spots are illustrated in Supplemental Files 5–9.

For the sake of clarity, in the following we refer to the magmatic enclaves under study as mafic enclaves, enclaves, or mme. Given the large area of the TIC (1100 km²), the 10 m.y. incremental growth history (Paterson et al., 2016), the varied spatial dimensions of units from meter-scale sheeting, to km-scale large lobes, to asymmetrically nested 10s of km-scale diapirs (Memeti et al., 2014), and the widespread internal magmatic structures associated with enclaves (Ardill et al., 2020), enclaves in the TIC can be expected to show wide variation in composition, textural features, and evolution. Field observations and bulk-rock analyses support this expectation. Enclave bulk compositions are mostly diorite/monzodiorite, but include quartz diorite, quartz monzodiorite, tonalite, and granodiorite (Table 1; Supplemental File 1 [^{footnote 1}]). Specific gravity of enclaves from Kuna Crest and Half Dome units varies from 2.6 to 2.75 gcm⁻³ (Link, 1969). The enclaves range from cm to decameter scales, but most are less than 0.5 m long, and most are elliptical, with long/short axis ratios between 1.2–4 (Link, 1969; Figs. 2A and 2B), although some irregular (Fig. 2C) and rarely quadruple-pronged enclaves are observed (Paterson et al., 2003). Double (i.e., one enclave within another) and rarely triple enclaves occur, and some enclaves have preserved mafic or felsic rims (Figs. 2D and 2E). Enclave-host contacts vary from sharp to gradational and are locally crossed by veining and crystal growth. Enclave microstructures and crystal shapes and sizes also vary. Some enclaves are equigranular (Figs. 2F and 2G), but most are porphyritic (Figs. 2A–2C), with phenocrysts varying between hornblende and plagioclase ± biotite ± quartz ± titanite (Table 1; Supplemental File 1 [^{footnote 1}]).

The settings of TIC enclaves are equally varied. Many are isolated (Figs. 2A, 2B, 2D, 2H, and 3A) suggesting that they formed elsewhere and were then distributed in TIC host magmas during ascent and emplacement. These isolated enclaves have the largest spatial density (smallest spacing) in the eHD and Kuna Crest units, are less common in Cathedral Peak units, and are rare in leucogranites. Spacing of isolated enclaves in Kuna Crest and Half Dome units varies from 0.2 to 2.7 m⁻² and in the Cathedral Peak unit to 0.1–0.01 m⁻² (Link, 1969; Paterson et al., 2016). Link (1969) described local increases in enclave abundance outward toward internal contacts, although our studies suggest this is not a general observation.

In contrast to isolated enclaves, examples of in situ TIC enclave formation are preserved where dikes disaggregated and mixed with adjacent host magma (Fig. 2I; Bateman, 1992). Enclaves formed in this fashion are much less common than isolated examples.

A third setting in which TIC enclaves are preserved is in local swarms (Figs. 2E–2G and 2J) and/or accumulations associated with local compositionally defined magmatic structures (e.g., Reid et al., 1993; Tobisch et al., 1997; Paterson, 2009; Ardill et al., 2020). These swarms and/or accumulations are typically associated with accumulations of phenocrysts, and less commonly with xenoliths and cognate inclusions (Figs. 2K and 2L). Thus, they are generally interpreted to reflect objects that formed in a variety of locations and subsequently collected at or near the emplacement site by physical flow sorting processes (Barbarin et al., 1989; Tobisch et al., 1997; Ardill et al., 2020). Enclave swarms occur in irregularly shaped pools (Figs. 2E, 2F, and 2J), dike-shaped bodies (Fig. 3D), local diapirs (Fig. 2L), and in various compositionally defined magmatic structures such as troughs, tubes, and pipes (Figs. 2G and 2M; Paterson, 2009; Ardill et al., 2020).

Enclaves are generally aligned with one of several magmatic fabrics (foliation + lineation) in the host (Figs. 2D, 2G, 2H, and 2N; Link, 1969; Paterson et al., 1998; Ardill et al., 2020). Internal enclave fabrics range in intensity of alignment and in geometric relationship to host fabrics. Some enclaves are associated with structures (folds, faults, and boudinage), which indicates strong magmatic deformation of the enclave and implies variable rheologic contrasts and, potentially, physical exchange between enclave and host magmas.

Estimates of enclave density established from field measurements are between 3 and 0.01 enclaves m⁻²; values that lead to estimates of several millions of exposed enclaves in the Kuna Crest, Half Dome, and Cathedral Peak units (Link, 1969; Paterson et al., 2016). Among these millions, the range of geologic and petrographic characteristics implies numerous different histories, locations of formation, and modes of collection and transport. This variety presents a challenge of how best to begin a study of the enclaves, since detailed mineral analysis of the entire range of magmatic enclave types and their host magmas is well beyond the scope of this work. We chose to begin by sampling from the eHD, the unit with the greatest number of enclaves, and in order to compare enclave mineral compositions with recent geochemical results on hornblende, plagioclase, and titanite from the host eHD granodiorite (Werts, 2019; Werts et al., 2020). We selected three isolated enclaves, one of which is a double enclave, one from an enclave swarm, and one from an enclave-rich dike (Table 1; Fig. 3). Samples were unaltered or weakly altered and displayed a range of phenocryst proportions and color indices. It should be emphasized that the results of this study provide a glimpse into the likely variability of enclave histories in the TIC, a topic we hope to explore in future projects.

Enclaves studied encompass a variety of rock types (diorite, quartz diorite, quartz monzodiorite, tonalite, and granodiorite; Table 1; Supplemental File 1 [^{footnote 1}]), color indices, and proportions of Hbl and biotite. Many enclaves contain crystals distinctly larger than their matrix. Although the term “large crystal” is appropriate in some samples, it does not encompass instances in which the larger grains are sub-millimeter in length. We therefore chose to refer to grains larger than their matrix as phenocrysts, without implication as to their origin. Some enclaves are equigranular (e.g., TCB-8i), but most are porphyritic, with phenocrysts of Hbl and plagioclase ± biotite ± quartz ± titanite. The groundmass consists of Hbl (acicular to prismatic), plagioclase, biotite, and magnetite ± quartz ± alkali feldspar ± titanite. Acicular apatite is common but not ubiquitous, and ilmenite occurs in some samples as inclusions in Hbl and titanite (Table 1). Groundmass textures vary from hypidiomorphic granular to idiomorphic granular, and groundmass Hbl is aligned in some samples. Hornblende, plagioclase, biotite, and apatite are ubiquitous groundmass phases, quartz and magnetite are common, and alkali feldspar and titanite are locally common (Table 1).

Plagioclase phenocrysts characteristically display intermediate to sodic core zones (An_46–33), commonly surrounded by a thin zone (mantle) of more calcic plagioclase (An_70–42; Fig. 4B), which is in turn surrounded by broad rims whose compositions are, on average, less calcic than the core (An_40–25). Details of plagioclase zoning and inclusion relationships are presented below.

Hornblende phenocrysts reach ∼1 cm in length. Some display fine oscillatory color zoning (Fig. 4A), whereas most lack such zoning. Inclusions of magnetite are common, with fewer inclusions of apatite, plagioclase, biotite, ilmenite, and scant zircon. In addition to prismatic phenocrysts (Fig. 4C), some samples contain mm-scale clusters of granular Hbl + magnetite + titanite (particularly the interior of enclave TCB-8; Fig. 4D), and some large Hbl grains display mottled color zoning. Hornblende phenocrysts and groundmass crystals are locally altered to and replaced by biotite (some altered to chlorite) ± magnetite ± titanite.

Biotite and titanite are common but not ubiquitous. Biotite occurs as phenocrysts, as poikilitic grains (Fig. 4E), and as small groundmass crystals (Fig. 4C). Titanite phenocrysts are locally present, and in some samples, titanite is poikilitic (Fig. 4F). Quartz and alkali feldspar, where present, are interstitial and poikilitic to interstitial, respectively.

The immediately adjacent host rocks to these enclaves are generally coarse grained, hypidiomorphic granular, and vary from tonalite to granodiorite (Supplemental File 1 [^{footnote 1}]). Hornblende and biotite are present in sub-equal proportions. Hornblende grains display a seriate size distribution, olive- to yellow-green pleochroism, and inclusions of apatite + plagioclase ± magnetite ± titanite. Hornblende is locally replaced by biotite, actinolite, and epidote. Books of brown biotite reach four mm in diameter and contain scant inclusions of plagioclase + apatite ± ilmenite ± magnetite ± titanite ± zircon. Chlorite is a minor alteration product. Plagioclase cores are typically weakly oscillatory-normal zoned, with some patchy and/or box-work zoning and rare calcic zones. The core zones contain very few fine-grained inclusions (apatite ± biotite ± Hbl ± Fe-Ti oxides ± zircon). In some samples, plagioclase rims enclose larger proportions of these phases. Compositions of the great majority of plagioclase grains range from An₄₀ to An₃₀, with extremely rare calcic zones (∼An₇₆), and with rims as sodic as An₂₀ (also Burgess, 2006; L. Oppenheim, written commun., 2020). Interstitial quartz contains Hbl and plagioclase inclusions. Alkali feldspar is interstitial to poikilitic and locally is finely perthitic. It encloses plagioclase, Hbl, magnetite, and titanite. A few Hbl inclusions in alkali feldspar are green to pale blue-green rather than olive. Euhedral titanite is common and reaches two mm in length. Interstitial accessory phases are titanite, magnetite, apatite, and zircon; myrmekite is rare. Enclave TLM-53 was collected from an enclave-rich dike (Fig. 3D), the matrix of which has a distinctly higher color index and a larger proportion of cm-scale Hbl (Fig. 3E) than the normal eHD host rocks (TLM-54).

Mafic enclave SiO₂ contents are mainly in the range 50–60 wt%, which overlaps with low-SiO₂ samples of the Kuna Crest and eHD units (Fig. 5). In most major-element plots, TIC samples plot in an approximately linear array. The mme compositions extend this array in terms of Al₂O₃ and TiO₂ (Fig. 5E). However, when compared to the overall TIC trend, mme display lower values of Mg/(Mg+Fe_t) (Fig. 5A), lower CaO and CaO/Al₂O₃ (Fig. 5B), and higher Na₂O and K₂O (Fig. 5C and 5D, respectively). Moreover, the mme show considerable scatter of Na₂O, K₂O, and Zr (Fig. 5F). The compatible trace elements Cr (Fig. 5G) and Ni (Fig. 5H) also display significant scatter. A few samples display Cr and Ni contents higher than nearly all TIC rocks, whereas in other samples, Cr and Ni contents are near detection limits (e.g., Fig. 5H). Similarly, Sr (Fig. 5I) and Ba (Fig. 5J) display scatter and lack discernable trends when plotted against SiO₂. Although Sr abundances in mme overlap the majority of eHD compositions (Fig. 5I), Ba abundances in the mme are much lower than in eHD samples and are not collinear with the eHD trend (Fig. 5J).

We calculated zircon saturation temperatures using bulk-rock major-oxide and Zr concentrations. The calibration of Watson and Harrison (1983) yielded values of 668–760 °C, with all but one value less than 745 °C. More recent calibrations (e.g., Boehnke et al., 2013) yield still lower values. Moreover, every enclave displays an “M” value [cation ratio (Na + K + 2Ca)/(Al x Si)] greater than 1.9, which places all enclave compositions outside the range of calibration of the geothermometer. In addition, among the samples described in detail here, all calculated temperatures are less than 720 °C (Watson and Harrison, 1983 calibration). In contrast, the apatite saturation thermometer (Harrison and Watson, 1984) yielded temperature estimates of 760–960 °C for all enclaves and 760–905 °C among the five enclaves under study.

Plagioclase phenocryst cores are intermediate to sodic (An_46–33). Inclusions of ferromagnesian silicates in core zones vary from sample to sample: they are lacking in TLM-61B, and they are biotite ± Hbl in enclaves TCB-8, YTSB-14B, and TLM-53. In enclave TCB-7, some cores contain augite and biotite inclusions, whereas others contain biotite ± Hbl. Most plagioclase cores also contain scant inclusions of tiny (less than a few tens of microns), elongate apatite, magnetite, ilmenite, and rare zircon. The plagioclase cores are commonly surrounded by a thin mantle of more calcic plagioclase (An_70–42; Fig. 4B), which is in turn surrounded by broad rims of An_40–25. The mantle and rim zones contain inclusions of Hbl, biotite, and acicular apatite ± Fe-Ti oxides ± titanite ± zircon. Thus, in general, plagioclase phenocryst cores are slightly more calcic than rims and contain fewer inclusions. Inclusion assemblages in rims are the same as groundmass phases, including acicular apatite, whereas core inclusions may be distinct from the groundmass.

Trace-element data for plagioclase are plotted against CaO in Figures 6–8. In order to directly compare trace-element to major-element compositions, the CaO values used in these diagrams are those determined on the same analytical spot by LA-ICPMS.

Plagioclase in enclave TLM-61B can be divided into three types (Fig. 6A): phenocryst cores or mantles with relatively high CaO and Sr (group 1), phenocrysts whose compositions are identical to plagioclase in the host rock (group 2), and enclave groundmass grains (group 3), some of which are more calcic than host plagioclase but have lower Sr abundances than the calcic phenocryst zones (Fig. 6A). The compositional variation in a core-to-rim traverse across a group 1 phenocryst is illustrated by red-dashed lines and indicates normal zoning in the core followed by reverse zoning to form a Ca-rich mantle, then normal zoning in the mantle and rim to low-Sr, sodic compositions. In addition, one group-2 phenocryst core is mantled by a Ca-rich zone with Sr abundance intermediate between group 1 cores and group 3 groundmass grains. These groups are also evident in plots of Ce (Fig. 7A) and Ba (Fig. 8A). It is noteworthy that the Ce and Ba contents in host plagioclase and group 2 phenocrysts extend to higher values than in group 1 plagioclase and are essentially identical to plagioclase in the host. This feature indicates that the calcic zones in group l phenocrysts must represent a distinct origin compared to group 2 phenocrysts, and potentially that group 2 phenocrysts were derived from the host magma.

One group of plagioclase phenocrysts in enclave TCB-7 contains inclusions of augite and biotite (grains 2 and 5), whereas the remaining analyzed crystals contain Hbl and biotite inclusions and display prominent calcic mantles (Supplemental File 1 [^{footnote 1}]). Compositions of grains 2 and 5 (Figs. 6–8) tend to overlap compositions of groundmass plagioclase, although interior zones of these crystals are slightly higher in Sr and Ce. Compositions of the remaining analyzed crystals are similar to those of grains 2 and 5, except that the calcic mantles are distinctly higher in Ce and generally lower in Ba contents (Figs. 7 and 8). Plagioclase crystals in the host to enclave TCB-7 are similar to enclave plagioclase in CaO, Ce, and Ba contents, but on average display lower Sr abundances (Figs. 6–8).

Compositions of phenocryst and groundmass plagioclase in enclave YTSB-14B overlap in terms of CaO and Sr and show decreasing Sr with decreasing CaO (Fig. 6B). Ce abundances scatter (Fig. 7B) and Ba increases with decreasing CaO (Fig. 8B). One phenocryst and one groundmass crystal display significantly higher Ba than other plagioclase in this enclave.

Enclave TCB-8 is a double enclave (Fig. 3C) with mm-scale poikilitic plagioclase in the interior and blocky phenocrysts as much as 5 mm long in the outer zone. In general, outer-zone phenocrysts and groundmass compositions overlap with those of plagioclase in the adjacent host, with large variations in Sr, Ce, and Ba. These wide ranges of trace-element abundance occur over a narrow range of CaO, except in crystal rims and groundmass grains, in which CaO and the trace elements are correlated (Figs. 6–8). In contrast, compositions of plagioclase oikocrysts in the interior zone plot in a trend that crosses the trend of outer-zone plagioclase. These inner-zone grains display decreasing Sr and Ce and slightly increasing Ba with decreasing CaO (Figs. 6–8).

Enclave TLM-53 was collected from an enclave-rich dike (Fig. 3D), the matrix of which has a distinctly higher color index and a larger proportion of cm-scale Hbl (Fig. 3E) than the normal eHD host rocks (TLM-54). The phenocryst cores are slightly more calcic than plagioclase in the dike matrix and are commonly surrounded by calcic mantles (Figs. 6–8). The Sr and Ce abundances in the phenocryst cores and mantles overlap with plagioclase in the dike matrix and partially overlap with plagioclase in the eHD host rocks, although the latter plagioclase compositions range to lower Sr and higher Ce contents (Figs. 6D and 7D). In contrast, Ba contents of enclave phenocrysts are lower than in plagioclase from the dike matrix, but overlap with the lowest Ba values in host plagioclase (Fig. 8D). Groundmass plagioclase in the enclave displays lower Sr and Ce than the phenocrysts.

Calcic amphibole in the enclaves and their host rocks is magnesio-hornblende (Hbl), except for a few grains with compositions that plot in the tschermakite field (Fig. 9A). When compared to Hbl compositions in typical eHD rocks, most enclave Hbl is slightly more magnesian, as is Hbl in the immediately adjacent, “local,” host rocks (within 1–2 cm of the enclave; Fig. 9A). Major-element compositions (as atoms per formula unit [apfu]) are generally well correlated with Si, and data from one enclave tend to overlap data from the others. The greatest distinction among enclaves is in Mg, which is illustrated, for example, in the lower Mg# (Mg/(Mg + Fe) of Hbl from enclave YTSB-14B and mainly higher Mg# of TLM-53 Hbl (Fig. 9A). It is noteworthy that Hbl Mg# is uncorrelated with bulk-rock Mg#. For example, enclave TLM-53 Hbl displays the highest Mg# at a given Si content, but the bulk-rock Mg# of this sample (0.43) is the lowest measured for the enclaves studied. Crystallization temperatures (calculated using Equation 5 from Putirka, 2016) reach 836 °C for enclave Hbl and 823 °C for host eHD Hbl (Fig. 9B). Although temperatures calculated from enclave and host Hbl overlap nearly completely, most low-T Hbl is from host eHD samples (Fig. 9B).

It is noteworthy that Hbl crystallization temperatures are significantly higher than zircon saturation temperatures (see above). This relationship is interesting because Hbl phenocrysts in the enclaves contain zircon inclusions, indicating that Hbl and zircon co-precipitated. Co-precipitation is also indicated by decreasing Zr content in Hbl with decreasing calculated temperature (not shown; see Barnes et al., 2019).

Hornblende trace-element abundances display considerably more diversity than do major elements (Figs. 10–12). In all these figures, trace-element abundances are plotted against Ti content, which is a proxy for temperature (Otten, 1984; Barnes et al., 2017; Werts et al., 2020). For example, Hbl from three enclaves (TCB-8, TLM-53, and YTSB-14B) is richer in Cr than Hbl in the other enclaves or in the host eHD (Fig. 10A). The latter display low, approximately constant Cr contents. The highest Cr values are all from double enclave TCB-8. Four of these data points are from granular Hbl clusters in the inner zone, interpreted to be the products of recrystallization of higher-T amphibole or augite, and three are from an elongate prismatic phenocryst. Zirconium in most grains is well correlated with Ti; however, some grains from enclave YTSB-14B contain much higher abundances of Zr than is typical (Fig. 10B; note that ablation spectra for each of these high-Zr data points were inspected for evidence of contamination by Zr-rich inclusions, and such evidence is lacking). Strontium is also correlated with Ti (Fig. 10C), although some enclave Hbl contains higher (e.g., TLM-53) or lower (e.g., TCB-8) Sr at a given Ti content.

Except for one grain (phenocryst 8; Fig. 11A), phenocryst and groundmass Hbl in enclave TLM-61B have lower Ni contents than the local host or typical eHD Hbl. This pattern is in contrast to Hbl in enclave TCB-7, in which phenocryst interior zones are comparable to the local host, but groundmass Hbl, phenocryst rims, and Hbl inclusions in a plagioclase phenocryst all display higher Ni contents (Fig. 11B). Hbl in enclave YTSB-14B displays higher Ni contents than in typical host eHD (Fig. 11B). One phenocryst (grain 1) is reversely zoned, with <28 ppm Ni in the core and >34 ppm Ni in rims; the latter values are similar to the other Hbl in the sample (Fig. 11B).

In the double enclave (TCB-8; Fig. 11C), three groups of Hbl are defined by Ni contents. Phenocrysts and groundmass in the interior of the enclave have the highest Ni, with the highest values in the Hbl cluster (Fig. 4D); these same grains display high Cr contents. Phenocryst and groundmass Hbl in the outer enclave have lower Ni contents, and one outer-zone phenocryst (grain 4) has Ni contents identical to those in Hbl from the host. This low-Ni phenocryst is also reversely zoned, with lowest Ti in the interior and highest Ti in outer zones (Fig. 11C). In enclave TLM-53, Ni contents in phenocrysts generally have lower Ni than Hbl in the groundmass. These values are similar to or lower than Ni contents in Hbl from the host dike and lower than Ni in Hbl in adjacent “normal” eHD (TLM-54; Fig. 11D). The most Ni-rich phenocryst (phenocryst 1) displays oscillatory color zoning, similar to Hbl in the host dike (Fig. 11D). However, the eHD host to the enclave dike (TLM-54) is unusual in having one Hbl grain with Ni > 25 ppm (Fig. 11D).

The Ni data indicate that most Hbl grains in mme are distinct from those in the adjacent host, but that a few are similar to host Hbl. In addition, some Hbl phenocrysts with distinct cores display rims with compositions identical to groundmass Hbl (e.g., TLM-61B grain 8 and YTSB-14B grain 1), whereas in others, cores and rims of phenocrysts are dissimilar to groundmass Hbl (e.g., TCB-8 grain 4). It is also noteworthy that in two of the five enclaves, Hbl phenocrysts have Ni contents lower than normal eHD Hbl.

Two compositional groups stand out in terms of Mn contents (Fig. 12). One group, with lower Mn contents, consists of all Hbl from YTSB-14B, groundmass and most Hbl phenocrysts from enclave TCB-7, and phenocrysts and groundmass from the interior of double enclave TCB-8. Within this group, Mn decreases with decreasing Ti. Samples with higher Mn consist of all Hbl from enclave TLM-61B, Hbl from the outer zone and host of enclave TCB-8, and Hbl from the host of enclave TCB-7. Similar bimodality is seen in enclave TLM-53, in which one phenocryst plus Hbl in the host dike display lower Mn contents than the remaining phenocrysts, whose higher Mn contents overlap typical eHD compositions (Fig. 12D).

The distinct bimodality of Mn contents cannot be simply a function of temperature, inasmuch as there is broad overlap in both temperature (Fig. 9B) and Ti contents (Figs. 10–12). Moreover, at a given Ti content, the high-Mn Hbl in enclave TLM-61B also yields relatively high temperature estimates (Fig. 9B). Instead, the high Mn contents are thought to reflect the relatively evolved nature of the melt from which the Hbl crystallized. A similar relationship was reported for amphibole from the English Peak pluton (Barnes et al., 2017), in which Mn contents increased from the outer, less evolved zones of the pluton to the inner evolved zone. This does not mean that the melts from which the low-Mn Hbl crystallized were mafic, because the low abundances of Sr, Zr, and Cr, combined with the low calculated temperatures, indicate crystallization from an evolved melt.

Additional detailed diagrams of trace-element variation, specifically plots of Ce and Sr versus Ti for individual enclaves, are given in Supplemental File 10, Figures SA and SB (^{footnote 1}). In the eHD host, Hbl Ce decreases with decreasing Ti, with a steep decrease between ∼8000–6000 ppm Ti, followed by a less dramatic decrease at lower Ti contents. Enclave Hbl is generally distinct in having lower Ce at a given Ti content (Supplemental File 10, Fig. SA). Exceptions are Hbl in the interior of double enclave TCB-8 and some Hbl phenocrysts in enclave TLM-53. In contrast, Sr contents in enclave Hbl overlap those of host eHD Hbl except for double enclave TCB-8, in which many Hbl phenocryst and groundmass crystals have lower Sr at a given Ti content (Supplemental File 10, Fig. SB).

Silica contents of the enclaves under study range from 50.4 wt% to 53.3 wt%. These silica contents suggest that the enclaves should represent mainly basaltic to basaltic andesite magmas. If this is the case, then evidence of the mafic parentage should be preserved in major- and trace-element contents of the constituent minerals. For example, if these enclaves had an H₂O-rich mafic parent, as suggested by silica contents and the abundance of Hbl, then enclave plagioclase would be expected to be calcic (≥An₈₀; e.g., Beard and Lofgren, 1991; Sisson and Grove, 1993; Lundstrom and Tepley, 2006). However, the most calcic plagioclase is ∼An₇₀, and plagioclase more calcic than An₄₀ occurs mainly as thin mantles on phenocrysts and in the groundmass of enclave TLM-61B, which coincidentally is one of two enclaves with the lowest Ni contents in Hbl. Thus, plagioclase compositions indicate crystallization from intermediate to felsic melts rather than mafic ones.

The CaO and SiO₂ contents of melts from which enclave plagioclase and Hbl crystallized were calculated using algorithms from Scruggs and Putirka (2018) and Zhang et al. (2017), respectively. The results indicate that the calcic mantles on plagioclase phenocrysts crystallized from melts with 60–65 wt% SiO₂ and ∼3–5 wt% CaO (Fig. 13). The remainder of the plagioclase and all Hbl phenocrysts crystallized from melts with ∼65–77 wt% SiO₂ and <2.5 wt% CaO (Fig. 13). Although both sets of calculations carry large uncertainties, the results are consistent in indicating that the bulk of minerals in these enclaves crystallized from dacitic to rhyolitic melts and that the calcic mantles on plagioclase cores crystallized from andesitic melts.

Hornblende compositional trends also indicate crystallization from evolved melts. For example, the positive correlations of Sr and Zr with Ti (Fig. 10) and calculated temperatures (Fig. 9) indicate that Hbl co-precipitated with plagioclase and zircon (e.g., Barnes et al., 2016, 2017), a condition not expected in a basaltic magma. Likewise, Cr and Ni contents of amphibole are not consistent with crystallization from a primitive basaltic magma. It is possible that the high Cr and Ni contents of some Hbl phenocrysts in YTSB-14B and TCB-8 (Figs. 10 and 11) represent crystallization from andesitic magmas. Use of the lowest partition coefficient (K_d) values for Cr and Ni in Hbl (6 and 4, respectively; Ewart and Griffin, 1994; Tiepolo et al., 2007) yield maximum enclave melt concentrations of 35 ppm Cr and 12 ppm Ni. Use of more appropriate K_d values (22 and 11.5 for Cr and Ni, respectively; Ewart and Griffin, 1994; Bachmann et al., 2005), yield maximum melt contents of ∼9 ppm Cr and ∼4 ppm Ni. By comparison, the average Cr and Ni contents of >1100 andesites and basaltic andesites are 78 ppm and 40 ppm, respectively (from database PetDB, 10 September 2018, https://www.earthchem.org/petdb). We therefore conclude that even the Hbl with the highest Cr and Ni contents crystallized from dacitic melts, rather than andesitic ones, in agreement with estimated melt SiO₂ and CaO contents (Fig. 13).

This conclusion does not negate the possibility that magmas parental to the enclaves were basaltic. Rather, it indicates that evidence for a primary basaltic parent is lacking in the major- and trace-element contents of the enclave minerals. Average compositions of primitive continental and oceanic arc basalts (Schmidt and Jagoutz, 2017) are plotted in Figure 5, which illustrates the discrepancy between primitive magmas and the enclaves under study. If the original enclave magmas were similar in composition to these primitive arc basalts, then the data indicate that such primitive magmas underwent substantial fractional crystallization, and/or hybridization with felsic magmas, prior to processes recorded in the enclave minerals. Such fractional crystallization and mixing are likely to have occurred in deeper parts of the magmatic system (e.g., Barnes et al., 1986; Annen et al., 2006; Coint et al., 2013) and in composite dikes (e.g., Collins et al., 2000).

Taken together, the data indicate that (1) the bulk of crystals in these mme crystallized from evolved magmas; (2) each enclave studied is the result of magma mixing and crystal-melt separation; (3) although the ultimate origin of enclave magmas was probably basalt, any evidence of the basaltic heritage is lost; and (4) it is not possible to identify just two end-member compositions that can explain the major- and trace-element contents and zoning patterns in Hbl and plagioclase. Instead, each enclave evidently records a distinct, complex history of hybridization, as summarized below, with reference to Figures 6–8, 12, and 13. In addition, these conclusions fail to explain the broadly basaltic to basaltic andesite bulk compositions of the enclaves (Fig. 5), a topic that will be addressed in the section on melt loss below.

Enclave YTSB-14B (from enclave swarm) displays the simplest history. Plagioclase phenocryst cores are slightly more calcic than typical eHD plagioclase, but enclave phenocryst and groundmass plagioclase compositions generally overlap with each other and with typical eHD plagioclase. With one exception, all Hbl grains display relatively high Ni and low Mn contents compared to typical eHD Hbl. The exception is the core zone of a phenocryst in which Ni contents are intermediate between other enclave Hbl and typical eHD Hbl. This Hbl core presumably represents mixing with a felsic magma followed by crystallization of a rim that has the same composition as other Hbl in the enclave. Thus, the bulk of Hbl and plagioclase in this enclave could represent direct crystallization from an intermediate-composition enclave magma followed by mingling with typical eHD magma and incorporation of low-Ni Hbl grains.

The double enclave TCB-8 provides an example of arrested hybridization. Plagioclase and Hbl in the inner zone of this enclave are distinct in their trace-element trends (plagioclase) and compositions (Hbl) compared to crystals in the outer zone or the host. In contrast, plagioclase phenocrysts in the outer zone are nearly identical to the host plagioclase. One outer-zone Hbl phenocryst is similar to host Hbl, whereas the remaining phenocrysts and groundmass Hbl in the outer zone are intermediate between the inner zone and host Hbl. We suggest that this enclave represents hybridization of a magma similar to the inner zone and carrying higher-T Hbl or augite phenocrysts (now clusters of Ni-rich Hbl) and scant plagioclase phenocrysts. Elongate, prismatic Hbl phenocrysts and poikilitic plagioclase in the inner zone crystallized after the enclave magma came into contact with its host. The outer zone represents a hybrid of the inner-zone magma with the host magma. Outer-zone hornblende crystallized during or after this hybridization, explaining its intermediate Ni contents (Fig. 11) and the presence of plagioclase phenocrysts inherited from the host.

In enclave TLM-53 (from enclave-rich dike), sodic plagioclase phenocryst cores and calcic mantles are distinct from plagioclase in the dike matrix (see especially Fig. 8D). This feature indicates that the andesitic-composition enclave magma was mixed with a rhyolitic magma from which relatively sodic plagioclase cores were inherited, after which calcic mantles crystallized, providing evidence for the andesitic nature of the enclave magma. The Ni-poor nature of Hbl phenocrysts may reflect inheritance from the same rhyolitic magma that provided sodic plagioclase cores, or post-mixing differentiation of the enclave magma prior to mingling into the enclave dike. This now-hybrid enclave was then incorporated into the dike magma, from which it inherited additional, large Hbl grains.

Enclave TLM-61B (isolated) contains plagioclase phenocrysts with cores that indicate crystallization from, and fractionation in, an intermediate magma, followed by mixing with magma more mafic than the original, and then by further differentiation. This already hybrid magma then mixed with a plagioclase- and Hbl-bearing felsic magma. This felsic end member was evidently more evolved than typical eHD magma, as indicated by the low Ni contents in Hbl. The Ni content of Hbl in the immediate host to this enclave is intermediate between Hbl in the enclave and Hbl in typical eHD, suggesting that the local host to the enclave is distinct from typical eHD magma. We note here that enclaves TLM-61B and TCB-8 were collected less than 30 m from one another, yet have very distinct plagioclase zoning trends and Hbl compositions (see below).

Two types of plagioclase phenocryst occur in enclave TCB-7 (isolated), both of which have sodic cores, indicating crystallization in felsic magmas. Cores of the first, larger phenocryst type (grains 2 and 5, Fig. 6B) contain augite and biotite inclusions. Crystals of the second phenocryst type are smaller, contain Hbl inclusions, and display calcic mantles, which were presumably formed after the cores were engulfed by enclave magma. The larger plagioclase grains (2 and 5) contain higher Sr contents than plagioclase in the adjacent host plagioclase with similar An contents (Fig. 6B). In addition, plagioclase in the adjacent host essentially lacks inclusions of any type. The rims of Hbl phenocrysts and groundmass Hbl contain higher Ni abundances than Hbl cores. This feature indicates growth of rims and groundmass from a hybrid magma. These data suggest that (1) Hbl phenocryst cores and the smaller plagioclase phenocryst cores were derived from a felsic magma that mixed with an andesitic enclave magma, after which the plagioclase was mantled by calcic plagioclase and Ni-richer rims grew on the hornblende. This enclave then mixed or mingled with a second rhyolitic magma, from which the large, sodic phenocrysts with augite and biotite inclusions (grains 2 and 5) were inherited. Finally, the enclave was engulfed in the present-day host magma, which is distinct from both of the felsic magmas involved in the prior mixing events.

Each of these five enclaves displays similar but distinct evolutionary histories, despite gross similarities in bulk composition and mineral assemblage. The similarity is the fact that mixing took place between intermediate (rather than mafic) magma and one or more felsic magmas. The distinctions lie in the differences in crystal assemblage, crystal proportions, and perhaps temperature between specific intermediate and felsic end members. For example, neither plagioclase nor Hbl phenocrysts share common trace-element compositions or uniform trends. Moreover, some felsic end-member magmas contained only plagioclase phenocrysts, whereas others contained plagioclase and Hbl (±biotite ± titanite) phenocrysts. Similarly, some intermediate magmas contained plagioclase phenocrysts, and at least one, the inner zone of TCB-8, contained Hbl phenocrysts. Thus, each enclave is the product of mixing of intermediate magmas with, in each case, somewhat distinct rhyolitic magmas. In addition, much of this mixing occurred prior to incorporation of the enclaves into their ultimate host eHD magma.

These observations highlight an unexpected outcome of this study that is illustrated in the Hbl data for enclaves TLM-61B, TCB-7, and TCB-8, and their immediately adjacent host rocks. We anticipated that local host-rock Hbl, analyzed adjacent to the enclave (in the same thin section) would overlap the compositional trend of regional eHD Hbl. However, in all three enclaves noted above, the local host Hbl is distinct from the regional host Hbl (Figs. 10–12; Supplemental File 10 [^{footnote 1}]). One possible explanation is that element diffusion between enclave and host modified the compositions of host Hbl in proximity to the enclave. We are inclined to discount this explanation because of the slow diffusivities of elements such as Ce, Y, and Nb (e.g., Supplemental File 10). A second possibility is that residual melts in the enclaves and their hosts mixed in close proximity to the enclave. Such mixing of residual enclave melts with the host could be accommodated by expulsion of interstitial enclave melts, for example by gas filter pressing (Anderson et al., 1984) as illustrated by Bacon (1986). However, compositional trends of local host Hbl do not define simple mixing lines, for example in terms of Ce and Y (Supplemental File 10), although simple mixing might not be expected in the case of expelled residual melt from the enclave mixing with adjacent host melt. A third alternative is that the local host to these enclaves actually “belongs” to the enclave, such that the obvious, dark enclave is partly or completely surrounded by a host that was attached to the enclave before or during transport to the site of emplacement. In this process, enclaves would commonly form in pods or swarms in which mixing would modify the local host magma composition and lead to an envelope of magma that is distinct from the average host magmas. Once a swarm is disrupted into individual enclaves, most evidence of this early stage would be lost. Resolution of this question awaits further field and laboratory study; however, the abundance of enclave-rich dikes in the TIC may provide some insight: disruption and/or mingling of an enclave-rich dike into “typical” granodioritic magma could result in a mesoscopically obvious mme surrounded by a much less obvious rind from the dike matrix.

If the third scenario is correct, then the enclaves must have formed in similar, yet distinct, rhyolitic magma bodies. Such bodies could represent truly isolated melt-rich pods in a more voluminous magma mush, or they could represent zones of rejuvenated mush with slightly different compositions and crystallinity, depending on the degree of re-melting. The fact that the rhyolitic end members were distinct from those of the typical eHD suggests that the exposed eHD rocks represent an amalgamation and hybridization (in space and time) of smaller magma reservoirs in which enclave formation occurred. The locations of these individual magma bodies are unknown. However, assuming that emplacement of eHD magmas was time-transgressive, it seems logical that these individual bodies existed deeper in the magma system.

The presence of Hbl phenocrysts distinct from Hbl in the ultimate host indicates that hybridization of some enclaves occurred between enclave magmas that were Hbl-bearing, but typically plagioclase-free, with plagioclase-bearing rhyolitic magmas (e.g., TCB-8). In such instances, the maximum temperature (T) of mixing should be indicated by T of crystallization of Hbl phenocrysts. The highest Hbl crystallization temperatures are ∼830 °C (Fig. 9; using Equation 5 of Putirka, 2016). It is noteworthy that the inner zone of enclave TCB-8 contains clusters of equigranular Hbl (Fig. 4D), which probably represent recrystallized relicts of higher-T amphibole or augite. Similar Hbl clusters are absent in the other enclaves, however.

Taken together, the textural and chemical data indicate the following: (1) at the time existing minerals crystallized, the enclave magmas were intermediate, not mafic, in composition, as indicated by compositions of calcic mantles on plagioclase cores, Hbl compositions, and calculated melt compositions. (2) Hbl phenocrysts were probably common in the enclave magmas, whereas plagioclase phenocrysts were less so, suggesting H₂O-rich enclave magmas. (3) The felsic end members were rhyolitic, and all contained plagioclase crystals (∼An₄₀). Some rhyolitic magmas also contained crystals of Hbl ± biotite ± titanite. (4) The rhyolitic end members were not uniform in chemical composition, crystal assemblages, or inclusion assemblages in plagioclase. (5) Some enclaves lack evidence of mixing with their present host, whereas others contain phenocrysts identical to the local host. (6) In the enclave with calcic plagioclase cores, zoning indicates that the enclave magma underwent mixing with other intermediate magmas prior to mixing with rhyolite. (7) Hbl in the immediately adjacent host to the enclaves is commonly distinct from Hbl in typical eHD granodiorite, suggesting that some enclaves may consist of the obvious dark enclave embedded in a granodioritic rind.

Major-element compositions of enclaves from three volcanic systems are plotted in Figure 5: Lassen volcano (Scruggs and Putirka, 2018), Soufrière Hills volcano (Plail et al., 2018), and the Oruanui eruptions of the Taupo system (Allan et al., 2017; Rooyakkers et al., 2018). Enclaves from the Lassen and Soufrière systems display major-element trends that are collinear with their host magma compositions (Fig. 5). In contrast, two distinct groups of mme are present in the Oruanui system—a calc-alkaline group with higher Mg/(Mg + Fe) and Cr and a tholeiitic group with higher FeO and TiO₂ (Rooyakkers et al., 2018). Enclave compositions from these systems plot in tightly constrained arrays in terms of the alkalis, Sr, and Ba (Fig. 5). In contrast, bulk compositions of mafic enclaves in the TIC display a wide range of major- and trace-element contents, particularly with regard to the alkalis, Sr, and Ba, with no discernable trend (Fig. 5).

Further comparisons with mafic enclaves from the Oruanui eruption are particularly instructive because the host magma to Oruanui enclaves is rhyolitic, the enclaves are variably porphyritic—and therefore some may reflect unmodified or weakly modified melt compositions—yet the enclaves display a wider compositional array than enclaves from Lassen or Soufrière. Moreover, although the Oruanui enclave magmas were mingled with rhyolite, no more than a few years before eruption (Allan et al., 2017), the interiors of many of the Oruanui enclaves contain variable proportions of large crystals and pale glass inherited from a rhyolitic host magma (Rooyakkers et al., 2018).

Since the Oruanui mme overlap TIC mme in SiO₂ contents (Fig. 5), one might expect other element compositions and trends to show similar behavior between the Oruanui and TIC enclaves. However, Na₂O, K₂O, Sr, and Ba concentrations in the Oruanui mme are considerably lower than in TIC mme, and in the Oruanui example, these elements plot in discrete arrays that are collinear with their rhyolitic host (Fig. 5). In contrast, for the same elements, the TIC enclaves are widely scattered and lack a discernable trend (Fig. 5). If the TIC enclaves resulted from mixing and/or mingling of intermediate magmas with rhyolitic host magmas, what explains the high abundances of Na₂O, K₂O, Sr, and Ba relative to typical mme from volcanic suites?

The compositions of groundmass plagioclase and Hbl in eHD enclaves are consistent with crystallization from rhyolitic melts (Fig. 13). However, these rhyolitic melts were probably not equivalent in composition to the host rhyolite. Mafic enclaves in volcanic systems also commonly contain interstitial rhyolitic glass that is distinct from the host rhyolite (e.g., Bacon, 1986; Allan et al., 2017; Rooyakkers et al., 2018; Humphreys et al., 2019). Thus, one explanation for relatively high concentrations of incompatible elements in enclaves is enrichment by fractional crystallization within the enclave magma (e.g., Streck and Grunder, 1999; Johnson and Grunder, 2000). If the original alkali contents of the enclave magmas were collinear with the trend of the TIC host rocks—as is seen in the volcanic suites (Fig. 5), then the high alkali contents of the TIC enclaves could be explained as the result of a combination of fractional crystallization of basaltic enclave magma(s) and mixing with more evolved rhyolitic magma.

Alternative explanations involve exchange of alkalis, Sr, and Ba in particular, between enclave and host. Such enclave-host exchange could have occurred under magmatic conditions, with significant interstitial melt present, or under subsolidus (deuteric) conditions. The nature of possible exchange may be addressed with the following observations: (1) The enclaves are characterized by unaltered or weakly altered Hbl and preservation of concentric zoning in Hbl phenocrysts (Fig. 4A). These features are in contrast to Hbl in the eHD host granodiorite, in which Hbl crystals display complex resorption and replacement by actinolitic amphibole as well as partial alteration to chlorite, epidote, and albite (Werts et al., 2020; but see Challener and Glazner, 2017). These replacement and alteration features are thought to reflect fluid-dominated deuteric alteration (Werts et al., 2020; but see Challener and Glazner, 2017). Thus, if fluid-dominated exchange was responsible for the increased alkalis, Sr, and Ba in the enclaves, we would expect to see similar alteration of magmatic Hbl in the enclaves. (2) Where present in the enclaves, alkali feldspar and biotite are typically poikilitic, and in some samples, titanite is also poikilitic. These habits suggest crystallization from a melt (e.g., Holness and Sawyer, 2008), which in turn suggests that the high abundances of alkalis, Sr, and Ba reflect exchange of interstitial melt in the enclave with adjacent rhyolitic melt, and not hydrothermal/deuteric alteration. Detailed comparison of biotite and alkali feldspar compositions in enclaves and host will be necessary to determine whether the rhyolitic melt was introduced during initial mixing events or insertion of the enclaves into their final host magma.

Melt-dominated exchange may occur by diffusion of elements in the melt (e.g., Cramer and Kwak, 1988; Baker, 1991) or by advection of melt through the enclave magma (e.g., Eberz and Nicholls, 1988; Petrelli et al., 2006), or both. Either process may act to homogenize enclaves over relatively short time scales compared to the longevity of the granodioritic host (Baker, 1991; Petrelli et al., 2006).

In the case of advection, groundmass phases and phenocryst rims should be similar in composition to the same minerals in the local host magma, for both rapidly and slowly diffusing elements. This is the case for plagioclase, in which slow (Ce) and fast (Sr, Ba) diffusing element trends merge at low CaO contents (Figs. 6–8). Evidently, the enclave magmas were capable of precipitating thin, relatively calcic mantles on plagioclase grains inherited from the local rhyolitic host, but with the exception of TLM-61B, calcic groundmass plagioclase is absent. This lack of continuous zoning from calcic mantles to sodic rims and/or groundmass implies that after mixing, the melt phase in many of the enclaves was rhyolitic.

However, we reach the opposite conclusion on the basis of groundmass Hbl compositions, which are commonly distinct from Hbl in both the local host and the regional eHD host (e.g., Fig. 11 and Supplemental File 10, Fig. SA [^{footnote 1}]), with the exception of Sr (Supplemental File 10, Fig. SB). One possible explanation for this discrepancy is that groundmass Hbl crystallized during hybridization. We propose that initial mixing of enclave and rhyolitic magmas was capable of precipitating Ca-rich mantles on sodic plagioclase inherited from the rhyolite. Continued mixing combined with crystallization, particularly of Hbl, quickly led to a rhyolitic or rhyodacitic melt composition. It was from this melt that the bulk of the Hbl and groundmass plagioclase crystallized, explaining the low temperatures of crystallization of Hbl (Fig. 9B), the nearly acicular habits of many groundmass Hbl grains (i.e., compositional undercooling), and the slightly higher Mg/(Mg + Fe) values of enclave Hbl compared to typical eHD Hbl (Fig. 9A).

Although this model suggests significant advective exchange of melt, it cannot rule out diffusive exchange between the host and interstitial enclave melt. Such exchange would explain the poikilitic alkali feldspar and biotite that are common in these enclaves, and therefore their high K₂O and Ba contents.

In the preceding discussion, the case is made that the mme magmas were andesitic at the time mixing occurred, despite the SiO₂ contents of the enclaves (50.4–53.3 wt%), which indicate mainly basaltic compositions (Fig. 5). The conclusion that enclave magmas were andesitic is supported by the fact that sodic plagioclase inherited from the rhyolitic mixing end members is mantled by intermediate plagioclase rather than calcic plagioclase (Fig. 6) and by the low abundances of compatible elements such as Ni and Cr in the bulk rock (Fig. 5) and Hbl (Fig. 11). If the enclave mineral compositions reflect hybridization between andesitic and rhyolitic end-member magmas, then the bulk-rock SiO₂ contents should be higher than observed. One explanation for the discrepancy between basaltic bulk-rock compositions and calculated andesite compositions is that the enclaves lost a rhyolitic melt after hybridization. There is geochemical evidence that many rocks in the TIC lost as much as 30% melt (Barnes et al., 2019), although these authors did not evaluate potential melt loss in enclaves. That said, the discrepancy between maximum Hbl crystallization temperatures compared to zircon saturation temperatures (see above) combined with the presence of zircon inclusions in Hbl is readily explained if the enclave magmas lost melt (Barnes et al., 2019).

The potential that enclaves also lost melt was tested with a simple calculation illustrated in Figure 14. The dashed line is the mixing line between the bulk composition of enclave TCB-7 and a rhyolitic melt. The melt composition was one calculated to be in equilibrium with Hbl in the enclave (Zhang et al., 2017). We then estimated the melt SiO₂ and CaO contents necessary to crystallize the most calcic plagioclase mantle composition observed in the enclave samples (An₇₀; cf. Scruggs and Putirka, 2018), which resulted in melt SiO₂ of ∼60–62 wt% and melt CaO of ∼5.2 wt%. According to the simple mixing line (Fig. 14), these values imply that the enclave bulk composition represents as much as 50% melt loss. This amount of melt loss is considerably higher than previous results (Barnes et al., 2019). We therefore suggest that the melt from which An₇₀ plagioclase mantles crystallized were themselves hybrids. For example, in volcanic mme, crystals derived from the host magma are commonly jacketed by rhyolitic glass similar in composition to the host (e.g., Bacon, 1986; Rooyakkers et al., 2018). During slow cooling, such rhyolitic jackets could hybridize with the enclave melt prior to crystallization of plagioclase mantles. We therefore conclude that many enclaves lost melt, but that the proportion of melt loss was probably comparable to melt loss calculated for the adjacent host magmas of ∼0%–30% (Barnes et al., 2019).

Expulsion of enclave melts into the immediately adjacent magma could explain the slight, but observable, differences in Hbl and plagioclase composition in host rocks in contact with the enclaves, compared to Hbl and plagioclase compositions in “normal” eHD samples (Figs. 6–9 and 11). One mechanism of melt loss could be expulsion by formation of a fluid phase (Anderson et al., 1984; Bacon, 1986; Sisson and Bacon, 1999; Bachmann and Bergantz, 2008). Alternatively, melt loss could be related to deformation during magma migration (Holness, 2018), to successive emplacement of new magma batches, or to postemplacement tectonic deformation (Bea et al., 2005). Alignment of groundmass phases in many enclaves supports the idea of syn-magmatic deformation as one cause of melt loss. However, none of these mechanisms can be adequately addressed with the small number of samples reported on here.

• (1) Although mineral assemblages in mme are generally identical to those in the adjacent host rocks, and although major-element compositions of these minerals are similar, trace-element abundances and trends in enclave minerals record complex, individual histories of mixing.

• (2) Most enclaves contain relatively sodic plagioclase phenocryst cores with calcic mantles, which represent one or more hybridization events (Fig. 15). The fact that mantling plagioclase is generally more sodic than An₇₀ suggests that the “mafic” (enclave) end members were andesitic, not basaltic. This conclusion does not negate ultimate basaltic parentage of enclave magmas; it simply acknowledges the lack of evidence for such parentage in the mineral and chemical features of the enclaves.

• (3) The plagioclase cores are not identical from one enclave to the next, and they are generally not identical to typical eHD plagioclase. We suggest that many enclaves formed by mixing with distinct, possibly isolated rhyolitic reservoirs, and only after this mixing, were these disparate bodies combined to form the present eHD unit.

• (4) Nevertheless, some enclaves contain phenocrysts identical in composition and zoning to typical eHD minerals, which indicates that at least some mme were capable of mixing during transport to, and potentially at, the level of emplacement. Moreover, many mme lost a melt phase during or after transport. These conclusions indicate that, at least in the eHD, the enclaves were not solids (e.g., Farner et al., 2014).

• (5) All of the porphyritic enclaves studied are hybrids, and most or all of them lost melt. Therefore, in agreement with many other authors, the assumption that bulk-rock compositions of these enclaves represent a common magma composition (e.g., parental magma) is unjustified.

We are pleased to dedicate this paper to Calvin Miller. Calvin’s work, along with his many students and colleagues, highlighted the importance of detailed investigations of igneous systems at every scale, from the field to mineral micro-analysis. We thank Ethan Backus for assistance in the field, Louis Oppenheim for providing sample YTSB-14B, and Guil Gualda and Shan de Silva for editorial handling and comments. Anita Grunder and Bill Collins are thanked for their thoughtful, insightful journal reviews. The research was supported by National Science Foundation grants EAR-1550969 (Barnes) and EAR-1550935 and EAR-1624854 (Memeti).