Trace Element Geochemistry of Li-Rich Pegmatites in the Carolina Tin-Spodumene Belt, North Carolina, USA: Implications for Petrogenesis and Exploration

Lithium is a critical metal necessary for the transition from a fossil fuel-based energy economy to the renewable energy-based economy of the future (Bradley et al., 2017). Currently, granitic pegmatites account for the majority of worldwide Li production, with brines being the other major Li resource (U.S. Geological Survey, 2024). Pegmatites will continue to be an important Li resource because of their higher ore grade and wider geographic distribution than brine deposits (Kesler et al., 2012; Bradley et al., 2017). Understanding the petrogenesis of Li-rich pegmatites, especially their mineral chemistry, and applying such knowledge to mineral exploration are necessary to ensure that a sustainable Li supply is available that can meet the anticipated large Li demand as society moves toward an energy economy reliant on Li batteries.

Conventional hypotheses for the petrogenesis of pegmatites enriched in rare elements include fractional crystallization from a parental granite and anatexis of crustal rocks (Černý, 1991a; Simmons et al., 1995; Müller et al., 2017). The parental granite hypothesis requires high degrees of fractional crystallization from a parental granitic melt that, in the case of Li-rich pegmatites, usually has an aluminous, S-type (i.e., a sedimentary source) character (Černý et al., 2012b). We refer to this hypothesis as the residual melts of granite magmatism (RMG) after Wise et al. (2022a). The anatexis hypothesis involves direct melting of a metaigneous or metasedimentary rock after which fractional crystallization, although still occurring, is not the main driver of elevated rare element concentrations (e.g., Müller et al., 2017). We refer to this hypothesis as the direct products of anatexis (DPA) after Wise et al. (2022a). Recent studies have expanded on these ideas with (1) a version of the RMG hypothesis that involves the magmatic volatile phase of an extremely fractionated melt forming the pegmatites instead of the fractionated melt (Troch et al., 2022) and (2) a two-stage version of the DPA hypothesis that involves first crystallizing melted metasediments into a granite and then melting that granite to form an enriched pegmatite (Koopmans et al., 2023). In all petrogenetic models for Li-rich pegmatites, the melting of metasedimentary rocks is involved to produce peraluminous liquids in which the melting of biotite, muscovite, cordierite, and staurolite releases the Li that is eventually concentrated to sufficiently high concentrations to form the Li phases spodumene, petalite, and/or lepidolite in pegmatites (Stewart, 1978; Dutrow et al., 1986; Černý et al., 2012b; Kunz et al., 2022; Koopmans et al., 2023).

In addition to recent studies considering petrogenetic mechanisms, a new classification system for pegmatites was recently proposed (Wise et al., 2022a) that addressed needs in pegmatite research not met by the classic pegmatite classification system (Černý, 1990, 1991a, b; Černý and Ercit, 2005). In the new classification system of Wise et al. (2022a), group 1 pegmatites correspond to the lithium-cesium-tantalum (LCT) pegmatite family, and group 2 pegmatites correspond to the niobium-yttrium-fluorine (NYF) pegmatite family of Černý and Ercit (2005). Group 3 pegmatites in the new classification do not relate to a family in the Černý and Ercit (2005) schema, because Al-rich group 3 pegmatites are formed only by a DPA mechanism, in contrast to groups 1 and 2, both of which can be produced by either a DPA or an RMG mechanism. In this study, we follow the new Wise et al. (2022a) classification system.

The Li-rich pegmatites in the Carolina tin-spodumene belt provide an ideal natural laboratory in which to understand the petrogenesis and geochemical evolution of group 1 pegmatites, and renewed interest in spodumene mining in the Carolina tin-spodumene belt makes this a timely study relevant to Li exploration and mining efforts in North Carolina and elsewhere. The Carolina tin-spodumene belt pegmatites were originally mined for Sn from cassiterite (SnO₂) in the late 19^th century, but Li from spodumene (LiAlSi₂O₆) proved to be the major resource with historical Li mining operations occurring from the 1950s to 1990s (Graton, 1906; Horton and Butler, 1981; Swanson, 2012). Economically, the Carolina tin-spodumene belt group 1 pegmatites are on par with other large Li pegmatite ore deposits around the world, with estimates of 0.32 to 0.42 million metric tonnes (Mt) of Li in two historical mines that are on the scale of the Greenbushes (Australia; 0.85 Mt Li), Jiajika (China; 0.48 Mt), and Manono-Kitolo (Democratic Republic of the Congo; 0.33 Mt) Li-rich pegmatites, among others (Kesler et al., 2012). Despite the Carolina tin-spodumene belt pegmatites being historically mined and studied for over a century, no detailed examination of their trace element geochemistry has been previously undertaken, and the petrogenetic question of how these pegmatites formed remains unclear. In this study, we use whole-rock and mineral trace element geochemistry to understand the petrogenesis and geochemical evolution of pegmatites in the Carolina tin-spodumene belt.

The Carolina tin-spodumene belt is a 3- to 5-km-wide belt of pegmatites hosted in amphibolite and mica schist country rocks that extends along a northeasterly trend approximately 40 km from the South Carolina border to near Lincolnton, North Carolina (Fig. 1; Hess, 1940; Kesler, 1942; Goldsmith et al., 1988; Horton, 2008). The belt is positioned on the west side of the Kings Mountain shear zone, a structural lineament that marks the boundary between the Laurentian/peri-Gondwanan Cat Square terrane of the Appalachian Inner Piedmont on the west and the peri-Gondwanan Kings Mountain sequence of the Carolina superterrane on the east (Horton, 1981; Horton and Butler, 1986; Hatcher, 2002; Hibbard et al., 2002, 2006; Hatcher et al., 2007; Merschat and Hatcher, 2007). The Kings Mountain sequence is the western edge of the Charlotte terrane and consists of Neoproterozoic metasedimentary and metavolcanic rocks intruded by the High Shoals and Churchland granitic plutons of Pennsylvanian age (Goldsmith et al., 1988; Hibbard et al., 2006; Horton, 2008). The Siluro-Devonian Cat Square terrane is the eastern edge of the Inner Piedmont and consists of predominantly metasedimentary and subordinate metaigneous rocks intruded by several Devonian and Mississippian granitoids including the Cherryville Granite (Hibbard et al., 2006; Merschat and Hatcher, 2007). The Cherryville Granite is a two-mica granite that occurs within 5 km to the west of the Carolina tin-spodumene belt (Fig. 1; Griffitts and Overstreet, 1952).

The Carolina tin-spodumene belt contains both fairly homogeneous, poorly zoned spodumene-free and group 1 spodumene-bearing pegmatites with spatial dimensions of 1 m to tens of meters wide and tens to hundreds of meters long (Hess, 1940; Kesler, 1942; Hodges, 1983). Most pegmatites in the Carolina tin-spodumene belt, either with or without spodumene, strike approximately northeast along the regional metamorphic wall-rock foliation trend, with dips varying between southeast, northwest, and horizontal (Kesler, 1942, 1961; Hodges, 1983; Swanson, 2012). Some Carolina tin-spodumene belt pegmatites strike northwest with northeastern dips that crosscut regional structural trends (Kesler, 1942; Hodges, 1983). Spodumene-free pegmatites are variably folded and sheared, whereas spodumene-bearing pegmatites are generally not deformed with the exception of shearing in some pegmatites (Swanson, 2012). Swanson (2012) distinguished between two families of spodumene-free pegmatite dikes and one family of spodumene-bearing dikes: (1) folded and sheared pretectonic dikes, without spodumene, (2) spodumene-bearing dikes without folding but sometimes sheared, and (3) post-tectonic dikes without spodumene that lack obvious deformation.

The main rock-forming minerals besides spodumene, when present, are plagioclase, quartz, K-feldspar, and muscovite, with the latter two subordinate to quartz and plagioclase, and plagioclase is dominantly albite, with a few recorded instances of oligoclase (Kesler, 1976; White, 1981; Hodges, 1983; Swanson, 2012). Group 1 spodumene-bearing pegmatites from the Carolina tin-spodumene belt are categorized as the albite-spodumene type of the rare element class of Černý (1990). Kesler (1961) estimated that the spodumene-bearing pegmatites contained 32% quartz, 27% albite, 20% spodumene, 14% K-feldspar, 6% muscovite, and 1% accessories based on average mine yields over several months at the historical Kings Mountain mine. As would be expected for pegmatites, modal proportions vary greatly depending on locality, with Griffitts (1954) estimating between 5 and 35% spodumene for different pegmatites throughout the Carolina tin-spodumene belt, and Hodges (1983) estimating 20 to 30% spodumene, 15 to 40% microcline, 5 to 35% albite, 5 to 10% muscovite, and 20 to 25% quartz at the historical Hallman-Beam mine. Accessory minerals in spodumene-bearing pegmatites include beryl, garnet, cassiterite, Mn-bearing fluorapatite, triphylite, columbite-group minerals, schorl, and zircon (White, 1981; Hodges, 1983; Swanson, 2012). Spodumene-free pegmatites, likely because of their lack of spodumene, have received less attention to detail in terms of accessory mineralogy, but Hodges (1983) observed garnet, apatite, tourmaline, and beryl in spodumene-free dikes.

Pegmatites of the Carolina tin-spodumene belt typically have a medium-grained to very coarse grained texture, but they lack the abundant meter-scale spodumene crystals present in other group 1 pegmatites (e.g., Etta, South Dakota, and Plumbago North, Maine; Rickwood, 1981; Simmons et al., 2020). Among the main rock-forming minerals, albite, K-feldspar, quartz, and muscovite are usually medium to coarse grained, and spodumene crystals have long dimensions typically less than 30 cm and rarely up to 1 m (Kesler, 1961; Hodges, 1983; Swanson, 2012). Spodumene crystals occur randomly oriented as well as perpendicular to the contact. In addition to the pegmatitic texture, albite and quartz constitute a fine-grained, saccharoidal texture that replaces the pegmatite fabric and infills fractures in larger minerals (Kesler, 1961; Swanson, 2012). Fine-grained muscovite also occurs as a replacement mineral healing fractures or replacing minerals (Swanson, 2012). Aplitic zones sometimes occur along the margins of spodumene-bearing pegmatites, with mineralogy similar to that of the spodumene-bearing pegmatites, except with less K-feldspar and variations between spodumene-poor and spodumene-rich zones (Hodges, 1983; Swanson, 2012).

The petrogenesis of Carolina tin-spodumene belt pegmatites is uncertain, with some studies proposing an RMG mechanism for their formation, with the nearby Cherryville Granite as the parental granite, and others arguing for formation via a DPA petrogenesis (Hodges, 1983; Kish and Fullagar, 1996; Swanson, 2012). The presence of the Cherryville Granite, a peraluminous, two-mica granite, just to the west of the Carolina tin-spodumene belt (Fig. 1), along with its mineralogical and geochemical similarities to Carolina tin-spodumene belt pegmatites, led some researchers to hypothesize that Carolina tin-spodumene belt pegmatites were either directly related to the Cherryville Granite or that they shared the same source (Kesler, 1942; Griffitts and Overstreet, 1952). Thus, early researchers noted the possibility that the Carolina tin-spodumene belt pegmatites could be derived either by an RMG process via the Cherryville Granite or by a DPA process because the peraluminous nature of both rocks indicated the involvement of crustal rock anatexis. Previous data point either weakly toward one hypothesis or can be used as evidence for either hypothesis.

In terms of physical evidence in the field, early studies reported spodumene-free pegmatites in and around the Cherryville Granite and spodumene-bearing pegmatites near the Cherryville Granite (Kesler, 1942, 1961; Griffitts and Overstreet, 1952). In the Kings Mountain mine, Kesler (1961) mapped spodumene-free and spodumene-bearing pegmatites in direct contact with each other, suggesting a nongradational contact ~60 m from the Cherryville Granite. A later investigation of the Hallman-Beam mine suggested that a garnet-bearing granite in the Hallman-Beam mine (“mine granite”) was genetically related to the Cherryville Granite, and this mine granite graded into garnet-bearing, spodumene-free pegmatites, which were spatially associated with spodumene pegmatites in the Hallman-Beam mine (Hodges, 1983). However, the gradational contact between the Hallman-Beam mine granite and spodumene-free pegmatites was only inferred and could not be traced in outcrop, and the mine granite was not shown to be definitively related to the Cherryville Granite (Hodges, 1983). The most recent detailed study on the Carolina tin-spodumene belt pegmatites by Swanson (2012) had access to both the Hallman-Beam mine and Kings Mountain mine and did not observe gradational contacts between spodumene and spodumene-free pegmatites. Based on this, field evidence seems to weakly suggest that the Carolina tin-spodumene belt pegmatites formed through an RMG process with the Cherryville Granite (or other granite) as the parent, but this evidence is inconclusive (Kesler, 1961; Hodges, 1983; Swanson, 2012).

The main line of evidence supporting a DPA origin for the Carolina tin-spodumene belt pegmatites is the low Li content of the Cherryville Granite (20–40 ppm), which authors argued would be too low to produce Li-rich, spodumene-bearing pegmatites (Kish, 1983; Kish and Fullagar, 1996). Existing whole-rock Rb-Sr geochronology of the Cherryville Granite and pegmatites yielded Mississippian ages of 351 ± 10 Ma for the pegmatites and 340 ± 5 Ma for the Cherryville Granite that can be taken to support either the RMG or DPA interpretation (Kish, 1983; Kish and Fullagar, 1996). High initial Sr isotope ratios for the Cherryville Granite (0.7121 ± 0.0016) and the pegmatites (0.7326 ± 0.0184) and high O isotope ratios for the Cherryville Granite (δ¹⁸O = 10‰) support a crustal anatexis origin for both rocks (Wenner, 1981; Kish, 1983; Kish and Fullagar, 1996). As with the existing geochronology, Sr and O isotope data leave room for either a DPA or an RMG petrogenesis. Both the Cherryville Granite and the Carolina tin-spodumene belt pegmatites could have formed through separate but similar DPA processes, or the Carolina tin-spodumene belt pegmatites could have formed through an RMG process of fractional crystallization of a Cherryville Granite melt by which the pegmatites inherit the anatectic isotopic signature of their parental magma. All previous evidence and data have not, therefore, resolved the initial observations of early researchers that either hypothesis could be valid (e.g., Kesler, 1942).

A total of 96 samples were collected from outcrop and drill core across the Carolina tin-spodumene belt (Fig. 1). This includes a sample of the Cherryville Granite, 87 pegmatite samples, and eight amphibolite wall-rock samples. Feldspar, quartz, muscovite, garnet, apatite, and spodumene were handpicked for geochemical analysis by electron probe microanalysis (EPMA) and laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS). Our suite of pegmatites included 32 spodumene-bearing samples and 55 spodumene-free samples. Detailed sample information is provided in Appendix Table A1. In addition, muscovite, K-feldspar, and spodumene from group 1 pegmatites elsewhere in the United States and around the world were analyzed in this study for comparison to Carolina tin-spodumene belt minerals (Table 1). Muscovites and K-feldspars from the Custer and Keystone pegmatite fields in South Dakota and the Oxford County pegmatite field in Maine were also analyzed. In the Custer field, we analyzed minerals from the Buster, Old Mike, Silver Dollar, Tin Mountain, and Tip Top pegmatites, and we analyzed minerals from the Dan Patch, Etta, Gap Lode, and Hugo pegmatites from the Keystone field. In the Oxford County field, we analyzed minerals from the BB#7, Bumpus, Bennett, Black Mountain, Cole, Emmons, GE, Hayes, Hibbs, Lord Hill, Mt. Marie, Mt. Mica, Plumbago North, Tamminen, Waisanen, and Willis Warren pegmatites. Among these mineral samples, we analyzed (1) two K-feldspar samples (Etta-NMNH164694 and Tip Top-NMNH150093) and one muscovite sample (Tin Mountain-NMNH97655) from the mineralogy collection of the National Museum of Natural History (NMNH), Smithsonian Institution, and (2) eight muscovites of the E. Wm. Heinrich research collection from six different pegmatites (Buster, Dan Patch, Gap, Hugo, Old Mike, and Silver Dollar). Spodumenes were analyzed from Oxford County, Etta, Lacorne (Canada), and Barroso-Alvão (Portugal) pegmatites. Muscovite, K-feldspar, and spodumene samples from outside the Carolina tin-spodumene belt that were not from the NMNH or the Heinrich collections are part of the research collection of coauthor Wise or provided by C. Fabre (University of Lorraine). Analytical results are presented with ±1σ standard deviations unless otherwise noted.

A portion of each rock sample approximately 15 cm in diameter was crushed to gravel size. Representative pegmatite whole-rock compositions are difficult to obtain because of their large grain sizes and the near impossibility of acquiring a homogeneous sample. To minimize this grain size effect, the very coarse grained parts of pegmatite samples were avoided for whole-rock analysis, instead using medium-grained pegmatite samples when possible. Crushed rocks were sent to American Assay Laboratories for major, minor, and trace element analysis. Because of contamination by the W mill during powdering, W and Co were omitted from the reported results. Loss on ignition was calculated by heating rock powders to 1,000°C. Rock powders were then fluxed with lithium borate prior to being fused into glass beads. Major and minor element compositions were determined using wavelength dispersive X-ray fluorescence (XRF) spectrometry. Prior to ICP-MS analysis for trace elements, hydrofluoric, perchloric, nitric, and hydrochloric acids were used for digestion, and boric acid was used for stabilization. Certified reference materials OREAS 552, 600b, 751, 752, 753, and 906 were run during XRF and ICP-MS analyses for quality assurance. Estimated errors for whole-rock XRF and ICP-MS analyses are <5%.

Minerals were extracted from medium-grained to very coarse grained pegmatite samples using a rock hammer and tweezers and then mounted in epoxy for the mineral chemistry analyses described below. The exception is one thin section of coarse-grained, spodumene-bearing pegmatite sample 22-AC-37, in which garnet, spodumene, and muscovite were analyzed in situ. Garnet is much less abundant in spodumene-bearing samples, so thin section 22-AC-37 was chosen because of the presence of coexisting garnet and spodumene. Care was taken to analyze only larger muscovite, quartz, and plagioclase crystals considered to be primary, but because of the nature of mineral extraction (crushing with rock hammer), some small amount of fine-grained replacement minerals (i.e., fish-scale mica and saccharoidal quartz and albite) might also have been analyzed.

Major and minor element concentrations of spodumene, feldspar, muscovite, apatite, and garnet were determined by EPMA on a JEOL JXA-iHP200F field emission electron probe at the Virginia Polytechnic Institute and State University. An acceleration voltage of 15 kV, a 10-nA probe current, and a 5-µm beam diameter were used for mineral analysis. Calculations for ZAF matrix-effect corrections were performed within the JEOL PC-EPMA program used for data collection. For feldspars and garnet, analyses with oxide mass totals of 100 ± 1.5 wt % were retained. Formula-specific cations were calculated for each mineral analysis as a further check on data quality. The standard minerals diopside (Si, Ca), albite (Al, Na), olivine (Mg), pyrope (Fe), rhodonite (Mn), orthoclase (K), fluorite (F), and apatite (P) were used for EPMA standardization. These standards, minus olivine, were analyzed periodically during each session for quality assurance. Uncertainties were estimated based on the relative standard deviation of these standard analyses, including 60 of albite, 96 of diopside, 59 of orthoclase, 54 of apatite, 52 of fluorite, 47 of rhodonite, and 22 of pyrope. Estimated uncertainties were ≤3% for all analyzed elements except for F (≤4%). Secondary standard analyses, uncertainty calculations, and EPMA configuration are provided as Appendix Table A2.

Analysis by LA-ICP-MS for trace element abundances in spodumene, feldspar, muscovite, quartz, apatite, and garnet was performed at the LionChron facility at Pennsylvania State University. Analyzed minerals, either in thin section or mounted in epoxy, were ablated using a Teledyne/Photon Machines Analyte G2 excimer laser ablation system with a Helex2 ablation cell coupled to a ThermoScientific iCAP-RQ ICP-MS system for trace element abundance determinations. The total Ar gas flow was 1.00 L/min, with total He gas flows from the laser at 0.44 L/min. Laser spot sizes were 50 μm for muscovite, spodumene, and garnet transects in thin section, and those for epoxy-mounted minerals were either 65 μm (quartz), 65 or 85 μm (spodumene), or 85 μm (all others). All samples were run in back-to-back sessions, with a10-Hz repetition rate, 180 to 250 shots, and a laser fluence of ~4 J/cm² at the sample surface, yielding pit depths on the order of ~10 µm. Quartz was run at higher energy, with a 15-Hz repetition rate, a laser fluence at the sample surface of ~10 J/cm², and 180 to 250 shots. The laser was first fired thrice with the same spot size to remove surface contamination, and this material was allowed to wash out for ~15 s. Analyses of unknowns were bracketed by analyses of trace element glass NIST612 (Pearce et al., 1997) and whole-rock glasses from the Max Planck Institute (Jochum et al., 2006) spanning a range of compositions, including Gorgona Island komatiite G132-G, Kilauea basalt KL2-G, Mauna Loa basalt ML3B-G, Alpine quartz diorite T1G, Mt. St. Helens andesite StHs6/80-G, and Icelandic rhyolite ATHO-G. Trace element-doped clinopyroxene glasses CPX666 and CPX777 were also analyzed during the spodumene analysis (Klemme et al., 2008). Standard KL2-G was used as the primary reference material for all analyses on all minerals except for quartz, which was reduced with NIST612 as the primary standard. For trace element quantification, ²⁷Al (using data collected from prior EPMA sessions on the same grains) was used as an internal standard for all minerals except apatite (⁴⁴Ca) and quartz (²⁹Si), with measured peaks on the iCAP-RQ at ⁷Li, ⁹Be, ²³Na, ²⁴Mg, ²⁷Al, ²⁹Si, ³¹P, ⁴⁴Ca, ⁴⁵Sc, ⁴⁹Ti, ⁵¹V, ⁵²Cr, ⁵⁵Mn ⁵⁹Co, ⁶⁰Ni, ⁶³Cu, ⁶⁶Zn, ⁷¹Ga, ⁷³Ge, ⁸⁵Rb, ⁸⁸Sr, ⁸⁹Y, ⁹⁰Zr, ⁹³Nb, ⁹⁵Mo, ¹¹⁵In, ¹¹⁸Sn, ¹²¹Sb, ¹³³Cs, ¹³⁷Ba, ¹³⁹La, ¹⁴⁰Ce, ¹⁴¹Pr, ¹⁴⁶Nd, ¹⁴⁷Sm, ¹⁵³Eu, ¹⁵⁷Gd, ¹⁵⁹Tb, ¹⁶³Dy, ¹⁶⁵Ho, ¹⁶⁶Er, ¹⁶⁹Tm, ¹⁷²Yb, ¹⁷⁵Lu, ¹⁷⁸Hf, ¹⁸¹Ta, ²⁰⁸Pb, ²³²Th, and ²³⁸U. Iolite version 4 was used to correct measured isotopic ratios and elemental intensities for baselines, time-dependent laser-induced interelement fractionation, plasma-induced fractionation, and instrumental drift (Paton et al., 2011). The mean and standard errors of the measured ratios of the backgrounds and peaks were calculated after rejection of outliers more than two standard errors beyond the mean. Using the same methods as applied to unknowns, and treating all whole-rock and clinopyroxene glasses besides KL2G as secondary reference materials, this routine yielded values accurate to <10% for all elements, except (1) Be, Cr, Ni, Co, and Ge, which were accurate to <20%, and (2) Mo, In, and Sn, which were often >30% inaccurate because of their low concentrations.

Prior to presenting our results and discussion, it is important to provide a note about our Carolina tin-spodumene belt pegmatite samples. First, pegmatites were classified as spodumene bearing or spodumene-free based on either macroscopic hand-sample or microscopic petrographic identification of spodumene. Due to their coarse-grained nature, it is possible that samples were collected without spodumene from a spodumene-bearing pegmatite. Because thin sections were not made for all 87 pegmatite samples, it is also possible that some microscopic spodumene was missed. Thus, there is the possibility that some samples classified as spodumene-free could be mislabeled, although this does not affect the main discussion points of this paper.

As previously noted, Carolina tin-spodumene belt pegmatites are high-silica, peraluminous granitic rocks (Fig. 2; Table 2). Spodumene-bearing samples have higher average SiO₂ contents (76 ± 2 wt %) than spodumene-free samples (74 ± 3 wt % SiO₂; Table 2). Molar Al₂O₃/Na₂O + K₂O + CaO (A/CNK) ratios range from 1 to 1.4 for spodumene-free pegmatites and extend up to 7 for spodumene-bearing pegmatites (Fig. 2A). The Cherryville Granite is also silica rich (74 wt % SiO₂) and peraluminous (A/CNK = 1.1), similar to the spodumene-free pegmatites (Fig. 2A; Table 2). Spodumene-bearing pegmatites typically have lower K/Rb ratios, higher Li contents, and lower Al/Ga ratios than spodumene-free pegmatites and the Cherryville Granite (Fig. 2). The K/Rb ratios of spodumene-free pegmatites average 102 ± 44 compared to only 45 ± 13 for the spodumene-bearing pegmatites (Fig. 2B, C). The average Li content of spodumene-free pegmatites is 102 ± 99 ppm compared to 5,096 ± 4,221 ppm for spodumene-bearing pegmatites (Fig. 2B; Table 2). Whole-rock Al/Ga ratios are also distinct for spodumene-free versus spodumene-bearing samples, with averages of 3,855 ± 790 and 2,749 ± 389, respectively (Fig. 2C). The analysis of the Cherryville Granite sample is similar to spodumene-free pegmatites, with K/Rb = 150, Li = 44 ppm, and Al/Ga = 3,695 (Fig. 2). Pegmatites and the Cherryville Granite are notably poor in Zr (1–54 ppm; Table 2).

Pegmatites and the Cherryville Granite are depleted in the rare earth elements (REEs), and all granitic rocks have lower REE contents than wall-rock amphibolite (Fig. 3; Table 2). When normalized to the average upper continental crust of Taylor and Mclennan (1995), the REE contents of pegmatites and the Cherryville Granite are less than the average for upper continental crust (<1 on Fig. 3), whereas the amphibolite analyses are near or slightly above this value. Abundances of REEs in many of the pegmatites were below the analytical detection level (no symbol present on Fig. 3), with this being more prevalent in spodumene-bearing samples, and the Lu level of the Cherryville Granite was below the detection limit. Spodumene-bearing samples, although overlapping with some spodumene-free samples, generally have lower REE concentrations than samples without spodumene (Fig. 3). The light REEs (LREEs) are less abundant than the heavy REEs (HREEs) for all samples, and many pegmatite REE patterns in Figure 3 display negative Ce anomalies and positive Tb anomalies (Fig. 3). Positive and negative Eu anomalies are present in the Carolina tin-spodumene belt pegmatite samples, and many pegmatites have Eu contents below the detection level (Fig. 3). The Cherryville Granite has slightly negative Nd and Eu anomalies, and the amphibolite samples mimic the pegmatites that have positive Tb and Eu anomalies (Fig. 3). Complete whole-rock geochemical results are provided in Appendix Table A3.

Plagioclase: Plagioclase in the Carolina tin-spodumene belt pegmatites is dominantly albite with compositions between Ab₉₁ and Ab₁₀₀ (where Ab = albite), with only two oligoclase analyses of Ab_84–85 recorded. Plagioclase from spodumene-bearing pegmatites is Na-rich albite between An₉₈ and An₉₉ (where An = anorthite), and plagioclase from spodumene-free pegmatites is mostly Na-rich albite, with five analyses <Ab₉₅, in addition to the two oligoclase analyses of Ab_84–85. Plagioclase from spodumene-bearing pegmatites has a lower average Ca content (664 ± 306 ppm), with an interquartile range (IQR) of 414 to 965 ppm, than spodumene-free pegmatites (4,086 ± 5,588 ppm), with an IQR of 879 to 5,823 ppm (Fig. 4A). Concentrations of Rb (0.1–46 ppm) and Li (0.6–118 ppm) are low in plagioclase, if detected at all. Plagioclase Ca and Sr contents, 156 ppm to 2.77 wt % and 4 to 181 ppm, respectively, are also low in the Carolina tin-spodumene belt pegmatites and display a weak positive covariation (Fig. 4A). The Sr content of plagioclase in spodumene-free pegmatites is higher on average (37 ± 38 ppm) than that of spodumene-bearing pegmatites (19 ± 14 ppm) in the Carolina tin-spodumene belt (Fig. 4A).

Plagioclase Eu content shows a strong positive correlation with Ca content, and plagioclase total REE (ΣREE; here considered as the sum of all lanthanides) content shows a weak positive covariation with Ca content (Fig. 4B, C). Plagioclase from spodumene-bearing pegmatites is lower in Eu and ΣREEs than spodumene-free pegmatites (Fig. 4B, C). Plagioclase Eu content ranges from not detected to 0.6 ppm, with only one detection in plagioclase from a spodumene-bearing sample (0.003 ppm Eu), and an average of 0.08 ± 0.13 ppm in spodumene-free pegmatites (Fig. 4B). The ΣREE content of Carolina tin-spodumene belt plagioclase ranges from not detected to 8.6 ppm, with an average of 0.06 ± 0.13 ppm in spodumene-bearing pegmatites and 0.97 ± 1.95 in spodumene-free pegmatites (Fig. 4C). The IQR of ΣREE content in plagioclase from spodumene-bearing pegmatites is 0.01 to 0.06 ppm, and that of plagioclase from spodumene-free pegmatites is 0.04 to 0.76 ppm (Fig. 4C). Plagioclase from the Cherryville Granite sample generally behaves like spodumene-free pegmatites in terms of its Sr, Ca, Eu, and ΣREE contents (Fig. 4). Plagioclase EPMA and LA-ICP-MS results are provided in Appendix Tables A4 and A5, respectively.

K-feldspar: K-feldspar in Carolina tin-spodumene belt pegmatites has slightly sodic compositions of Or₉₃ to Or₉₉ (where Or = orthoclase), and K-feldspar contents in both spodumene-bearing and spodumene-free pegmatites span that range. K-feldspars from Carolina tin-spodumene belt pegmatites have K/Rb ratios ranging from 31 to 197 with Li contents that vary from 3 to 193 ppm and Cs contents ranging from 22 to 238 ppm (Fig. 5A, C). Samples with spodumene are characterized by lower K/Rb ratios and higher Li and Cs contents than samples without spodumene (Fig. 5A, C). Spodumene-bearing samples have K/Rb ratios averaging 52 ± 17 and average Li contents of 47 ± 52 ppm, whereas K/Rb ratios average 138 ± 40 in spodumene-free samples with lower average Li contents of 14 ± 21 ppm (Fig. 5A). The IQR of Li content in K-feldspar is 17 to 47 ppm in spodumene-bearing samples and 6 to 14 ppm in spodumene-free samples (Fig. 5A). The Cs contents of K-feldspar in spodumene-free samples average 49 ± 37 ppm compared to a higher average of 74 ± 54 ppm in spodumene-bearing samples (Fig. 5C). K-feldspars from the Cherryville granite sample have higher K/Rb ratios (193 ± 10) and lower Li (2–4 ppm) and Cs contents (8–9 ppm) than the Carolina tin-spodumene belt pegmatites (Fig. 5A, C). An interesting data gap exists in K-feldspar K/Rb ratios between values of 70 to 100, in which only one analysis plots, and this gap differentiates between spodumene-bearing and spodumene-free pegmatites of the Carolina tin-spodumene belt.

The K/Rb ratios, Li, and Cs contents in K-feldspar from Carolina tin-spodumene belt pegmatite samples are comparable to other group 1 pegmatites from Maine and South Dakota with some exceptions (Fig. 5B, D). The pegmatites with K-feldspar geochemistry outside the range of Carolina tin-spodumene belt pegmatites are Etta and Black Mountain with lower K/Rb ratios and higher Cs contents, Bennett and Tip Top with higher Cs contents but similar K/Rb ratios, and Bumpus with lower Cs contents but similar K/Rb ratios (Fig. 5B, D). Interestingly, the K/Rb data gap also occurs for K-feldspars in pegmatites from Maine and South Dakota (Fig. 5B). K-feldspar EPMA and LA-ICP-MS results are provided in Appendix Tables A6 and A7, respectively.

Quartz: Quartz analyses from Carolina tin-spodumene belt pegmatites yield K/Rb ratios between 22 and 323 with Li contents that range from 0.5 to 330 ppm (Fig. 6A, B). Similar to K-feldspar, quartz from spodumene-bearing samples has lower K/Rb ratios and higher Li contents than quartz from spodumene-free samples (Fig. 6A). Quartz from spodumene-bearing samples has K/Rb ratios averaging 66 ± 44, whereas K/Rb in quartz of spodumene-free samples averages 158 ± 83 (Fig. 6A). The Li content of quartz from spodumene-bearing pegmatites averages 66 ± 58 ppm compared to 21 ± 37 ppm for spodumene-free pegmatites (Fig. 6A, B). In samples with spodumene, the IQR of Li in quartz is 29 to 90 ppm compared to an IQR of 7 to 19 ppm Li in quartz from spodumene-free samples. The IQR of K/Rb ratios in quartz is 30 to 77 in spodumene-bearing samples and 86 to 215 in spodumene-free samples. Quartz in pegmatite samples with spodumene typically has higher Al, with an IQR of 164 to 568 ppm, compared to an IQR of 100 to 172 ppm Al in pegmatites without spodumene (Fig. 6B). Quartz from spodumene-bearing samples also has higher average Rb (1.6 ± 3.1 ppm) than quartz in pegmatites without spodumene (0.6 ± 2.2 ppm; Fig. 6C). Quartz LA-ICP-MS results are provided in Appendix Table A8.

Muscovite: Micas from Carolina tin-spodumene belt pegmatites and the Cherryville Granite plot in the muscovite field of Tischendorf et al. (2007) and contain very little Mg and Li, whereas micas from the Maine and South Dakota pegmatites range from muscovite to lithian muscovite (App. Fig. A1). Muscovites from Carolina tin-spodumene belt pegmatites have K/Rb ratios ranging from 9 to 187 with Li contents varying from 67 to 4,686 ppm (Fig. 7A). Muscovites from Carolina tin-spodumene belt pegmatite samples with spodumene have K/Rb ratios averaging 26 ± 10 and average Li contents of 621 ± 502 ppm. In spodumene-free Carolina tin-spodumene belt pegmatite samples, muscovite K/Rb ratios average 71 ± 40 and Li contents average 421 ± 232 ppm (Fig. 7A). The IQR of muscovite K/Rb ratios in spodumene-bearing samples is 19 to 31 compared to 39 to 94 in spodumene-free samples (Fig. 7A). Muscovites from the Cherryville Granite sample have much less chemical variation, with an average K/Rb ratio of 104 ± 10 and an average Li content of 240 ± 58 ppm (Fig. 7A).

Muscovites from Carolina tin-spodumene belt pegmatites have K/Rb ratios and Li contents similar to the Maine and South Dakota pegmatites with some exceptions (Fig. 7B, C). The Carolina tin-spodumene belt muscovites do not reach Li contents as high as or K/Rb ratios as low as the lithian muscovites of Black Mountain and Tin Mountain that plot in the highly fractionated, high-Li domain of K/Rb < 10 and Li > 500 ppm (Fig. 7B, C). Muscovites of the Willis Warren and Silver Dollar pegmatites extend to lower Li contents than those of Carolina tin-spodumene belt pegmatites, and lithian muscovites of the Plumbago North and Bennett pegmatites reach higher Li contents at comparable K/Rb ratios (Fig. 7B, C). Muscovite EPMA and LA-ICP-MS results are provided in Appendix Tables A9 and A10, respectively.

Spodumene: Spodumenes of the Carolina tin-spodumene belt pegmatites have low trace element contents. The REE, U, Th, and Pb contents in spodumene are mostly below detection limits, and detected trace elements are often low (10s to 100s of ppm). Detected trace elements include Be, Na, Mn, Mg, P, Ca, Sc, Ti, Ga, and Ge, among others. Spodumenes of the Carolina tin-spodumene belt have Li/Mn ratios between 13 to 101, with an average of 39 ± 19, and Ga contents that range from 31 to 75 ppm with an average of 56 ± 10 ppm (Fig. 8A). In Li/Mn versus Ga space, Carolina tin-spodumene belt spodumene overlaps with the Plumbago North, Black Mountain, and Etta pegmatites, although the Etta spodumene reaches the highest Li/Mn ratios (Fig. 8A). Spodumenes from the BB#7, Mt. Marie, Mt. Mica, Tamminen, Lacorne, and Barroso-Alvão pegmatites have a higher Ga content than those documented for the Carolina tin-spodumene belt, Plumbago North, Black Mountain, and Etta pegmatites (Fig. 8A). Spodumene Li/Ge ratios for the Carolina tin-spodumene belt extend to higher values than other spodumene, and spodumenes from the Black Mountain, Mt. Marie, Mt. Mica, and BB#7 pegmatites in Maine have lower Li/Ge ratios (Fig. 8B). The Li/Ti ratios of Carolina tin-spodumene belt spodumene are the lowest measured in this study, and the Etta and Black Mountain pegmatites reach the highest Li/Ti ratios (Fig. 8C). Spodumene EPMA and LA-ICP-MS results are provided in Appendix Tables A11 and A12, respectively.

Garnet: Garnet is an accessory phase in the Carolina tin-spodumene belt pegmatites and occurs in both spodumene-bearing and spodumene-free pegmatites. This is the first published study to document garnet in the Carolina tin-spodumene belt spodumene-bearing pegmatites. Garnet is less abundant in spodumene-free pegmatites than those containing spodumene. Garnets from spodumene-bearing pegmatites range in composition from Alm₄₉ to Alm₅₈ (where Alm = almandine), with three analyses of Fe-rich spessartines with compositions <Alm₅₀, and those from spodumene-free pegmatites range from Alm₆₄ to Alm₇₄. Garnets from spodumene-bearing samples are clearly differentiated from their spodumene-free counterparts on the basis of MnO and FeO abundances, with the former enriched in MnO (21 ± 1 vs. 13 ± 1 wt %) and depleted in FeO (24 ± 1 vs. 30 ± 1 wt %) compared to the latter (Fig. 9A). Similar to K-feldspar K/Rb ratios, garnet MnO has a data gap at 15 > MnO > 18 wt %, and an FeO boundary can be observed at FeO = ~27 wt % between spodumene-bearing and spodumene-free samples (Fig. 9A). Garnets from spodumene-bearing pegmatites have Ga contents averaging 18 ± 4 ppm and Li contents averaging 212 ± 95 ppm, whereas garnets from spodumene-free samples have average Ga contents of 17 ± 2 ppm and average Li contents of 254 ± 59 (Fig. 9B). Although the respective ranges of Ga and Li contents of both types of garnet are similar, Carolina tin-spodumene belt garnets exhibit a positive covariation in Ga versus Li space, with those from spodumene-bearing pegmatites displaying slightly elevated Ga abundance for a given Li content compared to spodumene-free pegmatites (Fig. 9B). Garnets from spodumene-free pegmatites contain higher ΣREE contents (41 ± 42 ppm) and lower Sn contents (17 ± 18 ppm) than those of spodumene-bearing pegmatites (0.07 ± 0.09 ppm ΣREE; 167 ± 110 ppm Sn; Fig. 9C, D). The IQR of the ΣREE contents in garnet is 13 to 61 ppm in spodumene-free pegmatites and 0.03 to 0.13 ppm in spodumene-bearing pegmatites (Fig. 9D). The IQR of Sn in garnet is 8 to 18 ppm in spodumene-free samples and 59 to 264 ppm in spodumene-bearing samples (Fig. 9D). Garnet EPMA and LA-ICP-MS results are provided in Appendix Tables A13 and A14, respectively.

Apatite: Apatite is present in both Carolina tin-spodumene belt pegmatite types as accessory Mn-rich fluorapatite having an F content that ranges from 2.5 to 5.6 wt % (App. Table A15). Apatite MnO and FeO contents of Carolina tin-spodumene belt pegmatites range from 1.3 to 3.2 and 0.14 to 0.84 wt %, respectively (Fig. 10A). Apatites from spodumene-bearing and spodumene-free pegmatites have similar MnO contents (2.5 ± 0.5 and 2.1 ± 0.3 wt %, respectively) but more distinct FeO contents (0.33 ± 0.16 and 0.60 ± 0.14 wt %, respectively; Fig. 10A). The FeO contents of apatite from spodumene-bearing samples have an IQR of 0.24 to 0.38 wt %, and those from spodumene-free samples have an IQR of 0.51 to 0.69 wt % (Fig. 10A). Apatites from Carolina tin-spodumene belt pegmatites have variable trace element contents, including REEs, Sr, U, and Th, and are depleted in Zr and Hf (Figs. 10–11; App. Table A16). Apatite Sr contents range from 48 to 1,300 ppm, and spodumene-free samples have a lower average (58 ± 7 ppm) than spodumene-bearing samples (290 ± 262 ppm; Fig. 10B). Apatite Sr contents display a data gap at 70 > Sr > 100 ppm (Fig. 10). The ΣREE contents in Carolina tin-spodumene belt apatites range from 8 to 729 ppm. Spodumene-free samples have higher average ΣREE contents of 388 ± 146 ppm with an IQR of 282 to 490 ppm compared to spodumene-bearing samples with an average ΣREE content of 119 ± 117 ppm and an IQR of 26 to 192 ppm (Fig. 10B). Apatites from spodumene-free samples have a higher average U content (857 ± 552 ppm) and a lower average Th content (16 ± 13 ppm) than samples with spodumene (246 ± 93 ppm U; 35 ± 18 ppm Th; Fig. 10C). The IQR of U in apatite is 543 to 1,166 ppm in spodumene-free samples compared to 165 to 311 ppm in spodumene-bearing samples (Fig. 10C).

When normalized to chondritic values of McDonough and Sun (1995), apatite REE patterns display a distinct tetrad pattern and a prominent positive Eu anomaly (Fig. 11). A tetrad pattern is when chondrite-normalized REE patterns show four distinct segments designated T1 to T4 tetrads by Irber (1999) that occur between La to Sm (T1), Sm to Gd (T2), Gd to Ho (T3), and Ho to Lu (T4). In T1, apatites from both spodumene-free and spodumene-bearing pegmatites show a negative slope, but those from spodumene-free samples have a concave-down shape compared to the fairly straight to slightly concave-up shape of spodumene-bearing samples. The positive Eu anomaly in apatite is larger in spodumene-bearing samples compared to spodumene-free samples, and some REE concentrations of spodumene-bearing samples dip below chondritic values in T2 and Nd concentrations. Both types of pegmatites have concave-down shapes in T3, but spodumene-free samples have a negative slope compared to the generally positive slopes of spodumene-bearing samples. Apatites from spodumene-free samples have negative slopes in T4, with variable concavity, and those from spodumene-bearing samples have positive slopes with variable concavity. Apatite EPMA and LA-ICP-MS results are provided in Appendix Tables A15 and A16, respectively.

Trace element modeling: One fundamental question pertaining to whether a peraluminous, two-mica granite like the Cherryville Granite can be the parental granite body to Li-rich pegmatites in the Carolina tin-spodumene belt is, does fractional crystallization of such a granite enrich Li sufficiently in the residual melt to saturate with respect to Li phases? This question can be generalized to other granite-pegmatite systems where a parental, fertile granite is hypothesized to produce Li-rich pegmatites, because the Li content of the Cherryville Granite (44 ppm Li) is comparable to such fertile granites in terms of its Li content. For instance, Li contents in granites of the Winnipeg River pegmatite district (Canada) range between 10 to 76 ppm; the Harney Peak Granite in the southern Black Hills (South Dakota) contains 25 to 28 ppm Li; and the Black Wonder granite of the Quartz Creek pegmatite district (Colorado) has 4 to 30 ppm Li (Černý, 1982). Pegmatites in each of these three localities contain Li-rich pegmatites that are hypothesized to be derived from the nearby granitic bodies. Thus, trace element modeling is relevant not only to the Carolina tin-spodumene belt but also to granites worldwide with similar Li contents.

To ascertain whether the Cherryville Granite could produce pegmatites that crystallize spodumene, we modeled Li evolution during granitic fractional crystallization. Fractional crystallization was explored rather than equilibrium crystallization because complete crystal-melt equilibrium is usually not achieved in silicic magmas, and experimental petrology suggests fractional crystallization is the dominant process controlling the genesis of granitic pegmatites (London, 1992; Pichavant et al., 2007). Lithium trace element modeling was performed using the Rayleigh crystal fractionation equation with three different starting Li contents and three different bulk partition coefficients (⁠$KDbulk$⁠). Initial Li contents were chosen based on the whole-rock Li content of the Cherryville Granite (44 ppm) from this study and two Li contents representing melted metasediments from Koopmans et al. (2023). Koopmans et al. (2023) modeled Li enrichment during the melting of pelite and graywacke metasediments at different crustal pressure-temperature conditions relevant to pegmatite melt generation, and we chose the lowest and highest Li values of the produced melts, 255 and 538 ppm Li, as starting values. We calculated $KDbulk$ values for a magma crystallizing (1) a three-mineral eutectic composition of one-third each of quartz and two feldspars (⁠$KDbulk=0.4$⁠), (2) a five-mineral assemblage of 32% each of quartz and two feldspars plus 2% each of biotite and muscovite (⁠$KDbulk=0.6$⁠), and (3) using the minimum individual partition coefficients for either three-mineral or five-mineral assemblage (⁠$KDbulk=0.03$⁠). Combined with the three initial Li contents, we ran a total of nine models (Fig. 12). Mineral-melt K_D values used for the modeling were based on literature values for silicic systems (Walker et al., 1986, 1989; Jolliff et al., 1992; Bea et al., 1994; Icenhower and London, 1995; Neukampf et al., 2019). Partition coefficients used here are provided in Appendix Table A17.

Trace element modeling results are shown in Figure 12 down to a melt fraction of 0.03. Enrichment of Li in residual melts during fractional crystallization of granitic magmas is least pronounced in models with low starting Li content (44 ppm), reaching maximum values of only ~1,300 ppm Li. Enrichment of Li is most pronounced in models with lower $KDbulk$ values (0.03 and 0.4) and higher starting Li contents (255 and 538 ppm), reaching maximum values of nearly 10,000 ppm Li (Fig. 12). Based on crystallization and dissolution experiments using granitic compositions, Li solubility ranges from 5,000 to 15,000 ppm in melts between 500° and 750°C, meaning that a granite needs to concentrate at least 5,000 ppm in the melt to saturate with respect to Li phases such as spodumene (Maneta and Baker, 2014; Maneta et al., 2015). Such a saturation level occurs only in models using the lowest $KDbulk$ (0.03) and higher initial Li contents of 255 and 538 ppm. The two models in our study that reach 5,000 ppm Li do so only after 90 to 95% crystallization (i.e., 5–10% residual melt), a stage of crystallization well past the rheological lockup of granitic systems (Marsh, 1981; Vigneresse et al., 1996). Mobilizing such high-viscosity magma would be challenging even with the high-volatile content of pegmatitic melts.

Notably, all models that use the starting Li content of the Cherryville Granite do not reach Li saturation, even the model with a $KDbulk$ of 0.03, suggesting that it would be difficult for the Cherryville Granite to be the parental granite to the Carolina tin-spodumene belt Li-rich pegmatites without another mechanism of Li enrichment. The same approach can be extrapolated to other hypothesized fertile parental granites that have similar Li contents. Using the lowest $KDbulk$ (0.03), a granite would need an initial Li content of ~170 ppm to reach 5,000 ppm with 3% of melt remaining, which is ~4 times the Li content of the Cherryville Granite. Interestingly, the average Li content of spodumene-free pegmatite samples is only 102 ± 99 ppm, lower than the requisite 170 ppm. If extreme fractional crystallization does not concentrate Li sufficiently to result in crystallization of Li phases, and these granites are in fact parental to Li-rich pegmatites, then there must be an influx of Li to a geochemically open crystallizing system between parental granite formation and pegmatite formation. Granites do exist with Li contents ≥170 ppm Li, such as the early Permian Variscan granites described by Simons et al. (2017).

A new take on the RMG model was proposed by Troch et al. (2022), in which pegmatites crystallize directly from the magmatic volatile phase that forms after extreme levels of fractionation when fluids saturate in the crystallizing magmatic mush, rather than directly from crystallization of melt. This model is attractive because Li fluid-melt partition coefficients can be compatible, meaning Li prefers to be in the exsolved magmatic volatile phase after fluid saturation (i.e., $KDfluid−melt>1$⁠). However, the detailed modeling of Troch et al. (2022) shows a maximum enrichment factor of two for Li from melt to magmatic volatile phase formation. Applying this to our modeling where extreme fractionation produces ~1,300 ppm Li in the residual melt after 97% crystallization of a Cherryville Granite Li content and a K_D of 0.03, this only increases Li in the volatile phase to ~2,600 ppm Li, which is still well below the 5,000-ppm level required for Li phase saturation. Therefore, even if pegmatites are formed from the magmatic volatile phase instead of from melt, the Cherryville Granite, and other granites with similar Li contents, are unsuitable as parents to Li-rich, mineralized pegmatites.

The alternative to both the traditional RMG hypothesis and the Troch et al. (2022) RMG-magmatic volatile phase hypothesis is a DPA process—either the traditional single-stage anatexis or the two-stage anatexis proposed by Koopmans et al. (2023). Koopmans et al. (2023) prefer a two-stage model in which the melted metasedimentary rocks initially crystallize into a granite and then are subsequently remelted to form a melt that crystallizes into a pegmatite. This process of remelting and subsequent crystallization as a pegmatite enriches the melt past the 5,000-ppm Li requirement. Koopmans et al. (2023) discount single-stage anatexis because their Li crystallization models do not reach 5,000 ppm Li at or before 90% crystallization. Because one of our models (K_D = 0.03, Li = 538 ppm) does reach this 5,000-ppm threshold at 90% crystallization, we consider both anatectic models viable.

Trace element mineral chemistry: In addition to the trace element chemistry of whole rocks, trace element mineral chemistry provides insight into pegmatite petrogenesis. Quartz trace element chemistry has been used as a record of magma evolution for decades (e.g., Wark and Watson, 2006), and Müller et al. (2021) provided an in depth discussion as it applies to pegmatite type discrimination and petrogenesis of group 1 and group 2 pegmatites. Using principal component analysis of quartz trace element chemistry from pegmatites throughout the world formed through different petrogenetic processes (i.e., DPA versus RMG), Müller et al. (2021) argued that RMG-derived group 1 pegmatites are characterized by higher Rb and Al contents in quartz than DPA-derived group 1 pegmatites. Quartz Rb and Al variations of Carolina tin-spodumene belt pegmatites are compared to the quartz data used in Müller et al. (2021) in Figure 13 (Müller et al., 2008, 2015; Beurlen et al., 2011; Dill et al., 2013; Garate-Olave et al., 2017). The IQR of quartz Rb contents in Carolina spodumene-bearing pegmatites is similar to that of the RMG-derived Tres Arroyos (Spain) and Hagendorf-Pleystein (Germany) pegmatites, as well as the DPA-derived San Luis (Argentina) pegmatites (Fig. 13C). Among all quartz pegmatite data compared in Figure 13C, only quartz from the Hagendorf-Pleystein field shows a significantly higher IQR of Al content (Fig. 13D). Given the ambiguities in Rb-Al systematics in quartz highlighted here, it is unclear if these elements are useful in distinguishing a DPA versus RMG petrogenesis for pegmatites of the Carolina tin-spodumene belt.

As traditionally incompatible elements, the REEs typically increase in the melt during crystallization of granitic magmas, and accessory phases like zircon, titanite, allanite, and apatite, with K_D > 1 for all or some REEs, accumulate REEs during progressive crystallization (e.g., Prowatke and Klemme, 2006; Rubatto and Hermann, 2007; Schaltegger and Davies, 2017). Within this framework, the reduction in whole-rock REE contents from the Cherryville Granite to the spodumene-free pegmatites and ultimately to the spodumene-bearing pegmatites with the lowest REE values is not compatible with a genetic relationship between these three granitic rocks in which the pegmatites are derived from the Cherryville Granite via fractional crystallization. In terms of why the whole-rock ΣREE contents are so low in spodumene-bearing pegmatites, we suggest three possible answers: (1) the grain-size problem of pegmatites prevented an accurate determination of ΣREE content in whole rock samples, (2) formation through a DPA process with high degrees of partial melting from a similarly REE-poor source, or (3) crystallization of magma in which the REE-poor, Li-rich pegmatites form after the REEs have been extracted into REE-rich accessory phases, forming a complementary and as yet unidentified REE-rich lithology somewhere in the Carolina tin-spodumene belt system. We suggest that option (1), although possible, is unlikely to be the main cause of low whole-rock ΣREE contents, based on the consistently low ΣREE contents in Carolina tin-spodumene belt pegmatites as well as their resemblance to the REE contents in similar group 1 pegmatites, such as Mt. Mica in Maine (Simmons et al., 2016; Webber et al., 2019). Option (2) also seems unfeasible because metasedimentary rocks typically have higher REE contents than those documented for the Carolina spodumene-free pegmatites, so even complete melting of such a rock would not decrease the REE contents sufficiently (Harris and Inger, 1992). Option (3) presents an appealing hypothesis from an economic point of view, since somewhere in the Carolina tin-spodumene belt there could be a complementary REE-rich rock of economic potential. The existence of such a rock is purely hypothetical at this point, and it is beyond the scope of this study to explore this possibility further.

Interestingly, the low REE contents observed in our whole-rock data are mimicked in plagioclase, garnet, and apatite (Figs. 4, 9–11). Plagioclase Eu and ΣREE contents are lower, or not even detected, in spodumene-bearing samples compared to spodumene-free pegmatites and the Cherryville Granite (Fig. 4B, C). Garnet and apatite ΣREE contents are also notably lower in spodumene-bearing samples compared to spodumene-free samples (Figs. 9C, D, 10B, 11). The discussion below considers why plagioclase, garnet, and apatite ΣREE contents are so low in spodumene-bearing pegmatites in the Carolina tin-spodumene belt.

In terms of mineral ΣREE contents, plagioclase, garnet, and apatite offer a unique view of a magmatic system because of the different ways in which they incorporate, or exclude, the REEs. Although the trivalent REEs are incompatible in plagioclase (Bindeman and Davis, 2000), the ΣREE contents displayed in Figure 4C show that plagioclase nevertheless hosts some REEs. Divalent Eu²⁺, which substitutes for Ca²⁺ in plagioclase and is responsible for the well-known Eu anomaly in many igneous systems (Weill and Drake, 1973), is only detected in one plagioclase analysis from a spodumene-bearing pegmatite (Fig. 4B). In a magma-crystallizing plagioclase, the residual liquid has less Eu than the original melt. Based on this, the observation of plagioclase Eu contents decreasing from the Cherryville Granite to the spodumene-free pegmatites to the spodumene-bearing pegmatites makes sense in terms of a typical crystallization sequence (Fig. 4B). However, the fact that abundances of ΣREEs, which are dominantly incompatible, in plagioclase follow the same pattern does not agree with that logic, because Eu is the only REE compatible in plagioclase. Instead, we suggest that the plagioclase Eu patterns described here are likely due to both the low Ca content of these plagioclase analyses and the low REE content of the melt from which the spodumene-bearing pegmatite crystallized.

Incorporation of REEs into garnet favors the heavy HREEs over the LREEs, with partition coefficients between garnet and melt (⁠$KDgrt−melt$⁠) generally <1 (incompatible) for the LREEs and $KDgrt−melt>1$ (compatible) for the HREEs (Rubatto and Hermann, 2007; Taylor et al., 2015). All REEs are compatible to some degree in apatite, with the middle REEs (MREEs) more compatible than the LREEs and HREEs (Watson and Green, 1981; Pan and Fleet, 2002). The observation that ΣREE contents in garnets and apatites from Carolina spodumene-bearing pegmatites are lower than those of spodumene-free pegmatites is consistent with a genetic relationship via crystallization sequence only if apatite and garnet are crystallizing in sufficiently large quantities to make the REEs compatible so that their concentrations progressively decrease in the melt. Although this might be possible with the HREEs because they are compatible in both apatite and garnet, the LREEs are only compatible in apatite and would likely remain incompatible overall. Therefore, LREE abundances should keep increasing in the melt, and the relationship observed between apatite LREEs in spodumene-free and spodumene-bearing pegmatites is not consistent with a genetic relationship between these two types of Carolina tin-spodumene belt pegmatites (Fig. 11).

One possibility to consider is the effect of fluid saturation on REE contents in the magma. Fluid-melt partition coefficients for REEs (⁠$KD−REEsfluid−melt$⁠) are mostly incompatible, with some studies showing compatibility for Sm and Tm (i.e., $KD−Sm,Tmfluid−melt>1$⁠) and others showing less incompatibility for Eu (Flynn and Burnham, 1978; London et al., 1988; Candela, 1990; Zajacz et al., 2008). This means that, taken as a group, the REEs will not suddenly leave the melt for the fluid phase once fluid saturation occurs, and continued enrichment in the melt should continue, with the potential exception of Sm and Tm. Therefore, fluid saturation does not explain the dearth of REEs in garnets and apatites in spodumene-bearing pegmatites compared to those in spodumene-free pegmatites.

Apatite REE patterns display prominent positive Eu anomalies for both spodumene-bearing and spodumene-free pegmatites (Fig. 11). Apatite Eu anomalies in Carolina tin-spodumene belt pegmatites span three orders of magnitude ranging from Eu/Eu* = 1 to Eu/Eu* = 189 (Fig. 14). Apatites from spodumene-bearing pegmatites in the Carolina tin-spodumene belt have higher average Eu anomalies (Eu/Eu* = 35 ± 46) compared to those from spodumene-free pegmatites (Eu/Eu* = 7 ± 3). Compared to apatite trace element geochemistry from the Central Iberian zone pegmatites of Spain and Portugal (Roda-Robles et al., 2022), Carolina tin-spodumene belt apatites have larger Eu anomalies (Fig. 14). The apatite Eu anomaly from Li-rich pegmatites of the Central Iberian zone is toward the lower end of Carolina tin-spodumene belt Eu anomalies (Fig. 14). Apatite data from barren pegmatites of the Central Iberian zone have lower Eu anomalies compared to those of spodumene-free pegmatites of the Carolina tin-spodumene belt (Fig. 14). Some researchers suggest that negative Eu anomalies in apatite can indicate reduced magmatic conditions (e.g., Miles et al., 2014), but oxidized magmatic conditions are harder to discern using apatite. An oxidized environment with a low Eu²⁺/Eu³⁺ ratio would not produce a positive Eu anomaly in apatite, because apatite prefers the oxidized Eu³⁺ state, which is the same as the other trivalent REEs (Watson and Green, 1981; Rakovan et al., 2001; Pan and Fleet, 2002; Bromiley, 2021). Instead, we suggest that the positive Eu anomalies in apatite were inherited from an Eu-enriched parental magma.

In addition to prominent positive Eu anomalies, apatites in Carolina tin-spodumene belt pegmatites show marked REE tetrad patterns when normalized to chondritic values (Fig. 11). We quantify these tetrad patterns using the TE_1,3 parameter of Irber (1999), who argued that TE_1,3 > 1.1 means that sample shows the tetrad effect. All of our Carolina tin-spodumene belt apatites display the tetrad effect using this criterion (Fig. 14). Apatites from Carolina tin-spodumene belt pegmatites have TE_1,3 values between 1.12 and 1.39, with those from spodumene-bearing pegmatites having lower values (1.22 ± 0.05) than those from spodumene-free pegmatites (1.35 ± 0.03; Fig. 14). In addition to apatite, whole-rock REEs of spodumene-free pegmatites have TE_1,3 values between 0.84 and 1.6, with an average of 1.22 ± 0.27, whereas TE_1,3 whole-rock values could not be calculated for spodumene-bearing pegmatites because of the many REE abundances below the analytical limit of detection. If whole-rock REEs mimicked the apatite trends, then spodumene-bearing TE_1,3 values would be less than those of the spodumene-free pegmatites, but this cannot be confirmed.

Irber (1999) observed that apatites have tetrad patterns similar to the whole-rock from which they were derived, arguing that these patterns are inherited from the melt and that typical fractional crystallization of a granitic melt does not cause the tetrad effects in whole rock or apatite. Other researchers have suggested that tetrad patterns in evolved granitic systems indicate a trace element behavior similar to that in aqueous systems where chemical complexation exerts a significant control on elemental fractionation in addition to charge and radius, but the acquisition of the tetrad pattern in relationship to the timing of fluid saturation and evolution is unclear (Bau, 1996; Irber, 1999; Monecke et al., 2002). According to these studies, tetrad patterns are observed in magmas significantly influenced by a fluid phase and are correlated with magmas rich in components like H₂O, CO₂, Li, B, F, P, and Cl. They also noted that TE_1,3 values increase with degree of evolution or fractionation. However, apatites from the Li-rich, spodumene-bearing pegmatites in the Carolina tin-spodumene belt that are arguably more evolved and fractionated have lower TE_1,3 values than those of spodumene-free pegmatites. One explanation of this is that fluid activity was higher in the less fractionated melts that produced spodumene-free pegmatites than in melts that produced the spodumene-bearing pegmatites. An alternative explanation is that both melts had a significantly high fluid activity based on TE_1,3 values >1.1 in both types of pegmatite, but the fluids involved had different compositions that variably affected their parental magmas. Based on the conclusion of this study that these two types of pegmatites are likely not related genetically, the latter explanation is favored. In our view, based on the differences in chemistries of each Carolina tin-spodumene belt pegmatite type, the two magmas developed independently, each reaching a sufficiently high fluid activity with fluids of different composition.

Compared to apatites from pegmatites in the Central Iberian zone in Spain and Portugal (Roda-Robles et al., 2022), apatites of the Carolina tin-spodumene belt have a more constrained spread of TE_1,3 values (Fig. 14). Apatites from the Li-rich pegmatites in the Central Iberian zone have TE_1,3 values <1.1, so they technically do not show a tetrad effect (Fig. 14). Apatites from P-rich pegmatites in the Central Iberian zone reach higher TE_1,3 values than those of Carolina tin-spodumene belt pegmatites (Fig. 14). An interesting similarity between the Carolina tin-spodumene belt and Central Iberian zone apatite data is that Li-rich pegmatites in both settings have lower TE_1,3 values than other pegmatites in each setting (Fig. 14).

Taken together, REE levels in plagioclase, garnet, and apatite show notable differences between spodumene-bearing pegmatites and spodumene-free pegmatites in the Carolina tin-spodumene belt that are not reconcilable with a genetic relationship via fractional crystallization. We argue that the lower REE contents in minerals from spodumene-bearing pegmatites compared to those from spodumene-free pegmatites and the variations in fluid activity based on apatite tetrad patterns reflect the differences in melt source, degree of partial melting, and/or crystallization history of the melts that produced each Carolina tin-spodumene belt pegmatite type.

K-feldspar and muscovite: Enrichment of incompatible elements during crystallization of granitic magmas provides a useful way to compare different group 1 pegmatites. For instance, the increase in Li, Rb, and Cs in residual melt during progressive crystallization due to their incompatibility in typical granitic minerals is the basis for using K/Rb-Li (or -Cs) systematics as a general indication of magma evolution or degree of fractionation in pegmatite whole rock or muscovite, as discussed and used in previous studies for over half a century (Gordiyenko, 1971; Černý and Burt, 1984; Černý, 1990; Smeds, 1992; Černý et al., 2012a; Maneta and Baker, 2019; Wise et al., 2024). To be clear, such fractionation can occur in either the RMG or DPA petrogenetic models, with the difference being the extent of fractional crystallization needed to attain the saturation of Li phases. The incorporation of Li, Rb, and Cs into muscovite and K-feldspar has been discussed and reviewed in numerous studies, with the mechanism being either direct substitution or interstitial incorporation (Černý et al., 1985; Maneta and Baker, 2019; Wise et al., 2024). The mechanism of Li incorporation into K-feldspar is still unclear, but it occurs based on the sometimes hundreds of parts per million of Li observed in K-feldspar (e.g., Fig. 5B).

The K content of K-feldspar and muscovite is relatively constant, so lower K/Rb ratios and increasing Li and Cs contents generally reflect the covarying increase of Rb, Li, and Cs contents of the compositionally evolving melt.

We chose to compare Carolina tin-spodumene belt pegmatites to group 1 pegmatites with various levels of mineralogical complexity from pegmatite fields in Maine and South Dakota that were mined historically and have potential for spodumene mining in the future (Page, 1953; Simmons et al., 2020). For this discussion, we divide pegmatites into three main categories similar to the divisions of Wise and Brown (2010): (1) relatively simple and homogeneous pegmatites with accessory tourmaline and garnet at most, (2) moderately evolved pegmatites containing beryl mineralization ± columbite ± phosphates, and (3) evolved pegmatites containing some form of Li mineralization that includes spodumene, petalite, lepidolite, and/or amblygonite-montebrasite. No simple pegmatites were analyzed in this study. The moderately evolved beryl pegmatites include Bumpus, Cole, Lord Hill, Hibbs, Willis Warren, GE, Dan Patch, Gap Lode, Buster, Old Mike, and Silver Dollar (Page, 1953; Jolliff et al., 1992; Wise and Brown, 2010). Among these, Cole, Hibbs, GE, Old Mike, and Dan Patch are of the beryl-columbite subtype, whereas Lord Hill and Bumpus are of the beryl-columbite-phosphate (e.g., triphylite-lithiophilite or triplite) subtype. The evolved, Li-mineralized pegmatites, besides those in the Carolina tin-spodumene belt, include BB#7, Black Mountain, Bennett, Mt. Marie, Mt. Mica, Emmons, Hayes, Plumbago North, Tamminen, Waisanen, Tip Top, Tin Mountain, Etta, and Hugo (Page, 1953; Norton et al., 1962; Norton, 1964; Shearer et al., 1985, 1986; Walker et al., 1986; Jolliff et al., 1992; Brown and Wise, 2001; Wise and Brown, 2010; Falster et al., 2019). Of these pegmatites, the Tamminen and Mt. Marie pegmatites are petalite subtypes with minor spodumene. Bennett, BB#7, Black Mountain, Emmons, Hayes, Mt. Mica, Waisanen, Plumbago North, Hugo, Etta, Tip Top, and Tin Mountain contain spodumene with no known petalite. Amblygonite-montebrasite is found at Plumbago North, Hugo, Tip Top, Tin Mountain, and Hugo, whereas both amblygonite-montebrasite and lepidolite are present at Black Mountain, Mt. Mica, Tamminen, and Mt. Marie. Importantly, we do not have information on the pegmatite zones from which these muscovites and K-feldspars from outside the Carolina tin-spodumene belt were sampled.

The Carolina tin-spodumene belt muscovites span most of the range of K/Rb ratios and Li contents observed for the Maine and South Dakota muscovites, with the exception of the lithian muscovites of Black Mountain, Tin Mountain, Bennett, and Plumbago North, and the low-Li muscovite of Willis Warren and Silver Dollar (Fig. 7B, C). These results are generally consistent with a model in which more fractionated pegmatites have lower K/Rb ratios and higher Li contents, as discussed in detail by Wise et al. (2024). Wise et al. (2024) divided the muscovite K/Rb-Li plot into poorly fractionated (K/Rb > 40), moderately fractionated (10 < K/Rb < 40), highly fractionated (K/Rb < 10), Li-rich (>500 ppm), and Li-poor (<500 ppm) domains. The evolved Black Mountain and Tin Mountain muscovites have the highest Li contents and lowest K/Rb ratios, placing them in the Li-rich, highly fractionated domain (Fig. 7B, C). This is consistent with those pegmatites having several Li phases, including spodumene and amblygonite-montebrasite in both, as well as lepidolite in Black Mountain (Walker et al., 1986; Brown and Wise, 2001).

The moderately evolved, beryl ± columbite ± phosphate pegmatites in Maine and South Dakota plot in both the poorly and moderately fractionated domains (Fig. 7B, C). According to Wise et al. (2024), beryl pegmatites plot in the poorly and moderately fractionated domains, but beryl-columbite-phosphate pegmatites concentrate in the moderately fractionated domain. Among the Oxford County beryl pegmatites, Willis Warren plots in the moderately fractionated, Li-poor domain and is the only beryl pegmatite without phosphates and columbite (Fig. 7B). The Hibbs beryl-columbite pegmatite plots in the poorly fractionated field, and the other beryl-columbite (Cole and GE) and beryl-columbite-phosphate (Lord Hill and Bumpus) pegmatites plot in the moderately fractionated, Li-rich domain (Fig. 7B). One explanation for the Hibbs pegmatite muscovite plotting in the poorly fractionated field is that it was collected from a less fractionated area of the pegmatite. Among the South Dakota pegmatites, three of the five beryl pegmatites plot solely in the poorly fractionated area, including the Silver Dollar, Buster, and Gap Lode that contain no columbite or phosphates (Fig. 7C). The Dan Patch and Old Mike beryl-columbite pegmatites plot partially or fully in the moderately fractionated domain (Fig. 7C).

Muscovites from evolved pegmatites with at least one Li phase plot in the poorly, moderately, and highly fractionated K/Rb fields (Fig. 7). As discussed above, the highly fractionated character of the Black Mountain and Tin Mountain pegmatites are consistent with saturation of multiple Li phases in those pegmatites, but micas from evolved pegmatites in the other fractionation domains require more explanation. All other evolved pegmatites have micas that plot in the high-Li, moderately fractionated field except Mt. Marie, whose mica data plot only in the poorly fractionated field (Fig. 7B). Taken alone, this might suggest that this is a simple or beryl pegmatite. However, given existing knowledge of the evolved nature of Mt. Marie, which contains petalite, pollucite, and minor spodumene, it is likely that these micas were acquired from less fractionated parts of a zoned pegmatite. Muscovites from the Hayes pegmatite have a wide range of Li contents (88–1,006 ppm) spanning the Li-poor to Li-rich fields, and those from the Plumbago North and Black Mountain pegmatites have a wide range of K/Rb ratios (11–48 and 7–51, respectively) and Li contents (487–6,146 and 520–15,630, respectively; Fig. 7B).

Of the Maine and South Dakota pegmatite suites, only K-feldspars from Bumpus, Black Mountain, Bennett, Mt. Mica, Tip Top, Etta, and Hugo pegmatites were analyzed for geochemistry (Fig. 5B, D). K-feldspars from the Bumpus beryl pegmatite have K/Rb ratios of 121 to 135 that place it above the K/Rb data gap, combined with low Li (3–5 ppm) and Cs contents (7–9 ppm; Fig. 5B, D). All other K-feldspars from Maine and South Dakota have at least some analyses that plot below the K/Rb data gap (Fig. 5B). K-feldspars from the Black Mountain and Etta pegmatites reach the lowest observed K/Rb ratios (~14 and 17–18, respectively), with Black Mountain K-feldspars having much higher contents of Li (71–131 ppm) and Cs (994–1,682 ppm) than those of Etta (5–10 ppm Li and 228–328 ppm Cs; Fig. 5B, D). The low K/Rb ratios and difference in Li contents likely reflect the fact that, whereas both pegmatites are evolved sufficiently to contain spodumene, the Black Mountain pegmatite also contains amblygonite-montebrasite, unlike Etta. Of all K-feldspars analyzed from the Maine and South Dakota pegmatites, only those of the Bennett pegmatite have analyses that plot on both sides of the K/Rb data gap, as is the situation for the Carolina tin-spodumene belt (Fig. 5B). The Bennett analyses in Figure 5B come from three different feldspars, and three clusters are present within the Bennett data. Based on the K/Rb systematics observed in Carolina tin-spodumene belt K-feldspars, it is possible that the Bennett K-feldspar with K/Rb < 70 comes from a zone of the pegmatite with Li mineralization, but these same analyses also plot in the <15-ppm Li domain. Without hand-sample confirmation of spodumene, greater confidence of the presence of spodumene in this sample would come in the form of analyses with K/Rb < 70 and Li > 15 ppm. The same could be said for the low-K/Rb and low-Li Etta K-feldspar data. We observe that the K/Rb data gap in K-feldspars from pegmatites in North Carolina, Maine, and South Dakota separates K-feldspars from pegmatites with Li mineralization (K/Rb < 70) and those without (K/Rb > 100). K-feldspars with K/Rb > 100 come from poorly to moderately fractionated pegmatites of the simple or beryl types, and K-feldspars with K/Rb < 70 come from highly fractionated pegmatites that have Li mineralization. The 15-ppm Li cutoff aids in distinguishing Li mineralization in pegmatites in the Carolina tin-spodumene belt pegmatites, but more data needs to be acquired to better understand this cutoff in zoned, Li-mineralized pegmatites like Etta and Bennett.

Quartz: The incorporation of elements such as K, Rb, Li, and Al that are incompatible in the quartz structure has been discussed and reviewed in numerous studies (e.g., Černý et al., 1985; Müller et al., 2021). Potassium and Rb are incorporated interstitially into the framework silicate structure, whereas Li⁺ and Al³⁺ reflect a coupled substitution for Si⁴⁺. Because univalent Li⁺ acts as a charge balance for Al³⁺, Li and Al are positively correlated, as observed in Carolina tin-spodumene belt quartz geochemistry (Fig. 6B). As noted by Müller et al. (2021), data on the Li versus Al plot for quartz should not have higher Li concentrations than the line defined by the atomic weight ratio of Li/Al (1:3.89) because of this codependence of Li and Al variation in quartz (Fig. 6B).

Quartz geochemistry of Carolina tin-spodumene belt pegmatites is compared in Figure 13 to that for other group 1 pegmatite fields (Müller et al., 2008, 2015, 2021; Beurlen et al., 2011; Dill et al., 2013; Garate-Olave et al., 2017). Note that this comparison excludes the quartz data of Müller et al. (2021) from group 2 pegmatites. Quartz geochemistry of the Carolina tin-spodumene belt pegmatites is comparable to other group 1 pegmatite areas around the world, with some notable differences. In terms of K/Rb ratios, Carolina tin-spodumene belt quartz has K/Rb IQRs similar to those of the Borborema (Brazil), San Luis, and Hagendorf-Pleystein pegmatite fields (Fig. 13A). Only one quartz K/Rb ratio from the Oxford County pegmatites is plotted in Figure 13A because most quartz analyses from that region in Müller et al. (2021) did not detect K. Quartz K/Rb ratios from the Tres Arroyos pegmatites are the lowest of all data compared here owing to lower K contents, given that the IQR of Rb content for Tres Arroyos quartz is similar to that of San Luis and Hagendorf-Pleystein quartz (Fig. 13A, C). The IQR for quartz Li content from Carolina spodumene-bearing pegmatites is comparable to that of Borborema, Oxford County, and Tres Arroyos, and Carolina tin-spodumene belt quartz Li contents extend to high values exceeded only by Oxford County quartz (Fig. 13B). In terms of quartz Rb content, the IQR of Carolina spodumene-bearing pegmatites is more comparable to the higher IQRs of Rb in quartz found in San Luis, Tres Arroyos, and Hagendorf-Pleystein quartz (Fig. 13C).

In addition to K, Rb, and Li, trace amounts of Al defects are incorporated into the quartz framework silicate structure (Weil, 1984; Müller and Koch-Müller, 2009; Müller et al., 2021). The IQR of quartz Al content in spodumene-bearing pegmatites of the Carolina tin-spodumene belt is similar to all other group 1 pegmatite fields shown in Figure 13D except that of the Hagendorf-Pleystein pegmatites, which have higher values than other pegmatites. The IQR of quartz Al content from Carolina tin-spodumene belt pegmatites indicates that Al > 165 ppm is likely to indicate Li mineralization (Fig. 13D), compared to the lower value of 100 ppm Al suggested by Müller et al. (2021). This difference is likely because we analyzed one pegmatite field in detail and found a certain Al content that predicts Li mineralization (165 ppm), whereas Müller et al. (2021) suggested the 100 ppm Al cutoff to distinguish Li-rich pegmatites from group 2 pegmatites as part of a broader study. As can be seen in Figure 13D, most pegmatites from the Carolina tin-spodumene belt, along with the data included from Müller et al. (2021), have quartz Al contents >100 ppm, as would be predicted by Müller et al. (2021) because they are group 1 pegmatites.

Garnet: Garnet geochemistry offers another opportunity to assess Carolina tin-spodumene belt pegmatites in the context of other group 1 pegmatites. We compare Mn, Ga, Li, and ΣREE contents of garnet in the Carolina tin-spodumene belt to those presented in Hernández-Filiberto et al. (2021) from the Oxford County pegmatite field, including the Mt. Mica, Berry-Havey, Emmons, Perham, and Stop-35 pegmatites, as well as the Palermo No. 1 pegmatite in New Hampshire (Fig. 15). The Palermo No. 1 pegmatite is a moderately evolved beryl-phosphate pegmatite, whereas both Perham and Stop-35 are barren pegmatites (Hernández-Filiberto et al., 2021).

Because of its incompatible nature in the main rock-forming minerals in granite (i.e., quartz and feldspars), Mn in garnet is often used as an indication of the degree of fractionation for a pegmatite system (Hernández-Filiberto et al., 2021). The barren, least fractionated pegmatites Stop-35 and Perham have garnet with the lowest IQR for MnO, and the more evolved pegmatites, including Mt. Mica, Berry-Havey, Emmons, and the spodumene-bearing pegmatites of the Carolina tin-spodumene belt, have higher IQRs for MnO (Fig. 15A). The moderately evolved, beryl-bearing Palermo No. 1 pegmatite has an IQR for MnO in garnet that is similar to that for the more evolved Mt. Mica pegmatite, suggesting that garnet MnO alone is not an indication of pegmatite Li phase saturation (Fig. 15A).

The mechanism of Li incorporation into the garnet structure is unclear, but Hernández-Filiberto et al. (2021) suggest that Li is incorporated in coupled substitution with P, based on the positive linear correlation between the two elements, in exchange for either Fe or Mn. The IQRs for Li content in garnet from Carolina tin-spodumene belt pegmatites are similar to those in all evolved, moderately evolved, and barren pegmatites of Hernández-Filiberto et al. (2021) except Stop-35, which has lower garnet Li contents (Fig. 15B). Garnets from Mt. Mica and the spodumene-bearing pegmatites of the Carolina tin-spodumene belt have the highest Li contents (>500 ppm; Fig. 15B). The observation that the barren Perham pegmatite and the beryl-bearing Palermo No. 1 pegmatite have garnet Li contents comparable to mineralogically more evolved pegmatites suggests that higher garnet Li contents alone are not an indication of pegmatite Li mineralization. This conclusion is strengthened by the additional observation that garnets from spodumene-bearing pegmatites of the Carolina tin-spodumene belt have a lower IQR for Li content than those from spodumene-free pegmatites (Fig. 15B).

The Ga content of pegmatites has been of interest for decades owing to its incompatible nature and resultant enrichment during fractionation combined with its chemical similarity to Al (Černý et al., 1985; Müller et al., 2022). Černý et al. (1985) noted the affinity of Ga for aluminous phases, including feldspars, micas, tourmaline, and garnet. The IQR of garnet Ga content in Carolina tin-spodumene belt pegmatites is higher than that of the Stop-35 barren pegmatite and Palermo No. 1 beryl pegmatite, but it is lower than that of the barren Perham pegmatite and the evolved Mt. Mica, Berry-Havey, and Emmons pegmatites (Fig. 15C). As a fractionation index, garnet Ga content is not consistent with mineralogy, because the barren Perham pegmatite has higher values than those of the Li-mineralized Carolina tin-spodumene belt pegmatites, and the Be-mineralized Palermo No. 1 pegmatite has the lowest garnet Ga contents in Figure 15C. Instead, we suggest that garnet Ga content is specific to each pegmatite system, indicative of its distinct petrogenetic history, and contributes to a unique fingerprint of that pegmatite system, similar to the “regional specialization” theory noted by Černý et al. (1985) for Ga in granitic systems.

As previously noted, the HREEs are compatible in garnet, whereas the LREEs are not (Rubatto and Hermann, 2007; Taylor et al., 2015). As is the situation with Mn, garnet ΣREE contents effectively distinguish between spodumene-bearing and spodumene-free pegmatites in the Carolina tin-spodumene belt (Figs. 9C, D, 15D). Garnet ΣREE contents of spodumene-bearing pegmatites are almost negligible, with values <1 ppm. The total range of ΣREE contents in Carolina tin-spodumene belt garnets to values <1 is only matched by that of the evolved Berry-Havey pegmatite, and Carolina tin-spodumene belt garnet reaches lower ΣREE contents than those of the Berry-Havey pegmatite (Fig. 15D). All other Maine and New Hampshire garnet data of Hernández-Filiberto et al. (2021) have higher IQRs of garnet ΣREE contents than spodumene-bearing pegmatites of the Carolina tin-spodumene belt (Fig. 15D). Most IQRs of garnet ΣREE contents are similar to those from spodumene-free samples in the Carolina tin-spodumene belt, but garnet data from the Palermo No. 1 beryl pegmatite plots between spodumene-free and spodumene-bearing Carolina tin-spodumene belt garnet data in terms of ΣREE content (Fig. 15D).

Spodumene: Spodumene trace element chemistry has received little attention in pegmatite studies, likely because all other avenues of research (i.e., whole-rock and mineral chemistry) have been used to find spodumene-bearing pegmatites, with little need to know the exact composition of spodumene trace elements. As noted above, Ga behaves similarly to Al and tends to incorporate into aluminous phases such as spodumene. Spodumene Mn, Ge, and Ti contents are considered here because of their incompatible nature in regard to Carolina tin-spodumene belt pegmatite mineralogy. Because Li is somewhat constant in spodumene, similar to K in muscovite, the ratios of Li/Mn, Li/Ge, and Li/Ti have been plotted versus Ga content in Figure 8 to evaluate trace element behavior in spodumene. Spodumene of the Carolina tin-spodumene belt pegmatites is compared to that from Oxford County, Etta, Lacorne, and Barroso-Alvão spodumene-bearing pegmatites. The Lacorne and Barroso Alvão pegmatites contain spodumene, petalite, and lepidolite Li mineralization (Martins et al., 2012; Mulja and Williams-Jones, 2018). The BB#7, Etta, and Carolina tin-spodumene belt pegmatites contain only spodumene in terms of Li mineralization. As with the K-feldspars and muscovites from outside of the Carolina belt, we note here that pegmatite zone sampling information is not available for spodumenes collected outside of the Carolina tin-spodumene belt.

Spodumenes from the Barroso-Alvão and BB#7 pegmatites have the highest Ga contents, and Carolina tin-spodumene belt spodumenes have the lowest observed Ga contents (Fig. 8). Spodumenes from Etta pegmatites have the highest Li/Mn ratios, and spodumenes from the Carolina tin-spodumene belt, BB#7, and La Corne pegmatites have the lowest Li/Mn ratios (Fig. 8A). This spread in Li/Mn ratios and Ga contents does not correlate with the number of different Li phases present. For instance, the Etta and BB#7 spodumenes have the highest and lowest Li/Mn ratios, respectively, even though they have a mineralogy similar to that of spodumene as the only Li mineral present. Pegmatites with multiple Li minerals, such as Plumbago North with spodumene and amblygonite-montebrasite and Tamminen with petalite and pollucite, have spodumene with intermediate Li/Mn ratios in Figure 8. Differences in melt Mn content could be the result of either Mn enrichment via fractionation or depletion through crystallization depending on the crystallization history of each pegmatite. In this sense, although Mn can be a useful fractionation index, the mineralogical complexity of these pegmatites highlights the need to pay attention to which phases are saturated and how that affects Mn content.

Spodumene Ge and Ti contents also appear to reflect the unique geochemical history of each pegmatite rather than being an indicator for the degree of fractionation. Black Mountain spodumene has the lowest Li/Ge ratios and some of the highest Li/Ti ratios, whereas Carolina belt spodumene, which has the highest Li/Ge ratios and lowest Li/Ti ratios, displays the opposite behavior (Fig. 8B, C). Spodumene from Tamminen, Mt. Marie, and Mt. Mica cluster in Li/Mn-Ga space, and Carolina tin-spodumene belt, Plumbago North, and Black Mountain spodumenes create a different cluster in the same space (Fig. 8A). However, spodumene analyses from these localities separate in Li/Ge or Li/Ti versus Ga space (Fig. 8B, C). For instance, Tamminen, Mt. Marie, and Mt. Mica have distinct Li/Ge ratios. Spodumene, as the end product of extensive crystallization and itself indicating saturation of a typically incompatible element, seems to incorporate other incompatible elements in ways that can be used to distinguish pegmatite localities.

Plagioclase, K-feldspar, quartz, muscovite, garnet, and apatite chemistry of Carolina tin-spodumene belt pegmatites with and without spodumene provide a multimineral, multielement methodology to determine the presence of spodumene. Based on the IQRs of Ca and ΣREE contents in plagioclase, the spodumene-bearing versus spodumene-free boundaries can be placed at 900 ppm Ca and 0.05 ppm ΣREE, with Ca < 900 ppm and ΣREE < 0.05 ppm in spodumene-bearing pegmatites (Fig. 4A, C). The significant gap between 70 to 100 in K/Rb ratios of K-feldspar shows that spodumene-bearing samples typically have K-feldspar with K/Rb < 70, whereas those lacking spodumene have K/Rb > 100 (Fig. 5). The IQRs of Li contents in K-feldspar define a boundary at 15 ppm Li, with spodumene-bearing pegmatites having Li > 15 ppm (Fig. 5A). The IQRs of K/Rb ratios, Li contents, and Al contents in quartz define boundaries at K/Rb = 80, Li = 25 ppm, and Al = 165 ppm, with spodumene-bearing pegmatites having lower K/Rb ratios together with higher Li and Al abundances (Fig. 6). The IQRs of K/Rb ratios of muscovite exhibit a boundary at K/Rb = 40, with spodumene-bearing pegmatites having K/Rb < 40 (Fig. 7A). For garnet, a data gap between 15 and 18 wt % MnO and a boundary at ~27 wt % FeO indicate that spodumene-bearing pegmatites have garnet with MnO > 18 wt % and FeO < 27 wt % (Fig. 9A). The IQRs of ΣREE and Sn concentrations in garnet define boundaries at 0.6 and 40 ppm, respectively, with ΣREEs < 0.6 ppm and Sn > 40 ppm in spodumene-bearing pegmatites (Fig. 9C, D). The IQRs of apatite FeO, ΣREE, and U contents define boundaries at 0.45 wt %, 250 ppm, and 400 ppm, respectively, with FeO < 0.45 wt %, ΣREE < 250 ppm, and U < 400 ppm in spodumene-bearing pegmatites (Fig. 10). A gap in apatite Sr levels between 70 and 100 ppm indicates that apatites in spodumene-bearing pegmatites have >100 ppm Sr (Fig. 10B). These mineral chemistry indices are summarized in Table 3.

Pegmatites in the Carolina tin-spodumene belt are poorly zoned and can be generalized as either spodumene-free or spodumene bearing, so this comprehensive assessment of mineral chemistry as spodumene indicators is useful to distinguish between truly spodumene-free pegmatites and spodumene-bearing pegmatites whose spodumene is either too small to see or weathered out in surface outcrops, both of which happen in the Carolina tin-spodumene belt. Using mineral chemistry to determine the presence of Li phases in pegmatites is not a new endeavor, with recent studies focusing on one to a few minerals (e.g., Maneta and Baker, 2019; Müller et al., 2021; Wise et al., 2024). This study includes six minerals and 12 different elements, ratios, or groups of elements (Table 3). Except for K/Rb-Li systematics in muscovite plus Li in quartz and K-feldspar, this is the first study to present these mineral chemistries as indicators for the presence of spodumene in group 1 pegmatites. In particular, this study presents data from both spodumene-free and spodumene-bearing pegmatites in order to observe any differences in mineral chemistry, rather than attempting to differentiate between group 1 and group 2 pegmatite families (e.g., Müller et al., 2021), and this study does not constrain itself to only spodumene-bearing pegmatites (e.g., Maneta and Baker, 2019).

The K/Rb-Li systematics of K-feldspar, quartz, and muscovite can be used in the Carolina tin-spodumene belt to predict spodumene mineralization in a given pegmatite, as well as the Al content of quartz, as demonstrated in Figures 5 to 7 and Table 3. Note that the coefficient of variation (1σ variation/average) of K/Rb ratios in these three minerals is generally lower than that of Li contents, indicating that, in the context of the Carolina tin-spodumene belt, the K/Rb ratios present a more useful geochemical index than Li abundance when exploring for spodumene-bearing pegmatites (Table 3). The obvious caveat is that both K and Rb need to be present in detectable abundances to be useful (i.e., see Oxford County quartz data in Fig. 13A). Muscovite K/Rb-Li systematics as they apply to Li mineralization were recently updated and discussed by Wise et al. (2024). Muscovites from spodumene-free pegmatites of the Carolina tin-spodumene belt plot dominantly in the poorly fractionated field (K/Rb > 40), with an average K/Rb ratio of 71 ± 40, and those from spodumene-bearing pegmatites plot mostly in the moderately fractionated field (10 < K/Rb < 40), with an average K/Rb ratio of 26 ± 10 (Fig. 7A). Only one muscovite data point plots in the highly fractionated field of K/Rb < 10 (Fig. 7A). Muscovites from Carolina tin-spodumene belt pegmatites are not as distinct in terms of Li content as with K/Rb ratios, displaying significant overlap in Li abundance. Muscovites from spodumene-bearing pegmatites plot in the high-Li domain (>500 ppm), with an average of 621 ± 502 ppm, more often than those of spodumene-free pegmatites, with an average of 421 ± 232 ppm (Fig. 7A).

This is the first study to look at K/Rb-Li together in K-feldspar and quartz, as recent previous studies only looked at Li or other elements besides K/Rb (Maneta and Baker, 2019; Müller et al., 2021). The additional consideration of K/Rb is important because it adds another dimension within which to assess Li content. For instance, Maneta and Baker (2019) analyzed Li content in K-feldspar, muscovite, and quartz from the spodumene-bearing Main Sill of the Moblan pegmatite in Canada. They recorded 9 to 350 ppm Li in K-feldspar, 9 to 365 ppm Li in quartz, and 567 to 6,710 ppm Li in muscovite, all from pegmatites that contain spodumene. Those Li contents are similar to those reported here in the same minerals of Carolina tin-spodumene belt pegmatites (Figs. 5–7), and the muscovite Li content >500 ppm of the Moblan pegmatite is in accordance with the Li-rich domain of Wise et al. (2024). However, the lower ends of the distribution of quartz and K-feldspar Li contents reported by Maneta and Baker (2019), if present in the Carolina tin-spodumene belt, could have either high or low K/Rb ratios, and therefore the presence of spodumene would still be in question. As such, the combination of K/Rb and Li is much more useful than just knowledge of Li in muscovite, K-feldspar, or quartz.

One of the main observations from K-feldspar K/Rb-Li data presented here is the gap in K/Rb ratios between 70 and 100, which is present in group 1 pegmatites of the Carolina tin-spodumene belt, South Dakota, and Maine (Fig. 5). Our results suggest that this break separates poorly to moderately fractionated pegmatites with at most beryl mineralization (K/Rb > 100) from highly fractionated pegmatites with Li mineralization (K/Rb < 70) and that it may be a widespread characteristic of group 1 pegmatites. Therefore, it could be a promising way to determine the presence of spodumene or other Li mineralization. Additional data from (1) barren, beryl-columbite, and beryl-columbite-phosphate pegmatites and (2) specific zones of well-zoned pegmatites would help elucidate the full nature of this feature in terms of its validity across the compositionally and mineralogically varied family of group 1 pegmatites.

Taken together, K-feldspar, quartz, muscovite, plagioclase, garnet, and apatite geochemistry are potentially useful in exploration efforts aimed at recognizing deposits of economic Li pegmatite ore (Figs. 4–7, 9, 10; Table 3). Minerals from spodumene-bearing pegmatites in the Carolina tin-spodumene belt have major, minor, and trace element compositional characteristics that are different from minerals in spodumene-free pegmatites, as discussed above. However, it is important to address the variation in these mineral chemistry data because considerable overlap exists between spodumene-bearing and spodumene-free pegmatites (Table 3). This variation could either reflect variation in the natural processes that generated both types of pegmatites or be caused by instances in which the presence of spodumene in a sample was not detected for various reasons, as addressed earlier. For instance, why do some analyses labeled spodumene-free (red) on Figure 7A have K/Rb ratios <20, which in theory should indicate spodumene? From an exploration point of view, these samples merit a more detailed assessment to determine if spodumene is truly absent from the whole sample and the whole pegmatite from which the sample was acquired. Even with such uncertainties, we observe these geochemical differences between spodumene-bearing and spodumene-free pegmatites described above in part due to the large number of analyses collected in this study.

Though the knowledge of these geochemical indicators is useful, an exploration campaign built around such data would presently require traditional EPMA and LA-ICP-MS techniques, which can be time intensive and expensive. Given the overlap in data discussed above, large numbers of analyses are required for each element or ratio to make an informed decision as to the potential for spodumene. With this in mind, analytical tools that can rapidly acquire large amounts of quantitative geochemical data in the field are ideal for Li pegmatite ore exploration. One example of such a method is handheld laser-induced breakdown spectroscopy (LIBS), a technique that recently has been shown to be useful in this specific context of muscovite in spodumene-bearing pegmatites of the Carolina tin-spodumene belt (Harmon et al., 2023). Quantitative field analysis using matrix-matched calibrations on handheld LIBS analyzers is well-suited for lighter elements such as Li (Fabre et al., 2022) but can also be applied to K and Rb (Wise et al., 2022b; Harmon et al., 2023). For the heavier elements in Table 3, including Mn, Fe, Sr, Sn, the REEs, Th, and U, matrix-matched calibrations are possible for either handheld LIBS or portable XRF analyzers, but such a calibration for a lighter element like Al is better suited on a handheld LIBS, even if theoretically possible on a portable XRF (Lemière and Harmon, 2021). Like calibrations for other analytical techniques, their utility and quality are dependent on the standards used, the range in concentrations for the elements of interest in both the standards and unknown samples, the ability to acquire data on the unknown sample in terms of mineral size versus handheld LIBS or portable XRF spot size, and the analyzer limits of detection for different elements.

Using such field tools for Li pegmatite exploration in conjunction with the geochemical knowledge presented in this study—especially handheld LIBS, which is better for Li detection and quantification—has the potential to transform Li pegmatite exploration (e.g., Harmon, 2024). Being able to confidently determine the presence of spodumene in a pegmatite is useful when spodumene is weathered out at surface exposure, not present in a drill core sample, too small to see in thin section, or otherwise missed by an exploration geologist because of observation of a relatively small portion of the overall pegmatite or other reasons. In these instances, time-consuming and expensive laboratory analyses are typically used to determine if a pegmatite is economically viable. With field tools able to determine quantitative geochemistry in real time, combined with detailed mineral chemistry studies like this one that provide the appropriate geochemical indices based on more robust geochemical analyses like LA-ICP-MS and EPMA, exploration efforts could become much more efficient. Use of this methodology could reduce a project’s overall costs, shorten the time between discovery and production, and minimize the environmental impacts of exploration and production.

This study presents geochemical indicators for finding spodumene in group 1 pegmatites of the Carolina tin-spodumene belt, which are poorly zoned pegmatites of the albite-spodumene type, and therefore should be most applicable on similar deposits for exploration. According to the U.S. Geological Survey Mineral Commodity Summaries, the major pegmatite sources of Li in 2023 were mines in Australia, Canada, China, and Zimbabwe (U.S. Geological Survey, 2024). Australian pegmatite mines include Greenbushes, Mt. Marion, Pilgangoora, Mt. Cattlin, Wodgina, Finiss, and Mt. Holland pegmatites (Partington et al., 1995; Sweetapple and Collins, 2002; Jacobson et al., 2007; Sweetapple et al., 2019; Phelps-Barber et al., 2022). Pegmatites mined for Li in Canada include Tanco and LaCorne; those in China include the Jiada and Jiajika pegmatites; and the Bikita pegmatite is mined in Zimbabwe (Crouse and Černý, 1972; Černý et al., 2003; Mulja and Williams-Jones, 2018; Fei et al., 2020; Huang et al., 2020; Zhang et al., 2021; Li et al., 2024). This list is not comprehensive in terms of Li pegmatite mines and exploration projects that are in the advanced stage but not yet mined. When compared to the Carolina tin-spodumene belt pegmatites, these other economic group 1 pegmatites are typically more complex in terms of pegmatite zonation and Li mineral diversity, with Carolina tin-spodumene belt pegmatites having the quasi-homogeneous nature described by Černý (1990) for the albite-spodumene type of group 1 pegmatites. The Mt. Marion pegmatites are the most similar to those of the Carolina tin-spodumene belt, containing only spodumene, with no other Li minerals, and being poorly zoned (Jacobson et al., 2007). The Pilgangoora, Wodgina, Mt. Cattlin, LaCorne, Jiada, and Jiajika areas contain poorly zoned, albite-spodumene pegmatites like those in the Carolina tin-spodumene belt (Sweetapple and Collins, 2002; Mulja and Williams-Jones, 2018; Sweetapple et al., 2019; Fei et al., 2020; Huang et al., 2020; Zhang et al., 2021; Li et al., 2024). However, these pegmatites also contain lepidolite, and some pegmatites in the LaCorne area are zoned, making them slightly more complex than the Carolina tin-spodumene belt pegmatites. Nonetheless, it is notable that most of the currently mined Li pegmatites are either albite-spodumene or a slightly more complex version of albite-spodumene. The Tanco, Greenbushes, and Bikita pegmatites are more complex pegmatites in terms of mineralogical zonation. In terms of Li mineral diversity, the Greenbushes pegmatite contains spodumene and lepidolite, whereas the Tanco and Bikita pegmatites have spodumene, lepidolite, petalite, and amblygonite-montebrasite (Crouse and Černý, 1972; Partington et al., 1995; Černý et al., 2003). Therefore, in terms of applications to economic-grade Li pegmatites, most of which are not extremely different from the group 1 Carolina tin-spodumene belt pegmatites, the geochemical results provided in this study regarding indications of spodumene-bearing versus spodumene-free pegmatites might prove relevant. An interesting area of future research would be to apply the same detailed major, minor, and trace element mineral chemistry approach to other relatively simple pegmatites (e.g., Mt. Marion), slightly more complex pegmatites (e.g., Pilgangoora, LaCorne), and much more complex pegmatites (e.g., Tanco and Bikita) in a systematic manner across minerals and zones if present. This would help answer questions such as these:

1. How do the geochemical cutoffs between spodumene-bearing and spodumene-free pegmatites in the Carolina tin-spodumene belt (Table 3) compare to similarly simple albite-spodumene pegmatites in different geologic settings?

2. How do these geochemical differences in the Carolina tin-spodumene belt compare to those in zoned albite-spodumene pegmatites that contain lepidolite?

3. How do our results translate to pegmatites with intense zonation and more complex Li-phase mineralogy?

This study presents a foundation of geochemical indices for spodumene in group 1 Carolina tin-spodumene belt pegmatites on which to build similar sets of geochemical indicators at other pegmatites with similar or more complexity. This is a research area of anticipated growth with direct application to Li pegmatite ore exploration. Although specific to the Carolina tin-spodumene belt system, and with the understanding that each pegmatite system has a unique petrogenetic history, we hope that this discussion will aid in Li pegmatite ore exploration and evaluation worldwide. The ability to apply results of this study in a real-time, comprehensive assessment of multiple minerals and multiple elements in the field could save considerable time and money in exploration efforts.

This study of trace element geochemistry provides a detailed view of the magmatic evolution of pegmatites from the Carolina tin-spodumene belt in North Carolina, United States. The peraluminous Carolina tin-spodumene belt pegmatites and two-mica Cherryville Granite formed through crustal anatexis. Trace element modeling shows that Li cannot sufficiently concentrate in the residual melt to saturate spodumene through fractional crystallization of the low-Li (44 ppm) Cherryville Granite, suggesting that the spodumene-bearing pegmatites of the Carolina tin-spodumene belt did not form through RMG processes with this protolith as the parental granite. Our calculations indicate that a starting Li content of at least ~170 ppm is needed to saturate Li phases through an extreme fractional crystallization process—a value higher than the Cherryville Granite and other granites that are nominally fertile sources for Li-rich pegmatites. In addition to the lack of a direct genetic connection between the spodumene-bearing pegmatites and the Cherryville Granite by fractional crystallization, we argue that the spodumene-bearing pegmatites of the Carolina tin-spodumene belt are not related to the spodumene-free pegmatites through fractional crystallization based on trace element mineral chemistry. Similar to whole-rock compositions, plagioclase, garnet, and apatite REE contents from spodumene-bearing pegmatites are notably lower than those in the same minerals from spodumene-free pegmatites. This observation is inconsistent with a petrologic connection through fractional crystallization. Instead, we favor either a one- or a two-stage DPA petrogenesis for the formation of spodumene-bearing pegmatites of the Carolina tin-spodumene belt. Apatite REE tetrad patterns further highlight the geochemical differences between spodumene-free and spodumene-rich pegmatites. Even though TE_1,3 values >1.1 suggest high fluid activity in both types of Carolina tin-spodumene belt pegmatites, TE_1,3 values are higher in spodumene-free pegmatites and suggest differences in the nature of this fluid activity.

Muscovite and K-feldspar K/Rb-Li systematics show that Carolina tin-spodumene belt pegmatites are moderately fractionated but do not reach the higher levels of fractionation seen in some pegmatites of the Oxford County, Custer, and Keystone pegmatite fields. Quartz in spodumene-bearing pegmatites of the Carolina tin-spodumene belt has high Li contents (>200 ppm) similar to quartz from the Borborema, Oxford County, and Tres Arroyos pegmatites. Compared to garnet chemistry of Oxford County pegmatites and the Palermo No. 1 pegmatite, garnets from spodumene-bearing Carolina tin-spodumene belt pegmatites have the lowest recorded ΣREE contents, comparable only to those of the Berry-Havey pegmatite, and some of the highest Li contents (>500 ppm) that are on par with those of the Mt. Mica pegmatite. Combined with Ga content, spodumene Li/Mn, Li/Ge, and Li/Ti ratios successfully discriminate among spodumene from individual pegmatites throughout the world, including those in the Carolina tin-spodumene belt, Maine, South Dakota, Canada, and Portugal. We suggest that spodumene and garnet Ga contents reflect the unique petrogenetic history of each pegmatite.

This study provides a comprehensive methodology using the geochemistry of six minerals to infer the presence of spodumene in the Carolina tin-spodumene belt. Spodumene can be either weathered out or too small to see in surface outcrops in the field, so these chemical indices are useful exploration tools. Compared to spodumene-free pegmatites, spodumene-bearing pegmatites have (1) plagioclase with lower Ca and ΣREE contents, (2), K-feldspar, muscovite, and quartz with lower K/Rb ratios and higher Li abundances, (3), quartz with higher Al concentrations, (4) garnet with higher MnO and Sn and lower FeO and ΣREE levels, and (5), apatite with lower FeO, U, and ΣREE quantities and higher Sr quantities. Based on our analysis of mineral trace element data, we propose that spodumene-bearing pegmatites will have plagioclase with Ca < 900 ppm and ΣREE < 0.05 ppm; K-feldspar with K/Rb < 70 and Li > 15 ppm; quartz with K/Rb < 80, Li > 25 ppm, and Al > 165 ppm; muscovite with K/Rb < 40; garnet with MnO > 18 wt %, FeO < 27 wt %, Sn > 40 ppm, and ΣREE < 0.6 ppm; and apatite with FeO < 0.45 wt %, Sr > 100 ppm, U < 400 ppm, and ΣREE < 250 ppm (Table 3). Though each pegmatite system is unique, many Li pegmatite ore deposits of economic importance are of the albite-spodumene type similar to those in the Carolina tin-spodumene belt, so these chemical fingerprints derived for the Carolina tin-spodumene belt pegmatites might have widespread applicability, assuming that the pegmatites being explored are in a group 1 pegmatite field. With the ability to generate quantitative geochemical data in situ in the field, tools like handheld LIBS and portable XRF analyzers have great potential for making Li pegmatite ore exploration much more efficient and less costly.

We thank the National Museum of Natural History, Smithsonian Institution, for providing some of the samples used in this study and the laboratory managers at the Electron Microprobe Laboratory at Virginia Polytechnic Institute and State University (Lowell Moore) and the LionChron facility at Pennsylvania State University (Joshua Garber) for assisting with data collection. Thanks to Piedmont Lithium, Inc., for providing access to outcrop and drill core on their Carolina Lithium prospect in the Carolina tin-spodumene belt and for providing Carolina tin-spodumene belt samples only. Thank you also to C. Fabre of the University of Lorraine for providing spodumene samples from Canada and Portugal. We also thank Myles Felch of the Maine Mineral and Gem Museum and one anonymous reviewer for their insightful and constructive reviews of this manuscript. Funding for this research was provided by Piedmont Lithium, Inc., and the North Carolina Policy Collaboratory.

Adam Curry is a professor in the Department of Marine, Earth, and Atmospheric Sciences at North Carolina State University, where he heads the Critical Minerals Research Lab. He completed his B.A. in geology from Pomona College (2010) and his M.S. in geological sciences at the University of North Carolina at Chapel Hill (2013). He completed his Ph.D. at the University of Geneva (2020) studying ignimbrites in the San Juan Mountains, Colorado, using petrology, geochemistry, and geochronology. Currently, he researches the petrogenesis of rare element-rich pegmatites and how to use this knowledge for critical mineral exploration efforts.