The magmatic-hydrothermal transition is key in controlling the fate of many economically important elements due to the change in partitioning when melt and magmatic fluid coexist. Despite its increasing economic importance, the behavior of lithium (Li) in such environments remains poorly known. We illustrate how compositionally unusual biotites from the rhyolitic Bishop Tuff (California, USA) and Kos Plateau Tuff (Greece) may contain a magmatic volatile phase trapped between layers of biotite crystals. Despite originating in pristine deposits and showing the expected X-ray diffraction spectra, these biotites return low (<95 wt%) analytical totals via electron microprobe (EMP) consistent with the presence of considerable amounts of light elements (non-measurable by EMP). Lithium contents and isotope ratios in these biotites are remarkable, with abundances reaching >2300 ppm, exceptionally light Li isotopic compositions (δ7Li as low as –27.6‰), and large isotopic fractionation between biotite and corresponding bulk samples (Δ7Libt–bulk as low as –36.5‰). Other mineral phases, groundmass glass, and melt inclusions from the same units do not support an extremely Li-rich melt prior to eruption. Biotites from phonolitic systems (Tenerife [Canary Islands] and Campi Flegrei [Italy]) do not show such extreme compositional differences, with biotite and melt showing roughly equivalent Li contents, underscored by significantly reduced Δ7Libt–bulk to a maximum of –10.9‰. We ascribe the difference in behavior to the near-liquidus appearance of biotite in alkaline magmatic suites, before widespread exsolution of a magmatic volatile phase in the magma reservoir, while in rhyolitic suites, biotite crystallizes at low temperature, trapping the coexisting exsolved fluid phase in the reservoir.

The ‘Green Revolution’ to address current global climate change, and growing demand for energy storage in batteries (e.g., for electric vehicles) have focused attention on the occurrence and global cycling of the elements required for this technology, such as lithium, cobalt, or tellurium. Although lithium (Li) found in economic deposits is generally thought to be sourced from silicic magmatism (Hofstra et al., 2013; Benson et al., 2017; Ellis et al., 2021), the actual mechanisms by which Li is enriched and ultimately removed from these magmas remain poorly constrained. Many economically important metals are concentrated by processes such as partitioning between silicate melt and magmatic fluids occurring at the magmatic-hydrothermal transition; therefore, understanding these processes has been a topic of intense interest from both experimental perspectives (e.g., Webster et al., 2004; Iveson et al., 2019) and natural samples (Zajacz et al., 2008). Most studies of natural materials have focused on quartz-hosted melt and fluid inclusions. However, these studies require samples in which both inclusion types were synchronously trapped, which is rare in nature (Fiedrich et al., 2020), and are complicated by the diffusivity of many phases through their host minerals, which can modify initial concentrations (e.g., Audétat et al., 2018). We investigated the mineral biotite as a recorder of the magmatic-hydrothermal transition in a suite of samples from rhyolitic and phonolitic magmas. Specifically, we focused on the abundance and isotopic composition of Li in biotites from the rhyolitic Bishop Tuff (California, USA) and Kos Plateau Tuff (Greece) previously inferred to contain magmatic fluids (Hildreth, 1977; Bachmann, 2010). These biotites yield low electron microprobe totals (<95 wt%) and are hereafter termed “low-total biotites” (LTBs). For comparison, we also investigated biotites that yielded normal (95–97 wt%) analytical totals (hereafter abbreviated to NTBs) from a rhyolitic system with involvement of cumulate melts (Caetano Tuff, Nevada, USA; Watts et al., 2016) and alkaline suites (Astroni pyroclastics and Campanian Ignimbrite from the Campi Flegrei, Italy [Forni et al., 2018] and the Granadilla ignimbrite from Tenerife, Canary Islands [Bryan, 2006]).

Pure biotite fractions were separated from bulk samples by crushing and handpicking under a binocular microscope. For in situ mineral analysis (all carried out at the Institute of Geochemistry and Petrology, ETH Zürich, Switzerland), biotites and groundmass glasses were embedded in epoxy, polished, and carbon coated prior to imaging using a JEOL JSM-6390 scanning electron microscope (SEM). Following imaging, samples were analyzed using JEOL JXA 8200 and 8230 Superprobe electron microprobes with conditions similar to those reported by Neukampf et al. (2019) for groundmass glass and biotites analyzed with operating conditions of 15 kV, 15 nA, and a 10 µm spot. Laser ablation–inductively coupled plasma–mass spectrometry (LA-ICPMS) trace elemental analyses were carried out following the methods of Neukampf et al. (2021). In addition to the traditional spot analyses, we performed rastering analyses to minimize the depth of ablation.

For bulk-mineral analyses, crystal separates were isolated from the bulk-rock samples, milled, and then analyzed with X-ray diffraction (XRD) using an AXS D8 Advance diffractometer equipped with a Lynxeye superspeed detector at the Institute of Geochemistry and Petrology, ETH Zürich. Oxygen isotope analyses were performed at the University of Cape Town (South Africa) using conventional analyses on bulk samples of 50–100 mg and laser fluorination for biotites following the methods described by Ellis et al. (2021). Lithium isotopic measurements were carried out at the Czech Geological Survey in Prague (Czech Republic) following methods outlined in Magna et al. (2004). Different aliquots of the same samples (picked material and full dissolution) were analyzed to ensure the robustness of the results. Final values are expressed relative to the standard reference material for Li isotopes, L-SVEC, with the analytical uncertainty better than 0.5‰ (2 s.d. [standard deviation]). All new data from this study, as well as the results from reference materials, are provided in the Supplemental Material1.

Biotite from all samples lacks visible alteration in hand specimens and under the binocular microscope. In backscattered electron imaging, the LTBs have distinctly swirly textures with irregular zones of brightness, while the NTBs are typically homogeneous (Fig. 1). Our data (Fig. 1) agree well with those of previous studies for both LTBs (Kos Plateau Tuff [KPT]) and NTBs (Caetano Tuff). In contrast to previous studies (Bachmann, 2010), we did not observe the high Na2O (to values >3 wt%) in LTBs from the KPT, but damage to biotite crystals is observable following microprobe analysis. Structural measurements using XRD reveal near-identical patterns between the NTBs and LTBs and indicate no additional crystalline or alteration phases are present. The broadening of the peaks in the LTB samples is consistent with a less structurally coherent biotite and consistent with the lower K + Na (a.p.f.u.[atoms per formula unit]) in the LTBs (Fig. S1 in the Supplemental Material). These data together with O isotopic compositions that indicate high-temperature equilibrium (Bachmann, 2010) argue against post-eruptive processes (e.g., Ellis et al., 2018) being responsible for the LTB compositions (Fig. S2).

Bulk-rock Li contents are unremarkable (Fig. S3), ranging from 10.7 to 33.3 ppm, and are unrelated to the biotite type (LTB versus NTB). The Li abundances of the groundmass glasses range from 2 to 70 ppm, with the highest values observed in the Astroni samples that contain NTBs and no relationship between groundmass Li content and biotite type (Fig. 2).

The Li contents of the LTBs are high, with the Kos samples ranging between 106 and 293 ppm (pumice) and between 180 and 510 ppm (granitic block) while the Bishop Tuff LTBs range from 848 to 2314 ppm. The elevated Li abundances in the LTBs were tested by both traditional point analyses and by rastering the laser across the biotite surfaces to minimize the depth effect of ablation. These methods yielded identical results (Fig. S4) without a clear relationship between biotite appearance and Li content. Bishop Tuff LTBs also contain elevated Cs (as much as 78 ppm), a feature previously interpreted as indicative of fluid involvement (Hildreth, 1977). In contrast, the NTB samples have significantly lower Li abundances (typically <25 ppm across the four NTB suites) and a more restricted range (Fig. 2A). The elevated Li contents in the LTBs are accompanied by remarkably low δ7Li, as low as −27.6‰ and never exceeding −18.3‰, while the NTB samples have δ7Li between −9.5‰ and 0.7‰ (Fig. 3). Within the KPT suite, the δ7Li of volcanic and granitic samples varies in bulk rocks by ~7.5‰ and in biotites by ~7‰, which may reflect the different degassing regimes of these samples and highlight the complexity of Li studies in the magmatic environment. The Δ7Libt–bulk (the difference between the δ7Li values of biotite and bulk rock) ranges from −18‰ to −36.5‰ in the LTBs, while it is ~−10‰ at maximum in the NTB units (Fig. 3A).

With lithium thought to be lost from the melt during syn-eruptive degassing (Neukampf et al., 2021), melt inclusions may have the potential to constrain the Li content of the melt prior to eruption. In the Bishop Tuff melt inclusions (Dunbar and Hervig, 1992; Wallace et al., 1999; Myers et al., 2019; Fig. 2B), Li and H2O contents broadly covary while showing little relationship to Rb contents (Fig. S5), suggesting that Li may be lost through the host quartz following entrapment. Considering only inclusions with water contents >5 wt% (i.e., minimally degassed), the average Li content is 86.2 ± 30.2 ppm (n = 63, 1 s.d.) with a maximum of 206 ppm. Melt inclusions from the KPT are within a similar range, emphasizing that the systems producing the LTBs did not contain extremely Li-enriched melts prior to eruption (Fig. S5).

Although the Li abundance and δ7Li values of the LTBs are extreme compared to those of the NTBs, it is noteworthy that previous studies have suggested that fluids exsolved from silicic magmas (Richard et al., 2018; Fiedrich et al., 2020) and pegmatites (Teng et al., 2006; Fan et al., 2020) trend toward high Li contents and low δ7Li. In pegmatites, Fan et al. (2020) interpreted the ranges of Li abundances and isotopic ratios as resulting from melt-fluid separation whereby the Li-rich pegmatites were generated in a H2O-rich and silicate-poor system; they proposed that 7Li was enriched in the more strongly bonded residual melt while 6Li was preferentially removed into the fluid. We argue that a similar process occurs in the LTBs, entrapping a Li-rich fluid preferentially residing between the biotite layers. Such a scenario explains the fragility of the LTBs under the microprobe beam, the unusual geochemistry, and the lack of readily visible fluid inclusions in conventional imaging. Bivariate plots of biotite compositions with Li (as a proxy of fluid involvement) against other elements are intriguing (Fig. S6). The variable behavior of elements of different geochemical affinities (e.g., potentially fluid-mobile elements such as Na, K, and Rb and those likely to be immobile such as La) indicates the complexity of fluid-melt-crystal partitioning in silicic magmatic systems and warrants further scrutiny.

Although medium- to low-K silicic magmas may or may not contain LTBs, such behavior has not been observed from high-K alkaline suites. In alkaline suites, biotite is typically a near-liquidus phase, as experimentally shown for Tenerife phonolites in which it may be the liquidus phase (Andújar et al., 2008). At Campi Flegrei, biotite is recognized as a high-temperature phase in magmatic evolution (Stock et al., 2016; Forni et al., 2018; Fig. 4A). Indeed, we note that the results from the Astroni pyroclastics and Campanian Ignimbrite samples are characteristic of the biotite-glass relationship in the Campi Flegrei, and data from 14 other deposits show similar features both for groundmass and biotite Li contents (Fig. S8).

In contrast, in medium- to low-K magmatic suites, biotite crystallizes much closer to the solidus, if at all. In a suite of water-saturated experiments (diamonds and dotted line in Fig. 4B) using a medium-K Adamello batholith (Italy) tonalite as starting material, Marxer and Ulmer (2019) showed biotite appearance at ~750 °C. To further investigate this behavior, we used rhyolite-MELTS modeling software (Gualda et al., 2012) to chart the relative appearances of biotite and a magmatic volatile phase (MVP) in magmatic systems containing the different biotite types (conditions for the models are provided in the Supplemental Material). As expected, in the alkaline suites, biotite appears at temperatures significantly (on the order 100–150 °C) above H2O saturation (Fig. 4C). In the Caetano Tuff, melting of a sanidine-dominated cumulate (Watts et al., 2016) resulted in biotite crystallization from a drier, low-Mg# [Mg# = 100 × molar Mg/(Mg + Fe)], high-Ba melt (Fig. 1; Fig. S7) likely in the absence of a separate MVP. In the LTB-bearing systems, the biotite crystallization is delayed, which results in biotite crystallizing dominantly in the presence of an exsolved MVP (Fig. 4C). This illustrates the potential of biotite to capture such potentially Li-enriched fluids.

The record of Li-rich fluids trapped in pristine magmatic biotites from unmineralized, young volcanic deposits has important implications for the evolution of magmatic systems in the shallow crust and the transition from magmatic to hydrothermal conditions. First, the partitioning of Li into a MVP provides a mechanism to concentrate Li, and the fate of this MVP may be an important control on the generation of economic Li deposits as either brine deposits (Munk et al., 2016) or pegmatites (Troch et al., 2021). Second, the potential for biotites to contain “invisible” fluid inclusions between crystal layers may play a role in the occurrence of “too-old” 40Ar/39Ar ages that have previously been reported both from the Kos Plateau Tuff (Bachmann et al., 2010) and elsewhere (Hora et al., 2010). Finally, the LTBs can provide unambiguous evidence of the presence of an exsolved, compressible MVP in the magma reservoir, with major consequences for many magmatic processes including the size and styles of potential eruptions from such reservoirs (Huppert and Woods, 2002; Popa et al., 2019).

1Supplemental Material. Supplemental Figures S1–S8 (additional compositional information relevant to this study), and a supplemental dataset (all new data for this study and reference materials). Please visit to access the supplemental material, and contact with any questions.

We gratefully acknowledge generous funding from the Swiss National Science Foundation projects 200021_166281 and 200020_197040 to B.S. Ellis. T. Magna contributed through the Strategic Research Plan of the Czech Geological Survey (grant DKRVO/ČGS 2018–2022). We are extremely grateful for the generous assistance of Wes Hildreth and Kathryn Watts for helping with sampling the Bishop Tuff and sharing samples of the Caetano caldera ignimbrite, respectively. We thank Martin Mangler and Alex Iveson for their thoughtful reviews that helped improved this work, and Marc Norman for efficient editorial handling.

Gold Open Access: This paper is published under the terms of the CC-BY license.