Lithium Inventory of the Cerro Galán Volcanic System (Argentina): The Role of Magmatism as a Source for Li-Bearing Brine Deposits Open Access

Lithium (Li) is a crucial ingredient in many industrial processes including synthesis of lubricating greases, glass manufacturing, and, most importantly, rechargeable Li-based batteries (Bibienne et al., 2020). As a result of a global transition from a dependency on fossil fuels to green energy, the demand for Li has consequently increased and is predicted to further increase in the coming decades (Martin et al., 2017; Bibienne et al., 2020; Jowitt and McNulty, 2021).

Economic deposits of Li occur in three primary forms: (1) pegmatites, (2) brine deposits, and (3) volcano-sedimentary deposits. Although pegmatitic deposits are currently the most exploited (Bowell et al., 2020), most of global Li reserves lie within the other deposit types. Pegmatite mining often requires additional processing to achieve Li desired purity (e.g., roasting and leaching) relative to brine extraction, which typically consists of a more straightforward process (Flexer et al., 2018; Karrech et al., 2020). Sterba et al. (2020) estimated that 70% of global reserves are found in brine deposits, but new discoveries of extremely high grades of Li in intracaldera hydrothermally altered clay deposits (e.g., Castor and Henry, 2020; Benson et al., 2023) have heightened interest in understanding the relationship between silicic magmatism and such economic Li deposits.

Although research into the formation mechanisms and mineralization potential of pegmatites has been ongoing for decades (Černý, 1991; London, 2014; Maneta et al., 2015; London and Morgan, 2017), the understanding of the processes by which Li is sourced, mobilized, and collected in the salar and volcano-sedimentary deposits currently lags behind. In the case of brine deposits, current models require the juxtaposition of numerous interrelated factors including tectonic subsidence, igneous or geothermal activity, denudation processes, sufficient time, suitably arid climate, and a closed hydrological system (Bradley et al., 2013; Munk et al., 2016).

Volcanic rocks and magmas, particularly those chemically evolved, are commonly implicated as potential sources of Li because of their elevated Li contents (Lowenstein and Risacher, 2009; Kesler et al., 2012; Godfrey et al., 2013; Hofstra et al., 2013; Munk et al., 2016; Ellis et al., 2022a). These Li contents may be more than an order of magnitude greater than those found in mid-ocean ridge basalts (MORB), which typically have between 4 and 8 ppm lithium (Tomascak et al., 2016; Marschall et al., 2017; Chen, C., et al., 2020). These elevated Li contents result from the low to moderate compatibility of Li in the common mineral phases formed during magmatic evolution (e.g., Tomascak et al., 2016; Iveson et al., 2018; Neukampf et al., 2019, 2023). Strontium isotope compositions of clastic and evaporitic materials (i.e., rocks, sediments, fluids) in closed basins where Li-bearing salars develop are strongly dominated by volcanic rock signatures (Godfrey et al., 2013; Munk et al., 2018; Meixner et al., 2022; Sarchi et al., 2023). Lithium isotope compositions of the same materials, however, imply that low-temperature processes should occur to fractionate enough of the Li isotope compositions of brines and sediments toward heavier and lighter values, respectively (Meixner et al., 2022; Sarchi et al., 2023).

Despite the seemingly pivotal role of evolved magmas in creating economic Li deposits, many of the magmatic and post-magmatic processes that can define the Li inventory remain poorly understood. In the magmatic realm, the partitioning of Li between minerals and coexisting melt remains relatively underconstrained, particularly for the systems of the greatest interest (i.e., evolved), with few experimental studies available (Icenhower and London 1995; Iveson et al., 2018). In evolved systems where a magmatic volatile phase may exist, the major control on the behavior of Li may be exerted by the partitioning of Li between the melt and such a magmatic volatile phase. Here we use “magmatic volatile phase” similar to Troch et al. (2022) to reflect an H₂O-dominated phase that contains minor amounts of other components and may be liquid, gaseous, or supercritical in nature. Constraining this partitioning is complicated; some studies attempted to utilize coexisting fluid and melt inclusions (e.g., Zajacz et al., 2008), whereas other workers have attempted to address this experimentally (Iveson et al., 2019). An alternative source of information was exploited by Ellis et al. (2022b), who combined textural, compositional, structural, and isotopic information to suggest that biotite crystals may entrap a magmatic volatile phase rich in Li. The exotic compositions they found are consistent with crystallization models that indicate biotite is a late-crystallizing phase that grows in the temperature-compositional space where a magmatic volatile phase would likely be exsolved.

Between the magma reservoir and the surface, the degassing of volatile species has long been appreciated. Currently few experimental studies have investigated the behavior of Li during degassing, and results based on surveys of natural samples indicate that the behavior of Li in silicic melts under such conditions may be complex (Charlier et al., 2012; Neukampf et al., 2022). Posteruptive processes have been studied in a greater detail. The rapid diffusivity of Li in both mineral phases (Marschall and Tang, 2020) and melts (Holycross et al., 2018) renders it one of the few elements that may retain mobility during the timescales of posteruptive cooling. The combination of slow cooling, dehydration of the system, and groundmass crystallization can lead to Li passing into the phenocrystic phases (best demonstrated so far within plagioclase, sanidine, pigeonite, and augite) within a deposit, increasing the Li content of these phases by as much as an order of magnitude while abundance inventory of slower-diffusing elements remains unchanged (Ellis et al., 2015, 2018; Cortes-Calderon et al., 2024). Equally, in the glassy groundmass of silicic volcanic deposits, Li may be mobilized through hydration of the glass (similarly to the well-known behavior of Na; Lipman, 1965) thus transferring Li to surface waters (Ellis et al., 2022a; Sarchi et al., 2023).

In this work we use these relatively new perspectives on the behavior of Li and apply them to the Cerro Galán volcanic system in South America, one of the largest caldera systems currently known within the watershed of the actively producing Salar del Hombre Muerto, which belongs to the Li triangle (Fig. 1).

The Cerro Galán volcanic system (25.93°S, 66.92°W), which contains one of the largest exposed calderas in the Andes (~35 × 20 km; Sparks et al., 1985), is found in northwestern Argentina between the Salta and Catamarca provinces, situated in the back-arc region of the Central volcanic zone of the Andes. Deformation and magmatism in the back-arc region has been attributed to several factors: (1) the subduction of the Nazca Plate beneath the South American Plate, inducing mantle melting by addition of fluids (Kay and Coira, 2009), (2) changes in the subduction angle associated to the subduction of Juan de Fuca Ridge (Kay and Coira, 2009), (3) the presence of NW- and NE-trending transverse structures in the basement (e.g., Culampajá, Archibarca, and Olicapato-El Toro lineaments; Richards et al., 2013), and (4) uplift and melting of the crust controlled by delamination of the lower crust (Chen, J., et al., 2020).

The Cerro Galán volcanic system spatially correlates with the Archibarca lineament (Richards et al., 2013). Such a NE-trending structure also seems to control the emplacement of the La Escondida porphyry Cu-Mo deposit and the Llullaillaco volcano, alongside additional magmatic expressions within the Andean orogen (Richards et al., 2013). The volcanic units within the Cerro Galán volcanic system coincide with the watershed of Salar del Hombre Muerto and overlie Ordovician sedimentary and igneous rocks and Precambrian metamorphic basement (Fig. 1; Folkes et al., 2011a). Francis and Baker (1978) identified the Cerro Galán caldera as a major magmatic system associated with a voluminous ignimbrite that was initially studied by Francis et al. (1983) and Sparks et al. (1985).

Folkes et al. (2011a) combined field work with biotite ⁴⁰Ar/³⁹Ar geochronology and paleomagnetism to derive a new stratigraphic sequence for the Cerro Galán volcanic system that we follow in this work (Fig. 1), including the discovery of two additional units in the older portion of the stratigraphy (the Pitas and Vega members). The stratigraphy of the Cerro Galán volcanic system (Fig. 1) encompasses (1) the Toconquis Group, comprising a minimum of eight main ignimbrites, (2) pre-Cerro Galán Caldera deposits, (3) the Cerro Galán ignimbrite, and (4) post-Cerro Galán ignimbrite deposits. The latter deposits coincide with caldera resurgence events, inducing doming of the intracaldera units (Folkes et al., 2011a), with the resulting dome structure still showing active hot springs (e.g., Agua Caliente; Fig. 1). The Cerro Galán volcanic system produced an accumulated volume of near 1,300 km³ (dense rock equivalent, DRE) of ignimbrites (Table 1; Folkes et al., 2011a). The ignimbrites from the Merihuaca Formation together with the Blanco ignimbrites only account for 70 km³ (DRE) of the total volume. Although the eruption of such ignimbrites could hardly produce a significant caldera collapse, the subsequent ignimbrites are considerably more voluminous, except the Vega ignimbrite, which consists of a relatively pumice-rich thin deposit (2–3 m; Folkes et al., 2011a). The most voluminous deposit in the Cerro Galán volcanic system is the Cerro Galán ignimbrite with an estimated volume of ~630 km³ (DRE), followed by the Real Grande and Pitas ignimbrites with ~390 and 190 km³ (DRE), respectively.

The geochronology presented by Folkes et al. (2011a, c), comprising ²⁰⁶Pb/²³⁸U zircon ages and ⁴⁰Ar/³⁹Ar biotite and sanidine ages, indicates that the oldest record of volcanic activity at Cerro Galán is at 5.72 ± 0.23 Ma (²⁰⁶Pb/²³⁸U zircon age) with the eruption of the Blanco ignimbrite (Toconquis Group). The youngest dated material yielded an age of 2.00 ± 0.03 Ma for the post-Cerro Galán deposits (ages relative to Fish Canyon tuff sanidine at 28.02 ± 0.16 Ma, Renne et al., 1998). Notably, the biotite ⁴⁰Ar/³⁹Ar ages are typically older than the sanidine ⁴⁰Ar/³⁹Ar from the same unit, often by as much as 0.5 m.y. (Hora et al., 2011; Kay et al., 2011). A further complication comes from Kay et al. (2011), who observed subtly different ages for Cerro Galán deposits in different geographic sectors surrounding the caldera, with an intracaldera sample returning an age of 2.126 ± 0.017 Ma, whereas extracaldera samples in the west and north returned ages that were distinguishably younger at 2.096 ± 0.016 and 2.056 ± 0.016 Ma, respectively. Zircon ²⁰⁶Pb/²³⁸U secondary ion mass spectrometry (SIMS) ages from the system are consistent with the eruption ages determined by ⁴⁰Ar/³⁹Ar geochronology but show a tail to older ages, suggesting a prolonged storage within the upper crust (Folkes et al., 2011c).

Compositionally, the products of Cerro Galán are predominantly rhyodacitic in bulk (Francis et al., 1989; Folkes et al., 2011b; Grocke et al., 2017; Lubbers et al., 2022) and rhyolitic in groundmass glass (Folkes et al., 2011b; Lubbers et al., 2022). All ignimbrites display crystal-rich (>40% modal void-free crystallinity; Table 1) juvenile clasts with a common mineralogy (Table 1) dominated by plagioclase, biotite, and quartz, with void-free modal abundances of ~21, ~12, and ~9%, respectively (Folkes et al., 2011b). The units vary in their crystal cargo in that the older units (the Toconquis Group, Fig. 1A) contain minor amphibole (typically <1% modal; Table 1) and very little sanidine, whereas the younger units contain up to 2% (modal, void-free) sanidine and typically lack amphibole (Folkes et al., 2011b). Wright et al. (2011) described the occurrence of two pumice populations within the Cerro Galán volcanic system, with the white pumice volumetrically dominant (95% of all pumice) and the volumetrically minor gray pumice interpreted as reflecting a deeper, more volatile-rich recharge magma. Grocke et al. (2017) used experimental petrology to document a transition from early amphibole-bearing magmas to the later sanidine-dominated magmas that reflect a shallowing of the magmatic system through time, leading to the magma of the Cerro Galán ignimbrite being stored at <100 MPa and resulting in homogenization of the system. Most recently, Lubbers et al. (2022) conducted a thorough study of compositional variations both within and between crystals through the lens of diffusional chronometry. They found that the magmatic system that led to the Cerro Galán ignimbrite primarily resided in the upper crust in a chemically heterogeneous, crystal-rich, relatively cool state for an extended period. Then the system underwent homogenization and remobilization in a matter of decades before eruption. This finding is in good agreement with the earlier zircon work that indicated a lengthy period of preeruptive zircon crystallization.

Here we perform a survey of the products of this volcanic system to investigate the Li budgets of the various components of the system and the mechanisms by which they could effectively source Li that is extracted in the nearby Salar del Hombre Muerto. Following previous work, we use the term “Cerro Galán ignimbrite” to refer to a specific unit that stratigraphically overlies the Toconquis Group; together these make up the Cerro Galán volcanic system.

Bulk samples were crushed and fused into glass beads using Li₂B₄O₇/LiBO₂ flux for subsequent analyses of bulk-rock major and minor element composition. The major elemental compositions of the bulk-rock samples were analyzed by X-ray fluorescence (XRF) at ALS global (Reno, Nevada). Handpicked mineral phases and groundmass glasses of the Cerro Galán deposits were investigated first by backscattered electron (BSE) images using a JEOL JSM-6390 laser ablation (LA) secondary electron microscope (SEM), housed at ETH Zürich, to understand compositional zoning and guide later in situ analyses. Within the biotite crystals this imaging particularly focused on the variations in textures within crystals and the occurrence of mineral inclusions that were later used for geochronology. Additionally, cathodoluminescence (CL) images were acquired using a Deben Centaurus CL detector fitted to the SEM for the quartz crystals.

The abundances of major and minor elements in minerals and groundmass glass were determined with a JEOL-JXA 8230 electron probe microanalyzer (EPMA) equipped with five wavelength dispersive spectrometers (WDS) at ETH Zürich, using an acceleration voltage of 15 kV. Following the method described in Cortes-Calderon et al. (2024), glasses were measured with 10-nA beam current and 20-µm beam diameter, a challenging task because of thin glass interstices between vesicles in some samples. Feldspar was analyzed at 20 nA and 10 µm, whereas analysis of biotite used a 15-nA and 10-µm beam, which ensured a better stability and intensity without compromising precision. Data was acquired and processed using Probe for EPMA software (Donovan et al., 2021). During all analytical sessions secondary reference materials were analyzed to ensure data quality (all available in App. 1). On-peak analysis time for each element was set to maximize precision without compromising material damage. The acquisition time for elements ranged from 20 to 30 s, except F, which was acquired for 60 s in glasses and biotites to obtain good intensity statistics. Potassium and sodium were analyzed first to minimize the potential effects of alkali migration during analyses. The backgrounds were not acquired off-peak but modeled based on a mean atomic number (MAN) background correction (Donovan and Tingle, 1996) after analyzing several primary standards not containing the analytes. Following this procedure, the analysis time is exclusively for on-peak count acquisition. Intensities of Si, K, and Na counts were recorded also multiple times during single-point acquisition to monitor changes in X-ray intensity through the analysis and, if necessary, apply a time-dependency intensity (TDI) correction (Nielsen and Sigurdsson, 1981).

For glasses, in situ trace element abundances of feldspar and biotite were determined similarly to Neukampf et al. (2019) using a 193-nm Resonetics Resolution S155 laser ablation system coupled to a sector field-inductively coupled plasma-mass spectrometer (SF-ICP-MS; Element XR) at the Institute of Geochemistry and Petrology, ETH Zürich. Laser spots were always circular, and their size varied from 29 (for glasses) to 43 μm (for minerals). Each analysis consisted of 30 s of gas blank acquisition followed by 30 s of ablation with a laser energy density of ~3.5 J cm⁻² and a pulse rate of 5 Hz. The synthetic glass NIST SRM610 (Jochum et al., 2011) was used as the primary reference material for trace element quantification and instrumental drift correction; the analytical reproducibility was checked by repeated measurements of the GSD-1G (Guillong et al., 2005) and BHVO-2G, ATHO-G, and GOR128-G (Raczek et al., 2001) glass reference materials. Data were reduced using Iolite 4 (Woodhead et al., 2007; Paton et al., 2011) with SiO₂ (taken from microprobe measurements) used as internal standard. For the Element XR, long-term (>7 year) precision of trace element analyses in glass and minerals, particularly of Li, is better than 5% relative at spot sizes >43 μm (Ellis et al., 2022a). Trace element abundances in quartz were determined with an Excimer 193-nm (ArF) GeoLas (Coherent) laser system coupled to a Perkin Elmer Nexion 2000 DRC quadrupole ICP-MS, using a spot size of 40 μm, a laser energy density of 15 to 20 J cm⁻², and a pulse rate of 10 Hz. NIST SRM610 silicate glass was used as external reference, and a natural quartz crystal analyzed by Audétat et al. (2015) was used as a secondary reference material. The raw data were reduced using the software package SILLS (Guillong et al., 2008). Silicon was used as internal standard for quartz (stoichiometric value). For quartz analyses, obtained with the quadrupole ICP-MS, reproducibility of concentrations in the secondary quartz reference material for all analytical sessions in this study was on average 6% relative. Laser energy for both laser systems was routinely calibrated using an external energy meter.

Trace element analyses by solution ICP-MS were performed at the Czech Geological Survey. Aliquots of bulk-rock samples and handpicked biotite were precisely weighed in Teflon vials and dissolved in a mixture of 27 M HF and 15 M HNO₃ (6:1 v/v) at 130°C for 48 h. Subsequently, the dried residues were refluxed three times with small quantities of 15 M HNO₃ and finally equilibrated in 6 M HCl. Trace element measurements in solutions were carried out using an Agilent 7900 ICP-MS. Analysis of the reference material BHVO-2 (basalt, U.S. Geological Survey) was conducted alongside the unknowns. The element abundances closely matched those reported in the GeoReM database to within ±5% (Jochum et al., 2005). Following the methodology outlined in Magna et al. (2004, 2006), Li purification was carried out using a two-stage ion exchange chromatography. The first stage used a BioRad AG50W-X8 (mesh 200–400) cation-exchange resin in 2.1-mL columns, employing an 80% methanol-1 M HNO₃ mixture for Li elution. In the second stage, 0.6 ml BioRad AG50W-X12 (mesh 200–400) cation-exchange resin and 0.5 M HNO₃ were used for additional purification of Li.

Following SEM imaging of biotite grains to identify zircon inclusions, we used a 193-nm Resonetics Resolution S155 laser ablation system coupled to a sector-field single-collector Thermo Element XR ICP-MS for zircon U-Pb dating (Guillong et al., 2014). Laser parameters included a circular spot size of 20 μm, a repetition rate of 5 Hz, and an energy density of ~2 J cm⁻². Laser energy was routinely calibrated with an external energy meter. The ablation aerosol was mixed in the fast washout S-155 ablation cell (Laurin Technic) with carrier gas consisting of He (~0.25 L min⁻¹), and makeup gas consisting of Ar (~1 L min⁻¹) and N₂ (2 mL min⁻¹). Detailed analytical parameters can be found in Appendix 1 following the community-derived guidelines (Horstwood et al., 2016). The zircon GJ-1 was used as a calibration reference material. Validating reference materials included were AUSZ7-1 (Kennedy et al., 2014), AUSZ7-5 (von Quadt et al., 2016), AUSZ8-10 (Lucaks et al., 2021), and 91500 (Wiedenbeck et al., 1995). The data were reduced using the software Iolite 4 (Paton et al., 2010, 2011) with VizualAge (Petrus and Kamber, 2012). No common Pb correction was applied; the data-reduction strategy to obtain a U-Pb crystallization age from zircons included in a single biotite population is detailed in the results section.

Unknown ages are reported as 2s absolute. Reproducibility of validation reference materials and systematic long-term uncertainty are propagated to the weighted average uncertainty by quadratic addition (i.e., total external uncertainty). The validating reference material results show the reproducibility of the method, which is in the range of 1.0%. The long-term external uncertainty is in the range of 0.5% for ²⁰⁶Pb/²³⁸U ages and is composed of the uncertainty from the applied corrections, uncertainty of the decay constants, the lack of common Pb correction, the uncertainty on the true ²⁰⁶Pb/²³⁸U ratio of the primary standard GJ-1, and possible uncertainty from matrix effects.

The Pb isotope compositions of biotites were determined using LA-ICP-MS using an Applied Spectra Resolution-LR 193-nm ArF excimer laser coupled to a ThermoFischer Element XR ICP-MS, housed at ETH Zurich. Ablation was performed with a spot size of 74 µm, laser fluence of 3.5 J cm⁻², and repetition rate of 4 Hz for all unknowns and primary (SRM NIST612; Jochum et al., 2011) and secondary reference materials (BCR-2G, GSD-1G, BHVO-2G; Raczek et al., 2001). Prior to ablation, five cleaning pulses were used to remove surface contamination. Counting times were 40 s on the signal and 30 s on the background. The ablation cell was flushed with an He carrier gas (~0.25 L min⁻¹) and mixed with 2 mL min⁻¹ of N₂ to increase signal sensitivity. Detailed parameters can be found in Appendix 1 following the community-derived guidelines (Horstwood et al., 2016). Conversion of raw laser data into isotopic ratios was performed using the Iolite 4.5 software (Paton et al., 2011) and an in-house Excel spreadsheet. Lead isotope ratios (²⁰⁷Pb/²⁰⁶Pb, ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb, ²⁰⁸Pb/²⁰⁴Pb) were corrected for background and Hg isobaric interference on mass 204 using ²⁰⁴Hg/²⁰²Hg = 0.2299 (Zadnik et al., 1989) and then offline drift and standard corrected using the accepted values of NIST-612.

Lithium isotope analyses were performed on biotite separates and bulk-rock powders from six different units, selected based on observed differences from microprobe measurements on biotites. The analyses were conducted using a Neptune multicollector (MC)-ICP-MS (ThermoFisher Scientific, Bremen, Germany) coupled with an Aridus 2 desolvating unit (Teledyne Cetac, Omaha, USA) at the Czech Geological Survey in Prague, Czech Republic. Purified Li fractions were adjusted to within ±5% of the signal of a 10 ppb L-SVEC solution to minimize instrumental bias (Magna et al., 2004). For the determination of Li isotope compositions in unknown samples, standard sample bracketing was utilized. The final results are reported in the conventional δ notation relative to the NIST RM 8545 reference material (L-SVEC; Flesch et al., 1973) and calculated as δ⁷Li (‰) = [(⁷Li/⁶Li)_sample/(⁷Li/⁶Li) _L–SVEC −1] × 1,000. Typically, the overall reproducibility of this analytical method was better than ±0.3‰ (2s). To ensure the reliability of measurements, analyses of international reference materials BHVO-2 and JGb-1 (gabbro, Geological Survey of Japan) were also conducted. Their results are consistent with published values (Magna et al., 2004, 2014; Seitz et al., 2004, John et al., 2012; Jochum et al., 2016). Complete analytical results and analyses of reference materials are available in Appendix 1.

Oxygen isotope measurements were performed on six units from the Cerro Galán system, with each unit analyzed for quartz, biotite, and feldspar to allow high-temperature equilibrium to be tested. Oxygen isotope measurements were conducted on quartz, feldspar, and biotite crystals that were isolated from crushed fresh pumice clasts using a binocular microscope. The determination of O isotope ratios was made at the Department of Geological Sciences, University of Cape Town, South Africa. Laser fluorination was employed utilizing a 20-W NewWave CO₂ laser in accordance with the methods outlined by Harris and Vogeli (2010), except for employing 20-kPa ClF₃ as a reagent. Typical sample sizes ranged from 2 to 3 mg, consisting of one to six crystals. Oxygen isotope ratios were analyzed offline on O₂ gas using a Finnegan DeltaXP mass spectrometer in dual inlet mode. Results are presented in standard form relative to Vienna-standard mean ocean water (V-SMOW), with δ¹⁸O (‰) = [(¹⁸O/¹⁶O)_sample/(¹⁸O/¹⁶O)_V-SMOW − 1] × 1,000. Raw data were converted to the V-SMOW scale using the Monastery garnet (MON GT) in-house standard (δ¹⁸O = 5.38‰, Harris and Vogeli, 2010) calibrated against UWG2 (Valley et al., 1995). The long-term average difference between duplicates of MON GT obtained by laser fluorination (2 MON GT per session) is 0.12‰ (n = 365 pairs), corresponding to a 2s (uncertainty) value of 0.16‰.

Hydrogen isotopes and water content were measured for biotite using the method described by Vennemann and O’Neil (1993). The samples were ground up, dried overnight at 110°C, and degassed at 200°C on a vacuum line, and their water was obtained by pyrolysis. The water was reduced to hydrogen gas by using low-blank Indiana zinc (Schimmelmann and DeNiro, 1993). Water abundances (as H₂O⁺⁾ were measured from the voltage on the mass-two collector of the mass spectrometer. This was calibrated against measured volumes of standard waters analyzed alongside the unknowns. The precisions for the δD and water content data were 2‰ (1s) and 0.10 wt % (1s), respectively, based on repeated analysis of the in-house standard Serina kaolinite (SB8; Harris et al., 1999), which has an accepted δD value of −57‰ and 12.4 wt % H₂O. The in-house water standards CTMP (δD, −7.4‰) and RMW (δD, −134.1‰) were used to calibrate raw sample data to the SMOW scale and correct for scale compression, and data were adjusted to a Serina kaolinite δD value of −57‰. The average δD value of Serina kaolinite before adjustment was −63‰ (n = 3, ±1s) with a water content of 12.1 (±0.3 1s) wt %.

Pumices are typically crystal rich (42.0–57.1 modal % void-free; Table 1) and show groundmass glasses (42.9–58.0 modal % void-free; Table 1) that consist of thin films of variable thickness (<5–50 µm) surrounding vesicles (41–54 modal %; Table 1; Fig. 2). Original totals of the groundmass glasses mostly range between 96 and 98 wt % (avg 96.70 wt %, n = 231). Groundmass glasses in this study are typically high-silica rhyolites (>75 wt % SiO₂ recalculated anhydrous, same for all further reported values), with total alkali contents (Na₂O + K₂O) ranging from 7.8 to 10.0 wt % and with FeO contents typically lower than 0.6 wt %, overlapping with available published data for groundmass glasses in white pumice clasts of the Cerro Galán volcanic system (Folkes et al., 2011b; Lubbers et al., 2022). Major element compositions of bulk pumice clasts are typically lower in SiO₂ (avg 67 wt %) relative to their groundmass glasses. Although there are subtle differences between the major element composition of groundmass glasses across the studied units and published data (Fig. 2), the most marked difference is found in two of the Cerro Galán ignimbrite samples that have the highest K₂O and lowest FeO and CaO of all the glasses measured here. These Cerro Galán ignimbrite samples are from a locality near Rio de los Patos (Fig. 1) and are also distinguished by their relatively higher bulk-rock SiO₂ contents (~69 wt %) and the low Cl contents in groundmass glasses (<0.02 wt %) relative to other samples reaching 0.19 wt % (App. 1, Tables A1, A2). Published glass data showing departures toward lower K₂O contents correspond to a single sample from the Cerro Galán ignimbrite unit reported by Lubbers et al. (2022). The Cerro Galán ignimbrite unit is chemically zoned (Lubbers et al., 2022) and represents the only unit saturated significatively in alkali-feldspar in the Cerro Galán volcanic system, which explains the occurrence of low-K₂O glasses. Across the whole data set there is no relationship between the analytical total and the K₂O/Na₂O ratio, suggesting that there has been minimal hydration of these glasses. Minor oxides (MgO, MnO, P₂O₅) are in all samples at, or slightly above, detection limits. The extremely vesicular nature of the Vega unit complicated EPM analyses, and we treat these few major element analyses with caution.

The Cerro Galán ignimbrite unit has the most evolved glass compositions with respect to trace elements, showing highest Rb and Nb (372–820 and 20–22 ppm, respectively) and lowest Sr and Ba contents (24–154 and 14–80 ppm, respectively). Trace element distribution in groundmass glass can be explained by crystallization of plagioclase, biotite, Fe-Ti oxides, apatite, zircon, and alkali-feldspar, which are present in the crystal cargoes of the studied pumices (App. 2, Fig. A1). Light rare earth elements (REEs) are relatively more abundant than heavy REEs and consistently display a negative Eu anomaly, reflecting extensive plagioclase fractionation (App. 2). Bulk-rock REE compositions show a less pronounced Eu negative anomaly and higher Ba contents, which is expected from the crystal-rich nature of the studied samples, dominated by plagioclase (App. 2, Fig. A1). The overall average Li content of groundmass glasses is 55.6 (n = 186; Fig. 2). Whereas Li contents in groundmass glass from units of the Toconquis Group range from 20 to 70 ppm, samples from the Cerro Galán ignimbrite unit show the highest Li contents (>90 ppm), irrespective of their locality, in a similar fashion to other incompatible elements. Bulk-rock Li contents are lower than groundmass glasses for Middle Merihuaca, Real Grande, and Cerro Galán ignimbrite units but relatively higher for Lower and Upper Merihuaca members (Fig. 2).

Plagioclase is the most common mineral phase across the studied units, whereas sanidine is found as an accessory phase in the Toconquis Group units and in higher proportions in the Cerro Galán ignimbrite unit. Note that we only analyzed feldspars from Real Grande, Pitas, Upper Merihuaca, and Blanco units. Plagioclase shows some minor mineral inclusions and is characterized by complex zoning (resorption and overgrowth) marked by differences in brightness in BSE images (Fig. 3). Plagioclase compositions within the analyzed units range in anorthite (An) molar content from 21 to 51% (n = 143; App. 1, Table A3), whereas the composition of three analyzed sanidine grains found in Blanco and Upper Merihuaca range in orthoclase (Or) molar content from 78 to 79%. This is in good agreement with previous studies of the Cerro Galán volcanic system that reported similar ranges and near identical average values of An₃₆ (Grocke et al., 2017) and An₃₈ (Lubbers et al., 2022).

In the four units studied here, the Upper Merihuaca member has the highest Li plagioclase contents reaching 68.5 ppm and averaging 48.0 ppm (n = 20; App. 1, Table A3). The other units (Real Grande, Pitas, and Blanco) average 14.5 ppm and have a maximum of 32.7 ppm Li (n = 48; App. 1, Table A3). Plagioclase from the Upper Merihuaca member exhibits the widest range in Li contents, with cores generally displaying higher Li contents than rims without a clear relationship with plagioclase An composition. Plagioclase from the Blanco unit also exhibits Li intragrain variability, albeit to a lesser extent. In contrast, Real Grande and Pitas plagioclase grains show the lowest Li contents and least variability between cores and rims. Lithium abundances in the two sanidine crystals analyzed by LA-ICP-MS in Blanco unit are ~1 ppm, (n = 4; App. 1, Table A3) as expected from published partitioning coefficients of Li in alkali feldspar (Neukampf et al., 2023).

CL images show concentric zonation patterns with sharp and/or corroded boundaries, with some grains having bright ~50-µm-thick rims. Titanium contents within quartz crystals range from 34.7 to 80.1 ppm, with higher Ti abundances corresponding to CL-brighter regions (Fig. 4) and with all units containing a similar range of Ti contents. Lithium contents within quartz range from 13.8 to 31.0 ppm across all samples from the system (App. 1, Table A4). Most units overlap in Li contents with the exception of the Vega unit (13.8–21.4 ppm, avg. 17.4 ppm), which is distinctly lower in Li, and some analyses of the Lower Merihuaca unit that fall within the same range. Aluminium contents range from 94.6 to 132.3 ppm (avg 108.3 ± 11.6 (2s), n = 198) with older units showing a limited shift toward higher Al contents. Lithium and aluminum show a positive linear correlation in quartz grains (Fig. 4). Such behavior is expected, as Li typically charge balances Al defects in quartz (Peterková and Dolejš, 2019; Rottier and Casanova, 2021).

Departures in Li contents from such correlation are found in quartz grains from Vega, Upper Merihuaca, and Lower Merihuaca, where Li increases toward the rims of the crystals relative to their cores, with no clear correlation with CL patterns or trace elements. Note that the Upper Merihuaca unit also shows a wide range in Li contents within plagioclase grains, whereas Real Grande and Pitas units, with homogeneous Li contents in quartz, show rather homogeneous Li contents in plagioclase. Other monovalent metals, such as Na and K, are typically below limit of detection (<1 and <6 ppm for Na and K, respectively).

Biotite crystals are found in all analyzed Cerro Galán samples, typically occurring as the dominant mafic mineral phase. Biotites are typically 500 μm or larger in diameter and frequently contain abundant mineral inclusions of Fe-Ti oxides, apatite, and zircon (Fig. 5). In BSE imaging the biotites show a variety of textures that can be simplified into two prominent end members, with different proportions of these end members occurring in different units. The first end member is euhedral and homogeneously bright in BSE images, with some examples showing slightly different zones of BSE brightness (Fig. 5A). The second biotite morphology is mostly euhedral to slightly subhedral and is distinguished by generally darker BSE images and swirly textures running throughout the crystals. The swirly zones are irregular and picked out by brighter and darker regions (Fig. 5A) similar to those observed in the Bishop tuff and Kos Plateau tuff (Ellis et al., 2022b). These crystals also have abundant inclusions of Fe-Ti oxides, apatite, and zircon. End-member morphologies may coexist within a single deposit, and the different samples are characterized by differing proportions of these biotite types.

The most striking feature of the biotite data is the variation in analytical totals, ranging from 90.6 to 96.9 wt % (with all Fe considered as FeO; App. 1, Table A5). Upon recalculation of major and minor oxides based on biotite stoichiometry (eight cations, 11 oxygens and two OH sites), recalculated totals range from 93.9 to 100.3 wt %. Biotites with originally low totals (~<95.5 wt %, hereafter “low total biotites”) are related to those with swirly textures and recalculated totals lower than 99 wt % (Fig. 5A), whereas the normal total biotites are the more homogeneous grains in BSE images (Fig. 5B) with near-ideal stoichiometry (i.e., totals greater than 99 wt %). Within the low total biotite population, a general correlation exists between BSE brightness and analytical total, with the brighter areas returning higher totals.

Compositionally, distinctions are immediately apparent between the major element content of low total biotites and normal total biotites, as most major elements seem to be diluted in low total biotites relative to normal total biotites when both biotite groups are found in each unit (Fig. 5C; App. 1, Table A5). In turn, Na₂O and CaO are consistently higher in the low total biotite average composition than for the normal total biotite average composition of each unit (Fig. 5C), with the exception of Vega unit. A notable feature of the major elemental compositions is that the Lower Merihuaca unit, where the low total biotites reach the lowest analytical totals, has the highest Na₂O contents of the entire Cerro Galán magmatic sequence, with values typically a factor of three greater than the normal total biotites in the same deposit.

Trace element systematics reveal further differences between the low and normal total biotite populations. The most striking difference is found in the Li contents reaching 689 ppm in low total biotites (Fig. 6). Notably the biotites from some units contain little to no Li; for example, the Real Grande unit has values beneath the detection limit of laser ablation analyses (<1 ppm; App. 1, Table A5). Generalization between low and normal total biotites in terms of trace elemental behavior is complicated by the classification of the biotite (low or normal total biotites) coming from microprobe analyses that involve a different analytical volume than the laser ablation analyses. We note that the preliminary data of Wright et al. (2011) for biotites from the Cerro Galán ignimbrite returned grains both with low (approximating 0 ppm) and high (713 ppm) Li contents. Cesium and lead covary with Li in the low total biotites (Fig. 6B, C), with Cs and Pb considered to be fluid mobile at shallow magmatic conditions (Zajacz et al., 2008). Strontium in biotite also varies inversely with the recalculated analytical total, suggesting a similar fluid-mobile behavior. Such observations also hold when comparing bulk biotite separate analyses by dissolution, with units with higher low total biotite proportions showing higher Cs, Pb, and Sr abundances, whereas contents of other elements remain unchanged (Fig. 6D; App. 1, Table A13). Thorium, which is not a fluid-mobile element, also varies in bulk biotite analyses (Fig. 6D); however, this is not related to variations in the low/normal total biotite ratio among units. For example, units with low ratios of low/normal total biotites (e.g., Middle Merihuaca and Cerro Galán ignimbrite) show the highest and the lowest Th contents among the studied samples.

Given the small size of the zircon inclusions, the in situ analyses give complex results, with U-Pb isotope compositions for each analysis representing a mixture of zircon and its biotite host (Fig. 7). We interpreted the results in a Tera-Wasserburg space. For each sample analyzed, the age of zircon crystallization is given by the lower intercept of a regression line that includes all analyses of zircon inclusions from a single sample. Regression lines are anchored by Pb isotope composition of the host biotite (Fig. 7; App. 1, Tables A8-A10).

To demonstrate the variable isotopic effect of host biotite material entrained during in situ analyses, we computed the surface area of analyzed zircons in Tera-Wasserburg diagrams (Fig. 7B, D). This exercise demonstrates a striking correlation between the size of analyzed grain and distance from concordia, with zircons with a surface area larger than the effective area of a laser ablation pit being nearly to fully concordant in Tera-Wasserburg space. Conversely, analyses of small zircons are dominated by the Pb isotope composition of the host biotite (Fig. 7B, D). The accuracy of zircon U-Pb analyses using this method is not trivial to assess, given the potential incorporation of small volumes of radiogenic lead produced by the nonzero U content in the biotite. Therefore, when possible, we also provide the age of the single largest concordant to nearly concordant grain for a sample.

The zircon crystallization ages obtained through the lower intercept method for each of the nine samples analyzed range between 4.81 ± 0.15 and 1.70 ± 0.13 Ma (2s uncertainty). These results, although largely overlapping with the previously published SIMS U-Pb zircon and Ar-Ar biotite and sanidine ages from same units as those investigated by Folkes et al. (2011a, c), provide zircon crystallization ages that are systematically younger than the existing ones (App. 2, Fig. A3). A notable exception is the Cerro Galán ignimbrite (i.e., sample 19035 CGS), which gives a U-Pb zircon crystallization age (2.33 ± 0.09/0.10 Ma; Fig. 7B) that is indistinguishable from published value. The systematic shift toward younger ages could be related to downhole fractionation overcorrection, with the laser beam not only ablating zircon material but partially ablating the enclosing biotite. At present we are not able to correct this effect, and the accuracy of the U-Pb method in this study is below what achievable under static ablation of pure zircon material. However, we highlight here that the aim of these analyses was not to obtain the most accurate and precise age (see Folkes et al., 2011a, c) but rather to tie the biotite crystals to the most recent episode of magmatism at Cerro Galán, thus excluding a xenocrystic origin.

The Pb isotope compositions of biotite were acquired for six units of the Cerro Galán system. The very low U/Pb ratios of biotites (<0.01; Ap. Table A5) indicate negligible ingrowth of radiogenic Pb; thus, the measured Pb isotope ratios should be a reliable proxy of their host magma. Our new Pb isotope data define a rather large compositional field that encompasses Pb isotope data from Francis et al. (1989) collected on alkali feldspar of the Cerro Galán ignimbrite, bulk-rock isotopic composition of large ignimbrites from the Altiplano-Puna Volcanic Complex (Folkes et al., 2013) and other small ignimbrites from the Central volcanic zone (Schnurr et al., 2007; Murray et al., 2015). The range of Pb isotope ratios for each sample shows no correlation with stratigraphic position (Fig. 8; App. 1, Table A12) or predominant type of biotite (i.e., low versus normal total biotites). The analyses outside of the range of literature values likely represent outliers related to the large analytical uncertainty of the method (Fig. 8E, F).

Lithium abundances in bulk-rock samples show a limited range (17.1–49.7 ppm; App. 1, Table A13), whereas the range of Li contents in bulk biotite separates is significantly larger (4.35–368.6 ppm). Lithium abundances from bulk biotite separates are in excellent agreement with average in situ Li contents of biotites and the relative proportion of low and normal total biotites of each unit (Figs. 6, 9). Units with pre-dominance of low total biotites show higher Li in their bulk-rock samples relative to in situ measurements on groundmass glasses, whereas units dominated by normal total biotites have bulk-rock Li contents lower than their respective groundmass glass. The bulk-rock samples display δ⁷Li values from −10.43 to 13.41‰, whereas biotite δ⁷Li values range from −20.13 to 0.12‰. A strong relationship (r² = 0.91) exists between the Li abundances and δ⁷Li values in the biotites, whereas Li abundance and isotope composition seem to be unrelated within the bulk rock (Fig. 9).

The new O isotope compositions from this study are listed in Table 2. Quartz δ¹⁸O values ranged from 9.40 to 9.83‰, whereas those of feldspar and biotite varied between 7.61 and 8.84‰ and from 7.03 to 7.42‰, respectively (Fig. 10). The differences between mineral values are consistent with high-temperature O isotope equilibrium with Δ¹⁸O_{quartz-biotite} between 2 and 2.4 compared to expected quartz-biotite differences of 2‰ and quartz-feldspar differences of 0.7 to 1.6‰ (Bindeman, 2008). This isotopic equilibrium is found in both units whether they are dominated by normal total biotites (e.g., Cerro Galán ignimbrite-Patos, Real Grande) or low total biotites (e.g., Upper Merihuaca, Lower Merihuaca). Our results are in good agreement with literature data (Bindeman and Valley, 2002; Kay et al., 2011; Folkes et al., 2013; Fig. 10B). The δD values for bulk biotite separates range from –92 to –78‰, within the range of unaltered biotite compositions and magmatic water (Taylor, 1986). The water contents of the bulk biotite separates vary from 2.30 to 2.77 wt % and roughly track the δD values, with lower water contents typically associated with more negative δD.

The biotite compositions within the Cerro Galán volcanic system are intriguing, and by combining independent observations from numerous sources we can better understand their origin. All samples that were studied here were of fresh appearance in the field, and the presence of unaltered glasses (Figs. 1, 2) in these samples reflects lack of alteration or significant posteruption devitrification. Indeed, as noted by Hora et al. (2011) the Galán region is one of the driest places on Earth, and posteruption alteration of groundmass glass by surface-derived water is therefore unlikely. In support of this observation is the fact that the oxygen isotope relationships (Fig. 8) of coexisting phases average at Δ¹⁸O_{quartz-biotite} of 2.27‰ and Δ¹⁸O_{quartz-feldspar} of 1.38‰, is consistent with high-temperature equilibrium (Bindeman, 2008; Vho et al., 2019), regardless of the type of biotite (low or normal total biotites). Additionally, hydrogen isotope compositions of bulk biotite separates support the pristine magmatic character of the studied biotites, regardless of their nature (low or normal total biotites).

The low total biotites found in the ignimbrites of the Cerro Galán system formed during the most recent (e.g., <7 Ma) episode of volcanism and do not represent a xenocrystic component, i.e., detrital from host rocks. This is supported by the preexisting ⁴⁰Ar/³⁹Ar biotite geochronology that, although slightly older than ages derived from sanidine (see later), is within this age constraint. Perhaps more compelling evidence derives from the inclusion of zircons in biotites that range in age between 5 and 2 Ma (Fig. 7), requiring that analyzed biotites grew in the youngest episode of magmatism.

The co-occurrence of low total biotites and normal total biotites within the same deposits and the moderate and relatively homogeneous Li contents of other mineral phases indicate that the low total biotites are not inheriting their high Li contents directly from the parental melt. Strontium contents in biotites behave in a fashion similar to lithium. Webster et al. (1989) and Zajacz et al. (2008) proposed that Sr is a fluid-mobile element in shallow magmatic systems. In turn, Iveson et al. (2019) showed experimentally that Sr is a fluid-immobile element at magmatic conditions where a single fluid phase is stable. It seems that the fluid-mobile character of Sr is mediated by the separation of magmatic volatile phase at lower pressures. The elevated levels of other fluid-mobile trace elements in the low total biotites (e.g., Cs and Pb) suggest that these high Li contents come from an exsolved magmatic volatile phase trapped within the biotites as was suggested by Ellis et al. (2022b) for biotites in similar systems (e.g., Kos Plateau tuff and Bishop tuff). A key component of that model was that the low total biotites occurred where biotite predominantly crystallizes late in the evolutionary history of the system at around the same time when an magmatic volatile phase is exsolving from the melt. Experimental petrology on Cerro Galán products suggests that the phase assemblage and composition is best reproduced at relatively low pressures (50–200 MPa), log f_O2 of NNO + 1 ± 0.5 (where NNO = nickel-nickel oxide buffer), and temperatures estimated between 805° and 815°C with biotite absent in any experiment at >850°C. Temperature estimates from ilmenite-magnetite thermometry for the Cerro Galán ignimbrite and similar magmas return values of 803° to 815°C (Grocke et al., 2017) and 790° to 820°C (Folkes et al., 2011b), suggesting that the Cerro Galán volcanic system was mostly at conditions where biotite was stable. Partition coefficients between bulk biotite and bulk rock for units where low total biotites are dominant tend to be higher than ~2 (Fig. 11A) and higher than ~1.5 between in situ analyses of biotite and groundmass glass (App. 1). Such values do not represent solid-melt Li partitioning at magmatic conditions in the Cerro Galán volcanic system, because the Li inventories of low total biotites are affected by the presence of Li-bearing magmatic volatile phase within the biotite. Similar high-partition coefficients, showing strong distribution of Li into biotite, in other cool and wet magmatic systems should be treated with caution.

The co-occurrence of normal and low total biotites across the studied samples allows the potential Li elemental and isotope composition of the magmatic fluids to be discussed. Units dominated by normal total biotites may be unaffected by the incorporation of the magmatic volatile phase, and this allows us to estimate the Li content of the pure biotite in that magmatic system. The mass deficits of low total biotites after stoichiometric recalculation represent an estimate of the mass of magmatic volatile phase present in the low total biotites. Lithium contents of both the low total biotites by laser ablation and the bulk biotite dissolution data involve the mixing of biotite and entrapped fluid. By making conceptual binary mixtures of solid biotite (normal total biotites) and potential magmatic volatile phase we can place some rough constraints on the Li systematics of the magmatic volatile phase (Fig. 11B). We use three end-member compositions of bulk normal total biotites in the binary mixing (normal total biotites from Real Grande and the two Cerro Galán ignimbrite units) to account for the variability in Li contents of melts from which biotite might crystallize. Two of the normal total biotite end members have similar Li isotope compositions but differ in their Li contents (i.e., tuff lithic [TL] Cerro Galán ignimbrite Patos and Real Grande). Biotites from the Real Grande unit show the lowest Li contents in normal total biotites across the studied samples in the Cerro Galán volcanic system (avg 2.1 ppm by LA-ICP-MS and 4.3 ppm by dissolution of bulk separates), whereas biotites from the Cerro Galán ignimbrite, sampled from Rio de Los Patos locality (Fig. 1), retrieved the highest Li contents (avg 28 ppm by LA-ICP-MS and 42 ppm by dissolution of bulk separates). The groundmass glasses of the latter unit show Li contents that are twice as high as those measured in the Real Grande unit. The other normal total biotite end member (i.e., Cerro Galán ignimbrite Patos) is isotopically lighter and has higher Li contents than the other normal total biotite end members (avg 45.2 ppm by LA-ICP-MS and 59 ppm by dissolution of bulk separates).

The Lower Merihuaca unit, which typically displays biotites with a mass deficit of ~3 wt % after stoichiometric recalculation, requires a trapped magmatic volatile phase that has a considerably high Li content (~10,000 ppm; Fig. 11B) and is at least as isotopically light as its bulk biotite Li isotope composition (approximately –20‰) when using the bulk biotite composition from Real Grande as a pure normal total biotite end member. According to our model, by increasing ~tenfold the Li content of the normal total biotite end member (i.e., using bulk biotite TL Cerro Galán ignimbrite Patos compositions), there is a decrease of 12% in the required Li concentration of the trapped magmatic volatile phase to reproduce the same bulk biotite Lower Merihuaca composition (Li ~9,000 ppm in magmatic volatile phase; Fig. 11B). Even if the lowest Li concentration found within the Lower Merihuaca member by in situ measurements is used instead of the bulk composition, the mixing still requires a magmatic volatile phase of at least 8,000 ppm lithium. In contrast, the Middle Merihuaca member, characterized by a higher low total biotite proportion (16%) than the Real Grande unit but lower than the Lower Merihuaca member (96% low total biotites), requires a fluid that has at least 1,000 ppm Li and is isotopically light (approximately −10‰). The bulk biotite Cerro Galán ignimbrite Patos composition lies on the mixing line between the bulk biotite TL-Cerro Galán ignimbrite Patos end member. Whether this composition is used as a normal total biotite end member or not, it does not affect the systematics explained from the binary mixing. When it is used as normal total biotite end member, the required fluid composition remains the same to reproduce the bulk biotite Lower Merihuaca compositions (Fig. 11B). If the bulk composition reflects a mixing of the TL-Cerro Galán ignimbrite Patos normal total biotite end member and a fluid, our mixing models show that the required fluid is less than 1%. The lack of low total biotites in some of the Cerro Galán volcanic system units does not imply that the magma reservoirs that fed those eruptions were undersaturated in magmatic volatile phase. Two samples from the Cerro Galán ignimbrite unit, one from Mirador Real Grande locality and the other from Patos locality, show different low/normal total biotite proportions, which might reflect the heterogeneous distribution of the magmatic volatile phase across the magma reservoir that fed the eruptions and/or fluctuations between the onset of biotite crystallization and magmatic volatile phase exsolution.

The susceptibility of Li concentrations and isotope ratios in melts and crystals to modifications persists until the magmas undergo efficient cooling, as indicated by previous studies (Ellis et al., 2018; Neukampf et al., 2021, 2022; Cortes-Calderon et al., 2024). Consequently, sampling within a single unit may reflect Li element and isotope inventories shaped by syn- to posteruptive processes. In this investigation, emphasis was placed on glassy rock types to minimize the potential effects of protracted posteruptive cooling on Li compositions both in glasses and mineral phases. Major and trace element composition of groundmass glasses indicates that posteruptive Li loss through hydration with surface waters, as typically observed on the basis of Na-H-Li exchange in volcanic glasses (Lipman, 1965), appears to be limited. However, to comprehensively assess the effects of hydration, we recommended conducting detailed sampling with respect to lithology at various localities, targeting both basal and upper portions of each ignimbrite. It should be noted that Li contents of groundmass glasses reported in this study might not accurately represent magma reservoir compositions, as Li can be exsolved from the melt during degassing (Neukampf et al., 2021). This assertion is reinforced by the presence of Li-rich magmatic fluids (Fig. 11B), as evidenced by the composition of biotites in the Cerro Galán volcanic system, and the presence of arrested Li element profiles found in analyzed quartz and plagioclase crystals.

The concentration of Li in quartz typically is linked to its Al content (Peterkova and Dolejs, 2019; Rottier and Casanova, 2021, and references therein). The higher the Al content in quartz, the greater the potential for Li uptake by quartz. However, the effective Li incorporation into quartz is constrained by competition with other monovalent cations, such as hydrogen, which may compensate for the charge deficit in Al defects (Rottier and Casanova, 2021). Consequently, higher H in quartz could explain deviations from the anticipated Li/Al isoatomic ratios in quartz toward lower Li contents for a given Al abundance (Fig. 4). Although significant departures toward lower Li contents relative to Al abundances in quartz have been observed in the Kos Plateau tuff (Fiedrich et al., 2020), a cool and wet rhyolite, hot and dry rhyolites such as the Huckleberry Ridge tuff typically show Li contents in quartz with near-expected values to charge-balance Al defects (Ellis et al., 2018; Neukampf et al., 2019). Interestingly, Li inventory in quartz grains from the Cerro Galán volcanic system, an evolved, cool (790°–820°C, Grocke et al., 2017), and wet (up to 6.6 wt % H₂O; Wright et al., 2011) system, does not entirely mirror what is observed in quartz from the Kos Plateau tuff. We attribute this inconsistency to the potential modification of hydrogen inventory in quartz during the degassing of melts, which facilitates the entry of Li into the quartz structure (Jollands et al., 2020) and results in distinct Li diffusion profiles within quartz grains. We correlate the varying Li distributions in quartz from the Cerro Galán volcanic system with different decompression rates among units. Faster decompression and subsequent cooling may preserve arrested diffusion profiles, whereas slower decompression rates could facilitate the reequilibration of quartz crystals. Nevertheless, there exists an upper limit to Li intake controlled by the Al content of the quartz. Notably, Li diffusion in quartz is among the fastest reported for magmatic phases (Jollands et al., 2020), potentially promoting the overprint of preeruptive magmatic Li compositions in quartz crystals. Apparent partition coefficients between quartz and groundmass glass might be strongly modified and might not represent magma reservoir systematics, as it would be representative of either partly degassed melt conditions or disequilibrium between quartz cores that did not fully reequilibrate with the partly degassed melt.

Apparent Li partition coefficients between plagioclase and groundmass glass (⁠$DLiplag/melt$⁠) agree with experimentally derived $DLiplag/melt$ by Iveson et al., (2018) when plagioclases show homogeneous Li contents (i.e., $DLiplag/melt=0.2−0.3$⁠). Apparent $DLiplag/melt$ from the Upper Merihuaca unit (avg 1.5) suggests that Li in plagioclase is not in equilibrium with the groundmass glass. Lithium inventories in plagioclase from our studied samples could potentially record diffusion processes, similar to quartz. Despite differences in absolute Li diffusion coefficients between quartz and plagioclase (Giletti and Shanahan, 1997; Jollands et al., 2020), units displaying homogeneous Li distribution in quartz also record homogeneous plagioclase contents, whereas units exhibiting rim to core variations in quartz also show variations in Li contents between rims and cores of plagioclases. In contrast to quartz grains, plagioclase crystals appear to lose rather than gain Li toward the rims, similar to what has been detailed in pumice samples from the Mesa Falls tuff (Neukampf et al., 2019, 2021, 2022). This difference could be reconciled by the potential substitution mechanisms involved in each mineral—for example, the need for charge balancing Al defects in quartz, which might not occur in plagioclase, or the Li loss in plagioclase grains at expense of Fe²⁺ oxidation in plagioclase (Neukampf et al., 2021). In order to better understand the redistribution of Li between plagioclase, quartz, and melt in the Cerro Galán volcanic system, a more extensive data set of Li isotope analyses of those phases is necessary.

The complex behavior of Li in volcanic products is highlighted by the Cerro Galán ignimbrite. The extensive data set (more than 3,700 analyses) of Lubbers et al. (2022) reveals a significant variability of Li contents in feldspar crystals (Fig. 12). Within their data set, plagioclase Li contents vary from 6.7 to 141.6 ppm (a factor of >20), and sanidine Li contents vary from below the detection limits of approximately 1 to 17.4 ppm (a similar magnitude). These large variations in Li contents occur without any significant change in contents of other trace elements in the feldspars (e.g., Sr, Rb, Ti, Ba) and, in the case of plagioclase, are independent of the anorthite content of the crystal. When we use the experimentally derived $DLiplag/melt$ of 0.25 (Iveson et al., 2018) and the average Li content in analyzed groundmass glasses from Cerro Galán ignimbrite samples, the expected average Li contents in plagioclase (in equilibrium) would be ~25 ppm, which is not the case for published data (Fig. 12). Although the chosen average Li content in groundmass glass for the Cerro Galán ignimbrite is limited to the data set (n = 67) in this study, the low-Li plagioclase population in Figure 12 seems to represent plagioclases reequilibrated with their groundmass glasses, in a similar fashion to plagioclases from Pitas and Real Grande units analyzed in this study (Fig. 12). The second plagioclase population with slightly lighter Li contents coincides with the Li inventory of plagioclases from Upper Merihuaca and could represent the same arrested disequilibrium distribution of Li in plagioclases. The last plagioclase population has the highest Li contents (Fig. 12), but decoupling of Li from other trace elemental signatures argues against this high Li abundance to primarily be controlled by a magmatic process. Instead, we interpret this variation in Li abundance to reflect posteruptive cooling rate variations in samples from different portions of the deposit. Similar features have been observed in the Snake River Plain (Ellis et al., 2018) and Iceland (Cortes-Calderon et al., 2024). Figure 12 shows the results from the Kilgore tuff as an example of an ignimbrite where differing cooling rates controlled the final distribution of Li in the deposit. An interesting point of comparison is the relative change in the Li content of crystals between Cerro Galán and other systems (Fig. 12), with Cerro Galán showing a much greater ingress of Li. We speculate that this may reflect the different tectonic regimes of the systems. Both Yellowstone and Krafla (where this behavior has been well documented) represent A-type felsic magmatic suites with low preeruptive water contents. Cerro Galán, by contrast, is a cooler (790°–820°C, Grocke et al., 2017) and water-rich (up to 6.6 wt % H₂O, Wright et al., 2011) system. If partitioning of hydrogen into feldspar is similar between the two systems, it might be expected that the feldspars from Cerro Galán had more hydrogen than the Snake River Plain and Iceland examples. Dehydration of the melt due to degassing upon eruption could induce the phenocrysts to attempt to reequilibrate with the lower water content surroundings, accomplished by exchanging hydrogen for Li during slow posteruptive cooling. The initially higher hydrogen content of the Cerro Galán feldspars would allow them to accommodate more Li, resulting in a distribution as observed in Figure 12. The plausibility of such a mechanism has recently been validated by the experimental work of Behrens (2023).

A number of geochronological studies of the Cerro Galán system have found that biotite typically yields older and less consistent ages than sanidine from the same sample (Hora et al., 2011; Folkes et al., 2011a, c; Kay et al., 2011). In these cases, the samples used for study were pristine, and so any potential posteruptive alteration preferentially affecting the biotite can be excluded. However, the retention of a magmatic volatile phase, or the dehydrated version of this fluid, within biotite crystals could help explain the older (and more variable) ages. Other too-old biotite ages have also been reported in other volcanic deposits of similar character, e.g., in the Kos Plateau tuff, (Bachmann et al., 2007), the Fish Canyon tuff (Charlier et al., 2007), and the Altiplano-Puna Volcanic Complex ignimbrites (Salisbury et al., 2011).

The Li element and isotope compositions of the magmatic volatile phase trapped in low total biotites retrieve considerably light isotopic values (down to −23‰). Although such values are common in low total biotites from other magmatic centers (e.g., Bishop tuff and Kos Plateau tuff; Ellis et al., 2022b), most experimental work and natural data on fluid Li isotope compositions suggest that magmatic fluids should be in turn isotopically heavier. Lithium isotope analyses of quartz-hosted fluid inclusions in Carlin-type gold deposits record heavier Li isotope compositions (δ⁷Li = 5.7–9.1‰ for magmatic fluids; Hu et al., 2024). Such fluid inclusions have low salinity (<6 wt % NaCl equiv) and low temperature (<350°C), and their Li element and isotope inventories should be treated carefully as they might have been affected by diffusive reequilibration with its host and external fluids (Zajacz et al., 2009; Lerchbaumer and Audetat, 2012). Experimental data on muscovite fluid isotope fractionation have shown contrasting results. Lynton et al. (2005) proposed that muscovite prefers ⁷Li more than fluid and more than quartz at temperatures <500°C, whereas Wunder et al. (2007) showed that aqueous fluids fractionate more ⁷Li than muscovite. These studies consider fluids at temperatures considerably lower than magmatic conditions at the Cerro Galán volcanic system. The samples studied here are all glassy, showing that quenching from magmatic conditions was effective and that posteruptive diffusion is not expected to affect the Li isotope compositions of low total biotites.

Ab initio molecular dynamics modeling shows that aqueous fluids between 300 and 6,000 MPa and at 726°C will tend to be isotopically heavier (Jahn and Wunder, 2009). Such models do not consider lower pressures at which shallow felsic magma reservoirs are typically stored (<200 MPa for Cerro Galán volcanic system; Grocke et al., 2017). The magma reservoirs that feed the Cerro Galán volcanic system eruptions were stored at conditions where the magmatic volatile phase most likely consists of two fluid phases—a vapor and a hydrosaline fluid (Driesner and Heinrich, 2007; Grocke et al., 2017)—which are not explored in ab initio modeling. The effect that fluid phase separation has on Li isotope fractionation at pressure-temperature conditions significant for evolved shallow magma reservoirs is still poorly understood, and perhaps this might be the reason magmatic volatile phases trapped in biotites have light isotope compositions. Lithium isotope diffusive fractionation has been shown to happen during bubble growth in melts, with strong ⁷Li depletions in the new-forming magmatic volatile phase (fractionation as large as −25‰; Watson, 2017) similar to isotopic compositions reported here for fluids trapped within the low total biotites.

The role of large caldera-forming volcanic systems: The Cerro Galán volcanic system consists of voluminous ignimbrites and lavas resulting from eruptions fed by a relatively shallow magma reservoir (Fig. 13A; Grocke et al., 2017). Eruptions of pyroclastic density currents, precursors of ignimbrites, often leave calderas with multiple fracture zones, providing pathways for magmatic and meteoric fluids to interact (Fig. 13B, C; Walter and Troll, 2001; Garden et al., 2017; Scott et al., 2022). Magmatic volatile phases with >8,000 ppm Li, as suggested by the biotites in this study, indicate that magmas in the Cerro Galán volcanic system contained Li-bearing fluids at depth throughout its nearly 7 m.y. history with the eruption of the Toconquis Group and the Cerrro Galán ignimbrites (Fig. 13B, C). Between major eruptions, such fluids could accumulate at the roof of the magma reservoir over time (Fig. 13A-C), as suggested by numerical and geophysical studies (e.g., Hurwitz et al., 2007; Jenkins et al., 2023). The occurrence of magmatic fluids is described by geophysical surveys, particularly magnetotelluric work conducted primarily in the Andes. These surveys have unveiled a general picture of electrical conductivity anomalies at various crustal depths, spanning deep, intermediate, and shallow levels beneath volcanoes in the Central volcanic zone and Southern volcanic zone in the Andes (Jenkins et al., 2023). The shallowest anomaly, typically manifesting at a few kilometers’ depth, is attributed to the existence of clay phases, whereas the deepest anomaly is associated with a water-rich melt-dominated body (Laumonier et al., 2017). An intermediate anomaly, generally situated at around 5-km depth, is interpreted as indicative of the coexistence of a silicate melt and a magmatic brine (Comeau et al., 2016; Afanasyev et al., 2018). These magmatic brines are proposed to persist for extended periods, potentially hundreds of thousands of years after volcanic activity ceases, harboring substantial quantities of magmatically derived metals (Blundy et al., 2021). It is noteworthy that the studied volcanoes in the Central volcanic zone are located in the back-arc region near or within the Li triangle, such as at Lastarria, Lascar, Uturuncu, and Paniri, with Lastarria and Lascar volcanoes being in the Olacapato-El Toro and Archibarca NE-trending lineaments, respectively. Hydrothermal systems within calderas play a crucial role in facilitating the remobilization of such exsolved magmatic volatile phase to shallower depths, where such magmatic fluids have the potential to mix with other fluids (Fig. 13B, C).

Lithium isotopes from sink to source: Although our modeled Li isotope compositions for the Li-rich magmatic volatile phase seem decoupled from brine isotope compositions (Godfrey et al., 2013), low-temperature processes in the hydrothermal system of the Cerro Galán volcanic system (like adsorption of the magmatic volatile phase) or reactions during the crystallization of secondary mineral phases (such as illite or smectite) within the caldera could lead to fractionation of Li isotope compositions. This could result in an isotopically heavier reservoir following clay formation or Li adsorption (Fig. 13B-D). Similar mechanisms have been utilized to explain how weathering of volcanic rocks contributes relatively isotopically heavier fluids to salars in the Andes (Álvarez-Amado et al., 2022; Meixner et al., 2022; Sarchi et al., 2023) and other locations (Munk et al., 2018). Araoka et al. (2014) proposed local hydrothermal activity linked to high-temperature rock-water interaction as a mechanism for acquiring δ⁷Li values within the crustal and volcanic range for evaporites in Nevada, United States. A combined effect of preferential scavenging of ⁶Li into neoformed secondary phases in hydrologically wet periods versus adsorption of Li onto ferrihydrite or smectite in dry hydrological periods, without attendant Li isotope fractionation, has been used to explain variations in δ⁷Li of evaporites, volcanic basement, and solute sources in the Salar del Hombre Muerto (Godfrey et al., 2013). Clays, especially, exhibit a preference for incorporating ⁶Li into their structure (Vigier et al., 2008; Hindshaw et al., 2019; Álvarez-Amado et al., 2022; Sarchi et al., 2023), resulting in reacted fluids with a more dilute Li concentration that is also heavier in composition (Δ⁷Li_{reacted solution-source solution} up to ~30‰; Pistiner and Henderson, 2003; Li and Liu, 2020).

Lithium isotopes have been proposed as a tracer of the potential source of Li found in the salars of South America and the hydrological systems feeding them (Godfrey et al., 2019; Meixner et al., 2022). Studies that involve Li isotope compositions of volcanic products near such salars typically show average compositions or ranges in δ⁷Li that do not exceed 5‰. The large δ⁷Li variability observed in our data set, especially considering that our samples only consider glassy rock types, calls this approach into question. The ignimbrites from the Cerro Galán magmatic system are rather similar in terms of bulk-rock compositions, mineralogy, and radiogenic (e.g., Pb, Fig. 8) and oxygen (Fig. 10) isotope composition. In contrast, the δ⁷Li values range from −10.4 to 13.4‰, a variability of more than 23‰. Indeed, even two spatially close samples from the Cerro Galán ignimbrite differ markedly in δ⁷Li by ~13‰ (Fig. 9), with one of them representing a pumice clast (lower δ⁷Li) and the other a tuff lithic (higher δ⁷Li). The difference in isotope composition between both samples might be a result of the interaction between the cold glassy lithic and the cooling of its host pyroclastic density current. Similar variations in the δ⁷Li of a deposit have been observed in ignimbrites of the Snake River Plain (USA) and rhyolitic lavas of Krafla volcanic center (Iceland). These variations were attributed to differences in posteruptive cooling histories within a single deposit allowing for degassing and redistribution of Li (Ellis et al., 2018; Cortes-Calderon et al., 2024). The existence of such isotopic variability within the potential Li source is further compounded by variable isotopic fractionation that the formation of secondary phases would produce, making direct source identification via this mechanism unlikely.

The role of deformation in back-arc systems: Back arcs encompass regions with diverse deformation styles, ranging from extension to compression, potentially leading to the formation of topographic lows at high elevations (Heuret and Lallemand, 2005). These topographic lows, or basins, may serve as depocenters for reworked material through weathering processes. The Cerro Galán volcanic system is located within the Andean back-arc region south of the Altiplano-Puna Volcanic Complex (Francis et al., 1978), where the Archibarca lineament and multiple caldera resurgence events during Cerro Galán volcanism shaped the current topography of the area (Fig. 13B-D; Folkes et al., 2011a). Two remarkable geomorphic features call for special attention: the elevated topographic peaks within the Cerro Galán volcanic system caldera (Fig. 13D) and paleobasins surrounding the caldera edifice, including those of the former Rio de los Patos and Salar del Hombre Muerto basins (Figs. 1, 13B). Importantly, these paleobasins are partly filled by the voluminous Cerro Galán ignimbrite unit and are currently linked through a hydrological system (Fig. 1).

Resurgence caldera events, such as the post-Cerro Galán ignimbrite resurgence dome within the Cerro Galán caldera (Folkes et al., 2011a), could cause an Li-rich magmatic volatile phase to rise to the surface during doming of the intracaldera units and favor downstream flow of magmatic-derived fluids to predecessor basins, like the modern-day Salar del Hombre Muerto, containing Li brines with ~650 to 800 mg L^–1 Li. (Kosinski and Cutler, 2023; Rosko et al., 2023). Similar dome-mediated fluid remobilization has been documented in other volcanic calderas, such as at McDermitt (Benson et al., 2023). The apparent lack of hydration in volcanic glasses and the Li loss observed in this study suggest that minimal posteruptive alteration of volcanic glass occurred in the Cerro Galán ignimbrite and much of the inventory in the Salar del Hombre Muerto comes from groundwater enriched in Li from Cerro Galán geothermal systems sourced from an exsolved magmatic volatile phase.

The use of biotite and other proxies for Li-rich magmatic fluids: Further work on the Li and O isotope compositions of modern hot spring systems at the Cerro Galán resurgent dome and the Aguas Calientes center on the northern ring fracture (Fig. 1) could provide further insight into the composition of those magmatic-derived fluids when reaching the surface of hydrothermal systems. Although biotite crystals in the Cerro Galán volcanic system provide a helpful vessel to trap the magmatic volatile phase, most other magmatic systems could be expected to become saturated with a magmatic volatile phase at some point in their history, and the challenge is to understand the proxies that may reveal this and how this can be exploited within an exploratory framework. For example, this study could be extended to the biotite-bearing tuffs of the Coranzuli and Cerro Aguas Calientes caldera centers (Petrinovic et al., 2010; Seggiaro et al., 2019) within the watersheds of the Li-bearing Cauchari-Olaroz and Pastos Grandes salars in Salta and Jujuy, Argentina (M. Dworzanowski, unpub. report, 2019; Burga et al., 2020).

This study represents the first detailed survey of the potential source rocks for Li deposits in the lithium triangle in the Andes. By combining microscale in situ work and bulk isotopic measurements we provide an improved understanding of the magmatic processes governing the availability of Li to the hydrological system. Groundmass glass compositions in the Cerro Galán volcanic system show no evidence of hydration by surface waters and suggest limited Li scavenging from the glasses to the hydrological system. Lithium inventories in plagioclase and quartz crystals suggest that Li contents in groundmass glass represent melt inventories after degassing during ascent of the Cerro Galán magmas. The evidence of higher Li contents in melts than those from groundmass glasses is shown in apparent arrested diffusion profiles in plagioclase grains in yet glassy rock types. Some of the units contain biotites with swirly textures in BSE imaging that return low totals (low total biotites) in microprobe analyses. These biotites tend to have extremely elevated concentrations of Li and other elements that are considered to partition into magmatic fluids (e.g., Cs, Pb). Such biotites coexist with normal total biotites, or normal total biotites within the same deposit, suggesting that whatever produces the low total biotites reflects a mechanical process rather than a process of melt-crystal equilibrium. Both types of biotites contain zircon inclusions that return U-Pb ages that are consistent with previous geochronology, indicating that these low total biotites were formed during the most recent episode of magmatism rather than being xenocrysts from crustal assimilation. Oxygen and hydrogen isotope analyses of mineral phases suggest high-temperature equilibrium. These results indicate that the unusual textures observed within the low total biotites are the result of high- (rather than low-) temperature processes. Binary mixing, using bulk Li isotope analyses of biotite, between fluids and normal total biotites suggests that Li-rich (up to 8,000–10,000 ppm Li), isotopically light (approximately −22‰), fluids are needed to reproduce the stoichiometric mass deficit of low total biotites and their element and isotopic compositions. The occurrence of low total biotites in some Cerro Galán units and not others suggests that the conditions of the magmatic system are fluctuating around those suitable to produce a dense magmatic volatile phase that can effectively sequester lithium to be later entrapped by biotite. This implies that expulsion of Li from the magmatic system in the form of brines may be pulsatory process providing lithium irregularly to the hydrological system throughout the history of a magmatic system. Our model suggests that magmatic fluids produced in back-arc regions in the Central volcanic zone, where large calderas overlap the watershed of closed basins, can provide enough Li to the hydrological system to form Li-bearing salars.

This research was made possible through the generous financial support provided by the Swiss National Science Foundation (SNSF) grants 200020_184867, 200021_166281 and 200020_197040 to B.S. Ellis and the Lithium Americas (Argentina) Corporation. T. Magna acknowledges support from the Czech Science Foundation project 23-07625S. We express our gratitude to O. Bachmann for thoughtful discussions, M. Casini and A. Bregman for their assistance in the field and to Z. Moser for her assistance during SEM sessions. Special thanks are extended to L. Meinert, D. Cooke, and A. Simon for editorial handling and to P. Ruprecht and A. Iveson for their insightful comments, which greatly contributed to refining this manuscript.

Alejandro Cortes Calderon holds a PhD in earth sciences from ETH Zurich (Switzerland) and currently works as an electron probe microanalyst at the Natural History Museum in London (UK). Alejandro is a volcanologist interested in the processes occurring deep within magma reservoirs and the chemical changes that take place as magma ascends and cools at the surface. Alejandro integrates field observations, geochemical analyses, and experimental studies to better understand the selective enrichment and transport of lithium in evolved magmas, which may lead to the formation of economically significant deposits.