Multi-stage arc magma evolution recorded by apatite in volcanic rocks

Protracted magma storage in the deep crust is a key stage in the formation of evolved, hydrous arc magmas that can result in explosive volcanism and the formation of economically valuable magmatic-hydrothermal ore deposits. High magmatic water content in the deep crust results in extensive amphibole ± garnet fractionation and the suppression of plagioclase crystallization as recorded by elevated Sr/Y ratios and high Eu (high Eu/Eu*) in the melt. Here, we use a novel approach to track the petrogenesis of arc magmas using apatite trace element chemistry in volcanic formations from the Cenozoic arc of central Chile. These rocks formed in a magmatic cycle that culminated in high-Sr/Y magmatism and porphyry ore deposit formation in the Miocene. We use Sr/Y, Eu/Eu*, and Mg in apatite to track discrete stages of arc magma evolution. We apply fractional crystallization modeling to show that early-crystallizing apatite can inherit a high-Sr/Y and high-Eu/Eu* melt chemistry signature that is predetermined by amphibole-dominated fractional crystallization in the lower crust. Our modeling shows that crystallization of the in situ host-rock mineral assemblage in the shallow crust causes competition for trace elements in the melt that leads to apatite compositions diverging from bulk-magma chemistry. Understanding this decoupling behavior is important for the use of apatite as an indicator of metallogenic fertility in arcs and for interpretation of provenance in detrital studies.


INTRODUCTION
The chemical diversity observed in the rock record of volcanic arcs is determined by a multitude of processes operating between the magma source region and the surface. A fundamental step in producing this variability is fractional crystallization, assimilation, and melting in the lower crust which drive magmas to more evolved and hydrous compositions (Hildreth and Moorbath, 1988;Annen et al., 2006;Davidson et al., 2007). During extensive fractionation of hydrous magmas in the lower crust, amphibole (±garnet) is stabilized in the fractionating assemblage and plagioclase is suppressed (Müntener et al., 2001), resulting in melts with elevated Sr, an absence of strong negative Eu anomalies (both elements being compatible in plagioclase), and depleted Y (compatible in amphibole and garnet). Such magma evolution is promoted in strongly compressional, thickened arc crust and is commonly associated with porphyry Cu ore deposits because impeded magma ascent from the lower crust facilitates volatile accumulation and enrichment in ore-forming components (Chiaradia, 2015;Chiaradia and Caricchi, 2017). The high Sr/Y values that result can be used to provide insights into arc magma evolution (Macpherson et al., 2006;Rodriguez et al., 2007), evaluate whether a magmatic system has the potential to form a porphyry-related ore deposit in exploration (i.e., is "metallogenically fertile"; Richards, 2011;Loucks, 2014) and track crustal thickness (Chapman et al., 2015). However, this deep fractionation history may be obscured due to differentiation and mixing upon ascent to the shallow crust (Reubi and Blundy, 2009). Because arc rocks are a product of this multi-stage, polybaric process, unravelling the complete history of magmatic evolution using bulk-rock chemistry alone can be challenging.
However, accessory minerals such as apatite are capable of capturing discrete periods of melt evolution during differentiation. For example, apatite has been shown to record the Sr content of the melt at the time of its crystallization, which has been used to reconstruct host-rock compositions in provenance studies (Jennings et al., 2011;Bruand et al., 2016).
In this study, we present a novel approach for understanding arc magma evolution by combining the trace element compositions of apatites from volcanic rocks with fractional crystallization modeling. We show that apatite records both deep crustal fractionation and shallow crustal crystallization processes, and tracks parameters valuable to understanding the generation of hydrous arc magmas and metallogenically fertile arcs.

STUDY AREA
The Andes of central Chile are an ideal region for a study of regional arc magma evolution and metallogeny because volcanic rock sequences ( Fig. 1) record an extended period of subduction-related magmatism that culminated in high-Sr/Y magmatism (Sr/Y > 50) and the genesis of three major porphyry Cu deposits between 12.3 and 4.3 Ma (Perelló et al., 2012;Toro et al., 2012;Spencer et al., 2015): Los Pelambres, Rio Blanco-Los Bronces, and El Teniente (Kay and Mpodozis, 2001). This transition in magma chemistry has been attributed to shallowing of the subducting slab, linked to subduction of the Juan Fernández Ridge from ca. 26 Ma which led to substantial compression and crustal thickening that peaked at ca. 5 Ma (Stern and Skewes, 1995;Kay et al., 1999).
The studied volcanic formations are all calc-alkaline, were deposited between 124 and 3.9 Ma (Fig. 1B;Hollings et al., 2005), and Published online 17 January 2020 represent magmas that followed different evolution paths in the arc crust. The Las Chilcas, Salamanca, Farellones, and Los Pelambres Formations show predominantly low Sr/Y (<50; Fig. 2) and pre-date the period of high-Sr/Y magmatism and mineralization (Fig. 1B). The Los Pelambres Formation displays some higher Sr/Y values that have been attributed to compression from localized shallowing of the subducting slab in this region (Yáñez et al., 2001). The La Copa Rhyolite is found in an immediately post-ore diatreme (ca. 4.5 Ma; Toro et al., 2012) at Rio Blanco-Los Bronces (Fig. 1A) and exhibits high Sr/Y, typical of the magmatism that generated mineralization.

SAMPLES AND METHODS
We present major and trace element analyses of apatites from a representative set of nine volcanic rock samples covering the period between the Cretaceous and Pliocene, both pre-and postdating porphyry mineralization. Samples were collected distal from mineralization centers to minimize the effects of alteration on wholerock and mineral chemistry. All samples have porphyritic texture, with compositions ranging from basaltic andesite to rhyolite (57-71 wt% SiO 2 ; Hollings et al., 2005). The studied samples are all derived from lavas with the exception of the La Copa Rhyolite diatreme. The occurrence of apatite as inclusions in clinopyroxene and amphibole, and apatite saturation tempera-tures (Harrison and Watson, 1984) that typically exceed 900°C, suggest that apatite is a relatively early-saturating phase in the samples. Major elements in apatite were determined by electron microprobe and trace elements by laser ablation inductively coupled plasma mass spectrometry at the Natural History Museum, London (UK). Detailed rock descriptions, data tables, and additional information on the analytical methods are provided in the GSA Data Repository 1 .

APATITE TRACE ELEMENT COMPOSITIONS
Apatite Sr and Y show a broad negative correlation ( Fig. 3) and exhibit both differences between samples (inter-sample) and within-sample trends (intra-sample). Apatite Sr/Y (Sr/Y ap ) varies between 0.1 and 11 (Fig. 4A), with low Sr/Y ap (<2) present in the most mafic (<61% SiO 2 ) and felsic (>70%) samples, and higher Sr/Y ap (>2) exclusively found in the intermediary whole-rock compositions (62%-66% SiO 2 ). The highest Sr/Y ap cannot be explained by high Sr alone, but requires reduced Y concentrations (<200 ppm; Fig. 3), and thus any mechanism to explain the presence of high Sr/Y ap must account for the simultaneous enrichment in Sr and depletion in Y. A correlation between Sr/Y ap and Eu/Eu* ap is present that forms a well-defined logarithmic curve, in which samples from the Farellones and Los Pelambres Formations display higher Eu/Eu* ap (Eu/Eu* ap >0.5) and the La Copa, Salamanca, and Las Chilcas samples exhibit lower values (Fig. 4A).
Experimental work shows that Mg in apatite (Mg ap ) is proportional to the Mg content of the melt (Prowatke and Klemme, 2006), which suggests that Mg ap can be used as a tracer of magmatic differentiation during apatite crystallization. Using this approach, our data reveal three systematic intra-sample trends in Sr/Y ap with magmatic differentiation (Fig. 4B). Within samples, decreasing Sr/Y ap with differentiation (decreasing Mg ap ) occurs where Sr/Y ap is initially elevated. Conversely, where Sr/Y ap is initially low, it remains low with differentiation.
Trends in igneous apatite chemistry can reflect: (1) changes in primitive melt composition; (2) changes in the trace element budget of the melt during fractional crystallization; (3) varying thermodynamic and compositional influences on apatite-melt partitioning; and/or (4) re-equilibration during near-solidus magma storage or hydrothermal alteration. Apatites analyzed here are predominantly fluorapatite with light rare earth element enrichment and 1 GSA Data Repository item 2020086, supplementary chemical diagrams, images, modeling methodology, and sample descriptions, Dataset DR1 (analyses of Durango apatite standard), and Dataset DR2 (whole-rock electron probe microanalysis and laser ablation-inductively coupled-mass spectrometry apatite data), is available online at http://www.geosociety. org/datarepository/2020/, or on request from editing@ geosociety.org.   negative Eu anomalies (Data Repository), consistent with an igneous origin (Piccoli and Candela, 2002). Cathodoluminescence images show igneous zoning features and not the irregular patchy zoning, porous textures, and secondary rare earth element mineral growth that characterize hydrothermal origins or alteration (Bouzari et al., 2016). Hence, we preclude hydrothermal effects as a major control on the observed trends.
Although melt composition and temperature influence apatite-melt partition coefficients, they do not significantly affect partition coefficient ratios. For example, (D Sr /D Y ) ap (D-distribution coefficient) does not vary systematically with changing melt composition ([D Sr /D Y ] ap = 0.5-1.0; Prowatke and Klemme, 2006) and therefore would have little systematic effect on Sr/ Y ap . Although we cannot exclude compositional variation in the primitive magmas involved, the restricted range in Sr/Y of the most primitive rocks in the region (Fig. 2) suggests that source variation is unlikely to have had a significant effect on Sr/Y ap . Thus, we consider the trends in our data as having been primarily controlled by melt evolution during crustal transit.

APATITE AS A TRACER OF DEEP CRUSTAL EVOLUTION
Because intra-sample trends of increasing Sr/Y ap with differentiation are not observed (Fig. 4B), we infer that elevated Sr/Y ap signatures in apatite are produced by a magmatic process occurring prior to apatite crystallization. Higher whole-rock Sr/Y (>20; Fig. 2) of the samples with high Sr/Y ap suggests a fractional crystallization process that affected both bulkmagma and mineral chemistry. High Sr/Y melt is readily generated in the deep crust through fractionation of an amphibole (±garnet) assemblage in high-pressure, hydrous magmas (Richards and Kerrich, 2007), and the lower compatibility of Eu over Sm and Gd in amphibole at typical oxygen fugacity, f O2 , (nickel-nickel oxide [NNO]; Nandedkar et al., 2016) produces a concomitant increase in Eu/Eu* melt . Although apatite is unlikely to crystallize under such conditions and compositions (Nandedkar et al., 2014), it may inherit these elevated Sr/Y melt and Eu/Eu* melt signatures upon near-liquidus crystallization in the shallow crust.
To investigate the sensitivity of apatite chemistry to deep crustal magmatic evolution, we developed a Rayleigh trace element fractionation model (Figs. 4A and 4B; see the Data Repository for further model details). We estimated an initial melt composition from the apatite compositions of the most primitive sample (CA-29) using apatite-melt partition coefficients (Prowatke and Klemme, 2006). We modeled progressive deep crustal fractionation from this primitive melt using phase proportions appropriate for the deep crust (66% amphibole, 33% clinopyroxene; Melekhova et al., 2014) and calculated the composition of residual melts. Lastly, we calculated new apatite compositions in equilibrium with the fractionated melts at each step (Figs. 4A and 4B). The maximum Sr/Y ap and Eu/Eu* ap values in our data are produced by fractionation of ∼60% of the deep crustal assemblage from the initial primitive melt, consistent with the typical degree of deep crustal fractionation calculated from experimental data and modeling (Richards and Kerrich, 2007;Nandedkar et al., 2014).

SHALLOW CRUSTAL DIFFERENTIATION CONTROLS ON APATITE CHEMISTRY
Although inter-sample differences in Sr/Y ap and Eu/Eu* ap in the earliest-crystallizing apatites in each sample can be explained by variable deep crustal fractionation, the intra-sample variations of these values with Mg ap are likely to have resulted from apatite co-crystallization with the phenocryst assemblages forming in the shallow crust. The compatibility of Sr and Eu in plagioclase, and the abundance of plagioclase as a phenocryst phase, mean that plagioclase crystallization is a likely mechanism for lowering Sr/ Y melt and Eu/Eu* melt during shallow crustal melt evolution. We tested this hypothesis by modeling the in situ evolution of three melts (pathways A,B,and C;Figs. 4C and 4D) and the apatite compositions in equilibrium with them. These trajectories were calculated by: (1) estimating a starting melt composition from the earliest-crystallizing apatite (highest Mg ap ) on each trend; (2) crystallization of successively increasing proportions of the phenocryst phases observed in the host rocks of these apatites; and (3) recalculating the apatite compositions in equilibrium with the residual melts at each step. In each case, the crystallization of the phenocryst assemblage in the rock produced apatite compositions consistent with the observed intra-sample Sr/Y ap and Eu/Eu* ap trends. The curvature of the modeled Sr/Y ap and Eu/Eu* ap relationship is sensitive to magmatic oxidation state (Fig. 4C inset), and best reproduces the natural apatite data at two log units (±1) above the fayalite-magnetitequartz (FMQ) redox buffer, a typical value for arc magmas (e.g., Carmichael, 1991).
Thus, we suggest that apatites crystallized closest to the liquidus capture the signature of deep crustal magma evolution that is coupled to whole-rock chemistry, and that this stage defines the inter-sample variations in our data set. Conversely, later-crystallizing apatites compete for trace elements with co-crystallizing mineral phases in the shallow crust and thereby track a chemical divergence of the melt from bulkrock compositions (e.g., Miles et al., 2013), as recorded by the intra-sample trends.

IMPLICATIONS FOR APATITE AS A PETROGENETIC TOOL
The utility of apatite in provenance studies relies on its ability to reflect host-rock chemistry (Belousova et al., 2002;Jennings et al., 2011;Bruand et al., 2016). Here, we show that although higher Sr/Y ap and Eu/Eu* ap in the Los Pelambres and Farellones samples are observed, reflecting the higher whole-rock Sr/Y and Eu/Eu* of the host samples (Fig. 2), this signature is preserved only in the earliest-crystallizing, near-liquidus apatites (high Mg ap ). The La Copa Rhyolite has the highest whole-rock Sr/Y, but consistently low Sr/Y ap which we interpret to reflect late apatite crystallization (low Mg ap ) after significant shallow plagioclase crystallization in the shallow crust (Figs. 4C and 4D). Consequently, caution should be exercised when using apatite to reconstruct rock compositions (e.g., for provenance or metallogenic fertility studies), because the ability of apatite to reflect bulk-rock and/or deep crustal chemical signatures is dependent on its timing of crystallization relative to other mineral phases. Therefore Mg ap is a key parameter in helping to properly interpret igneous apatite chemistry data in petrogenetic studies.

SUMMARY AND CONCLUSIONS
We demonstrate that apatite in volcanic rocks can track crustal-scale magmatic evolution by recording both deep crustal bulk-magma differentiation and shallow crustal melt evolution. Apatite tracks parameters (Mg, Sr/Y, and Eu/Eu*) that indicate the development of the evolved, hydrous and oxidized magma compositions thought to be required for metallogenic fertility in arcs. We suggest that our approach is widely applicable for unravelling the composite evolution of arc magmas and studying magmatic processes conducive to porphyry ore deposit formation.

ACKNOWLEDGMENTS
We thank J. Spratt, W. Brownscombe, and E. Brugge for analytical assistance. Our modeling benefited greatly from discussions with J. Blundy. This work was financially supported by the Department of Earth Science and Engineering, Imperial College London, and by the Imaging and Analysis Centre and LODE Laboratory, Natural History Museum, London. Wilkinson, Buret, and Sievwright acknowledge funding under Natural Environment Research Council grant NE/P017452/1, "From arc magmas to ores (FAMOS): A mineral systems approach". We thank D. Brown for editorial handling and A. Miles and two anonymous reviewers for their constructive reviews.