Constraining the sub-arc, parental magma composition for the giant Altiplano-Puna Volcanic Complex, northern Chile.

The Andean continental arc is built upon the thickest crust on Earth, whose eruption products reflect varying degrees of crustal assimilation. In order to robustly model magma evolution and assimilation at subduction zones such as the Andes, the compositions of parental magmas feeding crustal magma reservoirs need to be defined. Here we present new olivine and clinopyroxene oxygen isotope data from rare mafic volcanic rocks erupted at the margins of the giant Altiplano-Puna Magma Body (APMB) of the Altiplano-Puna Volcanic Complex, Central Andes. Existing olivine and pyroxene δ18O values for the Central Andes are highly variable and potentially not representative of sub-arc parental compositions. However, new olivine (n = 6) and clinopyroxene (n = 12) δ18O values of six Central Andean volcanoes presented here display a narrow range, with averages at 6.0‰ ± 0.2 (2σ S.D.) and 6.7‰ ± 0.3 (2σ S.D.), consistent with a common history for the investigated minerals. These data allow us to estimate the δ18O values of sub-arc, parental melts to ca. 7.0‰ ± 0.2 (2σ S.D.). Parental melts feeding the APMB and associated volcanic centres are postulated to form in the felsic continental crust following assimilation of up to 28% high-δ18O basement rocks by mantle-derived magmas.

(average = 6.2‰, n = 39), respectively (see Supplementary Tables S2 and S3). In contrast, typical mantle rocks show very limited variations in their δ 18 O values (e.g. olivine = 5.2 ± 0.3‰; clinopyroxene = 5.6 ± 0.4‰; ref. 11 ). The spread in the Central Andean literature data suggests that the oxygen isotope ratios of some of these lavas were modified by various processes post-dating the formation of the parental melt (e.g. extensive fractionation, late-stage assimilation, mixing of isotopically diverse magmas, or alteration). The challenge, therefore, is to constrain the parental melt value, before extensive fractionation or late-stage assimilation has taken place. We aim to achieve this goal by analysing the δ 18 O values of single minerals from rare, weakly differentiated lavas with relatively low silica contents (SiO 2 = 54.6 to 57.2 wt% 12 ) and relatively low Sr and high Nd isotope ratios ( 87 Sr/ 86 Sr = 0.70554 to 0.70669; 143 Nd/ 144 Nd = 0.51234 to 0.51251; ref. 12 ). The samples selected for this study are from six individual volcanoes (La Poruña, San Pedro, Paniri, La Poruñita, Palpana and Chela), which were active at different times and are all located around the western margin of the Altiplano-Puna Magma Body (APMB) melt anomaly in the Altiplano-Puna Volcanic Complex (Fig. 1). Based on their radiogenic isotope compositions, our samples have experienced limited degrees of crustal modification (e.g. assimilation) and are therefore ideally suited to obtaining the parental δ 18 O values locked in early-formed crystals. The crystal-focused approach we employ here offers critical new insights that have until now been under-explored due to the limitations inherent in whole-rock geochemical approaches (e.g. the susceptibility of whole-rock samples to secondary alteration and the fact that the δ 18 O values of whole-rock samples represent averages of the various phases that constitute the sample, cf. 1 ). Our results therefore contribute to filling the gaps in our knowledge of subduction related parental magma compositions feeding the largest continental magma system on Earth.  19 . Inset map created using "GeoMapApp" (www.geomapapp.org) 37 .
Study area and sample selection. Common volcanic products in the Central Andes include stratovolcanoes and extensive ignimbrite deposits, but several monogenetic volcanoes of mafic character also exist 12,13 . Within the volcano-tectonic ignimbrite province of the Altiplano-Puna Volcanic Complex 14 , the location of mafic volcanism is largely confined to the borders of the partial melt anomaly termed the Altiplano-Puna Magma Body (APMB 12 ; Fig. 1). The APMB is the largest known zone of partial melting in the continental crust throughout the world, with an estimated melt volume of 500,000 km 3 and spanning a region of ca. 200 km in diameter 15,16 (Fig. 1). Based on geophysical surveys, this anomaly, located in the upper crust, shows an increasing melt fraction from its margin (ca. 4 vol%) to its centre (up to 25 vol%) (e.g. [16][17][18]. In this region, volcanoes outside the limits of the APMB are composed of lava that is more primitive than the volcanoes situated directly above the APMB 12 . The volcanoes included in this study comprise, in order of increasing eruption age, La Poruña, San Pedro, Paniri, La Poruñita, Palpana and Chela, all of which are situated within the Altiplano-Puna Volcanic Complex but peripheral to the proposed APMB reservoir 19 (Fig. 1). In this region, the ascending parental basaltic-andesite magma is thought to have avoided significant contamination by evolved melts from the APMB as demonstrated by the lowest 87 Sr/ 86 Sr and highest 143 Nd/ 144 Nd being towards the borders of the large felsic body 12,20 . The studied volcanoes (Fig. 1), together with the other Pliocene to Quaternary andesitic-to-dacitic stratovolcanoes, dacitic domes and monogenetic cones, overlie Miocene rhyodacitic-to-rhyolitic ignimbrite sheets 21 .
La Poruña (21°53′S; 68°30′W) is a well-preserved 140 m high scoria cone 100 ka in age 22 situated on the west flank of the 6000 m San Pedro stratovolcano complex (21°53′S; 68°24′W). La Poruña is composed of pyroclastic material and an extensive basaltic-andesite to andesite lava flow that extends up to 8 km to the south-west of the main vent, whereas San Pedro is a composite stratovolcano formed by two superimposed coalescent cones 21 . The entire La Poruña volcano represents a monogenetic, relatively small to medium volume and short-lived singular eruption, whose magmatic evolution has been described as a two-stage evolutionary process involving minor assimilation and fractionation, followed by selective assimilation during turbulent ascent 22 . In contrast, San Pedro is a >100 km 2 andesitic-to-dacitic volcanic field, with a long-lived (from ca. 510 ka to present) but episodic eruptive centre, whose recent mafic activity (<160 ka) is genetically similar to La Poruña 22 . Paniri (22°03′S; 68°14′W) is a stratovolcano constructed during four separate stages between 1.4 Ma to 100 ka, whose most primitive activity is represented by isolated basaltic-andesite to andesite lava flows erupted at ca. 400 kyr ago 23 . La Poruñita (21°17′S; 68°15′W), situated in the northernmost part of the projected APMB, is a scoria cone ca. 600 ka in age of about 700 m in diameter 24 , similar in shape and composition to La Poruña 12 . Palpana (21°32′S; 68°31′W) is a conical stratovolcano built up of mafic andesite lava flows. The summit of the volcanic edifice has a crater morphology (dimensions 1.8 km by 1.3 km) that is truncated by the last-erupted dome 21 . Chela volcano (21°24′S; 68°30′W) is very similar in shape and composition to Palpana. The shape and relatively monotonous composition have been related to rapid construction of the volcanic edifices at ca. 4.1 Ma for Chela and ca. 3.8 Ma for Palpana, followed by restricted erosion and limited duration of magmatic differentiation 24 .
Olivine-and pyroxene-phyric lava and scoria are ubiquitous at La Poruña, San Pedro, Paniri, La Poruñita, Palpana and Chela and vary from basaltic-andesite to andesite in composition, with whole-rock elemental and Sr and Nd isotope compositions that range from e.g., SiO 2 = 54.6 to 62.9 wt%, MgO = 1.6 to 6.1 wt%, Sr = 389 to 885 ppm, Cr = 5 to 625 ppm, 87 Sr/ 86 Sr = 0.705541(10) to 0.707656 (10), and 143 Nd/ 144 Nd = 0.512337(12) to 0.512513(50) (see 12,22,23 ). Recent work on these volcanoes utilised whole-rock elemental and Sr and Nd isotope data to construct an evolutionary model, in which limited magmatic differentiation occurred at mid-upper crustal levels 12,22 . Lavas of these selected volcanoes may thus represent the composition of parental magmas feeding volcanism within the Altiplano-Puna Volcanic Complex, as the magmas feeding these mafic eruptions largely escaped assimilation of APMB felsic melts during ascent 12 . In this study, we focussed on sample material containing suitable mafic mineral phases for single mineral oxygen isotope analysis.

Results
petrography. Basaltic-andesite lavas from La Poruña contain ca. 30 vol.% phenocrysts (plagioclase > olivine > clinopyroxene > orthopyroxene) and Fe-Ti oxides set in a microlite-rich groundmass of plagioclase and pyroxene and a small percentage of remaining glass. Olivine (up to 2.5 mm in size; ≤12 vol.%) textures include subhedral crystals, embayments and skeletal textures. Clinopyroxene is the most common pyroxene phase and occurs as euhedral to subhedral individual crystals (up to 2 mm; ≤10 vol.%) or as reaction rims on orthopyroxene phenocrysts. Besides occurring individually, clinopyroxene crystals occur as glomerocrysts with plagioclase, olivine and orthopyroxene (Fig. 2).
Our olivine and pyroxene δ 18 O values display substantially narrower ranges than the available data for the Central Andes (Fig. 3). Published olivine δ 18 O values 6-10 tend to have either relatively high (>6.5‰) or mantle-like δ 18 O values. Notably, our olivine oxygen isotope data from La Poruña, San Pedro and Chela volcanoes have among the lowest δ 18 O values (δ 18 O = 5.7% to 6.2‰) with respect to all olivine data reported thus far for the Central Andes (cf. Parinacota 6 ; Cerro Galán 7 ; San Pedro 8 ; Fig. 4). Our clinopyroxene data (δ 18 O = 6.3% to 7.2‰) overlap with the δ 18 O values previously obtained for pyroxene from the Central Andes (5.5‰ to 8.7‰ 6,[8][9][10] ) and are at the higher end of the previously reported data range (excluding one exceptionally high value reported for Toconce volcano 8 ; Fig. 3).

Discussion
Available whole-rock geochemistry for the studied samples (Table 1) reveal that clinopyroxene-phyric (e.g. PAL-02) and olivine-pyroxene-phyric lavas (e.g. POR-06) have higher SiO 2 contents than samples that only contain olivine as phenocrysts (e.g. CHE-03). It is thus possible that pyroxene crystallised at a higher crustal level than olivine and might record late-stage crustal assimilation (cf. [26][27][28]. We also note that arc lava pyroxenes frequently contain inclusions of plagioclase (which would have higher δ 18 O values) and/or oxides (lower δ 18 O values) (e.g. Fig. 2; see also Fig. 4 in Deegan et al. 26 ). Because of the very dark appearance of pyroxene under the binocular microscope, it is both difficult to determine if inclusions are present and what they are. The wider variation of pyroxene δ 18 O values in this study, compared to olivine may, therefore, be due to either late-stage crustal assimilation or inclusions of various types in the analysed material.
It is possible to estimate the δ 18 O values of the equilibrium melt by using mineral-melt fractionation factors appropriate for basaltic-andesite (SiO 2 average 55.8 wt% among our samples; Table 1). These are calculated to be www.nature.com/scientificreports www.nature.com/scientificreports/ Δ olivine-melt = −1.3 and Δ pyroxene-melt = −0.7, using the silica-based equations in Bindeman et al. 29 . Olivine with δ 18 O values of 5.7 to 6.2‰, therefore, crystallised from magma having a δ 18 O value of 7.0 to 7.5‰ (average = 7.3‰ ± 0.17, n = 6). Clinopyroxene with δ 18 O values between 6.3 and 7.2‰ similarly calculates to magma δ 18 O values of 7.0 to 7.9‰ (average = 7.4‰ ± 0.29, n = 12). These magma δ 18 O values are within error of each other but are up to 2.0‰ higher than the accepted values for normal mid-ocean ridge basalts (N-MORB) (δ 18 O = 5.4‰ to 5.8‰ 30 ) and MORB glass (δ 18 O = 5.4 to 5.8‰ 31 ). They are also higher than previously reported δ 18 O values from mantle-derived rocks in subduction zones elsewhere (e.g. δ 18 O ≤ 6.3‰ 5,26,31,32 ). Given that our samples have relatively high SiO 2 contents and Mg numbers that range from 54.6 to 57.2 wt% and 49 to 61, respectively (Table 1), they are unlikely to represent primary or primitive mantle-derived magmas. Indeed, the O-isotope data presented here suggest assimilation of e.g. high-δ 18 O felsic continental crust resulting in an 18 O-enriched parental magma.
The high calculated melt δ 18 O values presented here cannot be explained by closed-system Rayleigh fractionation (see calculated curve in Fig. 5) as this would only increase primitive δ 18 O values by 0.2 to 0.3‰ (e.g. 29 ). Pre-Mesozoic felsic metamorphic and plutonic complexes form the Central Andean basement of northern Chile at ca. 18°S to 25°S have δ 18 O values that range between 6.4‰ to 11.8‰ 33 . If it is assumed that the mantle-derived magma had a δ 18 O value of 5.7‰ (e.g. 34 ), a minimum of approximately 21% assimilation of local crust with a δ 18 O of 11.8‰ would be required to reach a magma value of 7.0‰, using simple mass balance calculations , where X is the amount of contamination as a fraction) and assuming equal oxygen content for all end-members. This estimated degree of assimilation agrees well with the estimates based on radiogenic isotope and trace element modelling using data from the same samples, which require about 12 to 28% assimilation (see Supplementary Information), in broad agreement with recent findings for the studied volcanoes (cf. ~13% to 23% 20 ). Binary mixing modelling shows that our data are best explained by interaction between primitive mantle-derived melt and continental crust with high 87 Sr/ 86 Sr ratios (>0.714) and δ 18 O values of 11.8‰ to 19.5‰ (Fig. 5), which is not unreasonable for felsic crust in the whole Central Andean region given that e.g. Damm et al. 33 reported δ 18 O values up to 15.2‰ for Precambrian basement rocks from northern Argentina. The isotope modelling so far assumes simple mixing, which probably approximates behaviour in a deep crustal hot zone, but models involving AFC would likely require greater overall assimilation for the same result, because high-δ 18 O material is removed in the cumulates. Notably, the steady increase in SiO 2 with  www.nature.com/scientificreports www.nature.com/scientificreports/ no change in δ 18 O value at ca. 7.0‰ (Fig. 5) is consistent with parental magmas that underwent closed-system fractional crystallisation after an initial stage of crustal assimilation by mantle-derived magmas.
We propose, therefore, a model of magmatic evolution for the Altiplano-Puna Volcanic Complex where mantle-derived (primitive) magmas are injected into the felsic continental crust. Upon stagnation, these mantle-derived magmas assimilated basement rocks with high-δ 18 O values and highly radiogenic Sr isotope ratios to form a parental magma with a δ 18 O value of ca. 7.0‰ (Fig. 5). Building on the model of González-Maurel et al. 12 for the western boundary of the Altiplano-Puna Volcanic Complex, parental melts ascended to mid to upper crustal storage levels, where they stalled, differentiated and fractionated 13 , avoiding significant further contamination by e.g. felsic melts derived from the APMB as these mafic melts by-passed the molten APMB body. At these crustal levels, olivine and subsequent clinopyroxene crystallisation occurred, which is consistent with recent thermobarometric estimations performed in Quaternary lavas from the southwestern border of the Altiplano-Puna Volcanic Complex 28 .
In conclusion, volcanic rocks from the most mafic volcanoes at the western border of the Altiplano-Puna Volcanic Complex of the Central Andes have the lowest reported δ 18 O values of 5.7 to 6.2‰ (average = 6.0‰, n = 6) for olivine, whereas clinopyroxene yielded higher δ 18 O values of 6.3 to 7.2‰ (average = 6.7‰, n = 12). These mineral data are consistent with crystallisation from a magma of the same O-isotope composition, allowing a robust δ 18 O estimate of 7.0‰ for the sub-arc, parental magma of the APMB and associated volcanic centres in the Altiplano-Puna Volcanic Complex. This composition may be representative of parental magmas in the wider Central Andean region.

Methods
Sample selection and preparation. In this study we analysed crystals from the least silicic materials iden-  36 and San Carlos olivine. The long-term average difference in δ 18 O values of duplicates of MON GT is 0.15‰, which corresponds to a 2σ S.D. value of 0.15‰. Laser fluorination data are given in Table 1. All analyses gave gas pressures of O 2 that were consistent with ~100% conversion of mineral to O 2.

Data availability
The authors declare that all relevant data are available within the article and its supplementary information files.