This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Porphyry Cu ± Mo ± Au and iron oxide copper-gold (IOCG) deposits share many similarities (e.g., Fe, Cu, and Au contents), but also have important differences (e.g., the predominance of sulfide minerals in porphyry deposits and iron oxides in IOCG deposits). Genetic comparisons are complicated by the broad definition of IOCG deposits; here we restrict our study to IOCG deposits that are related to igneous intrusive systems. In the Mesozoic Coastal Cordillera of northern Chile, both porphyry and IOCG deposits occur in close spatial and temporal proximity, offering the chance to examine what controls their different modes of formation. From detailed examination of the timing, geochemistry, and tectonic setting of associated igneous rocks, based on new and published data, we find that rocks associated with mid-Cretaceous IOCG deposits (~125–110 Ma) are largely indistinguishable from those associated with slightly earlier (>125 Ma) and later (<110 Ma) porphyry Cu ± Mo ± Au deposits. Magmas related to IOCG deposits were formed during a brief period of back-arc transtension in the mid-Cretaceous and are, on average, somewhat more mafic (dioritic), locally alkaline, and isotopically primitive compared to granodioritic magmas associated with porphyry deposits formed during normal contractional arc tectonics in the later Cretaceous. However, these compositional ranges overlap, and the differences are not clear enough to be diagnostic.

We measured the SO3 content of igneous apatite from selected samples of these rocks to test the hypothesis that the difference in sulfur content of the ore deposits was due to differences in sulfur content of the associated magmas. Early igneous apatite crystals occurring as inclusions in silicate phenocrysts from the Carmen de Andacollo porphyry Cu-Au deposit (Re-Os molybdenite ages of 103.9 ± 0.5 Ma, 103.6 ± 0.5 Ma) are significantly richer in S (0.25 ± 0.17 wt % SO3, n = 69) than similar apatite crystals from two IOCG deposits (Candelaria, Casualidad) and a sample of regional mid-Cretaceous igneous rock from near Productora (0.04 ± 0.02 wt % SO3, n = 76). Using published partition coefficients for S between apatite and oxidized silicate melt, we semi-quantitatively estimate corresponding magmatic sulfur contents of ~0.02 wt % S in the Carmen de Andacollo magmas versus ~0.001 to 0.005 wt % S in the IOCG-associated magmas. This is an order of magnitude difference, and the opposite of what would be expected if the difference were due to bulk magma composition (sulfur solubility is generally higher in mafic magmas, whereas here the S content is higher in the more felsic porphyries). We conclude that the porphyry-forming magmas indeed had higher S contents than the IOCG-related magmas and suggest that these differences reflect different petrogenetic processes. During normal subduction, magmas derived from the metasomatized mantle wedge are hydrous, moderately oxidized, and S rich, and have the potential to generate S-rich porphyry-type deposits. In contrast, in back-arc extensional settings, upwelling asthenospheric melts carry a weaker subduction signature, including lower S contents. Interaction of these S-poor magmas with previously subduction modified upper plate lithosphere is more likely to give rise to S-poor IOCG deposits.

Introduction

Sulfur-rich porphyry Cu ± Mo ± Au deposits are formed by the precipitation of sulfide minerals (pyrite, chalcopyrite, molybdenite) from hydrothermal fluids exsolved from shallowly emplaced calc-alkaline magmas in volcanic arcs, typically generated in response to oceanic lithosphere subduction (Burnham, 1979; Richards, 2003; Cooke et al., 2005; Sillitoe, 2010). In contrast, sulfur-poor iron oxide copper-gold (IOCG) deposits include a wide range of different deposit types, broadly linked by the prevalence of hydrothermal Fe oxides (as opposed to Fe sulfides), with or without Cu and Au mineralization (Hitzman, 2000; Williams et al., 2005, 2010; Hunt et al., 2007; Groves et al., 2010). The broadness of this definition, as well as the capacity of oxidized saline fluids to transport Fe ± Cu ± Au in a variety of geologic settings, has led to controversy over the origin(s) of this group of deposits (e.g., Barton and Johnson, 1996, 2000; Pollard, 2000; Williams et al., 2005; Williams, 2010; Barton, 2014). However, within this group there is a subset of deposits that is more clearly associated with igneous rocks and fluids of possible magmatic-hydrothermal origin (Pollard, 2000; Sillitoe, 2003). Richards and Mumin (2013a, b) have referred to these as magmatic-hydrothermal IOCG deposits, but, for simplicity, we use the general term IOCG below. They share many features with porphyry systems (e.g., Cu-Au-Mo-Fe metal association and broad tectonic setting and magmatic affinity) but also have important differences (such as more extensive development of high-temperature sodic, sodic-calcic, and potassic iron alteration envelopes, and more restricted development of lower-temperature phyllic and argillic alteration in IOCG systems compared to porphyry systems; Hitzman et al., 1992; Mumin et al., 2010).

Richards and Mumin (2013b) explained some of these differences, in particular the smaller acidic alteration zones in IOCG deposits, as reflecting lower abundances of S (SO2) in the ore-forming fluids compared to porphyry fluids. As S-rich porphyry fluids cool, the SO2 disproportionates to form H2S and H2SO4 (sulfuric acid), leading to widespread development of acidic alteration at shallow levels (Burnham, 1979; Candela, 1992; Field et al., 2005; Richards, 2011b, 2015); this happens to a lesser extent in S-poor IOCG fluids. Another important difference is the broader range of metals found in IOCG versus porphyry deposits, including, in some deposits, the presence of abundant U, rare earth elements (REEs), P, Co, Ni, and Bi. The increased variety of metals is attributed to the much greater infuence of fluid reactions with crustal rocks and the resultant metal fluxing that occurs in giant IOCG hydrothermal systems (in addition to magmatic contributions), and the more common occurrence of mafic country rocks around many IOCGs (Somarin and Mumin, 2012; Richards and Mumin, 2013b; Barton, 2014).

The Mesozoic Coastal Cordillera of northern Chile is uniquely suited to compare the characteristics and controls on ore formation in porphyry and IOCG deposits because both deposit types occur in a broadly coeval (Cretaceous), 50- to 80-km-wide belt that runs parallel to the coast for over 1,000 km (Fig. 1a; Sillitoe, 2003). The largest IOCG deposits within this part of the belt are Candelaria (116–110 Ma; 501 Mt at 0.54% Cu, 0.13 g/t Au, and 2.06 g/t Ag) and Mantoverde (121–117 Ma; 440 Mt at 0.56 % Cu and 0.12 g/t Au). The largest porphyry Cu-Au deposit is Carmen de Andacollo (104 Ma; proven and probable reserves of 417 Mt at 0.34% Cu and 0.12 g/t Au; Table 1).

Small Early Cretaceous (>125 Ma; Berriasian-Barremian) porphyry Cu-Au deposits occur at 22°S (e.g., Antucoya, 142 Ma; Tovaku, 132 Ma; Maksaev et al., 2006) and at 33°S (e.g., Colliguay, ~129 Ma; Maksaev et al., 2010) and appear to have formed during a short period of synchronous transpression at 22°S (Maksaev et al., 2006) or continental arc extension at 33°S (Creixell, 2007). In contrast, IOCG (and magnetite-apatite) deposits formed predominantly in the mid-Cretaceous (~125–~110 Ma; Aptian-Albian; Gelcich et al., 2005; Arévalo et al., 2006; Rieger et al., 2010; Tornos et al., 2010) and are located in a tectonically distinct but spatially superimposed belt from 25° to 34°S. These deposits have been described as spatially and temporally related to the culmination of a period of back-arc extension that developed along the continental margin from the Late Jurassic to mid-Cretaceous (Oyarzun et al., 1999; Grocott and Taylor, 2002; Sillitoe, 2003) or, alternatively, as having formed in response to the initiation of basin inversion in the mid-Cretaceous (Chen et al., 2013). Porphyry Cu-Au-(Mo) deposits, including the large Carmen de Andacollo porphyry Cu-Au deposit (104 Ma), again began to form in the later Cretaceous (<110 Ma; Cenomanian-Turonian) between 25° and 32°S, following the resumption of arc magmatism during or after basin inversion (Maksaev et al., 2010). We group igneous rocks and associated ore deposits into three broad temporal groupings that relate to tectonic setting as follows: early Cretaceous (extension), mid-Cretaceous (transtension), and late Cretaceous (contraction). Because the tectonic setting changed diachronously from north to south over intervals of 10 to 15 m.y., these terms do not correspond exactly to formal stratigraphic period subdivisions (epochs), and there is some temporal overlap between groups.

Evidence for a magmatic hydrothermal origin for several of these Chilean IOCG deposits has been reported, including at Candelaria and Mantoverde (Marschik and Kendrick, 2015) and Tropezon (Tornos et al., 2010, 2012), and several smaller IOCG deposits occur in close proximity to coeval intrusions with alteration patterns apparently centered on those plutons (e.g., El Trapiche veins, Creixell et al., 2009; El Espino, López et al., 2014). In contrast, Barton and Johnson (1996, 2000) argue that the ore-forming fluids were basinal brines, albeit with convection driven by the heat from coeval magmatism.

In this paper, we adopt the hypothesis that the Chilean IOCG and porphyry deposits are both of magmatic-hydrothermal origin, and that their contrasting styles of ore formation relate to differences in the tectonic setting and chemistry of the associated magmas. We combine new and published geochro-nological and geochemical data for igneous rocks associated with Cretaceous porphyry and IOCG deposits in the Coastal Cordillera with published structural information to show that porphyry and IOCG deposits formed in distinct tectono-magmatic settings. The bulk compositions of the associated igneous rocks are almost indistinguishable, but we present data from analysis of igneous apatite to suggest that the porphyry-forming magmas were S rich compared to their IOCG counterparts, consistent with the difference in S content of porphyry (S rich) and IOCG (S poor) ore deposits.

Mesozoic Tectonomagmatic Setting and Metallogeny of Northern Chile

Arc magmatism related to subduction has taken place along the Chilean segment of the Gondwana supercontinental margin since the late Paleozoic (Parada et al., 2007). Mid-Jurassic to Cretaceous magmatism mostly developed ~100 km to the west of the Paleozoic arc (Parada et al., 2007). Extensional tectonics affected the arc from the mid-Jurassic to early Cretaceous, with intra-arc and back-arc basin formation occurring in the mid-Cretaceous, followed by tectonic inversion in the mid- to late Cretaceous (Charrier et al., 2007). These tectonic changes are interpreted to reflect alternating periods of subduction coupling and decoupling between the down-going and overriding plates, with the generation of periods dominated by contractional and extensional tectonics, respectively, in the upper plate (Scheuber and Gonzalez, 1999). Of particular relevance to this study, early Cretaceous extension and mid-Cretaceous transtension occurred during a period of low convergence rate and weak plate coupling that may reflect slab rollback. This was followed by contraction in the mid- to late Cretaceous, caused by an increase in convergence rate in response to global-scale plate reorganization (Matthews et al., 2012).

Early Cretaceous magmatism in the Coastal Cordillera between 25° and 34°S was generated in a broadly extensional tectonic regime that caused rifting and crustal thinning (Parada et al., 2007). It was characterized by relatively primitive calcalkaline to shoshonitic lavas that built thick (5–10 km) subaerial volcanic sequences (Vergara et al., 1995; Morata and Aguirre, 2003; Parada et al., 2005; Charrier et al., 2007; Girardi, 2014). By the mid-Cretaceous, the tectonic setting had become predominantly transtensional (Brown et al., 1993; Arévalo et al., 2003) and was characterized by episodic volcanism and sediment deposition in intra-arc and back-arc shallow marine and continental basins (Fig. 1b; Morata and Aguirre, 2003). The transition from extensional to transtensional tectonics was slightly diachronous from north to south, and represents the progression of continental arc rifting southward with time, beginning in the north in the Barremian (~130 Ma) and reaching its maximum in the Aptian-Albian (~120 Ma). Plutonic complexes were emplaced during both tectonic stages, controlled by crustal-scale fault zones that evolved into the Atacama fault system in the Valanginian-Barremian (144–126 Ma; Brown et al., 1993). The Atacama fault is an early dip-slip and later sinistral transtensional N-trending fault system that extends for more than ~1,000 km parallel to the continental margin (Fig. 1a) and is interpreted to have formed in response to oblique subduction (Scheuber and Andriessen, 1990; Brown et al., 1993; Palacios et al., 1993); it is the primary structural control on the location of IOCG deposits in Chile (Sillitoe, 2003; Creixell et al., 2009). Crustal extension ended in the Albian-Cenomanian (110–95 Ma) with a return to contractional tectonics that produced crustal shortening and thickening, basin closure, rapid uplift, and a marked eastward shift of magmatism (Parada et al., 2002, 2005; Arancibia, 2004; Maksaev et al., 2010).

The spatiotemporal distribution of Cretaceous IOCG and porphyry Cu-Au deposits in northern Chile between 25° and 34°S is shown in relation to these tectonomagmatic periods in Figure 1b. Here it can be seen that, although the deposits roughly overlap spatially, IOCG deposits are broadly separated in time from younger porphyry Cu-Au deposits at ~110 to 100 Ma, which also marks the time of the major tectonic change from transtension to contraction. During the initial stages of arc rifting in the Late Jurassic-early Cretaceous, only a few small porphyries were developed (e.g., Antucoya, Colliguay). This was followed by the early formation of magnetite-apatite deposits (Gelcich et al., 2005; Creixell et al., 2009) and then IOCG deposits during the main stage of rifting and back-arc basin development in the mid-Cretaceous (~125–~110 Ma, Aptian-Albian; Oyarzun et al., 1999). This tectonic relationship appears to be consistent with IOCG deposits globally, which are commonly found to be associated with extensional events, including postcollisional, intracontinental, back-arc, and intra-arc rift settings (Williams et al., 2005; Corriveau and Mumin, 2010; Skirrow, 2010). Porphyry Cu-Au-(Mo) deposit formation returned in the late Cretaceous with the resumption of subduction-related magmatism during and after tectonic inversion. Apparent exceptions to this three-part deposit distribution (e.g., the small, mid-Cretaceous Totora, Pajonales, and Cachiyuyo porphyries, and the late Cretaceous Casualidad and El Espino IOCG deposits; Fig. 1) may reflect uncertainties in age determination, lithospheric heterogeneities, or delayed response to tectonic changes.

Sampling

Fifty-four samples of volcanic and intrusive igneous rocks associated with the Carmen de Andacollo, Frontera, and Dos Amigos porphyry Cu ± Au ± Mo deposits, the transitional Productora deposit, and the Candelaria, Casualidad, Espino, and Mantoverde IOCG deposits from the Coastal Cordillera of northern Chile between 26° and 32°S (Fig. 1) were collected from drill core and outcrops in July 2015. Where possible, intrusive rocks thought to be directly related to mineralization (coeval, cospatial) were collected; other samples were from broadly coeval intrusions or volcanic rock outcrops in the vicinity of the deposits (based on regional geologic maps). Least-altered samples were targeted for collection, but most of the rocks have undergone either low-grade regional metamorphism (resulting in minor chloritization of ferromagnesian silicate minerals and partial saussuritization of plagioclase) or hydrothermal alteration due to proximity to the ore deposits. Two samples of quartz-molybdenite veins were collected from the Carmen de Andacollo porphyry Cu-Au deposit for Re-Os dating.

The rocks were studied petrographically to determine the degree of alteration and to seek unaltered accessory minerals such as zircon and apatite for electron microprobe analysis.

Analytical Methods

Whole-rock geochemistry

Thirty-nine least-altered samples from the collected suite were submitted to Activation Laboratories Ltd. (Ancaster, Ontario) for analysis using the 4E-Research analytical package, which combines instrumental neutron activation analysis and lithium metaborate/tetraborate fusion inductively coupled plasma-mass spectrometry for determination of 62 elements. Analyses of standards and duplicates indicate that accuracy for major elements is typically within 5 relative %, and 10 relative % for minor and trace elements. The ferrous iron (FeO) content of the rocks was also determined by titration.

Electron microprobe analyses

Polished thin sections of all samples were examined for the presence of igneous accessory minerals such as zircon and apatite. Because of the relatively mafic nature of many of the samples, few contained zircon in thin section, so zircon chemistry was not attempted. A larger number of samples contained apatite, although care was needed to distinguish between igneous and hydrothermal (or late-stage) apatite. The latter was common in the groundmass of altered samples, and clearly hydrothermal apatite (intergrown with hydrothermal minerals such as quartz and sulfides) typically had thin, elongated prismatic habits (≤200-µm length; Fig. 2a). Apatite crystals with unequivocally igneous origin were distinguished by their inclusion within phenocrysts (mainly plagioclase and biotite). The igneous apatite crystals had stubby prismatic habits (20–50-µm length; Fig. 2b-d).

Compositional data were acquired with a Cameca SX100 electron microprobe using wavelength-dispersive spectroscopy and Probe for EPMA software (Donovan et al., 2015). For sessions in which seven elements (F, Na, Si, P, S, Cl, and Ca) were measured, the following conditions were used: 10-kV accelerating voltage, 20-nA beam current, and 5- to 10-µm beam diameter. Total count times of 30 s were used for both peaks and backgrounds. The X-ray lines, analyzing crystals, and standards were as follows: F Kα, PC0, topaz; Na Kα, TAP, tugtupite; Si Kα, TAP, topaz; P Kα, PET, fluorapatite; S Kα, PET, anhydrite (intensity data aggregated from two spectrometers; Donovan et al., 2011); Cl Kα, PET, tugtupite; and Ca Kα, PET, fluorapatite. The calculated limits of detection (as element, rounded to the nearest 10 ppm) at 99% confidence were as follows: F, 570 ppm; Na, 130 ppm; Si, 120 ppm; P, 170 ppm; S, 80 ppm; Cl, 290 ppm; and Ca, 280 ppm. To check for interference from third-order P Kα on F Kα, the signal from synthetic GaP was examined with the PC0 crystal: no significant interference from P on F was detected under the conditions of analysis.

For sessions in which nine elements (F, Si, P, S, Cl, Ca, Mn, Fe, and As) were measured, the following conditions were used: 15-kV accelerating voltage, 20-nA beam current, and 5-µm beam diameter. The X-ray lines, analyzing crystals, standards, and count times (seconds) on both peaks and backgrounds were as follows: F Kα, PC0, topaz, 60 s; Si Kα, TAP, topaz, 40 s; P Kα, PET, fluorapatite, 30 s; S Kα, PET, anhydrite, 30 and 40 s (intensity data aggregated from measurements on two spectrometers; Donovan et al., 2011); Cl Kα, PET, tugtupite, 40 s; Ca Kα, PET, fluorapatite, 30 s; Mn Kα, LIF, spessartine, 40 s; Fe Kα, LIF, spessartine, 40 s; and As Lα, TAP, synthetic GaAs, 40 s. The calculated limits of detection (as element, rounded to the nearest 10 ppm) at 99% confidence were as follows: F, 460 ppm; Si, 110 ppm; P, 260 ppm; S, 70 ppm; Cl, 150 ppm; Ca, 170 ppm; Mn, 170 ppm; Fe, 160 ppm; and As, 180 ppm.

In all sessions, time-dependent intensity corrections for F and Cl were carried out (peak count times divided into six intervals) with Probe for EPMA software (Donovan et al., 2015), following Nielsen and Sigurdsson (1981), Stormer et al. (1993), and Henderson (2011). Intensity data for all elements were reduced following the methods of Armstrong (1995). Oxygen was calculated by stoichiometry and included in the data reduction, as was the correction for oxygen equivalence of the halogens (F and Cl).

Re-Os dating

A molybdenite mineral separate was made for each sample by metal-free crushing followed by gravity and magnetic concentration methods described in detail by Selby and Creaser (2004). The 187Re and 187Os concentrations in molybdenite were determined by isotope dilution mass spectrometry using Carius tube, solvent extraction, anion chromatography, and negative thermal ionization mass spectrometry techniques. A mixed double spike containing known amounts of isotopically enriched 185Re, 190Os, and 188Os was used (Markey et al., 2007). A ThermoScientific Triton mass spectrometer with a Faraday collector was used for isotopic analysis. Total blanks for Re and Os are less than 3 and 2 pg, respectively, which are insignificant for the Re and Os concentrations in molybdenite. The molybdenite HLP-5 (Markey et al., 1998) was analyzed as a standard, and over a period of two years an average Re-Os date of 221.56 ± 0.40 Ma (1 s.d. uncertainty, n = 10) was obtained. This value is within the uncertainty of the 221.0 ± 1.0 Ma age reported by Markey et al. (1998).

Re-Os Age of the Carmen de Andacollo Porphyry Cu-Au Deposit

The Carmen de Andacollo porphyry Cu-Au deposit has previously been dated at 104 ± 3.3 Ma by U-Pb analysis of zircons in the ore-forming porphyry intrusions (Maksaev et al., 2010), and at 104 ± 3 and 98 ± 2 Ma by K-Ar analysis of phyllic- and potassic-altered rocks, respectively (Reyes, 1991). We have dated two samples of molybdenite from B-type quartz veins in the deposit, which yielded statistically indistinguishable ages of 103.9 ± 0.5 and 103.6 ± 0.5 Ma (2s errors; Table 2), in good agreement with the U-Pb date for magmatism of Maksaev et al. (2010).

Cretaceous Igneous Geochemistry

Whole-rock major and trace element geochemical data for 125 igneous rocks coeval with either IOCG or porphyry deposits from the Coastal Cordillera between 25° and 34°S were compiled from the literature and combined with our 40 new analyses (new data are listed in Table 3 and Supplementary Table S1, and data from the literature are listed in Supplementary Table S2). These rocks are divided into three groups, based on the tectonomagmatic periods described above: early Cretaceous, related to a few small porphyry Cu-Au deposits and early stages of arc rifting; mid-Cretaceous, related to IOCG deposits and back-arc transtension; and late Cretaceous, related to porphyry Cu-Au-(Mo) deposits and a return to contractional tectonics and arc magmatism. Note that, because of diachronous changes in tectonic style from north to south over periods of 10 to 15 m.y., the terms “early,” “mid-,” and “late Cretaceous” do not correspond exactly to formal stratigraphic divisions, but are approximately separated at ~125 and ~110 Ma, respectively.

Volcanic and plutonic rocks from the three tectonomagmatic groups are mostly metaluminous, calc-alkaline to high-K calc-alkaline, and range in composition from andesite (diorite) to dacite (granodiorite). Samples of igneous rocks directly associated with porphyry and IOCG deposits show a similar compositional range on a total alkali-silica diagram (Fig. 3), although it is evident that the porphyry-related rocks are mostly more felsic (diorite to granodiorite) compared to those associated with IOCG deposits (mostly gabbroic diorite to diorite). The apparently alkaline composition of several of these samples (especially from IOCG deposits) might be due to variable degrees of hydrothermal alteration (sodic-calcic or potassic), which was unavoidable in these suites of ore-associated rocks. However, on a plot of immobile element ratios (Zr/Ti versus Nb/Y; Fig. 4; Winchester and Floyd, 1977), a small number of mid-Cretaceous and IOCG-related samples do plot in more alkaline fields (alkali gabbro and syenite), suggesting that some of these samples are genuinely alkaline in composition. Nevertheless, the majority of the samples overlap in the gabbro-diorite-granodiorite fields in Figure 4, with no clear distinction between age groups or association with porphyry and IOCG deposits.

On primitive mantle-normalized extended trace element (Fig. 5) and chondrite-normalized REE (Fig. 6) diagrams, the porphyry and IOCG suites display almost indistinguishable patterns: enrichments in large-ion lithophile elements (LILEs: Rb, Ba, Th, U, K) and light rare earth elements (LREEs); negative anomalies for Nb, Ta, and Ti; relative depletions in compatible elements and middle to heavy rare earth elements (MREEs, HREEs); and flat to listric-shaped patterns from MREEs to HREEs. Such patterns are typical of subduction-related igneous rocks and reflect enrichments in fluid-mobile LILEs, retention of Nb, Ta, and Ti in insoluble Fe-Ti oxides, and fractionation of amphibole (which preferentially partitions MREEs; Gill, 1981; Green and Pearson, 1985; Klein et al., 1997). The only noticeable difference between these suites is the slightly greater depletion of MREEs-HREEs in the porphyry-related suite, which can be attributed to the more felsic (fractionated) nature of these rocks. This characteristic is also observed in the ratios of Sr/Y and La/Yb, which are generally more elevated in the felsic porphyry suite compared to the more mafic IOCG suite (Fig. 7).

In an attempt to assess the relative oxidation states of the various suites of rocks, we have assessed whole-rock Fe2O3/FeO ratios where reported. While many of these samples are altered to varying degrees, which will clearly affect the Fe2O3/FeO ratio, most of the samples have values ranging from 0.18 to 5.10, corresponding to moderately or strongly oxidized rocks (using the criteria of Blevin, 2004). Importantly, there is no clear difference in oxidation state between the three temporal-tectonomagmatic groups and porphyry- and IOCG-related rocks, which all appear to be similarly oxidized.

Published Sr-Nd isotope compositions for 53 Cretaceous igneous rocks from the Coastal Cordillera between 25° and 34°S (Supplementary Table S3) are plotted in Figure 8. Early Cretaceous rocks have evolved isotopic compositions suggesting extensive crustal contamination. In contrast, the mid- and later Cretaceous rocks cluster at relatively primitive compositions, albeit still with some crustal contamination.

Apatite Compositions

Electron microprobe analyses of magmatic and hydrothermal apatite from three samples from the Carmen de Andacollo porphyry and seven samples from or near IOCG deposits (Productora, Candelaria, and Casualidad) are listed in Supplementary Table S4, and SO3 analyses of magmatic apatite are summarized in Table 4 (along with data from the literature). The results show a clear difference between igneous apatite from rocks associated with the Carmen de Andacollo porphyry deposit (0.25 ± 0.17 wt % SO3, n = 69) compared with those from IOCG deposits (0.04 ± 0.02 wt % SO3, n = 76). A similar relationship is observed for apatite from porphyry deposits (0.12–0.60 wt % SO3) and IOCG deposits (0.07–0.13 wt % SO3) reported in the literature (Table 4). Hydrothermal or late-stage igneous apatite (e.g., edges of apatite microphenocrysts; Fig. 2d) showed lower and more variable SO3 contents (Supplementary Table S4), consistent with observations elsewhere in the literature (Streck and Dilles, 1998; Van Hoose et al., 2013).

The SO3 content of apatite varies as a complex function of magmatic temperature, oxidation state, and sulfur fugacity (Peng et al., 1997; Parat and Holtz, 2005; Parat et al., 2011), and an accurate calculation of magmatic sulfur content from apatite SO3 compositions is not currently possible. However, the apatite-melt partition coefficient formula of Peng et al. (1997), which is derived for relatively oxidized arc magmas, can be used to obtain a semiquantitative estimate of magmatic S content. We have estimated the apatite saturation temperature of four samples from the Carmen de Andacollo porphyry, the Candelaria and Casualidad IOCG deposits, and a sample of regional mid-Cretaceous igneous rock near Productora (using the equation of Piccoli and Candela, 1994, 2002, which is derived from the data of Harrison and Watson, 1984) and used this temperature in the equations of Peng et al. (1997) and Parat et al. (2011) to derive estimates of magmatic S content. The data reported in Table 5 suggest that the average S content of the magma associated with the Carmen de Andacollo porphyry deposit was ~0.02 wt % S (up to 0.06 wt % S)—significantly higher than the average values for magma associated with IOCG deposits in the region (0.001–0.005 wt % S; Table 5). An alternative method for calculating magmatic sulfur content from apatite SO3 compositions is provided by Parat et al. (2011), and these values are also listed in Table 5. The results differ somewhat in absolute values compared to the results using the Peng et al. (1997) formula, but not in the relative enrichment in S of the porphyry-related magmas compared to the IOCG-related magmas.

In order to eliminate the possibility that contrasting magmatic oxidation state was responsible for this difference in apatite SO3 content, we have estimated magmatic fO2 values for these samples from the average MnO contents of apatite using the equation of Miles et al. (2014, Table 5). These data confirm that all of the rocks are moderately oxidized (DFMQ ≈ 0.4–2.7), although we note that this fO2 calculation relies on electron microprobe analyses of Mn in apatite at concentrations close to the detection limit, and so cannot be considered highly accurate. There is also debate about the applicability of this method to rocks of different magmatic temperature and composition (Marks et al., 2016; Miles et al., 2016), although the samples compared here are of broadly similar composition and likely temperature. Consequently, we consider that the formula of Peng et al. (1997), derived for oxidized arc rocks, should be approximately valid, and the calculated abundances of magmatic S should be relatively correct, even if the absolute values are questionable (cf. Streck and Dilles, 1998). These data support the hypothesis that the porphyry-forming magmas were richer in S than the IOCG-related magmas.

Discussion

Igneous geochemistry

Whole-rock geochemical compositions of Cretaceous igneous rocks from the Coastal Cordillera of Chile between 25° and 34°S show minimal differences beyond those expected from normal fractionation processes (Figs. 35). Rocks associated with early Cretaceous arc rifting (with minor porphyry Cu-Au deposits), mid-Cretaceous back-arc transtension (with IOCG deposits), and later Cretaceous contractional arc magmatism (with porphyry Cu-Au-(Mo) deposits) have compositions typical of subduction-related magmatism. On average, however, later Cretaceous arc rocks are more felsic (Fig. 3), with pronounced listric-shaped REE patterns (Fig. 6) and higher Sr/Y and La/Yb ratios compared to mid-Cretaceous rocks associated with IOCG deposits (Fig. 7). There is also a suggestion that some of the mid-Cretaceous magmas and those associated with IOCG deposits may have had slightly more mafic, alkaline compositions (alkali gabbros to syenites).

We interpret these data to indicate that there was little fundamental change in the source of magmatism throughout the Cretaceous, except for a hint of more mafic alkaline magmatism during the mid-Cretaceous transtensional rifting phase. Published Sr and Nd isotope compositions of Cretaceous rocks from the Coastal Cordillera indicate mixing between depleted, mantle-derived magmas and isotopically evolved continental crust, with early Cretaceous magmas showing the highest degrees of contamination, likely by late Paleozoic lower and mid-crustal rocks (Fig. 8, Damm et al., 1990; Lucassen et al., 1999, 2001, 2002). Although the isotopic ranges of mid- and late Cretaceous rocks overlap, there is a significant clustering of mid-Cretaceous data at the most primitive end of the range (εNd ≈ 6, 87Sr/86Sri ≈ 0.7033). We suggest that back-arc extensional tectonics in the mid-Cretaceous may have facilitated the greater involvement of primitive, mildly alkaline magmas derived from partial melting of upwelling mantle asthenosphere during crustal extension and thinning (Fig. 9). Nevertheless, a significant amount of crustal processing is evident for all of these Cretaceous magmas, as is commonly observed in continental arc rocks (Hildreth and Moorbath, 1988). The more felsic nature and higher Sr/Y and La/Yb ratios of the later Cretaceous porphyry-associated magmas (Fig. 7) may reflect longer residence times in deep crustal magma chambers under contractional tectonic conditions (Hildreth and Moorbath, 1988; Haschke et al., 2002; Richards and Kerrich, 2007), with more extensive fractionation of amphibole (Lang and Titley, 1998; Richards and Kerrich, 2007; Richards, 2011a).

We co nclude that it is not possible uniquely to distinguish between igneous rocks associated with porphyry and IOCG deposits in the Coastal Cordillera of Chile using their whole-rock geochemical or isotopic compositions alone, although the IOCG-associated rocks are, on average, more mafic (locally alkaline) and isotopically slightly more primitive than the porphyry-related rocks, consistent with their back-arc as opposed to main arc tectonomagmatic setting.

Magmatic sulfur content

Porphyry and IOCG deposits are most fundamentally differentiated by the much greater abundance of sulfur in the former (predominantly fixed as pyrite in phyllic alteration zones) compared to the latter (where Fe oxides, magnetite and hematite, predominate). We have hypothesized that this might reflect a difference in the chemistry of magmas associated with these contrasting deposit types, but the analysis presented above shows that no clear differences can be found in their major and trace element compositions, although the IOCG-related rocks are somewhat more mafic than the porphyry intrusions.

Two other possible explanations are differences in the oxidation state or sulfur content of the original magma, but it is diffcult to constrain these parameters due to the widespread effects of oxidation during hydrothermal alteration in ore-associated rocks and the loss of S (as SO2) from igneous rocks as they crystallize. Nevertheless, Fe2O3/FeO ratios of igneous rocks from these different associations show no clear difference, and all are moderately to strongly oxidized (acknowledging that many of these rocks are hydrothermally altered, which likely leads to some secondary oxidation). We would expect the oxidation state of the magma to make a difference to the behavior of dissolved sulfur only if the rocks were significantly reduced (resulting in S being dominantly present as sulfide) compared to oxidized (with S dissolved as sulfate; Carroll and Rutherford, 1985; Jugo et al., 2005). All of the rocks in these suites appear to be at least moderately oxidized, as expected in arc rocks, so S should have behaved in a similar way in the magma.

In order to test the hypothesis that the magmas that formed S-rich porphyry deposits were more S rich than those forming S-poor IOCG deposits, we analyzed the compositions of igneous apatite crystals trapped as inclusions in silicate phenocryst phases. Our results show that apatite from intrusive rocks associated with the Carmen de Andacollo porphyry Cu-Au deposit contains significantly more S (0.25 ± 0.17 wt % SO3; Table 4) than that from rocks associated with or near IOCG deposits (Productora, Candelaria, and Casualidad; 0.04 ± 0.02 wt % SO3; Table 4). No reliable formula currently exists to convert apatite SO3 compositions to magmatic S contents, but the partition coefficients of Peng et al. (1997) can be used to provide a semiquantitative estimate. The results of these calculations suggest that Carmen de Andacollo porphyry magmas contained ~0.02 wt % S, compared to 0.001 to 0.005 wt % S in the IOCG-associated magmas—an order of magnitude difference (Table 5). This is the opposite to what would be expected if the difference were simply due to the felsic versus more mafic nature of the porphyry-versus IOCG-related rocks, because S solubility decreases with decreasing FeO content of magmas (Wallace and Carmichael, 1992). We conclude that the IOCG-related magmas were significantly S undersaturated compared to the porphyry-forming magmas and, consequently, could only generate S-poor magmatic hydrothermal fluids and ore deposits.

Metallogenic implications

The differences in tectonic setting and S content of magmas associated with porphyry and IOCG deposits in the Mesozoic Coastal Cordillera of Chile suggest contrasting models for their formation (Fig. 9). Tectonomagmatic processes that give rise to magmas with the potential to form porphyry Cu ± Mo ± Au deposits are well established (Richards, 2003; Cooke et al., 2005) and involve the evolution of hydrous, oxidized, S-rich, calc-alkaline magmas derived from partial melting in the metasomatized suprasubduction zone asthenospheric mantle wedge (Fig. 9a, c). A continuous flux of oxidized sulfur and water (and other fluid-mobile components such as Cl and LILEs) from the subducting slab supplies the large amount of S that is typically present in Phanerozoic arc magmas and gives them the potential to form S-rich porphyry deposits. An input of oxidized sulfur is specifically required because reduced sulfur will separate early from the magma as sulfide liquid and will deplete the magma of chalcophile elements prior to their potential segregation into late-exsolving magmatic hydrothermal fluids (Hamlyn et al., 1985; Spooner, 1993; Richards, 1995, 2003).

If this supply of oxidized sulfur ceases (e.g., subduction stops due to arc migration, reversal, or collision) or moves away from the region of magma production (e.g., subduction rollback), then derived magmas will be relatively S poor (Wallace and Edmonds, 2011). This may occur in back-arc settings, where slab rollback and back-arc extension result in the generation of magmas from upwelling asthenosphere distal to the subduction zone. These magmas have compositions reflecting more primitive intra-plate characteristics, although they retain some arc geochemical features (Kimura and Yoshida, 2006; Caulfield et al., 2012; Timm et al., 2012).

We hypothesize that this may have been the case during the mid-Cretaceous back-arc extensional phase in northern Chile, and that magmas formed at this time had lower magmatic S contents than normal arc magmas due to their distance from the subduction zone. Interaction of these relatively primitive, S-undersaturated asthenospheric magmas with previously subduction metasomatized upper plate lithosphere may have resulted in remobilization of metals and volatiles by dissolution of residual Cu-Au–rich sulfides and melting of hydrous silicates from lower crustal arc cumulates (Keays, 1987; Ackerman et al., 2009; Richards, 2009). Derivative magmas would be compositionally similar to normal arc magmas (being largely derived from subduction-modified lithosphere) but would only have the potential to form S-poor magmatic-hydrothermal ore deposits upon upper crustal emplacement (Fig. 9b; Core et al., 2006; Richards, 2009; Pettke et al., 2010; Richards and Mumin, 2013a, b). Deposits formed under such conditions would share many traits with normal arc porphyries but would be deficient in sulfur and sulfide minerals, and rich instead in Fe oxides. We refer to these as IOCG deposits and note that they share many similarities with magnetite-rich, alkalic-type porphyry Cu-Au deposits formed in back-arc settings (Harris et al., 2013), suggesting a continuum of broadly subduction related magmatic-hydrothermal deposit types, differentiated largely by their sulfur contents and the sulfur contents of the generative magmas (Richards and Mumin, 2013b).

Conclusions

By studying an area where porphyry and IOCG deposits occur in relatively close spatial and temporal proximity, we have shown that small changes in tectonic conditions can generate magmas with broadly similar bulk compositions but substantially different sulfur contents, such that distinctive ore deposit types are formed. Specifically, porphyry Cu ± Mo ± Au deposits form from relatively S rich arc magmas under conditions of upper plate contraction or transpression (e.g., the late Cretaceous period in Chile) and suboptimally under extension (early Cretaceous; Tosdal and Richards, 2001). In contrast, magmas generated in back-arc (extensional) settings distal to the subduction zone are likely to be relatively S poor and somewhat more mafic (locally alkaline, and isotopically primitive) in composition. Where such magmas interact with previously subduction modified lithosphere during ascent, they may remobilize metals, leading to the potential formation of S-poor ore deposits, such as IOCG and alkalic-type porphyry Cu-Au deposits. Some overlap in space and time between these deposit styles may result from local mantle and crustal heterogeneities, as well as delayed magmatic responses to tectonic changes. We suggest that other deposit types in the wider IOCG family diverge from this point of intersection with porphyry systems by the greater involvement of external fluids and crustal metal fluxing.

Acknowledgments

This paper is based in part on the published findings of numerous researchers, whose work we hope we have appropriately cited. This research was funded by a Discovery Grant (RGPIN203099) and an Engage Grant (EGP 485693-15) from the Natural Sciences and Engineering Research Council of Canada to JPR, and a postdoctoral fellowship from the University of Alberta to GPL. Claire Chamberlain and Andrew Davies from Teck Resources Ltd. are particularly thanked for their support of the Engage Grant and for access to and accommodation at the Carmen de Andacollo mine. We also thank all the geologists who helped us at the mine or project sites, particularly Andres Castillo from Teck Resources Ltd., Cristian Vasquez from Hot Chili Limited, Osvaldo Beltran from Pucobre, Cristian Leal from the Dos Amigos Mine, Bernardino Garay and Orlando Rivera from Codelco Chile, and Jose Armando Rodriguez from Anglo American. Special thanks go to Brian Townley for lending us regional maps. Finally, we are also grateful to managers Jorge Skarmeta (Codelco Chile), Vicente Irarrazabal (Anglo American), Rodrigo Arce (Anglo American), Wilfredo Tabilo and Marcelo Bruna (Pucobre), Rodrigo Diaz (Hot Chili), and Melanie Leighton (Hot Chili) for giving permission to access their sites and to sample drill core. Jared Geiger is thanked for his help with sample preparation at the University of Alberta. Garry Davidson is thanked for his careful review of an earlier version of the manuscript, and Murray Hitzman and John Dilles are thanked for their editorial contributions, which have helped to improve its quality.

Sources of Data for Figure 1

Tectonostratigraphic time boundaries in Figure 1b (gray lines) were defined based on correlation of similar formations and their interpreted tectonic settings, as follows:

Black lines represent the time boundaries of different tectonic settings based on a compilation of kinematics and timing of movement on structures along the Coastal Cordillera, as follows:

  1. Atacama fault system at 25.5° to 27°S (Grocott and Taylor, 2002);

  2. Atacama fault system at 25° to 27°S (Brown et al., 1993; Taylor et al., 1998);

  3. Atacama fault system at 26° to 27.5°S (Dallmeyer et al., 1996);

  4. Atacama fault system at 29°S (Arévalo et al., 2003);

  5. Dos Amigos fault system (Almonacid, 2007);

  6. Bahia Agua Dulce shear zone at 32°S (Ring et al., 2012);

  7. El Romeral fault at 30°S (Emparan and Pineda, 2006);

  8. Silla del Gobernador fault at 32°S (Arancibia, 2004);

  9. Magnetic fabric in Illapel plutonic complex (Ferrando et al., 2014);

  10. Dike swarms at 33°S (Creixell et al., 2011);

  11. Rapid exhumation of intrusions at 33°S (Gana and Zentilli, 2000).

*
A digital supplement to this paper is available at http://economicgeology.org/ and at http://econgeol.geoscienceworld.org/