Hypogene mineralization in porphyry Cu deposits is typically associated with crustal thickening and rapid exhumation, whereas supergene enrichment requires slow exhumation to allow sufficient time for leaching and downward transport of Cu before it is lost to surface erosion. Therefore, spatial and temporal patterns of exhumation within a metallogenic belt can highlight favorable locations for hypogene mineralization, supergene enrichment, and preservation. Here, we determine average pluton exhumation rates along an ~730-km segment of the middle Eocene-early Oligocene metallogenic belt in northern Chile (17.8°–24.2°S). By combining zircon U-Pb geochronology with Al-in-hornblende geobarometry, we pinpoint the time and depth at which each pluton was emplaced and use the age of overlying cover units or supergene minerals to date its arrival at the surface (or near-surface) environment.

Uranium-lead zircon ages for 49 samples from plutons and porphyries range from Carboniferous to Eocene (~314–35 Ma). Al-in-hornblende emplacement depths for 19 plutons are ~4–7 km, with one Carboniferous pluton emplaced at ~12 km. Two phases of net exhumation are identified: early Permian-Middle Triassic and middle Eocene-late Oligocene, with an intervening period of net burial. The highest exhumation rates (>0.30 km/m.y.) derive from the second phase, coeval with the Incaic orogeny and the main phase of hypogene mineralization. Present-day preservation of plutons and porphyry Cu deposits required low post-Oligocene average exhumation rates of <~0.01 km/m.y.—favorable for the development of many world-class supergene blankets. However, spatial variability in exhumation and burial across the belt led to poor conditions for supergene development locally: enrichment was hampered in some places by rapid exhumation after hypogene mineralization (e.g., ≥~4 km at El Abra), by burial beneath significant cover (e.g., Ministro Hales, Queen Elizabeth), or, in the Inti region of northernmost Chile, by a combination of the two.

Porphyry copper deposits are the world’s principal source of Cu and Mo (Sillitoe, 2010). The largest and highest-grade deposits commonly result from a two-stage process involving primary hypogene ore formation and secondary supergene enrichment, each favored by different geologic controls. During the hypogene stage, metal sulfides precipitate from magmatic-hydrothermal fluids to form an orebody, usually in the upper ~5 km of the crust (Seedorff et al., 2005; Wilkinson and Kesler, 2007; Singer et al., 2008; Yanites and Kesler, 2015). This is often associated with a compressional tectonic regime where crustal thickening and exhumation rates of several 100 m to over 1 km/m.y. prevail (Chivas et al., 1984; Kurtz et al., 1997; Sillitoe, 1998, 2010; Masterman et al., 2005; Yanites and Kesler, 2015; Sillitoe et al., 2019). Faster exhumation promotes efficient exsolution of metal-bearing fluids from magmas (Sillitoe, 1998; Cooke et al., 2005) and may facilitate telescoping, whereby early porphyry-style mineralization is overprinted and enriched by high-sulfidation mineralization at shallower crustal levels (Sillitoe, 1994, 1999; Masterman et al., 2005; Hervé et al., 2012; Maydagán et al., 2015; Sillitoe et al., 2019). Conversely, supergene enrichment occurs when hypogene ore sulfides, such as pyrite and chalcopyrite, are oxidized by meteoric water, and Cu is leached and transported in solution to the reducing conditions below the water table and reprecipitated in higher-grade sulfides such as chalcocite and covellite (Emmons, 1917; Sillitoe, 2005; Reich and Vasconcelos, 2015). In contrast to hypogene mineralization, this process requires slow exhumation that does not exceed the rate of downward Cu transport (Emmons, 1917; Brimhall et al., 1985; Hartley and Rice, 2005; Sillitoe, 2005; Riquelme et al., 2017; Sanchez et al., 2018). Supergene enrichment can more than triple the hypogene Cu grade (Sillitoe, 2005, 2010) and is often key to making a deposit economically viable. Constraining spatial and temporal patterns of exhumation during hypogene and supergene mineralization in porphyry Cu belts has potential as an exploration tool.

The Precordillera of northern Chile (Fig. 1A) hosts one of the most important porphyry Cu belts in the world, comprising middle Eocene-early Oligocene deposits, some of which have ore reserves exceeding 1 Gt (Sillitoe and Perelló, 2005). Several deposits exhibit both high hypogene Cu grades, often enhanced by late-stage high-sulfidation mineralization (Sillitoe and Perelló, 2005, and references therein), and world-class supergene enrichment zones (Sillitoe, 2005). These features suggest that exhumation rates were favorable for both hypogene and supergene mineralization, as well as for their subsequent preservation at near-surface levels. However, in northernmost Chile (north of ~19.5°S), there is scant evidence for porphyry Cu mineralization, perhaps due to extensive late Oligocene-early Miocene volcanic cover (e.g., García et al., 2017). The possibility of undiscovered porphyry Cu deposits in this area makes it attractive for exploration, but the potential for supergene enrichment is low, as volcanic deposition likely terminated supergene processes before significant ore-grade enrichment could occur (Sillitoe and McKee, 1996; Sillitoe, 2005).

This paper presents a way of tracking exhumation by dating and determining the emplacement depth of outcropping granitoid plutons: the average exhumation rate can simply be calculated as the emplacement depth divided by the time interval between emplacement and arrival at the surface (or near-surface). The latter can be constrained by supracrustal cover rocks in depositional contact with the pluton or by dated supergene minerals that constrain the time at which a pluton reached the subsurface water table. Unlike thermochronology-based exhumation studies conducted in the region (Maksaev and Zentilli, 1999; McInnes et al., 1999; Barnes and Ehlers, 2009; Juez-Larré et al., 2010; Reiners et al., 2015; Avdievitch et al., 2018; Sanchez et al., 2018; Stalder et al., 2020), this approach does not require knowledge of the geothermal gradient and its fluctuations through time, which can be difficult to accurately model, particularly in terranes experiencing simultaneous exhumation and magmatism (Murray et al., 2018).

New U-Pb zircon laser ablation-inductively coupled plasmamass spectrometry (LA-ICP-MS) ages of 49 intrusive rocks (43 plutons and six porphyries) from the middle Eocene-early Oligocene metallogenic belt in the Precordillera of northern Chile between 17.8° and 24.2°S (Fig. 1B) were obtained. The intrusions represent four different age groups: Carboniferous-Permian, Triassic, Late Cretaceous-Paleocene, and Eocene. Detailed petrography and whole-rock geochemistry identified a subset of 20 pluton samples suitable for Al-in-hornblende geobarometry, and emplacement depths were determined using the calibration of Mutch et al. (2016). When combined with minimum exposure ages from supracrustal cover rocks or nearby supergene mineralization, an average exhumation rate was determined for each pluton. In this way we were able to deduce a 300-m.y. history of exhumation and burial along this segment of the Precordillera and constrain average exhumation rates during periods of hypogene and supergene mineralization.

Tectonomagmatic evolution of northern Chile

Northern Chile, situated on the western margin of the Central Andes, comprises six morphotectonic units: Coastal Cordillera, Central depression, Precordillera, Western Cordillera, and the Altiplano and Puna plateaus (Fig. 1A). The region has experienced active subduction since at least the Ordovician, which has given rise to several magmatic arcs (Charrier et al., 2007; Niemeyer et al., 2018). Ordovician rocks are sparse, but the Carboniferous to Triassic arcs are well represented in the Precordillera by large intrusions (Fig. 1B) and their volcanic counterparts (Maksaev et al., 2014, and references therein). Slab rollback and possible steepening of the subduction angle in the Triassic caused extension in the overriding plate and eventually a westward shift in arc magmatism, commencing in the Coastal Cordillera in the Jurassic (del Rey et al., 2016; Oliveros et al., 2020). A marine back-arc basin developed in the present-day Central depression and Precordillera at this time (Vicente, 2006). The magmatic arc migrated progressively eastward from the Coastal Cordillera due to subduction erosion at the continental margin and progressive shallowing of the subduction angle (Kukowski and Oncken, 2006), generating volcanic and plutonic rocks with an eastward-younging trend (Coira et al., 1982). By the Eocene, the arc had reached the Precordillera, and intrusive rocks were emplaced into dominantly Jurassic-Cretaceous and Carboniferous-Triassic stratigraphic units (e.g., Tomlinson et al., 2001). Since the late Oligocene, the main axis of the magmatic arc has lain in the Western Cordillera, but in many locations volcanic deposits from this arc unconformably overlie the Eocene-Oligocene arc in the Precordillera, particularly north of ~19.5° S (García et al., 2004, 2017; van Zalinge et al., 2016; Platzman et al., 2020).

Several orogenic events have affected northern Chile since the Carboniferous. The middle Permian San Rafael orogenic event (Rapalini, 1989; Llambías and Sato, 1990) caused major reorganization of fault-bounded crustal blocks in the Precordillera (Tomlinson et al., 2001, 2012; Tomlinson and Blanco, 2008), before a long, dominantly extensional period from the Triassic to the middle Eocene. During this period, abundant supracrustal units were deposited in the Central depression and Precordillera, first in the Jurassic-Early Cretaceous marine back-arc basin and later in continental basins (Tomlinson et al., 2001; Cornejo et al., 2003; Charrier et al., 2007). The extensional regime was interrupted by the Late Cretaceous Peruvian (Mégard, 1987; Scheuber et al., 1994) and K-T (Cornejo et al., 2003) orogenic events, which produced regional unconformities. The Andean orogeny (Eocene to Present), which is dominated by tectonic shortening, crustal thickening, mountain building, and exhumation (Maksaev and Zentilli, 1999; Haschke et al., 2002; Armijo et al., 2015), commenced with the Incaic orogenic event between ~45 and ~33 Ma (Tomlinson et al., 2001; Mpodozis and Cornejo, 2012). The Precordillera at this time was bounded to the east and west by two arc-parallel thrust fault systems with opposite vergence, causing pop-up-style uplift and significant erosion of the range (Maksaev and Zentilli, 1999; Mpodozis et al., 2005; Amilibia et al., 2008; Charrier et al., 2013; Armijo et al., 2015; Reiners et al., 2015). Around 2–3 km of surface uplift of the Western Cordillera and Precordillera relative to the Central depression has taken place since the Incaic orogenic event (Armijo et al., 2015) in a tectonic regime that has remained largely compressional with several deformational peaks (Kay et al., 1999).

Patterns of exhumation

In the area around Chuquicamata (Fig. 1B), rock units representing contrasting Carboniferous-Permian paleodepths occur in different fault-bounded structural domains (Tomlinson et al., 2001, 2012; Tomlinson and Blanco, 2008). Erosional contacts with Triassic supracrustal rocks indicate juxtaposition and exposure of these different paleodepths by this time. This suggests that a period of exhumation occurred between the Permian and Triassic, possibly correlated with the San Rafael orogenic event, during which spatially variable exhumation rates prevailed (Tomlinson et al., 2012). In Sierra del Medio and at Cerro Jaspe (Fig. 1A), Carboniferous-Permian rocks are dominantly supracrustal units intruded by epizonal granitoids. The Cerros de Chuquicamata metaplutonic complex (Fig. 1B) represents a deeper crustal level with schistose and gneissic Carboniferous-Permian granitoids. Close to Chuquicamata, high-pressure metamorphism of the Limón Verde metamorphic complex (Fig. 1A) reached 13 ± 1 kbar at 270 ± 12 Ma (Lucassen et al., 1999), providing clear evidence for deep Permian burial in this location. It has also been suggested that Sierra de Moreno represents a deeper crustal level than Sierra del Medio, based on exposed basement consisting of Precambrian protoliths metamorphosed to amphibolite (and locally granulite) facies in the Cambrian-Ordovician (Damm et al., 1990; Tomlinson et al., 2001; Loewy et al., 2004).

Regional low-temperature thermochronology studies of Cenozoic exhumation in the Central Andes using zircon and apatite fission track (ZFT and AFT) and zircon and apatite (U-Th)/He (ZHe and AHe) analyses have consistently identified an eastward-younging trend in cooling ages (Barnes and Ehlers, 2009; Reiners et al., 2015; Stalder et al., 2020). This has been interpreted as reflecting an eastward progression of exhumation, starting in the Precordillera and Western Cordillera between ~60 and 40 Ma and propagating to the eastern side of the Altiplano and Puna plateaus, where exhumation continues today (Barnes and Ehlers, 2009; Reiners et al., 2015). Based on low-temperature thermochronology and 1-D thermokinematic modeling, two recent studies determined long-term mean exhumation rates in the Precordillera of <0.20 km/m.y. since ~50 Ma (Avdievitch et al., 2018) and 0.25 km/m.y. since ~80 Ma (Stalder et al., 2020). Neither study can, however, rule out a pulsed exhumation history with periods of higher rates. Maksaev and Zentilli (1999) used AFT analysis to determine rates of 0.10–0.20 km/m.y. from 50 to 30 Ma associated with the Incaic orogenic event, and ~0.05 km/m.y. from 30 Ma to present. Sanchez et al. (2018) estimated <2.5 km of exhumation in the Centinela district since the early Oligocene, based on AHe ages, yielding an exhumation rate of <~0.09 km/m.y. since then. They also infer that rates were higher than this during the Incaic orogenic event. Lower exhumation rates from at least the early Miocene onward are supported by widespread exposed Eocene plutons (e.g., Tomlinson et al., 2001) alongside late Oligocene-early Miocene supracrustal units, such as exceptionally well preserved ignimbrites (van Zalinge et al., 2016). Furthermore, the presence of extensive late Oligocene-Miocene erosion surfaces (pediplains) across northern Chile (Galli-Olivier, 1967; Mortimer, 1973; Dunai et al., 2005; Kober et al., 2007; Bissig and Riquelme, 2009; Evenstar et al., 2017) suggest decreased exhumation rates from this time onward (Riquelme et al., 2017; Carretier et al., 2018; Sanchez et al., 2018). This is consistent with a compilation by Carretier et al. (2018) of erosion rates across Chile north of 27°S, over shorter time spans of <100 years (from suspended sediments or shortlived isotopes), 100–100,000 years (cosmogenic nuclides, lacustrine sediment budgets), and <~20 m.y. (cosmogenic nuclides), which are predominantly less than 0.01 km/m.y. These determinations are generally taken from gently sloping surfaces; erosion rates an order of magnitude higher occur locally where slopes are steeper, such as in rivers with steep knickzones or on volcanic edifices (Carretier et al., 2018, and references therein).

The long-term exhumation rates determined for northern Chile are relatively low for an orogenic belt—at least 10 times lower than the highest rates observed in the Himalayas (e.g., Thiede and Ehlers, 2013; Adams et al., 2015). This difference has been attributed to the arid climate that has prevailed in the region since the Triassic or Jurassic (Hartley et al., 2005; Clarke, 2006). The present hyperarid climate is thought to have persisted since at least the middle-late Miocene (Mortimer, 1973; Alpers and Brimhall, 1988; Sillitoe and McKee, 1996; Jordan et al., 2014; Cooper et al., 2016; Rech et al., 2019) and possibly since the late Oligocene (Dunai et al., 2005; Evenstar et al., 2009, 2017). Rates of weathering and erosion in modern hyperarid climates are orders of magnitude lower than in more humid regions (Jordan et al., 2014, and references therein), which could explain the extremely low Miocene-Recent erosion rates.

Hypogene ore formation in the Precordillera

Northern Chile hosts two major porphyry Cu belts: the Paleocene-early Eocene belt, which mainly occupies the eastern part of the Central depression, and the middle Eocene-early Oligocene belt in the Precordillera. Additional porphyry Cu deposits associated with Permian-Triassic and Middle Jurassic-Cretaceous magmatism are of minor economic importance (Sillitoe and Perelló, 2005; Maksaev et al., 2007; Zentilli et al., 2018, and references therein). The Permian-Triassic and middle Eocene-early Oligocene belts are superimposed, and both occur within the study area (Fig. 1B). The middle Eocene-early Oligocene belt hosts some of the largest porphyry Cu deposits in the world, including Chuquicamata, Escondida, and Rosario (e.g., Sillitoe and Perelló, 2005). The Permian-Triassic porphyry Cu deposits have characteristics resembling their younger counterparts and do not seem to differ in terms of level of exposure but lack sufficient metal endowment to be economically viable (Cornejo et al., 2019). Hypogene mineralization in the younger belt took place between ~45 and ~31 Ma, and many deposits formed along strike-slip faults of the Domeyko fault system (Fig. 1A; Reutter et al., 1991, 1996; Lindsay et al., 1995; Cornejo et al., 1997; Sillitoe and Perelló, 2005, and references therein; Mpodozis and Cornejo, 2012). Several of the deposits have ore reserves exceeding 1 Gt and high hypogene Cu grades (Sillitoe and Perelló, 2005). The bulk of the ore is typically hosted within the early formed sulfide assemblage of chalcopyrite, pyrite, bornite, and molybdenite, associated with potassic alteration in the cores of the systems, but the highest Cu grades commonly occur where a late-stage high-sulfidation assemblage associated with phyllic and/or advanced argillic alteration is telescoped onto the potassic assemblage during rapid exhumation (Sillitoe, 1994, 1999; Masterman et al., 2005; Hervé et al., 2012; Maydagán et al., 2015; Sillitoe et al., 2019). Prime examples include Chuquicamata (Ossandón et al., 2001), Rosario (Masterman et al., 2005), Escondida (Padilla Garza et al., 2001; Hervé et al., 2012), Ministro Hales (Sillitoe et al., 1998; Boric et al., 2009), and Escondida Norte-Zaldivar (Hervé et al., 2012; Fig. 1B).

Supergene enrichment in the Precordillera

The middle Eocene-early Oligocene metallogenic belt hosts several world-class supergene enrichment zones (Sillitoe, 2005). Favorable conditions for enrichment require a balance between exhumation and relative water table descent, allowing enough time for Cu to be leached and transported down to the water table before it is lost at the surface (Emmons, 1917; Brimhall et al., 1985; Hartley and Rice, 2005; Sillitoe, 2005; Riquelme et al., 2017). Most major mature enrichment zones developed over 3–9 m.y. of supergene activity (Sillitoe, 2005) and are thought to require relatively stable landscapes and low exhumation rates, like those that prevail in semiarid to arid regions during pediplain formation (Bouzari and Clark, 2002; Riquelme et al., 2017; Sanchez et al., 2018).

However, exhumation rate is not the only factor affecting supergene enrichment. The presence of pyrite is essential, since the sulfuric acid generated by pyrite oxidation drives further leaching and Cu transport. Host rocks with low acidbuffering capacity help keep Cu in solution and prevent precipitation of Cu oxide minerals above the water table (Emmons, 1917; Chávez, 2000; Sillitoe, 2005; Reich and Vasconcelos, 2015). Rock permeability, which enables efficient descent of solutions, and the presence of bacteria that facilitate sulfide oxidation, sulfate reduction, and metal adsorption are also crucial (Sillitoe, 2005, and references therein; Enders et al., 2006).

The timing of supergene enrichment in northern Chile has generally been constrained via K-Ar and 40Ar/39Ar dating of the K-bearing sulfate mineral alunite (KAl3(SO4)2(OH)6), which forms in the acidic conditions resulting from pyrite oxidation (Vasconcelos, 1999). Early studies in northern Chile found alunite ages ranging from ~34 to ~14 Ma and linked the cessation of alunite formation to the onset of hyperaridity, which reduced available meteoric water to drive enrichment (Alpers and Brimhall, 1988; Sillitoe and McKee, 1996). Subsequent studies have broadened this age range to ~44–~5 Ma (Marsh et al., 1997; Mote et al., 2001; Rowland and Clark, 2001; Bouzari and Clark, 2002; Perelló, 2003; Quang et al., 2003, 2005; Arancibia et al., 2006; Bissig and Riquelme, 2010; Riquelme et al., 2017). The oldest ages (>~25 Ma) derive from the Paleocene-early Eocene metallogenic belt, where deposits were unroofed earlier than in the middle Eocene-early Oligocene belt (Sillitoe and McKee, 1996), and experienced favorable conditions for supergene enrichment during the Incaic orogenic event, which mainly affected the Precordillera (Bouzari and Clark, 2002; Maksaev and Zentilli, 1999). The youngest supergene ages are found in southern Peru and to the south of our study area (~26°S), which were likely the last areas affected by desiccation (Arancibia et al., 2006). Supergene ages within our study area define a supergene window between ~25 and ~12 Ma, during which conditions were favorable for supergene enrichment.

Forty-nine samples from intrusive rocks (43 from plutons and 6 from porphyries) were collected over an ~730-km-long segment (17.8°–24.2°S) of the middle Eocene-early Oligocene metallogenic belt (Fig. 1). The study area is split into six sections, each named after a major porphyry Cu deposit or district within it, apart from Inti, a name given by BHP to the area close to the border with Peru (Fig. 1A), which contains only minor, subeconomic Eocene mineralization (Ataspaca stockwork and skarn on the Peruvian side of the border; Clark et al., 1990).

Sampling focused on outcropping unaltered felsic granitoid rocks, which were the most abundant intrusive rock type in most parts of the study area. Samples were mainly collected from Eocene plutons, but Carboniferous-Permian, Triassic, and Late Cretaceous-Paleocene plutons were also sampled. Eocene mineralized porphyries were sampled at the Queen Elizabeth, Conchi, and Centinela deposits as well as from the San Lorenzo porphyries within the Los Picos-Fortuna pluton. A Triassic mineralized porphyry dike was sampled within the Caballuno pluton north of the Collahuasi district (Fig. 1B). Samples C91 and C92 were collected from exploration holes in the Inti region at depths of 1,272 and 947 m, respectively, immediately beneath the thick Oxaya Formation ignimbrites.

Petrography

Detailed thin section examination of each sample combined with point counting analysis (1,000 points per thin section) established that the plutons range in composition from quartz gabbro to monzogranite (Fig. 2). Representative images of the different age groups and areas are shown in Figure 3A-J. Mineralized porphyries are heavily altered, with an aphanitic matrix. Therefore, apart from the relatively coarse grained sample SID-16-23 from the San Lorenzo porphyries, they were not point counted, and classification is based on estimated primary phenocryst proportions and grain size. Complete petrographic descriptions can be found in Appendix 1; general features are summarized below.

All samples contain the major primary mineral phases plagioclase, quartz, and K-feldspar in variable proportions, with minor and accessory zircon, apatite, and Fe-Ti oxides (Table 1). Most samples contain biotite and amphibole, but in the Centinela and Escondida areas, clinopyroxene tends to be the dominant mafic mineral. Titanite occurs as a minor phase in some samples. In all but one sample, K-feldspar is orthoclase with or without thin, poorly developed, perthite exsolution lamellae. Carboniferous-Permian granodiorite sample SID-17-02 from Sierra de Moreno contains microcline K-feldspar (Fig. 3F), which may reflect slow cooling. Granophyric intergrowths between K-feldspar and quartz occur occasionally (Fig. 3H). The order of crystallization of the major phases is generally as follows: plagioclase (± clinopyroxene where present) > amphibole, biotite > interstitial quartz and K-feldspar. K-feldspar is typically the last phase to crystallize and, in clinopyroxene-rich rocks of the Escondida area, commonly occurs as poikilocrysts enclosing mafic minerals and plagioclase (Fig. 3J).

Of the pre-Eocene plutons, the Carboniferous-Permian age group was sampled in four locations within the Collahuasi and El Abra-Chuquicamata areas. Samples comprise relatively coarse grained (0.5–4 mm), hypidiomorphic hornblende-biotite granodiorites and granites, typically coarser than samples from the other age groups. Two Triassic samples, from the Longacho (SID-16-05) and Caballuno (SID-16-06) plutons (Fig. 1B) in the Collahuasi area, are equigranular, medium-grained (1–3 mm) biotite-hornblende granodiorite and monzogranite, in places containing minor clinopyroxene, commonly rimmed by hornblende. Petrographically, these plutons strongly resemble neighboring intrusions of Eocene age. Late Cretaceous-Paleocene plutons sampled from seven locations in the Collahuasi, El Abra-Chuquicamata, Centinela, and Escondida areas vary between quartz gabbro and monzogranite. Hornblende and biotite are the dominant mafic minerals in the medium- to coarse-grained plutons of Collahuasi and El Abra-Chuquicamata, whereas clinopyroxene is more common in the finer-grained, sometimes porphyritic, plutons of Centinela and Escondida.

Thirty Eocene pluton samples were collected across all six areas and range in composition from quartz diorite to monzogranite. In general, they are coarser grained and more felsic in the northern areas of Inti to El Abra-Chuquicamata than in the southern areas of Centinela and Escondida. In the Late Cretaceous-Paleocene samples, clinopyroxene is more abundant in plutons of the Centinela and Escondida areas than in those farther north. Clinopyroxene-dominated, more mafic plutons do occur between Inti and Chuquicamata (e.g., SID-16-10, Collahuasi area), but biotite and hornblende are more commonly the dominant mafic minerals. Plutons containing these minerals were sampled preferentially because of their suitability for Al-in-hornblende geobarometry. Eocene plutons tend to be equigranular and hypidiomorphic, but porphyritic textures also occur. Miarolitic cavities containing euhedral quartz crystals growing from the walls and a later infill of epidote ± chlorite occur in the Centinela (sample SID-16-26; Fig. 3I) and Queen Elizabeth (sample GEP-001, Yabricoya pluton) areas. Miarolitic cavities likely indicate shallow epizonal emplacement depths (Candela, 1997), but there is no definitive pressure limit for their formation.

Most plutons have experienced secondary alteration, but its intensity varies greatly between samples. The style of alteration is generally propylitic, represented by sericite-altered feldspar, chlorite-altered biotite, and hornblende and pyroxene replaced by actinolite ± chlorite. Epidote-altered mafic minerals and calcite occur locally. Clinopyroxene-bearing samples tend to be the most intensely altered, particularly Eocene samples from the Escondida area, where the mafic minerals have been completely replaced by actinolite and chlorite. In Centinela sample SID-16-88, some hornblendes have been partly replaced by biotite instead of actinolite ± chlorite, indicative of higher-temperature potassic, rather than propylitic, alteration. Also, in some porphyry samples relict hornblende has been completely replaced by shreddy biotite (samples SID-16-08, SID-16-23, and GEP-004). In some mineralized porphyries, phyllic (quartz-sericite-pyrite) alteration locally overprinted earlier potassic-altered zones (e.g., Conchi sample SID-16-17).

Geochemistry

Powders of <25-μm grain size were prepared for each sample at the University of Bristol by milling a portion of ~2 kg of crushed material in a Si3N4 ball mill. For major and minor element oxide determinations, rock powders were mixed with a flux in known proportions, then melted in Pt crucibles, poured into a Pt casting dish, and quenched into fusion beads for analysis by X-ray fluorescence using a Panalytical Axios Advanced spectrometer at the University of Leicester. A fraction of each powder was also heated in a 950°C furnace for 1–1.5 h to determine loss on ignition. However, analysis of sulfide-bearing, mineralized porphyry samples may damage the Pt equipment used to prepare the fusion beads. Therefore, the only analyzed porphyry sample is the weakly mineralized SID-16-23 from the San Lorenzo porphyry in the Los Picos-Fortuna pluton, where the Cu-bearing minerals occur in oxidized form (mainly chrysocholla). Accuracy (±5%) and precision (<2% 2SD) were determined by analyzing reference materials. Analytical results are presented in Appendix 2.

The sampled plutons range between ~50 and ~75 wt % SiO2 (normalized to 100% anhydrous) and geochemically resemble intrusive rocks elsewhere in this part of the Precordillera (Fig. 4). In Harker diagrams, the plutons mostly plot within the fields delineated by volcanic rocks erupted in the Andean arc between 17° and 25°S (data from http://georoc.mpch-mainz.gwdg.de/georoc/), showing that the plutons likely represent liquid compositions rather than cumulates. K2O behaves as an incompatible element in the suite, showing a general increase with increasing SiO2, in agreement with the observation that K-feldspar in most samples is the last phase to crystallize. Samples from the Late Cretaceous-Paleocene group commonly exhibit higher K2O contents than the general trend—a signature previously attributed to derivation from mafic magmas generated in an extensional setting by low degrees of partial melting of the mantle wedge inboard of the Late Cretaceous-Paleocene arc (Richards et al., 2001; Haschke et al., 2002; Ireland, 2010). The most SiO2-rich published data from (or near to) mineralized porphyries also tend to plot at higher K2O than the trend defined by the volcanic rocks, likely due to potassic alteration. Al2O3, FeOT, TiO2, and P2O5 decrease with increasing SiO2 and therefore behave as compatible elements. The lack of a fractionation peak for P2O5 indicates that apatite was saturated throughout crystallization. This, together with overall lower P2O5 concentrations, suggests relatively cool magmas compared to the volcanic rocks, since apatite solubility decreases with decreasing temperature (Chappell et al., 2004).

Methodology

Zircons were separated from the 49 samples using conventional heavy mineral separation techniques. Around 100–150 zircons were picked randomly from each sample to reflect the true distribution of grain sizes and morphologies. Grains were mounted in epoxy resin, polished, and imaged at the University of Bristol using a Centaur cathodoluminescence (CL) detector on a Hitachi S-3500N scanning electron microscope to investigate the internal morphology of the crystals, with particular focus on different growth zones.

Zircons were analyzed by LA-ICP-MS U-Pb geochronology at the Geochronology and Tracers Facility, British Geological Survey, Keyworth, using a Nu Instruments, Nu Plasma HR, multicollector-inductively coupled plasma-mass spectrometer (MC-ICP-MS). Laser sampling was performed with a New Wave Research 193ss laser ablation system. Zircon reference material 91500 was used as primary reference material and analyzed at regular intervals during each session. Two other zircon reference materials (GJ1 and either Plesovice or Mud Tank) were also analyzed during each session as secondary and tertiary validation materials.

Grains were randomly selected for analysis, but continuous growth zones on grain rims were targeted in order to sample the youngest domains of the crystal and, hence, the best approximation of the emplacement age. Around 20–30 zircon grains were analyzed per sample.

Data were processed using the Iolite software package (Paton et al., 2011), and weighted mean interpretations were calculated using 206Pb/238U isotope ratios rather than dates (Horstwood et al., 2016), after rejecting all data points with >5% discordance between the 206Pb/238U and 207Pb/235U systems and the inbuilt rejection criteria for 206Pb/238U weighted mean dates in Isoplot 4.15 (Ludwig, 2012). Components relating to the systematic uncertainty of the method were then added to the analytical uncertainties using an in-house Excel calculation. Finally, the weighted mean isotope ratio interpretations were converted into 206Pb/238U ages using the algorithms of Isoplot 4.15 (Ludwig, 2012). Full details of the method are provided in Appendix 3.

Results and U-Pb age interpretations

Age interpretations were derived from weighted mean 206Pb/238U dates of the data following the rejection criteria described above. Where the mean squared weighted deviation (MSWD; Wendt and Carl, 1991) of the remaining data points exceeded the statistically acceptable limit for a single population at the 95% confidence level, one or more interpretations that met the acceptable MSWD for the population size were evaluated. Our first weighted mean interpretation includes the youngest data point and the largest data population that yields a statistically acceptable MSWD (WM1 in Table 2). This interpretation is made on the assumption that resolvable, older dates reflect geologic dispersion induced by antecrystic zircon growth and/or subtle common Pb components within the analyses. The latter biases 206Pb/238U dates toward older values. Where samples contained a dominant central population of dates with both older and younger outliers, a second evaluation was conducted by rejecting the oldest and youngest data points until a statistically acceptable MSWD was obtained (WM2 in Table 2). This second interpretation is made on the assumption that the observed dispersion in dates is induced by the effects attributed to WM1, in addition to a component of Pb loss in the youngest dates.

Where present, the offset between the two age interpretations in the samples is typically below 1%—too small to significantly affect the geologic interpretations. For consistency, WM1 was chosen as the preferred age interpretation and used in the exhumation rate calculations. The preferred emplacement age interpretations of the 49 samples range between 314.5 ± 1.7 and 35.26 ± 0.34 Ma (Table 2; Fig. 5) and define four different age groups: Carboniferous-Permian, Triassic, Late Cretaceous-Paleocene, and Eocene.

Due to limitations in the absolute accuracy of the LA-ICPMS method of ~2% (Schaltegger et al., 2015), because data are not corrected for initial 230Th disequilibrium (equating to an additional ~90 k.y. on each age interpretation), and due to the offset between age interpretations of complex U-Pb data for individual samples, we do not attempt to resolve geologic events at a temporal resolution beyond a conservative ~1 m.y., despite the uncertainties of individual statistically derived age calculations being discernably lower. Analytical results including beam intensities, isotope ratios, dates, and concordance for single data points are presented in Appendix 4 along with weighted mean interpretations of 206Pb/238U isotope ratios.

Principle

The total Al content of igneous hornblende is governed by several pressure- and temperature-sensitive exchange reactions:

tremolite+phlogopite+2albite+2anorthite=2pargasite+6quartz+orthoclase
(1a)

phlogopite+2quartz+2anorthite=tschermakite+orthoclase
(1b)

tremolite+albite=edenite+4quartz
(1c)

edenite+albite=richterite+anorthite
(1d)

In general, those involving octahedral aluminum (AlVI), such as equations (1a, b), are more pressure sensitive (Hollister et al., 1987; Johnson and Rutherford, 1989), and those involving tetrahedral aluminum (AlIV) (eq. 1c, d) are more temperature sensitive (Blundy and Holland, 1990; Holland and Blundy, 1994). The Al content of amphibole is also affected by mineral assemblage, bulk composition, and oxygen fugacity (e.g., Hammarstrom and Zen, 1986; Anderson and Smith, 1995). Hammarstrom and Zen (1986) and Hollister et al. (1987) argued that for low thermodynamic variance granitic systems close to the H2O-saturated solidus (which is roughly isothermal above 2.5 kbar), hornblende composition is dependent chiefly on pressure. This dependence is most pronounced in the total Al content (AlT = AlIV + AlVI) leading to calibration of an Al-in-hornblende geobarometer for granitic rocks calibrated against pressure constraints from their metamorphic aureoles. Subsequent calibrations (Johnson and Rutherford, 1989; Thomas and Ernst, 1990; Schmidt, 1992; Anderson and Smith, 1995; Ridolfi et al., 2010; Ridolfi and Renzulli, 2012; Mutch et al., 2016) used amphibole data from high pressure-temperature (P-T) experiments. The different geobarometer calibrations yield variable pressures for the same amphibole composition. It is therefore important to assess which calibration provides the most accurate results for a specific problem, ideally by comparing Al-inhornblende geobarometry pressures with an independent reliable pressure estimate.

The low thermodynamic variance approach requires that a buffering assemblage of minerals (plagioclase, hornblende, biotite, quartz, K-feldspar, magnetite, ilmenite/titanite, apatite; Mutch et al., 2016) and water-saturated melt coexist in equilibrium. However, even in the presence of the buffering assemblage, there are rival temperature and pressure controls on the AlT content of amphiboles. Consequently, it is necessary to confirm that the amphibole compositions correspond to temperatures of the H2O-saturated solidus. This obviates the problem of suprasolidus amphiboles being used for barometry, which will lead to overestimates of the pressure. Subsolidus alteration, on the other hand, may cause formation of low-Al amphiboles and an underestimation of pressure.

Methodology

To evaluate the available Al-in-hornblende barometer formulations, we performed a test using amphibole rim compositions from the Peach Spring tuff (USA). This tuff has the appropriate mineral assemblage, and both amphibole analyses and modeled pressures of equilibration (using rhyolite-MELTS) are available (Pamukcu et al., 2015). Rhyolite-MELTS modeling yields a preeruption storage pressure between 1.85 and 2.30 kbar. Applying the nine Al-in-hornblende barometers (Hammarstrom and Zen, 1986; Hollister et al., 1987; Johnson and Rutherford, 1989; Thomas and Ernst, 1990; Schmidt, 1992; Anderson and Smith, 1995; Ridolfi et al., 2010; Ridolfi and Renzulli, 2012; Mutch et al., 2016) to Peach Spring tuff amphibole rims yielded a pressure range of –4.2 to 8.5 kbar (App. 5). The closest match to the rhyolite-MELTS model is provided by Mutch et al. (2016), i.e., 1.66–2.51 kbar, and this formulation was preferred here.

In order to be suitable for geothermobarometry using the above approach, samples must meet the following criteria: (1) presence of hornblende with rims in immediate contact with plagioclase, (2) presence of the buffering mineral assemblage plagioclase, hornblende, biotite, quartz, K-feldspar, magnetite, ilmenite/titanite, and apatite in apparent textural equilibrium, (3) low degrees of hydrothermal alteration, (4) bulk composition between 55 and 80 wt % SiO2, and (5) equilibration at or close to the appropriate water-saturated solidus.

Polished thin sections of each sample were examined under transmitted and reflected light in order to assess the first three criteria; the fourth was determined from the whole-rock geochemistry data. Criterion 5 was evaluated using geothermometer A of Holland and Blundy (1994) and is described further below. Scanning electron microscopy (SEM) was used to confirm the phase assemblage, and mineral microtextures were identified using backscattered electron (BSE) imagery. Energy dispersive spectrometry (EDS) element maps were made of some amphibole grains to identify potential element zoning. Major element compositions of amphibole and plagioclase were determined quantitatively by electron probe microanalysis using wavelength dispersive spectroscopy (WDS) on a Cameca SX100 at the University of Bristol with operating conditions of 20 kV, 10 nA, and 1-μm beam diameter. Coupled amphibole-plagioclase analyses were made on touching rims of grains in textural equilibrium. Amphibole mineral formulae were determined following Holland and Blundy (1994).

Results and interpretation

Twenty samples met all suitability criteria: all four Carboniferous-Permian samples, one Triassic sample, three Late Cretaceous-Paleocene samples, and 12 Eocene samples. Representative amphiboles are shown in Figure 6A-D. All but one of the suitable samples is from the four northern areas: Inti, Queen Elizabeth, Collahuasi, and El Abra-Chuquicamata. Where primary amphibole occurs in Centinela samples, many grains have been partly resorbed and have breakdown rims, demonstrating they were not in equilibrium at the conditions of magma emplacement (Fig. 6E, F). In the Escondida area, most samples lacked primary amphibole or were strongly altered, making it difficult to identify the precursor mafic minerals. Only one hornblende-bearing sample from this area (granodiorite SID-16-31) was suitable for barometry. North of Centinela, many samples contain hornblende that has been completely replaced by actinolite ± chlorite ± magnetite and in rare cases biotite (Fig. 6G, H). In less altered samples, hornblende is preserved but commonly exhibits a patchy zoning pattern, caused by partial replacement by actinolite, evident as dark zones on BSE images (Fig. 6I) that correspond to zones low in aluminum (Fig. 6J).

Mineral chemistry: Amphibole rims in all but four samples range between magnesiohornblende and actinolite (Leake et al., 1997). Amphibole rims of Carboniferous-Permian sample SID-17-02 (Fig. 6A) from Sierra de Moreno differ from the others and classify mostly as ferrohornblende. Amphiboles with breakdown rims in Centinela samples SID-16-72 (Fig. 6E) and SID-16-88 (Fig. 6F) are also exceptions, classifying as magnesiohastingsite and edenite, respectively. Quartz gabbro sample SID-17-14, which was discarded from the barometry study because of its mafic composition (50.1 wt % SiO2), has amphiboles ranging from edenite to magnesiohastingsite. Plagioclase rim compositions are relatively anorthite (An) poor; mean XAn ranges between 0.11 and 0.38 in all samples, except for the Late Cretaceous quartz gabbro (SID-17-14), which has a mean of 0.53. This compositional range is consistent with the experimental data used in the barometer calibration of Mutch et al. (2016), An25–58. Full analytical results are presented in Appendix 6.

Pressure determination: When touching-pair amphiboleplagioclase analyses from an individual pluton are plotted in P-T space, they define a linear trend that crosscuts the solidus curves of granitoids (Fig. 7) as taken from Piwinskii and Wyllie (1970) and Holtz and Johannes (1994; parameterized by Mutch et al., 2016). Amphibole-plagioclase pairs thus equilibrated at temperatures both above and below solidus, consistent with the wide stability of amphibole in granitic magmas. A linear regression was made through the paired analyses; the intercept between the regression and the relevant solidus is then taken to approximate the emplacement pressure. To account for uncertainties, a cloud of 1,000 randomly distributed points was generated around each paired analysis according to the uncertainties on the barometer (±16%) and thermometer (±35°C). One thousand linear regressions were then drawn through the point clouds, and the emplacement pressure and its standard deviation were determined as the mean and standard deviation of the intercepts between the regression lines and the appropriate solidus. The effect of the analytical uncertainty (around 1% AlT) on the final pressure estimate is insignificant compared to that from the barometer and thermometer calibrations. Even though amphibole-plagioclase analyses from samples containing only subsolidus temperatures can show a linear trend in P-T space, calculated emplacement pressures are not considered reliable unless the sample contains analyses that yield near-solidus temperatures.

Emplacement pressures and depths

Pressures obtained from the 20 suitable samples are similar regardless of age and location and lie between 1.16 ± 0.10 and 1.87 ± 0.15 kbar, except at Sierra de Moreno, where Carboniferous granodiorite SID-17-02 has a significantly higher emplacement pressure of 3.20 ± 0.28 kbar (Table 3). Pluton emplacement depths were calculated assuming a uniform upper crustal density in the Precordillera of ~2,700 kg/m3 (Götze and Kirchner, 1997; Lucassen et al., 2001; Prezzi et al., 2009). Potential lateral density variations are ignored, and it is also assumed that the pressure experienced by a crystallizing pluton is entirely lithostatic. The emplacement depth corresponding to the calculated pressure for sample SID-17-02 is 12.1 ± 1.1 km and between 4.43 ± 0.67 and 7.05 ± 0.58 km for all other plutons (Fig. 8).

The emplacement age and depth of 20 plutonic rocks have been precisely constrained by geochronology and geobarometry. With the exception of the two Inti drill core samples, all plutons are exposed at the surface today, so the average exhumation rate from emplacement to the present can be calculated by simply dividing emplacement depth by emplacement age. In many locations exhumation histories can be further refined by dated cover rocks, which provide minimum exposure ages for the plutons. Similarly, dated supergene minerals that formed within associated porphyry Cu deposits can constrain when they reached near-surface levels. However, caution must be taken since formation of supergene alunite can take place over considerable depth ranges (Alpers and Brimhall, 1988). For example, Chuquicamata hosts supergene alunite veins to depths of 1,000 m along high-permeability pathways created by faults (Sillitoe and McKee, 1996). Only alunite or jarosite that formed in the leached cap at the top of the weathering profile of a porphyry Cu deposit are therefore considered as reliable minimum exposure constraints. Leached caps in the Central Andes are typically a few tens to 200 m thick but reach 500 m at Escondida and El Salvador (Sillitoe, 2005, and references therein). This introduces an uncertainty in the depth the minimum exposure age captures. Likewise, in cases where a sample location is some distance away from the relevant cover unit, some exhumation may have happened at the sample location after the cover was deposited. However, these effects are considered to be minor and are not accounted for in the exhumation rate calculations unless stated.

For each pluton with a determined emplacement age (tEmp) and depth (zEmp), a minimum exposure age has been determined by the oldest dated cover unit or nearby supergene age. The average exhumation rate (RE) is then calculated for each pluton (illustrated in Fig. 9):

RE=zEmp/tEmptExp
(2)

Uncertainty on RE is calculated by full error propagation. The results provide information on the timing of periods of net exhumation along the Precordillera as well as average exhumation rates during these periods at each sampled location.

Minimum exposure ages

Fifteen minimum exposure ages were estimated from dated cover units and four from dated supergene minerals (Table 3; Fig. 10). Sixteen plutons were found to have been exposed by the late Oligocene-Miocene, while three Carboniferous-Permian plutons were exposed in the Triassic-Jurassic. Only one pluton lacks cover or supergene ages that can be used to determine a minimum exposure age. All plutons are assumed to have reached the surface by the time of their minimum exposure age, apart from Queen Elizabeth samples SID-16-02 and SID-16-03, which require further explanation. These samples were taken from the floor of a valley approximately 300 m below the contact with the 25.7 ± 1.0 Ma Utayane Formation (biotite K-Ar age; Muñoz, 1991, as cited in Sellés et al., 2016), which is the oldest dated cover unit in the area and was deposited on a pediplain surface following the Incaic orogenic event (Morandé et al., 2015). For this reason, exhumation rates for samples SID-16-02 and SID-16-03 are calculated assuming that they were 300 m below the surface at 25.7 ± 1.0 Ma. Subsequently, the area has experienced continued deformation, uplift, and erosion, (Morandé et al., 2015; Tomlinson et al., 2015; Sellés et al., 2016). A 12.9 ± 0.6 Ma volcanic deposit (whole-rock K-Ar age; Baker and Francis, 1978) sourced from the Cerro Patara volcano reaches the valley floor, suggesting that the final ~300 m plus the cover thickness must have eroded by this time, exposing the samples. See Appendix 7 for a detailed justification of all exposure ages.

Exhumation rates

The timings and average rates of pluton exhumation throughout the study area are presented in Figure 11. Average rates for the Carboniferous-Permian plutons range from 0.018 ± 0.002 to 0.125 ± 0.021 km/m.y., with the highest for sample SID-16-14 from the Collahuasi area, which had been exposed by the Triassic. The Triassic-Paleocene plutons range between 0.023 ± 0.001 and 0.113 ± 0.007 km/m.y., though none showed evidence of pre-Miocene surface exposure. For the Eocene plutons, average exhumation rates range between 0.097 ± 0.018 and 0.381 ± 0.049 km/m.y., and exhumation is generally bracketed between the late Eocene and late Oligocenemiddle Miocene. The highest exhumation rates are recorded in the Queen Elizabeth area where there is only ~15 m.y. between emplacement and near-surface arrival. Samples SID-16-02 and SID-16-03 experienced an additional pulse of at least 300 m of exhumation between deposition of the oldest cover unit (at 25.7 ± 1.0 Ma) and final surface exposure (by 12.9 ± 0.6 Ma) at a minimum average rate of ~0.023 km/m.y.

Pluton emplacement ages

The emplacement ages determined in this study are generally consistent with previously published ages with three notable exceptions: one sample of granodiorite from the Longacho pluton (SID-16-05) and two samples from the Caballuno pluton (granite SID-16-06 and mineralized feldspar porphyry SID-16-08) in the Collahuasi area (Fig. 1B) yielded Triassic ages rather than the expected Eocene ages (Longacho: 45.3 ± 0.4 Ma; Caballuno: 45.7 ± 4.6 Ma; Vergara and Thomas, 1984). This apparent discrepancy is probably due to the presence of texturally and compositionally similar Triassic and Eocene phases in these bodies, as noted elsewhere along the Precordillera (Zentilli et al., 2018). Overall, Eocene plutons in the Collahuasi, Centinela, and Escondida areas are older than those in the Inti, Queen Elizabeth, and El Abra-Chuquicamata areas.

Pluton emplacement depths

The consistency of pluton emplacement at epizonal depths of ~4–7 km, regardless of age, suggests an intrinsic physical control. Magmas can stall at a specific crustal depth for a number of reasons, such as magma viscosity, buoyancy, local stress regime, and crustal architecture (Chaussard and Amelung, 2014, and references therein). Annen et al. (2006) introduced the concept of “viscous death,” whereby buoyant hydrous magmas ascend through the crust until reaching their H2O-saturated liquidus, whereupon H2O exsolves and crystallization starts. Both fluid exsolution and crystallization increase the viscosity of the magma, which inhibits further ascent. In this model, the depth at which a magma crystallizes is primarily controlled by its H2O content. A wet magma would stall at a deeper level than a dry magma. If viscous death is the primary control on pluton emplacement depth in the study area, it follows that all plutons with similar emplacement depths had similar magmatic H2O contents. It is possible that the plutons in the Centinela and, in particular, Escondida areas with relatively dry clinopyroxene-rich mineralogies stalled at shallower levels than the amphibole-biotite–rich plutons farther north. Hornblende breakdown is commonly observed in volcanic rocks that have equilibrated at pressures below the minimum stability of this mineral (~1 kbar for dacite magmas; Rutherford and Hill, 1993).

The only exception to epizonal emplacement within the study area is the 12.1 ± 1.1-km-deep Carboniferous pluton at Sierra de Moreno (SID-17-02). The greater depth is supported by the geologic context: in contrast to the Carboniferous-Permian plutons of Cerro Jaspe and Sierra del Medio farther east, which intruded a thick package of approximately coeval volcanic and sedimentary rocks, the pluton at Sierra de Moreno intruded early Paleozoic metamorphic rocks (Damm et al., 1990; Tomlinson et al., 2001; Loewy et al., 2004). These metamorphic rocks represent a lower structural level, inferred to be part of the basement onto which the Carboniferous-Triassic surface units were deposited (Tomlinson et al., 2001; Tomlinson and Blanco, 2008).

Temporal exhumation patterns

The ages and exhumation rates determined in this study delineate two regionally important periods of exhumation: early Permian to Middle Triassic and middle Eocene to late Oligocene.

The fact that several Carboniferous-Permian plutons exposed at the surface by the Late Triassic or Jurassic are preserved alongside Eocene plutons with 4–7 km of postemplacement exhumation requires a period of net burial that returned the Carboniferous-Permian plutons to depths of 4–7 km prior to Eocene emplacement. This burial could have been accomplished by the abundant supracrustal rocks deposited across northern Chile during regional extension between the Late Triassic and middle Eocene (Cornejo et al., 2003; Charrier et al., 2007; Armijo et al., 2015). Jurassic-Early Cretaceous sediments deposited in the back-arc basin that developed in the Central depression and Precordillera probably reached thicknesses of at least 10 km in places (Armijo et al., 2015, and references therein). Several Late Cretaceous-middle Eocene volcano-sedimentary units, such as Tolar, Cerro Empexa, Quebrada Mala, Augusta Victoria, and Icanche, were subsequently deposited in continental basins along the Precordillera (Charrier et al., 2007). The extensional regime was interrupted by relatively short-lived compressional events, most notably the ~90–80 Ma Peruvian (Scheuber et al., 1994) and the ~65–62 Ma K-T (Cornejo et al., 2003) orogenic events, which caused deformation and depositional hiatuses (Tomlinson et al., 2001, 2015; Morandé et al., 2015; Sellés et al., 2016). Although some exhumation probably occurred during these events, it is unlikely to have been enough to counteract the overall net burial.

The Eocene plutons were all exhumed from depths of 4–7 km and, in several locations, show evidence of exposure by the late Oligocene-early Miocene. Average exhumation rates in several places exceed 0.30 km/m.y., although they are likely to encompass periods of faster and slower exhumation. For example, faster exhumation could have taken place at an early stage, correlated with the ~45–33 Ma Incaic orogenic event (Tomlinson et al., 2001; Mpodozis and Cornejo, 2012), in agreement with thermochronometric studies by Maksaev and Zentilli (1999) and Sanchez et al. (2018), who proposed slower rates after the early Oligocene (<~0.05 and <~0.09 km/m.y., respectively). Pre-Eocene and Eocene plutons likely experienced contemporaneous exhumation, which is supported by Eocene AFT ages of Carboniferous-Triassic plutons in Sierra del Medio (Maksaev and Zentilli, 1999), indicating residence beneath the AFT partial annealing zone until this time. Most AHe ages obtained in the study area fall between the middle Eocene and late Oligocene (Stalder et al., 2020, and references therein), underpinning that rocks exposed at the surface today were at shallow crustal depths already at this time (AHe closure happens at depths <~2.5 km, assuming a closure temperature of ~70°C, Farley, 2000; and a geothermal gradient of ~30°C/km, Maksaev and Zentilli, 1999).

Early Miocene and later exhumation rates in the Precordillera must have been relatively low, since rocks exposed at this time are still preserved at the surface today. Slow exhumation is consistent with an increasingly dry climate since the early Miocene (Dunai et al., 2005; Evenstar et al., 2009, 2017; Rech et al., 2019), the formation of regionally extensive late Oligocene-late Miocene geomorphic surfaces, often interpreted as pediplains (Galli-Olivier, 1967; Mortimer, 1973; Dunai et al., 2005; Kober et al., 2007; Bissig and Riquelme, 2009; Evenstar et al., 2017; Riquelme et al., 2017) and low erosion rates (<0.01 km/m.y.) calculated from cosmogenic surface exposure dating (Dunai et al., 2005; Kober et al., 2007; Evenstar et al., 2009, 2017).

Spatial exhumation patterns

Early Permian-Early Jurassic: Constraints on pre-Eocene exhumation are confined to the center of the study area where older plutons are exposed within different crustal blocks. The apparent variation in exhumation rate between Carboniferous-Permian samples SID-16-14 from Sierra del Medio, SID-17-03 from Cerro Jaspe, and SID-17-10 from Cordon del Millo (Fig. 1A), between 0.021 ± 0.002 and 0.125 ± 0.021 km/m.y., is due to differences in cover unit ages that constrain their time of exposure (Fig. 11). It is possible that the differences are real, but the plutons could instead have been exhumed simultaneously at the same rate and then been covered diachronously. Another possibility is that cover units are not preserved in every location. Sample SID-17-02 from Sierra de Moreno must have had a different exhumation history than the other Carboniferous-Permian samples for a much deeper crustal level to have been exposed by the Jurassic. Higher exhumation rates at SID-17-02 are supported by the amphibolite-granulite facies metamorphic rocks exposed in this location (Damm et al., 1990; Tomlinson et al., 2001; Loewy et al., 2004) and by the variability in exhumed thicknesses between crustal blocks during the Permian San Rafael orogenic event (Tomlinson et al., 2001, 2012; Tomlinson and Blanco, 2008).

Middle Eocene-late Oligocene: Average exhumation rates for Eocene plutons during this period were ~0.30–0.40 km/m.y., and the total amount of exhumation was ~5–6 km. As with the older intrusions, spatial variations in exhumation rate are possibly a function of the pluton and minimum exposure ages bracketing the period of exhumation. In reality, all locations may have experienced a period with an average exhumation rate >0.30 km/m.y. between the late middle Eocene and early Oligocene, which is only captured where a suitable late middle Eocene pluton and a late Oligocene lower bracket are available. Nevertheless, local variations are suggested by differences in the ages of exposed plutons. For example, near El Abra, sample SID-16-20 was emplaced at 36.8 ± 0.2 Ma and 5.20 ± 1.10-km depth, whereas Collahuasi samples SID-16-09 and SID-16-13 were emplaced at similar depths of 5.10 ± 0.42 and 4.92 ± 0.57 km, respectively, but are substantially older (45.61 ± 0.31 and 45.03 ± 0.33 Ma, respectively). Magmatism at Collahuasi continued until at least ~34 Ma, as indicated by the age of the Rosario porphyry Cu deposit (Masterman et al., 2004), but the apparent lack of exposed plutons <40 Ma may indicate that the deep plutonic environment of this age has not yet reached the surface. Masterman et al. (2005) estimated ~1 km of erosion in ≤2 m.y. during formation of the Rosario deposit, suggesting that at least one pulse of relatively fast exhumation has occurred in the Collahuasi district, but the total exhumed thickness since the late Eocene was probably less than at El Abra. Likewise, the abundance of potentially shallow, clinopyroxene-dominated plutons at Centinela and Escondida suggests relatively limited exhumation.

Early Miocene-Present: From the early Miocene onward, exhumation rates across the Precordillera were consistently low but appear to have been slightly higher north of ~20°S, as indicated by the evidence for deformation, uplift, and erosion in the Queen Elizabeth area between the late Oligocene and middle Miocene (Morandé et al., 2015; Tomlinson et al., 2015; Sellés et al., 2016).

Summary and interpretation

Based on the Eocene and later pluton exhumation histories and other geologic constraints discussed above, it is possible to divide our studied section of the Precordillera into three segments: a northern segment comprising areas north of and including Queen Elizabeth, a central segment between and including Collahuasi and El Abra-Chuquicamata, and a southern segment comprising Centinela and Escondida (Fig. 12A). The northern segment experienced relatively fast exhumation throughout the Eocene-late Oligocene, was covered to a greater extent than areas farther south by thick late Oligocene-early Miocene volcanic deposits, and shows evidence of greater post-early Miocene deformation and exhumation. The central segment records slower exhumation, even though relatively young and deeply emplaced plutons are locally exposed (e.g., El Abra). The southern segment comprises abundant fine-grained, clinopyroxene-bearing (rather than amphibole-bearing) plutons, which may indicate shallow emplacement depths and less post-Eocene exhumation than in the northern and central segments.

The boundary between the northern and central segments is located at a dividing line between late Oligocene-early Miocene cover rocks that largely overlap with the axis of the Eocene-Oligocene arc to the north, and younger cover rocks mostly located east of the Eocene-Oligocene arc to the south (Freymuth et al., 2015). Early-middle Miocene flat-slab subduction south of ~20°S has been suggested as the explanation for the lack of volcanic rocks of this age in this area (e.g., Wörner et al., 2000; Ramos and Folguera, 2009). The northern segment comprises a series of N-trending anticlines, formed by mainly Oligocene-Miocene deformation, which has led to uplift and erosion (Farías et al., 2005; Sellés et al., 2016; van Zalinge et al., 2017). The Queen Elizabeth area is located in the core of one of the anticlines (Sellés et al., 2016), which may partly explain its relatively high exhumation rates during this time.

The boundary between the central and southern segments coincides with the Antofagasta-Calama lineament, a fault zone that initiated in the Eocene and has been interpreted as the boundary between two basement terranes—Antofalla to the north and Chilenia to the south—based on different paleomagnetic rotation patterns and isotope signatures (Arriagada, 2018; Bunker, 2020). It is likely that the different rotation patterns reflect differing histories of deformation and thus exhumation either side of the fault zone.

Our study suggests that within the Precordillera of northern Chile, periods of exhumation in the early Permian-Middle Triassic and Eocene-late Oligocene, separated by a Late Triassic-middle Eocene period of net burial, led to preservation at near-surface levels of epizonal rocks from time periods spanning ca. 300 m.y. Consequently, a belt of subeconomic Permian-Triassic porphyry Cu deposits, which formed prior to the phase of net burial, is preserved alongside major porphyry Cu deposits of the superimposed middle Eocene-early Oligocene metallogenic belt (Cornejo et al., 2019). The Permian-Triassic porphyry Cu deposits have characteristics similar to those of their younger counterparts but lack economic metal endowments (Cornejo et al., 2019).

Patterns of post-Eocene exhumation must have been conducive for the formation, enrichment, and preservation of the middle Eocene-early Oligocene porphyry Cu deposits, as previously suggested by Maksaev and Zentilli (1999) and Sanchez et al. (2018). Hypogene ore formation benefited from relatively high exhumation rates, resulting in large, high-grade orebodies, of which several were telescoped and enhanced by a late-stage high-sulfidation overprint, e.g., Chuquicamata (Ossandón et al., 2001), Escondida (Padilla Garza et al., 2001; Hervé et al., 2012), Escondida Norte-Zaldivar (Hervé et al., 2012), Rosario (Masterman et al., 2005), and Ministro Hales (Sillitoe et al., 1998; Boric et al., 2009). The subsequent slowing of exhumation from the late Oligocene-early Miocene onward created ideal conditions in many locations for supergene enrichment and surface preservation of the orebodies (Sillitoe, 2005). However, the belt also includes several poorly enriched deposits that were either dominated by in situ oxidation or experienced only minor supergene development (Fig. 12A).

Enriched porphyry Cu deposits (Fig. 12B): Porphyry Cu deposits with significant enrichment zones occur in the central and southern segments of the study area (Fig. 12A) and tend to have developed in orebodies where pyrite-rich alteration zones are preserved. Some host significant oxidized zones above the secondary sulfides that developed when acidity was insufficient to mobilize all the Cu from the hypogene ore (e.g., Toki cluster, Rivera et al., 2009) or from early formed enrichment zones (e.g., Chuquicamata, Ossandón et al., 2001; Polo Sur, Perelló et al., 2010). Enrichment appears to have been particularly efficient in telescoped deposits, where a high-sulfidation assemblage overprinted the potassic zone, introducing abundant acidity-generating pyrite and altering host rocks to phyllic or advanced argillic assemblages with low acid-buffering capacity. Examples include Chuquicamata, Escondida, and Escondida Norte-Zaldivar, which all have enrichment factors ≥3 (Sillitoe, 2005). Enrichment is also well developed in less telescoped porphyry Cu deposits with phyllic zones overlying potassic zones, such as Quebrada Blanca and Ujina, which have enrichment factors of ~2.4 (Sillitoe, 2005). Preservation of the relatively shallow phyllic and advanced argillic alteration zones probably requires fairly low postmineralization exhumation rates. In the Collahuasi district, this is supported by fluid inclusion pressure estimates from the Rosario porphyry Cu deposit, suggesting firstly that relatively fast synmineralization exhumation removed ~1 km of overburden between early potassic (~34.4 Ma) and late high-sulfidation (~32.6 Ma) epithermal assemblages, and secondly that <500 m of surface denudation occurred thereafter (Masterman et al., 2005). Slow postmineralization exhumation resulted in prolonged preservation of the upper parts of the hypogene orebodies, which enabled the formation of welldeveloped enrichment zones and may explain the apparent absence of outcropping plutonic rocks with ages <40 Ma in the vicinity of the porphyry Cu deposits. The preservation of shallowly emplaced, clinopyroxene-rich plutons in the Centinela and Escondida areas also supports slow postmineralization exhumation.

In situ oxidized porphyry Cu deposits (Fig. 12C): Some deposits within the study area lack major secondary sulfide zones but have been affected by significant in situ oxidation, such as at El Abra, Radomiro Tomic, Esperanza, and Telegrafo. These deposits are pyrite poor, and many are dominated by potassic alteration (Ambrus, 1977; Cuadra and Camus, 1998; Perelló et al., 2004, 2010; Sillitoe, 2005), which may indicate that pyrite-rich phyllic and advanced argillic zones once present above the potassic zone have been eroded, exposing relatively deep levels (Ambrus, 1977; Sillitoe, 2005). The 36.8 ± 0.2 Ma, 5.20 ± 1.10-km-deep sample SID-16-20 from the Pajonal-El Abra pluton supports deep erosion of the El Abra deposit. The pluton overlaps in age with the ~38–36 Ma El Abra porphyries (Ballard, 2001; Campbell et al., 2006; Correa et al., 2016), indicating ≥4 km of postmineralization exhumation. This is substantial, given that most porphyry Cu deposits are thought to form in the upper 5 km of the crust (Seedorff et al., 2005; Wilkinson and Kesler, 2007; Singer et al., 2008; Yanites and Kesler, 2015). Esperanza and Telegrafo are characterized by concentric alteration zoning patterns where pyrite-poor potassic cores (hosting the bulk of the Cu) presently reach near-surface levels and grade laterally to more pyrite-rich sericite-chlorite and phyllic alteration (Perelló et al., 2010). Erosion of potentially pyrite rich zones once present above the potassic cores may have prevented significant enrichment. Fluid inclusion data from Esperanza suggests that quartz veins formed 1–2 km below the water table (Perelló et al., 2004), which suggests a shallower erosional level but also a shallower emplacement depth compared to El Abra. Shallow emplacement is consistent with the high Au content at Esperanza and Telegrafo (Perelló et al., 2010) compared to the primarily Cu-Mo deposits that dominate the rest of the belt, since Cu-Au deposits are typically emplaced at shallower levels than Cu-Mo deposits (Cox and Singer, 1992; Sillitoe, 2000). Shallow emplacement is also supported by the abundant fine-grained intrusions in the Centinela district, which lack hornblende in equilibrium. The Radomiro Tomic deposit is deeply oxidized but poorly enriched and also exposed at a level dominated by potassic alteration with only minor structurally controlled zones of phyllic alteration (Cuadra and Camus, 1998).

Porphyry Cu deposits with minor supergene development (Fig. 12D): The metallogenic belt also includes porphyry Cu deposits affected by only minor supergene modification. The Queen Elizabeth area in the northern segment (Fig. 12A) hosts relatively small, low-grade porphyry Cu deposits (Camus, 2003) that have experienced only minor supergene development. This is probably due to the presence of late Oligocene-early Miocene volcanic cover units that currently surround, and may once have covered, the Yabricoya and Alantaya plutons. The ~38 Ma (this study) Queen Elizabeth porphyry Cu deposit was already covered by volcanic rocks in the late Oligocene, before major supergene enrichment in other parts of the belt, and this likely inhibited supergene development. Erosion continued in the area after cover rocks were deposited (Morandé et al., 2015), resulting in partial reexposure of the deposit, but this may have happened outside or too close to the end of the supergene window. The ~40 Ma La Planada deposit (Gardeweg and Sellés, 2013; Tomlinson et al., 2015) is older than Queen Elizabeth and has been interpreted as a deeply eroded system (Eggers and Fuentes, 2019), which may partly explain the poor supergene development. However, the deposit may also once have been concealed beneath late Oligocene-early Miocene cover. Ministro Hales, on the western side of the West fault in the Chuquicamata district, has a poorly developed supergene profile compared to Chuquicamata and Radomiro Tomic on the eastern side (Sillitoe, 2005). Ministro Hales was exhumed to near-surface levels by 20.8 ± 0.6 Ma, as indicated by a supergene alunite K-Ar age (Sillitoe and McKee, 1996), but then covered by >50 m of alluvial gravels, indicating that it became located in a depositional environment that probably did not experience the required water table drop (Sillitoe et al., 1998). At the same time, Chuquicamata and Radomiro Tomic on the eastern side of the fault likely experienced relative water table drop and deep weathering (Sillitoe and McKee, 1996; Sillitoe et al., 1998; Sillitoe, 2005).

Porphyry Cu deposit potential in northernmost Chile: Age and emplacement depth data from plutons in the Inti area suggest that Eocene plutons representing a crustal paleodepth similar to that of areas farther south in the metallogenic belt were exposed at the surface prior to late Oligocene-early Miocene volcanism. Thus, potential middle Eocene-early Oligocene porphyry Cu deposits could also have reached the surface prior to deposition. The ~35 Ma plutons (samples C86 and C88) suggest that an even greater thickness of rock has been exhumed at a later stage at Inti than in other parts of the belt in order to have exposed such young plutons. However, these plutons lack emplacement depth constraints due to an absence of hornblende and could have been emplaced at shallow levels (<4 km). Despite exhumation histories similar to those of economically richer parts of the belt, the volcaniclastic units, which likely terminated supergene processes, were deposited already in the early Miocene. The only possible window for supergene enrichment of potential porphyry Cu deposits is therefore prior to deposition of the cover. The requisite time is not known, but, based on the span of supergene alunite ages from Escondida and Chuquicamata, supergene enrichment processes lasted for at least ~3 to ~5 m.y. (Alpers and Brimhall, 1988; Sillitoe and McKee, 1996). Taking 3 m.y. as a minimum time to develop a mature supergene blanket in an arid environment would require porphyry Cu deposits at Inti to have been exposed at or near the surface by ~25 Ma and located in a favorable environment for supergene enrichment until cover deposition at ~22 Ma. Sample C91 suggests that the Eocene magmatic arc in the Inti area underwent exhumation at an average rate of 0.326 ± 0.032 km/m.y. between ~40 and ~22 Ma, which is probably too high to enable appreciable supergene enrichment in an arid environment. A situation where the rate of exhumation was higher toward the start of the time interval and significantly lower just prior to the time of burial beneath volcanic cover is possible and could have created favorable supergene conditions toward the late Oligocene. However, the reconstructed late Oligocene paleolandscape suggests that the Precordillera in this area was cut by ~500-m-deep EW-directed paleochannels that transported conglomerates of the Azapa Formation into the Central depression (van Zalinge et al., 2017). Such a landscape is different from a pediplain of the type known to be favorable for supergene enrichment (Bouzari and Clark, 2002; Riquelme et al., 2017; Sanchez et al., 2018), suggesting that the Precordillera in northernmost Chile was a highly erosive terrane until the volcanic cover was deposited. Inti therefore appears to be an area where extensive postexhumation cover was detrimental to supergene development.

Despite having Carboniferous-Eocene emplacement ages spanning almost 300 m.y., most plutonic rocks in the middle Eocene-early Oligocene metallogenic belt in the Precordillera of northern Chile (17.8°–24.2°S) have remarkably similar emplacement depths of ~4 to ~7 km. This is inconsistent with a unidirectional exhumation model and likely reflects a more complex history of exhumation and burial.

Two periods of net exhumation can be defined: (1) early Permian-Middle Triassic, when Carboniferous-Permian plutons were exhumed and exposed at the surface and (2) middle Eocene-late Oligocene, when Eocene plutons were exhumed and exposed. Preservation of both Carboniferous-Permian and Eocene plutons at the surface today requires a period of net burial between the Late Triassic and middle Eocene, during which a dominantly extensional regime prevailed in northern Chile. Preservation of plutons today that were exposed by the late Oligocene-early Miocene also requires low exhumation rates since this time.

Hypogene mineralization in the middle Eocene-early Oligocene porphyry Cu deposits occurred at a time of relatively fast exhumation (>0.10 km/m.y.), consistent with previous thermochronology studies and with models suggesting that high exhumation rates favor hypogene ore formation by promoting efficient fluid exsolution and facilitating telescoping. Supergene mineralization, on the other hand, occurred when average exhumation rates were lower (<~0.01 km/m.y.). Slow exhumation would have created a favorable balance between surface denudation and relative water table descent, allowing enough time for Cu to be leached and transported down to the paleowater table before it was lost to surface erosion. This likely enabled the formation and preservation of supergene enrichment zones.

Even though the regional exhumation history of the middle Eocene-early Oligocene belt was favorable for both hypogene and supergene mineralization, supergene enrichment appears to have been hampered where the exhumed crustal thickness following hypogene mineralization was substantial (e.g., ≥~4 km at El Abra) or where porphyry Cu deposits were buried beneath significant cover (e.g., Queen Elizabeth, Ministro Hales). In the Inti region, extensive late Oligocene-early Miocene volcanic cover would have terminated nascent supergene enrichment processes within any near-surface porphyry Cu deposits, while exhumation rates prior to cover deposition were probably too high to enable effective supergene enrichment to initiate.

This work was funded by BHP. We are grateful to all the BHP staff in Chile who helped us with field work, particularly Carlos Galdames, Aldo Vásquez, Jamie King, Chris Ford, Guilherme Andrade Santos, and Walter Jimenez. All sampling was conducted with written permission from mining and exploration companies holding exploration permits. We thank Anglo American, Antofagasta Minerals, Codelco, Collahuasi, ENAMI, Freeport-McMoran, Glencore, Minera Escondida, Minera Meridian, Sumitomo Corporation, Teck Resources, and Vale for granting us sampling permission. We thank Edward Bunker, Luke Neal, and Anne Mather for field assistance; Dan Condon at the British Geological Survey for field assistance and helpful discussions about U-Pb geochronology; Stuart Kearns and Ben Buse for technical assistance with the SEM and electron microprobe; Vanessa Pashley and Matthew Horstwood at the British Geological Survey for help with the U-Pb analyses; Tom Knott and Lin Marvin-Dorland at the University of Leicester for help with whole-rock geochemistry analyses; and Isa Witick for assistance with compiling the final manuscript. We appreciate the constructive reviews provided by Jose Piquer, Eduardo Campos, and David Cooke, and we thank Larry Meinert and David Cooke for their editorial handling of the manuscript.

Simon Dahlström received his Ph.D. degree in earth sciences from the University of Bristol, United Kingdom, in 2020. His Ph.D. project was fully funded by BHP and investigated how granite emplacement and exhumation in the Central Andes can be tied to porphyry copper deposit formation, enrichment, and preservation. Before moving to Bristol, he obtained his M.Sc. degree at Åbo Akademi University in Turku, Finland, and worked on exploration projects for several companies in northern Sweden and Finland. Simon currently works as a project geologist at Terrafame Ltd. in Sotkamo, Finland, investigating the Paleoproterozoic Kolmisoppi Ni-Zn-Cu-Co deposit.

Gold Open Access: This paper is published under the terms of the CC-BY-NC 3.0 license.

Supplementary data