Detrital zircon record of Phanerozoic magmatism in the southern Central Andes

Subduction zones are fundamental components of Earth systems that link lithospheric dynamics, mountain building, surface processes, magmatism, and climate–carbon cycles. Spatial, temporal, and compositional changes in continental arc magmatism in Cordilleran-type orogenic systems have been shown to be a function of variations in subduction dynamics and overriding plate processes (Dickinson, 1975; Haschke et al., 2002, 2006; Kay et al., 2005; Trumbull et al., 2006; Kirsch et al., 2016; Ancellin et al., 2017; de Silva and Kay, 2018; Chapman and Ducea, 2019). The southern Central Andes, spanning the modern retroarc region of central Argentina and forearc region of Chile (Fig. 1), represent one of the few places on Earth where a long-lived (>100 m.y.) convergent margin provides the magmatic record necessary to evaluate spatial-temporal relationships among geodynamic processes, upper-plate deformation, and subduction- related magmatism.

The South American plate at 28°S–33°S is composed of numerous north-south–trending terranes of variable composition that may have induced a spatial control on the geochemical signatures of subsequent Andean tectono-magmatic regimes (Chapman et al., 2017). A critical debate persists over the timing and geologic expression of subduction initiation along the South American margin after Paleozoic terrane accretion. Traditional models have attributed magmatic compositional changes to Triassic subduction shutoff during intraplate anorogenic and/or extensional conditions followed by Early Jurassic subduction initiation (Kay et al., 1989; Mpodozis and Kay, 1992; Llambías and Sato, 1995; Kleiman and Japas, 2009). Andean magmatic records that span from the Early Jurassic (ca. 200 Ma) to present provide a foundation for interpreting the history of Andean subduction and orogenesis (Haschke et al., 2002, 2006; Trumbull et al., 2006; Folguera and Ramos, 2011; Balgord, 2017; Faccenna et al., 2017; Schellart, 2017; Butler et al., 2019; Sundell et al., 2019; Spencer et al., 2019; Chen et al., 2019). Although best known for Cenozoic contractional orogenesis, the western margin of South America has experienced varied tectonic regimes throughout the Phanerozoic, potentially preserving an earlier phase of subduction-related magmatism that dates back to the Carboniferous (Alasino et al., 2012; Dahlquist et al., 2013; del Rey et al., 2016). For long-lived (>100 m.y.) Cordilleran orogenic systems, variations in the position of the magmatic arc have been variably linked to changes in the dip of the subducted slab, the rate of plate convergence, and/or magnitude of forearc erosion (Ramos et al., 2002; Kay et al., 2005; Stern, 2011). Subduction angle has been interpreted to modulate mechanical coupling along subduction margins, control retroarc basin evolution, and drive phases of arc migration that are reflected in isotopic records (Ramos and Folguera, 2009; Horton, 2018a, 2018b; Chapman and Ducea, 2019). Evaluation of the operative spatial and temporal scales of trends in magmatic arc composition provides a means to assess proposals of high-flux events and magmatic lulls that may have generated cyclical or punctuated phases of arc magmatism (de Silva et al., 2015; Paterson and Ducea, 2015; Kirsch et al., 2016).

This study presents new detrital zircon U-Pb ages, U-Th ratios, and Lu-Hf isotopic data from modern river sands and sandstones throughout the retroarc region of central Argentina and forearc region of Chile (Fig. 1 to evaluate long-term (>100 m.y.), regional (>100 km scale) geologic patterns and changes in magmatic composition along the Andean margin. The detrital record provides valuable temporal controls on the evolution of upper-crustal composition (Condie et al., 2011; Cawood et al., 2012; Augustsson et al., 2016; Pepper et al., 2016; Balgord, 2017; Dhuime et al., 2017; McKenzie et al., 2018; Barber et al., 2019). Additionally, the spatial and temporal patterns in magmatism derived from compiled bedrock geochronological data sets provide constraints on magmatic arc width, from which we estimated average slab dip values for past subduction systems. Subduction angle can be calculated for a continental magmatic arc when the width of the arc and depth of melting are known (Fig. 2), assuming dehydration melting at a specific depth and vertical melt ascent (Keith, 1978, 1982; Peacock et al., 1994). To assess the relationships among arc width, subduction angle, and melting depth, we first characterized the modern geometry of the Nazca subduction zone, the spatial pattern of Andean arc magmatism, and the depth of melt generation. Integrated spatial, temporal, and compositional trends in magmatism enabled us to evaluate the spatial control on crustal geochemical signatures, timing of subduction initiation, and estimation of subduction angle variations along the transition between the central and southern Andes.

Igneous and associated metamorphic rocks in Argentina and Chile (Fig. 1) can be divided into four distinct groups: (1) Ediacaran to Devonian basement rocks of the Sierras Pampeanas and eastern craton; (2) Carboniferous–Triassic rocks of the Frontal and Principal Cordilleras; (3) Jurassic–Paleogene rocks of the Coastal Cordillera; and (4) Cenozoic volcanic rocks that get younger eastward from the Principal Cordillera toward the Argentina foreland.

Latest Ediacaran to Devonian igneous and meta-morphic belts are exposed across the Sierras Pampeanas. Key phases include the Pampean (555–515 Ma) and Famatinian (495–460 Ma) magmatic arcs, generated during east-dipping subduction (Ramos et al., 1986; Bahlburg et al., 2009). The Precambrian to Ordovician history included accretion of Laurentian terranes (including Cuyania) to the Gondwana margin, followed by the poorly constrained accretion of the Chilenia terrane (Ramos et al., 1984; Thomas et al., 2015; Rapela et al., 2016; Martin et al., 2019). Isolated granitic plutons of Devonian age across the Sierras Pampeanas region are interpreted to have been emplaced during extensional conditions (Dahlquist et al., 2013; Moreno et al., 2020).

Carboniferous–Triassic igneous complexes form much of the High Andes, with the Choiyoi igneous complex spanning the Principal Cordillera along the Chile-Argentina border and flanking regions to the west and east (Frontal Cordillera and parts of the Andean foreland). Batholith emplacement in the northern Frontal Cordillera involved many phases of Carboniferous to Triassic orogenic and postorogenic magmatism (Hervé et al., 2014; Sato et al., 2015). Late Carboniferous to early Permian magmatism was synchronous with growth of the NW-NNW–trending Gondwanide orogenic belt during east-dipping subduction (Giambiagi et al., 2014; Hervé et al., 2014). After the cessation of Gondwanide shortening and crustal thickening, the emplacement of calc-alkaline to alkaline bimodal intrusive suites and exceptionally thick (>5–10 km) ignimbrites of the Choiyoi Group (ca. 280–248 Ma) is consistent with crustal melting and possible postorogenic collapse prior to and during the earliest stages of Gondwana breakup (Mpodozis and Kay, 1992; Kleiman and Japas, 2009; Sato et al., 2015; Nelson and Cottle, 2019).

East-dipping subduction generated the Jurassic to Cenozoic Andean magmatic arc, which is defined by north-trending belts involving principally granite/granodiorite intrusions and andesite flows. These belts show a systematic eastward decrease in age, from Jurassic (200–165 Ma) and Cretaceous (130–90 Ma) rocks along the Chilean coast, to Paleocene–Eocene (67–38 Ma) and Eocene–Oligocene (27–18 Ma) units along the western flank of the Principal Cordillera, and finally Miocene (17–10 Ma) volcanic rocks in the eastern Principal Cordillera, Frontal Cordillera, and Argentine foreland (Parada et al., 1988, 1999; Kay et al., 1991; Haschke et al., 2006; Jones et al., 2016). Miocene magmatism was concurrent with significant (~150 km) retroarc crustal shortening largely focused in the Precordillera fold-and-thrust belt (Jordan et al., 1993; Allmendinger and Judge, 2014; Fosdick et al., 2015; Mosolf et al., 2019; Mackaman-Lofland et al., 2020). The modern Pampean segment of the Nazca–South American plate boundary is characterized by shallow subduction, an ~500-km-long gap in active volcanism, and foreland basement uplifts of the Sierras Pampeanas that persist >700 km toward the craton, potentially driven by subduction of the Juan Fernandez Ridge (Barazangi and Isacks, 1976; Jordan et al., 1983; Ramos et al., 2002; Ramos and Folguera, 2009; Alvarado et al., 2009).

The angle of subduction influences the position of the magmatic arc, thermal structure of the mantle wedge, and stress conditions in the overriding plate. To constrain long-term variations in past subduction angle, we utilized the time-space distribution of arc magmatism following the simplified approach of Keith (1978, 1982) in which dehydration melting occurs at specific depths during vertical ascent of magma through the overriding plate (Fig. 2). Dehydration melting occurs in the upper mantle over a narrow range of depths defined as a melting window with finite upper and lower limits (Fig. 2). From this relationship, subduction angle can be calculated for a continental arc with a known arc width and melting depth. The arc width to slab dip relationship does not require estimates of the arc-trench gap (distance between the magmatic front and subduction trench-slope break), which is a function of both subduction duration and subduction angle (Dickinson, 1973; Coney and Reynolds, 1977; Jacob et al., 1977; Jarrard, 1986; Isacks, 1988).

To identify the relationship of arc width to subduction angle, we first constrained the limits of the melting depth window beneath the modern Andean arc. Pleistocene to Holocene arc volcano locations were compared with the geometry of the subducted Nazca slab to extract values of arc width, slab dip, and slab depth directly below all volcanic locations (Fig. 3). Volcano locations (n = 1344; see Table S2¹) were derived from the Smithsonian Institution database (Global Volcanism Program, 2013) and a compilation of South American bedrock radiometric ages (Pilger, 2018). The geometry of the Nazca slab was constrained by geophysical data, which integrated active-source seismic data interpretations, receiver functions, local to regional seismicity catalogs, and seismic tomography models to characterize active global subduction zone geometries (Fig. 3; Slab2 model: Hayes et al., 2018). Comparisons between the position of Andean arc volcanic centers and Nazca slab depth provided the upper and lower limits on the depth of dehydration melting. From the compiled Andean arc segments, we derived a mean melting depth of 91–150 km (Fig. 3A; Table S3). This value is in agreement with the observed locations of modern dehydration melting and volcano locations above the expected slab depth determined from tomographic models of the subducting Nazca plate (Gilbert et al., 2006; Porter et al., 2012; Bianchi et al., 2013). Importantly, this data synthesis demonstrates a clear relationship for the modern magmatic arc in which progressively lower values of slab dips correlate with progressively greater width of the magmatic arc (Fig. 3B; Table S3).

We utilized an extensive data set of igneous ages and locations between 28°S and 33°S and between 65°W and 72°W to reconstruct arc width through time. The spatial distribution of magmatism was constrained by an extensive compilation (n = 1409; Table S4) of Phanerozoic bedrock radiometric ages (Fig. 4A; Pilger, 2018). Using longitudinal variations in igneous age, the width of the magmatic arc can be estimated for Cenozoic and Mesozoic igneous activity, but Paleozoic reconstructions are hampered by limited constraints on Paleozoic deformation. The sample density allowed for Cenozoic–Mesozoic arc width to be calculated for 2 m.y. intervals, with the exception of the Early Jurassic (190–170 Ma) magmatic arc (Fig. 4B). These calculations included palinspastic restoration of Cenozoic Andean shortening, including ~100 km of shortening in the Precordillera fold-and-thrust belt, 10 km of shortening along the Frontal Cordillera, 15 km of shortening in the Principal Cordillera, and 10 km of shortening along the Coastal Cordillera (Cristallini and Ramos, 2000; Farías et al., 2008; Allmendinger and Judge, 2014). Restorations of shortening were integrated over the entire data set, assuming shortening occurred during the major phase of Andean shortening (since 20 Ma). From this relationship, subduction angle was calculated for the Mesozoic–Cenozoic Andean continental arc using the restored arc widths and melt depths of 91–150 km (Fig. 4C; Table S5).

The long-term relationship between reconstructed Mesozoic to Cenozoic subduction angle and magmatic arc width shows the following trends: narrow (<50 km) arc widths associated with >50° subduction angles; moderate (50–150 km) arc widths associated with 20°–50° subduction angles; and wide (>300 km) arcs associated with <20° subduction angles (Figs. 4B and 4C). A clear temporal decrease in slab dip is apparent from the calculations (Fig. 4C). Triassic to Jurassic slab dips were the highest (50°–70°), followed by decreasing Cretaceous to early Cenozoic values (from 60° to 25°), and thereafter low Oligocene–Miocene values (from 15° to 10°) decreasing to 5° by Miocene–Pliocene time (Fig. 4C), in agreement with estimates for the modern Pampean flat slab geometry (Fig. 3B).

Although the width of a magmatic arc strongly correlates with subduction angle, our simple reconstruction involves several caveats. (1) The calculations assumed subvertical ascent of fluids and magmas from the zone of melting, which may be oversimplified for shorter time frames (Schurr et al., 2003; Trumbull et al., 2006). (2) Andean arc migration has been associated with extensive (35–85 km) forearc erosion and removal of trench-proximal (mostly Mesozoic) materials (Rutland, 1971; Coira et al., 1982; Kay et al., 2005; Stern, 2011), potentially resulting in an underrepresentation of true arc width for Mesozoic time steps. This approach does not explicitly account for episodes of forearc erosion, which would affect the position of the arc, but not the width. However, we note that any episode of forearc erosion that was accompanied by a shallowing in the dip of the subducted slab would be registered as an increase in arc width, per the relationship in Figure 4. (3) Older igneous materials may be covered or overprinted by younger volcanic events, limiting the extent of older outcrops and associated ages, which may limit arc width estimates for older time steps. (4) Major phases of intraplate Mesozoic volcanism occurred coeval with continental arc activity and were excluded from the calculations because their genesis far inboard from the subduction boundary was not related to slab dehydration melting (Fig. 4A). We used geochemical evidence (Rossel et al., 2013) and the spatial distribution to remove Jurassic (175–140 Ma) backarc magmatism from the arc width reconstruction. (5) Present-day flat slabs are principally characterized by the cessation of arc magmatism, such that our modern comparison between subduction angle and arc width lacks constraints for flat slab (<10° dip) segments. These caveats suggest that our calculations provide minimum values for arc width, which would generate modestly higher values of past subduction angle.

Zircon is a highly refractory accessory mineral that most commonly forms in intermediate to felsic igneous assemblages and high-grade metamorphic rocks and incorporates relatively high concentrations of U and Hf (typically hundreds of parts per million) with very low amounts of common Pb (Gehrels, 2014; Schoene, 2014). Zircon Th/U ratios increase with decreasing SiO₂ content, such that zircons from granitic rocks have distinctly lower ratios (0.55–0.80) than zircons from gabbroic rocks (0.8–1.2; Kirkland et al., 2015; Yakymchuk et al., 2018). Zircon Th/U values can also serve as proxies for tectonic regime. Extensional phases have been associated with zircons with wide a range of Th/U values, including significant populations of elevated Th/U zircon (>1.0) produced by hotter magmas (700 to >1000 °C) with increased fractionation from more primitive mafic melts (Kirkland et al., 2015; McKay et al., 2018). In contrast, low Th/U values (<0.5) with limited geochemical variability correspond to phases of contractional orogenesis (Kemp et al., 2009; Rubatto, 2017; McKay et al., 2018).

Zircon U-Th-Pb analyses were conducted on Miocene sandstone samples from the Andean foreland. They record erosion from an extensive drainage area that includes Cenozoic volcanic rocks, Carboniferous–Triassic igneous rocks, and Paleozoic sedimentary rocks (Figs. 5A and 5B; Capaldi et al., 2020). Mineral separation included crushing, grinding, and water table, heavy-liquid density, and magnetic susceptibility separations. Non magnetic heavy mineral separates were poured onto double- sided tape on 2.5-cm-diameter epoxy resin mounts, and then zircon grains were chosen randomly for analysis by laser ablation–inductively coupled plasma–mass spec-trometry (LA-ICP-MS) to determine Th/U isotopic ratios and obtain zircon U-Pb ages. Sample mounts were loaded into a large-volume Helex sample cell and ablated using a PhotonMachine Analyte G.2 excimer laser for analysis with a single- collector, magnetic sector Element2 ICP-MS at the University of Texas at Austin UTChron laboratory (Horton et al., 2015; Odlum et al., 2019). Corrections for depth- dependent, elemental and isotopic fractionation were performed using zircon standard GJ1 (600.4 ± 0.1 Ma; Jackson et al., 2004), yielding a concordia U-Pb age of 600.46 ± 0.7 Ma (mean square of weighted deviates [MSWD] = 0.52; n = 67). Secondary standards were analyzed periodically to verify the accuracy of the analyses; these included the Plesovice zircon (PL-1; 337.2 ± 0.4 Ma; Sláma et al., 2008), yielding a concordia U-Pb age of 339.3 ± 3.6 Ma (MSWD = 5.5; n = 9), and Pak-1 (43.03 Ma; in-house standard), yielding a concordia U-Pb age of 41.7 ± 0.5 Ma (MSWD = 2.8; n = 8). Zircon U-Pb ages and 2σ errors are reported as ²⁰⁶Pb/²³⁸U ages and ²⁰⁶Pb/²³⁸U versus ²⁰⁷Pb/²³⁵U discordance for analyses with less than 10% ²⁰⁶Pb/²³⁸U uncertainties, less than 20% discordance, and less than 5% reverse discordance (Table S6).

Zircon Hf isotopic compositions are sensitive to source rock type and crystallization environment, and they have been used to reconstruct the geo-chemical evolution of continental crust (Belousova et al., 2002, 2010; Kemp et al., 2006). Zircon Hf isotopic data are reported in epsilon units (ε) and presented in Hf evolution diagrams as εHf values representing the isotopic composition at the time of crystallization U-Pb age in reference to chondritic uniform reservoir (CHUR; Bouvier et al., 2008) and depleted mantle (DM; Vervoort and Blichert-Toft, 1999). Positive εHf values reflect juvenile crustal material originating from melts of depleted mantle origin, while negative εHf values are indicative of enriched melts derived from older recycled crustal material (Hildreth and Moorbath, 1988; Belousova et al., 2002; Hawkesworth and Kemp, 2006; Kemp et al., 2009). Studies of convergent margins have argued that a relative increase in zircon εHf is indicative of an extensional system characterized by outboard arc retreat during slab rollback and associated crustal thinning (Kemp et al., 2009; Spencer et al., 2019).

To characterize the isotopic composition of Phanerozoic magmatism, Hf geochemical analyses were conducted on detrital zircon grains with new and previously reported U-Pb age constraints (Figs. 5C; Tables S3 and S4; Levina et al., 2014; Capaldi et al., 2017, 2020). Numerous zircon grains were selected from each major U-Pb age component and then subjected to isotopic analysis by multicollector (MC) LA-ICP-MS at the University of Hong Kong. Zircon Lu-Hf-Yb isotopes were measured on a Nu Plasma high-resolution (HR) MC-ICP-MS coupled with a 193 nm excimer laser-ablation system (RESOlution M-50) by placing a 55-μm-diameter spot over the original U-Pb ablation pit following methods outlined in previous studies (Xia et al., 2011), and using zircon standards GJ1 (Elhlou et al., 2006) and 91500 (Wu et al., 2006; Table S7). Zircon standard 91500 yielded an average ¹⁷⁶Hf/¹⁷⁷Hf value of 0.282307 ± 0.00007 (MSWD = 0.45), and zircon standard GJ-1 yielded an average ¹⁷⁶Hf/¹⁷⁷Hf value of 0.282013 ± 0.000005 (MSWD = 0.34). Additional analytical metadata for U-Pb and Lu-Hf analyses can be found in the supporting information (Text S1 [see ^{footnote 1}]).

Zircon U-Pb age distributions incorporated data from Phanerozoic-aged grains from river sands, sandstones, and metasedimentary rocks in Argentina (Figs. 6A and 6B; Cristofolini et al., 2012; Capaldi et al., 2017, 2020; this study) and beach sands, river sands, and metasedimentary rocks in Chile (Figs. 6A and 6B; Table S8; Willner et al., 2008; Pepper et al., 2016). Felsic to intermediate melt compositions recorded by zircon Th/U ratios were compared with a synthesis of new detrital zircon Hf isotopic results (Table S7), previously published detrital zircon Hf results (Fig. 6C; Willner et al., 2008; Pepper et al., 2016; Otamendi et al., 2017), and zircon Hf isotopic results from bedrock samples (Fig. 6C; Table S9; Dahlquist et al., 2013, 2016, 2018; Hervé et al., 2014; Jones et al., 2015; Otamendi et al., 2017; Rapela et al., 2018). The zircon U-Th-Pb and Hf isotopic data sets were integrated with an extensive compilation of bedrock geochronologic data to define temporal, spatial, and compositional variation in Phanerozoic magmatism along the western margin of South America.

The compiled zircon U-Pb and isotopic results from Argentina and Chile (28°S–33°S) record 550 m.y. of arc magmatism with the following age modes and associated isotopic signatures (Figs. 7A and 7B): (1) a gradual decrease in the relative abundance of Ediacaran–early Devonian (550–385 Ma) ages coincident with low εHf values (−4 to −2 mean values) and Th/U ratios (0.48–0.65 mean values); (2) persistent Late Devonian–Triassic (380–208 Ma) age components characterized by a positive isotopic shift in both εHf values (+5 units) and Th/U ratios (+0.3); and (3) Jurassic–Neogene (190–4 Ma) ages that display recurring 20–30 m.y. high-flux events and lulls in zircon age abundances during continental arc magmatism and exhibit a negative isotopic shift in both εHf values (−5 units) and Th/U ratios (−0.2) around 25 Ma. General agreement between the detrital zircon and bedrock compilation for central Argentina and Chile indicates that U-Th-Pb and Lu-Hf data derived from detrital samples provide a powerful tool for interrogating long-lived records of magmatism and effectively fill potential data gaps in the available bedrock record. Compiled detrital zircon U-Th-Pb–εHf results are summarized in Table 1.

Time-space variations in magmatism were tracked using a compilation of Phanerozoic igneous bedrock radiometric ages (Fig. 7C; Pilger, 2018). Ediacaran–Early Devonian (550–385 Ma) magmatism occurred principally in the eastern (68°W–63°W) regions of the present-day Sierras Pampeanas and jumped westward with sequential terrane accretion events. Late Devonian–Triassic (380–208 Ma) magmatism was focused at 70.5°W–70°W, along the present-day Principal Cordillera, with the exception of an eastward expansion of magmatic activity during Permian–Triassic (280–238 Ma) time. Jurassic–Neogene (190–4 Ma) magmatism recorded a gradual eastward advance involving Jurassic to Paleocene–Eocene volcanism in the Coastal Cordillera (71°W–70°W) followed by a rapid eastward shift during the Oligocene–Pliocene.

Trends in magmatic activity, zircon geochemical signatures, and locations of arc magmatism were integrated to reconstruct the geodynamic evolution of the plate margin (Figs. 7 and 8). Major tectonic phases included: (1) Ediacaran–Devonian terrane accretion, (2) Carboniferous to Triassic Gondwanan subduction, and (3) Mesozoic–Cenozoic Andean orogenesis.

Zircon geochemical signatures in the eastern craton region correspond with Cambrian–Ordovician (550–440 Ma) igneous and metamorphic belts that exhibit evolved magmatic patterns and metamorphic zircon growth, which correspond to the collision of Laurentian continental blocks with the Gondwanan margin of South America (Fig. 8A; Schwartz et al., 2008; Ducea et al., 2010; Rapela et al., 2016). The relatively negative εHf isotopic values are indicative of evolved magma generation through thick Mesoproterozoic continental crust, a pattern consistent with shortening and crustal thickening during Pampean (555–515 Ma) and Famatinian (460–440 Ma) orogenesis (Coira et al., 1982; Astini et al., 1995; Casquet et al., 2001; Ramos, 2004, 2010; Schwartz et al., 2008; Charrier et al., 2015; Augustsson et al., 2016). The significant ~300 km westward shift in early Paleozoic magmatism over ~30 m.y. is compatible with subduction migration following these terrane accretion events. Diminished detrital zircon U-Pb age components at ca. 440 Ma suggest arc cessation occurred immediately following accretion of the Cuyania terrane (Fig. 8A; Thomas et al., 2015; Rapela et al., 2016; Martin et al., 2019).

Low relative proportions of Silurian–Devonian (435–385 Ma) zircon grains are associated with a magmatic lull and formation of metamorphic zircon supported by relatively enriched εHf values and low (<0.1) Th/U ratios. These low-abundance zircons may record magmatic quiescence related to continued collisional orogenesis linked to possible terrane accretion, prolonged flat slab subduction, or development of a passive or transform margin (Fig. 8B; Ramos et al., 1984; Sims et al., 1998; Bahlburg et al., 2009; Giambiagi et al., 2014; Cawood, 2005; Ramos, 2009; Rapalini, 2018; Dahlquist et al., 2018). We favor a model of continued shortening during final amalgamation of terrane accretion due to the lack of subduction-related magmatism and predominance of metamorphic zircon.

The Late Devonian–early Carboniferous geo-dynamic setting of the pre-Andean margin of Gondwana is unresolved, with some arguing for a passive-margin setting (Bahlburg et al., 2009; Hervé et al., 2013, 2014) and others suggesting postcollisional magmatism related to terrane accretion (Willner et al., 2012). The mid-Paleozoic lull in igneous activity was followed by Late Devonian magmatism located throughout the Sierras Pampeanas provinces (Fig. 7). Late Devonian to early Carboniferous (380–320 Ma) magmatism was expressed as widely distributed (>500 km) granitic plutons emplaced during extensional conditions (Fig. 8C; Dahlquist et al., 2013; Moreno et al., 2020). This is consistent with isotopically enriched magmatism with a mean εHf value of −2 and mean Th/U ratio of 0.6, linked to magmas derived from lithospheric mantle and recycled continental crustal sources with possible upper-crustal contamination (Dahlquist et al., 2013). This type of magmatism has been attributed to incipient subduction during slab rollback and upper- plate extension (Alasino et al., 2012; Dahlquist et al., 2013; del Rey et al., 2016). We interpret late Paleozoic (re)establishment of a convergent subduction margin along the western edge of South America on the basis of Carboniferous magmatism (Fig. 8C). The proposed initiation of subduction was associated with westward (trenchward) retreat of magmatic activity, and this is supported by petrographic and metamorphic age relationships within the Paleozoic forearc subduction complex that record convergence initiating between ca. 343 and 310 Ma (Willner et al., 2008, 2012). Carboniferous magmatism was interrupted by the punctuated Gondwanide compressional event (284–270 Ma), which was recorded by intense folding and thrusting along a NW-NNW–trending orogen in the Argentinian Frontal Cordillera and San Rafael uplift further south (Ramos et al., 1986; Llambías and Sato, 1995; Franzese and Spalletti, 2001; Kleiman and Japas, 2009; Ramos and Folguera, 2009; Giambiagi et al., 2014; del Rey et al., 2016).

The spatially extensive Permian–Triassic magmatic trend does not reflect a decrease in slab angle and associated magmatic arc widening; rather, it is exemplified by the Choiyoi igneous complex, a suite of intrusive and extrusive igneous rocks associated with an extensive ignimbrite flareup along the southwestern margin of Gondwana (Fig. 8D; Kay et al., 1987; Llambías et al., 2003; Nelson and Cottle, 2019). The Choiyoi igneous complex progressed from granitic/dacitic magmas during convergent arc magmatism (290–265 Ma) to intermediate or bimodal compositions during extensional magmatism (265–240 Ma) as the upper plate transitioned to postorogenic collapse (Mpodozis and Kay, 1992; Kleiman and Japas, 2009). Mechanisms that would have generated spatially broad, trenchward-retreating volcanism include subducting slab rollback and lithosphere delamination beneath Gondwana, with resulting crustal thinning and upper-plate extension driving ignimbrite flareup phases (Fig. 7D; Kay et al., 1987; Pankhurst et al., 2006; Nelson and Cottle, 2019).

Triassic (238–208 Ma) zircons record positive mean εHf values and a systematic increase in Th/U ratios to 0.9, suggestive of increased depleted components related to asthenospheric upwelling triggered by extensional thinning. The petrography, geochemistry, and isotopic composition of Triassic volcanic rocks show progressive changes over time yet indicate dominant subduction-related characteristics since the Carboniferous (Vásquez et al., 2011; Poma et al., 2014; del Rey et al., 2016; Coloma et al., 2017; Oliveros et al., 2018; González et al., 2018). In this manner, a combination of postorogenic collapse and backarc extension during slab rollback marked the onset of a principally extensional to neutral tectonic regime that characterized much of western South America during the Mesozoic (Fig. 8E; Rossel et al., 2013; Oliveros et al., 2018).

Reconstructed subduction angle estimates from the Late Triassic to present can address potential temporal linkages between variations in the subducted plate and associated magmatic geochemical signatures and the tectonic setting (Fig. 9). The distribution of arc magmatism reveals several shifts in subduction angle estimates that broadly decrease throughout the Mesozoic–Cenozoic (Fig. 9A), which are paralleled by a similar trend in decreasing zircon isotopic signatures (Fig. 9B). During the 230 m.y. span of Triassic to present arc magmatism, 8 high-flux events are recorded by zircon U-Pb age peaks at 230–204, 184–160, 156–136, 118–105, 102–82, 74–52, 46–32, and 28–4 Ma, which typically span ~10–25 m.y. and are interrupted by magmatic lulls that decrease in duration through time from Jurassic (20 m.y.) to Miocene (<5 m.y.). Increased arc frequency appears to correspond with a decrease in subduction angle and more negative εHf magmatic signatures.

Compositional and temporal patterns in arc magmatism were compared to the retroarc basin record to highlight how observed variations in magmatism temporally correlate with periods of Andean extension and orogenesis. A synthesis of Mesozoic–Cenozoic sediment accumulation histories for Andean extensional and compressional retroarc basins records the onset of major phases of subsidence and temporal shifts in accumulation, which provide constraints for the onset and pace of extension, contractional orogenesis, and tectonic subsidence (Fig. 9C; Kent et al., 2014; Fosdick et al., 2017; Mackaman-Lofland et al., 2019). Relationships among Andean subduction, arc magmatism, and retroarc basin evolution are related to first-order plate kinematics of the South American plate (Fig. 9D). We show global plate kinematic reconstructions for varied oceanic plates of the Pacific Ocean basin relative to a fixed South American plate from 170 Ma to present alongside our data to emphasize the role played by plate dynamics in Andean subduction (Seton et al., 2012; Maloney et al., 2013).

Triassic to Jurassic magmatism was coeval with relatively high subduction angle (55°–65° slab dip), which reflects a prolonged phase of slab- rollback and crustal extension (Fig. 9A). Zircons recorded consistently positive ~5 mean εHf values and >0.8 mean Th/U values, reflecting phases of juvenile arc and associated backarc magmatism indicative of slab-dehydration melting in the upper mantle (Fig. 9B; Vergara et al., 1995; Charrier et al., 2002, 2015; Rossel et al., 2013; Balgord, 2017). Time versus thickness diagrams display initial Triassic sediment accumulation curves that show progressively diminished sedimentation rates from 230–200 m.y. to 200–100 m.y. (Fig. 9C). This concave-upward sediment accumulation pattern is attributed to fault-controlled and subsequent thermal subsidence in the Triassic Cuyo and Ischigualasto basin systems (Ramos and Kay, 1991; Currie et al., 2009; Barredo, 2012; Kent et al., 2014) and Jurassic La Ramada backarc basin (Alvarez and Ramos, 1999; Oliveros et al., 2012; Mackaman-Lofland et al., 2019). The trench-normal absolute velocity of the South American plate reveals initial Triassic to Jurassic east-directed motion away from the trench, coeval with the breakup of Pangea (Fig. 9D; Storey, 1995; Ramos, 2009).

An Early Cretaceous (ca. 140–120 Ma) lull in Andean arc magmatism was associated with intraarc transtensional faulting along the Atacama fault system in Chile (Seymour et al., 2020). Early Mesozoic extensional basin subsidence was followed by Early Cretaceous neutral stress conditions and protracted thermal subsidence (Horton, 2018b; Mackaman-Lofland et al., 2019). Cretaceous (118–82 Ma) magmatism along the Coastal Cordillera of Chile records decreased subduction angle from 55° to 40° and isotopic decrease in εHf composition (Fig. 9B) that was coeval with a Late Cretaceous phase of contractional deformation (Parada et al., 2005; Litvak et al., 2007; Charrier et al., 2015; Horton et al., 2016; Mackaman-Lofland et al., 2019). Basin thickness versus time profiles exhibit a Late Cretaceous inflection around 105 Ma that corresponds with a convex-upward sediment accumulation curve in the La Ramada basin and initiation of deposition in the distal Bermejo foreland (Fig. 9C). Accelerated accumulation of sediment defines the onset of initial Andean shortening and deposition in a proximal Andean foreland basin (Fosdick et al., 2017; Mackaman-Lofland et al., 2019). A major shift in absolute plate motion at 110–90 Ma suggests the South American plate transitioned from a net retreating subduction system during Jurassic–Early Cretaceous time (including initial opening of the South Atlantic Ocean) to a net advancing system during the Late Cretaceous, contemporaneous with early Andean shortening (Figs. 8E and 9D; Torsvik et al., 2009; Seton et al., 2012; Horton, 2018a).

A latest Cretaceous magmatic lull (82–74 Ma) was associated with transpressional faulting in the Coastal Cordillera (Taylor et al., 1998). Paleogene subduction angle remained relatively consistent at 25°–35° eastward dip with minimal deformation recorded in the upper plate (Fig. 9A). Latest Cretaceous–Eocene (74–32 Ma) increased εHf and Th/U values are indicative of magmatism produced by slab-derived fluids and enhanced melting in the mantle wedge (Fig. 9B; Litvak and Poma, 2010; Jones et al., 2016). Subsequent Eocene to Oligocene volcanism records diminished εHf and Th/U values, suggestive of mantle wedge enrichment due to increased crustal assimilation or subducted sediment associated with increased forearc erosion (Kay and Mpodozis, 2002; Kay et al., 2005; Litvak et al., 2007; Jones et al., 2016; Lossada et al., 2017). The latest Cretaceous to early Paleogene interval was characterized by a depositional hiatus in the proximal La Ramada basin and slow accumulation in distal retroarc regions, which may reflect negligible shortening and orogenic loading during a neutral stress regime (Fig. 9C; Horton and Fuentes, 2016; Mackaman-Lofland et al., 2019). Reduced westward motion of the South America plate characterized the Paleogene, which is compatible with relatively diminished shortening (Fig. 9D).

Zircon age populations for the Oligocene–Pliocene (32–4 Ma) record broadening of the magmatic arc and an isotopic decrease in both zircon εHf and Th/U ratios that corresponds with the shallowing of the subduction angle (Figs. 8F and 9A; Kay and Mpodozis, 2002; Jones et al., 2016). Oligocene volcanic rocks yielded geochemical signatures characteristic of partial melting of mantle wedge material (Kay and Abbruzzi, 1996; Jones et al., 2016; Litvak et al., 2018) and were associated with intra-arc oblique faulting (Piquer et al., 2017; Mosolf et al., 2019; Mackaman-Lofland et al., 2019). Middle Miocene arc volcanism was defined by isotopic signatures indicative of crustal thickening (>50 km; Kay et al., 1991; Jones et al., 2015). Arc broadening, flareup, and crustal thickening were synchronous with shortening and erosion along the Frontal Cordillera and early foreland basin development across the retroarc (Fosdick et al., 2017; Buelow et al., 2018; Pinto et al., 2018; Mackaman-Lofland et al., 2020). The rapid thickening of the Andean crust and relative advancement of the arc in the late Miocene produced more evolved arc magmatism as the flat slab propagated eastward post–12 Ma (Fig. 9B; Kay and Mpodozis, 2002; Allmendinger and Judge, 2014; Jones et al., 2016; Capaldi et al., 2020). Following continued flattening of the subducted slab, major arc activity within the Principal Cordillera ceased by ca. 5 Ma, and diffuse volcanic centers across the Precordillera and Sierras Pampeanas continued until ca. 2 Ma (Kay and Abbruzzi, 1996; Ramos and Folguera, 2009). The main phase of Andean orogenesis corresponded to rapid Neogene sedimentation, where 3–8-km-thick fill denotes major flexural subsidence in foreland settings (Fig. 9C; Jordan et al., 1993; Fosdick et al., 2015).

The Pliocene to Quaternary crustal response to flat slab subduction involved a shift in deformation from a thin-skinned thrust belt to series of basement-involved foreland uplifts and eastward progression of coarse-grained deposystems within the broken foreland basin system (Capaldi et al., 2020). Initiation of shortening in the early Miocene may have been associated with a shift in Nazca plate motion, leading to enhanced trench-normal convergence, potential shallowing of subduction angle, and initial subduction of the Juan Fernandez Ridge, which rapidly migrated southward (~20 cm/ yr) along the South American margin (Fig. 9D; Yáñez et al., 2001; Martinod et al., 2010). Nazca plate reconstructions indicate that late Miocene (ca. 11 Ma) arrival of the relatively thick, younger oceanic crust of the Juan Fernandez Ridge to the Chile Trench at 31°S induced Pampean flat slab subduction (Fig. 9E; Pilger, 1981, 1984; Kay et al., 1991; Yáñez et al., 2001; Ranero et al., 2006). Neogene magmatism displays isotopically enriched signatures that correspond with subduction angle decrease, crustal thickening, and magmatism that advanced toward the craton and intruded thicker continental lithosphere.

Integration of Phanerozoic zircon U-Pb age distributions, U-Th and Lu-Hf geochemical data, and the spatial distribution of magmatism provides new insights into the potential drivers of the magmatic compositional variability along the South American margin. Here, we (1) discuss spatial trends in zircon isotopic signals with respect to variable crustal composition, (2) propose variations in subduction angle as the main driving mechanism for Mesozoic–Cenozoic magmatic arc migration, and (3) evaluate our results in the context of competing Cordilleran models for the South American margin.

The South American crust at 28°S–33°S contains north-south–trending terranes with variable composition that may have induced a spatial control on crustal geochemical signatures inherited by subsequent Andean tectono-magmatic regimes. Coupled zircon geochemical data and correlative bedrock locations exhibit a similar spatial pattern in both U-Th-Pb and Lu-Hf systems, such that magmatic activity in the western Principal Cordillera and Coastal Cordillera was associated with higher zircon isotopic values, in contrast with relatively lower values in eastern magmatic sources (Fig. 10). The parallel trends in zircon geochemistry appear to follow the proposed location of the suture between the Chilenia and Cuyania terranes at ~69°W. West of 69°W, Mesozoic to Cenozoic zircon εHf averages are mostly >0, and the Th/U ratio averages exceed ~0.75 (Fig. 10). Trenchward retreat of the magmatic arc commenced after extensive Choiyoi magmatism, with reduced crustal assimilation during magma ascent through thinner, possibly juvenile crust and melting of upwelling, depleted asthenosphere of the Chilenia terrane. East of 69°W, zircon εHf values range between −5 and −2 with Th/U ratios less than 0.7 (Fig. 10). Magmatism farther from the trench is characterized by more enriched radiogenic isotopic signatures due to the composition of eastern Precambrian terranes (i.e., Cuyania and Sierras Pampeanas). These trends highlight a spatial relationship that reflects contrasting source regions, with asthenospheric (juvenile) compositions near the trench and more lithospheric (evolved) contributions toward the craton (Hildreth and Moorbath, 1988; Pepper et al., 2016; Chapman et al., 2017). Thus, inboard arc migration produces a shift to more enriched compositions (Chapman and Ducea, 2019). Variation in subduction angle is a potential mechanism that affects Cordilleran arc migration, arc widening, and shifts in isotopic composition.

Spatial, temporal, and compositional changes in continental arc magmatism along the Andean margin are shown to be related to variations in subduction angle (Ramos et al., 2002). Here, we compared Mesozoic to Cenozoic (220–0 Ma) slab dip estimates with time-averaged εHf and Th/U zircon results (Fig. 9), which show decreasing (enriched) magma compositions with increasing arc width (Fig. 11A) and decreasing slab dip (Fig. 11B). Narrow (<75 km) magmatic arcs correspond to steep (>50°) subduction, with elevated εHf and Th/U zircon values. In contrast, wider (>300 km) magmatic arcs correlate with low-angle (<15°) subduction and diminished εHf and Th/U zircon values (Fig. 11). Moderate subduction angles (~20°–50°) generally show a decrease in εHf and Th/U zircon values with decrease in slab dip, albeit with greater variability. Others have demonstrated that arc migration driven by an increase in subduction rates and forearc erosion may also impart a more enriched geochemical signature in Andean magmatism (e.g., Kay et al., 2005; Stern, 2011). However, arc migration driven exclusively by subduction erosion cannot account for the observed range in preserved widths for the Mesozoic–Cenozoic Andean arc. Here, we suggest that more evolved magma compositions are associated with a decrease in slab dip that induces cratonward arc migration with enhanced contributions from Precambrian lithosphere, whereas slab steepening and arc retreat involve magmatic sources from depleted asthenospheric mantle of the accreted Chilenia terrane.

We compared the spatial, temporal, and compositional variations in the magmatic history of the southern Central Andes with numerous geodynamic, tectonic, and autogenic Cordilleran models to better understand the processes driving Andean orogenesis and subduction-related magmatism.

Recent geodynamic models predict that variations in upper-plate deformation are a function of the subducted plate interacting with and anchoring into the lower mantle (Husson et al., 2012; Faccenna et al., 2017; Schellart, 2017; Chen et al., 2019; Spencer et al., 2019). During incipient subduction, these models predict viscous coupling between the retreating slab and the upper mantle that induces a return flow that drags against the trench- proximal overriding plate toward the retreating subduction zone, inducing surface extension (Schellart et al., 2007; Husson et al., 2012; Schellart, 2017). Anchoring of the subducted slab in the lower mantle generates large-scale mantle flow that induces basal drag along the overriding plate toward the trench, which then drives orogenesis progressively on short 10–20 m.y. (Faccenna et al., 2017; Chen et al., 2019) to longer 100 m.y. (Schellart, 2017) time scales. In the southern Central Andes, the increase in juvenile magmatic sources and extensional basin development during Carboniferous subduction initiation and trenchward arc retreat (Figs. 7 and 8C) support the upper-mantle convection and slab retreat geodynamic model. Carboniferous subduction initiation was followed by the punctuated Gondwanide compressional event, potentially driven by slab anchoring and increased plate convergence. However, these geo-dynamic models predict subduction initiation in the Jurassic or Cretaceous followed by slab anchoring over 10–20 m.y., which is at odds with the magmatic record in the southern Central Andes that preserves a clear Carboniferous subduction initiation event. Longer subduction histories (~300 m.y.) along the South American margin require reevaluations of the timing of subduction initiation and/or reinitiation, which is critical to assessing when slab anchoring and subsequent orogenesis are predicted to occur in Cordilleran systems.

One class of tectonic models relates the degree of mechanical coupling between the overriding and downgoing plates to associated tectonic regimes operating during decoupled, neutral, and coupled subduction (Jarrard, 1986; Heuret and Lallemand, 2005; Martinod et al., 2010; Ramos, 2009; Horton, 2018a, 2018b). Subduction angle, in particular, has been interpreted to modulate overriding plate processes, such that high subduction angle and slab rollback settings, where overriding plate motion is away from the subduction trench, manifest as a retreating subduction zone, trenchward arc migration, crustal thinning, and extensional basin development (Royden, 1993). In the southern Central Andes, Triassic–Jurassic juvenile magmatism indicates trenchward arc retreat and associated slab steepening that occurred coeval with a phase of backarc extension and magmatism, which may reflect a phase of decoupled subduction (Fig. 9). The protracted phase of Cretaceous and Paleogene subduction exhibited constant arc width and relatively invariant radiogenic isotopic values that are consistent with a Cordilleran system that was neither contractional nor extensional; rather, the boundary conditions reflect neutral plate coupling (Horton and Fuentes, 2016; Mackaman-Lofland et al., 2019). The Neogene decrease in slab dip angle was coeval with retroarc crustal thickening and widening of arc magmatism toward the foreland region (Ramos et al., 2002; Haschke et al., 2006; Capaldi et al., 2020). The predicted increase in mechanical coupling along the subduction interface would have been related to the relative advance of the subducting slab toward the overriding plate, inducing shallowing of subduction angle (Schellart, 2017, 2020). Additionally, the accelerated relative convergence following the Jurassic to Cretaceous opening of the South Atlantic Ocean and far-field effects from the subduction of positively buoyant oceanic lithosphere appear to be first-order drivers of coupled subduction systems (Horton, 2018a).

Segments of Cordilleran systems that sustained high (>150 km) margin-perpendicular shortening and major crustal thickening have exhibited fluctuations (possibly cyclical) in deformation and magmatism, including rapid orogenic wedge propagation, high-flux magmatic events, lithospheric removal, and extensional collapse (Haschke et al., 2002, 2006; DeCelles et al., 2009, 2015; Wells et al., 2012; Carrapa and DeCelles, 2015; Ducea et al., 2015; Anderson et al., 2018). Though Cordilleran cyclicity models do not invoke external mechanisms such as variation in subduction angle, the models do provide predictions that may explain short-term (>20 m.y.) trends observed in our zircon compositional data. Deviations from the general trend of decreasing isotopic values with decrease in subduction angle occurred at 60 Ma and 35 Ma—time points that show increases in εHf and Th/U zircon values during shallowing of the Nazca plate (Figs. 9A and 9B). Such variance in isotopic trends may indicate intra-arc processes exhibited as mixed modes of deformation consisting of possibly cyclical phases of extension or neutral stress conditions (Folguera et al., 2006; Ramos et al., 2014; Giambiagi et al., 2015), potentially driven by dynamic convective removal of lower-crustal arc roots (Currie et al., 2015). Additionally, phases of intra-arc strike-slip deformation occur during periods of subduction angle change, which may drive tectonic reorganization and structural configurations that can induce changes to the bulk compositions of the lower crust–mantle lithosphere regions beneath arcs that lead to rejuvenation of melting and high-flux events (DeCelles et al., 2015; Ducea et al., 2015).

The Phanerozoic magmatic history of the southern Central Andes preserves a strong relationship between magmatism composition and arc migration, which corresponds with subduction angle variations. Magmatic perturbations over shorter (<50 m.y.) time scale may have been associated with intra-arc deformation processes and lithospheric removal. The protracted subduction angle decrease throughout the late Paleozoic to Cenozoic may have been linked to large-scale slab anchoring and global plate kinematics associated with opening of the Atlantic Ocean. Subduction angle decrease is shown to not only drive arc migration inboard, but also enhanced plate coupling and shortening within the overriding plate. Coupled arc advance into thicker, older lithosphere and crustal thickening during orogenesis are reflected as enriched isotopic signatures and evolved magmatic phases.

1. Phanerozoic detrital zircon U-Th-Pb and Hf results coupled with spatial shifts in magmatism from the southern Central Andes (28°S–33°S) show three significant trends: (i) Ediacaran–Silurian magmatism was characterized by enriched Hf isotopic values associated with crustal thickening followed by growth of metamorphic zircons during continued accretion of exotic terranes. The spatial distribution of coeval magmatic centers further reflects westward arc migration associated with terrane accretion along the Gondwana margin. (ii) Devonian–Carboniferous zircons record enriched Hf isotopic values and inboard magmatism during subduction initiation, followed by a Permian–Triassic phase with higher, depleted isotopic signatures located along the South American margin as the upper crust extended and lower crust delaminated during slab rollback. (iii) The westernmost Jurassic arc recorded depleted magmatic signatures, which shifted to moderately juvenile as magmatism advanced progressively inboard. Neogene magmatism recorded a dramatic isotopic decrease to enriched signatures as the Nazca plate shallowed and volcanism swept inboard from the trench.

2. Zircon U-Th-Pb and Hf analyses displayed similar spatial patterns in geochemical trends such that eastern (craton) magmatism exhibit 0 to −5 εHf and >0.7 Th/U average values, and western (Andean) magmatism exhibit 0–8 εHf and >0.7 Th/U average values. The observed spatial isotopic trends, with more radiogenically enriched magmatic signatures in the orogenic interiors, reflect the nature of the melt-source region, with greater asthenospheric (juvenile) contribution near the trench and more lithospheric contributions to magma compositions (evolved) away from the trench. Such geochemical trends highlight the spatial relationships among crustal age and composition in the eastern Precambrian terranes (i.e., Cuyania and Sierras Pampeanas) and more juvenile western terrane (Chilenia), which imparted a significant control on subsequent magmatism composition.

3. To calibrate the relationship between arc width and slab dip, we compared Pleistocene to Holocene Andean arc volcano locations with present-day Nazca plate geometries to extract values of arc width, slab dip, and slab depth. The spatial distribution of Mesozoic–Cenozoic (220–0 Ma) igneous rocks preserves Andean magmatic arc widening and narrowing through time, from which we estimated past subduction angle. Comparisons between Mesozoic–Cenozoic slab dip calculations and time- averaged zircon results exhibited a clear trend of decreasing εHf and Th/U values, indicating enriched magma compositions, associated with arc width increase and slab dip decrease. These compositional trends in zircons reflect the spatial heterogeneity of the numerous crustal terranes affected as the arc migrated relative to the trench, and subsequent modulations in crustal thickness during orogenesis and extension along the South American margin.

This research was supported by U.S. National Science Foundation grant EAR 1348031 awarded to N.R. McKenzie and B.K. Horton, a Jackson Distinguished Postdoctoral Fellowship to McKenzie, and student research grants from the Geological Society of America, American Association of Petroleum Geologists, Society for Sedimentary Geology (SEPM), and the Jackson School of Geosciences awarded to T.N. Capaldi. We thank Susanne Kay, Victor Ramos, Margo Odlum, Sarah George, Kristina Butler, Mark Helper, Emily Cooperdock, Angela Hessler, Lisa Stockli, and Jean Wong for helpful discussions. Constructive reviews by Andrea Stevens-Goddard, Joel Saylor, and Associate Editor Christopher Spencer significantly improved the manuscript.