Lithospheric foundering has drawn increasing attention as an important contributor to continental plateau formation, especially as a driver for increased elevation, extension, and mafic magmatism. This contribution focuses on the mafic magmatism that led to the creation of monogenetic volcanoes throughout the Puna Plateau of NW Argentina. Lavas from these volcanoes provide a means to evaluate the recent petrotectonic development of the plateau and, in combination with basement intrusive rocks, determine the isotopic composition and long-term evolution of the lithosphere beneath the central Andean back-arc domain. Mafic samples have trace-element concentrations and isotopic values typical of an enriched magma source region. We propose that the mafic magmas originated from an aged, metasomatized subcontinental lithospheric mantle. The lavas have isotopic values nearly identical to those of Early Ordovician Famatinian gabbro and granodiorite. We suggest the most primitive Puna lavas and Famatinian magmas originated from the same subcontinental lithospheric mantle. This implies that at least a thin portion of the subcontinental lithospheric mantle has remained intact beneath NW Argentina for the past ~485 Ma. A comparison to coastal Jurassic igneous rocks and mantle xenoliths from the nearby Salta rift system suggests that the sub-continental lithospheric mantle is chemically decoupled from the depleted mantle to the west and east. This has been the case for hundreds of millions of years despite long-term tectonomagmatic activity along the proto-Andean and Andean margin and within the continental interior. Our data almost certainly rule out large delaminating bodies, suggesting instead partial or piecemeal removal of the lithosphere beneath the Puna Plateau.


Research into the formation of continental plateaus has increasingly focused on the role of lithospheric loss in generating uplift, extension, and magmatism, particularly in northern Tibet (England, 1993; Molnar et al., 1993; Haines et al., 2003), the central Andean Plateau (Garzione et al., 2006; Ehlers and Poulsen, 2009), and in eastern Anatolia (Göğüs and Pysklywec, 2008a). Early research in the southern part of the Andean Plateau, the Puna of NW Argentina, was critical in establishing the links among high elevation, thinned lithosphere, extensional faulting, and small-volume, mafic volcanism and lithospheric loss (Kay and Kay, 1993; Kay et al., 1994). More recently, attention has refocused on this problem because of stable isotope paleoaltimetry, which suggests the rapid rise of the northern Andean Plateau, the Altiplano, and links this increase in elevation to convective removal of mantle lithosphere (Ghosh et al., 2006; Garzione et al., 2006). However, evidence in the Puna Plateau has not been significantly reevaluated for nearly two decades. This contribution focuses on geochemical data from the mostly mafic, small-volume, late Miocene, Pliocene, and Quaternary lava flows that were originally instrumental in shaping ideas about lithospheric loss. Additional geochronological data have been produced in this region in the intervening decades (e.g., Risse et al., 2008), but new geochemical data from mafic rocks are scarce. Our new geochemical data and interpretations challenge previous analysis, painting a more complex picture of the evolution of the subcrustal lithosphere in a long-lived subduction-zone setting.

The young (ca. 7 Ma to present), mostly mafic magmatism that led to the creation of numerous, small-volume monogenetic and simple polygenetic volcanoes throughout the Puna Plateau is of particular interest to this study. Lavas from these volcanoes provide a means to evaluate the petrotectonic development of the region and, in combination with basement intrusive rocks exposed throughout the plateau, determine the chemical composition and long-term evolution of the lithosphere beneath the Central Andes. Specifically, mafic volcanic rocks that are demonstrably derived from the upper mantle and have not undergone interaction with the overlying crust are of great importance in deciphering recent modifications of the uppermost mantle in response to various tectonic processes such as lithospheric foundering (e.g., Kay et al., 1994).

Here, we present new whole-rock major- and trace-element data and Sr-Nd-Pb isotope values for lavas sampled from the Antofagasta and Pasto Ventura Basins in the southern Puna Plateau. We compare these isotopic data to Sr and Nd isotope values from mafic and intermediate intrusive rocks of the Early Ordovician Famatinian arc (Otamendi et al., 2009), demonstrating their striking similarity. This important finding suggests that parts of the subcontinental lithospheric mantle and lower crust have remained intact and nearly invariant in composition near an active continental margin for ~485 Ma. This also implies that recent foundering of lower crustal and mantle lithospheric material as proposed by Kay and Kay (1993) and Kay et al. (1994) may have only thinned the subcontinental lithospheric mantle behind the active volcanic arc rather than removing it completely. Finally, it suggests that the “subduction signature” found in recent back-arc central Andean volcanism may not only be the result of Neogene subduction and subduction erosion, but rather a composite signal inherent to the subcontinental lithospheric mantle from long-lived subduction beneath central South America. A comparison of our data to coastal Jurassic volcanic and plutonic data supports this conclusion: the isotopic dissimilarity between our data and both the Jurassic magmatism and Cretaceous mantle xenolith–bearing basalts from the Salta rift system shows that the back-arc region of the central Andean volcanic zone is underlain by a subcontinental lithospheric mantle keel that is an isolated, relatively homogeneous chemical domain intact since the Early Ordovician, despite long-term periodic tectonomagmatic activity and postulated recent lithospheric loss.


The Altiplano-Puna Plateau is the second largest modern plateau on Earth and the largest formed in the absence of continental collision (Fig. 1). The plateau is primarily the result of extensive crustal shortening and thickening of the overriding plate as Nazca subducts beneath South America (Isacks, 1988; Allmendinger et al., 1997, and references therein). Magmatism may have contributed to crustal thickening (Beck and Zandt, 2002). Loss of the lithospheric root, whether through delamination or convective removal, may have contributed to the modern elevation, structural development, and magmatism of the plateau as well (Kay et al., 1994; Garzione et al., 2006). The geography of the southern segment of the plateau, the Puna Plateau, is characterized by fault-bounded ranges and intervening, high-elevation basins (Allmendinger et al., 1997; Coutand et al., 2001). Crustal thickness beneath the Puna Plateau is on average 50–55 km, as opposed to 75 km beneath the Altiplano Plateau (Yuan et al., 2000). The crust is thought to be mostly intermediate to felsic in composition (Beck and Zandt, 2002; Tassara et al., 2006). The total thickness of lithosphere is only 60 km, suggesting a relatively thin (5–10 km) mantle lithosphere (Tassara et al., 2006). This value is considerably less than expected considering the amount of shortening (>150 km) evidenced in the upper crust (Baby et al., 1997; Kley and Monaldi, 1998).

Miocene-Pliocene deformation on the Puna Plateau is characterized by thrust faults that accommodate northwest-southeast shortening and vertical extension (Allmendinger et al., 1989; Marrett et al., 1994). A compressional tectonic regime continues today around the margins of the plateau, where there is clear evidence for high-magnitude shortening in the Subandean zone in Peru and Bolivia (Hindle et al., 2005; McQuarrie et al., 2005) and in the Sierra Pampeanas of northwest Argentina (Allmendinger, 1986; Coutand et al., 2001). However, at high elevations, an overall northeast-southwest to north-south extensional regime now exists, with evidence for approximately N-S extension along normal and strike-slip faults since at least Quaternary time (Allmendinger et al., 1989; Marrett et al., 1994). These extensional faults may reflect the increase in gravitational potential energy accompanying detachment of foundering lithosphere, but they could also reflect an incipient gravitational extensional collapse of the plateau as a whole (Schoenbohm and Strecker, 2009).

Arc-related volcanism in the Andes has generated a frontal, linear (orogen-parallel) arc in the Western Cordillera that takes the form of large stratovolcano complexes composed primarily of andesite to dacite lavas, domes, and pyroclastic flows (de Silva and Francis, 1991; Allmendinger et al., 1997, and references therein), and a more diffuse domain of large calc-alkaline ignimbrite centers known as the Altiplano-Puna volcanic complex (de Silva, 1989). The Altiplano-Puna volcanic complex is the world's largest continental volcanic field, covering an area of at least 50,000 km2 in the southern Altiplano and northern Puna Plateaus. In the Puna Plateau, the Altiplano-Puna volcanic complex is marked by widespread andesite and dacite eruptions that occurred from 16 to 12 Ma. Proliferation of magmatism was temporally linked to uplift in the southern Puna Plateau (Coira et al., 1993). Large-volume ignimbrite eruptions began during late Miocene time (12–5 Ma) and continued until Pliocene time (Coira et al., 1993). They were likely the result of extensive crustal melting caused by mantle-derived magmas entering the continental crust (Francis et al., 1989). In general, arc-related volcanism coincides with the timing of compression in the region (Allmendinger et al., 1997).

Late Miocene to recent volcanism is typified by andesite to dacite stratovolcanicdome complexes and relatively small-volume, mafic volcanoes in the Puna Plateau. Mafic volcanism takes the form of monogenetic and simple polygenetic cinder cones, lava cones, and lava flows. In the southern and central Puna, most of these lavas were erupted after 3.5 Ma, but some date back to 7.3 Ma (Risse et al., 2008). Lava flows are basalt to andesite and are thought to be dominantly mantle derived (Kay et al., 1994). Many of the volcanic centers in the southern Puna Plateau appear to relate to local strike-slip and dip-slip, generally N-S–trending faults (Marrett and Emerman, 1992; Marrett et al., 1994) (Fig. 1). In general, individual volcanoes and related lava flows cover an area less than a few km2. The lava flows are usually no more than 3–4 m thick. Lavas are dominantly calc-alkaline with minor shoshonitic composition in areas of thickest lithosphere. Kay et al. (1994) argued based on trace-element compositions that some younger, mafic flows have intraplate-like (ocean-island basalt [OIB]) geochemical signatures. They point to a spatial correlation between the generally OIB geochemical signature and location of possible lithospheric foundering beneath the Puna Plateau.


Seventeen lava samples were collected from the southern Puna Plateau along a N-S transect in April and May of 2007 (Fig. 1). We collected some samples from lava flows previously analyzed by Kay et al. (1994). We also provide analyses of never before sampled lavas in the Antofagasta Basin and Pasto Ventura, the southernmost region of the plateau. Mafic samples share common petrographic and mineralogical features (Table 1). They have low phenocryst counts (mostly <10%), and many exhibit a glassy texture. Olivine is the most common and largest phenocryst. Clinopyroxene is also present in many samples. Clinopyroxene is typically subordinate to olivine, smaller in size, and often clustered. Both olivine and clinopyroxene are typically euhedral. Olivine and clinopyroxene phenocrysts often have glass and Fe-Ti–oxide inclusions. Elongate plagioclase phenocrysts are present in some samples. The matrix consists of olivine, clinopyroxene, plagioclase, and glass. Two of our samples, P07-ANT02 and P07-J201, are more felsic (“evolved”), suggesting that their precursor magmas must have interacted with the local crustal basement. They have very low phenocryst counts (<5%) and are glassy. Orthopyroxene as well as apatite and Fe-Ti oxides are present in the evolved samples.

Risse et al. (2008) presented a compilation of existing ages for Puna volcanism and added 22 new 40Ar/39Ar ages to the database. Ages range from 7.3 Ma to samples that give zero ages. The older samples are less mafic and possibly relate to the arc volcanism in the area. The majority of mafic volcanics were erupted subsequent to 3.5 Ma. Three of our samples are from lava flows dated by Risse et al. (2008): P07-ANT01 (0.3 Ma), P07-CH01 (0.8 Ma), and P07-J401 (4.5 Ma). Some of our other samples can be assigned relative ages based on edifice degradation and crosscutting relationships (see Table 1). No reliable age data exist for lavas from the Pasto Ventura region. Elemental concentrations and isotopic ratios do not show variation that correlates with age except that older volcanism (>7 Ma) in the Puna tends to be more felsic in composition then the younger, generally mafic lavas.

Major and Trace Elements

Whole-rock samples were analyzed by X-ray fluorescence (XRF) and inductively coupled plasma–mass spectrometry (ICP-MS) at the Geoanalytical Laboratory in the School of Earth and Environmental Sciences at Washington State University for major and trace elements (Table 2202) (Knaack, 2003; Johnson et al., 1999). The whole-rock samples were essentially liquids because of their low phenocryst count. Justification for whole-rock analysis includes low phenocryst count, sample homogeneity, and lack of alteration. The samples show a range of compositions, including basalt, trachybasalt, basaltic andesite, basaltic trachyandesite, andesite, and dacite (Fig. 2). A plot of MgO versus SiO2 shows a strong correlation (Fig. 3). Other major element trends compared to MgO include increasing FeO and CaO and decreasing Na2O with increasing MgO (Fig. 4), which support a model of contamination of primitive magmas by means of fractional crystallization and/or assimilation of crustal material to produce the more evolved compositions. Two distinct sample groups (high and low) are seen on the Na2O versus MgO plot, which reflect the “trachy” composition of some of our samples (see Fig. 2). TiO2 and Al2O3 show no correlation with MgO (Fig. 4). The lack of correlation between TiO2 and MgO is primarily the result of six mafic samples that contain a relatively high concentration of Ti at a given MgO value. These six samples also have higher Nb and Ta concentrations compared to our other samples but are still depleted in Nb-Ta relative to Th and La (Fig. 5). The variable Ti-Nb-Ta negative anomaly is discussed later herein in more detail. K/Ti versus MgO shows a fairly strong correlation (Fig. 4), especially if the Ti anomaly is taken into consideration. This correlation suggests that crustal material is being assimilated during differentiation of basalt toward more evolved compositions.

Figure 5 shows the overall trace-element composition of the three most primitive samples (P07-J101, P07-J301, and P07-SUR101) from our data set normalized to normal mid-ocean-ridge basalt (NMORB) (Sun and McDonough, 1989). In this case, normalization to NMORB allows for identification of elemental enrichment and depletion by subduction processes of a typical, depleted upper mantle composition. Figure 6 shows the rare earth element (REE) composition of the three most primitive samples as well as the most evolved sample compared to NMORB and normalized to depleted MORB mantle (DMM; Workman and Hart, 2005). Important trends in these plots include: enrichment of the lighter elements compared to depleted upper mantle compositions; fractionation between fluid and/or melt mobile large ion lithophile elements (LILEs) and immobile high field strength elements (HFSEs); fractionation between light rare earth elements (LREEs) and heavy rare earth elements (HREEs); and negative Ti-Nb-Ta anomalies.

The fluid and melt mobile elements (Cs, Rb, Ba, Th, La, Ce, Pb) are enriched compared to mantle values and the HFSEs and HREEs. These elements have only a weak correlation with wt% MgO (R value for Rb versus MgO = 0.88, Cs = 0.82, Th = 0.79) or none at all (Ba = 0.42, Ce = 0.27, La = 0.22), suggesting that their enrichment is a signature of the magma source region rather than a local crustal influence. Thus, there is a clear subduction signature in the mafic lavas, which is exemplified by the similarity between LILE concentrations and GLOSS (Plank and Langmuir, 1998) seen in Figure 5. GLOSS is a global average subducted sediment composition and is a useful tool for determining if fluid and/or melt from the sediment have contributed to the composition of the magma source region, typically by causing enrichment in LILEs and LREEs. Certain HFSEs (Nb and Ta) do not correlate with MgO. There is significant variation in depletion of these elements in rocks of similar SiO2 and MgO concentrations. Therefore, the Ti-Nb-Ta signature is also considered to be a characteristic of the magma source region. The HREEs are slightly enriched compared to DMM, possibly due to fractional crystallization, but are depleted relative to NMORB. This depletion suggests long-term melt extraction from the mantle source region. Further, the evolved samples show lower middle (M) REE and HREE compositions than the basalts, indicating a strong crustal component.


We analyzed fourteen of our samples for Sr, Nd, and Pb isotopes (Table 2). The isotopic ratios of 87Sr/86Sr and 143Nd/144Nd were measured by thermal ionization mass spectrometry (TIMS) on whole-rock samples. Fresh whole-rock samples were crushed to a fine powder in a shatterbox. Rock powders were put in large Savillex vials and dissolved in mixtures of hot concentrated HF-HNO3 or, alternatively, mixtures of cold concentrated HF-HClO4. Strontium and the bulk of the REEs were separated in cation columns containing AG50W-X4 resin, using 1 N to 4 N HCl. Separation of Sm and Nd was achieved in anion columns containing LN Spec resin, using 0.1 N to 0.5 N HCl. Strontium samples were loaded onto single Ta filaments with Ta2O5 slurry in order to keep the Sr cut loaded onto the middle third of the filament. Neodymium samples were loaded onto single Re filaments using resin beads.

Mass spectrometric analyses were carried out at the University of Arizona on an automated VG Sector multicollector instrument fitted with adjustable 1011 Ω Faraday collectors and a Daly photomultiplier (Ducea and Saleeby, 1998). Typical runs consisted of acquisition of 100 isotopic ratios. Fifteen analyses of standard Sr987 during the course of this study yielded mean ratios of: 87Sr/86Sr = 0.710255 ± 7 and 84Sr/86Sr = 0.056316 ± 12. Fifteen measurements of the La Jolla Nd standard were performed during the course of this study. The standard runs yielded the following isotopic ratios: 142Nd/144Nd = 1.14184 ± 2, 143Nd/144Nd = 5.118530 ± 11, 145Nd/144Nd = 0.348390 ± 2, and 150Nd/144Nd = 0.23638 ± 2. The Sr isotopic ratios of standards and samples were normalized to 86Sr/88Sr = 0.1194, whereas the Nd isotopic ratios were normalized to 146Nd/144Nd = 0.7219. The estimated analytical ±2σ uncertainties for samples analyzed in this study are: 87Sr/86Sr = 0.0014%, and 143Nd/144Nd = 0.0012%. Procedural blanks averaged from five determinations were: Rb 10 pg; Sr 150 pg; Sm 2.7 pg; and Nd 5.5 pg.

Washes from the cation column separation were used for separating Pb in Sr–Spec resin (Eichrom, Darien, Illinois) columns using protocol developed at the University of Arizona. Samples are being loaded in 8 M HNO3 in the Sr–Spec columns. Pb elution was achieved via 8 M HCl. Pb isotope analysis was conducted on a GV Instruments multicollector inductively coupled plasma–mass spectrometer (MC-ICP-MS) at the University of Arizona (Thibodeau et al., 2007). Samples were introduced into the instrument by free aspiration with a low-flow concentric nebulizer into a water-cooled chamber. A blank, consisting of 2% HNO3, was run before each sample. Before analysis, all samples were spiked with a Tl solution to achieve a Pb/Tl ratio of ~10. Throughout the experiment, the standard National Bureau of Standards (NBS)-981 was run to monitor the stability of the instrument.

All results were Hg-corrected and empirically normalized to Tl by using an exponential law correction. To correct for machine and inter-laboratory bias, all results were normalized to values reported by Galer and Abouchami (2004) for the NBS-981 standard (206Pb/204Pb = 16.9405, 207Pb/204Pb = 15.4963, and 208Pb/204Pb = 36.7219). Internal error reflects the reproducibility of the measurements on individual samples, whereas external errors are derived from long-term reproducibility of NBS-981 Pb standard and result in part from the mass bias effects within the instrument. In all cases, external error exceeds the internal errors and is reported here. External errors associated with each Pb isotopic ratio are as follows: 206Pb/204Pb = 0.028%, 207Pb/204Pb = 0.028%, and 208Pb/204Pb = 0.031%. We did not perform any age correction to the measured isotopic ratios, given the young age of the volcanic rocks.

The samples show two clusters in 87Sr/86Sr versus 143Nd/144Nd space (Fig. 7). One grouping consists of the two evolved lavas at high 87Sr/86Sr and low 143Nd/144Nd. The other grouping has a range of 87Sr/86Sr values between 0.70553 and 0.70704 and 143Nd/144Nd values between 0.51240 and 0.51260. The 87Sr/86Sr and 143Nd/144Nd values for the samples display only a weak correlation with wt% MgO. Further, there is no mineralogical evidence for crustal interaction/magma mixing in our most primitive samples. Sr and Nd isotopic compositions from our basalts are therefore interpreted to be representative isotopic values for the magma source region. The 206Pb/204Pb, 207Pb/204Pb, and 208Pb/204Pb values have ranges of 18.8012–19.0365, 15.6380–15.6900, and 38.8310–39.1885, respectively (Fig. 8). These values, including those for our basalts, are relatively radiogenic compared to values for a depleted mantle magma source region. Four samples are not included in our Pb isotope diagrams due to failure to collect the correct Pb cut in the Sr–Spec columns.


Small-volume lavas have erupted on the Puna Plateau over approximately the past 7 Ma (e.g., Kay et al., 1994; Risse et al., 2008). Our samples span the majority of this time (Table 1). Major-element chemistry shows that 15 of our samples are basalt, trachybasalt, basaltic andesite, and basaltic trachyandesites, some with <50.0 wt% SiO2 and >8.0 wt% MgO, which are indicative of a mantle-derived melt that has experienced little to no crustal interaction. One sample, P07-SUR101 (47.52 wt% SiO2, 10.30 wt% MgO, 505 ppm Cr, and 163 ppm Ni), is the most primitive basalt yet reported for the central Andean Plateau region in NW Argentina. Unlike some of the basaltic andesite, this and other basalt samples do not have crustal xenoliths or quartz/feldspar xenocrysts. Further, P07-SUR101 contains only ~8% phenocrysts, which are mostly olivine (Table 1). Removal of phases reduces the MgO content by ~2%, leaving a mantle-type MgO concentration of >8 wt%. The basalts experienced minimal interaction with the overlying crust during ascent because they show no chemical or petrographic evidence for assimilation of felsic rock. Therefore, the elemental and isotopic composition of our basalt and trachybasalt samples is useful for evaluating the chemical characteristics of the magma source region beneath the Puna Plateau. Two of our samples are a silicic andesite and a dacite with low wt% MgO. They likely have an extensive crustal component and thus provide insight into the composition of possible crustal assimilant. For breadth of study, we have included all previous isotopic analyses from the Puna Plateau in Figures 7, 8, 9, and 10.

Importantly, LILEs and LREEs in basalt samples show enrichment compared to NMORB and enrichment relative to HFSEs and HREEs in the same sample (Figs. 5 and 6). This is a widely documented phenomenon in convergent-plate settings, and it suggests that Puna back-arc magmas originated from a magma source region that was enriched during a time when the mantle currently beneath the Puna Plateau was part of the mantle wedge fluidization zone. Spatial and temporal variations in Th, Ba, La, and other LILE and LREE concentrations indicate a complicated mixture of fluid and possibly melt addition to the magma source region. Samples also show variable HFSE negative anomalies (Fig. 5). This occurs most prominently with Nb and Ta. Most samples show large negative Nb-Ta anomalies compared to other incompatible elements, which are common in convergent-margin magmas. However, there is variation in the degree of the Nb-Ta negative anomalies and concentration at similar SiO2 and MgO wt%. For example, the Nb concentration range in basalt and trachybasalt samples from our data set is 14.1–31.3 ppm. There is no correlation between Nb-Ta and MgO. Thus, we consider Nb and Ta concentrations and related anomalies to be inherited from the magma source region, not the overlying continental crust. Lavas with higher Ti-Nb-Ta concentrations, and thus smaller negative anomalies relative to neighboring elements on a trace-element diagram, are scattered throughout the Puna Plateau. They also do not show a temporal pattern, making variations in recent melt removal an unlikely explanation. It is possible they were derived from source regions with increased amounts of amphibole, titanite, illmenite, and/or rutile. However, REE concentrations and patterns for basalts do not show supporting evidence for this hypothesis (Fig. 6), such as the expected negative Dy anomalies if amphibole were responsible for variation in the HFSE anomalies. Variable magma source region trace-element characteristics are common for spatially extensive monogenetic volcanic fields, and clarification of the Ti-Nb-Ta anomaly in the Puna Plateau data set warrants further study.

Puna lavas have Sr and Nd isotopic compositions that are typical for enriched continental mantle lithosphere in South America and elsewhere (e.g., lithosphere-derived “western Great Basin” basalts discussed in Kempton et al., 1991; for review of physical and chemical properties of continental mantle, see Carlson et al., 2005) (Fig. 7). This observation includes our basalts and published data from mafic samples (e.g., Kay et al., 1994), which have 87Sr/86Sr values greater than 0.70550 and 143Nd/144Nd less than 0.51265. As pointed out by Kempton et al. (1991) and many others, it is impossible for magma from the depleted mantle to obtain these Sr and Nd isotope values through crustal assimilation or assimilation and fractional crystallization (AFC) processes and maintain a basalt major-element composition, particularly such high wt% MgO and Ni and Cr concentrations. Further, Pb isotope values for Puna basalts are also radiogenic (Fig. 8) compared to what is expected for magmas derived from a depleted mantle. Based on both the trace-element and isotopic evidence, we suggest that mafic Puna magmas originated from an aged subcontinental lithospheric mantle that had also been previously metasomatized.

Plutonic rocks of the Ordovician Famatinian arc are exposed throughout northwest Argentina. Geochemistry, geochronology, and examination of field relations over the past 20 years have shed some light on the evolution of this magmatic arc (Ramos et al., 1986; Dalla Salda et al., 1992; Rapela et al., 1992; DeBari, 1994; Kleine et al., 2004; Pankhurst et al., 1998; Pankhurst et al., 2000). In short, the arc was formed by plate subduction on the margin of the Gondwana foreland. It is generally agreed that the major phase of magmatism took place during the Early Ordovician (ca. 490–480 Ma) in both the south (Pankhurst et al., 1998, 2000) and the north (Pankhurst et al., 2000; Kleine et al., 2004). Later intrusions of peraluminous granitoids occurred until the mid-Ordovician. Preserved rocks are dominantly calc-alkaline and include gabbro, gabbronorite, norite, tonalite, monzonite, granodiorite, and granite, some with mafic enclaves.

We compare Sr and Nd isotopic data of Puna volcanics to the most recently acquired data for the Famatinian arc (Stair, 2008; Otamendi et al., 2009) in Figure 7. Mafic (gabbro and gabbronorite) Famatinian samples have initial Sr and Nd isotopic values nearly identical to mafic lavas from the southern Puna Plateau. The limited Sr and Nd isotopic data in the literature, such as that from Kleine et al. (2004) for samples from the central and northern Puna Plateau, support this conclusion. Therefore, we suggest that Famatinian gabbro and Puna basalt were generated from the same subcontinental lithospheric mantle magma source region. This implies that at least a thin portion of the subcontinental lithospheric mantle has remained intact beneath NW Argentina during the past ~485 Ma. Radiogenic growth of 87Sr and 143Nd over the ~500 Ma time period elapsed since the formation of the Famatinian arc is minor, if concentrations typical of the subcontinental lithospheric mantle are used for Rb, Sr, Sm, and Nd (Carlson et al., 2005). We conservatively estimate an ~0.0001 increase in the 87Sr/86Sr and ~0.5 ϵNd of the mantle over 500 Ma.

Our silicic andesite and dacite samples, P07-ANT02 and P07-J201, respectively, and the more evolved samples from the literature have Sr and Nd isotopic ratios that trend toward and overlap granodiorite samples in Figure 7. The evolved Puna lavas have a strong elemental crustal signature. Because Puna lavas and Famatinian gabbro and intermediate rocks produce an obvious mixing trend in Sr-Nd isotopic space (Stair, 2008), we suggest that the relationship between Famatinian mafic and intermediate magmas was magma mixing during the Early Ordovician and the relationship between evolved Puna lavas and Famatinian arc rocks is recent crustal assimilation involving contamination of a basaltic magma derived from the subcontinental lithospheric mantle.

Binary mixing between our most primitive basalt P07-SUR101 and a high-87Sr/86Sr, low-143Nd/144Nd Famatinian granodiorite sample (CC1, Fig. 7) produces a crustal assimilation line that explains the evolution of many Puna lavas. Puna basaltic andesite magmas have mantle:crust mixtures of ~80:20. More evolved andesite and dacite magmas show mixtures with greater amounts of granodiorite crust. This contamination may result from interaction of ascending basaltic magma with lower crust or a mid-crustal melt layer (e.g., Yuan et al., 2000) that is at least partly composed of Famatinian granodiorite. Extremely evolved magmas are likely crustal melts of Famatinian granodiorite, granite, or metamorphosed rock.

A Famatinian granodiorite crustal end member is not appropriate for explaining the evolution of all Puna magmas. Therefore, we created a hybrid crustal end member that contains the lowest Sr (0.713478) and Nd (0.512038) isotopic ratios among Famatinian samples (CC2, Fig. 7). The hybrid end member better explains the evolution of some Puna magmas with relatively low 87Sr/86Sr. This isotopic composition could be part of the lowermost crust, possibly a more mafic layer (thus the less radiogenic 87Sr/86Sr) that was contaminated prior to Puna magmatism by ascending melts with low 143Nd/144Nd compositions.

Primitive basalts have not erupted in the active Andean arc at latitudes similar to the Puna Plateau. Only one volcanic sample with <57% SiO2 and >5% MgO has been analyzed for both Sr and Nd isotopes. This sample has 56.2 wt% SiO2, 5.81 wt% MgO, and isotopic ratios of 87Sr/86Sr 0.70624 and 143Nd/144Nd 0.51247 (Matthews et al., 1994). Volcanic samples from the Puna Plateau with similar major-oxide chemistry tend to have more radiogenic Sr isotope values and less radiogenic Nd isotope values. Assuming crustal assimilation or AFC magma evolution, this suggests that the magma source region composition beneath the active arc is depleted compared to that for Puna magmas. However, arriving at a sound conclusion based on this inference is impossible due to lack of arc basalts in the region. We have therefore compiled published data from Jurassic volcanic (47–70 wt% SiO2) and plutonic (43%–78% SiO2) rocks from the continental margin to the west (Lucassen et al., 2006) and Cretaceous mantle xenolith data from the Salta rift system to the east and southeast (Lucassen et al., 2005) for comparison to our Puna volcanic samples (Figs. 8, 9, and 10). Figure 9 shows that the Mesozoic volcanic and plutonic rocks used here as a proxy for the active arc have lower 87Sr/86Sr and higher 143Nd/144Nd at a given wt% MgO. Coastal Jurassic samples and mantle xenoliths from the Salta rift are more similar in isotopic composition to MORB (Figs. 8 and 9) than an aged subcontinental lithospheric mantle, trending toward a depleted mantle at mafic compositions. We also plotted isotopic ratios versus longitude for mafic Puna Plateau volcanics and mafic coastal Jurassic samples to illustrate the clear separation in Sr, Nd, and Pb isotopic values (Fig. 10). Thus, Figures 8, 9, and 10 show that the lithospheric magma source region for the Puna Plateau is chemically decoupled from the depleted mantle directly to the west and east. This has been the case for hundreds of millions of years despite long-term tectonomagmatic activity along the proto-Andean and Andean margin and within the continental interior.


Elemental and isotopic data from Puna Plateau lavas indicate that parts of the subcontinental lithospheric mantle and presumably the lower crust are in place beneath the plateau. Comparison to isotopic data from Early Ordovician Famatinian plutonic rock suggests that the preserved lithospheric mantle domain has been inherited from the Famatinian arc. However, the lithosphere is anomalously thin, and the lithospheric mantle has been reduced to <10 km, as evidenced by geophysical data (e.g., Tassara et al., 2006; McGlashan et al., 2008). Recent foundering of the mantle lithosphere and potentially lowermost crust, whether through gravitationally driven convective removal or delamination, has been proposed to explain the thinned lithosphere beneath the Puna Plateau, N-S extension within the plateau, and the late Miocene to present mafic volcanism (Kay and Kay, 1993; Kay et al., 1994; Whitman et al., 1996). This is supported by tomographic data, which suggest the presence of a detached lithospheric block beneath the Puna Plateau (Schurr et al., 2006). This apparent conflict between geochemical data and geophysical and structural data requires an explanation.

One possible explanation is that the radiogenic Sr and Pb ratios and low Nd ratios in the Puna Plateau basalts could reflect the signature of crustal material emplaced along the base of the remaining lithosphere as the result of Neogene subduction erosion (Kay et al., 2008). The occurrence of this tectonic process has been discussed for the Chilean margin (e.g., von Huene and Scholl, 1991; Kay et al., 2005). Subduction erosion involves the removal of trench sediment and forearc crust from the overriding plate and introduction of this material into the subduction zone. For the Puna Plateau, the eroded component should be Jurassic igneous material from the continental margin immediately to the west. Figures 8, 9, and 10 show isotopic data from Jurassic volcanic and plutonic rocks located along the central South American margin at latitudes equivalent to the Puna Plateau (21°40′S–27°0′S) (Lucassen et al., 2006). Clearly, mixing between these isotopic values and depleted mantle cannot produce the isotopic ratios present in primitive lavas from the Puna area. Therefore, although material from subduction erosion could be altering the composition of the mantle wedge beneath the active Andean arc, it is not entering the mantle beneath the Puna Plateau.

This brings us back to an explanation involving removal of the mantle lithosphere. Geophysical and mass balance arguments (e.g., Beck and Zandt, 2002) make lithospheric foundering inescapable, and delamination/convective removal of some form is favored in the area due to the occurrence of mafic volcanism with approximately OIB trace-element characteristics and a rapid change in tectonic regime (Kay and Kay, 1993; Kay et al., 1994). As discussed in the following paragraphs, however, this loss may have occurred partially or in a piecemeal fashion rather than as rapid removal of the entire mantle lithosphere.

Evidence against wholesale removal of the subcontinental lithospheric mantle includes the lack of mafic lavas erupted with asthenospheric isotopic compositions postdating the hypothesized major delamination event. If either convective dripping or delamination had occurred, then the isotopic composition of primitive basaltic volcanism should have a depleted mantle chemical signature, perhaps similar to that of mantle xenoliths from the Salta rift system (e.g., Lucassen et al., 2005). Sudden removal of large amounts of lithospheric material via convective dripping should also be accompanied by upwelling of asthenospheric mantle and relatively large amounts of decompression-related melting. For example, if the thickness of the lithosphere was 120 km and sudden foundering of a spherical body of the lower lithosphere (60 km in diameter) were to occur, one would expect about a 15-km-thick melt column from decompression melting of the asthenosphere replacing the space alone (using the parameterization of McKenzie and Bickle, 1988). There is no indication of such large volumes of basalt either at the surface or at depth. Interestingly, the phenomenon of only minor amounts of surface basalts derived from the lithosphere also occurs in other places in which delamination has been proposed (Farmer et al., 2002—Sierra Nevada, California; Manthei et al., 2009—Coast Mountains Batholith, British Columbia and southeast Alaska).

The issue of only minor volumes of surface basalt is difficult to address, as it is possible that a large volume of more mafic magmas may be present at depth as intrusive bodies. The isotopic data are more helpful in providing an alternative explanation to large-scale lithospheric foundering. The exclusively enriched isotopic composition of the existing mafic product has two explanations: (1) the melting portion is the downgoing mantle lithosphere because it is hydrated (Elkins-Tanton, 2005), and/or (2) the size of delaminating domains is small (e.g., 1 km or less), which in turn requires dripping to be slow, thus not allowing major upwelling and partial melting of asthenosphere. Therefore, while the chemical and isotopic composition of the mafic rocks studied here almost certainly rules out large delaminating bodies, our data do not rule out other scenarios for lithospheric foundering in which only part of the lithosphere is removed (partial removal) or in which it has proceeded gradually by smaller foundering events and the entire subcontinental lithospheric mantle has not yet been removed (piecemeal removal). Past lithospheric foundering models in subduction or collisional environments (e.g., England and Houseman, 1989; Nelson, 1991; Zandt et al., 2004) assume drip sizes of tens of kilometers in diameter, to a certain extent because geophysical observational evidence is coarse enough to only allow detection of large anomalies. More recent modeling results demonstrate that delamination, in which the mantle lithosphere peels away from the overlying crust, inherently removes the entire lithosphere, but that convective Rayleigh-Taylor dripping can leave a remnant of subcontinental lithospheric mantle, depending on the rheologic properties of the model (Göğüs and Pysklywec, 2008b). Thus, we favor a scenario in which lithospheric foundering has occurred in the Puna Plateau, possibly since the late Miocene, but as a piecemeal or partial process rather than the wholesale removal of a large, coherent lithospheric body.


We have shown that the central Andean back-arc Puna Plateau is underlain by a mantle domain (of unconstrained thickness) that resembles the isotopic composition of the Ordovician Famatinian volcanic arc, the products of which are abundantly found throughout NW Argentina in surface exposures. The Famatinian arc was a continental arc that shaped much of the western margin of proto–South America before outboard accretion of terranes took place, and it was later subject to Andean tectonism and magmatism. This domain has presumably inherited its isotopic composition from the development of the Famatinian arc, although further melt depletion and metasomatism through subsequent tectonomagmatic events cannot be ruled out. What can be ruled out, however, is that no mantle-derived melt products with asthenospheric compositions are found on the Puna Plateau. The striking similarity of isotopic ratios between widespread Ordovician surface igneous rocks and all late Cenozoic mantle-derived mafic magmas further suggests that the mantle source was removed from convection for at least ~500 Ma; the subplateau mantle that supplied Cenozoic Puna region volcanics must therefore be enriched, old, subcontinental lithospheric mantle.

Further, this domain has remained chemically distinct from the depleted mantle. Recently proposed lithospheric foundering for the central Andean Puna region (Kay et al., 1994) seems inescapable based on thermal, mass balance, and other constraints. However, we show here that only a portion of the subcontinental lithospheric mantle has been removed, and we infer that lithospheric loss took place as a piecemeal or partial process rather than wholesale removal of a large, coherent lithospheric body. Finally, although subduction erosion may be contaminating the mantle wedge to the west, it is clearly not affecting the chemical composition of the Puna Plateau subcontinental lithospheric mantle domain.

This project was funded by National Science Foundation grant EAR-0635584 to L. Schoenbohm. M. Ducea acknowledges generous support for this project from ExxonMobil's COSA fund to the University of Arizona. S. Drew thanks J.A. Palavecino for assistance in the field during sample collection. We wish to acknowledge K. Putirka and B. Cousens for helpful reviews that improved the manuscript considerably.

1GSA Data Repository Item 2009236, Table DR1, is available at www.geosociety.org/pubs/ft2009.htm, or on request from editing@geosociety.org, Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301-9140, USA.