Geochemistry and geology of spatially and temporally associated calc-alkaline (I-type) and K-rich (A-type) magmatism in a Carboniferous continental arc setting, Pataz gold-mining district, northern Peru

The Pataz–Parcoy gold-mining area, in the Eastern Andean Cordillera of northern Peru, is located on the western margin of the Amazonia craton, in a transtensional jog on the Cordillera Blanca Fault. Episodic subduction, accretion and rifting have taken place in the Eastern Andean Cordillera since the Mesoproterozoic. In the Pataz district, the Cambrian–Ordovician Vijus and Atahualpa formations comprise dacitic and rhyodacitic volcaniclastic rocks that formed by fractionation and/or crustal assimilation of a regionally more abundant andesitic parent melt. Mississippian magmatism began with enriched tholeiitic magmas that lack a Ti anomaly and formed by melting undepleted metasomatised asthenosphere. These magmas assimilated variable amounts of continental crust and were emplaced as Vista Florida Group volcaniclastic rocks, and dioritic plutons of the Pataz batholith. Ponding of these mantle-derived magmas in andesitic lower crust caused partial melting, which generated the volumetrically dominant, granodioritic component of the calc-alkaline (I-type) Pataz batholith. Within a few million years but following rifting, and uplift of the Pataz batholith, relatively K-rich (A-type) magmas formed by melting of a mid-crustal tonalitic source, and were emplaced in the upper crust as the latitic to rhyolitic Lavasen Volcanics and Esperanza subvolcanic complex. Anatexis in the mid crust was promoted by a second batch of magma derived from the undepleted metasomatised asthenosphere, evidence of which is preserved as mafic components of the Esperanza subvolcanic complex and post-Esperanza dolerite dykes. A K-rich magma chamber, the source of the Lavasen Volcanics and the Esperanza subvolcanic complex, is proposed as a possible source of the ore fluid, which deposited gold in veins hosted by the Pataz batholith.


INTRODUCTION
The Pataz-Parcoy gold field is located near the western boundary of the Eastern Andean Cordillera, on the western margin of the Archean to Paleoproterozoic Amazonia craton ( Figure 1). The Pataz district, in the northern part of the Pataz-Parcoy gold-mining area, is a rift and graben terrane resulting from Carboniferous and Cenozoic extensional tectonic events superimposed on a continental margin subduction zone (Witt et al. 2013a). Two major igneous episodes occurred during the Mississippian (early Carboniferous), and relevant geochronological data for these episodes are summarised in Table 1. These intrusive and extrusive rocks were emplaced into or onto latest Neoproterozoic to Ordovician basement and volcano-sedimentary units of the Eastern Andean Cordillera. The first Mississippian magmatic episode resulted in emplacement of the Pataz batholith, an elongate, NNW-trending body of granodiorite and diorite that hosts the majority of the auriferous quartz-carbonatesulfide veins (Haeberlin et al. 2004). SHRIMP (Witt et al. 2013a) and ID-TIMS (Schaltegger et al. 2006) U-Pb in zircon geochronology indicates an age of 338-336 Ma for the emplacement of the Pataz batholith in the Pataz district. A U-Pb in zircon age of 329 Ma (Table 1) from granodiorite in the Parcoy district suggests intrusive activity may have migrated southwards over time. The second Carboniferous igneous event followed shortly after emplacement of the Pataz batholith and resulted in deposition of the Lavasen Volcanics in the Lavasen rift at ca 334 Ma (Witt et al. 2013b). The Esperanza subvolcanic complex, which gives a SHRIMP zircon age within error of that for the Lavasen Volcanics, was emplaced into the western margin of the rift, adjacent to the Pataz batholith. *Corresponding author: wittww@iinet.net.au There are limited published whole-rock geochemical data for the Pataz batholith. Schreiber et al. (1990) presented some basic classification diagrams and concluded that the batholith comprised calc-alkaline, I-type granites. Their samples fall in the volcanic arc field on an Rb vs YþNb plot (Pearce et al. 1984), although Schreiber et al. (1990 suggested that, based on structural setting of the intrusions, they should be classified as post-collision granites. Macfarlane et al. (1999) reported Pb and Nd isotope data for the Pataz batholith, which they interpreted to show that the batholith-forming magmas had interacted substantially with older continental crust. Miskovic et al. (2009) reported whole-rock geochemical data from a regional study of granitic intrusions in the Eastern Andean Cordillera, with samples collected between 6 S and 16 S. Of these, several samples came from the Pataz batholith, but none were reported from the Pataz district. Carboniferous granites of the Eastern Andean Cordillera (such as the Pataz batholith) were classified as Cordilleran-type, continental-arc-related intrusions by Miskovic et al. (2009). In this contribution, we present whole-rock geochemical data for three groups of igneous rocks in the Pataz district (the northern half of the Pataz-Parcoy gold-mining area). The earliest group comprises early Paleozoic felsic to intermediate volcanic rocks of the Vijus and Atahualpa formations. The other two groups are both Mississippian in age and comprise the calc-alkaline Pataz batholith and the slightly younger, K-rich volcanic and subvolcanic rocks (the Lavasen Volcanics and Esperanza subvolcanic complex, respectively). We discuss the implications of these data for the tectonic evolution of the Pataz district and for the origin of the batholith-hosted gold mineralisation.

REGIONAL SETTING
The Pataz-Parcoy gold-mining area is located in a transtensional jog on the regional Cordillera Blanca Fault (Petford & Atherton 1992;Witt et al. 2013a), which separates the Eastern Andean Cordillera from the Western Andean Cordillera (Megard 1984). The transtensional jog coincides with a north-striking section of the mainly NNW-striking Cordillera Blanca Fault, and the Pataz district is dominated structurally by NW-to NNW-striking faults and NE-to ENE-striking lineaments ( Figure 2). The main gold-mining areas at Pataz are located in Mississippian intrusions and Neoproterozoic basement, at the western margin of the Eastern Andean Cordillera. Targeting methods for gold in these areas are described in a companion paper (Witt et al. 2013b).
Although the Western Andean Cordillera is dominated by the effects of the Cenozoic Andean orogeny, earlier orogenic activity was focused along the Eastern Andean Cordillera, which was the leading continental margin of western Amazonia for at least 900 Ma (Miskovic & Schaltegger 2009). Intrusive rocks of the Eastern Andean Cordillera are predominantly the products of major pulses of magmatic activity in the Carboniferous to early Permian (340-285 Ma continental arc magmas) and again in the middle Permian to Early Jurassic  Ma extension-related granites), with volumetrically subordinate plutons related to modern Andean subduction ).

GEOLOGY OF THE PATAZ DISTRICT WITH EMPHASIS ON THE IGNEOUS ROCKS
The geology and structural evolution of the Pataz district was described by Witt et al. (2013a), and the principle igneous units of interest are summarised in Table 2. The Mississippian Pataz batholith and its early Paleozoic country rocks are exposed in a central NNWtrending horst (Figure 2). The Pataz batholith was emplaced at ca 338-336 Ma (Table 1; Schaltegger et al. 2006), during the Gondwanide orogeny. Emplacement of the batholith was facilitated by intermittent extension in the Pataz-Parcoy transtensional jog during NE-SW Gondwanide shortening, which resulted in dextral strike-slip movement on the Cordillera Blanca Fault. Differential uplift occurred shortly after emplacement of the batholith. At this time, the batholith-hosted auriferous veins formed, and the Lavasen Volcanics were deposited in a graben (the Lavasen graben) to the east of the central horst (Witt et al. 2013a; Figure 2). An allochthonous sedimentary sequence, deposited in the Western Andean Cordillera, was thrust over the Eastern Andean Cordillera during the Upper Cretaceous Peruvian phase of the Andean orogeny (Megard 1984). In the Pataz  district, this allochthonous sequence has mostly been removed by erosion, but is locally preserved in the Chagual graben, to the west of the central horst. The Chagual graben formed during a Cenozoic, extensional phase of the Andean orogeny. The country rocks of the Pataz batholith belong to the Eastern Andean Cordillera and comprise late Neoproterozoic to early Cambrian basement (the Maranon Complex) and overlying Cambrian-Ordovician volcanosedimentary units (from oldest to youngest: the Vijus Formation, the Contaya Formation and the Atahualpa Formation; Figure 2 and Witt et al. 2013a). Whereas the Contaya Formation is predominantly sedimentary, the Vijus and Atahualpa formations are felsic to intermediate volcaniclastic rocks that contain millimetre-scale quartz-and biotite-rich aggregates, possibly devitrified vitric clasts, and crystal fragments of plagioclase, quartz and biotite. Typical outcrops of the Vijus Formation preserve igneous textures but are strongly fractured and contain variable amounts of hydrothermal sericite, chlorite, clay minerals and iron oxides. The Atahualpa Formation has a stronger ductile deformation fabric but lithic or devitrified vitric clasts and mineral clasts are easily recognised within the foliated rock. The Atahualpa Formation was deposited shortly after 466 Ma (Table 1).
Field and petrographic features of the Pataz batholith are described by Schreiber et al. (1990), Macfarlane et al. (1999, Haeberlin et al. (2004) and Witt et al. (2013a). The main component of the batholith is massive, medium-to coarse-grained, equigranular, biotite (AEhornblende) granodiorite and less common tonalite. Sharp contacts between granodiorite and tonalite have not been observed. There is minor textural variation within the granodiorite but internal contacts are rare, and gradational. The granodiorite locally contains hornfelsed xenoliths of the Contaya Formation. The other major component of the batholith is dioritic, ranging from medium-to coarse-grained diorite and quartz diorite to diorite porphyry and microdiorite. Hornblende is typically dominant over biotite in these rocks, and relict cores of clinopyroxene have been observed locally. The dioritic component has been emplaced in and adjacent to NE-ENE lineaments that cut the batholith (Figure 2). Medium-to coarse-grained biotite monzogranite and dykes of aplite and pegmatite are minor components of the batholith. Sharp xenolithic contacts generally indicate that the dioritic component has been intruded by the dominant granodiorite component, but magmamingling structures have also been observed, indicating that the two magmas were partly contemporaneous (Witt et al. 2013a).
The newly recognised Vista Florida Formation (Witt et al. 2013a) comprises mafic to intermediate volcaniclastic sandstone, siltstone and breccia, and minor lava flows. This unit outcrops to the west of the Pataz batholith, between the Cedro and Suyubamba lineaments ( Figure 2) and is separated from the batholith by the Vijus fault. The western contact is a back-thrust, which has emplaced the Vista Florida Formation over the younger allochthonous sedimentary rocks (Witt et al. 2013a). The volcaniclastic rocks are host to a minor component of microdiorite and diorite porphyry, which are petrographically indistinguishable from equivalent rocks that form parts of the batholith.
The Lavasen Volcanics comprise plagioclase-and K-feldspar-phyric latitic to rhyolitic pyroclastic rocks, resedimented pyroclastic rocks and volcanogenic sedimentary rocks. Over 3000 m of volcaniclastic deposits accumulated in the Lavasen graben. Geological mapping in the Misquichilca area ( Figure 2) has shown the volcanic sequence in that area to be gently folded and subdivided by NNW-striking faults into a series of second order grabens and horsts (Witt et al. 2013a  contrary to previous estimates of Cenozoic deposition (Reyes & Wilson 1964;Wilson & Reyes 1997). The Esperanza subvolcanic complex was emplaced along the western margin of the Lavasen graben and has a faulted contact with the Pataz batholith ( Figure 2). It is a high-level, bimodal complex, comprising latite porphyry and quartz latite porphyry, and a subordinate component of dolerite (Witt et al. 2013a). Latitic rocks contain up to 15% but mostly <5% phenocrysts of plagioclase and K-feldspar. Quartz phenocrysts in the quartz latite porphyry are partially resorbed. Flow banding and spherulitic textures are well developed in the latite and quartz latite porphyries, and the complex is interpreted as a possible flow dome complex. Weak potassic and argillic hydrothermal alteration are widespread. Cognate clots of coarse-grained plagioclase with pyroxene, Fe-rich dark green amphibole and green (Fe 3þ -rich) biotite suggest derivation of the latitic rocks from a subjacent magma chamber. Dolerite dykes intrude the latitic porphyries. Dark green amphibole and pleochroic relict clinopyroxene in the dolerite indicate Fe-rich mineral compositions, confirmed by EDS analysis of amphiboles using a SEM at the Centre for Microscopy, Characterisation and Analysis, University of Western Australia (Fe þ Mn þ Ti/Fe þ Ti þ Mn þ Mg ¼ 0.5 to 0.7; Witt et al. 2013a). Vari-textured dolerite contains discontinuous, internalised veinlets of plagioclase-magnetite interpreted as late-stage mesostasis. The increasingly calcic composition of plagioclase from early formed phenocrysts (albite) through groundmass (andesine-labradorite) to late-stage plagioclase (bytownite)-magnetite veinlets (Witt et al. 2013a) suggests enrichment of a volatile-rich phase during crystallisation of the dolerite magma (cf. Panjasawawong et al. 1995).

WHOLE-ROCK GEOCHEMISTRY OF IGNEOUS ROCKS IN THE PATAZ DISTRICT Introduction
Thirty-seven samples were submitted for whole-rock geochemical analysis in two batches. Although most samples came from the Pataz batholith, one sedimentary rock sample from the Contaya Formation was analysed to assess the role of local country rock assimilation on the chemistry of the batholith magmas. A full list of samples with rock type and location, and analytical methods, accuracy and precision, are described in the Supplementary Papers.
Every attempt was made to collect samples with minimum hydrothermal alteration. However, the effects of widespread but mostly low-intensity silicification and argillic alteration of Lavasen Volcanic samples are clearly reflected in the whole-rock geochemical data where analyses for SiO 2 extend beyond the maximum expected for unaltered acid igneous rocks ($76 wt%;Cox et al. 1980) to as much as 82.7 wt% SiO 2 (Figure 3). Some low-SiO 2 batholith surface outcrop samples display weak to moderate chloritisation and calcite veins (districtwide propylitic alteration), and this is reflected in scattered Na 2 O ( Figure 3a) and Sr values. Schreiber et al.
(1990) described intense and extensive alteration of plagioclase to sericite in most samples from the Pataz batholith, and attributed the sericitisation to deuteric alteration by K-rich residual magmatic fluids. The petrographic observation is confirmed by this study. However, limited dispersion on a K 2 O vs SiO 2 plot (Figure 3b) suggests that sericitic alteration did not have a major impact on the whole-rock chemistry of most granitoids. Sodium, Sr and, possibly, Ba appear to be the elements most affected by hydrothermal alteration, whereas most other major and trace elements are considered to reflect magmatic compositions. Three samples each of the Vijus and Atahualpa formations form relatively tight clusters on most plots. Their whole-rock chemistry does not appear to have been significantly affected by hydrothermal alteration with the possible exception of alkalis (Na 2 O, K 2 O), particularly in the Vijus Formation ( Figure 3).
Because whole-rock compositions of the Lavasen samples have probably been modified by hydrothermal alteration, relatively immobile elements (particularly Ti, Th, Hf, Zr, Nb, V, Ga) are used to support interpretations based on major elements. The geochemistry of the main igneous groups (Vijus and Atahualpa formations, Pataz batholith and the Vista Florida Group, Lavasen Volcanics and Esperanza subvolcanic complex) is described in turn, from oldest to youngest.

Geochemistry of the Vijus and Atahualpa formations
Samples of the Vijus and Atahualpa formations occupy an SiO 2 range between 65.8 wt% and 70.5 wt% (Supplementary Papers Table A2). They are subalkaline and plot in the calc-alkaline field of an AFM diagram (Figure 4a).
The A/NK vs A/CNK plot (after Shand 1943) shows the Atahualpa samples to be mildly peraluminous whereas the Vijus samples are strongly peraluminous (Figure 4b), possibly reflecting hydrothermal leaching of Ca and alkalis and the formation of secondary sericite, chlorite and clay minerals during metamorphism or hydrothermal alteration. Both groups belong to the high-K calc-alkaline series, according to the K 2 O vs SiO 2 plot (Figure 4c, after Pecerillo & Taylor 1979), although the Vijus Formation data are scattered, again perhaps reflecting metamorphism or hydrothermal alteration. As shown in Figure 5a, both groups plot within the rhyodacite/dacite field on an immobile element-based plot developed by Winchester & Floyd (1977), but the Atahualpa Formation plots near the andesite boundary whereas the Vijus Formation plots closer to the rhyolite field.
Discriminative plots developed by Frost et al. (2001), classify the Vijus and Atahualpa formations as ferroan ( Figure 6a). Most Atahualpa samples plot near the boundary between the calc-alkalic and alkali-calcic associations but the Vijus samples are highly dispersed, again possibly reflecting mobility of alkalis during metamorphism or hydrothermal alteration (Figure 6b). The Vijus and Atahualpa samples are tightly clustered on plots incorporating the immobile elements V, Th and Zr, which effectively distinguish the two groups ( Figure 5c).
The Vijus and Atahualpa formations display mildly concave upward REE curves ( Figure 7a) with steep light  Figure 4b is the maximum limit for I-type granites according to Chappell & White (1974). ESC, Esperanza subvolcanic complex. rare earth elements (LREE), a moderate negative Eu anomaly and relatively flat heavy rare earth elements (HREE). The (La/Sm) N ratios range between 2.56 and 3.66 and (La/Yb) N ratios between 5.25 and 7.61. Compared with the Atahualpa Formation, the Vijus samples have slightly deeper Eu anomalies and lower HREE but the LREE abundances are almost identical. The two groups of samples display very similar NMORB-normalised multi-element spidergrams ( Figure 8). The spidergrams show enrichment of incompatible elements (left side of

Geochemistry of the Pataz batholith and Vista Florida Group
Collectively, the Pataz batholith and Vista Florida Group samples define an SiO 2 range between 45.0 wt% and 76.4 wt% (Supplementary Papers Table A2). The data plot in the calc-alkaline field on an AFM diagram, although low-SiO 2 samples straddle the calc-alkaline/ tholeiite boundary (Figure 4a). Schreiber et al. (1990) classified granites of the Pataz batholith as I-type, consistent with the widespread presence of hornblende in diorite and tonalite, and, less consistently, in granodiorite. The I-type classification is also consistent with an 1974,1992) for the bulk of the analyses ( Figure 4b). Most of the data straddle the moderate to high-K calc-alkaline fields in an SiO 2 vs K 2 O plot ( Figure 4c). On plots developed by Frost et al. (2001), low-SiO 2 samples, including the Vista Florida Group, are magnesian, whereas high-SiO 2 samples straddle the magnesian/ferroan boundary ( Figure 6a); low-SiO 2 samples are scattered, but the high-SiO 2 samples straddle the boundary between calcic and calc-alkalic associations ( Figure 6b). The classification illustrated in Figure 6 shows that the Pataz batholith trends are similar to those of calc-alkaline Cordilleran batholiths such as the Tuolomne Suite and the northern Idaho batholith. These are subduction-related magmas typical of continental margin arcs (Bateman & Chappell 1979;Hyndman 1984).
Harker variation diagrams effectively subdivide the majority of batholith samples into two suites: a low-silica suite and a high-silica suite (Figures 9, 10), separated by a 'silica gap' between approximately 55 and 60 wt% SiO 2 . The high-silica suite comprises granodiorite and monzogranite, as well as three tonalite samples. The low-silica suite comprises samples of diorite, quartz diorite and diorite porphyry, as well as one tonalite sample. WWG2, a member of the low-silica suite, is a cumulate-textured metapyroxenite and therefore plots away from the main group of low-silica samples on several of the plots in Figures 9 and 10 (see particularly MgO). The two mafic volcaniclastic samples from the Vista Florida Group consistently plot with diorite and microdiorite of the low-silica batholith suite, supporting field evidence for a co-magmatic relationship between the two (Witt et al. 2013a). An andesite dyke that intrudes granodiorite of the batholith near the Glorita mine entrance ( Figure 2) lies at the high-SiO 2 end of the low-SiO 2 suite and contains relatively high La and P 2 O 5 , and may belong to a later intrusive event but is here discussed with the low-SiO 2 suite.
Both suites display decreasing MgO, CaO, Fe 2 O 3 (total), Co and V, and increasing K 2 O, Rb and K 2 O/Na 2 O ratio with increasing SiO 2 , although with some scatter particularly among the low-SiO 2 suite (e.g. Figure 9b).
The high-SiO 2 suite also displays decreasing trends of P 2 O 5 , TiO 2 and Sc with increasing SiO 2 , but these elements produce considerable scatter in the low-SiO 2 suite (e.g. Figure 9g). The elements Ba, Th, Zr and La display contrasting trends between suites, producing an overall increase with SiO 2 in the low-SiO 2 suite and a weak to pronounced decrease with increasing SiO 2 in the high-SiO 2 suite (Figure 10c, e, g, h). Copper produces scatter among the low-SiO 2 suite samples and a decreasing trend in the high-SiO 2 suite ( Figure 10a).
REE patterns for the low-silica suite are variably enriched in LREE, with the exception of two samples (WWG24 and WWG28), which are distinctly depleted in LREE ( Figure 7c). These two samples are among the most primitive of the low-SiO 2 suite, as indicated by low SiO 2 values and high Mg# ( ¼ MgO/MgO þ Total Fe as Fe 2 O 3 ). All low-SiO 2 suite samples lack a Eu anomaly. The (La/Yb) N ratios range between 0.54 and 7.78 and are broadly correlated with SiO 2 . REE patterns for the highsilica suite are similar to those of low-silica suite samples but contain higher total REE, lack examples equivalent to the flatter curves generated by the low-SiO 2 suite, and display a weak negative Eu anomaly ( Figure 7b). The (La/Yb) N ratios range between 2.48 and 13.54 with a . For comparison, the transparent grey envelope encloses four basalts that erupted through thick continental crust in the central Chilean arc (Pearce 1983). ESC, Esperanza subvolcanic complex. Gaps marked bld are values below the analysis detection limit.
poor correlation between (La/Yb) N and SiO 2 . Both suites show pronounced positive Pb anomalies on multielement spidergrams normalised to an NMORB composition, but these are higher in the low-SiO 2 suite and highest in WWG24 and WWG28 (Figure 8b, c). Most lowsilica suite samples display strong enrichment of the incompatible elements (e.g. Cs, Rb, Ba, Th, U, K), and weak negative P (AESr) anomalies, but any Ti anomaly is minor or absent (Nb was not analysed). The most primitive samples of the low SiO 2 suite (e.g. WWG24 and WWG28) are less enriched in La, Ce, K, U (below detection) and Th than other samples, but are similarly enriched in Ba, Rb and Cs. The high-silica suite shows a more pronounced negative P (AESr) anomaly and similar

Geochemistry of the Esperanza-Lavasen suite
Felsic rocks of the Lavasen Volcanics and the Esperanza subvolcanic complex span the range 68.1 to 82.7 wt% SiO 2 (Supplementary Papers Table A2) with the highest values clearly affected by hydrothermal alteration. Samples plot in the alkali-rich part of the calc-alkaline field on an AFM diagram (Figure 4a), and are mildly to distinctly peraluminous according to the A/CNK ratio ( Figure 4b). The more peraluminous compositions of some Lavasen samples may have been influenced by hydrothermal alteration, in which case most samples have A/CNK ratios 1.1. Apart from one K-depleted sample of Lavasen Volcanics (silicified sample 206773 with 82.7 wt% SiO 2 ), all samples plot in the high-K calcalkaline series field on an SiO 2 vs K 2 O plot ( Figure 4c). The samples are subalkaline and plot in the rhyolite and rhyodacite/dacite fields on a Zr/TiO 2 vs Nb/Y diagram ( Figure 5a). Two anomalous samples of latite porphyry (samples 195022 and 195026) that plot within the rhyodacite/dacite field are less silicic than the bulk of the Esperanza-Lavasen suite, and these samples plot away from the rest of the samples in most other geochemical plots. On plots developed by Frost et al. (2001), the Lavasen Volcanics and the Esperanza subvolcanic complex samples are distinctly ferroan, and most samples plot within the high alkali part of the calc-alkalic association ( Figure 6).
Apart from latite porphyry samples 195022 and 195026, the Esperanza and Lavasen samples are tightly clustered on plots incorporating V, Th and Zr, which effectively distinguish the two groups, and also distinguish the Lavasen and Esperanza samples from the Atahualpa and Vijus formations samples (Figure 5c). The Lavasen samples have a slightly higher Th/Zr ratio than the Esperanza samples, even though these two groups overlap on most other plots (e.g. Figures 4, 5a,b). Similar triangular plots to Figure 5c show that the Lavasen Volcanics also have a slightly higher Hf/Zr ratio than the Esperanza latitic rocks.
Potential mobility of silica (SiO 2 ) requires the use of an alternative parameter to monitor the degree of fractionation within the Esperanza and, more especially, Lavasen sample groups. The parameter chosen for this purpose is mg# (moles MgO/moles MgO þ moles Fe 2 O 3 ; Fe 2 O 3 is an expression of total Fe). Using mg# is preferable to using SiO 2 to monitor fractionation but introduces its own problems. In particular, sample 206773 contains very low Fe 2 O 3 and consequently has a high mg# that is difficult to reconcile with the felsic appearance of the sample in the field and thin-section. This sample also shows evidence of silicification (82.7 wt% SiO 2 ) and may have lost Fe as a result of hydrothermal alteration. The immobile trace-element character of this sample suggests it is an altered rhyolite similar to other felsic Lavasen Volcanic samples (Figure 5a).
Assuming sample 206773 is a rhyolite with a prealteration mg# similar to other rhyolites, Lavasen Group felsic samples have a more restricted range of mg# compared with the Esperanza subvolcanic complex, but generally overlap with the more felsic members of the latter (Figures 11, 12). The extended compositional range of the Esperanza samples is caused by the two relatively mafic latite samples (195022 and 195026). These two samples are characterised by relatively high mg#, as well as elevated Al 2 O 3 , CaO, P 2 O 5 , MgO, Fe 2 O 3, TiO 2 , V and Co, but also Zr and Hf (Figure 12c, d, f). Interestingly, these two samples do not contain higher Ni and Cr than the rest of the Esperanza-Lavasen suite, although the absolute abundances are near the threshold of detection. Within the main group of samples (at mg# <20), Al 2 O 3 decreases, and CaO, MgO, TiO 2 and P 2 O 5 remain more or less constant, with decreasing mg#. There is little variation in the abundance of most trace elements with mg# ( Figure 12). REE patterns show that latitic rocks of the Esperanza subvolcanic complex are tightly grouped and characterised by flat HREE (Ho to Lu) and steep MREE to LREE, interrupted by a prominent negative Eu anomaly (Figure 7d). Lavasen Volcanic samples display similar REE patterns, except that the range of REE concentrations is displaced to slightly lower values, and the Two samples of dolerite from the Esperanza subvolcanic complex plot in the tholeiite field on an AFM diagram ( Figure 4a). Both samples plot as basalt on the Zr/TiO 2 vs Nb/Y diagram, but one sample plots in the subalkaline field and the other just within the alkaline basalt field (Figure 5a). The two samples display distinctive REE curves, much less fractionated than the associated latitic rocks and lacking Eu anomalies (Figure 7f). Spidergrams normalised to an NMORB composition are also relatively flat. Although most elements are 10Â those of typical NMORB compositions, Rb and Cs are

Geochemistry of the Estrella monzonite porphyry
The monzonite porphyry sample from Estrella (near Santa Filomena in Figure 2) is clearly differentiated from the high-silica batholith suite and the Esperanza-Lavasen suite by virtue of its high K 2 O (Figure 4c), Ba, Rb, Th, high total alkalis ( Figure 6b) and high K 2 O/Na 2 O ratio ( Figure 3c). This is the only sample that clearly plots in the shoshonitic field on Figure 4c. As for the Esperanza-Lavasen suite, relatively high K, Ba and Rb in this intrusion is consistent with petrographic evidence for early crystallisation of K-feldspar. The REE curve for the Estrella monzonite porphyry (WWG11) is very similar to those of Esperanza subvolcanic complex latites but at the upper end of the range (Figure 7d). The spidergram is also similar in form to those of the Esperanza subvolcanic complex but is relatively enriched in Ba and Rb, and depleted in Pb, Zr, Y and the REE (Figure 8d).

Petrogenesis of the Vijus and Atahualpa formations
The LREE-enriched REE curves and enrichment of incompatible elements on chondrite-and NMORBnormalised spidergrams, respectively, are features typical of moderate-to high-K calc-alkaline magmas from continental arc environments (Pearce 1983;Wilson 1989). The negative Nb and Ti anomalies (Pearce 1983(Pearce , 1996Jenner 1996) and positive Pb anomalies (Elliott et al. 1997) are also typical features of island and continental arc rocks. These types of magmas are generally considered be generated in the metasomatised mantle wedge, above subduction zones (Pearce 1983;Wilson 1989) with the negative Nb anomaly reflecting a previous melt extraction event in the source region and addition of low volume melts enriched in large ion lithophile elements (LILE) and LREE (McCulloch & Gamble 1991). Although the overall form of the spidergrams is similar to that of calc-alkaline magmas, the degree of enrichment of all elements, especially the alkalis and LREE is high compared with typical continental arc basalts (Figure 8a), and more comparable to those erupted on thick continental lithosphere (Pearce 1983) or the 'mature continental arcs' of Brown et al. (1984). The relatively felsic bulk composition of the volcanic rocks, particularly the Vijus Formation (Figure 9a), and the prominent negative Eu and Sr anomalies (Figures 7a, 8a), and low Cr (<50 ppm) and Ni (<20 ppm) suggest a role for plagioclase fractionation and indicate that these are not pristine mantle melts. The Vijus and Atahualpa magmas could have been generated in metasomatised asthenosphere or subcontinental lithosphere and attained their more felsic compositions through fractional crystallisation and/or crustal assimilation. Alternatively, they may have formed more directly by partial melting of the lower continental crust, providing the source had an appropriate (andesitic) composition inherited from an earlier period of subduction (e.g. Chappell & Stephens 1988). Additional wholerock isotopic data are required to allow a distinction between these processes. The presence of Mesoproterozoic zircon xenocrysts in the Atahualpa Formation (Witt et al. 2013a) does not provide critical evidence for either argument because these could be interpreted as restite carried from the lower crustal source or refractory components of a crustal contaminant. Miskovic & Schaltegger (2009) describe the Vijus Formation as andesitic, which suggests that the felsic rocks in the Pataz district may be a minor component of the regional volcanic sequence. Therefore, we favour an origin for the Vijus and Atahualpa formations as fractionation products of subduction-related arc magmas with a metasomatised mantle or subcontinental lithosphere source. A significant component of fractional crystallisation as indicated by elevated Ba (500-2000 ppm), Th (13-16 ppm) and Zr (280-340 ppm) ( Figure 12) and depleted Sr and P ( Figure 8a) may also be attributed to fractional crystallisation.
Atahualpa Formation volcaniclastic rocks contain higher Fe 2 O 3 , MgO, TiO 2 , P 2 O 5 , Zr, Th, V, Ni, Co and Cr, but lower Al 2 O 3 and Ba, than Vijus Formation samples (Figures 11, 12). They are probably not related to Vijus Formation volcanics by fractional crystallisation, as they tend to plot in separate clusters on variation diagrams and in ternary diagrams such as Figure 5c, and are separated stratigraphically by the metasedimentary Contaya Formation (Haeberlin et al. 2004;Witt et al. 2013a). Compositional differences may reflect different source compositions and/or degrees of crustal contamination for the Vijus Formation.

Petrogenesis of the low-SiO 2 suite of the Pataz batholith and Vista Florida Group
The depleted LREE patterns of least evolved samples WWG24 and WWG28 from the low-SiO 2 batholith suite resemble those for tholeiitic NMORB basalts, and these two samples plot in the tholeiitic field of an AFM diagram ( Figure 4a). However, the NMORB-normalised spidergrams for these samples display pronounced enrichment of Cs, Rb and Ba, and more modest enrichment of K and Th, as well as a pronounced positive Pb anomaly. This suite of enriched elements is generally believed to be derived from the subducted sedimentary layer in oceanic subduction zone settings (Pearce 1982(Pearce , 1983Elliott et al. 1997). In contrast to most continental subduction zone magmas (Pearce 1982(Pearce , 1983Brown et al. 1984), LREE, MREE and Zr are not enriched in these primitive members of the low-SiO 2 batholith suite. Furthermore, although the absence of an Nb anomaly cannot be explicitly demonstrated (because Nb analyses are not available), the absence of a Ti anomaly contrasts with typical continental arc magmas (Pearce 1982(Pearce , 1983Brown et al. 1984). The absence of negative Ti anomalies on NMORB-normalised spidergrams suggests that the parent magma of the low-SiO 2 suite may not have arisen from a typical continental arc source region (i.e. the metasomatised asthenospheric wedge, Wilson 1989;or subcontinental lithosphere, Pearce 1983). According to the model of McCulloch & Gamble (1991), this could mean that the mantle source for the low-SiO 2 suite had not been depleted by a previous melting event. Nevertheless, even the most primitive of low-SiO 2 samples (WWG24, WWG28) display enrichment in alkalis and Th, and a prominent positive Pb anomaly, suggesting that the mantle source had been metasomatised, probably by low-volume melts derived from the slab or deeper parts of the asthenospheric mantle. The Cr and Ni concentrations in the low-SiO 2 suite samples are well below those predicted for primary mantle-derived melts (900 ppm and 350 ppm, respectively; Cox et al. 1980) indicating that the parent magma had evolved, probably by fractional crystallisation and/or crustal contamination, since initial formation in the mantle source region.
Other samples of the low-SiO 2 batholith suite display pronounced enrichment of LREE (La and Ce are 10Â to 100Â chondrite values), Th and U, and modest enrichment of Nd, Sm and Zr, and a positive Pb anomaly ( Figure 8c). Compared with WWG24 and WWG28, these samples are further enriched in K, Th, U, LREE, P, Nd, Sm and Zr but are depleted in HREE and Y. The resulting spidergrams for evolved low-SiO 2 samples are similar to those of the Vijus and Atahualpa formations and other high-K calc-alkaline magmas from continental arcs on thick lithosphere (Brown et al. 1984;Wilson 1989;Price et al. 1999). Compositional variation within the low-SiO 2 suite could be explained by contamination or assimilation of magma similar to that which produced WWG24 and WWG28, by continental crust containing elevated alkalis, LREE, Th, U and P, but similar Pb, lower HREE and Y than the parent low-SiO 2 magma. Mesoproteozoic rocks in the central segment of the Eastern Andean Cordillera reported by Miskovic &Schaltegger (2009) andMiskovic et al. (2009) have the appropriate composition to cause this contamination.
The negative correlation of MgO, CaO and Fe 2 O 3 and V (total) with increasing silica is consistent with assimilation of crustal material with higher SiO 2 than the parent low-SiO 2 magma, or fractional crystallisation. However, trace elements such as Sr and Ti do not demonstrate a regular negative correlation against alkali elements, as would be expected of fractional crystallisation. Fractional crystallisation of clinopyroxene and amphibole would be expected to produce steep enrichment trends for Zr, similar to that defined by other incompatible elements such as Th and Rb (Figure 10). The weak enrichment trends for Zr are consistent with assimilation of material with much higher Rb and Th but only slightly higher Zr, as is the case for the Mesoproterozoic crust beneath the Eastern Andean Cordillera (Miskovic & Schaltegger 2009;Miskovic et al. 2009). Therefore, fractional crystallisation is assigned a relatively minor role for compositional variation within the low-SiO 2 suite, most of which is attributed to variable assimilation of Mesozoic continental crust of heterogeneous but overall intermediate to felsic composition.

Petrogenesis of the high-SiO 2 Pataz batholith suite
The high-SiO 2 suite of the Pataz batholith, dominated by granodiorite (with subordinate monzogranite and tonalite), has the geochemical characteristics of calc-alkaline I-type granites, as described by previous investigators (Schreiber et al. 1990;Macfarlane et al. 1999). The REE curves and NMORB-normalised spidergrams for high-SiO 2 batholith suite samples, incorporating high LREE/ HREE ratios and enrichment of incompatible elements, are typical of mature continental arc magmas (Brown et al. 1984). The positive Pb anomaly seen in the Vijus and Atahualpa formations and the low-SiO 2 batholith suite is also present, although weaker, in the high-SiO 2 samples, caused mainly by relative enrichment of LREE and Sr (Figure 8b). A negative Ti anomaly on the NMORBnormalised spidergram (Figure 8b) is also a feature of calc-alkaline magmas in arc environments. Although Nb analyses are not available for the high-SiO 2 suite, Miskovic et al. (2009) also noted the presence of negative Nb anomalies in their regional study of Permian-Carboniferous granitoids of the Eastern Andean Cordillera.
As for other calc-alkaline, I-type granodioritedominated suites, the Pataz high-SiO 2 suite melt is interpreted to have formed by partial melting of an older Cordilleran-type andesite or diorite source in the lower crust (cf. Chappell & Stephens 1988;Chappell & White 1992), consistent with the radiogenic Pb and Nd isotopes reported by Macfarlane et al. (1999) and, more generally, with the Lu-Hf isotope data of Permian-Carboniferous granitoids in the Eastern Andean Cordillera ). The partly contemporaneous, mantle-derived low-SiO 2 magma probably provided the heat source required to melt the lower crustal source rocks.
Lead and Nd isotopic data (Macfarlane et al. 1999), and Lu-Hf isotope data (Miskovic & Schaltegger 2009) indicate that felsic crustal material played an important role in the chemistry of the Pataz batholith and associated dykes. This conclusion is consistent with the model presented here, which derives the Pataz batholith granodiorite from melting of older, arc-related andesitic or dioritic source rocks in the lower crust. Possible candidates for this andesitic crust are the Mesoproterozoic San Ignacio-Sunsas orogen, and equivalents of the Vijus and Atahualpa formation magmas that may have accumulated in the lower crust.
A distinct parent magma composition for the high-SiO 2 suite could not be recognised. Sample WWG7 (tonalite porphyry from Karola) has the lowest SiO 2 (61.9 wt%) and highest mg# (0.64) of all high-SiO 2 suite samples. Samples WWG26 and WWG31 have slightly higher SiO 2 (62.3 wt% and 64.8 wt%, respectively) and lower mg# (0.53 and 0.49, respectively). These three tonalite samples plot away from the main trend defined by high-SiO 2 suite samples, which otherwise define a depletion trend in Zr and LREE with increasing fractionation (Figure 10g, h).
Magma mingling structures Witt et al. 2013a) warrant assessment of magma mixing between low-and high-silica suite magmas. Binary plots involving trace-element ratios should produce curved trends where compositional variation is caused by magma mixing (Cox et al. 1980). The Pataz batholith data plot along straight-line trends (Figure 13a, b) and are not consistent with magma mixing as a significant process leading to compositional variation. Magma mixing between low-and high-SiO 2 magmas is also inconsistent with the behaviour of Al 2 O 3 , P 2 O 5 , Ba, Zr, La and Ce (Figures 9, 10). It can therefore be concluded that any mixing or assimilation between mantle-derived and crustal magmas took place in the source region and helped define the composition of the parental high-SiO 2 magma.
Harker variation diagrams suggest a role for fractionation within the high-silica suite (Figures 9, 10).
Depletion of MgO, TiO 2 , Fe 2 O 3 (total), CaO, Ni, Co, V and Sc with increasing SiO 2 , and negative Eu anomalies, in the high-silica suite are consistent with fractionation of amphibole, plagioclase and biotite. Depletion of Zr and LREE with increasing SiO 2 (Figure 10g, h) is attributed to fractionation of accessory phases such as zircon and apatite, the latter supported by negative P anomalies on Figure 13 Trace-element plots to distinguish genetic processes for the Pataz batholith. Plots of (La/Yb) CN vs (a) Ti/Zr and (b) Pb/Ce. Magma mixing between two end-members should generate a curved array of data on these plots; the approximate straight-line relationship, especially in Figure 17a, is not compatible with magma mixing between the low-and high-silica suites. Log-log plots of Sr vs Rb (c) and V vs Th (d). Inset in (c) is applicable to (c, d) and presents the generalised form of trends defined by fractional crystallisation and partial melting (after Cocherie 1986). For fractional crystallisation, i is the initial composition of the magma; A is the compositional trend of the magma caused by fractional crystallisation; B and C are the compositional trends of the magma caused by incremental equilibrium crystallisation where increments of magma are removed after 10% (B) and 20% (C) crystallisation. For partial melting, i is the initial bulk composition of the source rock; a is the compositional trend of the magma caused by batch partial melting; c is the compositional trend of the magma caused by fractional partial melting with continuous removal of the melt; b is an intermediate process between perfect batch melting and perfect fractional melting. Compositions of least fractionated (closest to starting composition) samples WWG24, WWG28 and WWG31 are shown; WWG24 and WWG28 contain <10 ppm Rb and are not plotted as individual points, but the area in which they would plot is shown as a grey circle. NMORB-normalised spidergrams (Figure 8b). The term fractionation, as applied here, could refer to fractional crystallisation of the magmas (during ascent or following emplacement), or variable partial melting within the source region. Log-log plots using compatible and incompatible trace elements can be used to assess the relative roles of these processes (Cocherie 1986). The relatively flat trends shown by the Pataz data on log-log plots involving Sr and V (compatible elements) and Rb and Th (incompatible elements) are not consistent with extended fractional crystallisation, whereby compatible elements are rapidly depleted, but could be explained by equilibrium crystallisation and/or variable batch partial melting (Figure 13c, d; Cocherie 1986). Rapid depletion of Zr and LREE may indicate a role for fractional crystallisation, but the data suggest that variable partial melting of continental crust dominated compositional variation within the high-SiO 2 suite. Some variation in the proportion of crustal and mantle-derived magmas in the source region may also have contributed to compositional variation among batches of parental high-SiO 2 magma. A fractionational crystallisation relationship between the low-silica suite and the high-silica suite is precluded by the silica gap between 55 and 60 wt% SiO 2 and the contrasting behaviours of Al 2 O 3 , P 2 O 5 , Ba, Zr, La and Ce (Figures 9, 10).
Data presented here for the Pataz batholith display weak and inconsistent convergence towards the sample of Contaya Formation on plots involving major and trace elements (Figures 9, 10). Although only one sample has been analysed, it is taken from the immediate country rock of the batholith and incorporates several inter-beds of sandstone and siltstone and is believed to be reasonably representative of the Contaya Formation, which is a fairly uniform sequence of sedimentary rocks in the Pataz district. The plots in Figures 9 and 10 suggest that assimilation of metasedimentary country rock has not had a major influence on the geochemistry of the Pataz batholith, despite the widespread presence of accidental xenoliths.

Petrogenesis of the Esperanza-Lavasen association and the Estrella monzonite porphyry
The Fe-rich Esperanza dolerite crystallised from a tholeiitic parent magma (Figure 4a). The relatively flat REE patterns (Figure 7f) are similar to those of basalts derived from slightly to moderately enriched EMORB mantle sources (Pearce 1983;Wilson 1989) but are approximately 10Â more enriched than average MORB (Figure 8f). The tendency towards alkaline compositions (Figure 5a) suggests that these mafic magmas were produced by relatively small volume partial melting of a metasomatised mantle source, possibly at greater depths than the source for low-SiO 2 suite batholith magmas (Jaques & Green 1980). The 10Â enrichment of Y and Yb, and most other elements, relative to average MORB (Figure 8f) also suggests small volume partial melting of the mantle source. Although REE curves and spidergrams are flat compared with other igneous rocks in the Pataz district, both dolerite samples display the positive Pb anomalies and enrichment of Cs and Rb suggesting the doleritic magma was produced by partial melting of a compositionally similar mantle source to that which yielded the least evolved low-SiO 2 batholith samples (i.e. a metasomatised mantle source). Also, the absence of negative Nb anomalies suggests this mantle source had not previously been depleted by partial melt extraction (McCulloch & Gamble 1991). The Mg#, Ni and Cr contents of the dolerites indicate that they are not primary mantle-derived melts, but the absence of negative Eu anomalies on REE curves precludes significant fractionation of plagioclase. Fractional crystallisation, which probably caused the iron-rich composition of the dolerites, was likely dominated by olivine and pyroxene.
Compared with the Pataz batholith, latitic to rhyolitic rocks of the Esperanza-Lavasen felsic suite are distinctly ferroan and alkali-rich ( Figure 6). These qualities identify the Esperanza-Lavasen felsic samples as having affinities with A-type or anorogenic felsic magmas (cf. Eby 1990;Frost et al. 2001). The A-type character of the samples is supported by plots developed by Dall'Agnol & de Oliveira (2007), which show that the Esperanza-Lavasen samples plot with moderately oxidised anorogenic granites (Figure 14a, b). The low CaO and Al 2 O 3 , and high K 2 O and Fe TOT /MgO, shown by the Esperanza-Lavasen felsic samples are characteristic of A-type magmas, as are the Fe-rich compositions of amphibole (inferred from the deep green colour) and relict pyroxene (Anderson 1983;Whalen et al. 1987;Eby 1990;Pitcher 1997;Dall'Agnol & de Oliveira 2007). The plots in Figures 6 and 14 are based on major elements, some including alkalis, which may have been susceptible to hydrothermal alteration. However, supporting arguments for an A-type classification are based on immobile trace element plots, such as those developed by Whalen et al. (1987). For example, on a FeO TOT vs 1000Ga/Al plot, all latitic samples (excluding the two most mafic Esperanza samples: 195022 and 195026) fall in the field of A-type granites (Figure 5b). Elevated high-field-strength element concentrations are a characteristic feature of A-type magmas (Eby 1990;Pitcher 1997;Dall'Agnol & de Oliveira 2007), and the Esperanza-Lavasen samples are generally enriched in Zr, Ta, Th and LREE compared with the Pataz batholith ( Figure 8; Supplementary Papers Table A2). The REE curves of the Esperanza-Lavasen felsic suite are also very similar to those of the A-type quartz latite-granitoid suite from the Roxby Downs area, South Australia (Creaser 1996), which hosts the Olympic Dam Cu-Au-U-REE deposit.
The large volume of Lavasen Volcanics (at least 2000 km 3 ) and the felsic nature of both the volcanic rocks and the Esperanza subvolcanic complex (including the very low Cr and Ni) imply a crustal source for the magmas, consistent with Y/Nb ratios >1.2 (Eby 1990). Models for the origin of A-type magmas include second melting of a residual source that had previously yielded a calcalkaline I-type melt (Collins et al. 1982) and low pressure melting of a calc-alkaline (tonalitic to granodioritic) source (Creaser et al. 1991;Roberts & Clemens 1993;Dall'Agnol & de Oliveira 2007), as well as fractionation of a tholeiitic magma (Turner et al. 1992;Frost & Frost 1997). Despite the close spatial and temporal association of the Esperanza-Lavasen suite with the I-type Pataz batholith and its residue in the lower crust, first-stage melting of a tonalitic to granodioritic source is preferred for reasons outlined by Creaser et al. (1991). In either case, high-temperature partial melting probably requires thermal input from a mantle-derived magma emplaced into the source region. The Fe-rich dolerite dykes post-date latitic rocks of the Esperanza subvolcanic complex, but their presence shows that mantlederived magma may have been present in the crustal source region to provide the required heat. Fractionation in this inferred tholeiitic magma chamber could have given rise to the A-type Esperanza and Lavasen magmas, but we have no evidence at the present level of erosion to suggest this occurred.
It was noted above that most Esperanza-Lavasen felsic samples plot within the A-type field in Figure 5b, but two samples (195022 and 195026) do not. These two samples have the lowest SiO 2 and the highest CaO, MgO, FeOt, MnO, TiO 2 , P 2 O 5 and MgO/Fe 2 O 3 ratio, indicating they are less evolved than the other samples. Most samples plot within a restricted range in mg# (0 to 20), but inclusion of samples 195022 and 195026 extends the range into the mid 20s. A crudely stratified magma chamber is envisaged for the mid-crustal source region of the Esperanza-Lavasen felsic suite, comprising a lower, mantlederived mafic melt and an upper, crustal melt of felsic Atype composition (cf. Barnes 1983;Sparks et al. 1984). The relatively mafic composition of samples 195022 and 195026 are interpreted to result from mixing of mantlederived and crustal melts across a transitional zone within this lower crustal magma chamber, prior to eruption. The Fe-rich dolerites are probably not representative of the lower, mafic magma in the hypothesised magma chamber because samples 195022 and 195026 do not lie on a mixing line between the Fe-rich dolerites and the latitic rocks (Figures 11, 12; see particularly Zr and Hf).
The remaining latitic samples of the Esperanza subvolcanic complex and those of the Lavasen Volcanics more likely represent unmodified melt derived from the upper part of the magma chamber, where crustally derived magma predominated. A more potassic, felsic (tonalite or granodiorite) source than that which produced the Pataz granodiorite is implicated (Chappell & Stephens 1988;Roberts & Clemens 1993). Similar to the source for the Pataz granodiorite, the positive Pb anomaly and some of the enrichment in Cs and Rb may have been inherited from this lower crustal source. This source rock must also have had negative Ti and Nb anomalies, inherited from the calc-alkaline source rocks in the mid crust. A suitable source for the Esperanza-Lavasen A-type magma is the compositionally variable Mesoproterozoic basement (e.g. the Satipo Tonalite) reported by Miskovic et al. (2009) andMiskovic &Schaltegger (2009), or possibly mid-crustal equivalents of the high-SiO 2 Pataz batholith suite.
The Estrella monzonite porphyry (near Santa Filomena in Figure 2) shows petrographic similarities to the Esperanza-Lavasen felsic suite, especially the presence of K-feldspar phenocrysts. The intrusion has the highest alkali content of all intrusions investigated in the Pataz district and is the only one to plot clearly within the shoshonite field in Figure 4c. A similar origin by partial melting of mid-crustal tonalite or granodiorite is proposed. However, the age of the Estrella monzonite porphyry is largely unconstrained following the failure of SHRIMP U-Pb dating of zircon to identify an unequivocal emplacement age (Witt et al. 2013a).

Tectonomagmatic evolution of the northern Pataz district
Previous geochronological and geochemical studies of granitoids in the Eastern Andean Cordillera of Peru have shown that this mainly Paleozoic belt represents a long-lived (up to 1.15 Ga) continental margin along the western edge of the Amazonia craton (Figure 1; Petersen 1999;Chew et al. 2007;Miskovic et al. 2009). Following the break-up of Rodinia in the Neoproterozoic (770-570 Ma), the history of the Eastern Andean Cordillera was characterised by almost continuous subduction, lasting into the late Triassic (Chew et al. 2007;Miskovic et al. 2009). The subduction zone setting of the Andean margin continues into the present, but the present locus of arc magmatism has shifted into the Western Andean Cordillera (Noble & McKee 1999;Petersen 1999). This study of igneous rocks in the northern Pataz district of the Eastern Andean Cordillera shows that intermediate to felsic magmas with arc-related continental margin geochemical signatures were emplaced into the Andean crust during the early Paleozoic (Vijus and Atahualpa formations) and Mississippian (Pataz batholith and Esperanza-Lavasen suite). The resulting volcanic and intrusive rocks consistently display alkali (Rb, Cs) enrichment, positive Pb anomalies, and most of the more felsic rocks exhibit a negative Nb anomaly or, where Nb is not available, a negative Ti anomaly (Figure 8). However, we contend that the more voluminous felsic Carboniferous magmas were generated by partial melting of the lower to mid-continental crust and inherited their continental arc signature from an earlier period of subduction, possibly that which produced the Vijus and Atahualpa formations during the Ordovician. This interpretation is consistent with the whole-rock chemistry discussed above and with the conclusions of Miskovic & Schaltegger (2009) who found a predominantly crustal Lu-Hf signature for Permian-Carboniferous granitoids in the Eastern Andean Cordillera of northern Peru. It is also consistent with the Pb and Nd isotope data of Macfarlane et al. (1999) who estimated a Mesoproterozoic crustal component of up to 70% for the Pataz batholith.
The origin of relatively primitive mafic intrusive rocks in the Pataz district (WWG24 and WWG28 of the low-SiO 2 batholith suite, and the Esperanza dolerites) is problematic. They display enrichment of alkalis Cs, Rb (AE Ba) and a prominent positive Pb anomaly, and Mg#, Cr and Ni values consistent with fractionated partial melts derived from a mantle source that was metasomatised by fluids derived from subducted oceanic crust and sediments. However, the dolerites at least lack the characteristic negative Nb anomaly of arc-related magmas, and the low-SiO 2 batholith suite lacks a Ti anomaly, both of which may indicate that the mantle source had not previously been subject to partial melt extraction (McCulloch & Gamble 1991). If the Vijus and Atahualpa formation magmas, which incorporated a negative Nb anomaly, were formed by partial melting of mantle that had previously yielded a partial melt, the source region would have had to be replaced by undepleted asthenosphere before it melted again to produce the low-SiO 2 batholith and Esperanza dolerite magmas. Mechanisms that might induce this history include periodic upwelling in incipient rifts and slab tears (Hale et al. 2009). Some of these mantle-derived magmas probably provided the thermal energy required to melt the continental crust and produce the Carboniferous felsic magmas.
The interpreted Ordovician to Mississippian tectonic evolution of the Andean margin in the vicinity of the northern Pataz district is summarised schematically in Figure 15. The volcanic rocks of the Atahualpa Formation are mid-Ordovician (SHRIMP U-Pb in zircon age ¼ 466.8 AE8.1 Ma; Witt et al. 2013a) and are stratigraphically younger than the rocks of the Vijus Formation, which is probably late Cambrian or Early Ordovician. The age of the Atahaulpa Formation is coincident with the age obtained from zircons in the Esterella monzonite porphyry (Witt et al. 2013a), which may represent a feeder to the Cambrian-Ordovician volcanic rocks in the northern Pataz district. Ordovician igneous rocks are not common in the Andean margin of northern Peru, but the Sitabamba granodiorite gneiss, about 50 km south of Pataz in the Eastern Andean Cordillera, has a slightly younger age of ca 448.  (2009), but this is inconsistent with our mid-Ordovician SHRIMP age for the overlying Atahualpa Formation. The mid-Ordovician age for the Atahualpa Formation suggests that Atahualpa Formation volcanism is related to renewal of subduction in a zone located west of the accreted Paracas-Arequipa Terrane (Figure 15a; cf. Miskovic et al. 2009). Paleoproterozoic and Mesoproterozoic xenocrysts in the Atahualpa Formation suggest some contamination of arc magmas by lower continental crust formed by the Amazonia craton and the San Ignacio-Sunsas orogen, respectively. These Ordovician igneous rocks may be the northward continuation of the early Paleozoic San Nicolas-Famatinia arc exposed in southwestern Peru and the Sierras Pampeanas in northwest Argentina (Bahlberg & Herve 1997;Miskovic et al. 2009).
During the mid-Mississippian (ca 338-336 Ma), the high-SiO 2 (mainly granodiorite) magma of the Pataz batholith formed as a result of partial melting of an andesitic or dioritic lower crustal source, and was emplaced in the mid-to upper crust (estimated a depth of 11 to 13 km; Haeberlin et al. 2004). A crustal source is indicated by the felsic bulk composition of the high-SiO 2 magma and geochemical data that point to variable partial melting rather than fractional crystallisation as the dominant genetic process. Positive Pb anomalies, enrichment of Cs and Rb, and implied negative Nb anomalies on NMORB-normalised spidergrams (Figure 8) are consistent with an arc-derived andesitic source in the lower continental crust. It is possible that the andesitic source was formed by the same magmas that formed the Vijus and Atahualpa formations but were underplated onto the base of the continental crust. Alternatively, the Mesoproterozoic San Ignacio-Sunsas orogen may again have provided the source. There are no xenocryst age data for the Pataz batholith known to the authors, but  Figure 2). Therefore, the San Ignacio-Sunsas orogenic belt may have been the lower crustal source for the high-SiO 2 batholith suite, or alternatively a contaminant of melts derived from a source equivalent to the Vijus and Atahualpa formations.
The heat required for melting of the lower crustal source was likely provided by emplacement of fractionated mantle-derived magmas similar in composition to those represented by the low-SiO 2 batholith suite ( Figure 15b). Mississippian magmatism commenced with the eruption of the low-SiO 2 batholith magma as the Vista Florida Group volcanic rocks, and emplacement of diorite, microdiorite and diorite porphyry into tensional fractures in the upper crust below the volcanic vents (Witt et al. 2013a). Enrichment of Cs and Rb, and positive Pb anomalies in the most primitive samples (WWG24 and WWG28) suggest the low-SiO 2 suite was derived by partial melting of metasomatised mantle, similar to that which yielded the parent magmas for the Vijus and Atahualpa formations, but with no negative Ti anomaly. This suggests that, whereas the mantle source for the Vijus and Atahualpa formation magmas was depleted by previous melt extraction, the source for the low-SiO 2 Pataz batholith suite was not, requiring regeneration of asthenospheric mantle to replace the depleted source with an undepleted source (Figure 15b). This could occur where upwelling of undepleted asthenosphere was initiated by incipient rifting, although any such rifts did not develop fully into back-arcs. Progressive incorporation of Mesoproterozoic lower crustal material to metasomatised but undepleted asthenospheric melt produced the more evolved samples of the low-SiO 2 suite and led to diminution of the Pb anomaly, but addition of Th, U, K, LREE, MREE and Zr. At some point, a change in the local stress field caused ponding of the low-SiO 2 magma in the lower crust, where partial melting occurred to produce the high-SiO 2 magma.
Within a few million years of the emplacement of the Pataz batholith, renewed crustal melting produced K-rich felsic magmas with A-type affinities, which were erupted as the Lavasen Volcanics and emplaced in a shallow crustal setting as latitic rocks of the Esperanza subvolcanic complex (Figure 15c). In comparison with the andesitic source for the high-SiO 2 batholith suite, the crustal source for the felsic Esperanza-Lavasen suite was relatively SiO 2 -and K-rich (cf. Roberts & Clemens 1993). The residual source is unlikely to be more SiO 2and K 2 O-rich following extraction of the Pataz batholith I-type granodiorite. The San Ignacio-Sunsas orogen embraces a sufficient compositional range to be an appropriate source for the felsic Esperanza-Lavasen suite, consistent with the presence of Mesoproterozoic zircon xenocrysts in the Esperanza subvolcanic complex (Witt et al. 2013a). Surface exposures of the compositionally diverse Mesoproterozoic orogen include peraluminous, calc-alkaline to alkali-calcic granitoids with negative Nb-Ta anomalies (Miskovic & Schaltegger 2009;Miskovic et al. 2009). However, Mesoproterozoic zircons are rare in the Esperanza-Lavasen felsic suite, as might be expected of alkali-rich magmas. Furthermore, unpublished Lu-Hf isotope data for zircon from a Lavasen sample require a much younger crustal source than the San Ignacio-Sunsas orogen (T. Kemp & T. Angerer, pers. comm., 2011). An alternative source for the Esperanza-Lavasen magmas could be provided by tonalite and granodiorite equivalent to those forming the high-SiO 2 Pataz batholith suite, if some of this magma ponded in the mid crust. This alternative model is consistent with alkali enrichment, positive Pb and negative Nb anomalies on spidergrams for the Esperanza-Lavasen felsic suite. Relatively deep Eu, Sr and P anomalies suggest residual plagioclase and apatite.
It is proposed that the heat required for crustal anatexis of the Esperanza-Lavasen felsic suite source was provided by emplacement of mantle-derived mafic magma. The true nature of this mantle-derived melt is not known, but the relatively mafic compositions of samples 195022 and 195026 provide some evidence for a more mafic component in the A-type magma chamber. The source of the magma was probably similar to that (enriched asthenosphere that had not previously melted), which yielded the low-SiO 2 suite of the batholith. The strong Cs and Rb anomaly, and positive Pb anomaly of the post-latitic Fe-rich dolerite provide some evidence that this asthenospheric source was melting at the time of A-type magmatism.
A-type granites are generally found in anorogenic settings in continental interiors, but also form at plate margins in the waning stages of subduction zone magmatism where they typically occur on the landward side of subduction-related magmas (Anderson 1983;Eby 1990;Dall'Agnol & de Oliveira 2007) (2009) interpreted these alkali-rich potassic granites as resulting from crustal underplating by juvenile, mantle-derived magmas during periods of extension. This model is essentially consistent with our proposed model for the Esperanza-Lavasen suite. The close temporal association of the calcalkaline Pataz batholith and the K-rich Esperanza-Lavasen suite is not unique in arc environments. In northern Luzon, Philippines, overlapping calc-alkaline and alkalirich magmatism takes place during a transitional period, from the former to the latter (Hollings et al. 2011).
The Carboniferous magmatic history of the Pataz district suggests rapidly alternating periods of extension allowing magma to reach the surface and erupt as volcanos, and compression that impeded the upward movement of mantle-derived magmas and promoted ponding and partial melting in the lower crust (Table 3). Alternations between compression and extension on time-scales of less than a million years are not unusual in the Andes (Mercier et al. 1992;James & Sacks 1999). Changes between normal and flat subduction have been proposed to account for similar alternations, but 'normal-flatnormal' cycles take ca 50 Ma (James & Sacks 1999;Kay et al. 1999). Thermal weakening of the crust, during a period of normal subduction and under a consistent external stress field, is proffered as a more viable explanation for rapid switching between extension and compression at Pataz (cf. James & Sacks 1999). During periods of extension and high heat flow that are typically associated with normal subduction, mafic magmas were generated in the mantle wedge and were erupted or emplaced into the upper crust. During these periods, incipient rifting of the arc may have occurred, providing a source for mantle-derived magmas without Nb or Ti anomalies. Over time, the passage of mantle-derived magma weakened the crust, and compressional deformation under the prevailing external stress field obstructed the upward passage of mafic magma, which was then ponded in the lower crust causing partial melting and generation of the more silicic magmas. When sufficient strain had accumulated to again strengthen the crust, extension was re-established, and another phase of magma emplacement or eruption in the upper crust would follow. Additionally, especially in the near surface environment, rapid switching between compressive and extensional tectonics could also be attributed to the transtensional setting of the Pataz-Parcoy gold field in a jog on the regional Cordillera Blanca Fault (Witt et al. 2013a).

Implications for gold mineralisation at Pataz
The auriferous quartz veins hosted by the Pataz batholith were emplaced during regional ENE-WSW shortening (Haeberlin et al. 2004). Crustal shortening prevented escape of the hydrothermal fluid through subvertical NNW-striking faults and promoted fracture and vertical extension under high fluid pressure, forming mineralised veins preferentially in the rheologically strong Pataz batholith (Witt et al. 2013a).
Previous workers have discounted the Pataz batholith as a source for the ore fluid at Pataz on the grounds of incompatibility of age data. However, these age constraints are based largely on Ar-Ar geochronology, which is prone to resetting by later thermal events (Kent & Hagemann 1996;Witt et al. 1996;Vielreicher et al. 2010). Haeberlin et al. (2004) published 40 Ar/ 39 Ar data from metasomatic white mica alteration halos around mineralised batholith-hosted veins. These data generated plateau at 314-312 Ma, which the authors interpreted as indicative of the age of gold mineralisation, although the 40 Ar/ 39 Ar data could as well be interpreted as the minimum age for mineralisation.
A recently published Re-Os date for molybdenite from a quartz-arsenopyrite-molybdenite vein in the Parcoy district (Szappanosn e-V ag o et al. 2010) provides an alternative age for gold mineralisation. Arsenopyrite is abundant in the batholith-hosted auriferous lodes and trace amounts of molybdenite were identified in veins from the northern Pataz district by qualitative SEM analysis (Witt, unpublished data) but this molybdenite has not been dated. The Re-Os age of 331.9 AE1.7 Ma for molybdenite from Parcoy is within error of the emplacement ages of both the Pataz batholith and the Esperanza subvolcanic complex but is closer to the mean age of the latter (334 Ma; Witt et al. 2013a). The batholith-hosted veins post-date high-temperature ductile deformation of batholith contacts (Witt et al. 2013a) and formed during subsequent uplift of the batholith (Haeberlin et al. 2004). Differential uplift of the Pataz batholith and its country rocks produced the Lavasen graben into which the Lavasen volcanic rocks accumulated, having been erupted from centres such as the Esperanza subvolcanic complex, at the margins of the rift. The age of the Esperanza subvolcanic complex (333.6 AE 2.9 Ma; Witt et al. 2013a) is therefore considered close to the age of the batholithhosted veins. The temporary switch from extension during the main Esperanza-Lavasen magmatic event to shortening during formation of the batholith-hosted veins may be attributed to the ongoing tectonic evolution of the Pataz-Parcoy transtensional jog.
It is further argued that, unlike porphyry deposits, the Pataz batholith does not display evidence of widespread hydrothermal activity (other than propylitic alteration) beyond the proximal illitic alteration zones surrounding the mineralised veins (Witt et al. 2013b). Furthermore, data presented here show that, unlike granites associated with intrusion-related gold deposits (Hart et al. 2004;Duuring et al. 2007), the Pataz batholith is not highly fractionated.
Not only is the interpreted timing of gold mineralisation broadly constrained as coeval with Esperanza-Lavasen magmatism, but field and petrographic observations suggest widespread hydrothermal activity, contemporaneous with and spatially related to the Esperanza-Lavasen magmatism, summarised as follows. Latitic rocks of the Esperanza subvolcanic complex display evidence for weak but widespread potassic alteration, secondary hematite and argillic alteration. The Fe-rich dolerites of the complex locally display calc-silicate alteration and minor disseminated pyrite and chalcopyrite. Furthermore, textural heterogeneity and the progressive enrichment of Ca in plagioclase suggest increasing water pressure during crystallisation of the dolerite magma (Witt et al. 2013a). The presence of late-stage plagioclasemagnetite veinlets in some dolerite samples provides evidence for the generation of a residual, Fe-rich, saline, hydrothermal fluid. This Fe-rich hydrothermal fluid may have contributed to the hydrothermal system that deposited gold-rich quartz veins, locally containing abundant iron and base metal (Pb, Zn) sulfides, in the Pataz batholith. The proposed genetic association between gold mineralisation and a vertically zoned magma chamber containing Esperanza-Lavasen suite magmas is consistent with the worldwide association between gold and K-rich igneous rocks derived from enriched (metasomatised) mantle sources (Muller & Groves 1995), a concept further developed by Hronsky et al. (2012). Mafic alkaline magmas in the Bingham and Tintic (Cu-Au-Mo, Ag-Pb-Zn) mining districts, Utah, provide an analogous model to that proposed above for an Fe-rich hydrothermal fluid derived from the mantlederived component of the Esperanza-Lavasen magma chamber. Keith et al. (1997) and Maughan et al. (2002) found that mafic alkaline dykes in these mining districts mixed and fractionated at shallow crustal levels to form ore-related latites and monzonites. They also proposed that resorption and oxidation of the abundant sulfide globules in the mafic magmas may have contributed metals to the ore fluid in these districts. The Bingham porphyry deposit is a different style of deposit to the metre-scale, batholiths-hosted veins at Pataz, but common processes may give rise to gold deposits encompassing a range deposit classifications (e.g. Hronsky et al. 2012). In fact, the overall form of NMORBnormalised spidergrams is very similar to that of mantle-normalised spidergrams for felsic magmas associated with gold mineralisation reported by Hronsky et al. (2012). However, analogues of the Pataz batholiths-hosted vein deposits are not commonly described in the geological literature. Orogenic veins, by definition (Groves et al. Perhaps the closest analogy for Pataz is the Charters Towers district (including Ravenswood) Queensland, Australia, where granitoid-hosted veins, locally containing 90% sulfides (pyrite, galena, sphalerite), are mined for gold (Kreuzer 2005). Like Pataz, the classification of these deposits remains controversial (e.g. Peters 1993; Kreuzer 2005;Bertelli et al. 2009).

SUMMARY AND CONCLUSIONS
A curious and distinctive feature of the Pataz igneous rocks is the persistence through time, extending at least from the Ordovician to the late Mississippian, with bulk compositions displaying positive Pb anomalies and enrichment of Cs and Rb. These persistent features suggest the presence of a metasomatised mantle source beneath the Eastern Andean Cordillera since at least Ordovician times. The earliest magmas derived from this mantle source, the Ordovician Vijus and Atahualpa formations, also display a negative Nb anomaly typical of subduction zone magmas and are interpreted to have been derived by partial melting of a depleted mantle source followed by fractional crystallisation and assimilation of continental crust (Figure 15a). The absence of a negative Nb or Ti anomaly in later (Mississippian) mantle-derived magmas requires convection of the asthenosphere or replacement of the depleted mantle source by undepleted asthenosphere (not previously depleted by extraction of a partial melt; Figure 15b, c). The strongest Pb anomalies are found in least evolved magmas of the low-SiO 2 suite of the Mississippian Pataz batholith. Assimilation of continental crust weakened the Pb anomalies in more evolved samples of the suite but added LREE, MREE, Th, U, Zr and P while simultaneously depleting HREE and Y.
Ponding of mantle-derived magmas is proposed as the heat source for melting of the lower to mid continental crust to produce the more felsic Mississippian magmas that constitute the bulk of the Pataz batholith and the slightly later Esperanza-Lavasen suite. These magmas show similar but stronger enrichment of LREE, MREE, Th, U, Zr and P than the contaminated low-SiO 2 suite magmas, and are interpreted to have formed by partial melting of calc-alkaline dioritic to andesitic lower crust in the case of the Pataz batholith, and of tonalitic middle crust in the case of the Esperanza-Lavasen A-type magmas. Negative Ti anomalies (and Nb anomalies in the case of the Esperanza-Lavasen association) in these magmas were inherited from earlier subduction-related lower to mid crustal source rocks. This model for formation of the high-SiO 2 batholith suite is consistent with the Pb and Nd isotope data of Macfarlane et al. (1999), which they interpret to indicate 35 to 70% crustal addition (assimilation of Maranon Complex basement) to a mantle-derived melt. In contrast, our model for the origin of the Mississippian felsic rocks envisages melting of crustal sources with a variable but subordinate to minor contribution from the mantle-derived melt that also provided a heat source.
Mesoproterozoic rocks of the San Ignacio-Sunsas orogen, locally exposed in the central section of the Eastern Andean Cordillera ), provide a suitable range of compositions for the generation of the high-SiO 2 suite magmas of the Pataz batholith and the Esperanza-Lavasen felsic suite. However, preliminary unpublished Lu-Hf isotope data for zircon from the Esperanza-Lavasen suite (T. Kemp and T. Angerer, pers. comm. 2011) require rapid remelting of mantle-derived source rocks. This could be achieved by melting of midcrustal tonalite and granodiorite formed by fractionation of magma equivalent to low-SiO 2 suite of the Pataz batholith, if some of that magma ponded in the mid crust. Similarly, the Pataz high-SiO 2 suite may have formed by remelting of lower crustal rocks equivalent to the Vijus and Atahualapa andesites, although Lu-Hf data are not available to support this model.
Although geochronological estimates for the emplacement of the Pataz batholith and the Esperanza-Lavasen suite both overlap with the Re-Os age of molybdenite in a mineralised vein from the Parcoy gold field, geological considerations indicate that the Pataz batholith was slightly earlier, and that the latitic to rhyolitic Esperanza-Lavasen suite magmas provide a viable source for the gold ore fluid. It is suggested that K-rich A-type magma chambers (unexposed but inferred as the source of Lavasen Volcanics and small A-type intrusions) belonging to this suite may be genetically related to gold in the Pataz district, but also more widely in the Eastern Andean Cordillera, consistent with the worldwide association between gold and K-rich igneous rocks (Muller & Groves 1995).