Detailed field mapping and petrographic characterization were completed for the intrusive units cropping out at La Invernada and La Resolana and their host rocks (Figure 4). The intrusive rocks at La Invernada have previously been treated as a single pluton, but here we consider them as a plutonic complex (La Invernada Plutonic Complex, LIPC), subdivided into six main facies (Figure 4). At La Resolana, in turn, two main intrusive units have been recognized.

The main fault systems observed in the study area are shown in Figure 4.

The fault plane database was analyzed to establish the preferred orientations of fault planes in the study area and in specific structural blocks, and kinematic and dynamic analyses were completed to better understand the conditions under which faulting occurred in different sectors and lithologies of the study area.

When the whole fault plane database is considered (N=227), it is possible to observe three preferred orientations (Figure 10), which are in agreement with the predominant orientation of large-scale faults represented in our geological map (Figure 4). The largest group shows WNW strikes, the second group shows NNW strikes, while a third noticeable population strikes NE to ENE. For all three groups, it is not possible to distinguish a preferred dip direction (Figure 10). When only faults with kinematic indicators are considered (N=132), the three main preferred strikes are preserved, although some preferred dip directions are apparent: WNW-striking faults dip preferentially to the NNE, NNW-striking faults tend to dip towards the WSW, while NE striking faults dip preferentially to the NW.

The kinematic analysis for the 132 fault planes with kinematic information, shows a very high dispersion of pressure and tension axes, evidencing a highly heterogenous deformation (Figure 10). This heterogeneity is confirmed by the results of the dynamic analysis using the Multiple Inverse Method (Figure 10), which shows that fault activation occurred under multiple stress states. The largest cluster (cluster A in Figure 10) is characterized by an E-to ESE-trending σ₁, and σ₃ varying from subhorizontal NNW to subvertical, with ϕ values close to 0 (σ₂ and σ₃ of similar magnitude). This is typical of a transpressive tectonic regime. A second cluster (cluster B in Figure 10) is defined by a σ₁ trending NE and commonly plunging at high angles, although with a high variability, and a subhorizontal, NW-trending σ₃, with ϕ values between 0.5 and 1. This can be considered a transtensional regime. Finally, a third cluster (cluster C in Figure 10) can be defined as a purely strike-slip regime, with σ₁ and σ₃ subhorizontal and trending N and E respectively, and ϕ values close to 0.5.

To better understand the link between fault systems and the flow of magmas and hydrothermal fluids, we also analyzed the preferred orientations and geometries of hydrothermal veins, dikes of variable compositions and volcanic morphologies.

A total of 138 hydrothermal veins were measured. The composition of the vein infill is variable; the most commonly observed mineral associations are epidote-chlorite, quartz, biotite, quartz-biotite, tourmaline and calcite. Epidote-chlorite and tourmaline veins commonly display albite haloes. Veins are commonly millimetric, although thicker veins locally occur. They are not related to any kind of known mineral deposit. They cross-cut the LIPC, and some of them also cut through the La Resolana intrusions. In the volcano-sedimentary host rocks, vein intensity is stronger close to the contact with the middle Miocene-Pliocene plutons. From all of this, a middle Miocene-Pliocene age can be inferred for the hydrothermal activity in the study area.

The analysis of preferred orientations (Figure 10) shows three main tendencies. The main preferred orientation is NNW, with variable dip directions. A second preferred orientation is close to E-W, dipping preferentially towards de north. Finally, a third preferred orientation is distinguishable, with veins in this group striking NE and dipping preferentially towards the NW.

21 dikes were measured, with predominantly andesitic to daci-andesitic compositions, although aplitic dikes were also measured. As with the hydrothermal veins, a late Miocene-Pliocene age is inferred for these dikes due to cross-cutting relationships. The preferred orientation of dikes (Figure 10) highlights two main tendencies. One dike system, striking broadly N-S and dipping at high angles towards the E, was observed at the Ventana III, Cipreses and Los Hornitos areas, while a second dike population strikes predominantly NE and dips towards the NW, observed mainly at La Resolana, and occasionally at Ventana III.

Based on the dikes mapped and measured in the field, complemented with the mapping of additional dike systems and volcanic morphologies using satellite images, a series of volcano-magmatic lineaments were identified (Figure 13).

The NE-to ENE-striking set of dikes identified at La Resolana, coincides with the elongation direction of the edifice of the Cerro Azul stratovolcano, the distribution of the La Resolana monogenetic vents, and the main crater of the Quizapu volcano (Figure 13). The N-S dike trend, identified mainly around the Cipreses River valley, is also reflected in the orientation of a dike system recognized through satellite images in the southern flank of the Cerro Azul stratovolcano (Figure 13). Finally, when considering the monogenetic vents at both Los Hornitos and La Resolana, a NNW alignment is apparent, coincident with one of the main preferred orientation of faults observed at La Invernada, Los Hornitos and La Resolana (Figures 12, 13).

Understanding the variations of magma chemistry in space and time, can provide crucial information for the interpretation of the tectonic setting of magma generation, ascent and emplacement, and also for the evaluation of the fertility of a magmatic suite, i.e., the capacity of a magma to form a giant porphyry-epithermal system. In particular, a direct relationship has been documented between “adakite-like” magmas and major porphyry-epithermal mineral deposits (Sajona and Maury, 1998; Oyarzún et al., 2001; Richards and Kerrich, 2007; Chiaradia et al., 2009), with these magmas being characterized by features such as high Sr/Y ratios and high HREE/LREE ratios (high slopes in REE diagrams, high La/Yb ratios). Initially, this adakitic signature was interpreted as the product of partial melting of the slab in subduction zones, but subsequently it has been shown that an identical geochemical signature can be obtained through several other processes, including partial melting of subducted sediments, partial melting of an amphibole-garnet bearing lower crust, and/or fractional crystallization of garnet and particularly hornblende during magmatic ascent and emplacement (Richards and Kerrich, 2007; Chiaradia et al., 2009). The observed relationship between adakite-like magmas and giant mineral deposits, has been explained by the highly hydrous character of the magmas generating them, which lead to abundant crystallization of hornblende and the temporal suppression of plagioclase formation, and by their large crustal residence times in thickened continental crust under a compressive tectonic regime (Richards and Kerrich, 2007; Chiaradia et al., 2009; Chiaradia et al., 2012; Chiaradia, 2015; Loucks, 2021). Loucks (2014) proposed the use of these geochemical characteristics to discriminate fertile magmas during mineral exploration.

Here we evaluate the Cenozoic spatial and temporal variabilities of these geochemical parameters in our study area and surrounding region, considering all the available geochemical and geochronological data, both from this work and from a compilation of existing publications (Vergara et al., 1999; Díaz et al., 2020). The rock units considered in our analysis correspond to the Colbún Formation (data from Vergara et al., 1999), the LIPC, the equigranular and porphyritic intrusions of the Brahma porphyry (data from Díaz et al., 2020) and the La Resolana intrusions.

The oldest rocks considered in our compilation belong to the Colbún Formation, which has been subdivided into two units separated by an erosional unconformity (Vergara et al., 1999). The lower unit has K/Ar and ⁴⁰Ar/³⁹Ar ages between 35.2 and 27.4 Ma, while the upper unit has K/Ar ages between 20.3 and 15.4 Ma. The latter strongly overlaps the age range of the Colbún Formation in our study area (Cabezas et al., 2018), and their lithologies are identical. Vergara et al. (1999) also documented the geochemistry and K/Ar ages of subvolcanic intrusions related to the Colbún Formation, with an age range of 35-15 Ma. The volcanic and subvolcanic rocks range in composition from basalt to rhyolite, and most of them show tholeiitic affinity in the AFM diagram (Vergara et al., 1999). In a chondrite normalized REE diagram (Figure 9; Vergara et al., 1999), the Colbún Formation igneous rocks show mostly flat patterns, with almost no enrichment of LREE relative to HREE (La_N/Yb_N ratios of 1–1.5), irrespective of their major elements composition and SiO₂ content. In the fertility diagrams of Loucks (2014; Figure 14) and in adakite-like rocks discriminating diagrams (Figure 14), they plot entirely within the field of normal magmatic arc rocks.

As highlighted before, the intrusive facies of the LIPC (∼14-15 Ma) show a calc-alkaline affinity and evidence of HREE depletion, with La_N/Yb_N ratios of 7–11, considerably higher than the partially coeval Colbún Formation. The slope in the REE diagram is slightly higher for the most differentiated (and later) facies. In fertility diagrams, the LIPC facies plot as weakly fertile magmas, while they plot as normal magmatic rocks in the adakite-like discriminate diagrams (Figure 14). All these features are consistent with a progressive increase in the degree of hydration and magma residence times during the evolution of the LIPC, allowing the fractionation of increasing amounts of hornblende.

When our results for the LIPC are compared with the intrusive units related to the nearby and partially coeval Brahma porphyry Cu-Mo deposit, important differences are apparent. Four main intrusive units have been recognized at Brahma (Díaz et al., 2020), two of them equigranular and two porphyritic; all of them have been dated by U-Pb in zircons. The oldest intrusion is a granodiorite (14.6 ± 0.4 Ma), followed by a felsic, hornblende-rich porphyry (14.1 ± 0.2 Ma) and a tonalite (13.4 ± 0.4, 13.8 ± 0.5 Ma). Finally, a second hornblende-rich porphyry was emplaced later (9.3 ± 0.2 Ma). All the intrusive units present at Brahma show steeper slopes in the REE diagram (Figure 9) and show a stronger fertile signature (Figure 14), when compared with the LIPC. The two equigranular units are moderately fertile, the difference with the LIPC being evident in the Sr/Y values (Figure 14). In the adakite-like discriminating plots, the equigranular units still plot in the normal magmatic arc field in the La/Yb versus Yb diagram (La_N/Yb_N 7–19), but they plot in the lower part of the adakite-like rocks in the diagram of Sr/Y versus Y. The two porphyries, in turn, show a similar pattern in terms of REE contents and fertility indices, irrespective of their age difference. They plot entirely within the field of adakite-like rocks (La_N/Yb_N 20–30), and their Sr/Y and V/Sc ratios are characteristic of strongly fertile magmas (Díaz et al., 2020; Figure 14).

Finally, the youngest intrusive units recognized in the study area are the La Resolana intrusions (∼4 Ma). This unit is more depleted in HREE (La_N/Yb_N 11.1) and has higher fertility indices than the LIPC (Figures 9, 14), but it is clearly less fertile and less HREE depleted than any of the Brahma intrusions. It plots within the normal magmatic arc rocks in the La/Yb versus Yb diagram, while in the Sr/Y versus Y diagram it plots in the transition zone between normal magmatic arc rocks and adakite-like rocks.

The different fault systems of the study area acted as pathways for the flow of different types of magmas and hydrothermal fluids. In general terms, there is a close correlation between the orientation of fault planes and hydrothermal veins (Figure 10), suggesting a strong structural control on their emplacement. Field mapping demonstrates that many of these veins are syn-tectonic (Figure 3), and that various generations of cross-cutting hydrothermal veins exist in the area. It is likely that hydrothermal fluids were released after the emplacement and progressive crystallization of each of the intrusive facies of the LIPC (∼15-12 Ma) and the La Resolana intrusions (∼4 Ma). A third hydrothermal pulse is likely to have occurred at ∼8-6 Ma, based on the three K-Ar ages obtained by Drake (1976). Our U-Pb zircon ages at the same localities demonstrate that these K-Ar ages do not correspond to crystallization ages, but they likely reflect the occurrence in the area of a major high-temperature hydrothermal alteration event, associated with the emplacement of magmas not exposed at the present-day surface, or cropping out in segments of the LIPC not covered by our fieldwork and sampling. This inferred magmatic-hydrothermal event would be equivalent in age to the Huemul-Risco Bayo plutonic complex (∼7.2-6.2 Ma; Schaen et al., 2017), located ∼20 km to the south of the LIPC.

The pattern shown by the orientation of dikes seems markedly different than the preferred orientation of fault planes and hydrothermal veins (Figure 10). However, when considering each fault system separately, several correlations are revealed.

The ∼NW striking fault systems (WNW to NNW) display evidence of having controlled the emplacement of some of the most differentiated intrusive facies of the study area. Indeed, a syenogranite dike of the LIPC was observed emplaced in a WNW-striking fault, to the north of the Los Hornitos vents (Figure 7). Also, the La Resolana intrusions show a clear NW elongation (Figure 4), while different felsic facies of the LIPC (monzogranite, granodiorite) display a NW-trending elongation axis (Figure 4). As already discussed, hydrothermal fluids were also channeled by these fault systems, as shown by the WNW and NNW preferred orientations of hydrothermal veins (Figure 10). Finally, these NW-NNW and WNW fault systems can be recognized regionally as morphostructural lineaments in satellite images, and the NW faults in particular, can be correlated to the SE with the Río Maule fault, which is in turn related to the emplacement of the rhyolitic Laguna del Maule volcanic complex (Kohler, 2016), while to the NW, it might be related to the emplacement of NW-elongated Miocene granitoids documented by SERNAGEOMIN (2002) in the Lontué River valley (Figure 1). These high-angle, NW-striking fault systems are oriented at high angles relative to the ∼ENE Neogene subduction direction (Somoza, 1998; Somoza and Ghidella, 2012; Bello-González et al., 2018), and also to the ∼ENE orientation of σ₁ calculated from the fault planes which cross-cut the different facies of the LIPC (La Invernada structural block, Figure 12), which is responsible for the sinistral-reverse sense of movement of the ∼NW-striking faults. This geometric configuration is not ideal for the quick ascent of magmas; instead, it favors a slow ascent and longer residence times, involving the formation of magma chambers in which magmas are allowed to differentiate and volatiles are accumulated, until their sudden release (Cembrano and Lara, 2009; Piquer et al., 2021). It has been proposed that the release of magmas and volatiles along these “misoriented” faults might happen during transient, co- and post-seismic reversals of fault kinematics, during and immediately after major subduction earthquakes (Mpodozis and Cornejo, 2012; Stanton-Yonge et al., 2016). Normal reactivations of NW-striking faults cross-cutting the LIPC observed at the Los Hornitos block, might be associated to this process. Misoriented faults also control magmatism in the Brahma porphyry deposit (Díaz et al., 2020), where a NW to WNW fault system dominates, and the felsic porphyritic intrusions and hydrothermal breccias show a strong NW-WNW elongation.

The ∼N-striking faults, in particular the Rio Cipreses fault, exert a strong control on the emplacement of most of the andesitic and daci-andesitic dikes recognized in the field (Figures 3, 10), whose orientation is parallel to the main fault plane (N strike, and high angle dips towards the E). Furthermore, several dikes were emplaced directly along the observed fault planes, evidenced by striated host rocks with growth of syn-tectonic hydrothermal minerals, and by faut gauge zones formed due to post-magmatic reactivations (Figure 3). Towards the north of the La Invernada Lake, satellite image interpretations allow us to identify a set of N-striking dikes at the southern flank of the Cerro Azul volcano (Figure 13), suggesting that the Cipreses fault is acting as one of the feeders of Quaternary volcanism in the study area.

The NE-to ENE-striking faults also exert a strong control on the emplacement of andesitic dikes within the study area (Figure 10). This is particularly evident in the La Resolana area, where a set of andesitic dikes, which intrude the Colbún Formation and some of them cross-cut also the La Resolana intrusions, are parallel to the ENE-to NE-striking faults recognized in the same area (Figure 4). Besides, the Los Hornitos vents show an ENE-trending alignment, the edifice of the Cerro Azul stratovolcano is strongly elongated in a NE to ENE direction, and it is aligned in the same direction with the La Resolana craters (Figure 13), all of which suggest that, as with the Los Cipreses fault, the NE-to ENE-striking structures also play an important role as feeders of ancient and active volcanism in the study area. This fault set is, in theory, the most favorably oriented for magma ascent in this Andean segment, as it is subparallel to the Neogene convergence direction of the Nazca and South American plates (Somoza, 1998; Somoza and Ghidella, 2012; Bello-González et al., 2018), and to the predominant post-middle Miocene maximum compression direction, as shown by the paleo-stress calculations from fault planes cross-cutting the LIPC (Figure 12).

Through an integration of our geochemical and structural interpretations, it is possible to infer the various factors which controlled the fate of Cenozoic magmatic products within the study area: in some cases, relatively primitive magmas were able to reach the surface, while in some other cases, magmas remained stationed in the upper crust, slowly cooling and crystallizing and achieving various degrees of differentiation. Some of these had the capacity to form porphyry Cu-Mo deposits in the vicinity of our study area.

The oldest Cenozoic magmatic rocks within and around our study area, correspond to the Oligocene-early Miocene volcanic and subvolcanic products of the Colbún Formation. As discussed before and as was highlighted by Vergara et al. (1999), the REE patterns, HREE/LREE ratios and fertility indices of these rocks, suggest that they correspond to normal (non adakite-like) magmatic arc rocks, formed under a relatively thin and extended continental crust. Hornblende or garnet fractionation were not significant during magma evolution; instead, magma ascent was rapid and crustal contamination was minimal. Similar patterns have been observed in equivalent volcanic rocks in various segments of the Principal Cordillera of Central and Southern Chile (Charrier et al., 2002; Kay et al., 2005; Piquer et al., 2017). Structural control on volcanism during this period is unclear, as the original geometry of volcanic and subvolcanic rocks has been strongly modified due to subsequent folding, uplift and erosion. It is possible that some of the ENE-striking dikes mapped in the La Resolana block, cross-cutting the Colbún volcanics but not the La Resolana intrusions, correspond to volcanic feeders of the final stages of middle Miocene volcanism. The extensional conditions reflected in our paleo-stress calculations for groups of faults cross-cutting these volcanic rocks, particularly in the Los Hornitos block, likely correspond in part to the preservation of this episode of intra-arc extension in the structural record.

The youngest volcanic products of the Colbún Formation are almost coeval with the oldest facies of the LIPC, and also with the oldest intrusions of the Brahma porphyry (Figure 14). In particular, the biotite granodiorite and the hornblende monzogranite-granodiorite of the LIPC are coeval (within error) with the oldest of the hornblende-rich porphyries of Brahma. However, major geochemical differences are evident when these intrusions are compared (Figs. 9, 14). The LIPC plutons are more depleted in HREE than the host volcanic rocks, but still within the normal range of arc-related calc-alkaline magmas. The coeval porphyritic units at Brahma, in contrast, show a much stronger HREE depletion, plotting entirely within the field of adakite-like and fertile rocks in discriminating diagrams. Structural mapping and the elongation of intrusive units suggest that in both cases (felsic facies of the LIPC and Brahma porphyries) magmas were channeled by NW-striking fault systems, although the NW-elongation of the magma bodies is more evident for the Brahma porphyries (Díaz et al., 2020). Thus, the higher fertility of the Brahma intrusions is not associated to a regional- or continental-scale control but must be found in another cause.

Variations in REE patterns and fertility indices are commonly inferred to be associated to continental-scale processes: crustal thickening under compression, subduction of oceanic ridges and fracture zones, or subduction and partial melting of sediments and crustal material (e.g., Hollings et al., 2005; Kay et al., 2005; Chiaradia et al., 2009; Piquer et al., 2017). In the case of our study area and the surrounding region, the transition between volcanism of the Colbún Formation and plutonism of the LIPC and Brahma porphyries, might be associated to continental-scale events of crustal thickening and compressive deformation, initiating at ∼15-14 Ma. A middle Miocene compressive event, producing folding of Mesozoic and Oligocene-early Miocene stratigraphic units, was already inferred by Spikings et al. (2008), Giambiagi et al. (2012) and Astaburuaga (2014). This change in stress regime is evident from our paleo-stress calculations for specific structural blocks, which shows that when pre-middle Miocene rocks are considered, deformation is heterogeneous, involving various stress states, while for the La Invernada block, which contains exclusively intrusive facies of the LIPC, all faults can be explained by a single stress solution, involving transpressive conditions with subhorizontal, ENE-trending σ₁. These results indicate that these conditions have been the predominant post-middle Miocene stress state. It is important to note that stress states with a ENE to NE orientation of σ₁ are registered in the La Resolana (cluster A) and Ventana III (cluster C) structural blocks. However, the remarkable geochemical differences between the coeval LIPC plutons and the Brahma porphyries, cannot be explained by this continental-scale phenomena. They are better explained by local differences in magma residence times and evolutionary history in magma chambers at different crustal depths, which in turn depend strongly on the orientation of magma pathways relative to the predominant stress tensor. For middle Miocene and later magmatic activity, if magmas are channelled exclusively by NW-striking faults, misoriented relative to the predominant ENE direction of maximum horizontal compression, they will tend to stall at various crustal levels, incorporating crustal material, becoming more differentiated, richer in volatiles, metals and other incompatible elements, and depleted in HREE due to hornblende fractionation. If magma pathways include more favourably oriented faults, magmas will be able to ascend more rapidly, and their adakite-like signatures and fertility indices will be more subdued. This might explain the differences observed between coeval intrusive units emplaced under the same overall tectonic context, separated by only 30 km (Figure 1). Immediately after the emplacement of the adakite-like and strongly fertile Brahma porphyry at 14.1 ± 0.2 Ma, a tonalite was emplaced at ∼14-13 Ma, with much weaker fertile signatures (Figure 14), although still higher than the LIPC. Local differences in magma pathways and residence times might also explain this high variability in REE patterns and fertility indices observed within very short periods of time, which would be hard to explain by major changes in subduction parameters or crustal thickness.

This is also consistent with what can be deduced from the structural controls and geochemical characteristics of active volcanism in the study area and surrounding region. As highlighted before, at Los Hornitos, La Resolana and Cerro Azul, ENE-striking faults, parallel to the predominant Neogene compression direction, strongly controlled the emplacement of post-Pliocene dike swarms and also the geometry and distribution of stratovolcanoes (Cerro Azul) and monogenetic vents (Los Hornitos). Magmatic products are primitive (andesites and basaltic andesites), and the compositional zonation observed in olivine phenocrysts at Los Hornitos, has been interpreted as indicative of rapid ascent and limited magma residence times in the crust (Salas et al., 2017). A similar situation was described by Sielfeld et al. (2019) at the Tatara-San Pedro-Pellado volcanic complex, located to the south of the study area (Figure 1), where a major ENE-striking fault system (Tatara Damage Zone) controlled the emplacement of dike swarms and the overall geometry of the volcanic complex and the distribution of minor volcanic vents, as well as the geometry of active geothermal systems. Towards the east, at the Laguna del Maule volcanic complex, Kohler (2016) established that the NW-striking Maule fault controls the distribution of explosive, rhyolitic volcanism, while the NE-striking Troncoso fault allows the ascent of basaltic magma. The products of all these volcanic centres, despite their various degrees of differentiation, show normal magmatic arc chemistry. In contrast, The Nevados de Longaví volcano, located ∼60 km to the SW of our study area, constitutes a remarkable anomaly. For equivalent SiO₂ contents, when compared with nearby volcanic centres, it is strongly depleted in HREE, and it shows very elevated La/Yb and Sr/Y ratios, plotting well within the field of adakite-like rocks. Amphibole is abundant in the lavas as phenocryst, irrespective of SiO₂ contents. It can be concluded, then, that the parental magmas of the Nevados de Longaví volcano are a modern equivalent to the strongly hydrated, porphyry-forming magmas present at Brahma and other porphyry Cu deposits elsewhere. It has been hypothesized that the anomaly of the Nevados de Longaví volcanic complex is caused by an abnormal input of fluids from the subducting slab, as the area is located over the projected trace of the subducted Mocha Fracture Zone (Sellés et al., 2004; Rodríguez et al., 2007). However, it is important to note that the Tatara-San Pedro-Pellado volcanic complex, which is of normal magmatic arc composition, is also located over the projection of the same subducted fracture zone (∼35 km towards the NE). Without discarding an anomalous input of slab fluids in the area, it is clear that a major role must be played by subsequent magma pathways and evolution, to explain the observed geochemical differences in nearby volcanic centres. This is totally consistent with our interpretations for the variability observed in Miocene intrusions and is a result which has strong implications for mineral exploration, as it demonstrates that magma fertility cannot be extrapolated to an entire metallogenic belt or magmatic arc segment, even when coeval magmatic centres are considered. This is shown by the fact that fertile magmas in Brahma were able to form a porphyry deposit, while the LIPC magmas of the same age (∼14 Ma) produced pluton-related alteration zones, but not a porphyry system.