The sampling of chrysocolla was carried out in four selected sites, all located close to the coastal cliff on the western metallogenic belt of the Coastal Cordillera: Michilla area (MI), Marimaca area (MA), Caleta El Cobre area (CC) and Paposo area (PA) (Fig. 2). The sites were selected considering the access to mining excavations located at different heights (from nearly 2,000 m a.s.l. to less than 500 m a.s.l.). A total of 154 samples were taken in-situ from mining excavations within the four study areas (Fig. 2) and chips of each sample were mounted in epoxy resin mounts (∅24 mm) and polished at the Universidad Católica del Norte, Chile.

μXRF geochemical maps for this study were produced on a Bruker Tornado M4 μXRF two detector system at the Institute for Geography of the University of Cologne (UoC), Germany. The mapping was performed under vacuum conditions using a 20 µm step size and 20 µm spot size at 15 ms per pixel, two frame counts and 50 kV acceleration.

Raman spectra were obtained with a Renishaw InVia Qontor Raman microscope at the Institute for Geology and Mineralogy, UoC, Germany. Spectrometer calibration was performed before analysis with a built-in silicon standard. Raman spectra of the polished samples were obtained after manually focusing the 532 nm Ne: YAG laser (100 mW) with a ×10 objective (NA = 0.25) on the sample surface. A Centrus 1AY701 front-illuminated CCD detector was used in combination with an 1800 lines mm⁻¹ grating and a slit opening of 65 µm. The estimated resolution is about 2.3 cm⁻¹. Spectra were recorded continuously from 100 to 4000 cm⁻¹ in the SynchroScan™ mode for 10 accumulations with an individual exposure time of 10 s. Mineral identification was performed by comparison with the RRUFF database (Lafuente et al., 2016). We used μXRF and Raman spectroscopy to enhance the analytical identification of minerals identified in our sample set.

In-situ U-Pb dating was carried out using laser ablation-inductively coupled plasma mass spectrometry (LA-ICP-MS) at FIERCE (Frankfurt Isotope and Element Research Centre, Goethe University Frankfurt). A ThermoScientific ElementXr sector field ICPMS was coupled to a RESOlution 193 nm ArF excimer laser (COMpexPro 102) equipped with a two-volume ablation cell (Laurin Technic S155).

Data were acquired in fully automated mode overnight during three analytical sessions. Analytical parameters can be found in Table S2. Raw data were corrected using an in-house VBA spreadsheet program (Gerdes and Zeh, 2006, 2009). No common-Pb correction was applied to the data.

To our knowledge, no chrysocolla reference material (RM) exists. Therefore, the standardisation strategy used to control the reproducibility and accuracy of the chrysocolla analyses was the following. NIST SRM612 reference glass was used as a primary RM to correct for instrumental mass bias, concentration calculations and to tune the instruments. To account for the matrix related bias on the 238U/206Pb ratios of chrysocolla we used another three RM of different and contrasting compositions, i.e.: zircon, monazite and titanite. Monazite was used as the matrix offset RM. We added to the uncertainty budget of each of the analytical sessions, an excess of scatter so that the weighted average of the 238U/206Pb and 207Pb/206Pb ratios of each of the secondary RM had an MSWD ≤1. We also calculated the range of dispersion of the 238U/206Pb ratios between the four RM and added this bias quadratically as a systematic uncertainty on all calculated dates.

We established a pre-screening strategy to target suitable samples, due to potential high initial Pb to radiogenic Pb concentrations in chrysocolla. Therefore, 66 mounts were prepared and pre-screened to enhance the success dating rate. This pre-scan method consists of quickly ablating different areas of the samples, while monitoring the 238U/206Pb and 207Pb/206Pb ratios, to determine which areas were prone to contain suitable and favourable ratios. After scanning the sample, if appropriate areas were found, the laser spots were placed for further analysis.

In the study areas chrysocolla can be found filling cracks in the host rock, as millimetre crusts covering the host rock, or filling amygdales within andesite (Fig. 3a–c). Also, it is possible to find it massive, but it is less common (Fig. 3d). Under the optical microscope, it is common to find banding and botryoidal texture in the chrysocolla (Fig. 4a–b). The banding texture is a consequence of physicochemical changes in the ore-forming fluids and the environment of mineralisation over time (Craig and Vaughan, 1981). Occasionally, chrysocolla can appear partially replacing malachite (Fig. 4c–d). Atacamite, the second most abundant mineral in our samples, shows replacement textures after chrysocolla. In hand specimens, atacamite is observed as a thin crust covering the chrysocolla (Fig. 5a–b). Microscopically, chrysocolla is cut by atacamite (Fig. 5c–d). Additional minerals identified in hand specimens include dioptase, copper pitch, gypsum, goethite, quartz, epidote and zeolite. Furthermore, sulphides such as pyrite, chalcopyrite, bornite, chalcocite and covellite are observed in some samples.

Nine samples were mapped using μXRF. In three samples we identified a chlorine-rich mineral surrounded by chrysocolla (Fig. 6a–c). The samples have variable silica, aluminium, calcium, iron, manganese and cobalt content (Fig. 6a–i). The μXRF analysis also shows that the sample CC2-3 has two kinds of chrysocolla that differ in the silica and copper content (Fig. 6h).

The mineral that is characterised by its striking chlorine content is clinoatacamite (Fig. 7a–c). Besides chrysocolla, atacamite and clinoatacamite, the copper supergene mineralisation is composed by brochantite, malachite and copper pitch (Fig. 7c–f). The silica, aluminium, calcium and iron content measured with μXRF is related to the host rock and some specific minerals (Fig. 7a–i). Silica is associated with quartz and amorphous silica (Fig. 7g–h). The calcium is related to calcite (Fig. 7f) and epidote (Fig. 7h). The iron content is associated with iron oxide such as hematite (Fig. 7g). All LA-ICP-MS analysis were targeted on chrysocolla and malachite.

A total of 66 samples were measured in five sequences to define the U-Pb ratios and the viability of U-Pb dating in chrysocolla. After these measurements, 26 samples were re-analysed, considering the texture and setting more laser spots per sample. Many samples reveal extremely young, inferred ages or present uncertainties larger than the apparent age. From these 26 samples, we obtained seven samples with reliable ages (Table 1) (Figs. 8 and 9). All dates are calculated from multiple spot analyses, ranging from 13 to 81 spots per date. The majority of the analyses have U concentrations between ∼1 and 355 µgg-1, with an average of 60 µgg-1 (see Table S3).

The obtained U-Pb ages in chrysocolla and malachite range from 0.045±0.027 to 8.0±1.2 Ma. Only for the Michilla area (Fig. 2b) we did not get any reliable U-Pb age (Table 1). In the Marimaca area, sample MA22-01-01b yielded an age of 0.33±0.11 Ma (Fig. 8a), whereas sample MA22-01-04b yielded three ages ranging from 0.39±0.30 to 1.0±0.2 Ma (Fig. 9a). In the Caleta El Cobre, sample CC22-06-01 yielded an age of 0.61±0.38 Ma (Fig. 8b), sample CC2-3 yielded three ages ranging from 0.48±0.35 to 8.0±1.2 Ma (Fig. 9b), and sample CC22-03-07a yielded two ages: 0.045±0.027 Ma and 0.17±0.11 Ma (Fig. 9c). In the Paposo area, sample PA23-01-11 yielded an age of 7.4±0.7 Ma (Fig. 8c), whereas sample PA23-04-07 yielded two ages: 0.12±0.11 Ma and 0.14±0.04 Ma (Fig. 9d). The sample CC2-3 presents two types of chrysocolla with different textures and different ages (Fig. 10). The massive chrysocolla has an age of 8.0±1.2 Ma (MSWD = 0.83; n=21; Fig. 10d) and the botryoidal chrysocolla an age of 0.48±0.35 Ma (MSWD = 0.54; n=14; Fig. 10f). All presented plots show only minor data dispersion; therefore, a single large-scale plot of one representative result is shown to illustrate the distance between the data and the concordia intersections (Fig. 11).

The Pourbaix diagram for copper minerals shows that chrysocolla and atacamite precipitate in the same thermodynamic conditions of other copper minerals and its precipitation is controlled by the concentration of Si and Cl in the system (Dold, 2006) (Fig. 12). The chemical composition of the mineralising fluid can explain why other copper supergene minerals, such as brochantite and malachite, are scarce in the studied samples (Fig. 7). This is because the precipitation of brochantite and malachite are controlled by the concentration of sulphur and carbonate in the fluid. Besides, chrysocolla and atacamite are the final minerals to precipitate in supergene copper profiles (Dold et al., 2023). In the Paposo area, we obtained an age of 7.4±0.7 Ma for malachite (PA23-01-11) and two ages for chrysocolla: 0.14±0.04 Ma and 0.12±0.11 Ma (PA23-04-07) (Fig. 13), indicating that chrysocolla precipitated after malachite. Textural relationship shows that chrysocolla in the studied areas replace malachite (Fig. 4c–d). In manto- and vein-type deposits in the Coastal Cordillera, it has been described that chrysocolla precipitate later than atacamite (Kojima et al., 2003; Trista and Kojima, 2003). Nevertheless, the textural relationship between chrysocolla and atacamite found in our samples indicates that the atacamite is formed later than the chrysocolla (Fig. 5). Besides, the ages obtained on chrysocolla are older than those yielded in atacamite by Reich et al. (2009) in deposits from the Coastal Cordillera (Fig. 14). The presence of atacamite in the samples is interpreted as the final stage for copper mineralisation in the Coastal Cordillera triggered by the reaction between Cl-rich fluid and chrysocolla described in Eq. (1). These fluids have two origins: (1) chloride-enriched fluid resulting from evaporation in a hyperarid climate (Reich et al., 2008; Lambiel et al., 2023); or (2) saline groundwater flowing through faults (Cameron et al., 2010; Dold et al., 2023). 1Cu2H2Si2O5(OH)4⋅nH2O+2CuCl++H2O→2Cu2(OH)3Cl+SiO2+SiO2⋅nH2O+2H+

Some of the obtained apparent U-Pb ages are extremely young (less than 1 Ma). A previous work conducting LA-ICP-MS U-Pb dating in Mn-rich chrysocolla, also known as copper pitch, suggests that the Pb-loss in the copper pitch occurs because the mineral structure is unable to keep the Pb after mineral precipitation (Kahou et al., 2021). Copper pitch is a mineraloid consisting of chrysocolla with co-precipitated birnessite ((Na,Ca)_0.5(Mn⁴⁺,Mn3+)2O4⋅1.5H₂O) (Schwartz, 1934; Throop and Buseck, 1971; Dold et al., 2023). The age spread observed in Mn-rich chrysocolla can be explained by a complex behaviour of Pb in this mineraloid during the interaction of post-crystallisation fluid events (Kahou et al., 2021). Nevertheless, in the studied samples there is no evidence of Pb-loss in chrysocolla and malachite, so the Pb-loss is discarded as an explanation for the obtained ages.

The sample CC2-3 shows the most complex result with three different ages (8.0±1.2; 5.5±1.3; 0.48±0.35 Ma) (Fig. 10) and two types of chrysocolla: one with a massive texture, a higher silica content (Fig. 10c) and the oldest ages (8.0±1.2; 5.5±1.3 Ma) (Fig. 10d–e), and other with a botryoidal texture, a lower silica content (Fig. 10c) and the youngest age (0.48±0.35 Ma) (Fig. 10f). The age dispersion measured within the massive chrysocolla (8.0±1.2 and 5.5±1.3 Ma) (Fig. 10d–e) might be explained by different precipitation process (Fig. 10d–e). According to the textural relationship, the botryoidal chrysocolla is the first to precipitate, because zoned monomineralic bands such as the observed on this chrysocolla are evidence of mineral growth in open spaces into fluid-filled voids (Craig and Vaughan, 1981). Nevertheless, according to the obtained U-Pb age, this botryoidal chrysocolla seems to be younger than the surrounding chrysocolla (Fig. 10f). In consequence, the ages younger than 1 Ma, obtained in this work and the age dispersion found in samples MA22-01-04b and CC2-3 (Fig. 9a–b) are consequence of recent supergene process in the Coastal Cordillera.

The study sites are located on the western side of the Coastal Cordillera (Figs. 1 and 2). The dated samples were taken from abandoned mine sites with an altitude that range from 821 to 2084 m a.s.l. (Table 1) in Marimaca, Caleta El Cobre and Paposo; no results were obtained from Michilla. Comparing the obtained ages vs latitude, it is possible to observe that young ages are present at all sampling sites where reliable U-Pb age were obtained, and the old ages are only in the southern sampling sites (Fig. 13a). The age vs elevation graph shows that in Caleta El Cobre and Paposo, where two samples were dated at different elevations, the older age corresponds to the higher altitude and the younger age to the lower one (Fig. 13b). In Marimaca, we did not get samples at different altitude, and both samples are below the maximum coastal fog influence (Fig. 13b). In the Caleta El Cobre area, the sample CC2-3 was taken at 1465 m a.s.l. and has an age of 8.0±1.2 Ma while the sample CC22-06-01 was taken at 821 m a.s.l. and have an age of 0.61±0.38 Ma (Fig. 13b). In the Paranal area the sample PA23-01-11 was taken at 2084 m a.s.l. and has an age of 7.4±0.7 Ma while the sample PA23-04-07 was taken at 1024 m a.s.l. and range from 0.12±0.11 to 0.14±0.04 Ma (Fig. 13b). In consequence, we suggest, based on our limited number of samples from the Coastal Cordillera, that the elevation of the samples have more influence on the mineralisation ages than the latitude (Fig. 13). This relation between elevation and age could be related to the persistent presence of coastal fog, locally named camanchaca, that is a characteristic of the Coastal Cordillera in northern Chile (Cereceda et al., 2002; Rech et al., 2003; Schween et al., 2022). The coastal fog provides fresh water that helps to develop ecosystems and drive surface processes (Cereceda et al., 2002; Dunai et al., 2020; Schween et al., 2022). The coastal fog is restricted to a height range from 300 to 1100 m a.s.l. in winter and from 600 to 1200 m a.s.l. in spring, and can reach 50 km inland through corridors formed within the Coastal Cordillera (Cereceda et al., 2002; Rech et al., 2003; Schween et al., 2022; Lobos-Roco et al., 2024). The majority of young ages are within these height ranges (Fig. 13b), and all dated samples are within the inland extent of coastal fog influence (Fig. 2). The fog occurrence and fog water availability along the Coastal Cordillera varies in time scales from diurnal to interannual (Larrain et al., 2002; Cereceda et al., 2008; Garreaud et al., 2008). Several studies along the coast of northern Chile have reported average annual fog water fluxes that range from 7 to 0.16 mm d⁻¹ (Larrain et al., 2002; Cereceda et al., 2008; Carvajal et al., 2022). The measurements vary within the year reaching the maximum fog water fluxes during winter and the minimum during summer (Cereceda et al., 2008; del Río et al., 2018; Carvajal et al., 2022; Schween et al., 2022). The fog cloud cover and fog water fluxes are controlled by the El Niño Southern Oscillation (ENSO) (Garreaud et al., 2008; del Río et al., 2018). To the north of ∼25° S during ENSO (+) (El Niño), there is a significantly higher fog cloud cover and fog water fluxes during both summer and winter when compared to ENSO (-) (La Niña) years (Garreaud et al., 2008; del Río et al., 2018). In Alto Patache (20°49^′ S, 70°09^′ W; 850 m a.s.l.), during September of 1997 (very strong El Niño) amount of fog-precipitation was collected of up to 28.4 mm d⁻¹ (∼852 mm per month), that is the highest month fog water collection registered in the area (Muñoz-Schick et al., 2001). Evenstar et al. (2024) propose that to develop supergene mineralisation, is required a precipitation rate above 120 mm yr⁻¹. Nevertheless, this value represents a long-term average estimated based on groundwater recharge; therefore, it does not necessarily represent the minimum value required to trigger supergene mineralisation. Under suitable conditions, a much lower MAR may be sufficient for the development of supergene mineralisation in the Coastal Cordillera. Besides, it has been proposed that the supergene mineralisation in the Atacama Desert could be active during arid conditions (Clarke, 2006), and that the supergene mineralisation is a still ongoing process after the onset of predominant hyperaridity (Reich et al., 2009; Bissig and Riquelme, 2010; Morales-Leal et al., 2023). Considering the amount of fog water collected in Alto Patache it is possible to propose that during very strong El Niño periods there might be enough water to trigger supergene mineralisation in the Coastal Cordillera.

If coastal fog is the main driver providing moisture to initiate supergene mineralisation, the older ages indicate that the supergene deposits were uplifted above the maximum fog height and therefore supergene mineralisation ceased. Uplift rates along the Coastal Cordillera vary significantly along strike. Reported uplift rates range from ∼45 m Myr⁻¹ during the Late Cretaceous to Paleocene (Juez-Larré et al., 2010) to ∼600 m Myr⁻¹ during the Pleistocene (Martinod et al., 2016). The obtained ages may therefore be influenced by heterogeneous uplift rates along the Coastal Cordillera, which could have uplifted samples above the zone of fog influence after supergene mineralisation, disconnecting them from any moisture supply capable of sustaining supergene processes. In addition, both fog intensity and fog height may vary through time in response to climatic changes. Consequently, supergene activity along the Coastal Cordillera is likely controlled by two main forcing factors: spatially variable uplift rates along strike and temporal variations in fog intensity and maximum fog height.

Supergene mineralisation studies, mainly from the Central Depression and Precordillera propose that the supergene mineralisation in the Atacama Desert was active from Eocene to Late Pleistocene, with the majority of supergene mineralisation ages from the Early to Middle Miocene. Based on the type of supergene mineralisation two general stages can be differentiated (Hartley and Rice, 2005; Arancibia et al., 2006; Reich et al., 2009; Evenstar et al., 2024) (Fig. 14). The first stage started 45 Ma and lasted until 5 Ma, with the majority of ages in the Early to Middle Miocene and with a scarce record of supergene minerals between 9 and 5 Ma (Arancibia et al., 2006; Reich et al., 2009). This scarce record between 9 and 5 Ma has been used to propose a Middle to Late Miocene onset of the hyperaridity in the Atacama Desert (Sillitoe and McKee, 1996). The second stage started at 2 Ma and lasted until the Late Pleistocene, and it is restricted to the atacamite precipitation (Reich et al., 2009). From both stages there are only four ages from the Coastal Cordillera: 21.1±0.6 Ma using K-Ar in supergene alunite (Sillitoe and McKee, 1996); 75.3±0.4, 84±11 and 143±29 ka using Th-U in gypsum intergrowth with atacamite (Reich et al., 2009) (Fig. 1). The Th-U ages has been described as the youngest age recorded in a supergene profile in the Atacama Desert (Reich et al., 2009). Nevertheless, gypsum is formed under arid conditions (Murray, 1964; Sofer, 1978) and is not directly related to the supergene copper mineralisation process. Furthermore, atacamite represents the final stage of the supergene mineralisation in the Coastal Cordillera, and its precipitation requires Cl-rich fluids produced by evaporation under hyperarid climate (Reich et al., 2008; Lambiel et al., 2023). If we only consider the ages obtained in copper supergene minerals, the obtained U-Pb chrysocolla ages in this work are the youngest ages recorded in supergene deposits from the Atacama Desert. The oldest ages obtained in this work (Sample CC2-3: 8.0±1.2, 5.5±1.3; PA23-01-11: 7.4±0.7 Ma; Fig. 13a) are in the time gap between 9–5 Ma, where the record of supergene minerals is scarce. The rest of the samples are in the second stage of mineralisation (Fig. 14). The youngest age of supergene mineralisation in a deposit is interpreted as the last time with sufficient moisture (above 120 mm yr⁻¹ as stated by Evenstar et al., 2024), so it should reflect the transition from arid towards hyperarid conditions (Alpers and Brimhall, 1988; Sillitoe and McKee, 1996; Hartley and Rice, 2005). However, even short-lived intervals with sufficient precipitation could have triggered supergene activity and the associated chrysocolla precipitation in the Coastal Cordillera within a predominantly hyperarid climate that has persisted for millions of years. The obtained ages on copper supergene minerals are young (from 8.0±1.2 to 0.045±0.027 Ma) and are consistent with the model that proposes that the supergene mineralisation is a still ongoing process after the onset of predominant hyperaridity in the Coastal Cordillera (Reich et al., 2009; Bissig and Riquelme, 2010; Morales-Leal et al., 2023). Furthermore, the results suggest that wetter periods occurred in the Coastal Cordillera but did not reactivate supergene activity in the Precordillera. This implies that the supergene mineralisation processes in the Coastal Cordillera and those in the Precordillera and Central Depression were most likely triggered by different water sources such as coastal fog.

In addition to the influence of coastal fog, the Middle to Late Miocene supergene mineralisation (sample CC2-3: 8.0±1.2, 5.5±1.3; PA23-01-11: 7.4±0.7 Ma) may record a (or multiple) previously unrecognised moisture event restricted to the Coastal Cordillera. This phase is not documented in palaeoclimatic or supergene archives from the Central Depression or the Precordillera to the east, suggesting a spatially limited hydrological signal confined to the coastal domain. However, the present study lacks samples from higher-elevation areas of the Coastal Cordillera, which limits our ability to further substantiate this interpretation. Future work will target these areas to better constrain the supergene mineralisation processes, their forcing factors, and the relative contributions of potential moisture sources.

Cosmogenic nuclides ages from the Coastal Cordillera at ∼ 19°34^′ S (Dunai et al., 2005) and at 21°30^′ (Ritter et al., 2018a) indicate the onset of predominant hyperarid conditions at the Oligocene/Miocene respectively for areas in the hyperarid core of the Atacama Desert between ∼19–22° S (Figs. 1 and 14). Erosion rate data at ∼24° S support a Late Miocene onset of hyperaridity in the Coastal Cordillera (Placzek et al., 2014) (Fig. 1). Sedimentological studies in the Coastal Cordillera at ∼26° S propose that pluvial activity capable to trigger stream incisions were reduced to insignificant levels in the late Pliocene to early Pleistocene as consequence of the onset of hyperaridity (Amundson et al., 2012) (Fig. 1). Nevertheless, cosmogenic nuclides data support that the Coastal Cordillera near the ∼24° S (Fig. 1) has remained geomorphologically active during the Pleistocene triggered by rainfalls that come from the south (Placzek et al., 2010) (Fig. 14).

Cosmogenic nuclides ages from the Central Depression support that, despite a predominantly hyperarid climate in the Atacama Desert, punctual and episodic wetter episodes occurred (Jordan et al., 2014; Evenstar et al., 2017; Ritter et al., 2018b). Proposed pluvial events in the Central Depression have been dated prior to 35–23 Ma, 17–16 Ma, 12–11 Ma, 8–7 Ma, 5.5–4.5 Ma, 4–3.6 Ma, 2.6–2.2 Ma, 2.92±0.24 Ma, 1.27±0.47 Ma, 540±160 ka, 392±37 ka and 274±74 ka (Jordan et al., 2014; Evenstar et al., 2017; Ritter et al., 2018b). Although some of these ages overlap with those obtained for chrysocolla and malachite in this study, the pluvial events are mostly associated with Andean-derived precipitation, whereas the copper supergene mineralisation analysed here more likely reflects episodes of pluvial and/or intense fog activity in the Coastal Cordillera.