In two simplified geological maps (Figures 3, 6), we integrated previous (González et al., 2015; Solari et al., 2017) and new stratigraphic and structural data (cross-cutting relationships along regional faults and folds, recognition of structural styles, and dip domains) of the northern and central parts of the Salar de Punta Negra and Sierra de Varas sectors in the Domeyko Cordillera. Both areas were selected because along these the Paleozoic to Cenozoic record is well preserved, thus having a good constrain of the stratigraphic units that they allow to reconstruct the architecture of the basin. The new structural and stratigraphic dataconsist of strike, dip, stratigraphic thickness measurements, inspection of the geometry and orientation of regional- and mesoscopic-scale faults and folds, type of geological contacts between stratigraphic units. The previous geological data consists of geochronological data (mainly U–Pb ages) determined for igneous and sedimentary rocks, and also we used the distribution of the stratified and igneous rocks to construct local cross-sections. The data were taken on outcrops well exposed along the main creeks and roads. The geological mapping was supported by the visualization of Landsat images at different scales (1:25,000, 1:50,000, and 1:100,000). We also used satellite images obtained from the Google Earth Pro platform to examine places that aredifficult to access.

The seismic data consist of a2D seismic grid of approximately 2,300 km² composed of N–S and WNW–ENE seismic profiles, which were provided by Empresa Nacional del Petróleo (ENAP)Sipetrol (Figures 2, 3A). To improve the previous seismic and structural interpretations of these data, this study increased the seismic amplitudes of those seismic reflectors related to angular unconformities and fault planes to obtain a cleaner image of these and refine the location of the tops and bottoms of the main tectonic sequences, the geometry of the faults and folds, and volcanic products. We choose two seismic profiles (1F014 and 2F004) located in the northern and central sections of the Salar de Punta Negra Basin (Figures 2, 3A) because these profiles exhibit a moderate-to-good quality. These were initially filtered and then migrated using two-way-travel time (TWT) to eliminate the seismic pitfalls (and/or artifacts) that prevent correct interpretation of the data. Considering that this study lacks an exploratory purpose as well as key geophysical information, such as density, velocity, and check shot logs to develop time–depth calibration, the seismic data were interpreted using TWT.

The top of the main geological units was constrained by surface data. Generally, geologic contacts are exposed well at the surface; therefore, their projection at depth in the first 3 seconds is quite reliable. Several contacts define angular unconformities with a high seismic amplitude that can be followed easily along the entire 2D seismic grid because they are marked by top-lap and/or on-lap terminations against the seismic reflectors, whereas the crystalline basement rocks (granitoids) are characterized by a chaotic and transparent seismic signal with low internal impedance contrasts. To understand the stratigraphic record under the basin, we correlated the tectonic sequences and unconformities observed with those previously interpreted under the Salar de Atacama Basin located 18 km to the north. We used the following four criteria to interpret the faults and folds:

(1) Identification of abrupt lateral terminations (cutoffs) and/or an offset of seismic reflectors

We used the Andino and StructureSolver software packages to visualize and interpret the digital seismic data.

We constructed a regional balanced cross section from the interpretation of the 2D seismic profile 2F004 (Figure 2), which shows the complete structure of the basin projecting from the eastern Domeyko Cordillera to the western volcanic arc. This section was elaborated by projecting to the subsurface the first-order faults and folds, the dips of the beds, and the contact between geological units, as presented in Figure 3A. The contacts between the geological units were correlated with the top and/or button of the different tectonic sequences previously interpreted in the seismic profile. Further, we used the stratigraphic thickness reported both in the literature (González et al., 2015; Solari et al., 2017) and estimated for us in the field to construct the stratigraphic packages following the structure identified in the 2F004 seismic profile.

To determine the magnitude and changes in the entire basin related to shortening and extension and to know the initial shape of this, the balanced cross section was sequentially retrodeformed. The retrodeformation was developed assuming the conservation of area before and after deformation as well as the conservation of bed length, which usually is represented by the top or bottom of the stratigraphic beds. The balanced cross section was sequentially restored in five steps, thus showing the progressive growth of the contractional structures from the Late Cretaceous to recent period, and their early extensional deformation from the Permian to the Jurassic. This allowed us to understand some relevant aspects, such as the initial configuration of the rift-related structures (also called pre-Andean structures), the initial stratigraphic architecture of the syn-rift deposits, and how the contraction of the basin occurred. The restoration was achieved using the structural component module of the StructureSolver software (www.structuresolver.com) by applying the simple-shear and flexural-slip algorithms, thus allowing the unfolding and reversing of the individual blocks bounded by faults, and them we tested this using the Andino software (www.andino3.com.ar) by some specifically forward models of the restored faults.

We applied apatite fission-track (AFT) and apatite (U–Th–Sm)/He (AHe) analyses and thermal modeling to characterize the exhumation patterns of the main ranges surrounding the Salar de Punta Negra Basin (Figure 2). These low-temperature thermochronologic methods are based on the thermally activated accumulation of radiogenic products within minerals (i.e., the isotope ⁴He and crystallographic damage caused by fission tracks) that result from the spontaneous decay of radioactive elements such as ²³⁸U, ²³⁵U, and ²³²Th (e.g., Reiners and Brandon, 2006). Three samples of Permian granitoids from the eastern Domeyko Cordillera, Pan de Azucar sector, and western Sierra de Almeida were analyzed by the AFT (samples CDFM2305-06, PAFM2305-05, and CPAFM12; Figure 3; see description in Section 2) and one sample (PAFM2305-03) for AHe.

The apatites were concentrated by conventional heavy-liquid and magnetic-separation procedures. Individual apatite grains were mounted in epoxy resin and etched in 5.5-M HNO₃ at 20°C ± 1°C for 21 s, based on the methods described in Sobel and Seward (2010). For two samples with abundant apatite, a second aliquot was mounted and sent for ²⁵²Cf irradiation to enhance the number of confined tracks available for measurement. Analytical procedures followed the external detector method (Gleadow and Duddy, 1981). The samples, assembled with dosimetry glass CN-5 were irradiated in the IEA-R1 research reactor at the Nuclear Research Institute (IPEN) at the University of São Paulo, Brazil, using a neutron fluence of 9 E15 n cm⁻². Following irradiation, induced tracks were etched with 40% hydrofluoric acid at 21°C for 45 min. Samples were counted at the Institute of Energy and Environment, University of São Paulo, with 1,250× magnification (dry objective) using an Olympus BX51 microscope with a drawing tube located above a digitizing tablet and a Kinetec^TM computer-controlled stage driven by the FTStage 4.0 program (Dumitru, 1993). Age and error calculations were performedusingthe zeta calibration method (Hurford and Green, 1983) witha zeta calibration factor of 350.9 ± 5.5 for CN-5 glass and were reported at the 1σ level (Table 1). Confined track lengths, the angle between the confined tracks and the c-crystallographic axis, and the kinetic parameter D_par were measured _{for each grain analyzed} (Table 1). The belonging of fission-track ages to a single population, i. e, concordance of the age, was assessed with the χ² test (Galbraith, 2005), with P (χ2) values > 5% considered as concordant ages.

For the AHe analyses, five individual apatite grains from sample PAFM2305-03 were hand-picked, photographed, and packed in 1-mm Pt tubes using an Olympus SZX16 stereo microscope at the Low-Temperature Thermochronology Lab of the University of São Paulo, following the procedure suggested by Farley (2002). Grain dimensions and the number of terminations were determined to calculate the ejection (Ft) correction factor (Farley et al., 1996). The apatites were degassed with an Alphachron He extraction and analysis system (Australian Scientific Instruments), which was equipped with a Pfeiffer Prisma 200 Quadrupole mass spectrometer (Pfeiffer Group, Germany) at the University of Potsdam, Germany. U, Th, and Sm were analyzed at the German Research Centre for Geosciences (GFZ) at Potsdam by isotope dilution inductively coupled plasma mass spectrometry (ICP-MS). Age calculations followed the procedure of (Meesters and Dunai, 2002). A 2σ uncertainty was reported for all ages. The potential effect of radiation damage on AHe data was assessed by comparing the effective uranium (e [U] = U + 0.235*Th) with individual ages as a proxy for radiation damage, following the model of Flowers et al. (2009).

Thermal history modeling was performed using the QTQt software (Gallagher, 2012) with AFT ages, track lengths, D_par measurements, and single-aliquot AHe data. The fission track annealing model of Ketcham et al. (2007) and the radiation-damage annealing model of Flowers et al. (2009) were using for the inversions. Data of two Permian granitoid samples from the hanging-wall block of the Imilac Thrust, located approximately 11 km apart (samples CPAFM-12 and CDFM2305-06) were modeled together. For these Permian granitoid samples, we used a high-temperature constraint (300°C ± 10°C) at the crystallization age derived from available U–Pbage data and a low-temperature constraint of 40°C/Ma ± 15°C/Ma associated with the nearby presence of the overlying Middle–Late Permian volcanic rocks of the La Tabla Fm. For sample PAFM2305-03 from the lower Eocene Naranja Formation, we used a low-temperature constraint using surface temperatures (20°C ± 10°C)at the maximum U–Pb U–Pbdepositional age available of 50 ± 2 Ma (Martínez et al., 2020). We ran 10,000 burn-in and 300,000 postburn-in models and provide the thermal histories and plots of measured vs. modeled ages for comparison.

On the surface, the main pieces of evidence for tectonic extension are found in the Sierra de Varas sector to the southwest of the Salar de Punta Negra Basin and along the northern section of the Sierra de Almeida (Figure 2). At Sierra de Varas (Figure 6), specifically at the footwall of the El Profeta fault, a serie of mesoscopic normal faults bounded tilted blocks are well recognized (Figure 7A). These consist of intraformational normal faults that affect the Jurassic depositsof the El ProfetaFormation. In neighboring sectors of the Sierra de Varas, these exhibit the greatest thickness at the hanging-wall blocks and form stratigraphic wedges that thicken along the fault planes related to syn-rift sequences (Espinoza et al., 2018). These syn-rift successions are incorporated into a narrow west-verging thick-skinned fold-and-thrust belt (Amilibia et al., 2008; Espinoza et al., 2018, Figures 6, 7B). Similar stratal geometries are observed on the footwall of the Imilac fault in the eastern Domeyko Cordillera (Figure 3). Here, Upper Triassic sedimentary syn-rift successions of the Pular Formation lie on the footwall block of the fault, thus indicating that, possibly, an ancient Mesozoic rift-related depocenter is hidden under this sector of the Salar de Punta Negra Basin. Other mesoscopic evidence related to Late Permian crustal extension is recognized in the Sierra de Almeida, where some poorly preserved normal faults directly affect the Upper Permian volcanic deposits of the La Tabla Formation (Figure 7C). Along these faults, the strata of the La Tabla Formation exhibit an apparent greater thickness on the hanging wall than that in the footwall fault blocks, suggesting that normal faulting occurred during its accumulation.

Half-graben structures and syn-rift stratigraphic wedges are well identified in the 2D seismic profiles (Figures 8, 9) located along the northern and central part of the basin, which correspond to a series of ancient and partially reactivated west- and east-dipping normal faults. The faults usually exhibit planar and semi-listric shapes with moderate dip angles in their upper sections (approximately 40°–60°) and penetrate down into the Paleozoic crystalline basement rocks. The fault planes are highlighted as thin and semicontinuous inclined reflectors that cut and separate seismic reflectors with different inclinations (Figures 8, 9) and with different seismic reflection patterns (e.g., chaotic seismic reflectors versus parallel seismic reflectors). The hanging-wall fault blocks consist of tilted blocks overlain by parallel seismic reflectors and asymmetrical stratigraphic wedges composed of divergent and intercalated low and high amplitude seismic reflectors (Figures 8, 9) related to Upper Permian–Jurassic deposits. The wedges get thicker near the fault planes and thinner over the shoulders of the half grabens, thus displaying a classical syn-rift stratigraphic architecture (Figure 9).

The shallow seismic reflectors of the youngest syn-rift sequences correlate with the Jurassic sedimentary successions of the El Profeta Formation (Figures 8, 9), whereasthe deeper ones correlate with the Permian stratified rocks of the La Tabla Formation (Figure 9). Some deepest parallel seismic reflectors could be correlated with the Devonian Zorritas Formation; however, there is no subsurface information (e.g., well data) that supports this interpretation. Based on the seismic patterns and the stratigraphic and structural relationships described here, we interpret that the region was initially affected by a regional-scale crustal extension from at least the Late Paleozoic (Permian) to the Jurassic times, thus creating an asymmetrical extensional system.

The main evidence of inverted structures is recognized from the 2D seismic profiles presented in this study. These data show that the syn-rift deposits in the hanging-wall fault blocks form either west- or east-verging asymmetrical anticlines, depending on the initial normal fault dip direction (Figures 8, 9). Along the anticlines, the tops of the syn-rift successions are elevated over their original positions (or regional data). Geometrically, these are composed of short and steeply dipping forelimbs and large and semi-horizontal back limbs, thus creating an asymmetrical anticline, such as those reported by the inversion anticlines created from the reverse-reactivation of normal faults (McClay, 2002; Figures 8, 9). A series of small cone-shape bodies also are recognized at the core of the anticlines (Figures 8, 9). Internally these exhibit chaotic seismic reflectors cutting the regional stratification with a flat base, a typical geometry of volcanic bodies. These could correspond to ancient and fossilized volcanic cones, considering that many of they lie well exposed to the southern part of the basin. The forelimbs are truncated against the fault planes, preferably along their upper segment. This is well recognized in the Almeida, Barrancas Blancas El Salado, and Punta Negra faults (Figures 8, 9). The 2D seismic profile 1F014 shows the existence of three west-verging anticlines under the northern section of the basin. Both are established on the hanging wall of previous Permian and Triassic extensional faults (e.g., the Almeida fault, Figure 8). The faults display normal slip in their lower segments and reverse slip along the uppermost segments, thus indicating a partial reverse reactivation following the classification provided by Bally (1984).

In the central section of the basin, the 2D seismic profile 2F004 also shows the existence of several west-and east-verging inverted structures separated by the Punta Negra structural high (Figure 9). In the western section, the hanging wall of the previous synthetic (e.g., El Salado fault) and antithetic normal faults (e.g., the Punta Negra fault) correspond to asymmetrical anticlines affecting the Permian–Jurassic syn-rift, which are expelled from basement half-graben structures (Figure 9). In the eastern section of the seismic profile, a series of east-verging anticlines are recognized (Figure 9). These mainly involve the Upper Permian deposits of the La Tabla Formation, which is tectonically elevated over the top of the Punta Negra structural high (Figure 9). The faults consist of semi-listric (e.g., the Barrancas Blancasfault) and planar faults. In this region, the Barrancas Blancas fault exhibit full reverse reactivation, considering that the top of basement pre-rift at the hanging wall block is elevated over its initial position on the footwall block, whereas the others were partially inverted, because maintain a reverse slip on the upper segment of the faults and a normal slip on the lower segments; furthermore, some minor faults were not reactivated (Figure 9).

In addition to the normal and inverted structures, several basement-involved east- and west-verging faults are well exposed along the western and northeastern margins of the basin (Figure 3). They consist of reverse faults (e.g., the Imilac, Pan de Azucar, Cachiyuyo, and Agua Amarga faults) that affect both the footwall and hanging wall of the inverted normal faults present in the subsurface and previously described. The hanging wall of the west-dipping Imilacfault is a kilometer-scale block superimposed on the Upper Triassic syn-rift deposits of the Pular Formation (Figure 3). Previous structural interpretations of2D seismic profiles through the fault indicated the presence of a blind and partially inverted Triassic half-graben structure in the footwall block (Martínez et al., 2020). Other structures, such as the west-verging Agua Amargafault (Figure 3), mark the eastern termination of the innermost part of the current volcanic arc. Its hanging-wall block is a broad west-verging basement-cored anticline along which the Paleozoic units of the Imilac Complex have been elevated over 2.5 km. The sedimentary beds of the Paleozoic Zorritas Formation near the fault-plane contact are highly sheared and steeply dipping (Figure 10), a sign of significant shortening experienced by these structures. Usually, the basement reverse faults cut the Pliocene–Pleistocene deposits that form the cover of the basin, thus revealing a Late Cenozoic activity. The2D seismic profile 1F014 (Figure 8) shows an important west-verging basement-involved reverse fault under the eastern section of the basin. This is affecting the hanging wall of a partially inverted normal fault with a style similar to the reported by by-pass faults (Bonini et al., 2012; Figure 8).

In the northern Salar de Punta Negra Basin, specifically at the footwall of the Pan de Azucarfault (Figure 3), contractional growth strata lie preserved along synclines involving the Paleocene–Eocene Naranja Formation (Figure 5A). Recent works (Solari et al., 2017; Martínez et al., 2019) have reported the U–Pb ages for detrital and volcanic zircons of this unit that range approximately between 60 and 55 Ma. Although the field evidence might be insufficient, the 2D seismic data (Figure 9) show shallow and intermediate seismic reflection patterns forming fan shapes over contractional structures, such as inversion anticlines and/or the forelimbs of basement reverse faults, which are the most robust evidence of synorogenic accumulation. The seismic reflectors exhibit an onlapping with growth strata over the crest and forelimbs of the anticlines (Figures 8, 9) and also over the flanks of the syncline folds. They form asymmetrical stratigraphic wedges that are thinner over the anticlines and thicker along the flanks, thus indicating that they were accumulated contemporaneously with contraction. Based on our field observations and the correlations with synorogenic successions exposed in neighboring basins (e.g., Salar de Atacama and Salar de Pedernales), we correlated the basal seismic reflectors of the synorogenic stratigraphic wedges with the Upper Cretaceous synorogenic deposits (e.g., Pajonales Formation; Figure 3), whereas those of the intermediate and shallow seismic reflectors are mostly related to Paleocene–Oligocene–Miocene synorogenic deposits.

The three Permian granitic rocks that we analyzed (PAFM2305-03, CDFM2305-06 PAFM2305-05; and CPAFM12; Table 1) were collected from the hanging-wall blocks of the Imilac and Pan de Azucarfaults described previously (see Section 4.3; Figure 3). The samples consist of granodiorites and monzogranites, with reported U–Pb crystallization ages between 281 and 298 Ma (Solari et al., 2017). The samples were collected from both the top and bottom of the ranges, with elevations between 3,000 and 3,500 m. Samples CDFM2305-06 and CPAFM12 (Figure 3), located ∼11 km apart in the hanging wall of the Imilac fault at 3023 and 3,116 m of elevation, respectively, yielded concordant AFT central ages of 49.4 ± 5.3 and 55.1 ± 7.4 Ma, with relatively long mean track lengths of 13.91 ± 1.34 μm (CDFM2305-06) and 13.10 ± 1.89 μm (CPAFM12) (Table 1; Supplementary Data S1). The AFT data define an age elevation trend, with older ages at higher elevations (Figure 11A). Thermal modeling shows that Permian granitic rocks were cooled rapidly and exhumed through the partial annealing zone between the Paleocene (∼60 Ma) and Eocene (∼40 Ma), as illustrated in Figure 11. Our thermochronological results are in agreement with the post–Late-Cretaceous uplift constrained mainly by the Upper Cretaceous synorogenic deposits reported in the footwall block of this structure in neighboring regions (e.g., Domeyko Cordillera, Salar de Atacama Basin; Arriagada et al., 2006; Bascuñán et al., 2016, Bascuñán et al., 2019; Henriquez et al., 2018; Martínez et al., 2018).

Sample PAFM2305-05 from the hanging-wall block of the Pan de Azucarfault (Figure 3; Table 1) yields a concordantAFT age of 48.0 ± 5.2 Ma, with a mean track length of 14.28 ± 2.16 μm. In addition, five aliquots analyzed for AHe yielded reproducible ages ranging between 44.4. ± 0.6 and 38.5 ± 0.5 Ma (Table 2). Despite a large variation in the effective uranium contents, from 83 to 155 ppm, reproducible ages indicated that limited radiation damage accumulated over the cooling history. Altogether, the thermochronological data suggest rapid cooling sometime between 60 and 35 Ma and limited cooling thereafter, which we interpret as the result of Eocene thrusting along the Pan de Azucar fault. The thermal models (Figure 11A) show time–temperature paths similar to the multisample model of CDFM2305-06 and CPAFM12, which suggests coeval reverse faulting along the Imilac and Pan de Azucar faults. Furthermore, this Eocene exhumation is also recorded by the Eocene synorogenic deposits of the Naranja Formation exposed on its footwall block (Figure 3; Martínez et al., 2020). There, sample PAFM2305-03, collected in the lower Eocene strata of the Naranja Formation, yields a discordant AFT central age of 93.9 ± 15.0 Ma with a shorter mean track length of 13.00 ± 2.43 μm in annealing-resistant apatites with a mean Dpar value of 2.80 μm.An AFT age older than the stratigraphic age documents that this age is, at most, partially reset, due to insufficient burial in the footwall of the Pan de Azucarfault. Such a scenario is captured in the thermal model presented in Figure 11C.

Our mesoscale field observations together with the analysis of 2D seismic profiles showed evidence of extensional structures, which were active during both periods. The pre-shortening restorations (Figures 12A–C) of a regional cross section extended from the eastern Domeyko Cordillera to the innermost part of the present-day volcanic arc show a pre-Andean (or precontraction) tectonic setting dominated by half-graben structures, along which the Permian La Tabla, Triassic Pular, and Jurassic El Profeta Formations were accumulated. Previous studies (Kay et al., 1989; Mpodozis and Kay, 1990, Breitkreuz, 1990; Charrier et al., 2007; Maksaev et al., 2014; Morandé et al., 2015, among others), mostly supported by U–Pb dating and geochemical analyses of Upper Paleozoic intrusive and volcanic rocks, have suggested that an extensional arc magmatic setting was active along the present-day Domeyko Cordillera during the Late Paleozoic, however; major structural evidence and details on the 2D geometry and kinematic evolution of the Permian extensional faults hidden in the pre-Andean basins of northern Chile have not been reported.

The causes of a Late Permian tectonic extension in northern Chile are a matter of debate. Some researchers considered a crustal extension related to geodynamic rollback subduction, which allowed the establishment of back-arc extensional basins (Mpodozis and Kay, 1990; Breitkreuz, 1991; Charrier et al., 2007; Maksaev et al., 2014). Others (Bahlburg and Breitkreuz, 1991; Semperé et al., 1997; Amilibia et al., 2008) interpreted a Late Paleozoic crustal extension as induced by enhanced mantle convection, similar to the pure shear mode of McKenzie (1978) and (Wernicke, 1981). Crustal extension developed large and narrow intracontinental rift systems extending from Central Perú (Arequipa Basin) to northern Chile (Tarapacá Basin), along which thick silica-rich volcanic products and continental sedimentary successions accumulated (Semperé et al., 2002). Regardless of both initial interpretations, we confirmed the existence of Upper Permian east- and west-dipping normal faults bounding halfgrabens under the pre-Andean Depression. Our confirmation supports the notion that Permian crustal extension dominated a significant part of the Late Paleozoic tectonic evolution of the Central Andes (Mégard, 1978; Tankard et al., 1995; Semperé et al., 2002; Ramos, 2009). Our tectonic reconstructions (Figure 12A) indicate that two large half-graben structures developed along the western and eastern sections of the study area.

The seismic tectonic sequences associated with the Upper Triassic (Pular Formation), and Jurassic (El Profeta Formation) deposits display similar stratigraphic and structural relationships to those shown by the Permian La Tabla Formation. Such relationships have also been recognized in several regions of northern Chile (Amilibia et al., 2008; Mpodozis and andRamos, 2008; Martínez et al., 2016; Fuentes et al., 2018; López et al., 2019; Martínez et al., 2019), thus indicating that initial Late Paleozoic crustal extension propagated through the Triassic and Jurassic. Following the kinematic evolution of rift systems proposed from field and experimental studies (McClay et al., 2002; Ziegler and Cloetingh, 2004; Corti, 2012; Corti et al., 2018, among others), these could have taken advantage of pre-existing weaknesses during the initial stages of crustal extension, such as ancient faults, fold axes, suture zones, and regional-scale foliations, among others, which are difficult to identify in the study area.

Accordingly, we propose that the Mesozoic syn-rift sequences were controlled tectonically by the negative reactivation of the previous Permian normal faults, such as is evidenced from the tectonic restorations (Figures 12B,C), and by the creation of new normal faults on the eastern Domeyko Cordillera and the central section of the Salar de Punta Negra Basin. All the interpretations discussed above significantly improve the recent interpretations by Espinoza et al. (2018), who proposed that the crustal extension in the region started ca. 240 Ma ago.

Considering various previous and recent interpretations in neighboring pre-Andean basins (e.g., Salar de Atacama, Salar de Pedernales, and Calama Basins; Muñoz et al., 2002; Mpodozis et al., 2005; Amilibia et al., 2008; Martínez et al., 2018, Martínez et al., 2020, Espinoza et al., 2018; López et al., 2019), where similar extensional structure-related rifts have been recognized, our interpretation is that the pre-Andean Depression was a region of significant pre-orogenic intracontinental weakness. This possibly acted as an elongated crustal depression bounded on one or both sides by basement-involved normal faults, and its axis possibly coincided with the present-day location of the salt flats. The presence of Lower Jurassic deposits along inverted half grabens in neighboring regions, such as the Domeyko Cordillera, Salar de Atacama Basin, Salar de Pedernales Basin, and the Frontal Cordillera of northern Chile, suggests that the crustal extension continued up to Early Jurassic times.

The basal seismic reflectors related to Upper Cretaceous tectonic sequences onlapping on the top of the inversion anticlines indicates that the crustal contraction began in the Late Cretaceous from the reverse reactivation of the previous Permian to Mesozoic half-graben structures (Figure 12D). This contraction was marked by the creation of doubly verging inversion anticlines that developed the early topographic relief in the region. Previous regional studies in northern Chile (e.g., studies of the Domeyko Cordillera, Coastal Cordillera, and pre-Andean Depression; Arriagada et al., 2006; Amilibia et al., 2008; Bascuñán et al., 2016; López et al., 2017, Bascuñán et al., 2019; Fuentes et al., 2018; Henriquez et al., 2018) have also reported a Late Cretaceous age for the onset of the Andean Orogenesis, as it is mainly related to the tectonic inversion of extensional structures of the Tarapacá Basin. Based on the abovementioned studies, the tectonic inversion is interpreted as a generalized mechanism active during the initial growth of the Central Andes of northern Chile. However, in the Salar de Punta Negra Basin, the basin inversion continued during the Cenozoic, as recorded by the Cenozoic synorogenic deposits covering the inversion anticlines (Figures 12D–F).

The tectonic restorations shown in Figure 12 indicate that much ofthe basin inversion was acquired during the Paleocene–Eocene. During this period, some faults located in the easternmost region of the basin were fully inverted and the faults located in the central section (e.g., El Salado fault, Figure 13) were reactivated, thus creating a contractional basin between them (Figures 12E,F, 13). The last suggests that the Cenozoic subsidence of the basin and the neighboring regions into the pre-Andean Depression was controlled by the progressive growth of the large inverted structures, such as it is presented in a tridimensional model (Figure 13; Supplementary Data S2), and not by tectonic extension or strike–slip faulting, as was initially proposed by Pananont et al. (2004) and Jordan et al. (2002), who interpreted the Cenozoic deformation in the region as the result of short-lived contractional and/or neutral and extensional tectonic conditions associated with periods of high and low convergence rates. The Late Cretaceous crustal shortening of the western Central Andes and the onset of the basin inversion are related to the absolute trenchward motion of the overriding plate (South American) from the opening of the Atlantic Ocean (Cobboldet al., 1995; Jaillard and Soler, 1996; Semperé et al., 1997), which led to changes in the regional field stress from extensional to contractional in the continental margin. Recently, similar tectonic conditions have been interpreted in several retroarc regions of the southern Central Andes (Gianni et al., 2018; Gianni and NavarreteandSpagnotto, 2019).

The final episodes of basin inversion were accompanied by significant basement deformation dominated by doubly verging reverse faulting, from which the current east and west tectonic bounds of the basin developed (Figure 13). The timing of this deformationin the region has been mainly constrained by the analysis of AFT data and also by the youngest Cenozoic synorogenic deposits exposed over the footwall blocks of the faults, indicating that the deformation started at least from the Eocene. The timing of reverse faulting was also detected from our AFT data, indicating that a significant cooling episode caused by exhumation of the kilometer-scale basement blocks occurred at this time. This episode is correlated with the so-called Incaic tectonic phase proposed by Steinman (1929), which is responsible for major basement deformation and uplift in the western Central Andes.

Recent U–Pbdating analyses of granitic clasts contained within Eocene–Oligocene and Miocene synorogenic successions exposed in the western Salar de Atacama Basin (Henriquez et al., 2018) have reported Permian ages, confirming that Permian granitic rocks were uplifted tectonically and eroded during the Cenozoic. The basement reverse faulting was favored possibly by the Late Cretaceous to the Miocene–Pliocene eastward migration of the magmatic arc (Bascuñan et al., 2016). This migration led to a rise in the geothermal gradient in the upper crust and, therefore, conditions most favorable to the development of basement-involved tectonics. Several large basement reverse faults cut Pliocene and Pleistocene volcanic and sedimentary rocks, indicating that the inner forearc and the pre-Andean Depression are currently under tectonic contraction.