The physical structure of the oceanic and continental plates have had an important role in the long- and short-term deformation process of the subduction margins. On the other hand, the tectonic activity has modified the internal structure and geometry of the tectonic plates (i.e. Bilek et al., 2003; Hackney et al., 2006; Hicks et al., 2014; Contreras-Reyes and Carrizo, 2011; Bassett and Watts, 2015; Poli et al., 2017). This geodynamical feedback is evinced by spatial correlations between the physical segmentation of the continental wedge, and ruptures of large megathrust earthquakes (i.e. Contreras-Reyes et al., 2010; Li and Liu, 2017; Martínez-Loriente et al., 2019; Molina et al., 2021). Examples of this are the spatial correlation between gravity (density) anomalies in the continental wedge and the location of high slip patches in large earthquakes (> 7.5–8 Mw, Song and Simons, 2003; Wells et al., 2003; Álvarez et al., 2014; Bassett and Watts, 2015; Bassett et al., 2016; Schurr et al., 2020), which suggests that changes in normal stresses on the seismogenic zone have a role on the seismic rate and slip propagation during large earthquakes (Tassara, 2010; Maksymowicz et al., 2015, 2018; Molina et al., 2021). On the other hand, changes of the continental wedge geometry have been associated with variations of the interplate boundary friction at the maximum slip patches of the large 2011 Mw 9.0 Tohoku-Oki, 2010 Mw 8.8 Maule and 1960 Mw 9.6 Valdivia earthquakes (Cubas et al., 2013a, b; Maksymowicz, 2015; Contreras-Reyes et al., 2017; Molina et al., 2021).

Diverse works have highlighted the importance of the transition between accretionary prisms (or highly fractured frontal units) and the more rigid rocks of the continental basement as a tectonic limit, controlling, at least partially, the upward propagation of coseismic slip, foreshocks and aftershocks during large megathrust earthquakes (Scholz, 1998; Contreras-Reyes et al., 2010; Moscoso et al., 2011; Kodaira et al., 2012; León-Ríos et al., 2016; Maksymowicz et al., 2017, 2018; Tsuji et al., 2017). At the same time, the downdip limit of the megathrust earthquakes has been related (among other factors) to physical properties of the mantle wedge and deep interplate boundary (Peacock and Hyndman, 1999; Seno, 2005; Wang et al., 2020), which are modified by fluid subduction, slab dehydration and the presence of basal accretionary complexes (Moreno et al., 2018; Menant et al., 2019). Nevertheless, less attention has been paid to the internal physical structure (and lithology) of the continental crust above the downward limit of the megathrust, even considering that all forearc units above the fragile–ductile limit should work as a part of the same mechanically coupled system (van Dinther et al., 2012; Comte et al., 2019).

In this context, we have explored the continental forearc density structure of the Nazca–South America subduction zone in a segment where the high slip patch of the giant 1960 Mw9.6 Valdivia earthquake ruptured (Fig. 1). As mentioned before, this slip patch correlates not only with a low gravity anomaly above the marine forearc (Wells et al., 2003) and low continental slope angles (Maksymowicz, 2015), but also, with a landward extension of Paleozoic metamorphic outcrops on the shore (Fig. 1a). Furthermore, ages and petrological data of continental basement rocks (metamorphic and plutonic rocks) suggest a complex ancient history of accreted terranes (Ramos et al., 1986; Rapalini, 2005) that constitutes the present continental crust in the area (Fig. 1b). Particularly, the recent proposal of an oceanic terrane accreted against the Gondwana margin during Devonian times (Chaitenia, Hervé et al., 2016, 2018, Ct in Fig. 1b) could determine changes in the internal structure of the continental forearc, southward of ∼ 40∘ S. However, the exact limits of this basement configuration remain poorly constrained. In order to reveal the crustal structure of this active portion of the Chilean margin, this work presents the results and interpretation obtained from regional 2D density models, extended from Nazca plate to the Andes Cordillera (Fig. 3), and a local 3D density inversion of the continental forearc (red rectangle in Figs. 2 and 3). The models include magnetotellurics (MT) and transient electromagnetic (TEM) measurements, as well as available independent geophysical and geological data to constrain forward modelling and 3D inversion.

The study zone, located between 38.5 and 42.5∘ S (Fig. 1), is part of the south-central Chilean margin, where the oceanic Nazca plate subducts beneath the continental South American plate. The current rapid convergence rate (∼ 6.6 cm yr-1, Kendrick et al., 2003; Vigny et al., 2009) determines high seismotectonic activity, including the occurrence of mega-earthquakes such as the giant 1960 Mw9.6 Valdivia earthquake (the largest instrumentally registered worldwide). In the long term, this subduction process has been continuously active since Jurassic times (Charrier et al., 2007), being superimposed to ancient tectonic processes of Gondwana structuration (Ramos et al., 1986), and generating the current configuration of the continental South American plate western border. Marine seismic studies, to the north and south of the study area, indicate that the structure of the continental wedge shows physical and tectonic segmentation from the trench to the coast, characterized by active accretionary prisms along the lower slope regions, compressional geometries, and the development of confined slope basins inside the middle and upper slope, while the shelf region exhibits forearc basins with a complex deformation style structured by normal and inverted faults (Bangs and Cande, 1997; Geersen et al., 2011; Becerra et al., 2013; Bangs et al., 2020). Consistently, Vp models derived from wide-angle seismic refraction, at 38∘ S (Contreras-Reyes et al., 2008) and south of 43∘ S (Contreras-Reyes et al., 2010), present changes in the deep structure of the continental wedge that can be interpreted as transitions between the accretionary prism, paleo-accretionary rocks and continental basement. Regarding the geometry of the marine forearc, the continental wedge shows a narrow continental slope (defined between the deformation front and shelf break; see Fig. 1) to the south of ∼ 41∘ S. This morphological change corresponds to a decrease in the slope angle at the northern region of the Valdivia earthquake rupture (38.5–41∘ S), which in turn can be interpreted as a decrease in the effective friction coefficient (μb∗) at the interplate boundary (Dahlen, 1984; Cubas et al., 2013b; Maksymowicz, 2015).

Onshore, three major trench-parallel morphostructural units from west to east can be observed (Fig. 1a): (1) the Coastal Cordillera (CC), where old rocks of a paired metamorphic belt are exposed (Hervé, 1988); (2) the Central Depression (CD) characterized by the presence of unconsolidated Quaternary sediments overlaying Cenozoic deposits (Jordan et al., 2001); and (3) the Principal Cordillera (PC), where the active volcanic arc is currently located. In a close spatial relation with the volcanic arc, the prominent Liquiñe–Ofqui Fault System (LOFS, Fig. 1) stretches along more than 1000 km between 37 and 46∘ S (Cembrano et al., 1996). This continental structure has been interpreted as right-lateral strike-slip system that currently concentrates most of the crustal intraplate seismic activity in response to oblique Nazca–South America convergence (Lange et al., 2008; Orts et al., 2012) and exhumation at these latitudes (Adriasola et al., 2005; Glodny et al., 2008). Moreover, numerous tectonic lineaments and fault zones have been described (SERNAGEOMIN, 2003; Melnick and Echtler, 2006), generally showing north-west and north-east orientations. According to Melnick et al. (2009), the kinematics of LOFS generates intense deformation in its northern limit, explaining the deformation associated with large north-west strike continental faults (as LFZ) and the eastward bending of the CC.

Accretionary metamorphic complexes, associated with late Paleozoic–early Mesozoic subduction, are exhumed along the study zone (Hervé, 1988; Duhart et al., 2001; Willner et al., 2004; Hervé et al., 2013). These units correspond to a paired metamorphic belt, which includes the Western and Eastern Series (WS/ES) formed under high P/T and low P/T conditions, respectively. WS has been interpreted as a basal accretionary complex while ES is interpreted as a frontal accretionary prism and/or as the shallow sedimentary units deformed by the basal underplating of WS units (Glodny et al., 2005; Willner et al., 2005). This paired metamorphic belt is observed continuously at the CC, but the width of their outcrops varies along the margin (see Fig. 1a). Between ∼ 38 and 40∘ S, and southward of ∼ 41.5∘ S, outcrops of WS are observed eastward, near the western limit of PC. Thus, between ∼ 40 and ∼ 41.5∘ S, the eastern limit of these units is not defined due to the presence of the CD deposits and could form most of the forearc basement, or it could be confined near the coast. Westward of accretionary metamorphic complexes and north of 38∘ S, the Coastal Batholith (late Paleozoic intrusive rocks) is observed along CC, but southward (in the study zone) the outcrops of this ancient volcanic arc bends to the southeast and becomes part of the PC. Younger Plutonic and intrusive rocks, related to magmatic arcs from Mesozoic to Cenozoic times (Andean tectonic cycle), are observed along the PC near the position of the active volcanic arc and the LOFS, forming the North Patagonian Batholith (Charrier et al., 2007; Hervé et al., 2018; SERNAGEOMIN, 2003; SEGMAR, 1997; see Fig. 1a).

The continental crust of the western border of South America was configured, during Paleozoic times, by collisions of allochthonous terranes against Gondwana (Rapalini, 2005). To the north of the study zone, Chilenia terrane (Ch in Fig. 1b) collided during Devonian times (Ramos et al., 1986; Hyppolito et al., 2014, and references therein), but its southern extension is roughly defined and could be present in the northern area of the study zone. Southward, the geodynamic evolution of the margin during Devonian to Triassic times has been explained with a double subduction system (Hervé et al., 2016). These authors proposed the development of an island arc (named as Chaitenia, Ct in Fig. 1b) parallel to the margin colliding with Gondwana during Carboniferous times (Hervé et al., 2016, 2018; Rapela et al., 2021, Ct in Fig. 1b). If this hypothesis is correct, the continental crust of the current forearc corresponds to Chaitenia, south of ∼ 40∘ S. However, it is important to point out that the limits between all these terranes are poorly constrained in the study zone owing to the scarcity of basement outcrops.

These obtained results exhibit a landward segmentation of the continental wedge density structure observed from trench to arc (see an interpretative schema at Fig. 8a). Offshore, the frontal portion of the continental wedge (to ∼ 25–35 km landward from the deformation front) presents low densities with a rapid horizontal increment of densities, interpreted as a compaction process in the active accretionary prism along the south-central Chilean margin (Maksymowicz et al., 2015). It is also evidenced by seismic studies in the region (Moscoso et al., 2011; Tréhu et al., 2019; Contreras-Reyes et al., 2008, 2010; Bangs and Cande, 1997; Bangs et al., 2020). To the east, below the sedimentary fill of slope and shelf basins, the continental wedge is characterized by a second unit of higher density and lower horizontal density gradient (middle wedge unit, MWU). This unit can be associated with fractured basement rocks and/or more compacted units of a paleo-accretionary prism. In this sense, Contreras-Reyes et al. (2008) at ∼ 38∘ S and Contreras-Reyes et al. (2010) at ∼ 43∘ S interpret this unit (in Vp-depth profiles) as a paleo-accretionary prisms of an undetermined age between Mesozoic and Tertiary. On the other hand, at ∼ 39 and ∼ 40.5∘ S, Bangs et al. (2020) suggest that the Paleozoic–early Mesozoic accretionary complex (WS/ES) can extend further seaward to the eastern limit of the active accretionary prism (seismic backstop), in accordance with the interpretation of marine seismic data (and boreholes) of González (1989) at ∼ 42∘ S. However, the exploration boreholes presented by González (1989) were drilled in the shelf basin area and therefore do not provide direct information about the age of the continental basement in the western portion of MWU.

Landward from MWU, the next segment correlates on the surface with the morphostructural domain of CC and shows a density increase with respect to the marine wedge, but lower densities compared to continental crust below the CD and PC. Therefore, this CC domain is clearly related to the Paleozoic–early Mesozoic accretionary complexes (WS/ES) and their continuity to depth. Gravity modelling techniques do not define a downward limit of WS/ES (without independent deep constraints). Nevertheless, interpretations of seismic reflection data at ∼ 38.25∘ S (Krawczyk et al., 2006; Ramos et al., 2018) showed the downward prolongation of WS/ES reaching deep levels near the continental Moho interface (∼ 30 km depth). As previously mentioned, the seaward limit of WS/ES is not defined by direct lithological observations; their presence beneath the shelf basin is confirmed by exploration boreholes (González, 1989). Thereafter, the relative rapid change in velocity associated with the transition between MWU and CC domain (dotted grey line in Fig. 8) is interpreted as a structural limit (rather than a lithological change of the basement). This structural limit is probably associated with the development of the shelf basin and a general seaward increase in fracturing within the continental wedge. This structural interpretation seems to be confirmed by Contreras-Reyes et al. (2008) at ∼ 38.25∘ S, where continental intraplate seismicity (located by Haberland et al., 2006) is aligned with this limit, as well as the intraplate seismicity located at 39.5∘ S by Dzierma et al. (2012c).

The CC domain extends landward to the contact with the D1 anomaly (dotted red line in Fig. 8). The eastern border of CC range at the surface correlates almost exactly with the western border of D1 in the 3D model (Figs. 6 and 7), which is also observed in 2D regional profiles (Fig. 4). Accordingly, we understand that the continental crust of the CC domain is deformed against a denser (and probably more rigid) block of the continental crust observed here as the D1 anomaly. As mentioned before, the lineament OL (Figs. 1b, 6 and 8) is continued to the south by a west-dipping reverse fault that limits CC and CD (SERNAGEOMIN, 2003; Melnick and Echtler, 2006; Hackney et al., 2006; Encinas et al., 2021). This is an example of the contractional deformation styles that could be generated in the eastern border of CC by the depth contact between the CC domain and D1. It is interesting to note that the onshore refraction seismic profile ZDO-001, located to the west of D1 (in the CC along the profile P3_Osorno; see location in Fig. 2), shows the inversion of an Oligo-Miocene normal fault, while the seismic profile Z5B-010A (located to the south of P4_Llanquihue profile) presents a minor contractional deformation in the CD sequences, above the D1 (Jordan et al., 2001). Deep below the CC, Maksymowicz et al. (2021) shows seismic reflectors at deep crustal levels with east-dipping angles, which is consistent with the geometry of the western border of D1 (Fig. 7) and supports a structural relation between the metamorphic complexes (WS/ES) and D1. However, the resolution of the density model is not enough to calculate the precise inclination of the western border of D1, and this structural relation is only suggested by the results.

In the northern profile P1_Toltén, D1 correlates with the high Vp and low Vp/Vs anomalies obtained by Dzierma et al. (2012b). Inside the region highlighted by white contours in Fig. 4b, these authors show Vp/Vs values lower than 1.74, contrasting with values higher than 1.78, eastward and westward. In same region Vp and Vs models reach values ∼ 4 % and ∼ 8 % higher than surrounding regions, respectively. Then, at least at the profile P1_Toltén, the correlation between D1 anomaly and the change in elastic properties is clearly observed. Considering this Vs velocity anomaly and an increase in density of about 0.05 g cm-3 (associated with the D1 anomaly, Figs. 4a and 7), we estimated an increase in shear modulus of the order of 20 % in comparison to the surrounding regions (at the same depth). To the south, this seismic velocity anomaly shows a clear continuity with the D1 geometry noticed in the 3D density model (Fig. 8b). This continuity indicates that D1 is a primary characteristic of the continental crust, southward from ∼ 39∘ S, and this supports the interpretation of D1 as a dense–rigid zone. The latitudinal analysis of these independent geophysical models establishes that those basement units associated with D1 are progressively shifted to the east (and taken away from the trench), northward from 41.5∘ S (Fig. 8b). In other words, the portion of the continental wedge formed by the MWU and CC domain is ∼ 50 km wider at 39.5∘ S compared to that observed at 42∘ S.

Outcrops of the late Paleozoic batholith near ∼ 40∘ S are observed in the western border of WS/ES and D1 anomaly, possible implying that D1 is the southward continuation of the late Paleozoic batholith. However, outcrops of this Batholith are described at ∼ 40.3∘ S (Deckart et al., 2014) and ∼ 42.5∘ S (SERNAGEOMIN, 2003) near to a volcanic arc, indicating a possible association of this unit with the D2 anomaly. In this case, south of 39∘ S, D1 should be a high-density basement unit located westward from the late Paleozoic (Pennsylvanian) batholith. An interesting candidate to fit these conditions is Chaitenia terrane (Ct in Fig. 1b), which is described as an island arc, accreted to the Gondwana margin during late Devonian times (Hervé et al., 2016, 2018). The northward limit of D1 (high Vp and Low Vp/Vs anomaly, Dzierma et al., 2012b) can be roughly defined by the MVFZ. This structure could be interpreted as a limit between Chaitenia and Chilenia terrane to the north (Fig. 8b). This interpretation raises an interesting question about the role of the Chaitenia–Chilenia limit in the observed westward shift of the late Paleozoic batholith southward of 38∘ S and its relationship with the continental deformation generated by the kinematics of LOFS (Cembrano et al., 1996; Melnick et al., 2009; Geersen et al., 2011). On the other hand, Plissard et al. (2019) observed that outcrops of mafic and ultramafic (serpentinites) rocks associated with WS (south of 39∘ S) show P–T patterns (and structural characteristics) that allow these units to be interpreted as rock located below an incipient back-arc basin during Devonian times (380–370 Ma), which were incorporated into the subduction channel, reaching depths of about 60 km downward in the interpolate boundary (during Carboniferous time), to be finally exhumed in the eastern border of an accretionary wedge during Permian times. Under this interpretation the eastern border of the Devonian island arc (associated with Chaitenia terrane by Hervé et al., 2016, 2018) corresponds to an incipient back-arc, rather than a subduction zone, but the process finally ends in the accretion of the Devonian island arc to the Gondwana margin. Again, these accreted units could be related to the D1 anomaly in the region.

Beyond the lithology and age of D1 and D2 anomalies it is necessary to highlight the spatial association between the active volcanism and the main lineaments of the LOFS (Lara and Folguera, 2006; Sánchez et al., 2013; Díaz et al., 2020). Figure 6 shows that most of the quaternary volcanoes are located above the local regions of relatively low-density contrast (in general < 0.0 g cm-3 below 5 km depth) inside D2. This local 3D pattern of density anomalies is not easy to interpret in 2D regional models, because they are averaging the density structure around the profiles, and shows that the 3D local inversion is a relevant methodology (complementary to 2D regional analysis) to observe medium-depth and shallow-density structures in the upper crust. These local regions of relatively low-density contrast could respond to more fractured regions of the upper crust as a response of deep structures associated with branches of LOFS and other continental structures presents below the CC and CD. In fact, relatively low-density zones may be related to active volcanic processes observed along this fault system in the Araucanía, Los Ríos and Los Lagos districts. As shown by Díaz et al. (2020), relatively low electric resistivity is found at depths between 7 and 15 km below the local trace of the LOFS, east of Osorno volcano, associated in this case with a zone of partial melt related to a deeper ascent of basaltic magmas enhanced by the LOFS, and therefore a lower density compared to its surroundings.

The upward migration of magmas should generate local weakening zones in the overriding plate, and consequently, the continental crust in the active volcanic zone should present pervasive fracturing, fluid migration and lower density. Hence, the basement extended to the east of D1 could correspond to a similar lithology but be affected by the pervasive fracturing and fluid migration processes associated with an active volcanic arc and LOFS. This interpretation is supported by the increase in Vp/Vs values to the east of D1 (Dzierma et al., 2012b), at least at the profile P1_Toltén (Fig. 4a). On the other hand, Kapinos et al. (2016) and Segovia el al. (2021) describe electrically conductive anomalies eastward from LOFS and high resistivity values beneath the CD basin (Fig. 4b and e). To the west of D1, these authors also establish conductive anomalies associated with the CC domain, reinforcing the interpretation of a transition from a highly deformed and fractured basement (related to deep units of WS/ES) to a denser/rigid basement below CD.

It is already known that the rupture propagation during large earthquakes, the interseismic deformation (including aftershocks and foreshocks), and the interplate locking are complex processes that depend primary on the frictional properties at the interplate boundary (subduction channel) and the stress field evolution (Scholz, 1998; Perfettini and Avouac, 2004; Tassara, 2010; Moreno et al., 2018; Im et al., 2020). As mentioned before, the segment of the continental wedge that includes MWU and CC domains, i.e. fractured and/or metamorphic basement units, is progressively wider to the north of 42∘ S. This structural change correlates with the patch of high coseismic slip of the 1960 Mw9.6 Valdivia earthquake (Fig. 8b), which added to the correlation with gravity anomaly L1 (Fig. 3a) and with changes in slope morphology, suggesting a link between the megathrust seismotectonics and physical properties of the overriding plate. In this regard, we propose that the MWU and CC domains correspond to a portion of the continental plate displaying a higher elastic and permanent deformation compared to the rigid basement landward (Chaitenia/Chilenia). Consequently, the change in the horizontal extension of this unit should modify the process of stress loading during the interseismic periods.

Due to the relatively scarce seismological data in the area and the long recurrence time for large events, it is difficult to conceptualize the complete seismotectonic story of the study zone. Nevertheless, some observations seem to support our hypothesis. Firstly, the rupture zone and aftershocks (including continental intraplate events) of the Mw 7.6 earthquake occurred in 2016 at 43.5∘ S. This event was the largest since 1960 in the rupture area of the Valdivia earthquake (Moreno et al., 2018; Lange et al., 2018) was located at the base and within the CC domain, in the western border of a high Vp-low Vp/Vs anomaly (Lange, 2008). This velocity anomaly is a clear continuation of D1 to the south of the studied area. On the other hand, historical (not instrumentally recorded) megathrusts events activated this segment of the margin in 1737 (Mw∼7.5) and 1837 (Mw∼8). They have been associated with ruptures extended to the south of ∼ 39∘ S (Kelleher, 1974; Lomnitz, 2004), indicating that the northern portion of 1960 Valdivia earthquake could have different mechanical properties.

Lithology and internal deformation style inside and at the base of MWU and CC domains can play a different but complementary physical role in the seismotectonic segmentation of the margin. The high slip patch of Valdivia earthquake also correlates with the segment where the geometry of the marine continental wedge (seaward from shelf break) is consistent with a decrease in the effective friction coefficient (μb∗) at the interplate boundary (Maksymowicz et al., 2015). This suggests oversaturated fluid conditions in the subduction channel, at least in the western portion MWU at the study zone. At the same time, according to Menant et al. (2019), the deformation style of basal accretionary complexes (typically an antiformal stack of duplexes) favoured upward fluid fluxes from the interplate boundary, generating dewatering and the increase in μb∗ in some adjacent regions of the subduction channel (mainly downward from basal accretionary complex). Several authors have suggested the presence of this deformation style in the deep zone of WS units (Krawczyk et al., 2006; Ramos et al., 2018; Moreno et al., 2018; Maksymowicz et al., 2021). Under this interpretation, the widening of MWU and CC domains to the north of ∼ 42∘ S could favour high friction in the deep region of the interplate boundary (below CC domain) and a relatively low friction in the seaward portion of MWU. Therefore, the position and horizontal extension of the WS could be linked to changes in the frictional properties along the megathrust. However, more studies should be done to explore the seaward limit of WS/ES and the internal structure of MWU and CC domain.

2D and 3D density models of the forearc show a landward and latitudinal segmentation of the continental wedge in the studied zone. Offshore, the active accretionary prism limit, with a more competent basement below the middle wedge and shelf, exhibits a landward increase in density, probably associated with a progressive decrease in fracturing. To the east, the Coastal Cordillera domain presents an increase in the upper crust densities but reaches lower values than those observed in the high-density anomaly below the Central Depression. Northward from ∼ 42∘ S, this high-density anomaly is seen progressively further from the trench, determining a northward widening of the middle wedge and Coastal Cordillera. This feature correlates with the high slip patch of the giant 1960 Mw9.6 Valdivia earthquake.

Based on geological information, we associate the middle wedge unit (at least its eastern portion) and Coastal Cordillera domain with the late Paleozoic–early Mesozoic accretionary complex, and the high-density anomaly below the Central Depression as geophysical evidence of Chaitenia terrane. The deformation style at the eastern border of the Coastal Cordillera and seismological studies support the hypothesis of a more rigid behaviour of the continental crust below the Central Depression. Accordingly, we propose that changes in the horizontal extension of the middle wedge unit and Coastal Cordillera domain should have modified the process of stress loading during the interseismic periods, and that changes in position and extension of the late Paleozoic–early Mesozoic accretionary complex could be linked to the frictional properties of the interplate boundary.

Our results highlight the role of the overriding plate structure in the seismotectonics process in subduction zones, but more studies are necessary to understand the changes in physical properties (elasticity, temperature, among others) associated with the geological story of the margin. This work motivates similar analysis of the continental basement in other subduction margins, as in the 2010 Mw 8.8 Maule earthquake and Mw 9.0 Tohoku-Oki rupture zones.