A grid of closely spaced, high-resolution multichannel seismic (MCS) reflection profiles was acquired in May 2012 over the outer accretionary prism up dip from the patch of greatest slip during the 2010 Mw 8.8 Maule earthquake (offshore Chile) to complement a natural-source seismic experiment designed to monitor the post-earthquake response of the outer accretionary prism. We describe the MCS data and discuss the implications for the response of the accretionary prism during the earthquake and for the long-term evolution of the margin. The most notable observation from the seismic reflection survey is a rapid north-to-south shift over a short distance from nearly total frontal accretion of the trench sediments to nearly total underthrusting of undeformed trench sediments that occurs near the northern edge of slip in the 2010 earthquake. Integrating our structural observations with other geological and geophysical observations, we conclude that sediment subduction beneath a shallow décollement is associated with propagation of slip to the trench during great earthquakes in this region. The lack of resolvable compressive deformation in the trench sediment along this segment of the margin indicates that the plate boundary here is very weak, which allowed the outer prism to shift seaward during the earthquake, driven by large slip down dip. The abrupt shift from sediment subduction to frontal accretion indicates a stepdown in the plate boundary fault, similar to the stepovers that commonly arrest slip propagation in strike-slip faults. We do not detect any variation along strike in the thickness or reflective character of the trench sediments adjacent to the change in deformation front structure. This change, however, is correlated with variations in the morphology and structure of the accretionary prism that extend as far as 40 km landward of the deformation front. We speculate that forearc structural heterogeneity is the result of subduction of an anomalously shallow or rough portion of plate that interacted with and deformed the overlying plate and is now deeply buried. This study highlights need for three-dimensional structural images to understand the interaction between geology and slip during subduction zone earthquakes.

The Tohoku earthquake (offshore Japan) in 2011 generated a devastating tsunami as slip extended to the subduction trench, and the structural signature of this process was captured in a remarkable pair of “before” and “after” seismic reflection images (Kodaira et al., 2012). Since then, a number of investigators have studied the structural characteristics of other subduction zones to try to infer tsunamigenic potential from the seismic reflection signature of the deformation front (e.g., Dean et al., 2010; Gulick et al., 2011; Cubas et al., 2016; Bécel et al., 2017; Han et al., 2017). As part of an experiment to monitor potential post-seismic deformation of the outer accretionary prism up dip from the patch of greatest slip during the 2010 Mw 8.8 Maule earthquake (offshore Chile) (Tréhu and Tryon, 2012), we acquired 1500 km of high-resolution multichannel seismic reflection data using a 600-m-long, 48-channel hydrophone streamer and two Generator-Injector (GI) guns in 45/105 mode (Fig. 1), referring to the volume (in cubic inches) of the generator and injector airguns. We also acquired coincident swath bathymetric, 3.5 kHz subbottom profiling, and gravity data. Although many of the early slip models for the Maule earthquake indicated that slip did not extend to the trench (e.g., Moreno et al., 2010; Delouis et al., 2010; Tong et al., 2010; Lorito et al., 2011; Vigny et al., 2011), the up-dip extent of slip is poorly constrained in most models, and recent studies suggest that slip may have locally reached the trench (Yue et al., 2014; Maksymowicz et al., 2017; Wang et al., 2017).

In this paper, we examine the seismic reflection data collected over the trench and accretionary prism north of the epicenter and near the northern boundary of slip from the Maule earthquake to evaluate whether along-strike variations in the structure of the deformation front can be related to variations in the response of the outer prism to plate boundary slip on both short and long time scales. Most models indicate that the largest slip during the earthquake occurred in this region. We document the presence of an along-strike transition over a short distance from nearly total sediment accretion to dominantly sediment subduction at the deformation front, and look both seaward and landward of the deformation front for possible mechanisms to explain this variability. We do not detect any systematic changes in the reflectivity structure or total thickness of the trench sediments that can explain this observation. Landward of the deformation front, we provide evidence for a transpressive boundary that separates the active accretionary prism from the metamorphic rocks of the paleo–accretionary prism, which acts as a backstop to subduction here (Moscoso et al., 2011; Contreras-Reyes et al., 2017) and note variations in prism and backstop morphology and structure that are correlated with the along-strike change in deformation front structure.

Our ultimate objectives are to explore whether structural patterns in the outer prism preserve a signature of consistent, long-term patterns of strain release and to investigate the implications of such observations for seismic and tsunami hazard evaluation. Integrating our observations with prior results from bathymetric, seismic, potential field, geodetic, and coastal uplift studies (e.g., Rietbrock et al., 2012; Lange et al., 2012; Métois et al., 2012; Hayes et al., 2013; Cubas et al., 2013; Lieser et al., 2014; de Moor, 2015; Maksymowicz et al., 2015, 2017; Bassett and Watts, 2015a, 2015b; Saillard et al., 2017), we conclude that sediment subduction is associated with up-dip propagation of slip during earthquakes in this region and speculate that the correlation between the along-strike transition in deformation front structure and the accretionary prism morphology results from subduction of a particularly shallow or rough, and now completely buried, part of the oceanic Nazca plate.

Plate Tectonic Setting

The south-central Chilean margin (32°–46°S) is characterized by the subduction of the oceanic Nazca plate beneath South America at a present rate of ∼66 mm/yr (Angermann et al., 1999) in a N78°E direction. This rate, determined from satellite data, is slower than the global plate model rate of ∼85 cm/yr (DeMets et al., 2010), which represents an average over several million years, suggesting a recent decrease in the convergence rate. Nonetheless, this rate is fast enough to drive a high level of historic interplate seismogenic slip. The Maule earthquake filled a well-documented seismic gap that had not ruptured in a major earthquake since A.D. 1835 (Campos et al., 2002; Lomnitz, 2004; Ruegg et al., 2009; Ruiz and Madariaga, 2018). Because of the length of the Peru-Chile trench and the systematic along-strike variation in parameters such as the age of the subducting plate, sediment thickness, and incoming plate roughness, many studies of this region have attempted to isolate the dominant parameters affecting the subduction process and associated seismogenesis (e.g., Contreras-Reyes and Osses, 2010; Contreras-Reyes et al., 2010; Contreras-Reyes and Carrizo, 2011).

Sedimentation in the Trench

Our seismic survey, acquired as part of ChilePEPPER (Project Exploring Prism Post-Earthquake Response; Tréhu and Tryon, 2012), was located just south of the intersection of the Juan Fernández Ridge with the trench at ∼33°S; this ridge acts as a barrier to northward transport of trench turbidites (Fig. 1), resulting in a sediment-starved trench to the north and a sediment-flooded trench to the south (von Huene et al., 1997). The high sedimentation rate in the trench between the Juan Fernández Ridge and the Chile triple junction since the Pliocene is linked to glaciation-deglaciation and rapid denudation of the Andes (e.g., Bangs and Cande, 1997; Rauch, 2005; Melnick and Echtler, 2006; Völker et al., 2006). Sediment is delivered from the continent to the trench through deep canyons and redistributed within the trench from south to north (Thornburg et al., 1990; Völker et al., 2008, 2013). Numerous submarine canyons cut across the continental shelf and slope and are offshore prolongations of the main rivers of Pleistocene glacial valleys (Gonzalez, 1990). Sediment fans are also commonly found where the canyons enter the trench (Thornburg et al., 1990; Völker et al., 2006).

Rauch (2005) estimated that the age of the current trench is at most 600 ka by comparing the progressive onlap of sediments on the western flank of the trench to the subduction rate and assuming a steady rate of trench fill and subduction. He documented a repeating pattern of high- and low-amplitude reflections in the trench, which he estimated to have been periodically deposited every ∼120 k.y. by cross-correlating reflection data with the temperature and CO2 concentration history of Earth’s atmosphere derived from ice cores. Völker et al. (2013) argued that during glacial periods, along-trench sediment transport from source regions in the glaciated south is dominant, whereas during interglacial periods, sediment input is dominated by input through the rivers and canyons and by slope instability, which recycles sediment from the accretionary prism back into the trench.

A striking feature of the Chile trench is its axial channel, which extends continuously for ∼1500 km, from ∼41°S to the northern limit of the filled trench near 32.75°S. It is interpreted to have initiated in the Pleistocene (Völker et al., 2006). Within the trench, the channel is narrow and deeply incised in places and broader in others as it meanders between the western and eastern boundaries of the trench (Fig. 1). In deeply incised areas of the trench, a thin layer of sediment is sometimes imaged filling the floor of the channel, suggesting episodic sedimentation activity. From 38°S to 40°S, a younger channel can be observed in the floor of the channel. Where deflected by sedimentary fans or slump deposits, the channel is commonly wider and more poorly defined. Sinuosity of the trench is anomalously high in the region of our study compared to segments to the north or south and suggests that tectonics influences channel dynamics.

Tectonics of the Accretionary Prism

The tectonics of the south-central Chilean forearc have been discussed by a number of investigators, including Bangs and Cande (1997), Contardo et al. (2008), Geersen et al. (2011), Moscoso et al. (2011), and Contreras-Reyes et al. (2017). Based on several seismic lines across the margin between 36°S and 40°S acquired from the R/V Conrad (Fig. 1), Bangs and Cande (1997) concluded that sediment accretion, nonaccretion, and subduction erosion along the margin were episodic and controlled by climatic as well as tectonic factors, with sediment accretion occurring during glacial periods when the rate of sediment input to the trench is high.

Contardo et al. (2008) analyzed forearc structure between 33.5°S and 36.5°S based on seismic reflection lines acquired by the R/V Vidal Gormaz using an airgun source in a region that overlaps the region covered by the ChilePEPPER reflection survey (Fig. 1). They described the slope basins as asymmetric half-grabens of variable size interpreted to result locally from subducted seamounts passing beneath the forearc and more generally from underplating of sediment packages beneath the margin, which results in uplift and tilting of the prism. The R/V Vidal Gormaz profiles stopped at the deformation front and did not cross the trench.

Geersen et al. (2011) focused on the forearc from 35°S to 40°S using an extensive swath bathymetric data set and seismic reflection profiles acquired as part of the SPOC (Subduction Processes Off Chile) project and identified four different tectonic segments, each characterized by a distinct morphology. In the northern and southern Concepción segments, located south of our field area (Fig. 1), they interpreted several thrust ridges interpreted to be the seafloor manifestation of splay faults accommodating some of the Nazca–South America plate motion. Like Contardo et al. (2008), they identified normal faults and tilted basins that they attributed to gravitationally driven extension driven by uplift in response to sediment underplating.

Deep Structure of the Forearc

The deeper structure of the forearc in our study region has been determined through tomographic inversion of travel times from marine sources recorded on ocean-bottom and onshore seismometers. Prior studies of the forearc crustal structure in our study region (Moscoso et al., 2011; Contreras-Reyes et al., 2017) indicate the presence of an outer prism with a P-wave velocity <3 km/s formed of accreted trench sediments; a middle prism formed of older accreted sediments (P-wave velocity ∼4 km/s) overlain by an apron of low velocity slope sediments; and continental framework rock with a velocity of >5 km/s (Fig. 2). The middle prism is a region of potential growth by underplating and internal deformation, resulting in the half-grabens and thrust ridges discussed in the previous section. The continental framework in this region is thought to represent a late Paleozoic subduction complex that has been metamorphosed in a low-temperature and high-pressure environment and later uplifted and is paired with exhumed roots of a magmatic arc onshore (Willner, 2005). Near the seafloor, the boundary between the middle prism and continental framework rocks is interpreted to coincide with the shelf break, which is characterized by a prominent fault scarp up to 2 km high. Contreras-Reyes et al. (2017) pointed out that the patch of greatest slip during the Maule earthquake occurred where the distance between the deformation front and the shelf edge is greatest (34.0°S and 35.5°S). They noted a similar relationship for the slip distribution in the great Chilean earthquake of A.D. 1960, which overlapped with the southern end of the Maule rupture and extended from 38°S to 45°S. Saillard et al. (2017) made a similar observation and concluded that regions of coastal uplift were correlated with creeping segments of the plate boundary.

Slip Distribution during the 2010 Maule Earthquake and Aftershock Distribution

Many different slip models have been published for the 27 February 2010 Mw 8.8 Maule earthquake offshore south-central Chile (e.g., Moreno et al., 2010; Lay et al., 2010; Delouis et al., 2010; Tong et al., 2010; Lorito et al., 2011; Vigny et al., 2011; Kiser and Ishii, 2011; Lin et al., 2013; Yue et al., 2014; Wang et al., 2017). While most indicate that the patch of greatest slip occurred northwest of the epicenter (Fig. 3), the cross-strike location of greatest slip, the relationship between slip during the main shock and aftershock activity, and whether slip propagated to the trench or was arrested beneath the outer accretionary prism vary among the models. This is due, at least in part, to different data sets used to infer slip and the lack of data to resolve coseismic seafloor deformation offshore. Resolving these details may shed light on the conditions under which coseismic slip can propagate beneath or through the accretionary prism and the implications for tsunamigenesis.

Sladen and Trevisan (2018) recently documented a global correlation between outer-rise aftershocks and apparent slip to the trench during megathrust earthquakes. The Maule earthquake was followed by pronounced outer-rise seismicity from 33.7°S to 35.0°S (Fig. 3) including a normal-faulting event with Mw 7.4 (Ruiz and Contreras-Reyes, 2015). Very few earthquakes were located within or beneath the outer accretionary prism, with apparent exceptions up dip of the epicenter near 36°S and near the northern end of the rupture zone. Although the earthquakes in Figure 3 were located using only onshore seismic stations, lack of aftershock activity within and beneath the outer prism has been supported by several ocean-bottom seismometer studies (Lieser et al., 2014; Hicks et al., 2014; de Moor, 2015). That the outer prism was seismically quiet after the main shock has been variably interpreted to indicate that this was the region of greatest slip during the mainshock, with a down-dip gradient of increasing seismicity marking the down-dip edge of slip during the main shock (e.g., Rietbrock et al., 2012), or that this region did not slip during the earthquake because of a velocity-strengthening rheology, implying that it should have deformed aseismically after the main shock (e.g., Lange et al., 2012). In the latter case, one might expect to observe very-low-frequency (VLF) earthquakes and non-volcanic tremor (NVT), as has been reported from the outer prism in the Nankai Trough (offshore Japan) (e,g., Obara and Ito, 2005). However, with the exception of a few possible small slow-slip events and tremor episodes (Tréhu et al., 2019), only limited evidence for VLF or NVT activity (compared, e.g., to the Nankai Trough) was found in broadband ocean-bottom data acquired during ChilePEPPER (de Moor, 2015), a negative result consistent with an interpretation in which the outermost prism slipped during the earthquake, releasing rather than increasing stress in the outer prism. Additional evidence that coseismic and/or rapid postseismic slip extended to the trench is supplied by comparing swath bathymetry acquired in 2008 and bathymetry along the same track acquired in 2011 and during cruise MV1206 on the R/V Melville, which resolved ∼5 m of uplift of the outer prism between 2008 and 2011 (Maksymowicz et al., 2017). Whether slip extended to the seafloor at the deformation front or was arrested within 6 km of the deformation front is not conclusively resolved by the bathymetric change data.

Multichannel seismic (MCS) data were acquired by the Scripps Institution of Oceanography (SIO) Shipboard Geophysical Group (SGG) using a 600 m, 48-channel streamer during cruise MV1206 on the R/V Melville. The number of channels, however, was decreased to 40 soon after the beginning of the cruise because shark bites penetrated two of the eight-channel streamer sections and only one spare section was available. The source was two GI-guns in a 45/105 cm3 configuration. The guns were mounted on a crossbar that maintained them at constant separation and were fired at an interval of 25 m as determined from GPS. Ship speed through the water was <5 kt to mitigate tow noise. The record length was 9 s for most of the survey, but was increased to 10 s near the end of the survey when the depth of penetration of the seismic energy and thickness of the trench sediments became apparent. Sample rate was 1 ms.

Marine mammal observers were on deck during all seismic acquisitions conducted during daylight hours, and reported numerous sightings during the cruise that required turning off the airguns (Tréhu and Tryon, 2012). When the airguns were shut down for <10 min, we continued along the profile and resumed shooting when the area was clear, leaving gaps in the data. For longer shutdowns, we turned and filled the gap once shooting could resume. In all, there were 110 separate sightings of 265 individual animals during the cruise. The most commonly sighted mammals were fur seals. A number of whales and dolphins were also seen.

Data were processed onboard through frequency-wavenumber migration using SIOSEIS software (https://sioseis.ucsd.edu). The processing sequence included sorting, normal moveout (NMO), stack, and frequency-wavenumber migration. A velocity of 1485 m/s was used for both NMO and migration. In general, each line was processed within a few hours of acquisition. Preliminary interpretations of the data guided subsequent acquisition. Seismic sections were input into a Kingdom Suite (https://ihsmarkit.com/products/kingdom-seismic-geological-interpretation-software.html) project while at sea for interpretation and integration with bathymetric and potential-field data. Data images shown here were generated either with Kingdom Suite or Seismic Unix (https://github.com/JohnWStockwellJr/SeisUnix).

Figure 4 shows the location of the ChilePEPPER seismic lines overlain on swath bathymetry. Because of the slow ship speed and good weather conditions during most of the cruise, data quality is excellent. The data were acquired in a grid of lines that covered the extent of ocean-bottom seismometers, absolute-pressure gauges, and seafloor flowmeters deployed to detect any local earthquakes, NVT, VLF earthquakes, seafloor uplift or subsidence, and diffuse fluid flow through the seafloor (Tréhu and Tryon, 2012). Additional profiles were acquired along the trench and across the deformation front. Although the spatial footprint of this survey is small, the lines are more closely spaced than those of other MCS surveys conducted to date along the Chile margin (Fig. 1), providing a unique opportunity to understand structural relationships in the accretionary prism in three dimensions along the segment of the margin that experienced the largest slip during the 2010 Maule earthquake.

Notable bathymetric features shown on Figure 4 include: normal faults nearly perpendicular to the trench, reflecting seafloor spreading fabric formed at the East Pacific Rise; horst-and-graben structure oriented slightly oblique to the trench resulting from plate bending; the Bucalemu seamount, which is cut by normal faults; many smaller seamounts; an axial channel that meanders from side to side of the sediment-filled trench; variations along strike in the development of protothrusts seaward of the primary deformation front; a distinct embayment in the deformation front; a thrust ridge identified by Geersen et al. (2011); and several deeply incised canyons. The Maule and Mataquito Canyons can be linked to major rivers onshore, similar to the Bio Bio Canyon, although high-resolution bathymetry is not available from the nearshore region to confirm this. In contrast, the Huenchullami Canyon is associated with a much smaller river that originates in the Chilean Coastal Range and ends abruptly ∼30 km east of the trench. Additional bathymetric features within the study region will be discussed in the next section in conjunction with their subsurface structure as imaged in the seismic data.

Figures 5A–5C show three representative seismic lines located ∼15 km apart. Detailed images of portions of these and other lines are discussed in this section, moving from west to east across the trench, deformation front, and accretionary prism. Raw data, migrated stacks, and images of each seismic line are available through the Academic Seismic Portal maintained by the University of Texas Institute for Geophysics (http://www-udc.ig.utexas.edu/sdc/). Line CP06 is coincident with large-aperture seismic profile P03, which was modeled and discussed by Moscoso et al. (2011) and is the basis for the simplified cross-section shown in Figure 2. It is also coincident with the bathymetric swath used by Maksymowicz et al. (2017) to infer uplift during or soon after the Maule earthquake. Figure 5D illustrates the along-strike topographic variability that characterizes this segment of the margin. Note the striking difference in the slope of the outer prism between profile CP19 and the other two profiles. The reflection character and topography of the middle prism also varies significantly along strike, as does the topographic boundary between the middle prism and the continental framework.

Lower Plate Structure and Trench Fill

Figure 6 shows seismic lines along and across the trench. The outer rise seaward of the trench along this segment exhibited a high level of aftershock activity in response to the 2010 Maule earthquake (Fig. 3). Outer-rise faulting here is approximately perpendicular to faulting that reflects the crustal fabric formed at the spreading center and is oblique to the trench (Fig. 4). The incoming plate is also marked by the presence of numerous seamounts of various sizes, including a large seamount west of the trench near 34°28.233ʹS, 73°52.457ʹW that is clearly cut by two outer-rise bending faults, forming a graben at the crest of the seamount (Fig. 4). The seafloor appears to be capped by a thin layer of pelagic sediment of variable thickness that follows the basement and is overlain by trench fill (e.g., seismic line CP11, Fig. 6). This layer can be clearly identified only when onlapped by trench sediments. Some of the basement blocks appear to also uplift trench sediments (e.g., line CP24, Fig. 6). This may explain the anomalously narrow width of the trench and apparent decrease in trench sediment volume estimated by Völker et al. (2013) along this segment of the margin because they included in their estimate only sediment east of the western boundary of the trench as defined by topographic contours.

Outer-rise normal faults and the resulting horsts, grabens, and tilted blocks show vertical offsets of 500 m or more and extend beneath the trench sediments. Offsets in the trench sediments overlying the basement ridges decrease with decreasing depth beneath the seafloor and have little or no offset at the seafloor, indicating that these faults are active as they continue into the trench but that normal-faulting activity decreases as the faults approach the deformation front (e.g., lines CP10 and CP24, Fig. 6). Although it is difficult to unambiguously trace individual ridges between seismic lines, even with this relatively dense line spacing, we identify two en echelon basement ridges, labeled BR1 and BR2 in Figures 4 and 6. On some profiles (e.g., lines CP10 and CP24, Fig. 6), BR1 appears as a tilted block bounded on its northwestern flank by a normal fault, whereas on line CP08a (Fig. 6), it appears as a fault-bounded horst. On line CP06 (Fig. 6), a composite basement high defined by an envelope of diffraction hyperbolae indicative of a rough surface adjacent to a small fault-bounded block is observed were we expect to see BR1. No basement high is observed on line CP21 (Fig. 4) at the corresponding position. We interpret the observations on lines CP06 and CP21 to indicate that the amplitude of the tectonic ridge identified as BR1 decreases to the northeast and is coincident with a small buried seamount at its northeastern terminus. A larger buried seamount overlain by a thin layer of pelagic sediment unconformably overlain by trench fill sediments is present at the northernmost end of the composite trench-parallel line (Fig. 6).

A distinctive feature of the study region is a thick shallow wedge (TSW) of sediment deposited immediately south of where the Maule canyon enters the trench. The ChilePEPPER data (Fig. 6) reveal the internal stratigraphy and thickness of this deposit and the erosional character of its western margin. The thickness of this deposit is shown in map view in Figure 7A, and the sediment thickness at the deformation front is shown in Figure 7B. Similar deposits have been documented associated with the intersection of other major canyons and the Chile trench (Thornburg and Kulm, 1987; Thornburg et al., 1990; Völker et al., 2006) and reflect complex interactions between sediment transported along the axis of the trench and sediment supplied locally to the trench from the Maule canyon.

The TSW has a strong effect on the depth and width of the axial channel. On seismic line CP11 (Fig. 6), it is ∼100 m deep and ∼1500 m wide with a flat floor and steep sides; no paleo-channel is observed in the trench sediments. Going north, the channel appears to be deflected to the west by the sediment lobe and is wide and shallow (lines CP25 and CP24, Fig. 6). The channel then swings abruptly to the east around the northern edge of the sediment wedge and is located at or near the deformation front (lines CP06 and CP08a, Fig. 6) before swinging back to the western edge of the channel, where it is again clearly incised into the trench floor (line CP18a, not shown). The competing effects of the TSW and tectonic uplift of the western edge of the trench may be responsible for the high sinuosity and indistinct cross-section of the axial channel along this segment of the trench.

Basement age in the study region is ca. 34 Ma. The closest cores available to constrain the age of the trench sediments is Ocean Drilling Program (ODP) Site 1232 (Mix et al., 2003), located near the western edge of trench near 40°S. At ODP Site 1232, ∼390 m of turbidites was recovered. Although porosity decreased rapidly from ∼90% to 55% in the upper 15 m of the core, the porosity was nearly constant below 15 m, indicating underconsolidation and high pore pressure. Paleomagnetic measurements indicated that this entire section was emplaced within the Bruhnes Chron (<0.78 Ma), indicating a sedimentation rate of at least 475 m/m.y. (Mix et al., 2003). Assuming this rate, age of the oldest trench sediments would be ca. 5 Ma. However, it is likely that the sedimentation rate for the TSW is significantly higher. If one assumes that the rate of 475 m/m.y. applies to sediments north of the TSW, the age of the oldest sediments would be ca. 3.4 Ma.

The Deformation Front

Figures 810 show examples of data crossing the deformation front. On line CP19 at the northern end of the survey (Fig. 8), the deformation front is characterized by a seaward-verging blind thrust fault that extends from the seafloor to near the base of the trench sediment. Similar structures are seen in SPOC line SO161-42 at 37°S (Völker et al., 2013). The offset in time on this fault decreases from ∼0.050 s at point A to 0.038 s at B to 0.025 s at C (Fig. 8). Assuming a velocity of 1800 m/s near the surface, increasing to 3600 m/s near the base of the sediment pile due to sediment compaction (Moscoso et al., 2011), the observations suggest a constant offset of ∼45 m on this fault, in contrast to the decreasing offset with decreasing depth observed for the normal faults in the trench. Whether this fault was developed in a single earthquake, in many small events, or through aseismic creep cannot be resolved from cruise MV1610 on the R/V Roger Revelle data, although the constant offset along the fault and seafloor manifestation of a frontal fold indicate that it was initiated relatively recently.

Landward of the frontal thrust, the outer 6 km of the accretionary prism on line CP19 (Fig. 8) appears to be formed primarily by folding of the entire sediment layer, and the top of the subducted oceanic crust is observed to ∼2 s two-way travel time (TWTT) (∼3 km assuming an average velocity of 3 km/s for the outer accretionary prism; Moscoso et al., 2011). In contrast, the R/V Vidal Gormaz data from line VG02-03 (Contardo et al., 2008), which is nearly coincident with line CP19 (Fig. 1), did not extend far enough west to image the frontal thrust fault; because line VG02-03 was processed only to 8.2 s TWTT, it also does not image the top of the subducting oceanic crust beneath the outer accretionary prism (Fig. 7). Similar structures are observed on line CP23 and on line VG02-02, which crossed the deformation front ∼30 km north of line CP19 (Fig. S1 in the Supplemental Material1).

In contrast to line CP19, lines CP21 and CP06 (located 10–15 km south of CP19) are characterized by a seaward frontal thrust that reaches the seafloor and subduction of all of the incoming trench sediment (Fig. 9; see Fig. S2 [footnote 1] for an uninterpreted version of this figure). The frontal thrust cuts through the upper 0.5 s TWTT (∼500 m) of the trench sediments before flattening and apparently following a stratigraphically controlled surface, or décollement, located ∼0.9 s TWTT (∼1 km) above the crust of the subducting plate. An unconformity between the uppermost subducted sediment and the décollement suggests that some of the subducted sediment is underplated to the base of the outer prism (Fig. 9A).

The décollement reflection and top of the subducted crust can be followed until ∼15 km landward of the deformation front. The position of the plate boundary, ratio of subducted to accreted sediment, and presence or absence of deformation within the subducted sediments deeper within the subduction zone are not constrained by these data. A possible earlier frontal thrust and décollement are observed ∼2.5 km landward of the deformation front (Fig. 9), suggesting a cyclical pattern of overthrusting and accretion at the deformation front as interpreted by Gutscher et al. (1998) for the Aleutian Islands (Northern Pacific Ocean).

A similar décollement beneath which most of the incoming sediment is subducted with no evidence for protothrusts in the trench sediments is observed on lines CP08a and CP10 (Fig. 6). The absence of protothrusts on these profiles indicates a weak and shallow décollement that does not transmit horizontal compressive stress to the trench sediments, in contrast to line CP19 and further north. Analysis of data acquired during cruise MGL1701 on the R/V Marcus Langseth with a large-volume airgun source array and a 15-km-long streamer coincident with line CP06 should provide more detail about the velocity structure and consolidation state of the underthrust sediments and the geometry of the front thrust at greater depth along this segment of the margin (Olsen et al., 2017; Bangs et al., 2017).

South of line CP10, protothrusts seaward of the frontal thrust are observed although, unlike on line CP19, they do not extend to the base of the trench sediments (Fig. 10). Instead, they appear to be confined to the sediments that form the TSW. Most of the trench sediments below the TSW are underthrust in a manner similar to what is observed from lines CP21 to CP10. We conclude that the shallow protothrusts on the southern lines are distinctly different from those on line CP19 and that there is a fundamental change in deformation front structure between the segment of margin crossed by CP19 and the margin segment to the south. The abrupt nature of this transition is evident in the left-lateral offset of the deformation front immediately north of line CP21 (Fig. 4).

The stratigraphic horizon into which the frontal thrust appears to sole on all lines south of line CP19 is observed as a strong, subhorizontal reflection at ∼7.5 s TWTT within the trench (Fig. 6). Figure 11 shows the portion of this horizon between lines CP21 and CP19. Although the event appears to have positive polarity, indicating a velocity increase, it is characterized by a package consisting of several cycles of bright reflections, which suggests interlaying of high- and low-velocity layers. No systematic change in the character of this reflection package is observed between line CP21, where the décollement merges with this horizon, and line CP19, where it is offset by the frontal thrust. The surface into which the frontal thrust on line CP19 soles corresponds to a bright reflection of apparently negative polarity. Although this reflection laps onto basement along the segment of profile shown in Figure 11, it is likely continuous with a reflection seen near the trench on line CP06 (Fig. 9). Why the structure of the deformation front changes so dramatically between lines CP19 and CP21 is not apparent from these data.

Figure 12 summarizes the observations of the deformation front and accretionary prism in map view. Where all of the trench sediment is being subducted, the seafloor slopes to the east and the axial channel is located at the deformation front. South of this segment, the axial channel is deflected to the west by the TSW; north of this segment, where all trench sediment is accreted, the axial channel swings to west side of the trench. The local eastward dip of the trench is likely related to plate flexure as sediment is subducted. Note that the uplifted trench sediment perched on top of outer rise horsts along the western margin of the trench is also restricted to this segment of the trench within our study region.

The Active Accretionary Prism

Figure 5 and Figure S3 (footnote 1) show the transition from the deformation front and to the continental framework on line CP06 with features labeled A–G keyed to bathymetric features labeled on Figure 12. Feature A corresponds to the discontinuous band of landward-dipping reflections tentatively interpreted to be an earlier frontal thrust in Figure 9; B and C are similar bands of landward-dipping reflections in an otherwise incoherent reflective background that can be correlated with small ridges that are parallel to the deformation front. The dashed pink line in Figure 12 delimits the region of the forearc characterized by such structures. This region also includes structures that are perpendicular to the trench-parallel ridges corresponding to A–C. Unfortunately, no seismic lines are oriented properly to image the structure underlying these subtle topographic features—which are more clearly seen in the slope map (Fig. 7)—in cross-section. Given the absence of offsets along the deformation front corresponding to these structures, we tentatively interpret them to be surficial erosional gullies, and note that this region does not show the impact of basement topography, consistent with formation through underplating of sediment subducted beneath a shallow décollement.

The feature labeled D on Figure 5 and Figure S3 (footnote 1) represents a transition from the highly deformed accreted sediments of the outer prism to a region of older accreted sediments that are overlain by distinct slope basins that we refer to as the middle prism. The middle prism is also characterized by a decrease in slope of the seafloor. These basins record a complex history with multiple episodes of extension, tilting, compression, and folding, as previously discussed by Contardo et al. (2008) and shown in Figure 13. The new data presented in this study show the complex, along-strike variation in structure in this region with higher resolution than was previously available. Line CP26 (Fig. 13) crosses a bathymetric ridge that is similar to and in line with the thrust ridges identified by Geersen et al. (2011) further south (Figs. 4, 12). This topographic ridge is bounded by near-vertical faults and resembles a flower structure formed by strike-slip motion in a transpressive stress field. Further north, lines CP15 and CP19 show faulting and folding in the sedimentary fill of the Huenchullami Canyon that extend nearly to the seafloor and also exhibit a flower structure (Fig. 14). Although there are no samples to provide direct age constraints on these structures, they appear to have formed recently and are likely still active.

The two regions where transpressive structures are observed are indicated by solid blue lines on Figure 12; dashed blue lines indicate the extension to the north and south of these structures based on bathymetry (Fig. 4). The offset between these segments is consistent with an extensional stepover. We therefore interpret the structural complexity of the middle prism in our study region to represent a pull-apart basin, outlined by a thin dashed line in Figure 12. The sense of motion implied by this structural analysis is consistent with the oblique plate motion vector (Fig. 1) as well as with bends in the Maule, Huenchullami, and Mataquito Canyons (Figs. 4, 12). At the current plate motion rate (66 mm/yr) and obliquity (30°), the 15-km-long north-south–trending segment of the Huenchullami Canyon that is aligned with an apparent strike-slip fault could have formed over ∼0.5 m.y.

In contrast, the middle prism in the segment of the margin crossed by line CP19 is generally deeper (Fig. 5D) and exhibits pronounced along- and across-strike bathymetric variability. This portion of the accretionary prism is located northwest of the solid pink line in Figure 12; local highs and lows within this apparently collapsed part of the middle prism are outlined by dotted lines labeled H and L. Unlike the Maule Canyon, which cuts across the entire margin to reach the trench, the Huenchullami Canyon has no bathymetric expression in this region. Further north, the Mataquito channel also does not reach the trench. We infer that this part of the prism accumulated through accretion of trench sediment along a décollement located near the top of the oceanic crust, and that the bathymetric heterogeneity in this region results from the sensitivity of this process to basement topography.

Evidence for Gas Hydrates

A bottom-simulating reflection (BSR) of variable amplitude is discontinuous but widespread throughout the accretionary prism in this region (Figs. 13, 15; Figs. S3, S5 [footnote 1]), indicating the presence of free gas underlying gas hydrate. However, no indications of significant fluid venting to the seafloor (e.g., gas chimneys in the subsurface, gas plumes in the water column, or seafloor mounds and bright spots) have been observed, in contrast to venting observed further south along the margin between 35°S and 37°S (e.g., Klaucke et al., 2012). The distribution of the BSR provides additional insights into the tectonic history of this segment of the margin. The seaward extent of BSR observations is indicated as a dashed black line in Figure 12. No BSRs are observed near the deformation front, similar to the central Cascadia margin (offshore western North America) (Phrampus et al., 2017). As in Cascadia, the observation of a gap between the deformation front and the onset of a BSR along this portion of the Chile margin is correlated with subduction of a significant fraction of the incoming sediment. We speculate that subduction of trench sediment delays upward migration of pore fluids into the gas hydrate stability zone. In contrast to what is observed in Cascadia, the seaward limit of the BSR is deflected even farther landward where all incoming sediment is currently being accreted. No BSR is observed beneath the morphologically distinct region north and northwest of the solid pink line in Figure 12. The lack of a BSR in this region is consistent with our interpretation that this region experienced widespread deformation and collapse due to subduction of seafloor topography, opening fluid pathways that allowed methane-rich fluids to escape.

The Inner Prism–Continental Framework

Figure 15 shows examples of data along the western edge of our study region (for location, see Fig. 12). On the basis of a few two-dimensional seismic refraction profiles perpendicular to the margin, Contreras-Reyes et al. (2017) proposed that the boundary between the middle prism and continental framework corresponds to the location of the shelf break, which is approximately defined by the 1000 m bathymetric depth contour along the central Chile margin (Figs. 1, 4), and noted a correlation between the northern limit of slip in 2010 and an abrupt landward deflection of this contour. The eastern end of our survey approaches this boundary and indicates considerable along-strike variation in the subsurface structure (Fig. 15).

The eastern ends of lines CP06 (Figs. 15A, 15D), CP08 (Fig. 15E), and CP20 (not shown) show basement rocks at the seafloor, which form a bathymetric and structural dome (labeled “basement uplift” on Figs. 12, 15). This uplift is overlies a subducted seamount with a radius of ∼15 km modeled by Maksymowicz et al. (2015) based on the gravity gradient (Fig. S4 [footnote 1]). The numerous normal faults that characterize the top of this local uplift and the slump scars along its western margin (Fig. 12) support an interpretation of rapid local uplift overlying a subducted seamount, although no distinctive wake indicative of seamount passage is observed to the west along the plate motion direction. We speculate that the characteristic signature of a subducted seamount that has plowed through the accretionary prism (uplift followed by subsidence parallel to the plate convergence direction) has been overprinted by the effects of transpression and transtension in this region.

North and south of the basement uplift, the eastern ends of lines CP15 (Fig. 15B), CP19 (Fig. 15C), and CP11 (Fig. 15F) show that the subsurface structure is characterized by faulted, folded, and tilted sediments similar to those observed in the transpressive zone to the west. These sediments overlie a highly deformed, acoustically transparent “basement” that has been uplifted relative to the basement to the west. The shallowest overlying sediments are subhorizontal and relatively undeformed. We tentatively interpret this transition to represent the transition from the active accretionary prism to the continental framework, which implies that this transition occurs west of where it was mapped by Contreras-Reyes et al. (2017) based solely on bathymetric data. Our proposed accretionary prism–continental framework transition is characterized by uplift and subsidence along strike, reflecting the structural complexity of this lithological contact. The northeastern extent of lines CP13a and CP15 (Fig. 16) are consistent with this interpretation because the most recent sediments were deposited in a broad basin rather than in multiple smaller basins bounded by faults. The dominant mechanism for stratigraphic disruption in these sediments is the complex interplay between canyon migration and instability of the canyon sides. Strata parallel to the basement overlain by a pronounced unconformity on line CP13a show evidence for a period of uplift on both sides of the basin followed by rapid sedimentation within the basin. Assuming that the basement rocks on these lines represent the continental framework, the top of basement on line CP15 is at a depth of ∼2400 m, in contrast to 1000 m at the crest of the uplifted dome. These results highlight the three-dimensional structure of the subduction backstop in this region.

Summary of New Stratigraphic and Structural Observations

The ChilePEPPER seismic reflection data provide a unique opportunity to examine the relationship between deformation front structure and slip during a major subduction zone earthquake. For the 2010 Mw 8.8 Maule earthquake, most models place the maximum amplitude of slip offshore near the northern end of an ∼500-km-long rupture plane. A primary observation of our study is that the structure of the deformation front changes abruptly from subduction of most of the incoming sediment up dip from the patch of greatest slip during the 2010 Maule earthquake to nearly complete accretion immediately to the north. This transition occurs over a distance of <10 km, as constrained by MCS lines ∼10 km apart (Figs. 8, 9), and is likely more localized, as suggested by an abrupt change in seafloor morphology. South of this transition, the frontal thrust appears to sole into a detachment surface, resulting in underplating of approximately one-third of the incoming trench sediment beneath the outermost prism and subduction of the remaining two-thirds to greater depth. Although the topographic signature of accretion reappears south of seismic line CP10, the MCS data indicate that accretion here is restricted to the sediments of a recent surficial thick shallow wedge (TSW) located upstream from where the Maule Canyon joins the trench (Fig. 10). Such sedimentary deposits are typical of areas where canyons enter the trench along the Chile margin (e.g., Thornburg and Kulm, 1987; Völker et al., 2006).

The frontal thrust on all ChilePEPPER lines, including and south of line CP21, appears to sole into the same stratigraphic horizon. In the trench, this horizon is a multicycle packet of strong reflectivity, suggesting that the décollement surface at depth beneath the outer prism is a smooth, stratigraphically controlled surface where the margin is dominated by sediment subduction (Fig. 11). Because the amplitude of this reflection does not change to the north, where it is cut by the frontal thrust in the region of near-complete sediment accretion and the décollement follow a deeper bright reflection, we do not a find an explanation for this along-strike structural change in the trench sediment structure.

A distinct change in the seafloor morphology can be traced from the transition in deformation front structure across the accretionary prism (Fig. 12). Northeast of where nearly all sediment is being accreted, the seafloor is deeper and more irregular than to the south, where sediment is being subducted. We interpret the persistence of the morphologic transition deep into the accretionary prism to indicate that the current along-strike change in the structure of the deformation front is not an ephemeral characteristic of the margin.

Along the landward edge of the middle prism at the southern and northern boundaries of the study area (Figs. 13, 14), we observe “flower” structures, which suggest strike-slip motion in a transpressive stress field that is consistent with the oblique plate motion vector in this region. An apparent pull-apart basin has formed in response to an extensional offset in this shear zone.

The basement rocks east of the shear zone are characterized by an uplifted dome where normal-faulted basement is exposed at the seafloor. This dome is approximately coincident with the patch of greatest slip and is flanked to the north and south by basins filled by tectonically undeformed sediments (although their stratigraphy has been disrupted by canyon incision and slumping of the canyon walls). Based on an apparent decrease in the amount of deformation in the slope basin sediments from west to east across the shear zone, we infer that the shear zone represents the boundary between the active accretionary complex and the continental framework, which is thought to be composed of older accretionary complex overlain by a slope basin in this region. This interpretation implies that the elevation (and presumably the thickness) of the continental framework basement here varies by at least 1.4 km along strike and that the strong curvature of the topographically defined edge of the continental framework near 34.5°S (Figs. 1, 4) may be due to local crustal thinning and subsidence rather than to a landward swing of the contact between the active accretionary prism and the continental framework.

Implications for Inferring the Up-Dip Extent of Seismogenic Slip from Deformation Front Structure

Based on seafloor geodetic evidence for slip extending to the trench in the 2012 Tohoku earthquake (Fujiwara et al., 2011) and on seismic images of the deformation front before and after the earthquake (Kodaira et al., 2012), which show a low-angle thrust fault extending into the trench in the post-earthquake image that was not present prior to the earthquake, other investigators have interpreted similar deformation front and trench structure to be a proxy for future tsunamigenic potential (e.g., Bécel et al., 2017). We now discuss our results in this context.

Figure 3 shows the 10 and 15 m slip contours from Yue et al. (2014), who suggested that slip locally extended to the trench along this segment of the Maule rupture. The Yue et al. (2014) slip contours straddle the transition from near-total accretion of trench sediment to significant sediment subduction with apparent deformation of the trench sediments approaching the deformation front. We consider two possible interpretations for this observation: (1) the deformation front structure does not influence whether slip in an earthquake reaches the trench and is therefore a poor proxy for past earthquake behavior, or (2) the slip contours are biased ∼20 km to the north or to the south. Noting that our results imply an offset in the plate boundary fault when viewed in cross-section (Fig. 17) and that studies of strike-slip faults indicate that such offsets often act as barriers to slip propagation (e.g., King and Nabelek, 1985; Oglesby, 2005), we consider interpretation 1 to be unlikely.

If the Yue et al. (2014) contours are not a precise indicator of where slip extended to the trench, then which structural segment is the most likely culprit? Maksymowicz et al. (2017) differenced swath bathymetry obtained before and after the earthquake, and concluded that the accretionary prism along seismic line CP06 experienced ∼5 m of uplift during or soon after the 2010 earthquake that extended to within 6 km of the trench. The results reported by Tréhu et al. (2019), who observed few aftershocks and scant evidence for slow earthquakes or seafloor fluid flow anomalies on a small-aperture ocean bottom seismometer network deployed on the outermost prism from May 2012 to March 2013, are consistent with this observation. Slow earthquakes and other indicators of fluid flow would be expected in response to at least 15 m of slip farther down dip if the outermost prism were characterized by velocity-strengthening behavior during the earthquake.

We conclude that the most likely scenario is that slip extended to the trench along the segment where the frontal thrust soles into a shallow detachment and nearly all of the trench sediment is subducted. The seismic reflection character of this segment of the accretionary prism, which shows no resolvable deformation within the trench except for within the shallow wedge deposited near the intersection with the Maule Canyon, indicates a very weak plate boundary thrust that does not transmit stress from the plate boundary to the lower plate. Contrasting seafloor topography between the segment of the margin characterized by sediment accretion and that characterized by subduction extends across the outer and middle accretionary prism, indicating that these bathymetric features are relatively long-lived features of the margin.

Undeformed trench sediment and sediment subduction are observed along much of the margin to the south in the region that slipped during the great A.D. 1960 earthquake south of our study area (Geersen et al., 2013; Olsen et al., 2017). However, thrust faults extending through the trench sediments from the seafloor to basement, similar to those on line CP19, are observed locally on seismic lines that cross the deformation front at 37.75°S (line SPOC-43, Geersen et al., 2013) and at 38.0°S (Olsen et al. 2017), adjacent to the Arauco Peninsula where the northern boundary of slip in the 1960 earthquake and the southern boundary of slip in 2010 overlap. Like the accreting segment in our study area, the segment of the margin between 37.75°S and 38°S is also characterized by seafloor topography indicative of fold development and outward building of the prism and a submarine canyon that is cut off before it reaches the trench (Fig. S5 [footnote 1]). These segments of the margin are correlated with high uplift rates onshore, which has been interpreted to be a geological proxy for upper plate deformation and long-term creep on the plate boundary, which accommodates some of the convergence (Saillard et al., 2017). We conclude that underthrusting beneath a shallow, lithologically controlled décollement with little deformation of trench sediments may be characteristic of slip to the trench in large subduction earthquakes in south-central Chile (and by inference, in similar subduction zones characterized by a moderate thickness of trench turbidites). Olsen et al. (2017) suggested that the subducting Mocha fracture zone (Fig. S5) may be responsible for deformation of the trench sediments near 38°S; however, no oceanic fracture zone is present in the Nazca plate near 34°S.

A similar (albeit less abrupt) change in deformation front structure from partial subduction to near-total accretion occurs along the Cascadia subduction zone (MacKay, 1995). Han et al. (2017) documented a change in the velocity structure of trench sediments along strike that is roughly correlated with this change in structure, with higher-velocity (and presumably stronger) sediments associated with sediment accretion offshore Washington (USA). Based on observations of strong sediments near the base of the 4-km-thick incoming sediment pile offshore Sumatra (e.g., Gulick et al., 2011), where the 2004 earthquake generated a devastating tsunami, Han et al. (2017) suggested that rupture may be more likely to propagate to the deformation front in northern Cascadia, where all sediment is being accreted. The relationship between sediment subduction or accretion and tsunamigenic potential proposed by Han et al. (2017) is the opposite of what we are proposing here. This may be due to different sediment thickness or composition, different underlying plate age (and thus temperature), different sediment consolidation state, or the ambiguity of relating structure to the up-dip extent of slip in large earthquakes where the details of the slip distribution are not known. Identifying the geologic signature of plate coupling that can be used to anticipate likely slip behavior in future plate boundary earthquakes in subduction zones is challenging, even in recently active and well-monitored regions (e.g., von Huene et al., 2019), and additional high-resolution observations of geologic structure in the source regions of large, well-documented subduction zone earthquakes is needed.

Reconciling the Long-Term Contrast in Margin Structure with the Contrast in Deformation Front Behavior

We have documented that an abrupt shift from sediment accretion to sediment subduction at the deformation front at the northern boundary of slip during the Maule earthquake may have controlled both up-dip and along-strike slip behavior. We have also noted that this contrast appears to be a long-term feature of the margin that has persisted over many earthquake cycles. Long-term persistence of the observed shift from sediment subduction to sediment accretion should lead to an increasing offset in the deformation front over time, which is not observed.

Two possible mechanisms that can result in a small offset in the deformation front even though there is a sharp transition from accretion to subduction of trench sediment are: (1) cyclical behavior of the deformation front, as elegantly proposed and modeled by Gutscher et al. (1998) to explain a similar along-strike contrast in deformation front behavior in the Aleutians; or (2) healing of an indentation in the deformation front caused by subduction of a topographic high, as proposed by Tréhu et al. (2012) to explain an anomalous deep basin on the Cascadia margin that is seaward of a subducted seamount inferred from topographic, seismic, and potential-field data. Although the cyclical model may locally be consistent with the observations on seismic line CP06 (i.e., apparent underplating of the upper 30%–35% of incoming sediment beneath the outer prism, near-constant thickness of the sediment subducted to greater depth, and the paleo–frontal thrust), it fails to explain why the structural change at the deformation front is coincident with a change in margin morphology that persists across the margin. We now examine the possibility that this margin has been affected by subduction of a region of anomalously thick crust and rough topography.

Numerous studies elsewhere have concluded that subducting plate topography interacting with the upper plate has a profound effect on forearc structure and seismicity (e.g., Cloos, 1992; Dominguez et al., 1998; Kodaira et al., 2000; Bilek et al., 2003; Bangs et al., 2006; Wang and Bilek, 2011; Tréhu et al., 2012; Morgan and Bangs, 2017). As mentioned earlier, Maksymowicz et al. (2015) inferred the presence of a subducted seamount that underlies the uplifted dome imaged in the MCS data based on the gravity gradient (Fig. S4A [footnote 1]). To explain the absence of a subducted seamount wake, they speculated that it had been sheared off of the subducting plate and obducted to the upper plate, forming an asperity that may have been responsible for the patch of greatest slip in the 2010 earthquake. We propose that the newly identified shear zone and pull-apart basin immediately seaward of the uplifted dome may have contributed to obscuring the seamount wake. Because no significant velocity anomaly is observed in the P-wave tomographic model of Hicks et al. (2014) beneath the offshore continental margin in this region, we conclude that the velocity of the subducted seamount is similar to that of the lower crust of the upper plate.

Onshore, Hicks et al. (2014) imaged high-velocity structures in the lower crust beneath the Chilean Coastal Range near 34°S and 36°S. The presence of these velocity anomalies has been correlated with slip segmentation in the Maule earthquakes and interpreted to be due to ultramafic intrusions related to Triassic extension and mantle upwelling (Bishop, 2018). To explain the margin-wide morphological and structural observations reported here, we suggest that a possible alternative origin for the anomaly beneath the coast at 34°S is subduction of a region of anomalously rough, shallow, and thick oceanic crust that has been entirely subducted.

Figure 18A shows the locations of these gravity and seismic velocity anomalies along with the locations of the major shallow structural boundaries identified in this study, seismicity before and after the 2010 Maule earthquake, and the region of critical prism topography determined by Cubas et al. (2013), which corresponds well with the transition in deformation front and accretionary prism structure. Not surprisingly, the region that ruptured in 2010 was relatively quiet prior to the earthquake, and many aftershocks were recorded in this region afterwards. As mentioned in the introduction, the outer prism was seismically quiet both before and after the earthquake. Although outer-rise aftershock activity extends along the entire study region, we note that it is centered on the segment where we document sediment subduction, uplift of trench sediments along the western edge of the trench, and deflection of the axial channel to the east, suggesting that loading of the subducting plate near the trench by the Maule earthquake is greatest here.

Just north of the uplifted dome is a swath of the outer and middle prism that is nearly devoid of aftershock activity. The lack of activity in this region was confirmed by the ChilePEPPER OBS study (Fig. 18A). This is also where the seaward edge of the continental framework swings sharply west when approximated by the 1000 m isobath (Fig. 1). We propose that the basement in this embayment may be continental framework crust that is anomalously thin and fractured because of prior subduction of an anomalously thick region of oceanic crust. It has been suggested that the fractured upper plate left in the wake of subduction of high relief would result in heterogenous stresses on the plate boundary, promoting the accommodation of plate motion through upper plate deformation and creep and small- to moderate-size earthquakes along the plate boundary (Wang and Bilek, 2011), consistent with the observations of pre- and post-earthquake seismicity and with the lower degree of interseismic locking reported for this segment by Métois et al. (2012). Such a zone of partial locking adjacent to a strongly locked patch on the plate boundary can act as a barrier to rupture propagation (Perfettini et al., 2010).

These observations suggest the plausible, albeit speculative, scenario shown in Figure 18B. Ages are estimated based on the plate convergence rate assuming that the structures depicted are still attached to the subducting plate and therefore represent minimum ages. In this model, a region of rough, anomalously thick oceanic crust that is now manifested as the Pichilemu seismic velocity anomaly (Hicks et al., 2014) was present seaward of the subduction zone by at least 2 Ma. Subduction of this structure resulted in uplift followed by subsidence and formation of a large embayment in the outer and middle prism at ca. 1 Ma. Continued subduction would have resulted in fracturing and erosion of the leading edge of the continental basement, resulting in the highly variable basement and basin structure observed landward of the active accretionary prism. The cross-section in Figure 18B shows the present structure along the leading edge of the continental framework implied by this scenario.

The large embayment left in the wake of subduction of a small oceanic plateau would have a subcritical topographic profile and would return to equilibrium through sediment accretion and internal prism deformation. We propose that the segment of margin characterized by accretion of nearly all trench sediment and heterogeneous morphology of the adjacent accretionary prism reflects healing of this embayment and restoration of the approximately linear structure of the Chile trench, which has persisted since the Cretaceous. Later subduction of a smaller seamount has resulted in the localized dome. We propose that along-strike transport of the outer and middle accretionary prism relative to the backstop along a dominantly strike-slip structure has obscured the expected subducted seamount wake. Similar processes may be occurring along other accreting segments of the trench, such as that near 38°S offshore from the Arauco Peninsula, where the process of margin disruption due to subducted topography may be at an earlier stage of evolution, with subduction of the Mocha fracture zone continuing at present. In contrast, we postulate that in our study region, the main topographic “disrupter” has been entirely subducted.

Maksymowicz et al. (2015, p. 276) concluded that: “Details of the relationship between local geologic structures in the forearc, inter-seismic coupling, slip during plate boundary earthquakes, and aftershock activity, however, remain ambiguous, and a process-based understanding will require that uncertainties in all of these characteristics of a subduction margin be resolved more accurately.” We hope that the speculations advanced here, which are suggested by the analysis and interpretation of new high-resolution seismic reflection data, represent a step forward toward a process-based understanding by presenting a model for the evolution of this apparent subduction zone segment boundary that can be tested through higher-resolution imaging of the deep structure of the margin and three-dimensional modeling of subduction zone dynamics.

High-resolution seismic reflection images of the continental margin near the northern boundary of slip in the 2010 Mw 8.8 Maule earthquake highlight the strongly three-dimensional structure of the forearc and the relationship between the crustal structure and along-strike and up-dip slip propagation in plate boundary earthquakes.

The seismic data demonstrate that the ratio of sediment accretion to subduction at the deformation front changes abruptly south of 34.43°S from subduction of most of the incoming sediment to accretion of nearly all the trench sediment on a seaward-verging frontal thrust fault to the north. Within the segment dominated by sediment subduction, which is located up dip from the patch of highest slip, 30%–35% of the subducted sediment appears to presently be underplated beneath the outermost prism and 60%–65% is subducted to greater depth. Only sediment in a local, shallow trench wedge formed near the confluence of the Maule Canyon and the trench is frontally accreted. The décollement at depth corresponds to the same bright, multicycle package of reflections observed in the trench along the entire imaged segment south of 34.43°S. The absence of resolvable faulting in the trench sediments seaward of the deformation front (with the exception of faulting within sediments composing the shallow trench wedge) indicates that the plate boundary is weak and does not transfer stress to the underlying sediments.

North of 34.43°S, a frontal thrust is observed that extends nearly to the top of the subducting oceanic crust, and the outer prism is formed of folded and faulted trench sediment. The transition in the deformation front structure is not correlated with any detectable changes in the thickness or seismic reflection character of the trench fill. It is, however, coincident with a distinct change in the morphology of the outer and middle accretionary prism, indicating that the along-strike variation in the mechanical response of the outer prism has been a persistent feature of the margin.

The abrupt nature of the transition from sediment subduction to accretion along strike implies a downward stepover in the plate boundary fault near the deformation front that may have acted as a barrier to northward propagation of slip in the Maule earthquake. We conclude that the apparently weak décollement overlying thick subducted sediment may have allowed slip to extend to the trench along this segment during the Maule earthquake. This conclusion is consistent with the pattern of outer-rise aftershock activity (Sladen and Trevisan, 2018) and with a change in elevation of up to 5 m in this region that occurred between 2008 and 2011–2012 (Maksymowicz et al., 2015). It implies that the signature of slip to the trench in Chile may be quite different from the signature observed in Tohoku, where deformation within the trench sediments appears to have resulted from the 2011 earthquake, or off Sumatra, where seismogenic slip on a deeply buried décollement is thought to have been enabled by the presence of unusually thick and stiff trench sediment. A better understanding of the impact of trench sediment thickness, composition, and temperature is needed before the structure of the deformation front can be used as a reliable proxy for slip to the trench and tsunami potential in the absence of a recent large earthquake.

Seismic data within the middle prism indicate a complex pattern of local uplift and subsidence, with several features interpreted to be “flower” structures indicative of a component of strike-slip motion along the eastern edge of this structural domain, consistent with the oblique plate convergence vector. A mid-slope basin in this segment may represent a pull-apart basin formed because of an extensional stepover in a transpressional fault system. In contrast to the margin to the north and south, the eastern boundary of this marginal domain is not clearly defined by the bathymetry, indicating a structurally complex boundary between the accretionary prism and the continental framework rock, interpreted to be pre-Cenozoic, metamorphosed accretionary prism. A dome-like structure of uplifted basement rocks is exposed on the seafloor locally, and the basement surface drops abruptly to the north and south of the dome and is buried beneath slightly folded forearc basin sediments. The uplifted basement dome is coincident with an anomaly in the vertical derivative of the gravity field that has been interpreted to be due to a subducted seamount (Maksymowicz et al., 2015). The absence of a “seamount wake” is attributed to disruption by the transpressional and transtensional deformation affecting the middle prism region.

We speculate that earlier passage of a larger welt of anomalously rough, thick, and shallow oceanic crust created a large embayment in the margin north of the uplifted basement dome and that the heterogeneous forearc morphology and truncation of the Huenchullami and Mataquito channels before they reach the trench that characterizes this segment of the margin are relicts of this history. In this model, a pronounced embayment in the edge of the continental shelf reflects an along-strike depression of the continental basement rather than a landward deflection of the boundary between active accretionary and continental basement rocks.

In summary, our detailed analysis of the forearc structure at the northern limit of rupture during the 2010 Maule earthquake reveals a strong relationship between rupture propagation during the main shock, aftershock seismicity, and long-term structural evolution of the forearc. Defining the structures at subduction zone segment boundaries in three dimensions and reconstructing their history requires dense grids of seismic reflection data and integration of these data with a wide variety of complementary data sets. While it is premature to interpret structural patterns as proxies for slip in future large subduction zone earthquakes, focused studies of three-dimensional structure tied to well-documented slip patterns in large earthquakes has promise for anticipating future earthquake behavior.

We thank the captain and crew of the R/V Melville for excellent ship handling and Lee Ellett and Jay Turnbull of the SIO Shipboard Geophysical Group for operating the seismic reflection and other geophysical acquisition instrumentation. Vern Kulm very kindly spent a morning bringing us up to speed on ideas about sediment processes in the Chile trench. The paper benefitted greatly from thoughtful and probing reviews from guest editor Laura Wallace and two anonymous reviewers. Funding for this work was provided by grants OCE1130013 and OCE1129574 from the U.S. National Science Foundation (NSF) to Oregon State University and the Scripps Institution of Oceanography, respectively, and by CONICYT/FONDECYT grant number 11170047 and CONICYT PIA/Anillo ACT172002 to the Universidad de Chile. Geophysical data were acquired using the NSF-supported Shipboard Geophysical Group at the Scripps Institute of Oceanography. All seismic data are archived at and are freely available from the U.S. Academic Seismic Portal maintained at the University of Texas at Austin (http://www-udc.ig.utexas.edu/sdc/). Other underway geophysical data are archived at the Rolling Deck to Repository data facility (https://www.rvdata.us/catalog/MV1206). Generic Mapping Tools (Wessel and Smith, 1998) and GeoMapApp (http://www.geomapapp.org) were used to compile the maps.

1Supplemental Material. Includes additional examples of seismic data, a gravity map of the study region, and a comparison of the morphology of the deformation front in the study region to that of the deformation front near 39°S. Please visit https://doi.org/10.1130/GES02099.S1 or access the full-text article on www.gsapubs.org to view the Supplemental Material.
Science Editor: Shanaka de Silva
Guest Associate Editor: Laura M. Wallace
Gold Open Access: This paper is published under the terms of the CC-BY-NC license.