We present a new tectonic plate reconstruction that suggests substantial revisions to events associated with development of the Peruvian flat slab and resolves several long-standing issues regarding the subduction of bathymetric highs in the region. The Tuamotu Plateau is widely considered to be the product of Easter Plume magmatism, and plate reconstructions suggest it formed following initial plume ascent at ca. 55 Ma. The Nazca Ridge is also linked to the Easter Plume and is an obvious candidate to be the spreading ridge conjugate to the Tuamotu Plateau. Models for the paired evolution of the two ridges, however, generally stop at ca. 33 Ma because of the inability of plate reconstructions to associate the two ridges across a spreading center prior to this time. In addition, seafloor magnetic data demonstrate that the Tuamotu Plateau developed at a complexly shaped and evolving mid-oceanic ridge that precluded development of a simple mirror image conjugate of the type commonly employed in Nazca Ridge reconstructions. Seafloor isochrons also suggest that a ridge jump separated the Tuamotu Plateau from its conjugate at ca. 42 Ma. Global plate models offer an alternative approach to assessing conjugate development, by showing how a hypothetical conjugate to the Tuamotu Plateau is built up over time. Using such a model, we found that the conjugate that developed during the main stage of Tuamotu growth (55 Ma to 42 Ma) cannot be the Nazca Ridge, which appears to have initiated at ca. 42 Ma, when the Easter Plume diverted volcanism southward. We named the newly recognized conjugate the Enigma Ridge. Importantly, subduction of this ridge starting from ca. 17 Ma on the north Peruvian trench can account for the missing slab buoyancy previously attributed to the hypothesized, but controversial, Inca Plateau. The Enigma Ridge must still be providing far more buoyancy over a much greater area than the Nazca Ridge, which only began to subduct rather recently.

The Peruvian flat slab is the longest in the world, and subduction of the buoyant Nazca Ridge is commonly invoked as a major factor in its development (Hayes et al., 2018; Bishop et al., 2017; Gutscher et al., 1999). Questions remain, however, about the extent to which buoyant crust generated by large igneous provinces controls flat slab development globally (Skinner and Clayton, 2013), and the Nazca Ridge’s capacity to account for the entire Peruvian flat slab (Gutscher et al., 1999; Ramos and Folguera, 2009). There are also many uncertainties concerning the Nazca Ridge itself, including the extent to which it has already been subducted and the details of its development as a conjugate to the Tuamotu Plateau (Hampel, 2002; Steinberger, 2002). Given these issues, we provide a new perspective on the development of the Peruvian flat slab based on mantle plume–plate boundary interactions as indicated by global plate reconstructions.

Despite the Tuamotu Plateau and Nazca Ridge (Fig. 1) being considered mid-oceanic-ridge conjugates, plate reconstructions have always presented challenges to resolving their early histories. Two approaches have been used to avoid this issue and model the later evolution of the Nazca Ridge. Some models, such as that of Pilger and Handschumacher (1981), considered only the currently unsubducted Nazca Ridge, which resembles a possible conjugate to the eastern end of the Tuamotu Plateau at ca. 33 Ma. The other approach adds a mirror image of some portion of west Tuamotu onto the eastern end of the Nazca Ridge to infer the size and history of the subducted part of the ridge. Pilger (1981) presented a model in which a ca. 55–43 Ma west Tuamotu conjugate was added to the northern end of the Nazca Ridge, and Hampel’s (2002) widely cited reconstruction matched dates for Pacific and Nazca Plate seafloor magnetic anomalies to deduce the size of the subducted conjugate to the western Tuamotu Plateau. Both papers “flipped” some portion of west Tuamotu and added it onto the present-day Nazca Ridge to infer the extent of the ridge’s subduction.

The challenge of resolving full development of the Tuamotu–Nazca system persists in recent global plate reconstructions. In the GPlates-based model used here (Müller et al., 2016), the earliest that any part of a hypothetical Nazca Ridge, projected back in time, lies on the Pacific-Farallon MOR is 43 Ma, but it does not align with the Tuamotu Plateau, and its implied Pacific Plate conjugate south of Tuamotu is missing. In other words, the plate reconstructions suggest that west Tuamotu cannot be added to the Nazca Ridge to establish its subducted counterpart.

Steinberger (2002) noted that while the more isostatically compensated northern part of the Nazca Ridge was plausibly formed at a spreading ridge, the less compensated southern ridge likely formed off-axis. Ar-Ar age dates (Ray et al., 2012) do suggest, for the GPlates reconstruction discussed here, that north Nazca Ridge volcanism occurred close to a spreading ridge–transform corner at ca. 31 Ma (Fig. 1B), whereas 25 Ma volcanism in the southern part of the ridge occurred far from any spreading center. These observations imply that a simple mid-oceanic ridge conjugate model is unlikely to apply even to the extant Nazca Ridge, and that extrapolations to any subducted counterparts based on such a model are suspect.

If the Nazca Ridge is not a simple conjugate to the Tuamotu Plateau, then there are consequences for the long-standing question of whether the buoyancy of the relatively narrow Nazca Ridge, subducting onto the southern part of the Peruvian flat slab, could account for the entire shallow subduction event (Pilger, 1984; Eakin et al., 2015). Gutscher et al. (1999) explained the latitudinal extent of the flat slab by invoking an “Inca Plateau” conjugate to a 43-Ma plateau that is speculated to underlie the <6 Ma Marquesas Plateau seamounts (Fig 1A). This scenario is far from universally accepted, with many authors attributing the seamounts to a young mantle upwelling (e.g., Guillou et al., 2014; Konrad et al., 2018). Additionally, Skinner and Clayton (2013) argued that even if the Inca Plateau did exist, the correct location of the conjugate is now too far east to continue supporting the flat slab (Fig.1A), because Gutscher et al.’s (1999) plate rotation model did not incorporate data from both sides of the mid-oceanic ridge, which has been spreading asymmetrically for at least 51 m.y. (Rowan and Rowley, 2014). Regarding the issue of slab buoyancy, it is also worth noting the relative volumes of the various crustal anomalies as indicated by their seafloor depth contours on crust of similar age (Fig. 1A). All of the bathymetric anomalies illustrated in Figure 1A were, or were thought to be, generated at similar times and in relatively close proximity. They are often portrayed based on their recognizable forms on the seafloor, with the Nazca Ridge and Marquesas seamounts defined by an outline at 4000 meters below sea level (mbsl), while the Tuamotu Plateau is commonly illustrated with an outline at 3000 mbsl. An Inca Plateau conjugate to the Marquesas Seamount 4000 mbsl feature would incorporate a far smaller volume of anomalous crust than the western Tuamotu Plateau and provide far less buoyancy. Given these issues and the uncertainties that propagate into assessments of the Peruvian flat slab, we employed a new approach by using a GPlates plate reconstruction to assess the evolution of a Tuamotu conjugate. This method also identifies other issues with previous studies, as outlined below.

As noted by Seton et al. (2023), choosing the best plate reconstruction model depends on the purpose of the study and the research question being addressed. We employed the most recent update of the Müller et al. (2016) tectonic plate model, which uses a hotspot mantle reference frame, in the GPlates 2.3 plate reconstruction software. The Matthews et al. (2016) model produced the same results, at least for the subject matter of this study. More recent GPlates models were developed for a variety of purposes. The Müller et al. (2022) model, for example, is optimized using a combination of a hotspot reference frame, net lithospheric rotation constraints, and trench-advance kinematics that is less appropriate for the subject of this study. The plate motions for the Pacific in the Müller et al. (2016) model utilize the relative motion from Croon et al. (2008) for 0 Ma to 47.9 Ma and Wright et al. (2015) for 47.9 Ma to 83 Ma. The default GPlates isochrons (Seton et al., 2020) are defined by picks from a global data set of >96,733 individual anomalies of the GEOphysical Data System Archive (GEODAS), which was developed by the U.S. National Geophysical Data Center (Sharman et al., 2001). The archive provides consistent and well-documented informati on on magnetic anomalies that are accessible to the public via a dedicated website (Müller et al., 2016; http://www.soest.hawaii.edu/PT/GSFML). An animation of events described in the paper (File S1 in the Supplemental Material1) and the files required to replicate the results in GPlates (File S2; see footnote 1) are provided. For simplicity, here we refer to the results of this particular collection of files as the “GPlates model” or “GPlates reconstruction.”

Isochrons from various studies of the Tuamotu Plateau and Nazca Ridge display comparatively minor differences in the study areas (e.g., Hampel, 2002). The picks used in the present study are mainly derived from Munschy et al. (1996) and illustrated in Figure 2A. The Munschy et al. (1996) set of picks provides the most complete coverage of the area. Other sets, such as those of Herron (1972), are not as extensive, and they may display slight offsets (~50 km) in the location of isochrons compared to those used here due to differences in techniques employed to geolocate the features on the oceanic floor and inherent margins of error in the data. Nonetheless, Skinner and Clayton (2013) showed that the application of five rotational models extending from Pilger (1981) through Müller et al. (2008) were equally good at reproducing chrons 10 (28 Ma) and 13 (33.1 Ma) to the north and south of the Nazca Ridge, using the locations of the same chrons in the vicinity of the Tuamotu Plateau.

Hampel (2002) assessed published figures for chrons 15–20 (34.7–43.8 Ma) in the vicinity of both the Tuamotu Plateau and the Nazca Ridge (Cande et al., 1989; Munschy et al., 1996; Cande and Haxby, 1991) and noted that whereas the chrons north and south of the Nazca Ridge were oriented near parallel, the published data suggested that chrons north and south of Tuamotu were less consistent. As a result, possible alignments of Nazca chrons could be made with either the north or south Tuamotu chrons, but not both. Hampel (2002) preferred matching the chrons north of the two bathymetric highs, and the resulting configuration and location of the subducted “mirror image” conjugate is widely cited in the literature (Bishop et al., 2017; Eakin et al., 2015). Hampel (2002) recognized some uncertainties in the location of chron 18 near Tuamotu but suggested that the impact on the overall reconstruction was minimal given that the model was also defined on the basis of chrons 15 and 16.

Other interpretations of the magnetic picks around the Tuamotu Plateau are possible. Although the concept of large igneous province conjugates at spreading ridges may suggest mirror-image bathymetric highs, this scenario is only feasible across straight spreading ridge segments and where spreading rates are symmetrical. In contrast, the Pacific–Farallon mid-oceanic ridge in the vicinity of the Tuamotu Plateau during the large igneous province’s formation was complex, continually evolving (Fig. 2A), and seafloor spreading is recognized to have been asymmetrical (e.g., Rowan and Rowley, 2014). In the GPlates reconstruction, the orientation of chron 18 (43.789 Ma) is not treated as anomalous. Instead, the distribution of seafloor magnetic picks is interpreted to record a ridge jump, or at least a rapid change in plate-boundary shape. This reconstruction negates the possibility that ideal conjugate mirror-image bathymetric highs formed across the Pacific–Farallon spreading center.

Hampel (2002) did not consider isochrons older than chron 21 (47.9 Ma) in detail because they are not preserved in the vicinity of the Nazca Ridge and were not considered essential for inferring the location of the subducted Tuamotu conjugate. In the GPlates reconstruction, however, there were ridge jumps, or changes in mid-oceanic-ridge shape, at 55.9 Ma and 40.1 Ma. In Figure 2, the locations of magnetic picks selected from GEODAS are shown along with the GPlates reference isochrons taken from Seton et al. (2020). The combination of pick locations and other features, such as faults or fractures and changes in the orientation of magnetic anomalies, provides strong support for the main features of the GPlates model and provides a basis for our Tuamotu conjugate model. Although the details of these plate boundary changes—as illustrated in the GPlates global plate reconstruction—may be simplified and not correct regarding every minor detail, they provide an approximation of events that is much more accurate than any reconstruction that ignores the jumps entirely.

Our reconstruction of the Tuamotu conjugate between ca. 55 Ma and 40 Ma is based on seafloor magnetic data, but it coincides with the formation of a mantle plume-related bathymetric high. The Easter Plume has long been considered the source of magmas for both the Tuamotu and Nazca bathymetric highs, particularly once the relative motion of the Hawaiian plume and other plumes was clarified (Pilger, 1981; Hampel, 2002; Steinberger, 2002). Although multiple suggestions were provided for the plume’s present location, there is general agreement that it lies in the vicinity of Easter Island and Salay Gomez, which are ~400 km apart. Our model used the location provided by Montelli et al. (2006) between these islands (27°S, 108°W) as the location of the plume conduit, but we note that Montelli et al. (2006) estimated that the present size of the upwelling in the shallow mantle is ~800 km in width. Tracking the Easter Plume back in GPlates provides a location at ca. 55 Ma that is consistent with geological evidence. For example, Steinberger (2002) inferred that the absence of a plume track on the Pacific Plate requires that initial ascent occurred either on the Farallon Plate or in the vicinity of the Pacific–Farallon spreading ridge. Our reconstruction places the plume in the vicinity of the mid-oceanic ridge and nascent Tuamotu Plateau as required, and the plume track eventually moves away from the plateau on a track that corresponds closely to the Plateau’s eastern “tail,” which is located on slightly older seafloor.

To resolve gaps in current conjugate models and to clarify how the Nazca Ridge relates to formation of the Peruvian flat slab, we considered the entire evolution of the Tuamotu Plateau from the time of Easter Plume ascent. Spreading ridge jumps will complicate application of a simple mirror-image conjugate model to any oceanic plateau, and they often occur when mantle plumes weaken overlying crust (Mittelstaedt et al., 2008; Sager et al., 2016). Multiple ridge jumps during development of the Tuamotu Plateau have long been suggested (Ito et al., 1995). We note that the ridge jumps occur as the Easter Plume track passes by, which is consistent with plume-induced crustal weakening (Fig. 1). Although Montelli et al. (2006) report that the present Easter Plume head has a radius of ~400 km, we illustrate a 500 km radius for the plume head (Figs. 3 and 4), given that its initial radius was likely larger.

The main body of the central–east Tuamotu Plateau and the northern end of the Nazca Ridge have compensated gravity signatures that require development on young crust in an on-axis, or near-axis, setting (Steinberger, 2002; Figs. 1B and 1C). An age of at least 51 Ma for the west Tuamotu Plateau was determined from late early Eocene shallow-water fossils (Fig. 3) in turbidites that overlie presumed plume basalts (Schlanger et al., 1976). Schlanger et al. (1984), however, also reported 47.4 Ma and 41.8 Ma Ar-Ar ages for basalt samples overlying the turbidites that were suggested to indicate a second plume (Ito et al., 1995). These magmas actually erupted near or within the region still underlain by the relatively young Easter Plume head (Fig. 3), which provides a more plausible source for the young magmas while also negating a multi-plume scenario. In any case, the younger magmatism has obscured the likely >51 Ma “near axis” gravity signature in the area.

As with previous workers, we employed mirror images to infer the configuration of the subducted conjugate. However, we applied the technique to segments of the Tuamotu Plateau to minimize the effects of ridge jumps and variably shaped spreading ridges. The successively younger segments and their mirror images are bounded by the locations of the Pacific-Farallon mid-oceanic ridge at various times, as defined by seafloor magnetic anomalies (Fig. 3). An approximation of the Tuamotu conjugate developed by this process is offset from the Nazca Ridge. We refer to this proposed feature as the “Enigma Ridge.”

Ascent of the Easter Plume head to the shallow mantle was coeval with local rearrangement of the Pacific–Farallon boundary from >56–55 Ma, which is represented in GPlates by the black and dark green lines in Figure 3. Considering local magnetic picks, we illustrate additional isochrons in Figure 2. The oldest of these, at 71.1 Ma and 64.7 Ma, differ slightly from the idealized GPlates 67.7 Ma isochron and appear to curve somewhat toward the northern transform (Fig. 2B). Based on the 71–56 Ma data, it is unclear whether the change in plate configuration involved a number of small discrete ridge jumps, as is argued for the Shatsky Rise (Sager et al., 1988), or occurred via continuous evolution. Nonetheless, the irregular and changing shape of the mid-oceanic ridge cannot have produced an ideal mirror image on the Nazca Plate.

In our conjugate reconstruction, we distinguish two sites of maximum eruption for these early magmas but note that the details for west Tuamotu are the most ambiguous part of the conjugate-forming event. The northern side of west Tuamotu appears to have developed from magmas that erupted from the newly formed ~N–S mid-oceanic-ridge segment but with magmatic volumes subsiding after ca. 53.3 Ma (Fig. 2B; dark orange isochron C24n.3n). Over this same time span, magmatism in the central–southern part of west Tuamotu appears to have been focused on the evolving mid-oceanic-ridge–transform corner to the south. It is possible that the lower crustal architecture of this corner feature provided a favorable pathway for the younger (ca. 47–42 Ma) magmas. The rest of the Nazca Plate conjugate is built from well-defined sections along the Tuamotu Plateau’s length, which are bounded by actual or interpolated mid-oceanic ridge from Gplates and with minor adjustments for slight overlaps or gaps resulting from asymmetrical spreading ridges or transforms.

Mittelstaedt and Ito (2005) demonstrated how plume volcanism may be focused on portions of nearby plate boundaries because of a combination of stress field complexities linked to plume ascent, variations in oceanic floor thickness, and uplift and plate-boundary interactions. They show that volcanic lineaments in the Galápagos area fan out from a central location toward segments of the nearby spreading center instead of focusing on the closest point along the plate boundary. Similar factors likely played a role in the distribution of volcanism that formed the Tuamotu Plateau and its conjugate. Figures 2 and 3 show that excess volcanism from Tuamotu was only associated with a fraction of the available plate boundaries in the vicinity of the Easter Plume. Between 55 Ma and 42 Ma, plume volcanism did not erupt directly above the calculated plume axis location, but instead at a slightly more distant sector of the spreading ridge, although still beneath the 500 km radius of the young plume head inferred here. Shifting the Easter Plume axis to underlie the plateau at, for example, 56 Ma would not result in the axis being consistently located beneath the plateau between 55 Ma and 40 Ma. Such a shift would also displace the plume axis from the east Tuamotu plume track and disassociate the plume from the Nazca Ridge at 33 Ma. Ridge jumps and the evident evolution of the mid-oceanic-ridge–transform boundary (Fig. 2B) may have contributed to focusing volcanic magmatism due to their impacts on subcrustal topography.

The context for initial north Nazca Ridge volcanism has never been described in detail, but the short interval between 42 Ma and 40 Ma is the only plausible time for generation of the on-axis gravity signature noted by Steinberger (2002) because its distance from the mid-oceanic ridge increases rapidly thereafter. On this basis, the reduction of magmatic supply to narrow east Tuamotu must broadly coincide with development of the north Nazca Ridge. In Figures 3 and 4, construction of the western end of the Enigma Ridge terminates during this same ca. 42–40 Ma interval and coincides with the diminished volume of volcanism on its Tuamotu Plateau conjugate, and spatial constraints imposed on the location of the western termination of the ridge by the plate model. Based on the plate reconstruction, the plume axis directly approached the portion of the spreading ridge where the Tuamotu and Enigma ridges were being constructed at ca. 42 Ma. Given that the axis was previously located away from the locus of plateau volcanism, we suggest that the eruption site of plume lavas was driven farther south at this time. In this scenario, the consistent earlier displacement of excess volcanism away from the plume axis provides a basis for the otherwise unusual 42 Ma to 40 Ma development of the north Nazca Ridge. The distinctive seafloor magnetic features for east Tuamotu over this time interval (Fig. 2C) also require a unique event. The isochrons defined by the 43.8 Ma and 40.1 Ma picks both deflect toward the plateau margin, resulting in a localized close association of picks of both ages. This evidence indicates a westward ridge jump and provides a basis for the idealized or simplified jump of the Gplates reconstruction. In the Enigma model, the ridge jump necessarily causes the separation of the Tuamotu and Enigma conjugates (Figs. 3 and 4A). The proximity of the plume axis to the mid-oceanic ridge likely caused this ridge jump. The presence of the developing Nazca Ridge may have also contributed to the jump and/or the resulting mid-oceanic-ridge configuration, but the jump eliminated any possibility of the Nazca Ridge being connected to the Tuamotu Plateau. The jump explains why there is no preserved Nazca conjugate on the Pacific Plate despite the north Nazca’s near-ridge gravity signature noted by Steinberger (2002). Because of the events at ca. 40 Ma, and mid-oceanic-ridge spreading rates, the Enigma Ridge was completely subducted relatively recently, in the last several million years, and the leading edge of the Nazca Ridge began to subduct at around the same time.

The plume axis was almost stationary relative to the spreading center between 42 Ma and 40 Ma, but the ridge jump subsequently placed it back under the Pacific Plate (Figs. 4A and 4B). It remained in proximity to the eastern Tuamotu Plateau while seafloor spreading moved the Enigma Ridge eastward and away from the plume magmatic system. From 40 Ma onwards, the relative location of the plume axis moved southeastward as it formed the narrow eastern Tuamotu Plateau and again approached the spreading ridge. Once the plume axis was located back under the mid-oceanic ridge, another ridge jump occurred in less than two million years. Accordingly, the easternmost Tuamotu is mainly a plume track that formed on slightly older crust and therefore has no conjugate on the Nazca Plate. The eastern Tuamotu Plateau terminates close to the location of a 33 Ma spreading ridge–transform corner and the southward-developing Nazca Ridge. This late plate-boundary corner appears to have guided the formation of the remaining Nazca Ridge until the transition to the Easter Island Chain. Considering this sequence of events makes clear why previous reconstructions of a Tuamotu conjugate typically begin at ca. 33 Ma. The Easter Plume was clearly capable of developing a major large igneous province, but the only apparent candidate for its conjugate was the Nazca Ridge, and there is no obvious way to push the reconstruction further back in time to create a continuous model for Tuamotu–Nazca Ridge development.

In the new model, the Enigma Ridge collided with South America at ca. 17 Ma, slightly earlier than the conjugate of Hampel (2002) at 11 Ma, or Rosenbaum et al. (2005) at 15 Ma. Although our collisional time is only slightly different from the latter, the projected collision much more closely matches the extent of the Peruvian flat slab. Establishing the precise shape of the Enigma Ridge is not feasible, but our reconstruction (e.g., Fig. 3; File S1) approximates the effects of spreading-ridge asymmetry that has tended to accumulate more crust on the Farallon Plate than on the Pacific Plate over the last 80 m.y. In the region of the Tuamotu Ridge during its formation between 60 Ma and 40 Ma, the Gplates model indicates that spreading on the Farallon Plate occurred at ~1.2 times the Pacific rate. For example, it indicates that the Farallon 67.7 Ma isochron is 1.19 times the distance from the 55.6 Ma mid-oceanic ridge than the Pacific 67.7 Ma isochron, where these features are oriented broadly parallel to each other. As constructed in Figure 3, the Enigma Ridge began to subduct at 5°S versus 10°S (Rosenbaum et al., 2005) and 11°S (Hampel, 2002) in the previous Nazca mirror image models (Fig. 2). The earlier timing closely corresponds to the “profound uplift” (Noble and McKee, 1999, p. 160) associated with the Quechua I phase deformational event in the same area and the south-trending formation of major porphyry copper deposits (Fig. 4, 16 Ma panel; Singer et al., 2008). The onset of sedimentation associated with flexural uplift of the Iquitos Arch, which emerged after 11 Ma, also occurred in the 17–11 Ma Pebas Formation (Roddaz et al., 2005), which is inboard of the initial site of contemporaneous Enigma subduction (Fig. 4, 16 Ma).

The new model has implications for ongoing debates associated with the Nazca Ridge. For example, Espurt et al. (2007) suggested that the Amazonian Fitzcarrald Arch (Fig. 4F) is linked to Nazca Ridge-induced anomalous mantle flow that modified mantle dynamic loading. Conversely, Bishop et al. (2018) attempted to account for the location of the arch above the steep slab in a model where basal shear transferred crustal thickening eastward in the lower crust, while acknowledging that similar features were not observed in association with the Carnegie or Iquique ridges. Dávila et al. (2019, p. 56), however, noted that the model “disagrees strongly” with their calculations of regional dynamic topography. The subduction history of the Enigma Ridge suggests that its southward movement on the subducted slab may have contributed to the southward progression of sedimentation associated with the arch.

Different subducted Nazca Plate topological models have appeared regularly. Some potentially important features, such as the slab tear inferred by Antonijevic et al. (2015) and Scire et al. (2016), are not observed in all studies (e.g., Lim et al., 2018), but all models display shallow slab inboard of the Nazca Ridge. The Scire et al. (2016) paper illustrated a box-like area of flat slab directly inboard of the Nazca Ridge. This potentially critical feature appeared to categorically support a model where shallow subduction is driven by the subducted Nazca Ridge, as is widely accepted. For example, Bishop et al.’s (2018) argument against a slab buoyancy effect as the cause of the Fitzcarrald Arch was based on the premise that a spatially limited flat slab lay too far to the south. More recently, however, Portner et al. (2020) employed a more geographically extensive P-wave tomographic model that replaces the feature from Scire et al. (2016) with a much longer flat slab region that closely resembles the Slab2 subduction zone geometry model of Hayes et al. (2018), as shown in Figure 4F. The Portner et al. (2020) results may eliminate the basis of many models that were reliant on the highly Nazca Ridge-localized flat slab, but it presents no obvious challenge to the proposed Enigma Ridge.

Where is the Enigma Ridge now? According to the model proposed here, it is located at the edge of a sag in the Nazca Plate. Unlike the slab inland from the Nazca Ridge, which lacks deep seismic activity (Bishop et al., 2017), this area exhibits high seismic activity at 120–200 km in the Pucallpa “seismic nest” recently recognized by Wagner and Okal (2019). They suggest that the cluster of seismic activity may be related to the subducted segment of the Mendaña fracture zone. In our plate reconstructions, however, the Enigma Ridge lies just south of the fracture zone and is a prime candidate for high seismic activity. Given that eclogitization of the slab and ridge was likely delayed where the slab was flat (Bishop et al., 2017), much of the ridge that is now on the very steeply dipping part of the slab is likely undergoing phase transformations that may be tearing or deforming the ridge.

1Supplemental Material. File S1: ZIP file containing an .mp4 animation and its constituent jpeg files, Tuamotu_Enigma_Na zca_Animation.zip. Animation starts at 55 Ma, when the western Tuamotu Plateau and its slightly longer conjugate have already formed, and proceeds to 0 Ma. File S2: ZIP file containing files for the GPlates Enigma Project to be opened in GPlates 2.X, which can be downloaded from https://www.earthbyte.org/category/gplates/. Please visit https://doi.org/10.1130/GEOS.S.24424516 to access the supplemental material, and contact editing@geosociety.org with any questions.
Science Editor: Christopher J. Spencer

John Cannon, Maria Seton, and Sabin Zahirovic are thanked for assistance with the use of GPlates during the study. The comments of several reviewers greatly helped to improve the manuscript.

Gold Open Access: This paper is published under the terms of the CC-BY-NC license.