Flat-slab subduction occurs when the descending plate becomes horizontal at some depth before resuming its descent into the mantle. It is often proposed as a mechanism for the uplifting of deep crustal rocks ('thick-skinned' deformation) far from plate boundaries, and for causing unusual patterns of volcanism, as far back as the Proterozoic eon. For example, the formation of the expansive Rocky Mountains and the subsequent voluminous volcanism across much of the western USA has been attributed to a broad region of flat-slab subduction beneath North America that occurred during the Laramide orogeny (80-55 million years ago). Here we study the largest modern flat slab, located in Peru, to better understand the processes controlling the formation and extent of flat slabs. We present new data that indicate that the subducting Nazca Ridge is necessary for the development and continued support of the horizontal plate at a depth of about 90 kilometres. By combining constraints from Rayleigh wave phase velocities with improved earthquake locations, we find that the flat slab is shallowest along the ridge, while to the northwest of the ridge, the slab is sagging, tearing, and re-initiating normal subduction. On the basis of our observations, we propose a conceptual model for the temporal evolution of the Peruvian flat slab in which the flat slab forms because of the combined effects of trench retreat along the Peruvian plate boundary, suction, and ridge subduction. We find that while the ridge is necessary but not sufficient for the formation of the flat slab, its removal is sufficient for the flat slab to fail. This provides new constraints on our understanding of the processes controlling the beginning and end of the Laramide orogeny and other putative episodes of flat-slab subduction.

Nazca subduction beneath South America is one of our best modern examples of long‐lived ocean‐continent subduction on the planet, serving as a foundation for our understanding of subduction processes. Within that framework, persistent heterogeneities at a range of scales in both the South America and Nazca plates is difficult to reconcile without detailed knowledge of the subducted Nazca slab structure. Here we use teleseismic travel time residuals from >1,000 broadband and short‐period seismic stations across South America in a single tomographic inversion to produce the highest‐resolution contiguous P wave tomography model of the subducting slab and surrounding mantle beneath South America to date. Our model reveals a continuous trench‐parallel fast seismic velocity anomaly across the majority of South America that is consistent with the subducting Nazca slab. The imaged anomaly indicates a number of robust features of the subducted slab, including variable slab dip, extensive lower mantle penetration, slab stagnation in the lower mantle, and variable slab amplitude, that are incorporated into a new, comprehensive model of the geometry of the Nazca slab surface to ~1,100 km depth. Lower mantle slab penetration along the entire margin suggests that lower mantle slab anchoring is insufficient to explain along strike upper plate variability while slab stagnation in the lower mantle indicates that the 1,000 km discontinuity is dominant beneath South America.

The dynamics of flat subduction, particularly the interaction between a flat slab and the overriding plate, are poorly understood. Here we study the (seismically) anisotropic properties and deformational regime of the mantle directly above the Peruvian flat slab. We analyze shear wave splitting from 370 local S events at 49 stations across southern Peru. We find that the mantle above the flat slab appears to be anisotropic, with modest average delay times (~0.28 s) that are consistent with~4% anisotropy in a~30 km thick mantle layer. The most likely mechanism is the lattice-preferred orientation of olivine, which suggests that the observed splitting pattern preserves information about the mantle deformation. We observe a pronounced change in anisotropy along strike, with predominately trench-parallel fast directions in the north and more variable orientations in the south, which we attribute to the ongoing migration of the Nazca Ridge through the flat slab system.

Summary The Andean Subduction Zone is one of the longest continuous subduction zones on Earth. The relative simplicity of the two-plate system has makes it an ideal natural laboratory to study the dynamics in subduction zones. We measure teleseismic S and SKS travel-time residuals at > 1,000 seismic stations that have been deployed across South America over the last 30 years to produce a finite-frequency teleseismic S-wave tomography model of the mantle beneath the Andean Subduction Zone related to the Nazca Plate, spanning from ∼5° N to 45° S and from depths of ∼130 km to 1,200 km. Within our model, the subducted Nazca slab is imaged as a fast velocity seismic anomaly. The geometry and amplitude of the Nazca slab anomaly varies along the margin while the slab anomaly continues into the lower mantle along the entirety of the subduction margin. Beneath northern Brazil, the Nazca slab appears to stagnate at ∼1,000 km depth and extend eastward sub-horizontally for > 2,000 km. South of 25° S the slab anomaly in the lower mantle extends offshore of eastern Argentina, hence we do not image if a similar stagnation occurs. We image several distinct features surrounding the slab including two vertically-oriented slow seismic velocity anomalies: one beneath the Peruvian flat slab and the other beneath the Paraná Basin of Brazil. The presence of the latter anomaly directly adjacent to the stagnant Nazca slab suggests that the plume, known as the Paraná Plume, may be a focused upwelling formed in response to slab stagnation in the lower mantle. Additionally, we image a high amplitude fast seismic velocity anomaly beneath the Chile trench at the latitude of the Sierras Pampeanas which extends from ∼400 km to ∼1000 km depth. This anomaly may be the remnants of an older, detached slab, however its relationship with the Nazca-South America subduction zone remains enigmatic.