Can you create a boundary without changing the properties of the material itself? Think of velvet: try brushing its fibers in different directions and a clear boundary appears, even though the fabric on either side is identical.
Well, something similar may happen deep inside the Earth.
Earth scientists have known for decades that the mantle has boundaries at 410 and 660 km depth. They are well understood as phase transitions in mantle minerals. Under increasing pressure, olivine and its relatives rearrange into denser structures, and the transformed rock changes properties. Deep in the lower mantle, somewhere below 660 km, nothing as clean is expected. Yet seismic waves keep bouncing off something near 1000 km depth, especially beneath subduction zones, where old oceanic plates sink into the mantle. The usual explanations are thermochemical in origin such as piles of recycled oceanic crust, unusual composition, or stalling of slab material due to jumps in viscosity at that depth.
We wondered about a different possibility. What if the mineralogic composition is the same on both sides, and only the way its crystals are lined up changes?
Our team focused on bridgmanite as it is the most abundant mineral in the average lower mantle. Its crystals are intrinsically anisotropic, meaning seismic waves travel through them at different speeds depending on propagation direction. In a random jumble of grains that make up the bridgmanite aggregate, this averages out. When rock flows however, grains deform by slipping along particular planes and rotate as they do so, and the whole aggregate develops a preferred orientation. Using the velvet analogy, that is the velvet pile. Mantle flows vigorously around a subducting slab and thus we expect plenty of brushing.
The twist is that the way the material responds to "brushing" is not fixed. Laboratory experiments done by my co-author Jeffrey Gay and colleagues showed that by squeezing and deforming pyrolite – the most common model composition for the mantle – in a diamond anvil cell, the easiest way for bridgmanite to slip depends on pressure. Between about 24 and 50 GPa (roughly 670 to 1250 km depth), the favored slip system gradually changes. Near 40 GPa which is about 1000 km depth, the old favorite and the new one are equally easy to accommodate deformation. Same mineral, same chemistry, but the same flow now builds a different fabric. Earlier models of mantle flow and anisotropy mostly assumed a single slip system throughout the lower mantle, so they could not see this.
As it turns out, bridgmanite behaves differently with increasing pressure and depth.
We put this to the test by allowing the self-consistent evolution of pressure-dependent bridgmanite fabric development into realistic models of mantle convection. Tiny tracers are carried along with the modeled flow, and at each step we calculate how grains would deform and re-orient, then convert the result into the elastic properties of the rock. When we map the total anisotropy, a thin interface appears around the slab corresponding to a local minimum in anisotropy strength despite smooth changes in slip-system activities with depth. This is where one fabric hands over to the other. The discontinuity in anisotropy starts near 1000 km, where the slip systems cross over. It then sinks toward the slab reaching about 1400 km directly beneath it, where the rock has been strained the most (Fig. 1: left panel). Strain also sharpens the boundary: in highly deformed regions the transition spans about 40 km in depth, compared with about 80 km where the rock has been deformed less.
A feature in a model is not worth much unless it can be seen. Hence, we asked an important question – is the deformation-induced boundary strong enough to reflect seismic energy and therefore be observed seismologically? For this, we first tested how strongly the boundary should reflect shear waves using first-order reflection coefficient modeling. Here, we found that the reflection changes with direction and can even flip polarity. We then ran full waveform simulations, meaning we propagate seismic waves through our model, and searched for the reflection among the SS precursors, the weak waves that bounce off structures at depth before the main SS arrival. The reflection is generated, albeit a tiny fraction of the main SS arrival. Nevertheless, the measured precursor amplitude is above the global detection threshold reported by earlier observational work.
The model also explains other things. The pattern of radial anisotropy (i.e., the dependence of seismic-wave speeds to polarization) it predicts resembles seismic observations beneath the Western Java subduction zone, where the slab currently ponds near 1000 km depth. We anticipate similar findings beneath subduction zones exhibiting deep flattening such as beneath Peru and Kamchatka.
Our proposed mechanism, however, may have its limitations. For example, a slab that stagnates at 660 km strains the deeper mantle too little, and the deformation-induced boundary is almost invisible (Fig. 1: right panel). Other geophysical mechanisms could be at play to explain the observed mid-mantle reflections beneath them. We therefore do not claim that our proposed mechanism replaces the thermochemical explanations, only that ours may co-exist with the latter.
Like fibers in velvet brushed in different directions, crystals in the deep mantle can acquire different orientations while remaining the same mineralogic composition. For seismic waves, that change in orientation can be enough to create a boundary – and in that sense, some of the structures we detect deep inside the Earth may be less like permanent layers and more like a record of how the mantle has been moving. Finally, a more thorough modeling scheme involves the feedback of anisotropy on the flow and viscosity, since aligned crystals may deform more easily in some directions. This opens up an interesting avenue for future work – as it reconciles mantle flow and rheology with mineral physics and seismic observations to help explain why some slabs flatten near 1000 km depth, whereas others further sink in the deep mantle.