A water transport system across the mantle transition zone beneath western North America as imaged by electrical conductivity data Open Access

Earth’s subduction zones extend for a total length of ~55,000 km (Stern, 2002) and are the planet’s largest recycling system. A huge amount of water can potentially be recycled into the deep mantle by subducting slabs in this system (Ribeiro and Lee, 2017). However, slab hydration is heterogeneous (Worzewski et al., 2010; Halpaap et al., 2019), suggesting variable water-transport capability of slabs. The thermal structure and other unknown factors control the hydration status of the slab (Syracuse et al., 2010; Ribeiro et al., 2015; Fujie et al., 2018), meaning that slabs are grouped into cold, warm, and hot types. Cold slabs have pressure and temperature conditions (P-T conditions) that are below the stability field of dense hydrous magnesium silicates (DHMSs) and can therefore transport water into the lower mantle (LM; Ohtani, 2020). In contrast, DHMSs are not stable in warm or hot slabs in the deep upper mantle, allowing strong dehydration reactions to occur, and generating almost dry lithosphere conditions in the slabs before entering the LM (van Keken et al., 2011). In this study, we provide geophysical observations that demonstrate the transport of water by a warm slab into the LM.

We chose the western United States, where the warm Farallon plate is subducting into the mantle, as a suitable research location (Ribeiro and Lee, 2017). Seismic imaging interpretation has led to the division of the subduction history of this plate into two stages: slab sinking into the LM at ca. 55 Ma, and the present steep-angle subduction, with the Juan de Fuca plate representing a remnant of the Farallon plate (Sigloch, 2011). This young plate has undergone slow subduction (~25 mm/yr). A super-imposed heat effect from a plume in front of the trench (Obrebski et al., 2010) is inferred to cause the subducting Farallon slab to reach a temperature of only ~300 K below the ambient mantle geotherm (van Keken et al., 2011) before entering the LM, as estimated from an ~2%–3% increase in seismic velocity (Sigloch, 2011). Recent seismic tomography data, including offshore observations over the plate, have identified an along-subduction-zone low-velocity feature near the trench, suggesting highly hydrated oceanic lithosphere (Gao, 2018). It is plausible that this slab could accommodate a considerable amount of water when entering the deep mantle. Consequently, the western United States is an ideal location in which to investigate the possible existence of a water transport system across the mantle transition zone (MTZ) based on geophysical observations.

Our research relied on data from 12 geomagnetic stations across North America (Fig. 1A; Table S1 in the Supplemental Material¹), retrieved from the World Data Centre (http://www.wdc.bgs.ac.uk/). The data from these “dense” observations can be converted into variation in mantle electrical conductivity, which can be interpreted in terms of the mantle hydration state, as minor amounts of water can considerably enhance electrical conductivity (Yoshino et al., 2008; Karato, 2011). This approach has been used to investigate the water distribution in northeastern Asia (Kelbert et al., 2009) and southern Europe (Yoshino et al., 2008). Here, our high-resolution three-dimensional model of mantle electrical conductivity was obtained by using the L1-norm regularization inversion (Li et al., 2020) involving high-quality data (derived from careful processing of long-time signal recording), allowing us to infer the water transport system across the MTZ beneath the western United States. Our data processing and inversion methods are described in the Supplemental Material.

The most prominent features in our electrical model are two high-conductivity anomalies in the lower MTZ and uppermost LM beneath North America (Figs. 1B–1D). One anomaly lies beneath the eastern United States and extends to Bermuda and the Sargasso Sea. Its conductivity is the greatest beneath the middle Appalachian Mountains, ~7 × the global average (Kelbert et al., 2009), and we refer to this as anomaly “E”. The other anomaly, which we term anomaly “W”, was detected beneath the southwestern United States. Of note, anomaly W passes vertically through the lower MTZ into the uppermost LM. This anomaly extends latitudinally along the southern Rocky Mountains from Colorado to New Mexico and longitudinally from the coastline landward to the Great Plains, where it joins anomaly E. Anomaly W in the MTZ measures ~10° × 10° but extends into the uppermost LM to form a 15° × 10° belt extending from Nevada to Kansas. The conductivity of anomaly W is slightly less than that of anomaly E, but is on average 2 × greater than the global conductivity. Farther north, both anomalies abruptly terminate close to the boundary between the United States and Canada, possibly because of the lack of constraints from effective high-latitude observations.

Anomaly E was discovered by a global electromagnetic induction inversion and has been associated with possible Bermuda lower-mantle upwelling (Li et al., 2020) or with recycled hydrothermal material upwelling related to subduction events (Mazza et al., 2019). We focused on anomaly W because it may be associated with a melting layer at the 410 km interface (Toffelmier and Tyburczy, 2007) and with mantle upwelling near the 660 km discontinuity (Schmandt et al., 2014). More importantly, this anomaly is unexpectedly accompanied by seismically imaged high-velocity anomalies beneath the 660 km discontinuity (Fig. 1D). Incorporation of all of these observations into our model should advance our understanding of the mechanism by which water can be transported into Earth’s interior, and underpins the concept of the water filter model (Bercovici and Karato, 2003).

The conductivity of the lower MTZ, such as that of anomaly W, depends on mineral composition, temperature, and the presence of impurities such as hydrogen (Karato, 2011). Considering a normal geotherm temperature (~1875 K at 660 km depth) and an average conductivity of ~0.5 S/m, we estimated that the water content in ringwoodite is ~0.8 wt% (Fig. S9A). This value is consistent with results obtained from a one-dimensional model for the region beneath Boulder (BOU), Colorado (Fig. 2A; Munch et al., 2020). The same conductivity can also be obtained by other variable combinations of temperature and water content, leading to different interpretations of W, e.g., as a hot region (up to 2500 K) in the dry lower MTZ (0.1–0.3 wt% water) or as a cold (1250 K) yet nearly water-saturated region (e.g., 1.5 wt% water) (Yoshino et al., 2008). However, in both interpretations, the derived temperature should cause an obvious anomaly in seismic wave velocity (Karato, 2011), which has not been seismically imaged (Sigloch, 2011; Schmandt and Lin, 2014; Wang et al., 2019). In particular, the first interpretation indicates the occurrence of partial melting (Iwamori and Nakakuki, 2013), which can decrease the seismic wave velocity. Therefore, the crucial constraint from velocity imaging rules out the temperature effect and strengthens the conclusion that anomaly W represents a water-ponding region in the lower MTZ that has a high-water content (~0.8 wt%; Fig. 2A).

Water could affect and therefore be recognized by the variation in the Clapeyron slope or the sharpness of the transition interfaces within the MTZ (Litasov et al., 2005; Tauzin et al., 2017). On this basis, pervasively heterogeneous water distributions have been inferred beneath North America, and this extensive hydration has been explained by a long-term stirring of the background mantle by the mixing of oceanic crust with harzburgite convected by upwelling from the LM (Cao and Levander, 2010; Wang et al., 2019). However, in our electrical model, two more water-enriched regions beneath the southern United States were identified, with the one located beneath Colorado containing at least two seismic negative scattering phases in the MTZ (Fig. 1D) that have been attributed to the presence of water (Tauzin et al., 2017). In the same region, an ~35 km depression of the 660 km discontinuity (Fig. 1D; Cao and Levander, 2010) requires ~2 wt% water in the lower MTZ (Litasov et al., 2005), which can also reproduce our observed elevated conductivity.

Overlapping anomaly W, there is a spectacular inverted triangle–shaped, high-velocity anomaly. Its western arm stretches upward to the subduction zone of the Gorda plate (Schmandt and Lin, 2014) and thus represents the subducted Farallon slab (Fig. 1D; Cao and Levander, 2010; Sigloch, 2011). This overlapping geometry leads us to consider that the lower-mantle portion of inverted triangle W is an electrical manifestation of the slab. Given that the P-T conditions of the Farallon slab in the LM are beyond the DHMS stability field, an alternative interpretation is needed for this anomaly. Stishovite contained in a subducted slab is stable at pressures up to 70 GPa along the mantle geotherm (Lin et al., 2019, 2022) and could thus survive within the slab. Stishovite can accommodate water contents up to 3.2–8.4 wt% (Nisr et al., 2020; Lin et al., 2022) and can hence deliver water to the LM. The velocity anomaly corresponding to the subducted slab indicates that a large amount of stishovite has stalled in the uppermost LM (Fig. 1D). The lithospheric structure of the Farallon slab shows that stishovite may account for ~5 vol% of the slab (see the “Methods” section of the Supplemental Material; Lin et al., 2019). Our modeling using the conductivity of the lower portion of anomaly W suggests that the stishovite in the anomaly region contains 2.5–8.0 wt% water (Figs. 2B and 2C; Fig. S9B; see the “Methods” section of the Supplemental Material).

In the subduction zone of the Farallon slab, electrical conductivity structures indicate dewatering of the slab beneath the forearc down to 40 km depth, and deeper arc-front dehydration beneath the arc volcanoes in the central Cascade area (McGary et al., 2014; Egbert et al., 2022). The occurrence of further deeper-slab dehydration in the upper mantle beneath the back-arc region has also been revealed by a deep electrical resistivity region across the northwestern United States (Meqbel et al., 2014), as supported by the following lines of evidence: (1) the mantle wedge overriding the slab is one of the most hydrated zones across the circum-Pacific subduction zone (Halpaap et al., 2019); (2) a largescale low-velocity structure in the upper mantle located approximately beneath the Rocky Mountains of Colorado has been attributed to hydration-induced lithospheric melting (Humphreys et al., 2003); and (3) the hydrous MTZ beneath the southwestern United States is inferred to represent the foundered hydrated Cascade arc root (Yu et al., 2020). These research results, together with our electrical model, suggest that the warm slab acts as a water transporter along its subduction pathway down to the LM (Fig. 3).

The proposed water transport system is regulated by the phase relations of hydrous minerals (Ohtani, 2020) and controlled by the corresponding dehydration and rehydration reactions in the subduction pathway (Stern, 2002; van Keken et al., 2011). Slab water participating in the process is initially contained in oceanic sediments, basaltic crust, and serpentinized lithospheric mantle (Stern, 2002). In the so-called “subduction factory,” the thermal structure of the Farallon slab (van Keken et al., 2011) causes the hydrous phases to undergo a series of low-degrade dehydration and rehydration processes at shallow depths, forming the series of observed electrical anomalies in the uppermost upper mantle. Most of the newly formed hydrous minerals are unstable and therefore absent in the deep upper mantle. However, starting at similar depths, DHMSs and hydrous stishovite can both form in the subducted slab and transport water. DHMSs are derived from peridotitic lithosphere, whereas stishovite is derived from basaltic components and is almost dry when it forms (Ohtani, 2020; Lin et al., 2022). For a warm slab, the P-T conditions for the stability field of DHMSs may be very narrow. Hence, as the slab descends farther into the upper mantle, the P-T conditions will exceed the stability boundary of DHMSs (Ohtani, 2020) and therefore trigger extensive dehydration reactions of these phases, forcing them to release water that hydrates the mantle wedge. Stishovites in the subducting slab can become rehydrated via this through-slab water percolation. As they cross the MTZ, these stishovites can be further hydrated by the top melting layer of the MTZ (Toffelmier and Tyburczy, 2007). After entering the LM, the high water-storage capacity of stishovites enables them to retain this water, forming the water reservoir imaged in the present study.

Although the water-storage capacity of stishovite remains poorly constrained, the mean value (~5.5 wt%) of experimental data (Nisr et al., 2020; Lin et al., 2022) is likely the best estimate. This value is within the range of the water content of stishovite estimated in this study (Fig. 2B). Stishovites could therefore release some of the water. The presence of water could decrease the solidus of the lowermantle minerals (Iwamori and Nakakuki, 2013), resulting in the formation of a slightly buoyant hydrous melt, which will percolate upward and return water to the MTZ (Schmandt et al., 2014). Part of this water will be absorbed by ringwoodite, forming the upper portion of anomaly W. Although the actual percolation mechanism remains debated (Karato, 2011), the transport time is likely to be geologically short, possibly no longer than the period of subduction of the Juan de Fuca plate.

We used a particular inversion strategy to obtain a high-resolution electrical conductivity model of the region beneath the western United States. The model reveals a conductive feature crossing the 660 km mantle discontinuity. We interpret this feature as a water reservoir in the lower MTZ and uppermost LM. Together with other research results, our model exposes the entire water transport and recycling system down to the uppermost LM associated with warm slabs, including upper-mantle dehydration and rehydration, trans-MTZ transport, lower-mantle ponding, mantle upwelling into the MTZ, and subsequent partial melting at the upper interface of the MTZ.

We are grateful to Qingyang Hu and Chen Cai for manuscript reviews, Wenliang Xu at Jilin University for insightful discussions, and Jilin Kingti Geoexploration Tech, Ltd (Changchun, China) for the computational resources. This work was supported by the National Natural Science Foundation of China (grants 42074080 and 42130302 to A.-H. Weng, and 42204076 to S.-W. Li), and by the Program for Jilin University Science and Technology Innovative Research Team (grant 2021TD-050).