“second asthenosphere” was recovered in inversions constrained by regional (oceanic) geoid anomalies (Kido et al., 1998). The effect of the low‐viscosity channel can modify predictions associated with GIA observables Milne et al. (1998), and in joint inversions of GIA, misfits can be significantly reduced for models that include such a low‐viscosity notch (Mitrovica and Forte, 2004). Finally, in global geodynamic models with prescribed plate motions, the behavior of slabs is broadly consistent with observations of stagnation when such a feature is included (Mao and Zhong, 2018; Lourenço and Rudolph, in review).
An increase in viscosity in the mid‐mantle or viscosity “hill,” which is a feature common to all of our viscosity inversions, has been suggested on the basis of geophysical inversions, and several potential mechanisms exist to explain such a feature. An increase in viscosity below 650 km depth has been recovered in many inversions constrained by the long wavelength geoid and GIA observables (e.g., King and Masters, 1992; Mitrovica and Forte, 1997; Forte and Mitrovica, 2001; Rudolph et al., 2015). An increase in viscosity would be expected to slow sinking slabs (Morra et al., 2010) and affect the dynamics of plumes. The correlation between subduction history and tomographic models has been used to test whether slabs sink at a uniform rate in the lower mantle. A recent study of the similarity between convergence patterns in plate reconstructions and patterns of mantle lateral heterogeneity from an average of VS tomographic models suggests that the data can neither confirm nor reject the possibility of a change in viscosity below 600 km (Domeier et al., 2016). On the other hand, an analysis of a catalog that relates imaged fast anomalies to specific subduction events does find evidence that the rate of slab sinking decreases across a “slab deceleration zone” between 650–1500 km (van der Meer et al., 2018); one explanation for such a deceleration zone is the increase in viscosity in the shallow lower mantle seen in all of our inverted viscosity profiles.
Several mechanisms exist that could produce an increase in viscosity in the mid‐mantle. Marquardt and Miyagi (2015) measured the strength of ferropericlase at pressures of 20–60 GPa (600–1,000 km) and observed an increase in strength across this range of pressures. Though ferropericlase is a minor modal component of the lower mantle, it could become rheologically limiting if organized into sheets within rapidly deforming regions, an idea supported by experiments with two‐phase analog materials (Kaercher et al., 2016) and with bridgmanite‐magnesiowüstite mixtures (Girard et al., 2016). If the lower mantle rheology is determined by the arrangement of distinct mineral phases, we expect history‐dependence and anisotropy of viscosity (Thielmann et al., 2020), further confounding our interpretations of viscosity in inversions. An increase in the viscosity of ferropericlase is also supported by experimental determinations of the melting temperature at mantle pressures (Deng and Lee, 2017), which show a local maximum in melting temperature for pressures near 40 GPa (1,000 km). Changes in the proportionation of iron could also alter the viscosity of bridgmanite across a depth range consistent with the inferred mid‐mantle viscosity increase. Shim et al. (2017) suggested that at depths of 1,100–1,700 km, an increase in the proportionation of iron into ferropericlase could depress the melting point of bridgmanite, increasing the viscosity predicted using homologous temperature scaling. These various mechanisms are not mutually exclusive and could operate in concert to produce an increase in viscosity near 1,000 km. Finally, we note that the deformation mechanisms of even single phases within the lower mantle remain uncertain. While the lower mantle has long been thought to deform by diffusion creep due to absence of seismic anisotropy at most lower mantle depths, recent calculations suggest that another deformation mechanism – pure climb creep, which is insensitive to grain size and produces no seismic anisotropy – may be active in bridgmanite at lower mantle conditions (Boioli et al., 2017).
1.5 CONCLUSIONS
We analyzed the long‐wavelength structure of four recent global tomographic models and compared the features of these models with predicted structures in five geodynamic models that incorporate surface velocity constraints from plate reconstructions. The long‐wavelength radial correlation functions of SEMUCB‐WM1 and SEISGLOB2 show strong evidence for a change in radial correlation structure near 1,000 km depth, whereas the most abrupt change in the RCFs for S362ANI+M and GLAD‐M15 occur at 650 km depth. The change in the RCF reflects a change in the pattern of long‐wavelength structure between the lower mantle and transition zone. The transition zone structure is correlated with more recent subduction history, whereas the long‐wavelength, lower‐mantle structure is more similar to ancient subduction history, in agreement with previous work. The long‐wavelength transition zone structure is dominated by seismically fast anomalies in the Western Pacific, and especially beneath the Philippine Sea Plate. This suggests that the change in the pattern of long‐wavelength heterogeneity in the transition zone and shallow lower mantle is controlled by subduction history and by the interaction of slabs with mantle phase transitions and rheological changes. The depth of changes in the pattern of heterogeneity in the models (and the associated RCFs) is sensitive to the data used to constrain the inversions and the radial parameterizations. Future studies that investigate whether these changes in long‐wavelength structure occur at 650 km depth or somewhat deeper within the lower mantle will have important implications for our understanding of mantle structure and dynamics. All of the tomographic models show a local minimum in the spectral slope at or slightly above 650 km, indicating concentration of power at longer wavelengths within the transition zone. This feature is most consistent with a global geodynamic model that includes a weakly endothermic (–2 MPa/K) phase transition at 660 km depth and a low viscosity channel below 660 km and a viscosity increase in the shallow lower mantle. New inferences of the viscosity profile (Figure 1.6) using both a whole‐mantle density model from full‐spectrum tomography (Moulik and Ekström, 2016) and a scaled VS model (SEMUCB‐WM1, French and Romanowicz (2014)), recover viscosity profiles that are compatible with the presence of a low‐viscosity channel below 660 km depth and a viscosity maximum in the mid‐mantle.
ACKNOWLEDGMENTS
The authors thank John Hernlund and Fred Richards for their reviews, which improved the quality and clarity of the manuscript. The authors thank Ebru Bozdağ for providing the GLAD‐M15 tomographic model and for helpful discussions about this model. All of the authors acknowledge support from the National Science Foundation through NSF Geophysics grant EAR‐1825104, and MLR acknowledges NSF CSEDI grant EAR‐1800450. Computational resources were provided through NSF Major Research Instrumentation grant DMS‐1624776 to Portland State University and by UC Davis. VL acknowledges support from the Packard Foundation.
REFERENCES
1 Austermann, J., B. T. Kaye, J. X. Mitrovica, & P. Huybers (2014). A statistical analysis of the correlation between large igneous provinces and lower mantle seismic structure. Geophysical Journal International, 197(1), 1–9, doi:10.1093/gji/ggt500.
2 Ballmer, M. D., N. C. Schmerr, T. Nakagawa, & J. Ritsema (2015). Compositional mantle layering revealed by slab stagnation at ∼1000‐km depth. Science Advances, 1(11), e1500,815, doi:10.1126/sciadv.1500815.
3 Becker, T. W., & L. Boschi (2002). A comparison of tomographic and geodynamic mantle models. Geochemistry, Geophysics, Geosystems, 3(1), 1003–n/a, doi:10.1029/2001GC000168.
4 Becker, T. W., C. O’Neill, & B. Steinberger (2014). HC, a global mantle circulation solver.
5 Boioli, F., P. Carrez, P. Cordier, B. Devincre, K. Gouriet, P. Hirel, A. Kraych, & S. Ritterbex (2017). Pure climb creep mechanism drives flow in Earth’s lower mantle. Science Advances, 3(3), e1601,958, doi:10.1126/sciadv.1601958.
6 Boschi, L., & A. M. Dziewonski (1999). High‐ and low‐resolution images of the Earth’s mantle: Implications of different approaches to tomographic modeling. Journal of Geophysical Research: Solid Earth, 104(B11), 25,567–25,594, doi:10.1029/1999JB900166.
7 Bower, D. J., M. Gurnis, & M. Seton (2013). Lower mantle structure from paleogeographically constrained dynamic Earth models. Geochemistry, Geophysics, Geosystems, 14(1), 44–63, doi:10.1029/2012GC004267.
8 Bozdağ,