Dynamics and structure of thermo-chemical mantle plumes: Are numerical models consistent with observations?

According to widely accepted models, plumes ascend from the core-mantle boundary and cause massive melting when they reach the base of the lithosphere. Most of these models consider plumes as purely thermal and predict flattening of the plume head to a disk-like structure, thin plume tails with a radius on the scale of 100 km and kilometer-scale topographic uplift before and during the eruption of flood basalts. However, several paleogeographic and paleotectonic field studies indicate significantly smaller surface uplift during the development of many LIPs, and seismic imaging reveals thicker plume tails as well as a more complex plume structure in the upper mantle including broad low-velocity anomalies up to 400 km depth and elongated low-velocity fingers. Moreover, geochemical data indicate a plume composition that differs from that of the average mantle and recent geodynamic models of plumes in the upper mantle show that plumes containing a large fraction of eclogite and therefore having very low buoyancy can explain the observations much better. Nevertheless, the question remains how such a low-buoyancy plume can rise through the whole mantle and how this ascent affects its dynamics. We perform numerical experiments in 2D axisymmetric geometry to systematically study the dynamics of the plume ascent as well as 2D and 3D models with prescribed velocity at the upper boundary to investigate the interaction between plume- and plate-driven flow. For that purpose, we use modified versions of the finite-element codes Citcom and Aspect. Our models employ complex material properties incorporating phase transitions with the accompanying density changes, Clapeyron slopes and latent heat effects for the peridotite and eclogite phase, mantle compressibility and a highly temperature- and depth-dependent viscosity. We study under which conditions (excess temperature, plume volume and eclogite content) thermo-chemical plumes can ascend through the whole mantle and what structures they form in the upper mantle. Modelling shows that high plume temperature and/or volume together with low content of eclogite result in plumes directly advancing to the base of the lithosphere. Due to the high eclogite density in a region between 300 and 400 km depth, plumes with slightly lower buoyancy (due to higher content of eclogite or lower temperature) pond there and form pools or a second layer of hot material. These structures become asymmetric when the plume interacts with the quickly moving overlying plate. Further reduction of buoyancy leads to plumes accumulating at 300 - 400 km depth range and never approaching the base of the lithosphere, but instead heating the adjacent mantle material above to form smaller secondary plumes. Our models also suggest that thermo-chemical plumes ascend in the mantle much slower compared to thermal plumes, have thicker plume tails and cause a much smaller (hundred meters scale) surface uplift. The conversion of plume excess temperatures to anomalies in seismic velocity shows that thermo-chemical low-buoyancy plumes can explain a variety of features observed by seismic tomography much better than purely thermal plumes.