Principal Investigator Matthew J Cordery Project r08

Geophysical Fluid Dynamics Group, Machine VP

Research School of Earth Sciences

Fluid Dynamics of Terrestrial Mantle Plumes

This project involves the numerical simulation of terrestrial mantle plumes and their interaction with both the crust and lithosphere of the Earth. Mantle plumes are essentially large masses (~600 km in radius) of thermally buoyant rock in the Earth's mantle. When plumes reach the surface of the earth, they begin to melt and the resulting magma migrates to the surface. In the ocean basins, mantle plumes are known to be the source of such features as the Hawaiian and Bermuda Islands which are the result of large outpourings of lava onto the ocean floor. When a mantle plume rises beneath a continent, however, these large outpourings of lava are known as continental flood basalts. Examples of continental flood basalts are the Deccan Traps of India and the Columbia River Basalts of the American Pacific Northwest.

What are the basic questions addressed?

What are the timing, lateral extent, and volume of melting when a mantle plume reaches the continental lithosphere? How do these correlate with surface uplift and gravity? How are they affected by low viscosities in the plume itself and by a low viscosity channel which feeds hot material to the plume from near the Earth's outer core? Does the chemically distinct continental lithosphere melt and if so under what conditions?

What are the results to date and the future of the work?

The code is now operational and has been altered to work in a cylindrical axisymmetric geometry. Also, after major effort, a 2D adaptive regridding algorithm has been developed in order to better resolve fine-scale structure in the solutions. A variety of preconditioning algorithms were tried in order to the improve the performance of the element-by-element iterative algorithm used to solve for velocity and pressure in the presence of large viscosity contrasts. It appears that preconditioners more sophisticated than simple diagonal preconditioning are not vastly superior. More likely, some form of multigridding is required. Also, an incremental batch melting scheme has been implemented for both mantle-like material and eclogite. Some first order results from using this code have been:

(1) melting of a plume head composed solely of mantle material is extremely difficult if the plume rises beneath old (150 Ma) lithosphere. This is largely due to the fact that plume head temperature anomalies are relatively modest and the stiff lithosphere causes the plume head to stagnate at a depth below its solidus. When melting does occur, insufficient amounts are generated and over a time period much too long to be in agreement with observations.

(2) a plume head composed of 10% pure eclogite can generate reasonable volumes of melt but it appears that melting occurs over too long a time period as well. It may be that some other structural control (i.e. variations in lithosphere thickness, the presence of a plume tail) may be required to match all observations.

What computational techniques are used and why is a supercomputer required?

Convective motions of the solid earth are modelled using finite element techniques on an unstructured mesh. The mesh is adaptive and responds to changes in the solution with time. New grid parameters are determined using the superconvergent patch recovery method. Remeshing is performed using a 2D version of the advancing front method developed by workers in the aerospace industry. All systems of equations are solved using iterative conjugate gradient solvers which are highly vectorized due to our use of the element-by-element method of data storage. A supercomputer is required because large spatial resolutions are required to adequately resolve strong gradients in viscosity, temperature, composition and strain rate. The small spatial scales for the finite element mesh also require a large number of time steps.