Plume zonation and modelling with "Fluent"


Basalts from hotspot ocean islands (e.g. the Hawaiian and Galàpagos Islands) are believed to be formed by melting of mantle plumes, which likely emanate from the core-mantle boundary layer. While it has long been recognized that lavas in these islands have different compositions, many geochemists believe that the variability within oceanic islands basalts reflects a source characteristic. The project is aimed at assessing whether a distribution of "heterogeneities" entrained in a plume conduit sheared by plate motion will maintain its identity. For transport in a vertical conduit, we expect simple stretching and thinning of heterogeneities. When horizontal shearing motion is superimposed on the buoyancy-driven vertical motion, the conduit is carried horizontally, becoming bent over. This leads to a complex recirculation in the conduit and entrainment of surrounding fluid. Thus "heterogeneities" at the source can be stirred, possibly to the extent that heterogeneities observed at the surface no longer reflect their distribution at the source. The amount of stirring depends on the buoyancy flux of the plume and the imposed shear, which together determine the angle of tilt of the conduit.


Principal Investigator

Catherine A. Meriaux
Research School of Earth Sciences
Australian National University

Project

x47

Facilities Used

SC

 

 

RFCD Codes

240502, 260200


Significant Achievements, Anticipated Outcomes and Future Work

The initial testing phase of the project has been completed and confirmed that the entrainment of surrounding fluid by a thermal plume which is sheared by a background flow can be investigated using the Gambit and Fluent software. The mean flow structure and dynamics of the plume were successfully captured by our first three-dimensional model and entrainment of the surrounding fluid by the thermal plume was observed.

However, a detailed analysis of the results showed that the computation has to be optimized. First, the buoyancy flux of the plume was not constant although the computation had presumably converged to the steady-state solution. We think that the convergence criterion defined by default in Fluent may actually not be appropriate in our case. Thus, we are currently testing it. Alternatively, the problem could result from the use of a mesh too coarse for the problem.

The second issue which arose from these first calculations is that the horizontal non-uniform shear inherent to our geometry introduced a non-symmetric component in the characteristics of the vortices present in the plume conduit which is driven by the entrainment. This could not be a posteriori corrected in Fluent. Therefore, we are now inclined to change the geometry of our model such that, without altering the physics of the problem, it ensures that the shear motion is horizontally uniform.

As soon as the problem regarding the buoyancy flux has been identified and fixed, a new series of experiments will start in the new geometry. We expect that this new geometry will allow for a better characterization of the entrainment into the plume conduit. The impact of stirring will be assessed by visualizing both the vorticity structure in the plume conduit and the pathlines of massless particles released in the problem domain. To complete our study, systematic calculations exploring a range of plume buoyancy fluxes and horizontal shearing motions will then be performed.

 

Computational Techniques Used

The calculations are performed by the computational fluid dynamics (CFD) packages Gambit and Fluent. The geometry and mesh of the model are initially built in Gambit while the flow and heat transfer are solved in Fluent. Fluent solvers utilize a finite-volume method based on fully unstructured meshes. Our application uses the segregated solver with a second-order implicit scheme (Gauss-Seidel-AMG) and the pressure-velocity coupling used is the PISO algorithm. The convection problem is solved within the Boussinesq approximation. Ultimately, we will take advantage of parallel processing which the Fluent solver allows for since our calculations are long and expensive.