The timescales of plume generation caused by continental aggregation

To understand the thermal evolution of the mantle following the aggregation of non-subductable thick continental lithosphere, we study a numerical model in which a supercontinent, simulated by high viscosity raft, HVR, covers a part of the top surface of a convection layer. We model infinite Prandtl number convection either in a three-dimensional (3D) spherical shell, 3D rectangular box (aspect ratios: 8 and 4) or two-dimensional (2D) rectangular box (aspect ratio: 8) and except for the HVR, we specify a constant viscosity. The KVR, which has a viscosity higher than that of its surrounding, is instantaneously placed on the top surface of a well-developed convection layer and its position is fixed. Our results from 3D spherical shell cases with and without phase transitions show the emergence of a large plume characterized by a long wavelength thermal anomaly (a degree one pattern) for a Pangea-like geometry. We analyze the volume averaged temperature under the HVR (=(T-C]) the remaining (oceanic) area (=(T-O]) and total area (=(T-M]) to determine the timescale of plume generation. The difference between [T-C] and [T-O](=Delta T-CO) and [T-M] show the existence of two characteristic timescales. Delta T-CO exhibits an initial rapid increase and may become constant or continue to gradually increase. Meanwhile, [T-M] shows a similar behavior but with a longer timescale. We find that these timescales associated with the increase of Delta T-CO and [T-M] can be attributed to the formation of large scale flow (i.e. plume) and response of the whole system to the emplacement of the HVR, respectively. For 3D spherical cases, we find that the timescale of plume generation is 1-2 Gyr, if the Rayleigh number is 10(6). To determine the effects of the viscosity of the HVR, 2D versus 3D modeling and the effects of the internal heating, we have also studied 2D and 3D rectangular box cases. A factor of about two variation exists in the timescale of plume generation. It appears that the timescale becomes greater for a smaller amount of internal heating. This may be attributed to the time-dependent flow caused by the internal heating. For 2D cases, we find that the timescale of the high Rayleigh number (10(7)) case is shortened by a factor of three to five when compared to the Ra = 10(6) case, which is consistent with the simple boundary layer theory. This may imply that well-developed plumes may arise with the timescale of 0.2 to 0.4 Gyr (for Ra = 10(7)). (C) 2000 Elsevier Science B.V. All rights reserved.