Van Keken, P.E.
Evolution of starting mantle plumes: a comparison between laboratory and numerical models.
Earth and Planetary Science Letters, 148, 1-14, 1997.

Abstract

Laboratory experiments on the initiation and development of mantle plumes suggest the formation of a bulbous plume head that grows in time by entrainment of surrounding material, followed by a thin feeder conduit. This type of plume is generally not seen in numerical models of mantle convection. To better determine the possible differences between numerical and laboratory experiments, we have numerically reproduced a published laboratory experiment on plume initiation. Theoretical predictions for the growth of the plume head have been verified and the experiments allowed for a new determination of a similarity constant. Models with a constant viscosity fluid show much less entrainment. Simulations using a plume rising from a thermal boundary layer with an olivine-type temperature-dependence show that most of the source material is contained in the plume head and that the conduit is largely composed of entrained material. Simulations with a more realistic non-Newtonian rheology show that plumes can rise much faster through the mantle, which will lead to higher temperatures of the plume head at the Earth's surface compared to what experiments with Newtonian rheology suggest.

Reprint (pdf)

Additional materials:

Figure 1:Plume initiation is studied numerically using a cylindrical, axisymmetric geometry.The mesh indicates the computational domain with dimensions 0.3 by 1 (corresponding to laboratory values of 14.4 by 48).

Figure 2. Snapshots of the evolution of the thermal plume shown every 100 s (laboratory scaling).The grey scales indicate temperature in degrees Celsius.The transition from source to surrounding material is indicated by the solid line.The inset focuses on the plume head at 400 s.

Griffiths and Campbell, Figure 1

cg90.mvSGI movie (1 frame = 10 lab seconds) 500 k
cg90-l.mvSGI movie (1 frame = 1 lab second) 5 M

Figure 3. a) A comparison of the growth of the volume of the plume head between theoretical predictions (dashed lines), two GC90 laboratory experiments (large open circles and crosses) and the numerical simulation of Figure 2 (solid circles). The straight bold line indicates the amount of volume that is supplied through the feeder conduit. The theoretical curves are obtained by numerical integration of (8) for three different values of the similarity constant C. The arrows indicate the time of transition from the linear to superquadratic regime.
b) Same, but now only for two experiments in a log-log diagram. Neither of the simulations reaches the superquadratic asymptotic regime.

Figure 4. A comparison of plume shape and size with different Ra and rheology. a) As Figure 2, but now with Ra=2.2e6. The plume rises more slowly and becomes much larger. b) As Figure 2, but now with Ra=4.4e7. The opposite effect occurs. Notice that the morphology of the plume head at comparable heights is similar for Figures 2, 4a and 4b.c) As Figure 2, but now with isoviscous rheology (Q=0). The plume conduit remains broaderand cools faster. The efficiency of recirculation in the plume head is much reduced.

fig5.jpg

Figure 5. The growth of the volume of the plume head as a function of time for the simulations shown in Figures 2 and 4. The models with temperature-dependent viscosity follow the theoretical predictions well if C<=0.5. The volume of the isoviscous plume head is always smaller than that supplied by the feeder conduit.

Figure 6. Simulation of a plume erupting from a thermal boundary layer. Snapshots are at 14 Myr intervals. The top markerchain indicates the position in the boundary layer where T=0.9. The other twomarkerchains are initially equidistantly below the top markerchain.

Figure 7. Same as Figure 6, but now with a temperature- and stress-dependent rheology. The strong strain-rate thinning effects cause detachment of the first plume head and the formation of a second one. The first snapshot is arbitrarily assigned t=0. The snapshots are not at equidistant time-intervals. The top of the model is indicated by the horizontal line.

Peter van Keken (keken@umich.edu) Last updated: February 2002