Skip Navigation | ANU Home | Search ANU | Directories
The Australian National University
Research School of Earth Sciences
Printer Friendly Version of this Document
Untitled Document


Thermal history and flow patterns in the mantle wedge above subducting plates.

C. Kincaid* and R. W. Griffiths Prof.

Chris Kincaid spent 2002 as a Visiting Fellow at RSES on sabbatical leave from the Graduate School of Oceanography, University of Rhode Island.

The aim of his visit was to carry out the first ever laboratory experiments modelling both the thermal evolution of subducting lithosphere and the three-dimensional circulation in the mantle wedge (the region between the sinking plate and the over-riding lithosphere) during 'rollback' subduction. The experiments were carried out in the new temperature-isolation facility in the Geophysical Fluid Dynamics laboratory. In the experiments a tank of glucose syrup was used to simulate the Earth's mantle (figure 4).

A plate of phenolic, or laminated plastic, was used to represent the oceanic lithosphere. Different modes of subduction were represented by driving the plate into the fluid using up to three hydraulic pistons to produce three components of motion: down-dip and rollback plate motions, and a mode of plate steepening with time. The plate was chilled to 5 degrees C and at the start of an experiment the tank of syrup is moved into the cold room so that a thermal boundary layer grows at the surface of the fluid. Over forty experiments were run covering a range in subduction parameters, including different thermal boundary layer thicknesses and the effect of back-arc spreading.

Figure 4. A photograph of the apparatus used for experiments with subducting plates. with the tank of glucose syrup that simulates the mantle. Three hydraulic piston-cylinders generate down-dip and roll-back velocities, and a variable dip angle. Rates are controlled with precision flow meters in the hydraulic lines. A Plexiglass sheet overlies the wedge region of the fluid to simulate an overriding plate, and migrates with the trench. Six rows of thermocouples (5 per row) record plate surface temperatures. The experiment is carried out in a low-temperature laboratory after a cold boundary layer has developed on the surface of the syrup.

The results with and without rollback plate motion show very different flow patterns in the wedge. When there is no rollback motion, the plate edges heated up faster than the centre of plate segments whereas, with rollback motion, plate centres heated up faster than plate edges. This was a response to focusing of mantle flow towards the centreline of the plate. Flow trajectories became steeper in the wedge (more favourable for decompression melting) in response to rollback motion and back-arc spreading, and when the dip of a plate steepened as it sank.

The experiments with and without rollback sinking also revealed distinct patterns in return flow of material into the wedge from behind the sinking slab. This has implications for models of geochemical recycling in arcs. The work is particularly timely given the recent explosion of research papers appealing to rollback plate motion in reconciling patterns found in geochemical and seismic data.


Sources of heat and helium in the dynamic mantle.

G. F. Davies

Top of page

Work has continued on the task of reconciling geochemical and geophysical constraints on mantle structure, dynamics and history. Two particularly puzzling questions regarding the geochemical constraints are where sufficient radioactive heat sources can be located and where anomalously unradiogenic helium is coming from.

Each of these questions has become so enigmatic that radical possibilities need to be explored. The argument has been developed that the concentrations of radioactive heat sources in much of the "depleted" mantle are still quite uncertain, and in particular that they could be substantially higher than has usually been inferred. Three factors are invoked to support this contention: (a) gravitational segregation, within the upper mantle, of buoyant depleted material from denser former oceanic crust, which implies that the immediate source of MORB is more depleted than average "depleted" mantle, (b) a bias towards the most depleted oceanic basalt compositions in conventional estimates and (c) refreezing of melt forming in a heterogeneous source region, which could trap a significant proportion of heat sources in "depleted" residual regions. Factor (a) is evident in recent numerical models. Factor (c) is the least constrained but potentially the most important.

An argument is also being developed that residual primitive helium (helium-3) could persist in subducted remnants of early mafic crust. Although such crust would have been substantially degassed during its formation, there would also have been higher abundances of noble gases very early in Earth's history, so the net effect could be that a significant amount of primitive noble gases are retained. The early mafic crust remnants could persist at the base of the mantle as part of the anomalous D" zone, which is plausibly interpreted as a reservoir of foundered basaltic oceanic crust, and which transforms to dense phase assemblages under mantle pressures. Recent numerical models (below) provide some initial support for the preservation of ancient subducted material in this way.


Dynamics of the Hadean Mantle.

G. F. Davies

Top of page

Initial numerical models of thermo-chemical convection in the early, hot mantle have yielded a compositionally stratified mantle through a novel interaction of compositional buoyancies with viscosity stratification. This work extends the modelling reported last year, in which models at present mantle conditions yielded long residence times and some gravitational segregation of simulated subducted basaltic oceanic crust. The basaltic component is included in the models using a large number of tracer points. The models assume an upper mantle temperature of 1550 degrees C, appropriate for the pre-4 Ga (Hadean) mantle. Because of the high temperature, the mantle viscosity is about 15 times lower than the modern mantle.

The lower mantle is assumed to be about 30 times the viscosity of the upper mantle, for which there is good independent evidence at present. The low viscosity in the upper mantle permits relatively rapid settling of the heavier basaltic component, which sinks into the lower mantle. However the lower mantle viscosity is still high enough that segregation there is slow, so most of the basaltic material remains entrained in the circulation in the lower mantle. The result is that two compositional zones develop (Figure 5): a depleted but cooler upper mantle and an enriched but warmer lower mantle.

The interface between these zones is quite sharp, although its position varies in space and time: the cooler upper layer forms large intrusions into the lower layer.

Figure 5. Images from a numerical model of the Hadean mantle. Upper panel: temperature and streamlines. Lower panel: tracers plotted over the viscosity structure. Viscosity is temperature-dependent with a superimposed increase by a factor of 30 entering the lower mantle. Tracers simulate basaltic component (orange), which melts to form oceanic crust (green) before being subducted (black). Blue tracers have passed through the lower 20 km of the model, which is a potential plume-feeding zone. Tracers have a density excess of 70 kg/m3, which causes them to settle out of the low-viscosity upper mantle, but not out of the higher-viscosity lower mantle. Two "fluids" have developed, a cooler depleted fluid over a hotter, enriched fluid.

The evolution of this system remains to be explored. It is possible that the layers could episodically mix if the upper layer cools and the lower layer heats sufficiently. In the longer term the layering is expected to dissipate, since models run at present mantle conditions do not develop this layering. However if there is a time lag in the dissipation, it is possible that some remnant could persist in the deep mantle at present.

This is a potential explanation for evidence from seismic tomography for large anomalous volumes below Africa and the Pacific (sometimes misleadingly called superplumes). The models potentially have important implications for the chronology of the accumulation of continental crust, for the geochemical evolution of the mantle and for interpreting the apparent episodic accumulation of the continental crust.


The structure of mantle plumes sheared by mantle flow.

C. Meriaux and R.C. Kerr

Top of page

This year we continued a series of laboratory experiments aimed at understanding the behaviour of "heterogeneities" rising in mantle plumes as they are sheared by overlying plate motion. In the experiments, a hot plume was generated by a circular hot plate on the base of a cylindrical tank of glycerol. The plume was then sheared by a horizontally rotating lid at the surface of the fluid. To observe the flow, two small tubes released dyed glycerol at opposite sides of the hot plate. The flow is characterised by 5 dimensionless parameters: a Rayleigh number, a Prandtl number, an aspect ratio, a viscosity ratio, and a ratio of the lid velocity to the plume rise velocity.

Our experiments are systematically varying the values of these parameters to elucidate how each affects the flow field. With no shear, we observed a vertical conduit where there was simple stretching and thinning of heterogeneities. When a weak shear was imposed, the plume was slightly tilted, and the dye stream from the upstream edge of the plate bifurcated dramatically. The hottest fluid rose close to the lid and was then dragged downstream over the downstream dye, while the cooler fluid initially flowed upstream before it then flowed down, out, and around the laterally spreading plume (figure 6a). When a strong shear was imposed, the plume was strongly tilted, which led to a cross-stream recirculation in the plume and rotation of the upstream dye to the sides of plume, while the downstream dye was dragged close to the surface (figure 6b).

Figure 6: Overhead views of sheared thermal plumes. Left: weak shear. Right: strong shear.


Patterns of Solidification in Long Channelized Lava Flows.

R.W. Griffiths, R.C. Kerr and K.V. Cashman

Top of page

In a project begun last year, and in collaboration with volcanologist Professor K.V. Cashman of the University of Oregon, we are exploring the behaviour of basaltic lava flowing through long channels, with the aim of understanding the factors influencing the amount of solidification of the flow, and the resulting flow behaviour.

Basaltic lava from large eruptions on volcanoes such as Hawaii is often channelled into rapidly-flowing rivers of melt. Channels are commonly 10-100 m wide and of the order of 10 km in length, with the flow being 2-10 m deep during its active period. Much longer channels, up to 750 km, were important in transporting lavas from large prehistoric flood basalt eruptions and spreading them over broad areas of the Earth. Lava tubes are also common. Tubes arise when the lava surface solidifies and forms a connected roof over the flow. The roof greatly reduces the rate of heat loss and hence enables the lava to flow much greater distances then would be possible if the roof were continuously disrupted or if there were little solid on the surface. These dynamics may be central to the interpretation of the rates and volumes of prehistoric eruptions. The physical processes that govern the formation of channels and the cooling of continued flow through them are complex.

Figure 7a

Figure 7b

Figure 7. Photographs of a steady state regime of solidification in a sloping channel flow of PEG cooled from above: a) from the side and 0 to 30 cm from the vent; b) from above, far downslope (0.8 m to 1.1 m from the vent). Channel is 8 cm wide. Solid wax is white, liquid PEG is transparent and the back wall and base of the tank are black. Flow is from left to right. The photographs correspond to distances from the entrance sluice gate of 0 to 60cm.

The experiments use polyethylene glycol (PEG) flows generated under cold water and flowing down a 3m-long, inclined channel. Our flows are laminar, having Reynolds numbers of 0.2 Ð 70 based on flow depth and centre-line speed, thus covering the range estimated for basalt channels. For a constant source volume flux we have found two steady state regimes, depending on the flow velocity and the temperatures of the flow and water relative to the freezing temperature of the PEG flow. Our results indicate that the condition for tubes is (U0ts/H)(Ra/R0)^1/3 < 25 ± 5, where U0 is the surface speed without cooling at the centre-line, ts is the predicted time for onset of solidification (in an idealised initial-value problem), W is the channel width, Ra is a Rayleigh number and R0 = 100 is a constant. At larger values of this parameter grouping, solid forms a raft along the channel centre-line on the flow surface, with open shear zones along the edges (figure 7). In basalt flows, the side wall shear zones expose incandescent material at the surface (figure 8) and much of the heat loss is from these zones.

Figure 8. A basaltic lava channel in Hawaii. The red stripes along the edges of the flow indicate exposure of hotter melt in those regions, whereas the dark surface along the centre of the channel is cooler solid. The flow is about 20 m wide.


Gravitational collapse of yield-strength materials - rapid emplacement of andesite flows.

A. Lyman, R.W. Griffiths and R.C. Kerr

Top of page

Motivated by the question of how rapidly large andesite flows may have been erupted and emplaced, new experiments were designed to explore the behaviour of lava flows having high effusion rates and intermediate composition. Earlier laboratory modelling using slurries having a yield-strength as well as surface cooling and solidification were focussed on slow effusion rates and the formation of highly silicic lava domes. In the new experiments mixtures of kaolin clay and polyethylene glycol (PEG) wax are suddenly released by removing a dam wall at the end of an inclined channel 1.8 m long and 0.15 m wide. The horizontal distance travelled and the elapsed times are recorded and, in the case of isothermal flows, are being compared with theoretical predictions. The first experiments allowed us to determine the range of ratios of clay to wax best suited to model lava flows.