Small amounts of melt are continuously generated by dynamic processes in the Earth's upper mantle, for example at mid-ocean ridges and subduction zones. The melt is initially distributed between the crystalline grains of the partially molten rock. Once melt is present, the physical properties of the rock, and with them the continuing melting process, are influenced by the grain scale melt geometry. In particular, this intergranular melt geometry is a key factor for the strength of partially molten regions, for the eventual segregation of the melt from the matrix by porous flow, and for our ability to detect partially molten regions by seismic or electromagnetic imaging. Since the melt generated at depth frequently leads to volcanic eruptions at the surface with consequences for our environment, understanding the melting process is important for recognizing and mitigating the attendant hazards.
In this project, the investigator proposes to determine the melt geometry in partially molten rocks by simulating upper mantle conditions in the laboratory. An important aspect of the experiments is that they will be conducted in a piston cylinder apparatus, which has the advantage that relatively long durations at high temperatures and pressures can be achieved. The experimental samples will be examined by repeated high resolution imaging (by Field Emission Scanning Electron Microscope) of a planar section and removal of a thin layer of material in order to reconstruct the three-dimensional pore geometry at the sub-micron scale. In particular, this procedure will address the question of whether wetted two-grain boundaries exist and how numerous they are. At a fixed melt fraction, wetted two-grain boundaries will have the largest effect on seismic velocities and attenuation, and the rheology. The effects of small amounts of water, as well as grain size effects on the melt distribution will also be examined.
Volcanic eruptions deliver molten rock to the surface of the Earth and release volatiles into the atmosphere. More than three quarters of the volcanic activity takes place under water, along ridges where new oceanic plates are generated. The molten rock originates at depths of 60 - 70 km in the Earth’s upper mantle, where upwelling mantle begins to melt. The melt constitutes only a small fraction of the upwelling material, and is distributed between solid crystalline grains forming a matrix. Because melt has very different properties compared to the solid matrix, its grain-scale distribution exerts a large influence on the behavior of the upwelling system. This grain-scale distribution was investigated by laboratory experiments replicating conditions in the Earth’s upper mantle in terms of pressure and temperature. An interesting aspect is that features at scales of tens of nanometers influence the dynamics of upwelling mantle at scales of tens of kilometers. Because of the small scale of the features of interest high resolution imaging is a key aspect of the evaluation of the experiments. Using a novel approach for high resolution imaging and serial sectioning of experimentally made samples, we have been able to reconstruct the three dimensional melt geometry for small amounts of melt. The Figure shows the melt geometry of one sample. The melt is shown in gray, the solid grains can be envisioned by the shape of the surrounding melt, which forms a mold for the grains. This figure illustrates that even at a content of 3.5%, melt coats or wets many grain boundaries. This will weaken the rock during upwelling when melt is present, and make partially molten regions more visible for seismic imaging. Further experiments showed how this geometry changes with melt content and size of the grain of the solid matrix. Determining the changes of the geometry with melt content is important because it allows estimation of the residual melt content of partially molten regions during the melting process. The results suggest that the matrix rapidly becomes more permeable (allowing segregation of the melt) at melt contents between 1 and 2%. Investigating the grain size dependence is necessary because experimental constraints require working with matrix grain sizes that are much smaller than grain sizes inferred for the mantle. The results show that the tendency of melt to wet grain boundaries increases with increasing grain size, so that wetted grain boundaries should be common in the mantle at melt contents of 1% or less. Partially molten regions in the Earth are therefore an important factor for the dynamics of mantle processes.
