The mechanism by which granitic melts are extracted from partially-melted crustal rocks is poorly understood. Experimental and theoretical models have shown that in equilibrated rock-melt systems, the viscosity of crustal melts is too high, and the permeability of the rock matrix is too low, to permit melt extraction. The constraint on the permeability of the rock matrix arises from the expectation that in texturally-equilibrated systems granitic melt occupies isolated pores, or a porosity connected by tube-like channels. In the past five years, recognition of this limitation has led many authors to suggest that externally-induced deformation perturbs the equilibrium rock-melt distribution so as to enhance the matrix permeability and facilitate melt extraction. An alternative cause of deformation is the volume change associated with the melting reaction itself. For many important crustal melting reactions, the volume change is positive, and this is a potential cause of micro-cracking. If the cracks generated by such reaction-induced embrittlement form a connected network, then melt extraction will be facilitated without, or in tandem with, external deformation. The intent of this study is to demonstrate the feasibility of this mechanism through melting experiments on natural rock samples.
The experimental starting material is a strongly foliated, but otherwise homogeneous, muscovite-bearing metaquartzite consisting of 88 vol% quartz and 12 vol% muscovite. The foliation is formed by the muscovite grains, which are 0.32 ± 0.1 mm in diameter and 0.04 ± 0.02 mm thick. Quartz occurs as equant grains 0.09 ± 0.02 mm in diameter. The experimental charges were 4.8 mm diameter cores, 6-10 mm long, cut either parallel or orthogonal to the foliation. The cores were dried at 340 K and sealed in platinum capsules. In some runs a 2-3 mm thick layer of a clean, sorted (0.125-0.150 mm), beach sand was added at one end of the charge to provide a drain for the melt. The experiments were performed in internally-heated Ar-pressurized vessels at 1123-1193 K and 8 kbar. At this pressure, the temperatures are estimated to be 100-150 K above the temperature of equilibrium melting. After each run, the charges were sectioned both parallel and orthogonal to the foliation, mounted in epoxy, chemically polished, and coated with carbon for electron microscopy. The most successful imaging technique for identifying the melt-filled cracks was found to be back-scattered electron microscopy with 7.5 nA beam current, a 15 kV accelerating voltage, and a working distance of 20 mm. In the resulting images, melt-filled cracks are easily distinguished from grain-boundaries and quench-induced cracks because the melt phase has a higher average atomic number than the quartz matrix and is therefore brighter. For purposes of characterizing the melt-filled crack population 2-4 images were made of each sample, each image being of a 1.6 mm by 2.4 mm area (4.1 mm2) with a resolution of 2048 by 2880 pixels.
In the experiments, melting initiates at muscovite-quartz grain boundaries and results in progressive replacement of muscovite by melt + mullite + biotite. Melting is completed in runs at 1193 K within 1 h, and is 30% complete within 4 h at 1123 K. For the modal composition of the rock a total dilation strain of 2.7% is necessary to accommodate the volume change for the complete reaction. In all samples, microcracks emanate from the melting sites, and show no preferred orientation. In sections orthogonal and parallel to the foliation, mean crack lengths are 116± 6 µm and 185± 8 µm, respectively - these lengths reflect the spacing between melting sites, which is greater in the foliation plane. The total crack length per unit area varies from 1000-4000 m/m2 and is not clearly correlated with any experimental variable. The crack widths are poorly constrained, but range from two to four microns.
A statistical model for crack connectivity indicates that the microcracks together with the hydraulic bonds represented by the melt pools are adequate to effect connectivity. This was verified through experiments in which a porous sand was loaded, together with the metaquartzite, to act as a drain for the melt. The microcrack permeabilities estimated from the statistical model are 10-14 m2, at least 4 orders of greater than those characteristic of metamorphic environments. Moreover, this permeability is isotropic, in contrast to that in texturally-equilibrated systems, where rock fabric leads to highly anisotropic permeability.
For reaction-induced microcracking to occur, it is necessary that reaction takes place on a time scale such that the associated volume change cannot be accommodated by plastic deformation (creep). Dimensional analysis suggests that this condition is fulfilled for a time scale of 105 - 107 years, which is comparable to the time scale of crustal melting events. Therefore reaction-induced microcracking is a feasible mechanism of permeability enhancement in geologic environments.