Abstract
Experiments were performed to determine melt concentration and strain distributions around basalt dykes in a San Carlos olivine matrix containing 10 wt% MORB (mid-ocean ridge basalt). Undrained triaxial compression tests were conducted at 1473 K and a confining pressure of 300 MPa, at constant stresses (80–160 MPa) and constant strain rates ranging from 5×10−5 to 3×10−4 s−1. Melt distribution in the dyke–matrix interface was determined by image analysis and chemical profiles. Melt migration appears to be enhanced by porosity of the microstructure and by the loading conditions. The presence of the dyke does not influence the bulk strength of the sample. Highest melt concentrations, and, presumably, the highest stress concentrations, are found at the tip of the dyke. The matrix deformation appears to be controlled by granular flow, but dilatancy occurs near the tip of the dyke, indicating coupled MORB transport and granular flow.
Dyke emplacement is a key mechanism of magma transport in the lithosphere (Fialko & Rubin 1999; Mège & Korme 2004). Magma migrates towards the surface, through a number of different mechanisms ranging from percolation through a permeable medium at great depth to hydraulic fracturing in the shallow layers. A complete dynamic model is not yet available, because the details of coupling between the stress, the country rock deformation and the role of fractures are not fully understood. However, field observations show that veins and dykes form and grow in partially molten rocks (Nicolas 1986; Kelemen et al. 1995; Rubin 1998). Theoretical calculation of basalt dyke ascent in partially molten systems (Rubin 1998) have inferred that the melt/rock viscosity ratio may drive dyke growth, according simply to a purely elastic process. In this case, flow can be treated as a diffusional process in a poroelastic medium. Local drops in pore pressure may be more important than elastic compressibility of the pore volume to enhance melt diffusion (Rubin 1998).
In order to elucidate which stress conditions among temperature and deviatoric load are the leading mechanism for driving dyke ascent in the upper mantle, we reproduced experimentally a melt-filled crack and investigated its migration in a partially molten matrix under pressure and temperature conditions representative of the upper mantle.
Laboratory experiments were carried out using a triaxial high-pressure, high-temperature Paterson apparatus, in which a sample assembly given from a cylindrical sample made of 90% San Carlos olivine and 10% basalt, including a vertical prismatic dyke inserted at the top end and composed by 100% mid-ocean ridge basalt (MORB), was heated up hydrostatically and loaded under a variety of stress conditions at high temperatures. We discussed elsewhere (Vinciguerra et al. 2004) results relative to mapping of CaO and FeO by X-ray fluorescence measurements. Here we present results relative to a full chemical analysis of all the material elements, as well as new mechanical and microstructural data, which allowed us a full investigation of the experimental conditions at which the migration of the dyke is enhanced, as well as the changes of the bulk strength and of the role of the porosity in the melt diffusion.
Material investigated, sample assembly and experimental procedure
The compositions of San Carlos olivine+MORB, and their rheology, in terms of partially molten system dynamics are well known (e.g. Hirth & Kohlstedt 1995a, b and references therein). Olivine is the dominant mineral in the upper mantle. Its rheological behaviour is crucial for the understanding of the deformation processes at a depth, where convective flow of the mantle is coupled to the movement of the lithosphere (Heidelbaach et al. 2005). Likewise, MORB glass is very representative of melt extracted from the lithosphere and is a well-characterized (e.g. Pan et al. 1998) natural basalt glass, without crystals and discontinuities.
Sample assembly and experimental procedure are reported in Figure 1. A cylindrical sample was isolated through alumina spacers from the loading pistons. The whole assembly was Fe jacketed (Fig. 1a). Powdered 100 wt% MORB (Fig. 1b), with a particle size of <15 mm, was cold-pressed into cylindrical Ni cans (26 mm long, 11.6 mm o.d., 10 mm i.d.) with a uniaxial pressure of 100 MPa and air dried for 12 h at 425 K, in a fO2 controlled-atmosphere furnace. Cold-pressed samples were subsequently encapsulated with Ni foils, and hot pressed at 1473 K±2 K and 300 MPa (about 20 km of depth) for 6 h in a Paterson apparatus (Fig. 1c).
Jackets were dissolved in a bath of nitric and hydrochloric acid. Prismatic bodies (length, 7 mm; thickness, 2.0 mm; width, 3.0 mm) were cut from the hot-pressed samples and were vertically inserted into the cold-pressed cylindrical samples of olivine (<35 µm)+10% MORB powder (Fig. 1d). The 10% MORB allows a partially molten aggregate to be obtained, where the melt is located at three and four junctions, in agreement with the upper-mantle conditions (Hirth & Kohlstedt 1995a, b). Samples were again hot-pressed at 1473 K±2 K and 300 MPa for 6 h (Fig. 1e). In preparation for deformation experiments, the surfaces of specimen were ground parallel with respect each other and normal to the long axis of the sample, and jacketed in a new Ni can (10 mm o.d., 9.4 mm i.d.).
(a) Experimental set-up. MORB powders cold-pressed into a cylindrical Ni can (b) were isostatically hot-pressed (c). Prismatic MORB bodies were cut after the HIP and inserted (d) in cold-pressed powders of olivine+10% MORB. This sample assembly was again isostatically hot-pressed and loaded (e).
Triaxial compression tests were then conducted at 1473 K±2 K and 300 MPa either at constant strain rate from 3×10−4 to 5×10−5 s−1, obtaining a constant strain of approximately 12–13%, or at constant differential stress, between 75 and 160 MPa, for 45 min in a servo-controlled internally heated gas-medium apparatus (Paterson 1990). Samples were encapsulated and experiments were not drained. Load v. time and displacement v. time data were converted to differential stress and axial strain rate by assuming that deformation was homogeneous and that the sample volume remained constant.
Results and discussions
Scanning electron microscope (SEM) images relative to hot isostatic pressing (HIP) and deformed samples, such as those reported in Figure 2a and b, allow three main features to be identified: the glass microstructure of inner dyke (Fig. 2c); the dyke–matrix interface (Fig. 2d); and the matrix area (Fig. 2e). The distribution of the melt phase in the olivine+10% MORB matrix (Fig. 2e) is in agreement with that documented for samples prepared by using the same procedure, where the melt phase is found at three- and four-grain junctions (Hirth & Kohlstedt 1995a, b and references therein). However, in the dyke–matrix interface melt concentration increases and it is possible to observe (Fig. 2d) how the diffusing melt penetrates inside the grain junctions, forming melt pockets and, in some cases, surrounding olivine crystals. Highest melt concentrations are found around the ‘dyke’ tip.
Images of (a) the HIP sample and (b) the deformed sample. Note that a high melt concentration forms around the dyke tip. (c)–(e) SEM images of dyke inside, dyke–matrix interface and matrix respectively.
In order to quantify whether/how melt migration occurred from the 100% MORB dyke into the 90% San Carlos olivine+10% MORB, we performed chemical profiles across the dyke, calculating the percentage of the main oxides, i.e. SiO2, Al2O3, FeO, MgO, CaO, as a function of the distance from the dyke (Fig. 3) . The chemical composition of the two starting materials used are reported in Table 1.
Profile of chemical composition in percentage across the dyke conducted by energy dispersive spectroscopy (EDS) measurement under de-focus mode with a spot size of 40 mm.
Chemical analyses of starting material
The chemical analysis reveals that melt diffused significantly in the matrix, forming a broad interface area (in grey in Fig. 3) beyond the dyke–matrix contact. The influence of the stress applied is suspected to have increased stress concentrations at the dyke tip, enhancing melt diffusion in the matrix. However, this is not straightforward for the experiments performed at constant strain rate, due in some cases also to the technical difficulty of polishing the dyke–matrix interface. Importantly, creep data (Fig. 4) show that the presence of the dyke does not significantly change the bulk strength of samples with respect to samples with the same matrix composition (i.e. San Carlos olivine 90%+MORB 10%) but dyke free (Hirth & Kohlstedt 1995a, b). This is in agreement with the further observation that samples are very dense and deformation occurred without dilation.
Comparison of creep data for samples containing a dyke with partially molten samples with 10% melt but without a dyke (Hirth & Kohlstedt 1995a, b).
However, by considering the set of deformation experiments performed at constant stress and hence for the same time, MORB ranges from 12–18% for 70 MPa constant stress applied to 14–22% for 130 MPa in the interface area (see also Vinciguerra et al. 2004), indicating therefore that stress concentration may have induced changes in melt diffusion rates.
Finally, it is important to observe that during deformation some samples showed low strength, resulting in strain weakening (Fig. 5a and b). As no apparent instrumental failures were revealed (no leaks occurred, no change in confining pressure), we support that some gas was trapped inside the sample or between the sample and the jacket, generating voids and bubbles up to the mm-scale, which are observed in the microstructure. Melt diffusion was then strongly enhanced by large cavities (Fig. 5c–f), which enhanced local pressure drops induced by dilatancy. As a consequence, melt migration zones are much broader than those in the denser samples obtained, which show high strength and strain-hardening deformation mechanisms.
(a) and (b) Plots of stress v. strain. Porous samples showed low strength and strain weakening, in comparison with dense samples which showed high strength, strain hardening and no dilation. SEM images of dense (c) and porous (d) samples. Chemical profiles for dense (e) and porous (f) samples.
Conclusions
Laboratory experiments were carried out to determine melt concentration and strain distributions around basalt dykes in a matrix of 90% San Carlos olivine containing 10 wt% MORB. Undrained triaxial compression tests were conducted at 1473 K and a confining pressure of 300 MPa, at constant stresses (80–160 MPa) and constant strain rates ranging from 5×10−5 to 3×10−4 s−1, using high-pressure, high-temperature Paterson apparatus.
Optical images and chemical analyses of the microstructures, as well as mechanical data, show that melt migration appears to be enhanced by increased porosity and higher loading conditions. Stress concentrations around the dyke tip appear to correlate with changes in melt concentration. Moreover, the presence of the dyke does not influence the bulk strength of the sample.
At the simulated conditions of our experiments the kinematics of migration are strongly influenced by plastic mechanisms and occur at some critical stress, favoured by local pressure drop during dilatancy.
Acknowledgments
G. Di Toro is warmly thanked for his optical analyses. G. Hirth and J. Renner are thanked for fruitful discussions.
- © The Geological Society of London 2008
References
Constraining melt concentration and strain distribution around basalt dykes in partially molten olivines
