The starting material in the lower part of sample as a liquid source was a mixture of 5 wt.% Ni-Cu sulfides, 10 wt.% basalt and 85 wt.% olivine. For a sulfide-free experiment, the mixture of 90 wt.% olivine and 10 wt.% basalt was used as the starting material of liquid source. And dunite consisting only of olivine crystals was placed in the upper part of sample as a depleted mantle peridotite. Natural olivine crystals were separated from fresh spinel lherzolite xenoliths collected from Damaping (Hannuoba region), North China and were ground to 10-20 μm grain size in an agate mortar. Tholeiitic basalt from the East Pacific Rise (102.704 4ºW, 2.649 61ºS) was ground to < 10 μm grain size. Ni-Cu sulfides were composed of pyrrhotite (59 wt.%), pentlandite (36 wt.%), and chalcopyrite (5 wt.%) separates, similar to the composition of base-metal sulfide aggregates in massif peridotites (Lorand et al., 2010), and were ground to < 10 μm. Compositions of these starting materials are listed in Table 1.
wt.% Olivine Basalt wt.% Ni-Cu sulfides Chalcopyrite (5 wt.%) Pyrrhotite (59 wt.%) Pentlandite (36 wt.%) SiO2 41.42(31) 51.73(24) S 35.29(14) 38.63(56) 34.06(41) TiO2 0.02(2) 2.30(6) Fe 30.23(27) 59.58(36) 33.04(88) Al2O3 0.01(1) 12.85(5) Ni 0.10(13) 0.13(3) 32.18(65) Cr2O3 0.01(1) 0b Cu 33.46(15) 0.01(1) 0b MgO 48.89(15) 4.62(3) The component of basalt was analyzed by XRF, and the components of olivine and three sulfide minerals were analyzed by EPMA using an accelerating voltage of 15 kV, a beam size of 1.0-2.0 μm, and a beam current of 20 nA, respectively. 1σ error is given in parentheses. a. Total FeO; b. lower the detection limit; c. not analyzed. CaO 0.04(2) 8.33(1) MnO 0.15(2) 0.23(1) FeOa 8.84(28) 14.60(36) Na2O 0.01(1) 3.59(4) K2O 0.01(1) 0.33(1) H2O+ -c 1.19(26)
Table 1. The chemical composition of starting materials
These two-layer reaction experiments were conducted at the State Key Laboratory of Geological Processes and Mineral Resources (GPMR) of China University of Geosciences using a 150 Ton non-end-loaded type piston-cylinder press. Starting materials were loaded into a 3.75-mm-diameter and 5-6-mm-high cylindrical platinum (Pt) capsule with a graphite inner sleeve (Fig. 1). A low friction assembly consisting of NaCl and Pyrex sleeves, a graphite heater, sintered MgO spacers, and an Al2O3 plug, was used for the piston-cylinder experiments. Pressure calibration was conducted against the quartz/coesite phase transition and is accurate to ±0.1 GPa. The temperature was monitored by a W/Re type C thermocouple located at the bottom of the capsule. These experiments were conducted under a constant pressure of 1.5 GPa and temperatures of 800 to 1 300 ℃ for 12-72 h prior to being quenched to room temperature by shutting off power to the furnace. All capsules were heated before being sealed at 120 ℃ in a vacuum oven for more than 12 h to remove absorbed water vapor in the specimens.
Polished sections cut parallel to the specimen axis were prepared from the recovered experimental specimens. The microstructure of experimental run products was observed using a Quanta 450 field-emission gun scanning electron microscope (FEG-SEM) at the GPMR with an accelerating voltage of 20 kV, a spot size of 6.0 and a working distance of ~12 mm. Quantitative compositional analyses were performed using an electron probe microanalyzer (EPMA, JXA-8100) at the Key Laboratory of Submarine Geosciences, State Oceanic Administration, China and the X-ray fluorescence spectrometry (XRF) at the ALS Minerals-ALS Chemex (Guangzhou) Co Ltd, China.
The crystallographic preferred orientations (CPOs) and shape preferred orientations (SPOs) of silicate minerals (olivine, clinopyroxene, and orthopyroxene) were analyzed by indexing of electron backscattered diffraction (EBSD) patterns using a Quanta 450 FE-SEM equipped with an HKL Nordlys EBSD detector at the GPMR. An accelerating voltage of 20 kV, a spot size of 6, a beam current of 6 nA and a working distance of 20-25 mm were used for automatic mapping measurements. And the grain size of silicate minerals can be also measured by the EBSD mapping analyses.
1.1. Starting Materials
1.2. Experimental Methods
1.3. Analytical Techniques
Several two-layer reaction experiments (Table 2) were conducted at a pressure (1.5 GPa), different temperatures (800 and 1 250 ℃) and durations (12-96 h) to investigate the mechanical transport of sulfide liquid during reaction infiltration of silicate melt. All back-scattered electron (BSE) images were analyzed using the freeware ImageJ developed by the National Institute of Health (NIH) (http://imagej.nih.gov/ij/, the analysis methodology is described in Wang et al., 2020) to measure area fractions of sulfide and silicate melt and the size of sulfide droplets.
No. Starting materials Experimental conditions Liquid distribution in the upper part of samples Upper part/lower part T (℃), t (h) PC350 Olivine/ olivine+basalt+sulfide 800, 10 Not molten PC343 1 250, 12 No melt channels PC340 1 250, 24 No melt channels PC347 1 250, 48 Several melt channels; more sulfide droplets PC362 1 250, 72 Apparent melt channels; more sulfide droplets PC389 1 250, 96 Some relics of melt channels PC401 Olivine/olivine+basalt 1 250, 48 The homogeneous distribution of silicate melt
Table 2. Summary of experimental conditions and liquid distributions in the upper part of samples
In a low temperature (800 ℃) experiment, no silicate melt was produced. Sulfide mineral grains were generally polygonal and homogeneously distributed between olivine and unmolten basalt glass in the lower part of the sample. No sulfide was observed in the upper part of the sample.
Under a higher temperature (1 250 ℃) condition, based on the calculation of mass balance, the basalt in the lower part of samples was molten to produce ~6.70 wt.%-8.96 wt.% silicate melt, which was consistent with these results (about 6%-8%) by analyzing BSE images. Silicate melt was mainly observed in melt junctions/pockets and melt channels between olivine-grain edges, forming locally interconnected melt networks (Figs. 2a-2c, 3a-3d). Sulfide liquid formed isolated and roundish droplets that were surrounded by silicate melt in silicate-melt-filled pockets and melt channels (Figs. 2a-2c, 3a-3d). A few large sulfide droplets were directly contacted with silicate minerals and squeezed into a wormlike shape, especially in these experiments with a long duration (Fig. 3d). With increasing duration, both the area fraction and average size of sulfide liquid in the lower part of samples were approximately constant within the error of measurement based on analyses of BSE images (~2.64%±0.20% and ~3.24±2.56 μm at the duration of 12 h; ~2.69%±0.11% and ~5.79±5.06 μm at the duration of 96 h). The average size of olivine grains was constant in the lower part of samples (~25.1±12.6 μm at 48 h, ~27.0±13.8 μm at 72 h) based on EBSD mapping analyses.
Figure 2. Microstructure of the distribution of sulfide and silicate melts in the lower part (a)-(c) and the upper part (d)-(i) of two-layer experiments under different duration (12-48 h) conditions (backscattered scanning electron microscopy). Red lines and red dotted lines are the interface between the upper and lower sample and melt channels in the upper part of sample, respectively. Mineral abbreviations: Ol. olivine; Opx. orthopyroxene.
Figure 3. Microstructure of the distribution of sulfide and silicate melts in the lower part (a)-(d) and the upper part (e)-(i) of two-layer experiments at the duration of 72 h (backscattered scanning electron microscopy). Symbols and mineral abbreviations are the same as Fig. 2.
In the upper part of samples, with increasing duration, more dendritic melt channels were formed, and more and larger sulfide droplets can be observed in melt junctions/pockets and channels (Figs. 2d-2i, 3e-3i, and 4). The area fraction and average size of sulfide droplets respectively increased from ~0.034% and 2.1±0.8 μm at the duration of 12 h to ~0.232% and 5.2±2.1 μm at the duration of 72 h (Fig. 4). The average size of olivine grains was approximately constant in upper samples with increasing duration from 48 (~30.7±18.4 μm) to 72 h (~35.3±22.5 μm) based on EBSD mapping analyses, which were both larger than those in the lower samples. Especially, when the duration increased to 72 h (experiment PC362), in the upper part of the sample, much more dendritic melt channels and larger sulfide droplets can be observed (Figs. 3e-3i and 4) and the area fraction of silicate melt is about 3.2%-6.1% based on BSE image analyses. About 21 melt channels can be distinguished in the upper part of experiment PC362 (Fig. 3i). The direction of these melt channels was approximately perpendicular to the interface between the upper and lower sample in 2-D sections, and their average length was ~221±101 μm (the longest one ~420 μm) with a space interval of about ~582±67 μm. However, at the duration of 96 h, there were only some relics of closed melt channels observed in the upper sample (red dotted lines in Fig. 5), and the area fraction and average size of sulfide droplets almost remain constant with those in the experiment with the duration of 72 h (Fig. 4). These both imply that the content of silicate melt in the liquid source was not enough to maintain the open of melt channels at a longer duration (96 h). In addition, the biggest size of sulfide droplets in BSE images of upper samples increased from about 3.7 μm at the duration of 12 h to about 10.7 μm at the duration of 96 h (Fig. 4).
Figure 4. The area fraction, biggest and average size of sulfide droplets in the upper part of samples with increasing duration. The statistical area for each sample is the same (~185 μm×1 000 μm in the upper part of samples).
Figure 5. Overview (backscattered scanning electron microscopy) of the distribution of two immiscible liquids in experiment PC389 at the duration of 96 h. The red line is the interface between the upper and lower sample, and red dotted lines represent the relics of melt channels.
To explore the effect of sulfide droplets as secondary phase particles on the migration of silicate melt, a sulfide-free experiment (PC401) with a duration of 48 h has been conducted under the same temperature (1 250 ℃) and pressure (1.5 GPa) condition. The area fraction of silicate melt in the upper part of sample was approximately the same as the sulfide-bearing experiment with the same duration (PC347, 48 h), but the distribution of silicate melt was more homogeneous in the sulfide-free experiment, and no apparent melt channel was observed (Fig. 6a). The size of olivine grains was almost the same in the upper part (~24.3±14.7 μm) and the lower (~26.4±17.3 μm) part of the sample, which were both similar with that (~25.1±12.6 μm) of the lower sample of experiment PC347 within error.
Figure 6. Microstructure (backscattered scanning electron microscopy) of the distribution of silicate melt in the upper part of sulfide-free experiment PC401. Silicate melt was observed in triple junctions or melt pockets.
Notably, no planar reaction layer was present between the upper and lower part of samples in all experiments.
The Mg# (=100×Mg/(Mg+Fe) in molar unit) of olivine in the starting material was ~90.9±0.2. When duration was shorter than 48 h, the olivine Mg# in the upper part of samples was approximately constant within errors (~90.6±0.2 at 12 h, ~90.5±0.2 at 24 h and ~90.6±0.4 at 48 h) and both higher than those in the lower part of samples (~89.7±0.4 at 12 h, ~89.7±0.2 at 24 h and ~89.8±0.3 at 48 h) (Fig. 7a). In contrast, under the long duration condition (72 h), the Mg# of olivine (~90.1±0.2) in the upper sample was almost the same as that (90.0±0.3) in the lower sample. With increasing duration to 96 h, the olivine Mg# in the upper part and lower part of the sample increased to 90.9±0.2 and 91.0±0.3, respectively, which were both the same as that of the starting material (Fig. 7a). With increasing duration, Ni contents of olivine in the upper part and lower part of samples were both almost constant, and the olivine Ni contents in the lower part were lower than those in the starting material and the upper part of samples (Fig. 7b). Silicate melt in the liquid source was disequilibrium with olivine grains, thus the melt-rock reaction can occur: melt1+Mg2SiO4 (Ol)←→Mg2Si2O6 (Opx)+melt2 (reaction 1), leading to the presence of small orthopyroxene (Opx) grains in the lower and upper part of samples (Figs. 2a-2c, 3a-3d). However, with increasing duration, silicate melt was gradually evolving toward the composition of melt2 saturating with Opx in the liquid source, then more melt2 can be migrated into the upper part of samples causing the occurrence of the left-hand of the reaction 1. Thus, at the duration of > 72 h, the content of Opx in the upper part of samples was decreasing, and only several Opx grains were observed in the 96 h experiment. The Mg# of Opx in the upper part of samples were higher than those in the lower part of samples (Fig. 7c). With increasing the duration from 12 to 96 h, the Mg# of Opx in the lower part of samples increased from 89.9±0.3 to 91.4±0.2, whereas those in the upper part of samples were approximately constant within errors (Fig. 7c). Additionally, no significant change in the Ni/S ratio of sulfide liquid was observed in the upper and lower part of samples within errors, and these values in the upper part of samples were commonly higher than those in the lower part of samples (Fig. 7d).