Al-Zoning of Serpentine Aggregates in Mesh Texture Induced by Metasomatic Replacement Reactions

Al-Zoning of Serpentine Aggregates in Mesh Texture Induced by Metasomatic Replacement Reactions Abstract Serpentinization of oceanic lithosphere commonly proceeds with the development of mesh texture. Examination of a serpentinized harzburgite and a plagioclase-bearing wehrlite revealed conspicuous zoning of Al in a serpentine mesh texture, with Al-rich cores and Al-poor rims, as well as Al-rich veins, indicating local transport of Al from plagioclase and pyroxene during serpentinization. To reveal the influences of Al on the reaction mechanisms and textural development during serpentinization, we conducted hydrothermal experiments in the olivine (Ol)–plagioclase (Pl)–H2O system and analyzed the variations in mineralogy and microtexture of olivine replacements as a function of distance from the Ol–Pl boundary. The Al–Si metasomatic zone, where Al-serpentine and a minor amount of Ca-saponite were formed, was developed in the Ol-hosted region close to the Ol–Pl boundary. Far from the Ol–Pl boundary, Al-free serpentine, brucite, and magnetite were formed, indicating the progress of an ‘isochemical’ reaction (separate from water). The aggregates of Al-serpentine after olivine in the metasomatic zone showed a characteristic zoning of Al. Microtextural evidence indicates that the zoning was produced in response to the migration of an Al metasomatic front that involved an early-stage of serpentinization with an Al-free solution, and the subsequent pseudomorphic replacement of olivine and simultaneous development of overgrowths as the amount of Al increased in the solution. The Al-bearing aqueous solution caused the formation of olivine pseudomorphs and this contrasts with the lack of preservation of original olivine outlines in the isochemical zones. Comparisons of zoning in natural and experimentally produced mesh textures suggest that Al-poor rims in the mesh texture form at the start of the serpentinization process, followed by the coupled formation of Al-rich mesh cores and Al-rich veins. Our experimental results indicate that Al-zoning in the mesh texture represent the transition from a closed to an open system during serpentinization. INTRODUCTION Serpentinization causes significant changes in the nature of the oceanic lithosphere, including its rheological properties (Escartín et al., 2003), thermal flux (Iyer et al., 2010), hydrogen production (Sleep et al., 2004), and microbial activity on the seafloor (Kelley et al., 2001). Serpentinized peridotite in the oceanic lithosphere can store up to 13 wt % chemically bonded water in a solid state and it may exert the primary control on fluid flux (Hyndman & Peacock, 2003; Nakatani & Nakamura, 2016; Plümper et al., 2017) and seismic activity in subduction zones (Wada et al., 2008; Boudier et al., 2010). Despite the influence of serpentinized peridotite on various processes, the in situ rate of serpentinization within oceanic lithosphere is still poorly constrained, since it is a complicated process involving volume change, deformation, and mass transfer, as well as the reaction itself. A mesh texture is the typical texture found in serpentinized peridotite from the oceanic lithosphere. The texture comprises polygonal domains divided by grain-scale fractures induced by tectonic, thermal, and/or reaction-induced stresses (O’Hanley, 1996; Plümper et al., 2012b; Rouméjon & Cannat, 2014; Okamoto & Shimizu, 2015; Shimizu & Okamoto, 2016). Mesh texture commonly shows conspicuous zoning that can be divided into core and rim. The core and rim of the mesh have different mineral assemblages and/or mineral compositions, and based on such textures, a two-stage model for serpentinization has been proposed (e.g. Viti & Mellini, 1998; Bach et al., 2006; Beard et al., 2009). In this model, the mesh rim forms first under near-isochemical conditions, and then subsequently the mesh core forms as external fluids infiltrate the system. The mesh zoning, therefore, represents temporal and spatial changes in the chemical environment during serpentinization. Some serpentines in the mafic and ultramafic rocks of the oceanic lithosphere contain large amounts of Al (up to 27 wt %) (O’Hanley, 1996; Viti & Mellini, 1998; Mellini et al., 2005; Beard et al., 2009; Andreani et al., 2013; Schwarzenbach et al., 2016; Nozaka et al., 2017). Based on hydrothermal experiments, Andreani et al. (2013) showed that the presence of aluminum accelerates the hydration of olivine, although the experimental product was chlorite rather than serpentine. Al is not a major component of olivine (<0·1 wt %) so potential sources of the Al are, therefore, pyroxenes, spinel and plagioclase in mafic and ultramafic rocks. Accordingly, the distribution of Al in partly serpentinized peridotite could be a useful tracer for the influx of external fluids and the transfer of elements, either on the scale of millimetres in the case of chlorite forming after olivine along the contacts between olivine and plagioclase (Frost et al., 2008), or on the scale of metres in the case that the assemblage tremolite + chlorite + serpentine is formed via Al–Si–Ca metasomatism of peridotite (Boschi et al., 2006). However, the distribution of Al in the mesh texture and the role of Al during serpentinization have not been studied systematically. In this study, we describe first the characteristics of Al zoning in the mesh texture within partially serpentinized mafic and ultramafic rocks of the oceanic lithosphere. Then, to understand the effects of Al transport on the mechanisms and textural development of serpentinization, we describe the results of hydrothermal experiments in the olivine (Ol)–plagioclase (Pl)–H2O system. Combining experimental data on the distribution of Al in the product mineral and observed Al distribution in two natural rocks, we discuss the implications for textual development and mass transfer during serpentinization within the oceanic lithosphere. METHODS Hydrothermal experiments We conducted hydrothermal experiments in the Ol–Pl–H2O system using a tube-in-tube vessel at 230°C and at a vapor-saturated pressure of 2·80 MPa (Fig. 1a, b). The experimental procedures were almost the same as those used in the Ol–quartz–H2O experiments of Oyanagi et al. (2015). Natural olivine (XMg (Mg/(Mg + Fetotal)) = 0·91) and plagioclase grains (Ca/(Ca + Na + K) = 0·97) were crushed and sieved to 25–53 μm size. The powders were then washed repeatedly with deionized water in an ultrasonic bath. In the main vessel, which was made of Inconel alloy (44 cm3), four inner tubes were set with 20 ml distilled water (Fig. 1a). In each inner tube, 40 mg of plagioclase powder was placed at the bottom, and 110 mg of powdered olivine was then placed on the plagioclase layer (Fig. 1b). The base of each inner tube was sealed and the tops left open. The thicknesses of the plagioclase and olivine layers were ∼10 and ∼30 mm, respectively. The water–rock mass ratio was ∼33. The vessels were then placed in ovens at 230°C. The temperature difference between the inside and the outside of the vessel was <1°C. The run times were 783, 2055, 3184, and 7980 h. Fig. 1. View largeDownload slide (a) Schematic illustration of the tubes placed in the reaction vessel. Four inner tubes were immersed in liquid. (b) Configuration of mineral powders within the inner tubes for the Ol–Pl–H2O experiments. Fig. 1. View largeDownload slide (a) Schematic illustration of the tubes placed in the reaction vessel. Four inner tubes were immersed in liquid. (b) Configuration of mineral powders within the inner tubes for the Ol–Pl–H2O experiments. For comparative purposes, Pl–H2O experiments were conducted with 1·0 g of the same plagioclase (<25 μm) and 10 ml of distilled water under the same P–T conditions as used for the Ol–Pl–H2O experiments. The run times for the Pl–H2O experiments were 158, 330, and 488 h. To investigate the effects of the Inconel alloy vessel on the solution chemistry, we carried out blank experiments (with distilled water only in the vessel) under the same experimental P–T conditions for 2224 h. Analyses of the experimental products After each individual run, the reaction vessel was cooled to room temperature within 1 h, and the solutions and four inner tubes, which contain solid materials (products + unreacted reactants), were taken from the vessel. Solution pH was immediately measured at room temperature of ∼25°C. After diluting the solutions with nitric acid, the concentrations of Si, Mg, Ca, Fe, and Al were measured by inductively coupled plasma–atomic emission spectrometry (ICP–AES; Thermo Scientific iCAP 6000) at Tohoku University, Japan. We calculated the fluid speciation of the analyzed fluids, the in situ pH, and the phase diagram in the CaO–Al2O3–MgO–SiO2–H2O system for the experimental P–T conditions using Geochemist’s Workbench (Bethke, 2007). Thermodynamic data in the default database thermo. tdat were customized on the basis of a log K value obtained from the SUPCRT92 database (Johnson et al., 1992) with revisions suggested by Zimmer et al. (2016). The primary thermodynamic data were taken mainly from Holland & Powell (2011) for minerals, from Tagirov & Schott (2001) for Al-bearing species in solution, and from Stefánsson (2001) for aqueous silica. After each individual run, the solid materials (products + unreacted reactants) from the four inner tubes were dried for >24 hours at 90°C. One of the inner tubes was encased in epoxy resin, and a polished thin section was made in the direction parallel to the long axis of the tube. The chemical compositions of the minerals in the polished thin section were analyzed using an electron probe micro-analyzer (EPMA; JEOL 8200) at Tohoku University with an accelerating voltage of 15 kV, beam current of 12 nA, and a focused 1–2 µm beam diameter. Standards used for calibrations were wollastonite (Si, Ca), rutile (Ti), corundum (Al), hematite (Fe), MnO, periclase (Mg), albite (Na), sanidine (K), and Cr2O3. JEOL software was used for the ZAF corrections. The minerals present in the polished thin section were observed with a field-emission scanning electron microscope (FESEM; JEOL JSM-7001F) at Tohoku University. The modal abundances of minerals and the porosities were obtained for selected thin sections by analyses of backscattered electron (BSE) images (102 × 385 µm in size). The resolution of the images used in the analyses were ∼1 μm2/pixel. Serpentine minerals were identified using a micro-Raman spectroscope (HORIBA XploRA PLUS) at Tohoku University equipped with a 532 nm laser and 2400 grooves/mm grating. Another inner tube was used for thermogravimetry (TG; Rigaku Thermo Plus EVOII TG8120) at Tohoku University to analyze the H2O contents of the solid samples. For the TG analyses, an inner tube was cut into eight segments of ∼5 mm length. Segments one and two were from the Pl-hosted region, and segments three to eight from the Ol-hosted region. During the TG analyses, temperatures were raised from room temperature to 1100°C at a rate of 10°C per minute. Loss on ignition (LOI) values for each segment were obtained in the temperature range 120–1100°C to exclude the weight loss associated with molecular water in the solid samples. The H2O contents of serpentine and brucite were determined separately from the weight losses in the temperature ranges 450–800°C and 320–450°C, respectively. The pseudosection in the FeO–MgO–SiO2–H2O–O2 system was calculated using Perple X 6·7·9 (Connolly, 2009) with the hp11ver.dat dataset (Holland & Powell, 2011), which includes thermodynamic data for lizardite, greenalite, brucite, hematite, magnetite, forsterite, fayalite, and talc. The thermodynamic data for ferric serpentine (Mg2 Fe23+SiO5(OH)4) and Fe-brucite were constructed following Evans et al. (2013). Brucite, and talc were modeled as ideal mixtures of Mg- and Fe-endmembers, while olivine activity–composition relationships were described by a symmetric formalism model (Powell & Holland, 1993) with a Wfo–fa value of 8 kJ/mol (Evans et al., 2013). Serpentines were modeled as ideal mixtures of the lizardite, greenalite and ferric serpentine. The activity of iron was assumed to be 0·25 as an analogue for an alloy phase such as awaruite (Evans et al., 2013). Analyses of natural serpentine mesh texture We analysed two rocks with a well-developed mesh texture: a serpentinized plagioclase-bearing wehrlite from the Izu–Ogasawara fore-arc and a partly serpentinized harzburgite from the Mineoka ophiolite complex in the Boso Peninsula, central Japan. We prepared polished thin sections and determined the chemical compositions of the minerals and the nature of the mineral zoning using EPMA (details of the analytical procedures are given above). Representative mineral compositions of the two rock specimens are summarized in Supplementary Data Table S1; supplementary data are available for downloading at http://www.petrology.oxfordjournals.org. ALUMINUM ZONING IN SERPENTINE MESH TEXTURE Plagioclase-bearing wehrlite from the Izu–Ogasawara forearc region The serpentinized plagioclase-bearing wehrlite sample (KH07–02-D31–101) was collected from the landward slope of the Izu–Ogasawara trench during a dredging cruise (KH07–02 leg 4) of the R/V Hakuho–Maru operated by the Japan Agency for Marine-Earth Science and Technology (Harigane et al., 2013). In previous studies (Morishita et al., 2011; Harigane et al., 2013) it was reported that peridotite, troctolite, pyroxenite, gabbro, dolerite and basalt were collected from 5293 to 5792 metres below sea level during dredging and remotely operated vehicle operations. Some of the doleritic samples exhibited a fore-arc tholeiitic basalt signature (cf. Ishizuka et al., 2011). The variations in the rock samples suggest that this trench wall represents the structure of an immature fore-arc mantle at the time of initiation of the subduction zone (Morishita et al., 2011; Harigane et al., 2013). The plagioclase-bearing wehrlite sample was covered by layers of brown weathered material about 10 mm thick (Supplementary Data Fig. S1). We made a polished thin section of the unweathered core of the sample. The serpentinized plagioclase-bearing wehrlite originally consisted of ∼84 vol % olivine, ∼11 vol % clinopyroxene, and ∼5 vol % plagioclase (Fig. 2a). More than 70 vol % of the olivine is serpentinized (Fig. 2b), with lizardite and magnetite being the main alteration products. Brucite and magnetite were not found at olivine–plagioclase contacts, but minor intergrowths of brucite (<5%) within serpentine and subhedral magnetite grains were found further away (>1·0 mm) from the contacts. The clinopyroxene is almost unaltered, whereas the plagioclase is intensely altered, mainly to chlorite and hydrogrossular, and partly to pumpellyite (Fig. 2a). Fig. 2. View largeDownload slide Mesh texture and distribution of Al in serpentinized plagioclase-bearing wehrlite from the Izu–Ogasawara trench, collected during the KH07–2 cruise. (a) Photomicrograph of the serpentinized plagioclase-bearing wehrlite. Olivine is highly serpentinized to form a mesh texture, and the plagioclase is almost completely altered to hydrogrossular and chlorite, along with some pumpellyite, while the clinopyroxene remains almost fresh. (b) Modal abundances of olivine in the area indicated by the rectangle in Fig. 2a. (c)–(e) Elemental maps of (c) Al, (d) Si, and (e) Ca, with a profile of relative integrated intensity (red line) obtained for each element in serpentine along the long axis of the area indicated by the rectangle in Fig. 2a. Olivine grains were removed from analysis by blackening BSE images. (f) Back-scattered image of serpentine mesh texture. (g) Elemental map of Al in the mesh texture shown in Fig. 2f. C, mesh core; R, mesh rim; V, mesh vein network. (h) Al content in serpentine (cations per seven oxygen formula unit) along line a–b in Fig. 2g. (i) Plot of Mg + Fe cations vs Si (per seven oxygen formula unit) shows the Tschermak substitution trend in the Al-rich mesh core and vein serpentine. The Al-poor mesh rim is characterized by R3+ cations in octahedral sites. Pl, plagioclase; Grs, hydrogrossular; Pmp, pumpellyite; Chl, chlorite; Cpx, clinopyroxene; Ol, olivine. Fig. 2. View largeDownload slide Mesh texture and distribution of Al in serpentinized plagioclase-bearing wehrlite from the Izu–Ogasawara trench, collected during the KH07–2 cruise. (a) Photomicrograph of the serpentinized plagioclase-bearing wehrlite. Olivine is highly serpentinized to form a mesh texture, and the plagioclase is almost completely altered to hydrogrossular and chlorite, along with some pumpellyite, while the clinopyroxene remains almost fresh. (b) Modal abundances of olivine in the area indicated by the rectangle in Fig. 2a. (c)–(e) Elemental maps of (c) Al, (d) Si, and (e) Ca, with a profile of relative integrated intensity (red line) obtained for each element in serpentine along the long axis of the area indicated by the rectangle in Fig. 2a. Olivine grains were removed from analysis by blackening BSE images. (f) Back-scattered image of serpentine mesh texture. (g) Elemental map of Al in the mesh texture shown in Fig. 2f. C, mesh core; R, mesh rim; V, mesh vein network. (h) Al content in serpentine (cations per seven oxygen formula unit) along line a–b in Fig. 2g. (i) Plot of Mg + Fe cations vs Si (per seven oxygen formula unit) shows the Tschermak substitution trend in the Al-rich mesh core and vein serpentine. The Al-poor mesh rim is characterized by R3+ cations in octahedral sites. Pl, plagioclase; Grs, hydrogrossular; Pmp, pumpellyite; Chl, chlorite; Cpx, clinopyroxene; Ol, olivine. The bulk Al content in the altered olivine region is high around the plagioclase grains; the Al contents are highest at contacts with plagioclase and decrease linearly with increasing distance from the plagioclase. The zone of Al metasomatism is about 0·9 mm wide (Fig. 2c) and in this zone the olivine is almost completely consumed (Fig. 2b). Up to 0·1 mm from the contacts of Ol and Pl, small amounts of tremolite and diopside were formed, as indicated by local increases in Si (Fig. 2d) and Ca (Fig. 2e). In most parts of the Al metasomatic zones, the Si content of the bulk-rock is lower than in regions outside the Al metasomatic zones (>1·0 mm; Fig. 2d). Most of the serpentinized olivine grains show a clear mesh texture, composed of polygonal units separated by veins (V in Fig. 2g). Each polygonal unit shows zoning, with a mesh core (C in Fig. 2g) and a rim (R in Fig. 2g) evident in BSE images (Fig. 2f). In the Al metasomatic zone, this mesh zoning corresponds to changes in the Al content of the serpentine aggregates (Fig. 2g), with Al-rich cores (Al = 0·6–0·8 per formula unit (pfu)), Al-poor rims (∼0·05 pfu), and Al-rich veins (0·5–0·8 pfu) (Fig. 2h). The composition of the mesh cores and veins are similar, and the variations can be explained by Tschermaks substitution (AlAl(Mg, Fe2+)-1Si-1) (Fig. 2i;Beard & Frost, 2016). The Al-poor composition in the mesh rims suggests minor amounts of R23+Si2O5(OH)4 component (R3+ is Al3+ and/or Fe3+; Fig. 2i). Harzburgite from the Mineoka ophiolite complex The Mineoka ophiolite complex, central Japan, is as an accreted fragment of ophiolite that formed at a mid-ocean ridge (Sato et al., 1999). The serpentinized harzburgite originally contained olivine (∼81 vol %), orthopyroxene (∼14 vol %), and clinopyroxene (∼4 vol %), along with minor spinel (∼1 vol %). The olivine grains are heavily serpentinized and serpentine (bastite) has replaced orthopyroxene. Raman spectra show the serpentine minerals to be lizardite and/or chrysotile (Katayama et al., 2010), and serpentine + magnetite + brucite were formed within the olivine regions (Hirauchi et al., 2010; Shimizu & Okamoto, 2016). Zoning of Al in the olivine mesh texture is displayed around grains of clinopyroxene (Fig. 3a). Clinopyroxenes have been replaced by serpentine without the formation of other Ca-bearing minerals, indicating removal of Ca from the system. The elemental map of Al indicates that the average Al content in the olivine region decreases with increasing distance from boundaries with clinopyroxene (Fig. 3a). Magnetite was commonly observed along olivine and mesh core boundaries, and in serpentine veins (Fig. 3b). Fig. 3. View largeDownload slide Serpentine mesh texture in harzburgite from the Mineoka ophiolite complex, Japan. (a) Al elemental map around the boundary between clinopyroxene and olivine· The white bold line and the dotted lines show a pseudomorph of clinopyroxene and the outlines of olivine relics, respectively. (b) Back-scattered image of mesh texture shows the distribution of magnetite in the mesh core and veins. (c) Elemental map of Al in the mesh texture shown in Fig. 3b. The Al-rich mesh core (C), Al-rich mesh vein network (V), and Al-poor mesh rim (R) are identified. (d) Al contents (cations per seven oxygen formula unit) along traverse a–b in Fig. 3c. (e) Plot of Si versus Mg + Fe cations (per seven oxygen formula unit) showing the brucite mixing trend in the mesh core and mesh rim. Srp, serpentine; Brc, brucite; Mag, magnetite. Other abbreviations are as in Fig. 2. Fig. 3. View largeDownload slide Serpentine mesh texture in harzburgite from the Mineoka ophiolite complex, Japan. (a) Al elemental map around the boundary between clinopyroxene and olivine· The white bold line and the dotted lines show a pseudomorph of clinopyroxene and the outlines of olivine relics, respectively. (b) Back-scattered image of mesh texture shows the distribution of magnetite in the mesh core and veins. (c) Elemental map of Al in the mesh texture shown in Fig. 3b. The Al-rich mesh core (C), Al-rich mesh vein network (V), and Al-poor mesh rim (R) are identified. (d) Al contents (cations per seven oxygen formula unit) along traverse a–b in Fig. 3c. (e) Plot of Si versus Mg + Fe cations (per seven oxygen formula unit) showing the brucite mixing trend in the mesh core and mesh rim. Srp, serpentine; Brc, brucite; Mag, magnetite. Other abbreviations are as in Fig. 2. A single mesh unit (Fig. 3c), which is separated from other units by Al-rich serpentine veins, shows clear zoning of Al with an Al-rich mesh core and an Al-free mesh rim. The Al content is high in the mesh core (C in Fig. 3d), gradually decreases towards the mesh rim (R in Fig. 3d), and increases in the veins (V in Fig. 3d). We note that even in the core and the veins, the Al content of serpentine in the harzburgite (Fig. 3d; Al = 0·03–0·04 pfu) is an order of magnitude lower than that of the serpentine in the plagioclase-bearing wehrlite (Fig. 2h; Al = 0·6–0·8 pfu). The chemical compositions of the Al-rich veins in the harzburgite probably indicate Tschermak substitution or a R23+Si2O5(OH)4 substitution trend (Fig. 3e). The SiO2 and Mg + ΣFe contents of mesh cores and rims lie on the brucite–serpentine mixing trend (Fig. 3e), suggesting the presence of brucite intergrowths within the serpentine aggregates of the mesh cores and rims. RESULTS OF HYDROTHERMAL EXPERIMENTS Mineral zones that formed during the experiments Figure 4a shows a photomicrograph of the products of the Ol–Pl–H2O experiment for a run time of 7980 h. Figure 4b is an enlargement of part of Fig. 4a. Hereafter, positions in the inner tube shown in this photograph are described relative to the Ol–Pl boundary (x = 0), so that the Pl-hosted region occurs from x = −10 to 0 mm and the Ol-hosted region from x = 0 to ∼30 mm. Fig. 4. View largeDownload slide (a) Photomicrograph of the experimental products for a run of 7980 h in the Ol–Pl–H2O experiments. The x values indicate the distance from the Ol–Pl boundary. (b) Enlarged photo around the Ol–Pl boundary of the area indicated by the rectangle in Fig. 4a. (c)–(d) Back-scattered electron images of reaction products and unaltered reactant minerals after 7980 h. The position where the BSE images were taken was shown in Fig. 4b. (c) and (d) Overviews in the plagioclase-hosted region at (c) x = −5 mm and (d) x = −1 mm. (e) Contact between Pl- and Ol-hosted regions. No reaction products were formed after the 7980 h reaction in the plagioclase-hosted region. (f) and (g) Overviews of the reaction products of (f) the Al-serpentine zone and (g) the serpentine + brucite + magnetite zone. Pl, plagioclase; Ol, olivine; Srp, serpentine; Al-Srp, Al-rich serpentine; Brc, brucite; Mag, magnetite. Fig. 4. View largeDownload slide (a) Photomicrograph of the experimental products for a run of 7980 h in the Ol–Pl–H2O experiments. The x values indicate the distance from the Ol–Pl boundary. (b) Enlarged photo around the Ol–Pl boundary of the area indicated by the rectangle in Fig. 4a. (c)–(d) Back-scattered electron images of reaction products and unaltered reactant minerals after 7980 h. The position where the BSE images were taken was shown in Fig. 4b. (c) and (d) Overviews in the plagioclase-hosted region at (c) x = −5 mm and (d) x = −1 mm. (e) Contact between Pl- and Ol-hosted regions. No reaction products were formed after the 7980 h reaction in the plagioclase-hosted region. (f) and (g) Overviews of the reaction products of (f) the Al-serpentine zone and (g) the serpentine + brucite + magnetite zone. Pl, plagioclase; Ol, olivine; Srp, serpentine; Al-Srp, Al-rich serpentine; Brc, brucite; Mag, magnetite. In the Pl-hosted region, no products were observed after the experiments. The color in the Pl-hosted region changed from brown at the Ol–Pl boundary to beige away from the boundary (Fig. 4a, b) and this color change was probably due to the the larger amount of epoxy resin-filled porosity and smaller grain size of Pl near the Ol–Pl boundary (x = ∼−1 mm; Fig. 4b, d) than in areas farther from the boundary (x = ∼−5 mm; Fig. 4b, c). In the Pl-hosted region, no secondary minerals (e.g. diopside, prehnite, epidote, and grossular) were observed. The color of the Ol-hosted region changes from pale blue at the Ol–Pl boundary to pale yellow in the Al metasomatic zone, and dirty green away from the boundary (Fig. 4a, b). The reaction products in the Ol-hosted region change systematically with distance from the Ol–Pl boundary (Fig. 4a, b) from an Al-serpentine + minor Ca-saponite zone (herein called the Al–Si metasomatic zone), to a serpentine ± magnetite zone (herein called the Si metasomatic zone), and finally to a serpentine + magnetite + brucite zone (herein called the isochemical zone). In the Al–Si metasomatic zone (x = 0–1·5 mm in Fig. 4b), Ol grains are highly reacted, and they show irregular outlines surrounded by product minerals (Fig. 4f). The chemical composition of the product aggregate is Al-rich with small amounts of Ca (< ∼1 wt % of CaO; Table 1). Raman spectra of the products show peaks at 227, 380, 684, 3685, and 3700 (trace) cm-1 (Supplementary Data Fig. S2), indicating that the Al-hydrous phase is lizardite (Enami, 2006). In addition, a peak at 191 cm-1 indicates the presence of talc or saponite in the Al-hydrous phase (Supplementary Data Fig. S2), whereas the peaks for chlorite (cf. the main peaks at 360, 552, 676, and 3675 cm-1 for clinochlore; Kleppe et al., 2003) are lacking. These observations suggest that the Al–Si metasomatic zone is composed mainly of Al-rich lizardite with lesser amounts of Ca-saponite (hereafter called Al-serpentine; Fig. 4e, f). The occurrences of Al-serpentine are described in more detail in a later section. Table 1 Representative chemical compositions of reactant and product minerals in the experiments Plagioclase Olivine Lizardite Brucite Magnetite Al-serpentine (Al > 0·1 pfu) domain. I domain. II domain. III Wt % SiO2 44·36 40·94 42·54 0·15 4·13 36·01 38·55 35·62 TiO2 0·01 0·00 0·00 0·00 0·00 0·03 0·02 0·00 Al2O3 35·70 0·02 0·08 0·12 0·00 10·64 1·73 13·65 Cr2O3 0·00 0·00 0·01 0·00 0·00 0·00 0·01 0·00 FeO 0·43 8·49 2·18 11·67 84·54 7·67 6·09 4·13 MgO 0·09 50·65 41·67 60·72 1·51 32·91 33·78 31·74 MnO 0·09 0·15 0·03 0·25 0·24 0·09 0·10 0·06 CaO 19·58 0·08 0·03 0·00 0·00 0·52 0·32 0·40 Na2O 0·33 0·03 0·02 0·07 0·05 0·00 0·19 0·00 K2O 0·00 0·00 0·00 0·04 0·00 0·00 0·00 0·00 Total 100·52 100·36 86·56 73·02 90·49 87·87 80·79 85·59 Oxygen 8·00 4·00 7·00 7·00 7·00 7·00 7·00 7·00 Si 2·04 0·99 2·00 0·01 0·36 1·72 1·98 1·70 Ti 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Al 1·94 0·00 0·00 0·01 0·00 0·60 0·11 0·77 Cr 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Fe 0·02 0·17 0·09 0·68 6·07 0·31 0·26 0·17 Mg 0·01 1·83 2·91 6·27 0·19 2·34 2·59 2·26 Mn 0·00 0·00 0·00 0·01 0·02 0·00 0·00 0·00 Ca 0·97 0·00 0·00 0·00 0·00 0·03 0·02 0·02 Na 0·03 0·00 0·00 0·01 0·01 0·00 0·02 0·00 K 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 total 5·00 2·99 5·00 6·99 6·65 4·99 4·98 4·92 Mg / (Mg + Fetotal) 0·91 0·97 0·90 0·88 0·91 0·93 Ca / (Ca + Na + K) 0·97 Plagioclase Olivine Lizardite Brucite Magnetite Al-serpentine (Al > 0·1 pfu) domain. I domain. II domain. III Wt % SiO2 44·36 40·94 42·54 0·15 4·13 36·01 38·55 35·62 TiO2 0·01 0·00 0·00 0·00 0·00 0·03 0·02 0·00 Al2O3 35·70 0·02 0·08 0·12 0·00 10·64 1·73 13·65 Cr2O3 0·00 0·00 0·01 0·00 0·00 0·00 0·01 0·00 FeO 0·43 8·49 2·18 11·67 84·54 7·67 6·09 4·13 MgO 0·09 50·65 41·67 60·72 1·51 32·91 33·78 31·74 MnO 0·09 0·15 0·03 0·25 0·24 0·09 0·10 0·06 CaO 19·58 0·08 0·03 0·00 0·00 0·52 0·32 0·40 Na2O 0·33 0·03 0·02 0·07 0·05 0·00 0·19 0·00 K2O 0·00 0·00 0·00 0·04 0·00 0·00 0·00 0·00 Total 100·52 100·36 86·56 73·02 90·49 87·87 80·79 85·59 Oxygen 8·00 4·00 7·00 7·00 7·00 7·00 7·00 7·00 Si 2·04 0·99 2·00 0·01 0·36 1·72 1·98 1·70 Ti 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Al 1·94 0·00 0·00 0·01 0·00 0·60 0·11 0·77 Cr 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Fe 0·02 0·17 0·09 0·68 6·07 0·31 0·26 0·17 Mg 0·01 1·83 2·91 6·27 0·19 2·34 2·59 2·26 Mn 0·00 0·00 0·00 0·01 0·02 0·00 0·00 0·00 Ca 0·97 0·00 0·00 0·00 0·00 0·03 0·02 0·02 Na 0·03 0·00 0·00 0·01 0·01 0·00 0·02 0·00 K 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 total 5·00 2·99 5·00 6·99 6·65 4·99 4·98 4·92 Mg / (Mg + Fetotal) 0·91 0·97 0·90 0·88 0·91 0·93 Ca / (Ca + Na + K) 0·97 Table 1 Representative chemical compositions of reactant and product minerals in the experiments Plagioclase Olivine Lizardite Brucite Magnetite Al-serpentine (Al > 0·1 pfu) domain. I domain. II domain. III Wt % SiO2 44·36 40·94 42·54 0·15 4·13 36·01 38·55 35·62 TiO2 0·01 0·00 0·00 0·00 0·00 0·03 0·02 0·00 Al2O3 35·70 0·02 0·08 0·12 0·00 10·64 1·73 13·65 Cr2O3 0·00 0·00 0·01 0·00 0·00 0·00 0·01 0·00 FeO 0·43 8·49 2·18 11·67 84·54 7·67 6·09 4·13 MgO 0·09 50·65 41·67 60·72 1·51 32·91 33·78 31·74 MnO 0·09 0·15 0·03 0·25 0·24 0·09 0·10 0·06 CaO 19·58 0·08 0·03 0·00 0·00 0·52 0·32 0·40 Na2O 0·33 0·03 0·02 0·07 0·05 0·00 0·19 0·00 K2O 0·00 0·00 0·00 0·04 0·00 0·00 0·00 0·00 Total 100·52 100·36 86·56 73·02 90·49 87·87 80·79 85·59 Oxygen 8·00 4·00 7·00 7·00 7·00 7·00 7·00 7·00 Si 2·04 0·99 2·00 0·01 0·36 1·72 1·98 1·70 Ti 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Al 1·94 0·00 0·00 0·01 0·00 0·60 0·11 0·77 Cr 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Fe 0·02 0·17 0·09 0·68 6·07 0·31 0·26 0·17 Mg 0·01 1·83 2·91 6·27 0·19 2·34 2·59 2·26 Mn 0·00 0·00 0·00 0·01 0·02 0·00 0·00 0·00 Ca 0·97 0·00 0·00 0·00 0·00 0·03 0·02 0·02 Na 0·03 0·00 0·00 0·01 0·01 0·00 0·02 0·00 K 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 total 5·00 2·99 5·00 6·99 6·65 4·99 4·98 4·92 Mg / (Mg + Fetotal) 0·91 0·97 0·90 0·88 0·91 0·93 Ca / (Ca + Na + K) 0·97 Plagioclase Olivine Lizardite Brucite Magnetite Al-serpentine (Al > 0·1 pfu) domain. I domain. II domain. III Wt % SiO2 44·36 40·94 42·54 0·15 4·13 36·01 38·55 35·62 TiO2 0·01 0·00 0·00 0·00 0·00 0·03 0·02 0·00 Al2O3 35·70 0·02 0·08 0·12 0·00 10·64 1·73 13·65 Cr2O3 0·00 0·00 0·01 0·00 0·00 0·00 0·01 0·00 FeO 0·43 8·49 2·18 11·67 84·54 7·67 6·09 4·13 MgO 0·09 50·65 41·67 60·72 1·51 32·91 33·78 31·74 MnO 0·09 0·15 0·03 0·25 0·24 0·09 0·10 0·06 CaO 19·58 0·08 0·03 0·00 0·00 0·52 0·32 0·40 Na2O 0·33 0·03 0·02 0·07 0·05 0·00 0·19 0·00 K2O 0·00 0·00 0·00 0·04 0·00 0·00 0·00 0·00 Total 100·52 100·36 86·56 73·02 90·49 87·87 80·79 85·59 Oxygen 8·00 4·00 7·00 7·00 7·00 7·00 7·00 7·00 Si 2·04 0·99 2·00 0·01 0·36 1·72 1·98 1·70 Ti 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Al 1·94 0·00 0·00 0·01 0·00 0·60 0·11 0·77 Cr 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Fe 0·02 0·17 0·09 0·68 6·07 0·31 0·26 0·17 Mg 0·01 1·83 2·91 6·27 0·19 2·34 2·59 2·26 Mn 0·00 0·00 0·00 0·01 0·02 0·00 0·00 0·00 Ca 0·97 0·00 0·00 0·00 0·00 0·03 0·02 0·02 Na 0·03 0·00 0·00 0·01 0·01 0·00 0·02 0·00 K 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 total 5·00 2·99 5·00 6·99 6·65 4·99 4·98 4·92 Mg / (Mg + Fetotal) 0·91 0·97 0·90 0·88 0·91 0·93 Ca / (Ca + Na + K) 0·97 The Si metasomatic zone was identified as a layer of serpentine ± magnetite in which brucite and Al-serpentine were absent (x = 1·5–3·2 mm in Fig. 4b). A thin layer composed solely of serpentine occurred at x = 1·5–2·4 mm (Fig. 4b). Serpentine in the Si metasomatic zone was also identified as lizardite (Supplementary Data Fig. S2). In the isochemical zone (x > 3·2 mm; Fig. 4b), lizardite formed as fine-grained aggregates of serpentine with a porous structure (Fig. 4g). Olivine was replaced in part by brucite (Fig. 4g), and discrete grains of brucite (∼20 μm) were also observed. Subhedral magnetite with a grain size of ∼10 μm was present (Fig. 4g). Narrow gaps (∼3 μm) were observed between olivine and serpentine, and the original outlines of the olivines were difficult to recognize (Fig. 4g). Extent of hydration The total H2O contents of the solid material (products and unreacted olivine/plagioclase) vary with distance from the Ol–Pl boundary (Fig. 5a). In the Pl-hosted region (segments 1 and 2), the H2O contents are less than 1 wt %, and there was no systematic increase of H2O content with time (Fig. 5a). In contrast, in the Ol-hosted region, hydration proceeded with time (Fig. 5a). The hydrous products were serpentine and brucite, and there was no signature of talc or tremolite in the run products. Far from the Ol–Pl boundary (segments 4–8), the H2O contents of the solids were almost constant throughout the inner tube, but they increased from ∼0·5 wt % at 783 h to ∼1·7–2·0 wt % at 2055 h, ∼5·8 wt % at 3184 h, and ∼8·9 wt % at 7980 h (Fig. 5a). At the contact between olivine and plagioclase at 2055 h, the H2O content of segment three (x = 0–5 mm) was higher than that of the other segments (segments four to eight) in the Ol-hosted region. Fig. 5. View largeDownload slide Temporal evolution of the extent of hydration (LOI: loss of ignition) in the Ol–Pl–H2O experiments, obtained by thermogravimetry. (a) Total weight loss and (b) weight loss for brucite. Brucite was formed in the 3184 h reaction throughout the reaction tube. Fig. 5. View largeDownload slide Temporal evolution of the extent of hydration (LOI: loss of ignition) in the Ol–Pl–H2O experiments, obtained by thermogravimetry. (a) Total weight loss and (b) weight loss for brucite. Brucite was formed in the 3184 h reaction throughout the reaction tube. At 783 and 2055 h, weight loss due to the breakdown of brucite was not observed anywhere in the tube (Fig. 5b). At 3184 and 7980 h, brucite was recognized in all segments of the Ol-hosted region (segments three to eight), but the amount of brucite at the contact with plagioclase (segment three; 0·58 wt % at 3184 h and 0·78 wt % at 7980 h) was lower than at the other sites (segments four to eight; ∼1·0 wt % at 3184 h and ∼1·2 wt % at 7980 h; Fig. 5b). Development of metasomatic zones within the Ol-hosted region Figure 6a–d shows the concentrations of Al (Fig. 6a), Si (Fig. 6b), and Ca (Fig. 6c), and the values of XMg (= Mg / (Mg + Fetotal)) in the products (serpentine ± saponite) (Fig. 6d) as a function of distance from the Ol–Pl boundary. The Al content of serpentine is highest at the Ol–Pl boundary (1·0–1·5 pfu) and decreases away from the boundary (Fig. 6a). The zone of Al metasomatism within the Ol-hosted region enlarged with time, with x = 0·8 mm at 2055 h, 0·9 mm at 3184 h, and 1·6 mm at 7980 h (Fig. 6a). Fig. 6. View largeDownload slide (a) Al contents, (b) Si contents, (c) Ca contents, and (d) XMg (= Mg/(Mg + Fetotal)) values for serpentine (cations per seven oxygen formula unit) as a function of distance from the Ol–Pl boundary. Dashed line shows XMg for olivine. (e) Migration of the Al front, magnetite appearance, and brucite appearance front with time. Each front migrated at a constant rate. The magnetite and brucite appearance front probably migrated in a similar manner. Fig. 6. View largeDownload slide (a) Al contents, (b) Si contents, (c) Ca contents, and (d) XMg (= Mg/(Mg + Fetotal)) values for serpentine (cations per seven oxygen formula unit) as a function of distance from the Ol–Pl boundary. Dashed line shows XMg for olivine. (e) Migration of the Al front, magnetite appearance, and brucite appearance front with time. Each front migrated at a constant rate. The magnetite and brucite appearance front probably migrated in a similar manner. The Si contents of the products (serpentine ± saponite) were lowest at the boundary (∼1·5 pfu), and with increasing distance from the boundary they increased at first, and then decreased gradually (Fig. 6b). The Si metasomatic zone within the Ol-hosted region was identified from the first appearance of euhedral brucite grains, and the Si front was found at x = 1·9 mm at 3184 h and 3·2 mm at 7980 h (Fig. 6b), which means that the Si front extended slightly further than the Al front. At 2055 h, the position of the Si metasomatic front is unclear due to the absence of brucite in the reaction tube. The Ca contents of the reaction products (serpentine ± saponite) were low (<0·1 pfu) at 2055 h throughout the tube (Fig. 6c). The Ca contents of the products increased at first to 0·13 pfu at x = 0·5 mm at 3184 h and 0·18 pfu at x = 0·4 mm at 7980 h, and then decreased away from the Ol–Pl boundary (Fig. 6c). The width of the Ca metasomatic zone within the Ol-hosted region (Fig. 6c) was almost the same as that of the Al metasomatic zone (Fig. 6a, e). The XMg values of the products (serpentine ± saponite) changed with time and space in association with a magnetite front. At 2055 h, the XMg values in the product were roughly constant (∼0·91) and similar to those of the olivine (Fig. 6d). At 3184 h and 7980 h, the spatial variations in XMg values in the products were similar. The XMg values in the product (∼0·91) were almost the same as those of the olivines around the Ol–Pl boundary, but increased with the appearance of magnetite and then became constant (∼0·95) with the appearance of brucite (Fig. 6b, d). The metasomatic fronts of Al and Ca migrated with time with the average migration rate of 0·154 μm/h from 2055 h to 7980 h. The metasomatic fronts of Si, which were recognized as the first appearances of brucite, migrated with time with the average migration rate of 0·259 μm/h from 2055 h to 7980 h (Fig. 6e). The position of the first appearance of magnetite is located between Al and Si front and migrated with the average migration rate of 0·257 μm/h. Zoning of Al in serpentine within the Al–Si metasomatic zone Al-serpentine in the Al–Si metasomatic zone occurs as fine-grained aggregates with clear zoning. The serpentine aggregates can be divided into three domains (Fig. 7a) based on differences in contrast and brightness in the BSE images and they can be described as light grey (domain I), dark grey (domain II), and light grey (domain III). Domain I commonly shows the straight outlines of the olivine grains (Fig. 7a) and most of the interfaces between domain I and the olivine are sharp with no visible gap, though occasionally we observed gaps several micrometres wide (Fig. 7a). Domain II is composed of fibrous serpentine crystals and contains more pores than domain I, although nano-scale pores were observed in domain I (arrows in Fig. 7b). Domain III is in contact with domain II but never in contact with domain I (Fig. 7a, b). Fig. 7. View largeDownload slide Characteristics of Al-serpentine. (a) The Al-serpentine domain I, II, and III are clearly seen in BSE images. (b) Microstructures of Al-serpentine at x = 0·8 mm. The domain II seems to be porous. Black arrows indicate secondary porosities. (c) Representative zoning of Al in an Al-serpentine domain. D, distance from olivine–domain I serpentine interface, L, length from olivine–domain I serpentine interface to domain III. (d) Normalized area of the core part with respect to that of the rim part. Both the core and rim were normalized with areas of olivine + domain I, and increased with time. One standard deviation (1σ) were also presented. (e) and (f) Plots of Mg + Fe vs Si for serpentine from the (e) 3184 h and (f) 7980 h reactions. Different substitutions were observed (see text for details). Fig. 7. View largeDownload slide Characteristics of Al-serpentine. (a) The Al-serpentine domain I, II, and III are clearly seen in BSE images. (b) Microstructures of Al-serpentine at x = 0·8 mm. The domain II seems to be porous. Black arrows indicate secondary porosities. (c) Representative zoning of Al in an Al-serpentine domain. D, distance from olivine–domain I serpentine interface, L, length from olivine–domain I serpentine interface to domain III. (d) Normalized area of the core part with respect to that of the rim part. Both the core and rim were normalized with areas of olivine + domain I, and increased with time. One standard deviation (1σ) were also presented. (e) and (f) Plots of Mg + Fe vs Si for serpentine from the (e) 3184 h and (f) 7980 h reactions. Different substitutions were observed (see text for details). Figure 7c shows the time evolution of Al-zoning at x = 0·8 mm. Even at the same position within the inner tube, the Al-zoning of serpentine changed systematically with time. Zoning of the serpentine aggregates was not developed in the early stages of the experiment (783 and 2055 h), but after 3184 h the zoning into domains I, II, and III became clear. At 3184 h, the Al zoning is characterized by high levels of Al (0·17 pfu) in domain I, a decrease to zero Al in domain II, and an increase to 0·10 pfu in domain III (Fig. 7c). After 7980 h, the Al content of the Al-serpentine in domain I was also high (0·55 pfu), decreased to ∼0·15 in domain II, and increased to 0·54 pfu in domain III (Fig. 7c). The areas of domains I and III were measured separately on BSE images at x = 0·8 mm, and the normalized areas of domains I and III with respect to the total area of olivine + domain I were obtained. Before 2055 h, the areas of domains I and III were set at zero, since zoning was not recognized (Fig. 7d). With time, the areas of domains I and III enlarged, and the normalized areas of domains I and III were 0·60 and 0·38, respectively, at 3184 h (Fig. 7d). At x = 0·8 mm after 7980 h, the normalized areas of domains I and III were 0·78 and 0·72, respectively (Fig. 7d). As described above, the composition of the Al-serpentines varies with distance from the Ol–Pl boundary (Fig. 6a–d). The composition also varies with domain. At 3184 h and x = 0·8 mm, the composition of the Al-serpentines within domain I are explained by the Tschermak substitution (AlAl(Mg, Fe2+)-1Si-1) and brucite mixing (Fig. 7e). The chemical compositions of the Al-serpentines in domain II also indicate the brucite mixing trend, while the Al-serpentines in domain III plot along the brucite and Fe22+Fe3+AlSiO5(OH)4 mixing trends (Fig. 7e). All the Al-serpentines have low contents of Ca (<0·1 pfu), indicating minor effects of saponite mixing. At 7980 h and x = 0·8 mm, the chemical compositions of the Al-serpentines in domains I, II, and III deviate from the Tschermak substitution trend (Fig. 7f), indicating the effects of a substitution of 3R2+(2R3+□)-1 in the M site (R3+ is Fe3+ or Al3+). The amounts of Fe3+ and Al3+ in the Al-serpentine could not be estimated solely from the EPMA data, due to the effect of the Ca-saponite mixing. Solution chemistry Table 2 lists the concentrations of cations in solutions after the Ol–Pl–H2O experiments, the Pl–H2O experiments, and the blank experiments. In the blank experiments, the concentration of Al, Ca, Mg, Fe and Si is below 0·002 mmol/kg (Table 2), which indicates that the effect of the leaching of cations from the vessels were negligible in our experiments. Table 2 Chemistry of solutions in the experiments Time hours Al Ca Fe mmol/kg Mg Si pH (at 25°C) pH (in situ) Ol–Pl–H2O  783 0·008 0·117 0·000 0·056 0·026 6·58 6·49  2055 0·000 0·099 0·005 0·005 0·004 7·15 6·80  3186 0·003 0·040 0·004 0·006 0·006 9·92 7·14  7980 0·000 0·074 0·005 0·003 0·008 10·53 7·66 Pl–H2O  158 0·070 0·306 0·004 0·016 2·155 6·84 7·58  330 0·015 0·351 0·010 0·017 2·392 7·06 7·66  488 0·025 0·457 0·018 0·022 2·407 8·50 7·82 Blank  2224 0·002 0·002 0·000 0·001 0·001 n.d. Time hours Al Ca Fe mmol/kg Mg Si pH (at 25°C) pH (in situ) Ol–Pl–H2O  783 0·008 0·117 0·000 0·056 0·026 6·58 6·49  2055 0·000 0·099 0·005 0·005 0·004 7·15 6·80  3186 0·003 0·040 0·004 0·006 0·006 9·92 7·14  7980 0·000 0·074 0·005 0·003 0·008 10·53 7·66 Pl–H2O  158 0·070 0·306 0·004 0·016 2·155 6·84 7·58  330 0·015 0·351 0·010 0·017 2·392 7·06 7·66  488 0·025 0·457 0·018 0·022 2·407 8·50 7·82 Blank  2224 0·002 0·002 0·000 0·001 0·001 n.d. n.d., not determined. Table 2 Chemistry of solutions in the experiments Time hours Al Ca Fe mmol/kg Mg Si pH (at 25°C) pH (in situ) Ol–Pl–H2O  783 0·008 0·117 0·000 0·056 0·026 6·58 6·49  2055 0·000 0·099 0·005 0·005 0·004 7·15 6·80  3186 0·003 0·040 0·004 0·006 0·006 9·92 7·14  7980 0·000 0·074 0·005 0·003 0·008 10·53 7·66 Pl–H2O  158 0·070 0·306 0·004 0·016 2·155 6·84 7·58  330 0·015 0·351 0·010 0·017 2·392 7·06 7·66  488 0·025 0·457 0·018 0·022 2·407 8·50 7·82 Blank  2224 0·002 0·002 0·000 0·001 0·001 n.d. Time hours Al Ca Fe mmol/kg Mg Si pH (at 25°C) pH (in situ) Ol–Pl–H2O  783 0·008 0·117 0·000 0·056 0·026 6·58 6·49  2055 0·000 0·099 0·005 0·005 0·004 7·15 6·80  3186 0·003 0·040 0·004 0·006 0·006 9·92 7·14  7980 0·000 0·074 0·005 0·003 0·008 10·53 7·66 Pl–H2O  158 0·070 0·306 0·004 0·016 2·155 6·84 7·58  330 0·015 0·351 0·010 0·017 2·392 7·06 7·66  488 0·025 0·457 0·018 0·022 2·407 8·50 7·82 Blank  2224 0·002 0·002 0·000 0·001 0·001 n.d. n.d., not determined. In the Pl–H2O experiments, the concentrations of Al ranged from 0·015 to 0·070 mmol/kg (Table 2). The concentrations of Ca increased with time from 0·306 mmol/kg at 158 h to 0·457 mmol/kg at 488 h, and the concentrations of Si increased from 2·155 mmol/kg at 158 h to 2·407 mmol/kg at 488 h. The pH at room temperature (∼25°C) increased with time from 6·84 at 158 h to 8·50 at 488 h (Table 2). The concentrations of Mg and Fe were higher than in the blank, probably due to contamination from clay minerals in the initial plagioclase grains. In the Ol–Pl–H2O experiments, the solutions were not taken from the inner tube, but their chemistry may represent the solution at the end of the Ol-hosted region (Fig. 1a, b). In the Ol–Pl–H2O experiments, the concentration of Al was relatively high (0·008 mmol/kg) at 783 h and became low (0·001–0·003 mmol/kg) at 2055–7980 h, which is an order of magnitude lower than in the Pl–H2O experiments and similar to the blank experiment (Table 2). The Ca concentration was also high (0·117 mmol/kg) at 783 h before decreasing sharply to near-constant values of 0·040–0·074 mmol/kg at 2055–7980 h, which is also an order of magnitude lower than in the Pl–H2O experiments (Table 2). The Si concentration was high (0·026 mmol/kg) at 783 h before decreasing sharply to near-constant values of 0·002–0·008 mmol/kg at 2055–7980 h, which is three orders of magnitude lower than in the Pl–H2O experiments (Table 2). The concentrations of Fe are extremely low (<0·005 mmol/kg). The concentration of Mg was high (0·056 mmol/kg) at 783 h and decreased to 0·002–0·008 mmol/kg at 2055–7980 h. The pH at room temperature was low (∼6·6) at 783 h and increased to ∼10·5 at 7980 h (Table 2). The solution chemistry for the Pl–H2O experiments is plotted in the activity-diagram of the CaO–Al2O3–MgO–SiO2–H2O system (Fig. 8). Here, only the stability fields of the Mg-endmembers of minerals are shown. The Fe-endmembers were omitted from the system since the boundaries of the minerals would not significantly affect the equilibrium level with the given amount of Fe in each phase. The activity diagram of log aSiO2(aq) versus log aCa2+/a2H+ is shown in Fig. 8a, and the log aSiO2(aq) versus log aAl3+/a3H+ diagram is shown in Fig. 8b. The calculated fluid speciation in the Pl–H2O experiments from 158 to 488 h is largely constant with a log aSiO2(aq) value of ∼3·0 and a log aCa2+/a2H+ value of ∼11·5 (Fig. 8a), and a log aAl3+/a3H+ value of ∼−2·2 (Fig. 8b). The compositions of the solution within the Pl-hosted region in the Ol–Pl–H2O experiments would have been similar to those in the Pl–H2O experiments. Fig. 8. View largeDownload slide Activity diagram in the CaO–Al2O3–MgO–SiO2–H2O system at 230°C and vapor-saturated pressure of 2·80 MPa. The diagrams were projected onto (a) the log aCa2+/a2H+ vs log aSiO2(aq) plane with log aAl3+/a3H+ = −2·5, which is the value for the 2055–7980 h reactions, and (b) the log log aAl3+/a3H+ vs log aSiO2(aq) plane with log aCa2+/a2H+ = 9·1 (grey dashed line) and 11·5 (black solid line), which are the values for the 783 h and 7980 h reactions, respectively. The speciated fluid compositions in the Ol–Pl–H2O and Pl–H2O experiments are shown. Liz, lizardite; Brc, brucite; Chl, Chlorite; Tr, Tremolite; Ca-Sap, Ca-Saponite; Di, Diopside; Tlc, Talc; Qtz, Quartz. Fig. 8. View largeDownload slide Activity diagram in the CaO–Al2O3–MgO–SiO2–H2O system at 230°C and vapor-saturated pressure of 2·80 MPa. The diagrams were projected onto (a) the log aCa2+/a2H+ vs log aSiO2(aq) plane with log aAl3+/a3H+ = −2·5, which is the value for the 2055–7980 h reactions, and (b) the log log aAl3+/a3H+ vs log aSiO2(aq) plane with log aCa2+/a2H+ = 9·1 (grey dashed line) and 11·5 (black solid line), which are the values for the 783 h and 7980 h reactions, respectively. The speciated fluid compositions in the Ol–Pl–H2O and Pl–H2O experiments are shown. Liz, lizardite; Brc, brucite; Chl, Chlorite; Tr, Tremolite; Ca-Sap, Ca-Saponite; Di, Diopside; Tlc, Talc; Qtz, Quartz. Speciated solutions in the Ol–Pl–H2O experiments, taken from the main vessel, were also plotted on the activity diagram of the CaO–Al2O3–MgO–SiO2–H2O system (Fig. 8). At an early stage of the Ol–Pl–H2O experiments (783 h), the fluid fell within the stability field of serpentine (Fig. 8a). The value of log aSiO2(aq) in the solution decreased from −4·5 at 783 h to −5·4 at 2055 h, and the aSiO2(aq) was an order of magnitude lower than in the Pl–H2O experiments (Fig. 8a). The value of log aSiO2(aq) after 2055 h was around -5·4, and fell to around the equilibrium line of serpentine–brucite (Fig. 8a). In the Ol–Pl–H2O experiments, log aCa2+/a2H+ at 783 h was 9·1, and the ratio aCa2+/a2H+ at 783 h was two orders of magnitude lower than in the Pl–H2O experiments. The value of log aCa2+/a2H+ in the Ol–Pl–H2O experiments showed a continuous increase with time from 9·1 at 783 h to 11·0 at 7980 h (Fig. 8a), and became similar to the value in the Pl–H2O experiments. The values of log aCa2+/a2H+and log aSiO2(aq) seem to have been evolving towards equilibrium among serpentine, brucite, and diopside (Fig. 8a). The value of log aAl3+/a3H+ at 783 h (∼1·0) was high in the Ol–Pl–H2O experiments, and it decreased to a constant value (∼ -2·5) from 2055 to 7980 h. Note that the concentration of Al in the Ol–Pl–H2O experiment was extremely low (on the order of a few μg/kg H2O), and thus the log aAl3+/a3H+ value in the Ol–Pl–H2O experiments has large uncertainties. DISCUSSION Metasomatic reactions in the olivine–plagioclase–H2O experiments The chemistries of solutions in the Pl–H2O and Ol–Pl–H2O experiments indicate that the Al and Si that were leached from plagioclase were consumed by olivine hydration within the reaction tube. The high concentration of Ca in the Ol–Pl–H2O solution (Table 1) suggests that the Ca released from Pl was partly consumed in the formation of Ca-saponite in the olivine-hosted metasomatic zone (Fig. 6c). The coupling of Al and Ca in distribution (Fig. 6e) took place because the Ca was contained in Ca-saponite (cf., Ca0·165Mg3Al0·33Si3·67O10(OH)2). As the Ca content of the metasomatic products (serpentine ± saponite) was very low (<0·2 pfu; Fig. 6c), we focus here on a discussion of reactions in the Al2O3–MgO–(FeO)–SiO2–H2O system. Isochemical zone In the early stages of the experiments (783 h), brucite and magnetite were not observed in the reaction tube (Fig. 5b). At this stage, the values of aSiO2(aq) in the solution were higher than the upper limit of the brucite stability field (Fig. 8a, b), which would have inhibited the formation of brucite thermodynamically. The solution at 783 h was supersaturated with lizardite (Fig. 8a, b), suggesting the following Si-adding overall reaction: 3 (Mg0.91, Fe0.09)2SiO4+ SiO2(aq)+ 4 H2Oolivine= 2 (Mg0.91, Fe0.09)3Si2O5(OH)4serpentine (1) Reaction 1 results in an increase in volume of ∼60%. After this induction stage, the solution chemistry moved to the stability field of lizardite + brucite (at 2055, 3184, and 7980 h; Fig. 8a, b), and brucite appeared (Fig. 5b), suggesting that serpentine + brucite were in equilibrium and that the following overall reaction had occurred: (Mg0.91, Fe0.09)2SiO4+ 1.41H2Oolivine→ 0.5 (Mg0.95, Fe0.05)3Si2O5(OH)4serpentine+ 0.368 (Mg0.90, Fe0.10)(OH)2brucite+ 0.04 Fe3O4+ 0.04 H2(aq)magnetite (2) Reaction (2) results in an increase in volume of ∼50%. The change from brucite-absent to brucite-present reactions is similar to that described in previously performed batch-type Ol-H2O experiments (Okamoto et al., 2011; Ogasawara et al., 2013; McCollom et al., 2016) and this suggests that the ‘isochemical’ Ol hydration reaction proceeded far from the Ol–Pl boundary in the Ol–Pl–H2O experiments. Si metasomatic zone The fact that the brucite breakdown front (Si metasomatic front) had advanced farther than the Al metasomatic front (Figs 4b and 6e) indicates that Si-metasomatism was followed by Al–Si metasomatism, forming a metasomatic sequence comprising the Al-serpentine + Ca-saponite zone, the serpentine zone, the serpentine + magnetite zone, and the serpentine + magnetite + brucite zone. When the Si-metasomatic front invaded the isochemical zone (serpentine + brucite + magnetite), brucite breakdown occurred to form serpentine + magnetite, as follows: 19/6(Mg0.90, Fe0.10) (OH)2+ 2 SiO2(aq)brucite → (Mg0.95, Fe0.05)3Si2O5(OH)4serpentine+ 1/18Fe3O4+ 10/9 H2O + 1/9 H+magnetite (3) Reaction 3 results in an increase in volume of ∼40%. The magnetite appearance front followed the brucite appearance front (Figs 4b and 6e) and the values of XMg in the serpentine increased from 0·91 to 0·95 where magnetite appeared (Fig. 6d). These observations suggest the formation of Fe3+-serpentine by the breakdown of magnetite. There are two possible Fe3+-serpentine which have been proposed; Fe23+Si2O5(OH)4 (Klein et al., 2009) and Mg22+Fe3+(Fe3+, Si)O5(OH)4 (Evans et al., 2013). In the case of our experiment, the component of Tschermak substitution looks greater than substitution of 3R2+(2R3+□)-1 in the M site (Fig. 7f), therefore, the magnetite breakdown reaction is written as follows; 4 Fe2+Fe3+2O4+ 3 SiO2(aq)+ 5 H2O + H2(aq)magnetite= 3 Fe2+2Fe3+(Fe3+, Si)O5(OH)4cronstedtite (4) The mineral assemblage during Si-metasomatism is strongly controlled by the chemical potentials of silica (μSiO2) and oxygen (μO2) (Evans et al., 2013; Frost et al., 2013). In our experiments, μSiO2 varied in time and space, induced by diffusion from the Pl-hosted region (Fig. 8a, b). No mineral redox buffers were used in our experiments, but it is reasonable to assume that μO2 would have been buffered by the iron-magnetite buffer (IM; -535 kJ/mol, log fO2 = -49·3), as a reaction vessel of Inconel alloy was used. The μSiO2 versus μO2 pseudosection for the experimental P–T conditions reveals that the mineral assemblage changed with decreasing μSiO2 along the IM buffer from talc to talc + serpentine, serpentine, serpentine + magnetite, serpentine + brucite + magnetite, and finally brucite + magnetite (Fig. 9a). This sequence is largely consistent with the sequence observed in our experiments, although Ca-saponite and Al-serpentine were formed in our experiments due to the effects of Ca and Al (Fig. 4b). The predicted XMg values of serpentine and brucite on the serpentine + magnetite + brucite equilibrium at the IM buffer are 0·95 and 0·91, respectively (Fig. 9b), and these values are similar to the XMg values of the serpentine (0·95; Fig. 6d and Table 1) and brucite (0·90; Table 1) in the isochemical zone. Moreover, the predicted XMg values of serpentine at high silica activity are 0·91 (Fig. 9b), which is similar to the serpentine without magnetite in the Si-metasomatic zone (XMg = 0·91, x = 1·5–2·4 mm at 7980 h; Fig. 6d). Fig. 9. View largeDownload slide (a) SiO2 versus O2 chemical potential pseudosection for olivine (XMg = 0·91) in the FeO–MgO–SiO2–H2O system with H2O saturation at 230°C and a vapor-saturated pressure of 2·80 MPa, as obtained using the software Perple X 6·7·9 (Connolly, 2009). Mineral redox buffers on iron–magnetite (IM) and fayalite–magnetite–quartz buffer (FMQ) were shown in the dashed line. (b) Predicted XMg values of serpentine (XMgSrp) and brucite (XMgBrc) at the IM buffer. XMgBrc was slightly higher than 0·91 because minor magnetite was formed. Srp, serpentine; Brc, brucite; Ol, olivine; Mag, magnetite; Hem, hematite; Tlc, talc. Fig. 9. View largeDownload slide (a) SiO2 versus O2 chemical potential pseudosection for olivine (XMg = 0·91) in the FeO–MgO–SiO2–H2O system with H2O saturation at 230°C and a vapor-saturated pressure of 2·80 MPa, as obtained using the software Perple X 6·7·9 (Connolly, 2009). Mineral redox buffers on iron–magnetite (IM) and fayalite–magnetite–quartz buffer (FMQ) were shown in the dashed line. (b) Predicted XMg values of serpentine (XMgSrp) and brucite (XMgBrc) at the IM buffer. XMgBrc was slightly higher than 0·91 because minor magnetite was formed. Srp, serpentine; Brc, brucite; Ol, olivine; Mag, magnetite; Hem, hematite; Tlc, talc. Al–Si metasomatic zone In the Al–Si metasomatic zone, compositional zoning that formed during the replacement of olivine produced Al-rich domain I, Al-poor domain II, and Al-rich domain III (Fig. 7a, b). We obtained the area proportions (%) of domains I, II, and III from BSE images. Although nano-scale pores could not be considered in our area analyses due to the resolution of the analyzed BSE images, they do not affect the following scenario regarding the development of each domain. The initial amount of olivine within the reaction tube was 45·6 ± 3·9 area % (mean ± one standard deviation, 1σ), which is greater than the area proportion of olivine + domain I (41·2 ± 2·7 area % at 7980 h) and smaller than the area proportion of olivine + domain I + domain II (72·3 ± 5·0 area % at 7980 h). These relationships indicate that the initial olivine grain outlines were located within domain II. Furthermore, the fact that the area of residual olivine + domain I was constant, regardless of the progress of the reaction (41·3 ± 3·0 area % at 3184 h, and 41·2 ± 2·7 area % at 7980 h) suggests that the Al-serpentine in domain I preserved the shape of the olivine at the time when the Al front arrived. In summary, our experiments revealed that (1) Al-free serpentine (domain II) formed at an early stage, and (2) when the Al metasomatic front reached the reaction site, Al-serpentine was formed both inwards (domain I) and outwards (domain III) (Fig. 10a, b). Fig. 10. View largeDownload slide Schematic image of (a) Al front migration and (b) reaction mechanism in the Ol–Pl–H2O experiments from t = t0 (before reaction), t1 (after reaction and before arrival of Al front to a reaction site), and to t2 (after arrival of Al front to a reaction site). (a) As the Al front migrates from t = t0 to t2, the Al concentration in the fluid at a reaction site increases. (b) Reaction mechanism for Al-serpentine formation from t = t0 (left-side diagram), t = t1 (diagram at center), and to t = t2 (right-side diagram). Narrow and bold black arrows indicate the transport of elements and the growth direction, respectively. (c) Schematic of variations in chemical potential of magnesium (μMg), silica (μSiO2), and aluminum (μAl) across domain III to olivine transect at t = t1 and t = t2 in Fig. 10(a) and (b). Fig. 10. View largeDownload slide Schematic image of (a) Al front migration and (b) reaction mechanism in the Ol–Pl–H2O experiments from t = t0 (before reaction), t1 (after reaction and before arrival of Al front to a reaction site), and to t2 (after arrival of Al front to a reaction site). (a) As the Al front migrates from t = t0 to t2, the Al concentration in the fluid at a reaction site increases. (b) Reaction mechanism for Al-serpentine formation from t = t0 (left-side diagram), t = t1 (diagram at center), and to t = t2 (right-side diagram). Narrow and bold black arrows indicate the transport of elements and the growth direction, respectively. (c) Schematic of variations in chemical potential of magnesium (μMg), silica (μSiO2), and aluminum (μAl) across domain III to olivine transect at t = t1 and t = t2 in Fig. 10(a) and (b). In domain I, the shape of the olivine was preserved, but this shape had already been modified from the original shape of the olivine before alteration during the earlier serpentinization that took place during the formation of domain II (Fig. 10b). As it was difficult to measure the exact volume of nano-pores, the following mass balance equations were obtained assuming an isovolume reaction (i.e. no nano-pores) during the formation of domain I as a first-order approximation. We note, too, that the isovolume reaction is only for the mass balance in domain I, and that the overall reaction for domains I + III involved an increase in the volume of the solid material. In the Al2O3–MgO–SiO2–H2O system, the pseudomorphic reaction in domain I can be written as follows: αMg2SiO4+βAl3++ (–2α+β+ 5) H2O + (4α– 2β+ 6) H+→olivineMg3–0.5βAlβSi2–0.5βO5(OH)4+ (2α+ 0.5β– 3) Mg2++ (α+ 0.5β– 2) SiO2(aq)Al−serpentine (5) where α and β indicate the coefficients of olivine and Al3+ required to produce 1 mol of Al-serpentine, respectively. The Fe and Ca components are ignored for simplicity. Assuming reaction (5) is an isovolume reaction, α is written as V-Al-srp/ V-Fo, where V-Fo and V-Al-srp are the molar volumes of forsterite and Al-serpentine, respectively. The V-Al-srp was obtained from the linear mixing of the molar volume of lizardite (Mg3Si2O5(OH)4) ( V-Liz) and amesite (Mg4Al4Si2O10(OH)8) ( V-Ame), and calculated as V-Al-srp =  V-Liz - 0·5 ( V-Liz - 0·5 V-Ame) β. The molar volumes of minerals were taken from Holland & Powell (2011). The value of β varies with the chemical composition of domain I Al-serpentine. We emphasize again that the estimated amounts of removed Mg and Si in reaction (5) are minimum estimates (no pores). Given the average Al content of domain I at 7980 h (α = 2·42 and β = 0·42), the typical reaction can be written as follows: 2.42 Mg2SiO4+ 0.42 Al3++ 0.58 H2O + 14.84 H+→Mg2.79Al0.42Si1.79O5(OH)4+ 2.05 Mg2++ 0.63 SiO2(aq) (6) This typical reaction of domain I proceeds with a supply of Al and with the removal of Mg and SiO2(aq) at the reaction front between domain I and olivine (Fig. 10b). The Mg and Si released from olivine during reaction (5) were transported outwards through the nano-scale pores, and used in the precipitation of Al-serpentine in domain III with a supply of Al and Si in the pore fluids from the Pl-hosted region, as follows: (3 – 0.5β) Mg2++ (2 – 0.5β) SiO2(aq)+βAl3++ (5 –β) H2O→Mg3–0.5βAlβSi2–0.5βO5(OH)4+ (6 – 2β) H+ (7) When we assume a balance between the Mg released in reaction (5) (domain I) and the Mg consumed in reaction (7) (domain III), the overall reaction for the formation of domains I and III could be rewritten as follows: (3 – 0.5β) Mg2SiO4+ 2βAl3++ (1 –β) SiO2(aq)+ (4β+ 4) H2O →Olivine2 Mg3–0.5βAlβSi2–0.5βO5(OH)4+ 8βH+Al−serpentine (8) Since the Al content in serpentine (β) ranges from 0–1·0 (Fig. 6a), this overall reaction represents the process where the serpentinization in the metasomatic zone proceeds with a supply of Si and Al from plagioclase, except at the Ol–Pl boundary, where the β value exceeds 1·0 (Fig. 6a). Therefore, at the Ol–Pl boundary, reaction (7) represents the overall reaction that proceeds with the supply of Al and the removal of Si. Figure 10c shows schematic profiles of the chemical potential along a traverse across domains I, II, and III. Since the source of Mg was olivine, the chemical potential of Mg (μMg) is high in contact with olivine, and decreases from domain I to domain II to domain III (Fig. 10c). Si and Al were supplied from plagioclase through macroscopic pores, and thus the chemical potential of Al (μAl) and silica (μSiO2) would be high at the margin of domain III, but decreases towards the olivine interface (Fig. 10c). The directions of Mg and Al transport, as inferred from the mass balance (Fig. 10b), are consistent with the downhill nature of the chemical potential by diffusion. In contrast, the transport of Si seems to occur in the direction opposed to the chemical potential gradient (Fig. 10b, c). Such uphill diffusion could have occurred during the formation of domain I due to the constant-volume constraint and the effects of diffusion of other elements (e.g. Nishiyama, 1998). Controls on pseudomorphic replacement One of the most important findings of our study is that the pseudomorphic replacements that occurred in domain I in the Al metasomatic zone were coupled with overgrowths in domain III (Figs 7a, b and 9b). Similar combinations of replacements and overgrowths in volume-increasing reactions have often been reported, as for example in the replacement of hematite by chalcopyrite (∼230% volume increase; Zhao et al., 2014) and magnetite by pyrite (∼65% volume increase; Qian et al., 2010). In contrast to the metasomatic zone, in the isochemical zone, the original olivine outlines were not preserved during serpentinization (Fig. 4g). The question is: what causes such differences in replacement textures, given that both reactions show similar increases in volume (40–50%)? Pseudomorphic replacements (i.e. preserving the outline of the parent mineral) are thought to proceed via interface-coupled dissolution–precipitation reactions (e.g. Putnis, 2009), and the textural clues to this process are porosity in the product phase and sharp interfaces between the parent and product minerals. Such textures are observed in the Al metasomatic zone, involving the interfaces between the olivine and domain I (Fig. 7a), and the porosity in the product phase (Fig. 7b). The macroscopic pores among the original olivine grains (∼10% porosity on the >10 μm scale) provided the main element transport path, but microscopic pores (nano-scale) in the replacement materials were also important for the progress of the reaction at the fluid–olivine interface. The process of serpentinization involves dissolution of the primary mineral, transport of the solute, and precipitation of secondary minerals; the relative rates of these processes are critical for pseudomorphism. Xia et al. (2009) showed that a high degree of pseudomorphism could be observed when the reaction is either transport-limited or dissolution-limited, but not precipitation-limited. Previous batch-type experiments of olivine hydration (reaction (2)) revealed that the rate of serpentinization increases with a decrease in the initial grain size of olivine, indicating that the rate-limiting process is the dissolution of olivine (Wegner & Ernst, 1983; Malvoisin et al., 2012). This seems to contradict our experimental results, where the replacement was not pseudomorphic in the isochemical region (Fig. 4d), and it is unclear why pseudomorphic replacement occurred only in domain I of the metasomatic zone. In both the metasomatic and isochemical zones, the rates of solute transport (rtrans) and precipitation of serpentine (rpre) would have been greater than the rate of olivine dissolution (rdis), as shown by the results of the powder experiments of olivine hydration undertaken by Wegner & Ernst (1983) and Malvoisin et al. (2012). We consider that the key to pseudomorphism during serpentinization is the relative magnitudes of rpre and rtrans. In the isochemical zone, solute transport would have been faster than the precipitation of secondary minerals (rdis < rpre < rtrans) and this would have allowed a dispersed distribution of secondary minerals outside the original outlines of the olivine (Fig. 4g). In contrast, when solute transport is slower than the precipitation rate (rdis < rtrans< rpre), the dissolved materials would be consumed immediately before their transport to distant sites, and this would result in the preservation of the outlines of parent minerals in dissolution-limited reactions. The change in the replacement textures between the metasomatic (pseudomorphic) and isochemical zones (non-pseudomorphic) implies that a supply of Al and Si triggered the change in the relative magnitudes of rtrans and rpre. It is known that hydrothermal solutions containing Al and Si can contain complexes such as Al(OH)3H3SiO4- (Salvi et al., 1998; Andreani et al., 2013). Figure 11 shows the abundance of Al species at 230°C and vapor-saturated pressure as a function of aSiO2(aq). With a low activity of silica around the serpentine–brucite equilibria (log aSiO2(aq) = −5·5), the dominant Al species is Al(OH)4−, and this decreases with increasing silica activity. In contrast, Al(OH)3H3SiO4− appears in the high aSiO2(aq) condition (log aSiO2(aq) > −5), and the fraction of Al(OH)3H3SiO4− exceeds Al(OH)4− at log aSiO2(aq) > −2·5. The silica activity inferred from the mineral assemblages (Fig. 9a) and the solution chemistry observed in our experiments (Fig. 8a, b) suggests that Al(OH)4- was dominant in the isochemical zone, whereas abundant Al(OH)3(HSiO4)- existed in the Al-Si metasomatic zone (Fig. 11). Andreani et al. (2013) reported that the formation of Al–Si complexes enhances the dissolution of olivine. In our experiments, the Al–Si complex would already have formed in the pore fluids derived from the Pl-hosted region. The large size of such a complex could have slowed the transport of Al to the reaction front through the nano-pores of domain I, similar to the mechanism suggested for silica complexes (Applin, 1987). Accordingly, the formation of Al(OH)3(HSiO4)– possibly changed the relative rates among the processes from rdis < rpre < rtrans in the isochemical zone to rdis < rtrans < rpre in the metasomatic zone. Further studies on the formation and diffusivities of Al–Si complexes, including molecular dynamic simulations, are needed to understand the impact of Al mobility on the overall reaction rates. Fig. 11. View largeDownload slide Molar fraction of dissolved Al-species at 230°C and vapor-saturated pressure of 2·80 MPa with pHin situ = 7·0 and aH2O = 1. The thermodynamic data are from Tagirov & Schott (2001), and references therein. Fig. 11. View largeDownload slide Molar fraction of dissolved Al-species at 230°C and vapor-saturated pressure of 2·80 MPa with pHin situ = 7·0 and aH2O = 1. The thermodynamic data are from Tagirov & Schott (2001), and references therein. Linkage of macroscopic Al diffusion and the development of Al zoning Al-Si metasomatism near the Ol-Pl contact caused the macroscopic spatial variation of Al content in the products (Figs 4a and 6a) and the temporal changes in Al concentrations in the solution. To establish the link between macroscopic Al transport during metasomatism and the development of Al zoning in the replacement products in the Ol-hosted region, we consider the one-dimensional reaction–diffusion model within a porous medium (Lasaga, 1998; Philpotts & Auge, 2009): ∂(φCi)∂t=∂∂xφDi∂Ci∂x + φ∑Rchem (9) where φ, Ci, Di and Rchem respectively indicate porosity, the concentration of species i in the solution (mol/cm3 H2O), the diffusivity of species i in porous media (cm2/s), the rate of gain or loss of species i by surface reaction (mol/cm3 H2O/s). For simplicity, we just considered a single overall reaction, and φ and Di were assumed to be constant during the reaction. When we assumed the local equilibrium of Al between minerals and fluid, the equilibrium fluid/solid partition coefficient by volume was defined as KV=Ci/CiSolid ( CiSolid is the concentration of species i in the total mineral (mol/cm3 solid)). Although KV varies with fluid composition, it was introduced as a constant. Under these assumptions, the reaction-diffusion equation in the Ol-hosted region used in the simulations of this study are written as follows: φKV∂CAlSolid∂t=φKVDAl∂2CAlSolid∂x2-(1-φ)∂CAlSolid∂t (10) which could be rearranged to yield ∂CAlSolid∂t=DAl*∂2CAlSolid∂x2 (11) where D*Al (cm2/s) is the apparent diffusion coefficient of Al in the porous medium, which was defined as DAl* =φKVφKV + (1-φ)DAl (12) Equation (11) can be solved analytically as follows: CAlSolid=CAl, 0Solid+(CAl, BSolid – CAl, 0Solid) erfcx2DAl*t (13) where the erfc is the complimentary error function, and the t (second) is the reaction time. The CAl, 0Solid and CAl, BSolid are respectively the concentration of Al in the solid phase at t = 0 (initial conditions), and the concentration of Al in the solid phase at the Ol–Pl boundary (boundary conditions). The values of CAl, 0Solid and CAl, BSolid were set as 0·00 (mol/cm3 solid) and 9·00 × 10-3 (mol/cm3 solid), respectively, based on the observed values of CAlSolid in our experiments. The values of CAlSolid in our experiments were calculated by mass balance using the Al content of the serpentine (β; Fig. 6a) and the modal abundance of Al-serpentine (VAl-srp), and by CAlSolid = β × (1/(1 − φ)) × (VAl-srp/ V-Al-srp). Since KV and DAl in equation (12) could not be obtained independently, the values of D*Al that fit the observed values of CAlSolid at both 3184 h and 7980 h were optimized by the least squares method. The result of simple reaction–diffusion modeling at 3184 and 7980 h are shown in Fig. 12a and b, respectively. The value of D*Al was estimated to be 2·63 × 10-7 (cm2/s), although the estimated diffusivity was on the order of that at ∼10% porosity. The model largely reproduced the Al concentration profile as a function of distance from the Ol–Pl boundary in our experiments (Fig. 12a, b). The model also reproduced the Al front, defined as the distance that satisfies CAlSolid/CAl, BSolid = 0·01, observed in our experiments (Fig. 12c). Figure 12d shows the time- evolution of the calculated CAlSolid at x = 0·5, 0·8, and 1·0 mm. With the progress of the metasomatic front, CAlSolidin the metasomatic zone at a fixed distance increased with time (Fig. 12d). For example, at x = 0·8 mm the value of CAlSolid was <0·1 (mmol/cm3 solid) until 2000 h, and it then increased from 0·1 (mmol/cm3 solid) at 2000 h to 2·0 (mmol/cm3 solid) at 8000 h (Fig. 12d). This increase in the Al content of the Al-serpentine was probably recorded as the growth zoning of domains I and III (Fig. 7c), although the directions of growth are opposite to each other (inwards and outwards, respectively; Fig. 10b). Fig. 12. View largeDownload slide Results of numerical simulations using a one-dimensional reaction–diffusion model. The numerical simulations were carried out from x = 0 mm to 2·0 mm in the Ol-hosted region. (a,b) Al concentrations in bulk solid (CSolidAl) after (a) 3184 h and (b) 7980 h reactions. The values of CSolidAl obtained from experiments (symbols), presented as the mean value of the Al concentration with one standard deviation (1σ), are compared with the calculated values from the numerical simulations (lines). (c) The Al fronts (symbols) obtained from experiments (Fig· 6e) and calculations (lines). (d) Calculated time evolution of CSolidAl at fixed distances of x = 0·5, 0·8, and 1·0 mm. Fig. 12. View largeDownload slide Results of numerical simulations using a one-dimensional reaction–diffusion model. The numerical simulations were carried out from x = 0 mm to 2·0 mm in the Ol-hosted region. (a,b) Al concentrations in bulk solid (CSolidAl) after (a) 3184 h and (b) 7980 h reactions. The values of CSolidAl obtained from experiments (symbols), presented as the mean value of the Al concentration with one standard deviation (1σ), are compared with the calculated values from the numerical simulations (lines). (c) The Al fronts (symbols) obtained from experiments (Fig· 6e) and calculations (lines). (d) Calculated time evolution of CSolidAl at fixed distances of x = 0·5, 0·8, and 1·0 mm. There are some discrepancies between the observed and calculated Al profile around x = 0–0·5 mm at 7980 h (Fig. 12b). Such discrepancies probably result from the effects of the porosity reduction and grain-scale compositional heterogeneity (Al zoning; Fig. 7c) in the experimental products, which were not included in the model. Further detailed modelling including porosity changes and surface reaction rate is needed to reproduce fully the CAlSolid observed in our experiments. Implications for natural hydrothermal alteration Millimetre-scale Al–Si metasomatism The results of our experiments are comparable to the natural alteration of Pl-bearing mafic or ultramafic rocks (including troctolites) in the oceanic lithosphere. Frost et al. (2008) reported that prehnite/grossular replaces plagioclase and chlorite replaces serpentine in olivine-rich troctolite. The estimated distances of movement of Si and Al from plagioclase to the olivine grains were ∼5 mm and ∼1 mm, respectively. In the plagioclase wehrlite from the Izu–Ogasawara forearc region (Fig. 2), the Si front inferred from the first appearance of brucite (>1·0 mm from the Ol–Pl boundary) is more advanced than the Al-front (a zone of Al-rich serpentine; ∼0·9 mm from the Ol–Pl boundary). This relative elemental mobility is consistent with that observed in our experiments (x = 3·2 mm for Si and 1·6 mm for Al after 7980 h; Fig. 6e). Moreover, the lack of magnetite around plagioclase grains in the plagioclase wehrlite is also consistent with that observed in our experiments. In the plagioclase wehrlite from the Izu–Ogasawara fore-arc region, olivine reacted at a faster rate in the Al–Si metasomatic zone (<1·0 mm from the Ol–Pl boundary; Fig. 2b) than in the isochemical zone (>1·0 mm from the Ol–Pl boundary; Fig. 2b). Andreani et al. (2013) and Pens et al. (2016) suggested that a supply of Al enhances hydration, whereas Ogasawara et al. (2013) showed that a supply of silica enhances the hydration rate of olivine with absence of brucite. Since plagioclase supplied Al and Si simultaneously to the olivine region (Fig. 2a), it is difficult to distinguish the effects of Si and Al on the olivine hydration rate. Al-zoning of serpentine aggregate in the mesh texture The Al-zoning in the natural mesh texture, with Al-rich cores, Al-poor rims, and Al-rich serpentine veins (Figs 2g and 3c), is essentially the same as that produced in our experiments (Fig. 7c, e). It is reasonable, therefore, to consider that the mesh rims were formed at the earlier stage under Al-free conditions and that the mesh cores and veins were formed at a later stage when Al was supplied from Al-bearing minerals such as plagioclase and clinopyroxene. This scenario is consistent with the proposed two-stage process of mesh formation, which involves a transition from a closed to an open system during serpentinization (e.g. Viti & Mellini, 1998; Bach et al., 2006). Natural rocks have low porosities compared with the reactants in our experiments and the elements being removed from the mesh core during pseudomorphic replacement would migrate into fractures to form Al-rich veins (Figs 2g, h and 3c, d). In our experiments, the Al contents in domain I (Fig. 7c) are similar to those in the mesh core of the Pl-wehrlite (Fig. 1h), but are an order of magnitude lower than in the mesh core of the harzburgite (Fig. 2d). As suggested by the reaction–diffusion modeling (Fig. 12), the differences in the Al contents of serpentine would have been controlled by the difference in the Al content between Al-source minerals ( CAlSolid; Cpx vs Pl). The preservation of Al-poor domains (domain II) or mesh rims in the zoning (Figs 2g and 3c) indicates that once serpentine was formed from Al-free solutions, the kinetics of re-equilibration was very slow, probably because such re-equilibration involved a two-site coupled substitution (i.e. Tschermak substitution; Anovitz, 1991). However, in our experiments, Al contents in domain II did increase with time to some extent (up to ∼0·2 pfu at 7980h; Fig. 7c), whereas the Al content of the mesh rim in the Pl-bearing wehrlite remained very low (∼0·05 pfu; Fig. 2h). We cannot explain this difference, but it is possibly due to the poor crystallinity of our experimental serpentine. The poor crystallinity of serpentine minerals that were formed in hydrothermal experiments has been reported previously, as judged from relatively high water contents rather than structural formulae (e.g. Kühnel et al., 1975; McCollom et al., 2016) and the poor crystallinity makes the mineral susceptible to chemical alteration via local dissolution and precipitation processes. A potential reference frame of volume changes during serpentinization Serpentinization often proceeds without the formation of brucite, but it is difficult to decide whether brucite-absent serpentinization is induced by a supply of Si, the removal of Mg, or the removal of both Si and Mg (e.g. Putnis, 2009; Putnis & Austrheim, 2010; Malvoisin, 2015) and this difficulty reflects the lack of a reliable frame of reference for mass balance analyses in natural systems. Our results presented here suggest that (1) mesh rims were formed under Al-free conditions, and (2) the mesh cores and veins were formed at a later stage when Al was supplied. During the formation of the mesh rims (the first stage of mesh texture formation; t = t1 in Fig. 10b), volume expansion would have occurred, but it is difficult to estimate accurately the extent of expansion because the initial olivine–fluid boundary is generally unclear. In this study we have shown that the shape of olivine was preserved during the formation of the Al-rich mesh cores (i.e. in the second stage of mesh texture formation; t = t2 in Fig. 10c), which means that a large loss of material occurred during serpentinization of the mesh cores. This provides a reliable physical reference frame for deciphering volumetric changes after the arrival of Al in natural serpentinized rocks, which involves reactions 5–8. With an assumption of isovolumetric replacement of the mesh core, the volume increase at this stage is equal to the volume of veins in the mesh texture. Accordingly, the volume increase ratio can be obtained easily by (area of mesh core + area of vein) / (area of mesh core). For example, the mesh core and vein formation in the harzburgite from the Mineoka ophiolite resulted in a volume increase of ∼13%, and Pl-bearing wehrlite from the Izu–Ogasawara forearc region shows a volume increase of ∼16%. Aluminum occurs in a number of different minerals, including pyroxenes, spinel and plagioclase, in both crustal and mantle rocks (Mellini et al., 2005; Boschi et al., 2006; Frost et al., 2008; Plümper et al., 2012a). In the Pl-bearing werhlite, radial fractures filled by Al-rich serpentine and chlorite were produced around the Pl grain in the metasomatic zone (Fig. 2a). The breakdown of the Al-bearing minerals could release aluminum as well as silica episodically and locally, which results in the heterogeneous progression of serpentinization. The preservation of Al-zoning observed in the mesh texture in the natural rocks (Figs 2g and 3c) suggests that the effective bulk composition of the peridotite could vary with the progression of serpentinization, since the serpentinized parts of the peridotite would become less reactive once they formed. The local increase in serpentinization rate around Al-bearing minerals could also cause reaction-induced fracturing by local volume expansion (Jamtveit et al., 2008; Kuleci et al., 2017). The reaction-induced fracturing would enhance fluid supply, thus leading to a positive feedback between hydration and fracturing. Such positive chemical–mechanical feedback induced by elemental transport provides an important means of enhancing the overall hydration rates around plagioclase–olivine grain contacts and, on a much bigger scale, the crust–mantle boundary. CONCLUSIONS (1) Prominent Al-zoning is present in the mesh textures of serpentinized olivine in samples of harzburgite and Pl-bearing wehrlite from the oceanic lithosphere. In both rocks, the mesh zoning is characterized by Al-rich mesh cores, Al-poor mesh rims and Al-rich veins. (2) Hydrothermal experiments were conducted in the olivine–plagioclase–H2O system at 230°C and under a vapor-saturated pressure. Metasomatism took place during the experiments so that the mineral assemblages changed systematically with distance from the olivine–plagioclase boundary, with Al-rich serpentine + Ca-saponite in the metasomatic zone changing to serpentine, serpentine + magnetite and serpentine + brucite + magnetite. This change of mineral assemblage is explained by decrease in activity of SiO2(aq). (3) In the metasomatic zone, the mesh texture of serpentine showed a clear zoning with an Al-rich domain I, an Al-poor domain II, and an Al-rich domain III. Microtextural and chemical features indicate that this zoning was produced by the initial formation of domain II in an Al-free solution, and the subsequent pseudomorphic replacement of olivine (domain I) and the development of overgrowths (domain III) as the Al metasomatic front migrated. During the pseudomorphic replacement, the transport of large amounts of various elements was facilitated by the nano-scale porosity of the products. The transition of the replacement texture from a non-pseudomorphic to a pseudomorphic product was probably triggered by the supply of Al. The formation of Al–Si complexes probably slowed the rate of diffusive transport rather than the precipitation of serpentine. (4) The similarities between Al zoning in natural mesh textures and the Al zoning in our experimental products suggest that the textures are initiated by the formation of mesh rims in solutions that are initially Al-free. Al is subsequently supplied to the solutions, and an Al metasomatic front develops and migrates with the simultaneous pseudomorphic replacement of olivine to form mesh cores and the precipitation of Al-rich serpentine veins. Such transition from closed to open system could be enhanced by fracturing induced by heterogeneous distribution of Al-Si-bearing minerals. ACKNOWLEDGMENTS We thank Masaoki Uno, Nobuo Hirano, Takayoshi Nagaya, Fumiko Higashino, and Otgonbayar Dandar for valuable discussions, and the captain and crew of the R/V Hakuo–Maru, Osamu Ishizuka, and the KH07–02 science party for their support and cooperation. We also thank Jeffrey Alt, Katy Evans, an anonymous reviewer, and editor James Beard for their helpful comments the manuscript. FUNDING This work was supported by a Grant-in-Aid for Research Fellows from the Japan Society for the Promotion of Science [JP16J04140 to R.O.], a Program for Leading Graduate Schools [to R.O.], a Grant-in-Aid for Scientific Research [16H06347 to A.O.], a Grant-in-Aid for Young Scientists [23684042 to A.O.], and a Grant-in-Aid for Specially Promoted Research [25000009 to N.T.]. SUPPLEMENTARY DATA Supplementary data are available at Journal of Petrology online. REFERENCES Andreani M. , Daniel I. , Pollet-Villard M. ( 2013 ). Aluminum speeds up the hydrothermal alteration of olivine . American Mineralogist 98 , 1738 – 1744 . Google Scholar CrossRef Search ADS Anovitz L. M. ( 1991 ). Al zoning in pyroxene and plagioclase: window on late prograde to early retrograde P–T paths in granulite terranes . American Mineralogist 76 , 1328 – 1343 . Applin K. ( 1987 ). The diffusion of dissolved silica in dilute aqueous solution . Geochimica et Cosmochimica Acta 51 , 2147 – 2151 . Google Scholar CrossRef Search ADS Bach W. , Paulick H. , Garrido C. J. , Ildefonse B. , Meurer W. P. , Humphris S. E. ( 2006 ). Unraveling the sequence of serpentinization reactions: petrography, mineral chemistry, and petrophysics of serpentinites from MAR 15°N (ODP Leg 209, Site 1274) . Geophysical Research Letters 33 , L13306 . Google Scholar CrossRef Search ADS Beard J. S. , Frost B. R. ( 2016 ). The stoichiometric effects of ferric iron substitutions in serpentine from microprobe data . International Geology Review 6814 , 1 – 7 . Google Scholar CrossRef Search ADS Beard J. S. , Frost B. R. , Fryer P. , McCaig A. , Searle R. , Ildefonse B. , Zinin P. , Sharma S. K. ( 2009 ). Onset and progression of serpentinization and magnetite formation in olivine-rich troctolite from IODP hole U1309D . Journal of Petrology 50 , 387 – 403 . Google Scholar CrossRef Search ADS Bethke C. M. ( 2007 ). Geochemical and Biogeochemical Reaction Modeling . Cambridge, UK : Cambridge University Press . Google Scholar CrossRef Search ADS Boschi C. , Früh–Green G. L. , Delacour A. , Karson J. A. , Kelley D. S. ( 2006 ). Mass transfer and fluid flow during detachment faulting and development of an oceanic core complex, Atlantis Massif (MAR 30°N) . Geochemistry, Geophysics, Geosystems 7 , Q01004 . Google Scholar CrossRef Search ADS Boudier F. , Baronnet A. , Mainprice D. ( 2010 ). Serpentine mineral replacements of natural olivine and their seismic implications: oceanic lizardite versus subduction-related antigorite . Journal of Petrology 51 , 495 – 512 . Google Scholar CrossRef Search ADS Connolly J. A. D. ( 2009 ). The geodynamic equation of state: what and how . Geochemistry, Geophysics, Geosystems 10 , Q10014 . Google Scholar CrossRef Search ADS Enami M. ( 2006 ). Mineralogical methods for identification of asbestos and their limitations . Japanese Magazine of Mineralogical and Petrological Sciences 35 , 11 – 21 . (Japanese with English abstract). Google Scholar CrossRef Search ADS Escartín J. , Mével C. , MacLeod C. J. , McCaig A. M. ( 2003 ). Constraints on deformation conditions and the origin of oceanic detachments: the Mid-Atlantic Ridge core complex at 15°45′N . Geochemistry, Geophysics, Geosystems 4 , 1067 – 1103 . Google Scholar CrossRef Search ADS Evans K. A. , Powell R. , Frost B. R. ( 2013 ). Using equilibrium thermodynamics in the study of metasomatic alteration, illustrated by an application to serpentinites . Lithos 168–169 , 67 – 84 . Google Scholar CrossRef Search ADS Frost B. R. , Beard J. S. , McCaig A. , Condliffe E. ( 2008 ). The formation of micro-rodingites from IODP hole U1309D: key to understanding the process of serpentinization . Journal of Petrology 49 , 1579 – 1588 . Google Scholar CrossRef Search ADS Frost B. R. , Evans K. A. , Swapp S. M. , Beard J. S. , Mothersole F. E. ( 2013 ). The process of serpentinization in dunite from New Caledonia . Lithos 178 , 24 – 39 . Google Scholar CrossRef Search ADS Harigane Y. , Michibayashi K. , Morishita T. , Tani K. , Dick H. J. B. , Ishizuka O. ( 2013 ). The earliest mantle fabrics formed during subduction zone infancy . Earth and Planetary Science Letters 377–378 , 106 – 113 . Google Scholar CrossRef Search ADS Hirauchi K. , Katayama I. , Uehara S. , Miyahara M. , Takai Y. ( 2010 ). Inhibition of subduction thrust earthquakes by low-temperature plastic flow in serpentine . Earth and Planetary Science Letters 295 , 349 – 357 . Google Scholar CrossRef Search ADS Holland T. J. B. , Powell R. ( 2011 ). An improved and extended internally consistent thermodynamic dataset for phases of petrological interest, involving a new equation of state for solids . Journal of Metamorphic Geology 29 , 333 – 383 . Google Scholar CrossRef Search ADS Hyndman R. D. , Peacock S. M. ( 2003 ). Serpentinization of the forearc mantle . Earth and Planetary Science Letters 212 , 417 – 432 . Google Scholar CrossRef Search ADS Ishizuka O. , Tani K. , Reagan M. K. , Kanayama K. , Umino S. , Harigane Y. , Sakamoto I. , Miyajima Y. , Yuasa M. , Dunkley D. J. ( 2011 ). The timescales of subduction initiation and subsequent evolution of an oceanic island arc . Earth and Planetary Science Letters 306 , 229 – 240 . Google Scholar CrossRef Search ADS Iyer K. , Rüpke L. H. , Morgan J. P. ( 2010 ). Feedbacks between mantle hydration and hydrothermal convection at ocean spreading centers . Earth and Planetary Science Letters 296 , 34 – 44 . Google Scholar CrossRef Search ADS Jamtveit B. , Malthe-Sørenssen A. , Kostenko O. ( 2008 ). Reaction enhanced permeability during retrogressive metamorphism . Earth and Planetary Science Letters 267 , 620 – 627 . Google Scholar CrossRef Search ADS Johnson J. W. , Oelkers E. H. , Helgeson H. C. ( 1992 ). SUPCRT92: a software package for calculating the standard molal thermodynamic properties of minerals, gases, aqueous species, and reactions from 1 to 5000 bar and 0 to 1000°C . Computers & Geosciences 18 , 899 – 947 . Google Scholar CrossRef Search ADS Katayama I. , Kurosaki I. , Hirauchi K. ( 2010 ). Low silica activity for hydrogen generation during serpentinization: an example of natural serpentinites in the Mineoka ophiolite complex, central Japan . Earth and Planetary Science Letters 298 , 199 – 204 . Google Scholar CrossRef Search ADS Kelley D. S. , Karson J. A. , Blackman D. K. , Früh-Green G. L. , Butterfield D. A. , Lilley M. D. , Olson E. J. , Schrenk M. O. , Roe K. K. , Lebon G. T. , Rivizzigno P. ( 2001 ). An off-axis hydrothermal vent field near the Mid-Atlantic Ridge at 30°N . Nature 412 , 145 – 149 . Google Scholar CrossRef Search ADS PubMed Klein F. , Bach W. , Jöns N. , McCollom T. , Moskowitz B. , Berquó T. ( 2009 ). Iron partitioning and hydrogen generation during serpentinization of abyssal peridotites from 15°N on the Mid-Atlantic Ridge . Geochimica et Cosmochimica Acta 73 , 6868 – 6893 . Google Scholar CrossRef Search ADS Kleppe A. K. , Jephcoat A. P. , Welch M. D. ( 2003 ). The effect of pressure upon hydrogen bonding in chlorite: a Raman spectroscopic study of clinochlore to 26.5 GPa . American Mineralogist 88 , 567 – 573 . Google Scholar CrossRef Search ADS Kühnel R. A. , Roorda H. J. , Steensma J. J. ( 1975 ). The crystallinity of minerals—a new variable in pedogenetic processes: a study of goethite and associated silicates in laterites . Clays and Clay Minerals 23 , 349 – 354 . Google Scholar CrossRef Search ADS Kuleci H. , Ulven O. I. , Rybacki E. , Wunder B. , Abart R. ( 2017 ). Reaction-induced fracturing in a hot pressed calcite-periclase aggregate . Journal of Structural Geology 94 , 116 – 135 . Google Scholar CrossRef Search ADS Lasaga A. C. ( 1998 ). Kinetic Theory in the Earth Sciences . Princeton, Chichester : Princeton University Press . Google Scholar CrossRef Search ADS Malvoisin B. ( 2015 ). Mass transfer in the oceanic lithosphere: serpentinization is not isochemical . Earth and Planetary Science Letters 430 , 75 – 85 . Google Scholar CrossRef Search ADS Malvoisin B. , Brunet F. , Carlut J. , Rouméjon S. , Cannat M. ( 2012 ). Serpentinization of oceanic peridotites: 2. Kinetics and processes of San Carlos olivine hydrothermal alteration . Journal of Geophysical Research: Solid Earth 117 , B04102 . McCollom T. M. , Klein F. , Robbins M. , Moskowitz B. , Berquó T. S. , Jöns N. , Bach W. , Templeton A. ( 2016 ). Temperature trends for reaction rates, hydrogen generation, and partitioning of iron during experimental serpentinization of olivine . Geochimica et Cosmochimica Acta 181 , 175 – 200 . Google Scholar CrossRef Search ADS Mellini M. , Rumori C. , Viti C. ( 2005 ). Hydrothermally reset magmatic spinels in retrograde serpentinites: formation of “ferritchromit” rims and chlorite aureoles . Contributions to Mineralogy and Petrology 149 , 266 – 275 . Google Scholar CrossRef Search ADS Morishita T. , Tani K. , Shukuno H. , Harigane Y. , Tamura A. , Kumagai H. , Hellebrand E. ( 2011 ). Diversity of melt conduits in the Izu–Bonin–Mariana forearc mantle: implications for the earliest stage of arc magmatism . Geology 39 , 411 – 414 . Google Scholar CrossRef Search ADS Nakatani T. , Nakamura M. ( 2016 ). Experimental constraints on the serpentinization rate of fore-arc peridotites: implications for the upwelling condition of the slab-derived fluid . Geochemistry, Geophysics, Geosystems 17 , 3393 – 3419 . Google Scholar CrossRef Search ADS Nishiyama T. ( 1998 ). Uphill diffusion and a new nonlinear diffusion equation in ternary non-electrolyte system . Physics of the Earth and Planetary Interiors 107 , 33 – 51 . Google Scholar CrossRef Search ADS Nozaka T. , Wintsch R. P. , Meyer R. ( 2017 ). Serpentinization of olivine in troctolites and olivine gabbros from the Hess Deep Rift . Lithos 282–283 , 201 – 214 . Google Scholar CrossRef Search ADS Ogasawara Y. , Okamoto A. , Hirano N. , Tsuchiya N. ( 2013 ). Coupled reactions and silica diffusion during serpentinization . Geochimica et Cosmochimica Acta 119 , 212 – 230 . Google Scholar CrossRef Search ADS O’Hanley D. S. ( 1996 ). Serpentinites. Records of Tectonic and Petrological History. Oxford Monographs of Geology and Geophysics . New York, Oxford : Oxford University Press . Okamoto A. , Shimizu H. ( 2015 ). Contrasting fracture patterns induced by volume-increasing and -decreasing reactions: implications for the progress of metamorphic reactions . Earth and Planetary Science Letters 417 , 9 – 18 . Google Scholar CrossRef Search ADS Okamoto A. , Ogasawara Y. , Ogawa Y. , Tsuchiya N. ( 2011 ). Progress of hydration reactions in olivine–H2O and orthopyroxene–H2O systems at 250°C and vapor-saturated pressure . Chemical Geology 289 , 245 – 255 . Google Scholar CrossRef Search ADS Oyanagi R. , Okamoto A. , Hirano N. , Tsuchiya N. ( 2015 ). Competitive hydration and dehydration at olivine–quartz boundary revealed by hydrothermal experiments: implications for silica metasomatism at the crust–mantle boundary . Earth and Planetary Science Letters 425 , 44 – 54 . Google Scholar CrossRef Search ADS Pens M. , Andreani M. , Daniel I. , Perrillat J. P. , Cardon H. ( 2016 ). Contrasted effect of aluminum on the serpentinization rate of olivine and orthopyroxene under hydrothermal conditions . Chemical Geology 441 , 256 – 264 . Google Scholar CrossRef Search ADS Philpotts A. , Auge J. ( 2009 ). Principles of Igneous and Metamorphic Petrology . Cambridge : Cambridge University Press . Google Scholar CrossRef Search ADS Plümper O. , Piazolo S. , Austrheim H. ( 2012a ). Olivine pseudomorphs after serpentinized orthopyroxene record transient oceanic lithospheric mantle dehydration (Leka ophiolite complex, Norway) . Journal of Petrology 53 , 1943 – 1968 . Google Scholar CrossRef Search ADS Plümper O. , Royne A. , Magraso A. , Jamtveit B. ( 2012b ). The interface-scale mechanism of reaction-induced fracturing during serpentinization . Geology 40 , 1103 – 1106 . Google Scholar CrossRef Search ADS Plümper O. , John T. , Podladchikov Y. Y. , Vrijmoed J. C. , Scambelluri M. ( 2017 ). Fluid escape from subduction zones controlled by channel-forming reactive porosity . Nature Geoscience 10 , 150 – 156 . Google Scholar CrossRef Search ADS Powell R. , Holland T. ( 1993 ). On the formulation of simple mixing models for complex phases . American Mineralogist 78 , 1174 – 1180 . Putnis A. ( 2009 ). Mineral replacement reactions . Reviews in Mineralogy and Geochemistry 70 , 87 – 124 . Google Scholar CrossRef Search ADS Putnis A. , Austrheim H. ( 2010 ). Fluid-induced processes: metasomatism and metamorphism . Geofluids 10 , 254 – 269 . Qian G. , Brugger J. , Skinner W. M. , Chen G. , Pring A. ( 2010 ). An experimental study of the mechanism of the replacement of magnetite by pyrite up to 300°C . Geochimica et Cosmochimica Acta 74 , 5610 – 5630 . Google Scholar CrossRef Search ADS Rouméjon S. , Cannat M. ( 2014 ). Serpentinization of mantle-derived peridotites at mid-ocean ridges: mesh texture development in the context of tectonic exhumation . Geochemistry, Geophysics, Geosystems 15 , 2354 – 2379 . Google Scholar CrossRef Search ADS Salvi S. , Pokrovski G. S. , Schott J. ( 1998 ). Experimental investigation of aluminum–silica aqueous complexing at 300°C . Chemical Geology 151 , 51 – 67 . Google Scholar CrossRef Search ADS Sato H. , Taniguchi H. , Takahashi N. , Mohiuddin M. M. , Hirano N. , Ogawa Y. ( 1999 ). The origin of the Mineoka ophiolite . Journal of Geography 108 , 203 – 189 . (Japanese with English abstract). Google Scholar CrossRef Search ADS Schwarzenbach E. M. , Caddick M. J. , Beard J. S. , Bodnar R. J. ( 2016 ). Serpentinization, element transfer, and the progressive development of zoning in veins: evidence from a partially serpentinized harzburgite . Contributions to Mineralogy and Petrology 171 , 1079 . Google Scholar CrossRef Search ADS Shimizu H. , Okamoto A. ( 2016 ). The roles of fluid transport and surface reaction in reaction-induced fracturing, with implications for the development of mesh textures in serpentinites . Contributions to Mineralogy and Petrology 171 , 73 . Google Scholar CrossRef Search ADS Sleep N. H. , Meibom A. , Fridriksson T. , Coleman R. G. , Bird D. K. ( 2004 ). H2-rich fluids from serpentinization: geochemical and biotic implications . Proceedings of the National Academy of Sciences of the United States of America 101 , 12818 – 12823 . Google Scholar CrossRef Search ADS PubMed Stefánsson A. ( 2001 ). Dissolution of primary minerals of basalt in natural waters: I. Calculation of mineral solubilities from 0°C to 350°C . Chemical Geology 172 , 225 – 250 . Google Scholar CrossRef Search ADS Tagirov B. , Schott J. ( 2001 ). Aluminum speciation in crustal fluids revisited . Geochimica et Cosmochimica Acta 65 , 3965 – 3992 . Google Scholar CrossRef Search ADS Viti C. , Mellini M. ( 1998 ). Mesh textures and bastites in the Elba retrograde serpentinites . European Journal of Mineralogy 10 , 1341 – 1360 . Google Scholar CrossRef Search ADS Wada I. , Wang K. , He J. , Hyndman R. D. ( 2008 ). Weakening of the subduction interface and its effects on surface heat flow, slab dehydration, and mantle wedge serpentinization . Journal of Geophysical Research 113 , B04402 . Google Scholar CrossRef Search ADS Wegner W. W. , Ernst W. G. ( 1983 ). Experimentally determined hydration and dehydration reaction rates in the system MgO–SiO2–H2O . American Journal of Science 283A , 151 – 180 . Xia F. , Brugger J. , Chen G. , Ngothai Y. , O’Neill B. , Putnis A. , Pring A. ( 2009 ). Mechanism and kinetics of pseudomorphic mineral replacement reactions: a case study of the replacement of pentlandite by violarite . Geochimica et Cosmochimica Acta 73 , 1945 – 1969 . Google Scholar CrossRef Search ADS Zhao J. , Brugger J. , Chen G. , Ngothai Y. , Pring A. ( 2014 ). Experimental study of the formation of chalcopyrite and bornite via the sulfidation of hematite: mineral replacements with a large volume increase . American Mineralogist 99 , 343 – 354 . Google Scholar CrossRef Search ADS Zimmer K. , Zhang Y. , Lu P. , Chen Y. , Zhang G. , Dalkilic M. , Zhu C. ( 2016 ). SUPCRTBL: a revised and extended thermodynamic dataset and software package of SUPCRT92 . Computers & Geosciences 90 , 97 – 111 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Petrology Oxford University Press

Al-Zoning of Serpentine Aggregates in Mesh Texture Induced by Metasomatic Replacement Reactions

Loading next page...
 
/lp/ou_press/al-zoning-of-serpentine-aggregates-in-mesh-texture-induced-by-0ZhXK1c6Yn
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
0022-3530
eISSN
1460-2415
D.O.I.
10.1093/petrology/egy039
Publisher site
See Article on Publisher Site

Abstract

Abstract Serpentinization of oceanic lithosphere commonly proceeds with the development of mesh texture. Examination of a serpentinized harzburgite and a plagioclase-bearing wehrlite revealed conspicuous zoning of Al in a serpentine mesh texture, with Al-rich cores and Al-poor rims, as well as Al-rich veins, indicating local transport of Al from plagioclase and pyroxene during serpentinization. To reveal the influences of Al on the reaction mechanisms and textural development during serpentinization, we conducted hydrothermal experiments in the olivine (Ol)–plagioclase (Pl)–H2O system and analyzed the variations in mineralogy and microtexture of olivine replacements as a function of distance from the Ol–Pl boundary. The Al–Si metasomatic zone, where Al-serpentine and a minor amount of Ca-saponite were formed, was developed in the Ol-hosted region close to the Ol–Pl boundary. Far from the Ol–Pl boundary, Al-free serpentine, brucite, and magnetite were formed, indicating the progress of an ‘isochemical’ reaction (separate from water). The aggregates of Al-serpentine after olivine in the metasomatic zone showed a characteristic zoning of Al. Microtextural evidence indicates that the zoning was produced in response to the migration of an Al metasomatic front that involved an early-stage of serpentinization with an Al-free solution, and the subsequent pseudomorphic replacement of olivine and simultaneous development of overgrowths as the amount of Al increased in the solution. The Al-bearing aqueous solution caused the formation of olivine pseudomorphs and this contrasts with the lack of preservation of original olivine outlines in the isochemical zones. Comparisons of zoning in natural and experimentally produced mesh textures suggest that Al-poor rims in the mesh texture form at the start of the serpentinization process, followed by the coupled formation of Al-rich mesh cores and Al-rich veins. Our experimental results indicate that Al-zoning in the mesh texture represent the transition from a closed to an open system during serpentinization. INTRODUCTION Serpentinization causes significant changes in the nature of the oceanic lithosphere, including its rheological properties (Escartín et al., 2003), thermal flux (Iyer et al., 2010), hydrogen production (Sleep et al., 2004), and microbial activity on the seafloor (Kelley et al., 2001). Serpentinized peridotite in the oceanic lithosphere can store up to 13 wt % chemically bonded water in a solid state and it may exert the primary control on fluid flux (Hyndman & Peacock, 2003; Nakatani & Nakamura, 2016; Plümper et al., 2017) and seismic activity in subduction zones (Wada et al., 2008; Boudier et al., 2010). Despite the influence of serpentinized peridotite on various processes, the in situ rate of serpentinization within oceanic lithosphere is still poorly constrained, since it is a complicated process involving volume change, deformation, and mass transfer, as well as the reaction itself. A mesh texture is the typical texture found in serpentinized peridotite from the oceanic lithosphere. The texture comprises polygonal domains divided by grain-scale fractures induced by tectonic, thermal, and/or reaction-induced stresses (O’Hanley, 1996; Plümper et al., 2012b; Rouméjon & Cannat, 2014; Okamoto & Shimizu, 2015; Shimizu & Okamoto, 2016). Mesh texture commonly shows conspicuous zoning that can be divided into core and rim. The core and rim of the mesh have different mineral assemblages and/or mineral compositions, and based on such textures, a two-stage model for serpentinization has been proposed (e.g. Viti & Mellini, 1998; Bach et al., 2006; Beard et al., 2009). In this model, the mesh rim forms first under near-isochemical conditions, and then subsequently the mesh core forms as external fluids infiltrate the system. The mesh zoning, therefore, represents temporal and spatial changes in the chemical environment during serpentinization. Some serpentines in the mafic and ultramafic rocks of the oceanic lithosphere contain large amounts of Al (up to 27 wt %) (O’Hanley, 1996; Viti & Mellini, 1998; Mellini et al., 2005; Beard et al., 2009; Andreani et al., 2013; Schwarzenbach et al., 2016; Nozaka et al., 2017). Based on hydrothermal experiments, Andreani et al. (2013) showed that the presence of aluminum accelerates the hydration of olivine, although the experimental product was chlorite rather than serpentine. Al is not a major component of olivine (<0·1 wt %) so potential sources of the Al are, therefore, pyroxenes, spinel and plagioclase in mafic and ultramafic rocks. Accordingly, the distribution of Al in partly serpentinized peridotite could be a useful tracer for the influx of external fluids and the transfer of elements, either on the scale of millimetres in the case of chlorite forming after olivine along the contacts between olivine and plagioclase (Frost et al., 2008), or on the scale of metres in the case that the assemblage tremolite + chlorite + serpentine is formed via Al–Si–Ca metasomatism of peridotite (Boschi et al., 2006). However, the distribution of Al in the mesh texture and the role of Al during serpentinization have not been studied systematically. In this study, we describe first the characteristics of Al zoning in the mesh texture within partially serpentinized mafic and ultramafic rocks of the oceanic lithosphere. Then, to understand the effects of Al transport on the mechanisms and textural development of serpentinization, we describe the results of hydrothermal experiments in the olivine (Ol)–plagioclase (Pl)–H2O system. Combining experimental data on the distribution of Al in the product mineral and observed Al distribution in two natural rocks, we discuss the implications for textual development and mass transfer during serpentinization within the oceanic lithosphere. METHODS Hydrothermal experiments We conducted hydrothermal experiments in the Ol–Pl–H2O system using a tube-in-tube vessel at 230°C and at a vapor-saturated pressure of 2·80 MPa (Fig. 1a, b). The experimental procedures were almost the same as those used in the Ol–quartz–H2O experiments of Oyanagi et al. (2015). Natural olivine (XMg (Mg/(Mg + Fetotal)) = 0·91) and plagioclase grains (Ca/(Ca + Na + K) = 0·97) were crushed and sieved to 25–53 μm size. The powders were then washed repeatedly with deionized water in an ultrasonic bath. In the main vessel, which was made of Inconel alloy (44 cm3), four inner tubes were set with 20 ml distilled water (Fig. 1a). In each inner tube, 40 mg of plagioclase powder was placed at the bottom, and 110 mg of powdered olivine was then placed on the plagioclase layer (Fig. 1b). The base of each inner tube was sealed and the tops left open. The thicknesses of the plagioclase and olivine layers were ∼10 and ∼30 mm, respectively. The water–rock mass ratio was ∼33. The vessels were then placed in ovens at 230°C. The temperature difference between the inside and the outside of the vessel was <1°C. The run times were 783, 2055, 3184, and 7980 h. Fig. 1. View largeDownload slide (a) Schematic illustration of the tubes placed in the reaction vessel. Four inner tubes were immersed in liquid. (b) Configuration of mineral powders within the inner tubes for the Ol–Pl–H2O experiments. Fig. 1. View largeDownload slide (a) Schematic illustration of the tubes placed in the reaction vessel. Four inner tubes were immersed in liquid. (b) Configuration of mineral powders within the inner tubes for the Ol–Pl–H2O experiments. For comparative purposes, Pl–H2O experiments were conducted with 1·0 g of the same plagioclase (<25 μm) and 10 ml of distilled water under the same P–T conditions as used for the Ol–Pl–H2O experiments. The run times for the Pl–H2O experiments were 158, 330, and 488 h. To investigate the effects of the Inconel alloy vessel on the solution chemistry, we carried out blank experiments (with distilled water only in the vessel) under the same experimental P–T conditions for 2224 h. Analyses of the experimental products After each individual run, the reaction vessel was cooled to room temperature within 1 h, and the solutions and four inner tubes, which contain solid materials (products + unreacted reactants), were taken from the vessel. Solution pH was immediately measured at room temperature of ∼25°C. After diluting the solutions with nitric acid, the concentrations of Si, Mg, Ca, Fe, and Al were measured by inductively coupled plasma–atomic emission spectrometry (ICP–AES; Thermo Scientific iCAP 6000) at Tohoku University, Japan. We calculated the fluid speciation of the analyzed fluids, the in situ pH, and the phase diagram in the CaO–Al2O3–MgO–SiO2–H2O system for the experimental P–T conditions using Geochemist’s Workbench (Bethke, 2007). Thermodynamic data in the default database thermo. tdat were customized on the basis of a log K value obtained from the SUPCRT92 database (Johnson et al., 1992) with revisions suggested by Zimmer et al. (2016). The primary thermodynamic data were taken mainly from Holland & Powell (2011) for minerals, from Tagirov & Schott (2001) for Al-bearing species in solution, and from Stefánsson (2001) for aqueous silica. After each individual run, the solid materials (products + unreacted reactants) from the four inner tubes were dried for >24 hours at 90°C. One of the inner tubes was encased in epoxy resin, and a polished thin section was made in the direction parallel to the long axis of the tube. The chemical compositions of the minerals in the polished thin section were analyzed using an electron probe micro-analyzer (EPMA; JEOL 8200) at Tohoku University with an accelerating voltage of 15 kV, beam current of 12 nA, and a focused 1–2 µm beam diameter. Standards used for calibrations were wollastonite (Si, Ca), rutile (Ti), corundum (Al), hematite (Fe), MnO, periclase (Mg), albite (Na), sanidine (K), and Cr2O3. JEOL software was used for the ZAF corrections. The minerals present in the polished thin section were observed with a field-emission scanning electron microscope (FESEM; JEOL JSM-7001F) at Tohoku University. The modal abundances of minerals and the porosities were obtained for selected thin sections by analyses of backscattered electron (BSE) images (102 × 385 µm in size). The resolution of the images used in the analyses were ∼1 μm2/pixel. Serpentine minerals were identified using a micro-Raman spectroscope (HORIBA XploRA PLUS) at Tohoku University equipped with a 532 nm laser and 2400 grooves/mm grating. Another inner tube was used for thermogravimetry (TG; Rigaku Thermo Plus EVOII TG8120) at Tohoku University to analyze the H2O contents of the solid samples. For the TG analyses, an inner tube was cut into eight segments of ∼5 mm length. Segments one and two were from the Pl-hosted region, and segments three to eight from the Ol-hosted region. During the TG analyses, temperatures were raised from room temperature to 1100°C at a rate of 10°C per minute. Loss on ignition (LOI) values for each segment were obtained in the temperature range 120–1100°C to exclude the weight loss associated with molecular water in the solid samples. The H2O contents of serpentine and brucite were determined separately from the weight losses in the temperature ranges 450–800°C and 320–450°C, respectively. The pseudosection in the FeO–MgO–SiO2–H2O–O2 system was calculated using Perple X 6·7·9 (Connolly, 2009) with the hp11ver.dat dataset (Holland & Powell, 2011), which includes thermodynamic data for lizardite, greenalite, brucite, hematite, magnetite, forsterite, fayalite, and talc. The thermodynamic data for ferric serpentine (Mg2 Fe23+SiO5(OH)4) and Fe-brucite were constructed following Evans et al. (2013). Brucite, and talc were modeled as ideal mixtures of Mg- and Fe-endmembers, while olivine activity–composition relationships were described by a symmetric formalism model (Powell & Holland, 1993) with a Wfo–fa value of 8 kJ/mol (Evans et al., 2013). Serpentines were modeled as ideal mixtures of the lizardite, greenalite and ferric serpentine. The activity of iron was assumed to be 0·25 as an analogue for an alloy phase such as awaruite (Evans et al., 2013). Analyses of natural serpentine mesh texture We analysed two rocks with a well-developed mesh texture: a serpentinized plagioclase-bearing wehrlite from the Izu–Ogasawara fore-arc and a partly serpentinized harzburgite from the Mineoka ophiolite complex in the Boso Peninsula, central Japan. We prepared polished thin sections and determined the chemical compositions of the minerals and the nature of the mineral zoning using EPMA (details of the analytical procedures are given above). Representative mineral compositions of the two rock specimens are summarized in Supplementary Data Table S1; supplementary data are available for downloading at http://www.petrology.oxfordjournals.org. ALUMINUM ZONING IN SERPENTINE MESH TEXTURE Plagioclase-bearing wehrlite from the Izu–Ogasawara forearc region The serpentinized plagioclase-bearing wehrlite sample (KH07–02-D31–101) was collected from the landward slope of the Izu–Ogasawara trench during a dredging cruise (KH07–02 leg 4) of the R/V Hakuho–Maru operated by the Japan Agency for Marine-Earth Science and Technology (Harigane et al., 2013). In previous studies (Morishita et al., 2011; Harigane et al., 2013) it was reported that peridotite, troctolite, pyroxenite, gabbro, dolerite and basalt were collected from 5293 to 5792 metres below sea level during dredging and remotely operated vehicle operations. Some of the doleritic samples exhibited a fore-arc tholeiitic basalt signature (cf. Ishizuka et al., 2011). The variations in the rock samples suggest that this trench wall represents the structure of an immature fore-arc mantle at the time of initiation of the subduction zone (Morishita et al., 2011; Harigane et al., 2013). The plagioclase-bearing wehrlite sample was covered by layers of brown weathered material about 10 mm thick (Supplementary Data Fig. S1). We made a polished thin section of the unweathered core of the sample. The serpentinized plagioclase-bearing wehrlite originally consisted of ∼84 vol % olivine, ∼11 vol % clinopyroxene, and ∼5 vol % plagioclase (Fig. 2a). More than 70 vol % of the olivine is serpentinized (Fig. 2b), with lizardite and magnetite being the main alteration products. Brucite and magnetite were not found at olivine–plagioclase contacts, but minor intergrowths of brucite (<5%) within serpentine and subhedral magnetite grains were found further away (>1·0 mm) from the contacts. The clinopyroxene is almost unaltered, whereas the plagioclase is intensely altered, mainly to chlorite and hydrogrossular, and partly to pumpellyite (Fig. 2a). Fig. 2. View largeDownload slide Mesh texture and distribution of Al in serpentinized plagioclase-bearing wehrlite from the Izu–Ogasawara trench, collected during the KH07–2 cruise. (a) Photomicrograph of the serpentinized plagioclase-bearing wehrlite. Olivine is highly serpentinized to form a mesh texture, and the plagioclase is almost completely altered to hydrogrossular and chlorite, along with some pumpellyite, while the clinopyroxene remains almost fresh. (b) Modal abundances of olivine in the area indicated by the rectangle in Fig. 2a. (c)–(e) Elemental maps of (c) Al, (d) Si, and (e) Ca, with a profile of relative integrated intensity (red line) obtained for each element in serpentine along the long axis of the area indicated by the rectangle in Fig. 2a. Olivine grains were removed from analysis by blackening BSE images. (f) Back-scattered image of serpentine mesh texture. (g) Elemental map of Al in the mesh texture shown in Fig. 2f. C, mesh core; R, mesh rim; V, mesh vein network. (h) Al content in serpentine (cations per seven oxygen formula unit) along line a–b in Fig. 2g. (i) Plot of Mg + Fe cations vs Si (per seven oxygen formula unit) shows the Tschermak substitution trend in the Al-rich mesh core and vein serpentine. The Al-poor mesh rim is characterized by R3+ cations in octahedral sites. Pl, plagioclase; Grs, hydrogrossular; Pmp, pumpellyite; Chl, chlorite; Cpx, clinopyroxene; Ol, olivine. Fig. 2. View largeDownload slide Mesh texture and distribution of Al in serpentinized plagioclase-bearing wehrlite from the Izu–Ogasawara trench, collected during the KH07–2 cruise. (a) Photomicrograph of the serpentinized plagioclase-bearing wehrlite. Olivine is highly serpentinized to form a mesh texture, and the plagioclase is almost completely altered to hydrogrossular and chlorite, along with some pumpellyite, while the clinopyroxene remains almost fresh. (b) Modal abundances of olivine in the area indicated by the rectangle in Fig. 2a. (c)–(e) Elemental maps of (c) Al, (d) Si, and (e) Ca, with a profile of relative integrated intensity (red line) obtained for each element in serpentine along the long axis of the area indicated by the rectangle in Fig. 2a. Olivine grains were removed from analysis by blackening BSE images. (f) Back-scattered image of serpentine mesh texture. (g) Elemental map of Al in the mesh texture shown in Fig. 2f. C, mesh core; R, mesh rim; V, mesh vein network. (h) Al content in serpentine (cations per seven oxygen formula unit) along line a–b in Fig. 2g. (i) Plot of Mg + Fe cations vs Si (per seven oxygen formula unit) shows the Tschermak substitution trend in the Al-rich mesh core and vein serpentine. The Al-poor mesh rim is characterized by R3+ cations in octahedral sites. Pl, plagioclase; Grs, hydrogrossular; Pmp, pumpellyite; Chl, chlorite; Cpx, clinopyroxene; Ol, olivine. The bulk Al content in the altered olivine region is high around the plagioclase grains; the Al contents are highest at contacts with plagioclase and decrease linearly with increasing distance from the plagioclase. The zone of Al metasomatism is about 0·9 mm wide (Fig. 2c) and in this zone the olivine is almost completely consumed (Fig. 2b). Up to 0·1 mm from the contacts of Ol and Pl, small amounts of tremolite and diopside were formed, as indicated by local increases in Si (Fig. 2d) and Ca (Fig. 2e). In most parts of the Al metasomatic zones, the Si content of the bulk-rock is lower than in regions outside the Al metasomatic zones (>1·0 mm; Fig. 2d). Most of the serpentinized olivine grains show a clear mesh texture, composed of polygonal units separated by veins (V in Fig. 2g). Each polygonal unit shows zoning, with a mesh core (C in Fig. 2g) and a rim (R in Fig. 2g) evident in BSE images (Fig. 2f). In the Al metasomatic zone, this mesh zoning corresponds to changes in the Al content of the serpentine aggregates (Fig. 2g), with Al-rich cores (Al = 0·6–0·8 per formula unit (pfu)), Al-poor rims (∼0·05 pfu), and Al-rich veins (0·5–0·8 pfu) (Fig. 2h). The composition of the mesh cores and veins are similar, and the variations can be explained by Tschermaks substitution (AlAl(Mg, Fe2+)-1Si-1) (Fig. 2i;Beard & Frost, 2016). The Al-poor composition in the mesh rims suggests minor amounts of R23+Si2O5(OH)4 component (R3+ is Al3+ and/or Fe3+; Fig. 2i). Harzburgite from the Mineoka ophiolite complex The Mineoka ophiolite complex, central Japan, is as an accreted fragment of ophiolite that formed at a mid-ocean ridge (Sato et al., 1999). The serpentinized harzburgite originally contained olivine (∼81 vol %), orthopyroxene (∼14 vol %), and clinopyroxene (∼4 vol %), along with minor spinel (∼1 vol %). The olivine grains are heavily serpentinized and serpentine (bastite) has replaced orthopyroxene. Raman spectra show the serpentine minerals to be lizardite and/or chrysotile (Katayama et al., 2010), and serpentine + magnetite + brucite were formed within the olivine regions (Hirauchi et al., 2010; Shimizu & Okamoto, 2016). Zoning of Al in the olivine mesh texture is displayed around grains of clinopyroxene (Fig. 3a). Clinopyroxenes have been replaced by serpentine without the formation of other Ca-bearing minerals, indicating removal of Ca from the system. The elemental map of Al indicates that the average Al content in the olivine region decreases with increasing distance from boundaries with clinopyroxene (Fig. 3a). Magnetite was commonly observed along olivine and mesh core boundaries, and in serpentine veins (Fig. 3b). Fig. 3. View largeDownload slide Serpentine mesh texture in harzburgite from the Mineoka ophiolite complex, Japan. (a) Al elemental map around the boundary between clinopyroxene and olivine· The white bold line and the dotted lines show a pseudomorph of clinopyroxene and the outlines of olivine relics, respectively. (b) Back-scattered image of mesh texture shows the distribution of magnetite in the mesh core and veins. (c) Elemental map of Al in the mesh texture shown in Fig. 3b. The Al-rich mesh core (C), Al-rich mesh vein network (V), and Al-poor mesh rim (R) are identified. (d) Al contents (cations per seven oxygen formula unit) along traverse a–b in Fig. 3c. (e) Plot of Si versus Mg + Fe cations (per seven oxygen formula unit) showing the brucite mixing trend in the mesh core and mesh rim. Srp, serpentine; Brc, brucite; Mag, magnetite. Other abbreviations are as in Fig. 2. Fig. 3. View largeDownload slide Serpentine mesh texture in harzburgite from the Mineoka ophiolite complex, Japan. (a) Al elemental map around the boundary between clinopyroxene and olivine· The white bold line and the dotted lines show a pseudomorph of clinopyroxene and the outlines of olivine relics, respectively. (b) Back-scattered image of mesh texture shows the distribution of magnetite in the mesh core and veins. (c) Elemental map of Al in the mesh texture shown in Fig. 3b. The Al-rich mesh core (C), Al-rich mesh vein network (V), and Al-poor mesh rim (R) are identified. (d) Al contents (cations per seven oxygen formula unit) along traverse a–b in Fig. 3c. (e) Plot of Si versus Mg + Fe cations (per seven oxygen formula unit) showing the brucite mixing trend in the mesh core and mesh rim. Srp, serpentine; Brc, brucite; Mag, magnetite. Other abbreviations are as in Fig. 2. A single mesh unit (Fig. 3c), which is separated from other units by Al-rich serpentine veins, shows clear zoning of Al with an Al-rich mesh core and an Al-free mesh rim. The Al content is high in the mesh core (C in Fig. 3d), gradually decreases towards the mesh rim (R in Fig. 3d), and increases in the veins (V in Fig. 3d). We note that even in the core and the veins, the Al content of serpentine in the harzburgite (Fig. 3d; Al = 0·03–0·04 pfu) is an order of magnitude lower than that of the serpentine in the plagioclase-bearing wehrlite (Fig. 2h; Al = 0·6–0·8 pfu). The chemical compositions of the Al-rich veins in the harzburgite probably indicate Tschermak substitution or a R23+Si2O5(OH)4 substitution trend (Fig. 3e). The SiO2 and Mg + ΣFe contents of mesh cores and rims lie on the brucite–serpentine mixing trend (Fig. 3e), suggesting the presence of brucite intergrowths within the serpentine aggregates of the mesh cores and rims. RESULTS OF HYDROTHERMAL EXPERIMENTS Mineral zones that formed during the experiments Figure 4a shows a photomicrograph of the products of the Ol–Pl–H2O experiment for a run time of 7980 h. Figure 4b is an enlargement of part of Fig. 4a. Hereafter, positions in the inner tube shown in this photograph are described relative to the Ol–Pl boundary (x = 0), so that the Pl-hosted region occurs from x = −10 to 0 mm and the Ol-hosted region from x = 0 to ∼30 mm. Fig. 4. View largeDownload slide (a) Photomicrograph of the experimental products for a run of 7980 h in the Ol–Pl–H2O experiments. The x values indicate the distance from the Ol–Pl boundary. (b) Enlarged photo around the Ol–Pl boundary of the area indicated by the rectangle in Fig. 4a. (c)–(d) Back-scattered electron images of reaction products and unaltered reactant minerals after 7980 h. The position where the BSE images were taken was shown in Fig. 4b. (c) and (d) Overviews in the plagioclase-hosted region at (c) x = −5 mm and (d) x = −1 mm. (e) Contact between Pl- and Ol-hosted regions. No reaction products were formed after the 7980 h reaction in the plagioclase-hosted region. (f) and (g) Overviews of the reaction products of (f) the Al-serpentine zone and (g) the serpentine + brucite + magnetite zone. Pl, plagioclase; Ol, olivine; Srp, serpentine; Al-Srp, Al-rich serpentine; Brc, brucite; Mag, magnetite. Fig. 4. View largeDownload slide (a) Photomicrograph of the experimental products for a run of 7980 h in the Ol–Pl–H2O experiments. The x values indicate the distance from the Ol–Pl boundary. (b) Enlarged photo around the Ol–Pl boundary of the area indicated by the rectangle in Fig. 4a. (c)–(d) Back-scattered electron images of reaction products and unaltered reactant minerals after 7980 h. The position where the BSE images were taken was shown in Fig. 4b. (c) and (d) Overviews in the plagioclase-hosted region at (c) x = −5 mm and (d) x = −1 mm. (e) Contact between Pl- and Ol-hosted regions. No reaction products were formed after the 7980 h reaction in the plagioclase-hosted region. (f) and (g) Overviews of the reaction products of (f) the Al-serpentine zone and (g) the serpentine + brucite + magnetite zone. Pl, plagioclase; Ol, olivine; Srp, serpentine; Al-Srp, Al-rich serpentine; Brc, brucite; Mag, magnetite. In the Pl-hosted region, no products were observed after the experiments. The color in the Pl-hosted region changed from brown at the Ol–Pl boundary to beige away from the boundary (Fig. 4a, b) and this color change was probably due to the the larger amount of epoxy resin-filled porosity and smaller grain size of Pl near the Ol–Pl boundary (x = ∼−1 mm; Fig. 4b, d) than in areas farther from the boundary (x = ∼−5 mm; Fig. 4b, c). In the Pl-hosted region, no secondary minerals (e.g. diopside, prehnite, epidote, and grossular) were observed. The color of the Ol-hosted region changes from pale blue at the Ol–Pl boundary to pale yellow in the Al metasomatic zone, and dirty green away from the boundary (Fig. 4a, b). The reaction products in the Ol-hosted region change systematically with distance from the Ol–Pl boundary (Fig. 4a, b) from an Al-serpentine + minor Ca-saponite zone (herein called the Al–Si metasomatic zone), to a serpentine ± magnetite zone (herein called the Si metasomatic zone), and finally to a serpentine + magnetite + brucite zone (herein called the isochemical zone). In the Al–Si metasomatic zone (x = 0–1·5 mm in Fig. 4b), Ol grains are highly reacted, and they show irregular outlines surrounded by product minerals (Fig. 4f). The chemical composition of the product aggregate is Al-rich with small amounts of Ca (< ∼1 wt % of CaO; Table 1). Raman spectra of the products show peaks at 227, 380, 684, 3685, and 3700 (trace) cm-1 (Supplementary Data Fig. S2), indicating that the Al-hydrous phase is lizardite (Enami, 2006). In addition, a peak at 191 cm-1 indicates the presence of talc or saponite in the Al-hydrous phase (Supplementary Data Fig. S2), whereas the peaks for chlorite (cf. the main peaks at 360, 552, 676, and 3675 cm-1 for clinochlore; Kleppe et al., 2003) are lacking. These observations suggest that the Al–Si metasomatic zone is composed mainly of Al-rich lizardite with lesser amounts of Ca-saponite (hereafter called Al-serpentine; Fig. 4e, f). The occurrences of Al-serpentine are described in more detail in a later section. Table 1 Representative chemical compositions of reactant and product minerals in the experiments Plagioclase Olivine Lizardite Brucite Magnetite Al-serpentine (Al > 0·1 pfu) domain. I domain. II domain. III Wt % SiO2 44·36 40·94 42·54 0·15 4·13 36·01 38·55 35·62 TiO2 0·01 0·00 0·00 0·00 0·00 0·03 0·02 0·00 Al2O3 35·70 0·02 0·08 0·12 0·00 10·64 1·73 13·65 Cr2O3 0·00 0·00 0·01 0·00 0·00 0·00 0·01 0·00 FeO 0·43 8·49 2·18 11·67 84·54 7·67 6·09 4·13 MgO 0·09 50·65 41·67 60·72 1·51 32·91 33·78 31·74 MnO 0·09 0·15 0·03 0·25 0·24 0·09 0·10 0·06 CaO 19·58 0·08 0·03 0·00 0·00 0·52 0·32 0·40 Na2O 0·33 0·03 0·02 0·07 0·05 0·00 0·19 0·00 K2O 0·00 0·00 0·00 0·04 0·00 0·00 0·00 0·00 Total 100·52 100·36 86·56 73·02 90·49 87·87 80·79 85·59 Oxygen 8·00 4·00 7·00 7·00 7·00 7·00 7·00 7·00 Si 2·04 0·99 2·00 0·01 0·36 1·72 1·98 1·70 Ti 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Al 1·94 0·00 0·00 0·01 0·00 0·60 0·11 0·77 Cr 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Fe 0·02 0·17 0·09 0·68 6·07 0·31 0·26 0·17 Mg 0·01 1·83 2·91 6·27 0·19 2·34 2·59 2·26 Mn 0·00 0·00 0·00 0·01 0·02 0·00 0·00 0·00 Ca 0·97 0·00 0·00 0·00 0·00 0·03 0·02 0·02 Na 0·03 0·00 0·00 0·01 0·01 0·00 0·02 0·00 K 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 total 5·00 2·99 5·00 6·99 6·65 4·99 4·98 4·92 Mg / (Mg + Fetotal) 0·91 0·97 0·90 0·88 0·91 0·93 Ca / (Ca + Na + K) 0·97 Plagioclase Olivine Lizardite Brucite Magnetite Al-serpentine (Al > 0·1 pfu) domain. I domain. II domain. III Wt % SiO2 44·36 40·94 42·54 0·15 4·13 36·01 38·55 35·62 TiO2 0·01 0·00 0·00 0·00 0·00 0·03 0·02 0·00 Al2O3 35·70 0·02 0·08 0·12 0·00 10·64 1·73 13·65 Cr2O3 0·00 0·00 0·01 0·00 0·00 0·00 0·01 0·00 FeO 0·43 8·49 2·18 11·67 84·54 7·67 6·09 4·13 MgO 0·09 50·65 41·67 60·72 1·51 32·91 33·78 31·74 MnO 0·09 0·15 0·03 0·25 0·24 0·09 0·10 0·06 CaO 19·58 0·08 0·03 0·00 0·00 0·52 0·32 0·40 Na2O 0·33 0·03 0·02 0·07 0·05 0·00 0·19 0·00 K2O 0·00 0·00 0·00 0·04 0·00 0·00 0·00 0·00 Total 100·52 100·36 86·56 73·02 90·49 87·87 80·79 85·59 Oxygen 8·00 4·00 7·00 7·00 7·00 7·00 7·00 7·00 Si 2·04 0·99 2·00 0·01 0·36 1·72 1·98 1·70 Ti 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Al 1·94 0·00 0·00 0·01 0·00 0·60 0·11 0·77 Cr 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Fe 0·02 0·17 0·09 0·68 6·07 0·31 0·26 0·17 Mg 0·01 1·83 2·91 6·27 0·19 2·34 2·59 2·26 Mn 0·00 0·00 0·00 0·01 0·02 0·00 0·00 0·00 Ca 0·97 0·00 0·00 0·00 0·00 0·03 0·02 0·02 Na 0·03 0·00 0·00 0·01 0·01 0·00 0·02 0·00 K 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 total 5·00 2·99 5·00 6·99 6·65 4·99 4·98 4·92 Mg / (Mg + Fetotal) 0·91 0·97 0·90 0·88 0·91 0·93 Ca / (Ca + Na + K) 0·97 Table 1 Representative chemical compositions of reactant and product minerals in the experiments Plagioclase Olivine Lizardite Brucite Magnetite Al-serpentine (Al > 0·1 pfu) domain. I domain. II domain. III Wt % SiO2 44·36 40·94 42·54 0·15 4·13 36·01 38·55 35·62 TiO2 0·01 0·00 0·00 0·00 0·00 0·03 0·02 0·00 Al2O3 35·70 0·02 0·08 0·12 0·00 10·64 1·73 13·65 Cr2O3 0·00 0·00 0·01 0·00 0·00 0·00 0·01 0·00 FeO 0·43 8·49 2·18 11·67 84·54 7·67 6·09 4·13 MgO 0·09 50·65 41·67 60·72 1·51 32·91 33·78 31·74 MnO 0·09 0·15 0·03 0·25 0·24 0·09 0·10 0·06 CaO 19·58 0·08 0·03 0·00 0·00 0·52 0·32 0·40 Na2O 0·33 0·03 0·02 0·07 0·05 0·00 0·19 0·00 K2O 0·00 0·00 0·00 0·04 0·00 0·00 0·00 0·00 Total 100·52 100·36 86·56 73·02 90·49 87·87 80·79 85·59 Oxygen 8·00 4·00 7·00 7·00 7·00 7·00 7·00 7·00 Si 2·04 0·99 2·00 0·01 0·36 1·72 1·98 1·70 Ti 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Al 1·94 0·00 0·00 0·01 0·00 0·60 0·11 0·77 Cr 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Fe 0·02 0·17 0·09 0·68 6·07 0·31 0·26 0·17 Mg 0·01 1·83 2·91 6·27 0·19 2·34 2·59 2·26 Mn 0·00 0·00 0·00 0·01 0·02 0·00 0·00 0·00 Ca 0·97 0·00 0·00 0·00 0·00 0·03 0·02 0·02 Na 0·03 0·00 0·00 0·01 0·01 0·00 0·02 0·00 K 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 total 5·00 2·99 5·00 6·99 6·65 4·99 4·98 4·92 Mg / (Mg + Fetotal) 0·91 0·97 0·90 0·88 0·91 0·93 Ca / (Ca + Na + K) 0·97 Plagioclase Olivine Lizardite Brucite Magnetite Al-serpentine (Al > 0·1 pfu) domain. I domain. II domain. III Wt % SiO2 44·36 40·94 42·54 0·15 4·13 36·01 38·55 35·62 TiO2 0·01 0·00 0·00 0·00 0·00 0·03 0·02 0·00 Al2O3 35·70 0·02 0·08 0·12 0·00 10·64 1·73 13·65 Cr2O3 0·00 0·00 0·01 0·00 0·00 0·00 0·01 0·00 FeO 0·43 8·49 2·18 11·67 84·54 7·67 6·09 4·13 MgO 0·09 50·65 41·67 60·72 1·51 32·91 33·78 31·74 MnO 0·09 0·15 0·03 0·25 0·24 0·09 0·10 0·06 CaO 19·58 0·08 0·03 0·00 0·00 0·52 0·32 0·40 Na2O 0·33 0·03 0·02 0·07 0·05 0·00 0·19 0·00 K2O 0·00 0·00 0·00 0·04 0·00 0·00 0·00 0·00 Total 100·52 100·36 86·56 73·02 90·49 87·87 80·79 85·59 Oxygen 8·00 4·00 7·00 7·00 7·00 7·00 7·00 7·00 Si 2·04 0·99 2·00 0·01 0·36 1·72 1·98 1·70 Ti 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Al 1·94 0·00 0·00 0·01 0·00 0·60 0·11 0·77 Cr 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 Fe 0·02 0·17 0·09 0·68 6·07 0·31 0·26 0·17 Mg 0·01 1·83 2·91 6·27 0·19 2·34 2·59 2·26 Mn 0·00 0·00 0·00 0·01 0·02 0·00 0·00 0·00 Ca 0·97 0·00 0·00 0·00 0·00 0·03 0·02 0·02 Na 0·03 0·00 0·00 0·01 0·01 0·00 0·02 0·00 K 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 total 5·00 2·99 5·00 6·99 6·65 4·99 4·98 4·92 Mg / (Mg + Fetotal) 0·91 0·97 0·90 0·88 0·91 0·93 Ca / (Ca + Na + K) 0·97 The Si metasomatic zone was identified as a layer of serpentine ± magnetite in which brucite and Al-serpentine were absent (x = 1·5–3·2 mm in Fig. 4b). A thin layer composed solely of serpentine occurred at x = 1·5–2·4 mm (Fig. 4b). Serpentine in the Si metasomatic zone was also identified as lizardite (Supplementary Data Fig. S2). In the isochemical zone (x > 3·2 mm; Fig. 4b), lizardite formed as fine-grained aggregates of serpentine with a porous structure (Fig. 4g). Olivine was replaced in part by brucite (Fig. 4g), and discrete grains of brucite (∼20 μm) were also observed. Subhedral magnetite with a grain size of ∼10 μm was present (Fig. 4g). Narrow gaps (∼3 μm) were observed between olivine and serpentine, and the original outlines of the olivines were difficult to recognize (Fig. 4g). Extent of hydration The total H2O contents of the solid material (products and unreacted olivine/plagioclase) vary with distance from the Ol–Pl boundary (Fig. 5a). In the Pl-hosted region (segments 1 and 2), the H2O contents are less than 1 wt %, and there was no systematic increase of H2O content with time (Fig. 5a). In contrast, in the Ol-hosted region, hydration proceeded with time (Fig. 5a). The hydrous products were serpentine and brucite, and there was no signature of talc or tremolite in the run products. Far from the Ol–Pl boundary (segments 4–8), the H2O contents of the solids were almost constant throughout the inner tube, but they increased from ∼0·5 wt % at 783 h to ∼1·7–2·0 wt % at 2055 h, ∼5·8 wt % at 3184 h, and ∼8·9 wt % at 7980 h (Fig. 5a). At the contact between olivine and plagioclase at 2055 h, the H2O content of segment three (x = 0–5 mm) was higher than that of the other segments (segments four to eight) in the Ol-hosted region. Fig. 5. View largeDownload slide Temporal evolution of the extent of hydration (LOI: loss of ignition) in the Ol–Pl–H2O experiments, obtained by thermogravimetry. (a) Total weight loss and (b) weight loss for brucite. Brucite was formed in the 3184 h reaction throughout the reaction tube. Fig. 5. View largeDownload slide Temporal evolution of the extent of hydration (LOI: loss of ignition) in the Ol–Pl–H2O experiments, obtained by thermogravimetry. (a) Total weight loss and (b) weight loss for brucite. Brucite was formed in the 3184 h reaction throughout the reaction tube. At 783 and 2055 h, weight loss due to the breakdown of brucite was not observed anywhere in the tube (Fig. 5b). At 3184 and 7980 h, brucite was recognized in all segments of the Ol-hosted region (segments three to eight), but the amount of brucite at the contact with plagioclase (segment three; 0·58 wt % at 3184 h and 0·78 wt % at 7980 h) was lower than at the other sites (segments four to eight; ∼1·0 wt % at 3184 h and ∼1·2 wt % at 7980 h; Fig. 5b). Development of metasomatic zones within the Ol-hosted region Figure 6a–d shows the concentrations of Al (Fig. 6a), Si (Fig. 6b), and Ca (Fig. 6c), and the values of XMg (= Mg / (Mg + Fetotal)) in the products (serpentine ± saponite) (Fig. 6d) as a function of distance from the Ol–Pl boundary. The Al content of serpentine is highest at the Ol–Pl boundary (1·0–1·5 pfu) and decreases away from the boundary (Fig. 6a). The zone of Al metasomatism within the Ol-hosted region enlarged with time, with x = 0·8 mm at 2055 h, 0·9 mm at 3184 h, and 1·6 mm at 7980 h (Fig. 6a). Fig. 6. View largeDownload slide (a) Al contents, (b) Si contents, (c) Ca contents, and (d) XMg (= Mg/(Mg + Fetotal)) values for serpentine (cations per seven oxygen formula unit) as a function of distance from the Ol–Pl boundary. Dashed line shows XMg for olivine. (e) Migration of the Al front, magnetite appearance, and brucite appearance front with time. Each front migrated at a constant rate. The magnetite and brucite appearance front probably migrated in a similar manner. Fig. 6. View largeDownload slide (a) Al contents, (b) Si contents, (c) Ca contents, and (d) XMg (= Mg/(Mg + Fetotal)) values for serpentine (cations per seven oxygen formula unit) as a function of distance from the Ol–Pl boundary. Dashed line shows XMg for olivine. (e) Migration of the Al front, magnetite appearance, and brucite appearance front with time. Each front migrated at a constant rate. The magnetite and brucite appearance front probably migrated in a similar manner. The Si contents of the products (serpentine ± saponite) were lowest at the boundary (∼1·5 pfu), and with increasing distance from the boundary they increased at first, and then decreased gradually (Fig. 6b). The Si metasomatic zone within the Ol-hosted region was identified from the first appearance of euhedral brucite grains, and the Si front was found at x = 1·9 mm at 3184 h and 3·2 mm at 7980 h (Fig. 6b), which means that the Si front extended slightly further than the Al front. At 2055 h, the position of the Si metasomatic front is unclear due to the absence of brucite in the reaction tube. The Ca contents of the reaction products (serpentine ± saponite) were low (<0·1 pfu) at 2055 h throughout the tube (Fig. 6c). The Ca contents of the products increased at first to 0·13 pfu at x = 0·5 mm at 3184 h and 0·18 pfu at x = 0·4 mm at 7980 h, and then decreased away from the Ol–Pl boundary (Fig. 6c). The width of the Ca metasomatic zone within the Ol-hosted region (Fig. 6c) was almost the same as that of the Al metasomatic zone (Fig. 6a, e). The XMg values of the products (serpentine ± saponite) changed with time and space in association with a magnetite front. At 2055 h, the XMg values in the product were roughly constant (∼0·91) and similar to those of the olivine (Fig. 6d). At 3184 h and 7980 h, the spatial variations in XMg values in the products were similar. The XMg values in the product (∼0·91) were almost the same as those of the olivines around the Ol–Pl boundary, but increased with the appearance of magnetite and then became constant (∼0·95) with the appearance of brucite (Fig. 6b, d). The metasomatic fronts of Al and Ca migrated with time with the average migration rate of 0·154 μm/h from 2055 h to 7980 h. The metasomatic fronts of Si, which were recognized as the first appearances of brucite, migrated with time with the average migration rate of 0·259 μm/h from 2055 h to 7980 h (Fig. 6e). The position of the first appearance of magnetite is located between Al and Si front and migrated with the average migration rate of 0·257 μm/h. Zoning of Al in serpentine within the Al–Si metasomatic zone Al-serpentine in the Al–Si metasomatic zone occurs as fine-grained aggregates with clear zoning. The serpentine aggregates can be divided into three domains (Fig. 7a) based on differences in contrast and brightness in the BSE images and they can be described as light grey (domain I), dark grey (domain II), and light grey (domain III). Domain I commonly shows the straight outlines of the olivine grains (Fig. 7a) and most of the interfaces between domain I and the olivine are sharp with no visible gap, though occasionally we observed gaps several micrometres wide (Fig. 7a). Domain II is composed of fibrous serpentine crystals and contains more pores than domain I, although nano-scale pores were observed in domain I (arrows in Fig. 7b). Domain III is in contact with domain II but never in contact with domain I (Fig. 7a, b). Fig. 7. View largeDownload slide Characteristics of Al-serpentine. (a) The Al-serpentine domain I, II, and III are clearly seen in BSE images. (b) Microstructures of Al-serpentine at x = 0·8 mm. The domain II seems to be porous. Black arrows indicate secondary porosities. (c) Representative zoning of Al in an Al-serpentine domain. D, distance from olivine–domain I serpentine interface, L, length from olivine–domain I serpentine interface to domain III. (d) Normalized area of the core part with respect to that of the rim part. Both the core and rim were normalized with areas of olivine + domain I, and increased with time. One standard deviation (1σ) were also presented. (e) and (f) Plots of Mg + Fe vs Si for serpentine from the (e) 3184 h and (f) 7980 h reactions. Different substitutions were observed (see text for details). Fig. 7. View largeDownload slide Characteristics of Al-serpentine. (a) The Al-serpentine domain I, II, and III are clearly seen in BSE images. (b) Microstructures of Al-serpentine at x = 0·8 mm. The domain II seems to be porous. Black arrows indicate secondary porosities. (c) Representative zoning of Al in an Al-serpentine domain. D, distance from olivine–domain I serpentine interface, L, length from olivine–domain I serpentine interface to domain III. (d) Normalized area of the core part with respect to that of the rim part. Both the core and rim were normalized with areas of olivine + domain I, and increased with time. One standard deviation (1σ) were also presented. (e) and (f) Plots of Mg + Fe vs Si for serpentine from the (e) 3184 h and (f) 7980 h reactions. Different substitutions were observed (see text for details). Figure 7c shows the time evolution of Al-zoning at x = 0·8 mm. Even at the same position within the inner tube, the Al-zoning of serpentine changed systematically with time. Zoning of the serpentine aggregates was not developed in the early stages of the experiment (783 and 2055 h), but after 3184 h the zoning into domains I, II, and III became clear. At 3184 h, the Al zoning is characterized by high levels of Al (0·17 pfu) in domain I, a decrease to zero Al in domain II, and an increase to 0·10 pfu in domain III (Fig. 7c). After 7980 h, the Al content of the Al-serpentine in domain I was also high (0·55 pfu), decreased to ∼0·15 in domain II, and increased to 0·54 pfu in domain III (Fig. 7c). The areas of domains I and III were measured separately on BSE images at x = 0·8 mm, and the normalized areas of domains I and III with respect to the total area of olivine + domain I were obtained. Before 2055 h, the areas of domains I and III were set at zero, since zoning was not recognized (Fig. 7d). With time, the areas of domains I and III enlarged, and the normalized areas of domains I and III were 0·60 and 0·38, respectively, at 3184 h (Fig. 7d). At x = 0·8 mm after 7980 h, the normalized areas of domains I and III were 0·78 and 0·72, respectively (Fig. 7d). As described above, the composition of the Al-serpentines varies with distance from the Ol–Pl boundary (Fig. 6a–d). The composition also varies with domain. At 3184 h and x = 0·8 mm, the composition of the Al-serpentines within domain I are explained by the Tschermak substitution (AlAl(Mg, Fe2+)-1Si-1) and brucite mixing (Fig. 7e). The chemical compositions of the Al-serpentines in domain II also indicate the brucite mixing trend, while the Al-serpentines in domain III plot along the brucite and Fe22+Fe3+AlSiO5(OH)4 mixing trends (Fig. 7e). All the Al-serpentines have low contents of Ca (<0·1 pfu), indicating minor effects of saponite mixing. At 7980 h and x = 0·8 mm, the chemical compositions of the Al-serpentines in domains I, II, and III deviate from the Tschermak substitution trend (Fig. 7f), indicating the effects of a substitution of 3R2+(2R3+□)-1 in the M site (R3+ is Fe3+ or Al3+). The amounts of Fe3+ and Al3+ in the Al-serpentine could not be estimated solely from the EPMA data, due to the effect of the Ca-saponite mixing. Solution chemistry Table 2 lists the concentrations of cations in solutions after the Ol–Pl–H2O experiments, the Pl–H2O experiments, and the blank experiments. In the blank experiments, the concentration of Al, Ca, Mg, Fe and Si is below 0·002 mmol/kg (Table 2), which indicates that the effect of the leaching of cations from the vessels were negligible in our experiments. Table 2 Chemistry of solutions in the experiments Time hours Al Ca Fe mmol/kg Mg Si pH (at 25°C) pH (in situ) Ol–Pl–H2O  783 0·008 0·117 0·000 0·056 0·026 6·58 6·49  2055 0·000 0·099 0·005 0·005 0·004 7·15 6·80  3186 0·003 0·040 0·004 0·006 0·006 9·92 7·14  7980 0·000 0·074 0·005 0·003 0·008 10·53 7·66 Pl–H2O  158 0·070 0·306 0·004 0·016 2·155 6·84 7·58  330 0·015 0·351 0·010 0·017 2·392 7·06 7·66  488 0·025 0·457 0·018 0·022 2·407 8·50 7·82 Blank  2224 0·002 0·002 0·000 0·001 0·001 n.d. Time hours Al Ca Fe mmol/kg Mg Si pH (at 25°C) pH (in situ) Ol–Pl–H2O  783 0·008 0·117 0·000 0·056 0·026 6·58 6·49  2055 0·000 0·099 0·005 0·005 0·004 7·15 6·80  3186 0·003 0·040 0·004 0·006 0·006 9·92 7·14  7980 0·000 0·074 0·005 0·003 0·008 10·53 7·66 Pl–H2O  158 0·070 0·306 0·004 0·016 2·155 6·84 7·58  330 0·015 0·351 0·010 0·017 2·392 7·06 7·66  488 0·025 0·457 0·018 0·022 2·407 8·50 7·82 Blank  2224 0·002 0·002 0·000 0·001 0·001 n.d. n.d., not determined. Table 2 Chemistry of solutions in the experiments Time hours Al Ca Fe mmol/kg Mg Si pH (at 25°C) pH (in situ) Ol–Pl–H2O  783 0·008 0·117 0·000 0·056 0·026 6·58 6·49  2055 0·000 0·099 0·005 0·005 0·004 7·15 6·80  3186 0·003 0·040 0·004 0·006 0·006 9·92 7·14  7980 0·000 0·074 0·005 0·003 0·008 10·53 7·66 Pl–H2O  158 0·070 0·306 0·004 0·016 2·155 6·84 7·58  330 0·015 0·351 0·010 0·017 2·392 7·06 7·66  488 0·025 0·457 0·018 0·022 2·407 8·50 7·82 Blank  2224 0·002 0·002 0·000 0·001 0·001 n.d. Time hours Al Ca Fe mmol/kg Mg Si pH (at 25°C) pH (in situ) Ol–Pl–H2O  783 0·008 0·117 0·000 0·056 0·026 6·58 6·49  2055 0·000 0·099 0·005 0·005 0·004 7·15 6·80  3186 0·003 0·040 0·004 0·006 0·006 9·92 7·14  7980 0·000 0·074 0·005 0·003 0·008 10·53 7·66 Pl–H2O  158 0·070 0·306 0·004 0·016 2·155 6·84 7·58  330 0·015 0·351 0·010 0·017 2·392 7·06 7·66  488 0·025 0·457 0·018 0·022 2·407 8·50 7·82 Blank  2224 0·002 0·002 0·000 0·001 0·001 n.d. n.d., not determined. In the Pl–H2O experiments, the concentrations of Al ranged from 0·015 to 0·070 mmol/kg (Table 2). The concentrations of Ca increased with time from 0·306 mmol/kg at 158 h to 0·457 mmol/kg at 488 h, and the concentrations of Si increased from 2·155 mmol/kg at 158 h to 2·407 mmol/kg at 488 h. The pH at room temperature (∼25°C) increased with time from 6·84 at 158 h to 8·50 at 488 h (Table 2). The concentrations of Mg and Fe were higher than in the blank, probably due to contamination from clay minerals in the initial plagioclase grains. In the Ol–Pl–H2O experiments, the solutions were not taken from the inner tube, but their chemistry may represent the solution at the end of the Ol-hosted region (Fig. 1a, b). In the Ol–Pl–H2O experiments, the concentration of Al was relatively high (0·008 mmol/kg) at 783 h and became low (0·001–0·003 mmol/kg) at 2055–7980 h, which is an order of magnitude lower than in the Pl–H2O experiments and similar to the blank experiment (Table 2). The Ca concentration was also high (0·117 mmol/kg) at 783 h before decreasing sharply to near-constant values of 0·040–0·074 mmol/kg at 2055–7980 h, which is also an order of magnitude lower than in the Pl–H2O experiments (Table 2). The Si concentration was high (0·026 mmol/kg) at 783 h before decreasing sharply to near-constant values of 0·002–0·008 mmol/kg at 2055–7980 h, which is three orders of magnitude lower than in the Pl–H2O experiments (Table 2). The concentrations of Fe are extremely low (<0·005 mmol/kg). The concentration of Mg was high (0·056 mmol/kg) at 783 h and decreased to 0·002–0·008 mmol/kg at 2055–7980 h. The pH at room temperature was low (∼6·6) at 783 h and increased to ∼10·5 at 7980 h (Table 2). The solution chemistry for the Pl–H2O experiments is plotted in the activity-diagram of the CaO–Al2O3–MgO–SiO2–H2O system (Fig. 8). Here, only the stability fields of the Mg-endmembers of minerals are shown. The Fe-endmembers were omitted from the system since the boundaries of the minerals would not significantly affect the equilibrium level with the given amount of Fe in each phase. The activity diagram of log aSiO2(aq) versus log aCa2+/a2H+ is shown in Fig. 8a, and the log aSiO2(aq) versus log aAl3+/a3H+ diagram is shown in Fig. 8b. The calculated fluid speciation in the Pl–H2O experiments from 158 to 488 h is largely constant with a log aSiO2(aq) value of ∼3·0 and a log aCa2+/a2H+ value of ∼11·5 (Fig. 8a), and a log aAl3+/a3H+ value of ∼−2·2 (Fig. 8b). The compositions of the solution within the Pl-hosted region in the Ol–Pl–H2O experiments would have been similar to those in the Pl–H2O experiments. Fig. 8. View largeDownload slide Activity diagram in the CaO–Al2O3–MgO–SiO2–H2O system at 230°C and vapor-saturated pressure of 2·80 MPa. The diagrams were projected onto (a) the log aCa2+/a2H+ vs log aSiO2(aq) plane with log aAl3+/a3H+ = −2·5, which is the value for the 2055–7980 h reactions, and (b) the log log aAl3+/a3H+ vs log aSiO2(aq) plane with log aCa2+/a2H+ = 9·1 (grey dashed line) and 11·5 (black solid line), which are the values for the 783 h and 7980 h reactions, respectively. The speciated fluid compositions in the Ol–Pl–H2O and Pl–H2O experiments are shown. Liz, lizardite; Brc, brucite; Chl, Chlorite; Tr, Tremolite; Ca-Sap, Ca-Saponite; Di, Diopside; Tlc, Talc; Qtz, Quartz. Fig. 8. View largeDownload slide Activity diagram in the CaO–Al2O3–MgO–SiO2–H2O system at 230°C and vapor-saturated pressure of 2·80 MPa. The diagrams were projected onto (a) the log aCa2+/a2H+ vs log aSiO2(aq) plane with log aAl3+/a3H+ = −2·5, which is the value for the 2055–7980 h reactions, and (b) the log log aAl3+/a3H+ vs log aSiO2(aq) plane with log aCa2+/a2H+ = 9·1 (grey dashed line) and 11·5 (black solid line), which are the values for the 783 h and 7980 h reactions, respectively. The speciated fluid compositions in the Ol–Pl–H2O and Pl–H2O experiments are shown. Liz, lizardite; Brc, brucite; Chl, Chlorite; Tr, Tremolite; Ca-Sap, Ca-Saponite; Di, Diopside; Tlc, Talc; Qtz, Quartz. Speciated solutions in the Ol–Pl–H2O experiments, taken from the main vessel, were also plotted on the activity diagram of the CaO–Al2O3–MgO–SiO2–H2O system (Fig. 8). At an early stage of the Ol–Pl–H2O experiments (783 h), the fluid fell within the stability field of serpentine (Fig. 8a). The value of log aSiO2(aq) in the solution decreased from −4·5 at 783 h to −5·4 at 2055 h, and the aSiO2(aq) was an order of magnitude lower than in the Pl–H2O experiments (Fig. 8a). The value of log aSiO2(aq) after 2055 h was around -5·4, and fell to around the equilibrium line of serpentine–brucite (Fig. 8a). In the Ol–Pl–H2O experiments, log aCa2+/a2H+ at 783 h was 9·1, and the ratio aCa2+/a2H+ at 783 h was two orders of magnitude lower than in the Pl–H2O experiments. The value of log aCa2+/a2H+ in the Ol–Pl–H2O experiments showed a continuous increase with time from 9·1 at 783 h to 11·0 at 7980 h (Fig. 8a), and became similar to the value in the Pl–H2O experiments. The values of log aCa2+/a2H+and log aSiO2(aq) seem to have been evolving towards equilibrium among serpentine, brucite, and diopside (Fig. 8a). The value of log aAl3+/a3H+ at 783 h (∼1·0) was high in the Ol–Pl–H2O experiments, and it decreased to a constant value (∼ -2·5) from 2055 to 7980 h. Note that the concentration of Al in the Ol–Pl–H2O experiment was extremely low (on the order of a few μg/kg H2O), and thus the log aAl3+/a3H+ value in the Ol–Pl–H2O experiments has large uncertainties. DISCUSSION Metasomatic reactions in the olivine–plagioclase–H2O experiments The chemistries of solutions in the Pl–H2O and Ol–Pl–H2O experiments indicate that the Al and Si that were leached from plagioclase were consumed by olivine hydration within the reaction tube. The high concentration of Ca in the Ol–Pl–H2O solution (Table 1) suggests that the Ca released from Pl was partly consumed in the formation of Ca-saponite in the olivine-hosted metasomatic zone (Fig. 6c). The coupling of Al and Ca in distribution (Fig. 6e) took place because the Ca was contained in Ca-saponite (cf., Ca0·165Mg3Al0·33Si3·67O10(OH)2). As the Ca content of the metasomatic products (serpentine ± saponite) was very low (<0·2 pfu; Fig. 6c), we focus here on a discussion of reactions in the Al2O3–MgO–(FeO)–SiO2–H2O system. Isochemical zone In the early stages of the experiments (783 h), brucite and magnetite were not observed in the reaction tube (Fig. 5b). At this stage, the values of aSiO2(aq) in the solution were higher than the upper limit of the brucite stability field (Fig. 8a, b), which would have inhibited the formation of brucite thermodynamically. The solution at 783 h was supersaturated with lizardite (Fig. 8a, b), suggesting the following Si-adding overall reaction: 3 (Mg0.91, Fe0.09)2SiO4+ SiO2(aq)+ 4 H2Oolivine= 2 (Mg0.91, Fe0.09)3Si2O5(OH)4serpentine (1) Reaction 1 results in an increase in volume of ∼60%. After this induction stage, the solution chemistry moved to the stability field of lizardite + brucite (at 2055, 3184, and 7980 h; Fig. 8a, b), and brucite appeared (Fig. 5b), suggesting that serpentine + brucite were in equilibrium and that the following overall reaction had occurred: (Mg0.91, Fe0.09)2SiO4+ 1.41H2Oolivine→ 0.5 (Mg0.95, Fe0.05)3Si2O5(OH)4serpentine+ 0.368 (Mg0.90, Fe0.10)(OH)2brucite+ 0.04 Fe3O4+ 0.04 H2(aq)magnetite (2) Reaction (2) results in an increase in volume of ∼50%. The change from brucite-absent to brucite-present reactions is similar to that described in previously performed batch-type Ol-H2O experiments (Okamoto et al., 2011; Ogasawara et al., 2013; McCollom et al., 2016) and this suggests that the ‘isochemical’ Ol hydration reaction proceeded far from the Ol–Pl boundary in the Ol–Pl–H2O experiments. Si metasomatic zone The fact that the brucite breakdown front (Si metasomatic front) had advanced farther than the Al metasomatic front (Figs 4b and 6e) indicates that Si-metasomatism was followed by Al–Si metasomatism, forming a metasomatic sequence comprising the Al-serpentine + Ca-saponite zone, the serpentine zone, the serpentine + magnetite zone, and the serpentine + magnetite + brucite zone. When the Si-metasomatic front invaded the isochemical zone (serpentine + brucite + magnetite), brucite breakdown occurred to form serpentine + magnetite, as follows: 19/6(Mg0.90, Fe0.10) (OH)2+ 2 SiO2(aq)brucite → (Mg0.95, Fe0.05)3Si2O5(OH)4serpentine+ 1/18Fe3O4+ 10/9 H2O + 1/9 H+magnetite (3) Reaction 3 results in an increase in volume of ∼40%. The magnetite appearance front followed the brucite appearance front (Figs 4b and 6e) and the values of XMg in the serpentine increased from 0·91 to 0·95 where magnetite appeared (Fig. 6d). These observations suggest the formation of Fe3+-serpentine by the breakdown of magnetite. There are two possible Fe3+-serpentine which have been proposed; Fe23+Si2O5(OH)4 (Klein et al., 2009) and Mg22+Fe3+(Fe3+, Si)O5(OH)4 (Evans et al., 2013). In the case of our experiment, the component of Tschermak substitution looks greater than substitution of 3R2+(2R3+□)-1 in the M site (Fig. 7f), therefore, the magnetite breakdown reaction is written as follows; 4 Fe2+Fe3+2O4+ 3 SiO2(aq)+ 5 H2O + H2(aq)magnetite= 3 Fe2+2Fe3+(Fe3+, Si)O5(OH)4cronstedtite (4) The mineral assemblage during Si-metasomatism is strongly controlled by the chemical potentials of silica (μSiO2) and oxygen (μO2) (Evans et al., 2013; Frost et al., 2013). In our experiments, μSiO2 varied in time and space, induced by diffusion from the Pl-hosted region (Fig. 8a, b). No mineral redox buffers were used in our experiments, but it is reasonable to assume that μO2 would have been buffered by the iron-magnetite buffer (IM; -535 kJ/mol, log fO2 = -49·3), as a reaction vessel of Inconel alloy was used. The μSiO2 versus μO2 pseudosection for the experimental P–T conditions reveals that the mineral assemblage changed with decreasing μSiO2 along the IM buffer from talc to talc + serpentine, serpentine, serpentine + magnetite, serpentine + brucite + magnetite, and finally brucite + magnetite (Fig. 9a). This sequence is largely consistent with the sequence observed in our experiments, although Ca-saponite and Al-serpentine were formed in our experiments due to the effects of Ca and Al (Fig. 4b). The predicted XMg values of serpentine and brucite on the serpentine + magnetite + brucite equilibrium at the IM buffer are 0·95 and 0·91, respectively (Fig. 9b), and these values are similar to the XMg values of the serpentine (0·95; Fig. 6d and Table 1) and brucite (0·90; Table 1) in the isochemical zone. Moreover, the predicted XMg values of serpentine at high silica activity are 0·91 (Fig. 9b), which is similar to the serpentine without magnetite in the Si-metasomatic zone (XMg = 0·91, x = 1·5–2·4 mm at 7980 h; Fig. 6d). Fig. 9. View largeDownload slide (a) SiO2 versus O2 chemical potential pseudosection for olivine (XMg = 0·91) in the FeO–MgO–SiO2–H2O system with H2O saturation at 230°C and a vapor-saturated pressure of 2·80 MPa, as obtained using the software Perple X 6·7·9 (Connolly, 2009). Mineral redox buffers on iron–magnetite (IM) and fayalite–magnetite–quartz buffer (FMQ) were shown in the dashed line. (b) Predicted XMg values of serpentine (XMgSrp) and brucite (XMgBrc) at the IM buffer. XMgBrc was slightly higher than 0·91 because minor magnetite was formed. Srp, serpentine; Brc, brucite; Ol, olivine; Mag, magnetite; Hem, hematite; Tlc, talc. Fig. 9. View largeDownload slide (a) SiO2 versus O2 chemical potential pseudosection for olivine (XMg = 0·91) in the FeO–MgO–SiO2–H2O system with H2O saturation at 230°C and a vapor-saturated pressure of 2·80 MPa, as obtained using the software Perple X 6·7·9 (Connolly, 2009). Mineral redox buffers on iron–magnetite (IM) and fayalite–magnetite–quartz buffer (FMQ) were shown in the dashed line. (b) Predicted XMg values of serpentine (XMgSrp) and brucite (XMgBrc) at the IM buffer. XMgBrc was slightly higher than 0·91 because minor magnetite was formed. Srp, serpentine; Brc, brucite; Ol, olivine; Mag, magnetite; Hem, hematite; Tlc, talc. Al–Si metasomatic zone In the Al–Si metasomatic zone, compositional zoning that formed during the replacement of olivine produced Al-rich domain I, Al-poor domain II, and Al-rich domain III (Fig. 7a, b). We obtained the area proportions (%) of domains I, II, and III from BSE images. Although nano-scale pores could not be considered in our area analyses due to the resolution of the analyzed BSE images, they do not affect the following scenario regarding the development of each domain. The initial amount of olivine within the reaction tube was 45·6 ± 3·9 area % (mean ± one standard deviation, 1σ), which is greater than the area proportion of olivine + domain I (41·2 ± 2·7 area % at 7980 h) and smaller than the area proportion of olivine + domain I + domain II (72·3 ± 5·0 area % at 7980 h). These relationships indicate that the initial olivine grain outlines were located within domain II. Furthermore, the fact that the area of residual olivine + domain I was constant, regardless of the progress of the reaction (41·3 ± 3·0 area % at 3184 h, and 41·2 ± 2·7 area % at 7980 h) suggests that the Al-serpentine in domain I preserved the shape of the olivine at the time when the Al front arrived. In summary, our experiments revealed that (1) Al-free serpentine (domain II) formed at an early stage, and (2) when the Al metasomatic front reached the reaction site, Al-serpentine was formed both inwards (domain I) and outwards (domain III) (Fig. 10a, b). Fig. 10. View largeDownload slide Schematic image of (a) Al front migration and (b) reaction mechanism in the Ol–Pl–H2O experiments from t = t0 (before reaction), t1 (after reaction and before arrival of Al front to a reaction site), and to t2 (after arrival of Al front to a reaction site). (a) As the Al front migrates from t = t0 to t2, the Al concentration in the fluid at a reaction site increases. (b) Reaction mechanism for Al-serpentine formation from t = t0 (left-side diagram), t = t1 (diagram at center), and to t = t2 (right-side diagram). Narrow and bold black arrows indicate the transport of elements and the growth direction, respectively. (c) Schematic of variations in chemical potential of magnesium (μMg), silica (μSiO2), and aluminum (μAl) across domain III to olivine transect at t = t1 and t = t2 in Fig. 10(a) and (b). Fig. 10. View largeDownload slide Schematic image of (a) Al front migration and (b) reaction mechanism in the Ol–Pl–H2O experiments from t = t0 (before reaction), t1 (after reaction and before arrival of Al front to a reaction site), and to t2 (after arrival of Al front to a reaction site). (a) As the Al front migrates from t = t0 to t2, the Al concentration in the fluid at a reaction site increases. (b) Reaction mechanism for Al-serpentine formation from t = t0 (left-side diagram), t = t1 (diagram at center), and to t = t2 (right-side diagram). Narrow and bold black arrows indicate the transport of elements and the growth direction, respectively. (c) Schematic of variations in chemical potential of magnesium (μMg), silica (μSiO2), and aluminum (μAl) across domain III to olivine transect at t = t1 and t = t2 in Fig. 10(a) and (b). In domain I, the shape of the olivine was preserved, but this shape had already been modified from the original shape of the olivine before alteration during the earlier serpentinization that took place during the formation of domain II (Fig. 10b). As it was difficult to measure the exact volume of nano-pores, the following mass balance equations were obtained assuming an isovolume reaction (i.e. no nano-pores) during the formation of domain I as a first-order approximation. We note, too, that the isovolume reaction is only for the mass balance in domain I, and that the overall reaction for domains I + III involved an increase in the volume of the solid material. In the Al2O3–MgO–SiO2–H2O system, the pseudomorphic reaction in domain I can be written as follows: αMg2SiO4+βAl3++ (–2α+β+ 5) H2O + (4α– 2β+ 6) H+→olivineMg3–0.5βAlβSi2–0.5βO5(OH)4+ (2α+ 0.5β– 3) Mg2++ (α+ 0.5β– 2) SiO2(aq)Al−serpentine (5) where α and β indicate the coefficients of olivine and Al3+ required to produce 1 mol of Al-serpentine, respectively. The Fe and Ca components are ignored for simplicity. Assuming reaction (5) is an isovolume reaction, α is written as V-Al-srp/ V-Fo, where V-Fo and V-Al-srp are the molar volumes of forsterite and Al-serpentine, respectively. The V-Al-srp was obtained from the linear mixing of the molar volume of lizardite (Mg3Si2O5(OH)4) ( V-Liz) and amesite (Mg4Al4Si2O10(OH)8) ( V-Ame), and calculated as V-Al-srp =  V-Liz - 0·5 ( V-Liz - 0·5 V-Ame) β. The molar volumes of minerals were taken from Holland & Powell (2011). The value of β varies with the chemical composition of domain I Al-serpentine. We emphasize again that the estimated amounts of removed Mg and Si in reaction (5) are minimum estimates (no pores). Given the average Al content of domain I at 7980 h (α = 2·42 and β = 0·42), the typical reaction can be written as follows: 2.42 Mg2SiO4+ 0.42 Al3++ 0.58 H2O + 14.84 H+→Mg2.79Al0.42Si1.79O5(OH)4+ 2.05 Mg2++ 0.63 SiO2(aq) (6) This typical reaction of domain I proceeds with a supply of Al and with the removal of Mg and SiO2(aq) at the reaction front between domain I and olivine (Fig. 10b). The Mg and Si released from olivine during reaction (5) were transported outwards through the nano-scale pores, and used in the precipitation of Al-serpentine in domain III with a supply of Al and Si in the pore fluids from the Pl-hosted region, as follows: (3 – 0.5β) Mg2++ (2 – 0.5β) SiO2(aq)+βAl3++ (5 –β) H2O→Mg3–0.5βAlβSi2–0.5βO5(OH)4+ (6 – 2β) H+ (7) When we assume a balance between the Mg released in reaction (5) (domain I) and the Mg consumed in reaction (7) (domain III), the overall reaction for the formation of domains I and III could be rewritten as follows: (3 – 0.5β) Mg2SiO4+ 2βAl3++ (1 –β) SiO2(aq)+ (4β+ 4) H2O →Olivine2 Mg3–0.5βAlβSi2–0.5βO5(OH)4+ 8βH+Al−serpentine (8) Since the Al content in serpentine (β) ranges from 0–1·0 (Fig. 6a), this overall reaction represents the process where the serpentinization in the metasomatic zone proceeds with a supply of Si and Al from plagioclase, except at the Ol–Pl boundary, where the β value exceeds 1·0 (Fig. 6a). Therefore, at the Ol–Pl boundary, reaction (7) represents the overall reaction that proceeds with the supply of Al and the removal of Si. Figure 10c shows schematic profiles of the chemical potential along a traverse across domains I, II, and III. Since the source of Mg was olivine, the chemical potential of Mg (μMg) is high in contact with olivine, and decreases from domain I to domain II to domain III (Fig. 10c). Si and Al were supplied from plagioclase through macroscopic pores, and thus the chemical potential of Al (μAl) and silica (μSiO2) would be high at the margin of domain III, but decreases towards the olivine interface (Fig. 10c). The directions of Mg and Al transport, as inferred from the mass balance (Fig. 10b), are consistent with the downhill nature of the chemical potential by diffusion. In contrast, the transport of Si seems to occur in the direction opposed to the chemical potential gradient (Fig. 10b, c). Such uphill diffusion could have occurred during the formation of domain I due to the constant-volume constraint and the effects of diffusion of other elements (e.g. Nishiyama, 1998). Controls on pseudomorphic replacement One of the most important findings of our study is that the pseudomorphic replacements that occurred in domain I in the Al metasomatic zone were coupled with overgrowths in domain III (Figs 7a, b and 9b). Similar combinations of replacements and overgrowths in volume-increasing reactions have often been reported, as for example in the replacement of hematite by chalcopyrite (∼230% volume increase; Zhao et al., 2014) and magnetite by pyrite (∼65% volume increase; Qian et al., 2010). In contrast to the metasomatic zone, in the isochemical zone, the original olivine outlines were not preserved during serpentinization (Fig. 4g). The question is: what causes such differences in replacement textures, given that both reactions show similar increases in volume (40–50%)? Pseudomorphic replacements (i.e. preserving the outline of the parent mineral) are thought to proceed via interface-coupled dissolution–precipitation reactions (e.g. Putnis, 2009), and the textural clues to this process are porosity in the product phase and sharp interfaces between the parent and product minerals. Such textures are observed in the Al metasomatic zone, involving the interfaces between the olivine and domain I (Fig. 7a), and the porosity in the product phase (Fig. 7b). The macroscopic pores among the original olivine grains (∼10% porosity on the >10 μm scale) provided the main element transport path, but microscopic pores (nano-scale) in the replacement materials were also important for the progress of the reaction at the fluid–olivine interface. The process of serpentinization involves dissolution of the primary mineral, transport of the solute, and precipitation of secondary minerals; the relative rates of these processes are critical for pseudomorphism. Xia et al. (2009) showed that a high degree of pseudomorphism could be observed when the reaction is either transport-limited or dissolution-limited, but not precipitation-limited. Previous batch-type experiments of olivine hydration (reaction (2)) revealed that the rate of serpentinization increases with a decrease in the initial grain size of olivine, indicating that the rate-limiting process is the dissolution of olivine (Wegner & Ernst, 1983; Malvoisin et al., 2012). This seems to contradict our experimental results, where the replacement was not pseudomorphic in the isochemical region (Fig. 4d), and it is unclear why pseudomorphic replacement occurred only in domain I of the metasomatic zone. In both the metasomatic and isochemical zones, the rates of solute transport (rtrans) and precipitation of serpentine (rpre) would have been greater than the rate of olivine dissolution (rdis), as shown by the results of the powder experiments of olivine hydration undertaken by Wegner & Ernst (1983) and Malvoisin et al. (2012). We consider that the key to pseudomorphism during serpentinization is the relative magnitudes of rpre and rtrans. In the isochemical zone, solute transport would have been faster than the precipitation of secondary minerals (rdis < rpre < rtrans) and this would have allowed a dispersed distribution of secondary minerals outside the original outlines of the olivine (Fig. 4g). In contrast, when solute transport is slower than the precipitation rate (rdis < rtrans< rpre), the dissolved materials would be consumed immediately before their transport to distant sites, and this would result in the preservation of the outlines of parent minerals in dissolution-limited reactions. The change in the replacement textures between the metasomatic (pseudomorphic) and isochemical zones (non-pseudomorphic) implies that a supply of Al and Si triggered the change in the relative magnitudes of rtrans and rpre. It is known that hydrothermal solutions containing Al and Si can contain complexes such as Al(OH)3H3SiO4- (Salvi et al., 1998; Andreani et al., 2013). Figure 11 shows the abundance of Al species at 230°C and vapor-saturated pressure as a function of aSiO2(aq). With a low activity of silica around the serpentine–brucite equilibria (log aSiO2(aq) = −5·5), the dominant Al species is Al(OH)4−, and this decreases with increasing silica activity. In contrast, Al(OH)3H3SiO4− appears in the high aSiO2(aq) condition (log aSiO2(aq) > −5), and the fraction of Al(OH)3H3SiO4− exceeds Al(OH)4− at log aSiO2(aq) > −2·5. The silica activity inferred from the mineral assemblages (Fig. 9a) and the solution chemistry observed in our experiments (Fig. 8a, b) suggests that Al(OH)4- was dominant in the isochemical zone, whereas abundant Al(OH)3(HSiO4)- existed in the Al-Si metasomatic zone (Fig. 11). Andreani et al. (2013) reported that the formation of Al–Si complexes enhances the dissolution of olivine. In our experiments, the Al–Si complex would already have formed in the pore fluids derived from the Pl-hosted region. The large size of such a complex could have slowed the transport of Al to the reaction front through the nano-pores of domain I, similar to the mechanism suggested for silica complexes (Applin, 1987). Accordingly, the formation of Al(OH)3(HSiO4)– possibly changed the relative rates among the processes from rdis < rpre < rtrans in the isochemical zone to rdis < rtrans < rpre in the metasomatic zone. Further studies on the formation and diffusivities of Al–Si complexes, including molecular dynamic simulations, are needed to understand the impact of Al mobility on the overall reaction rates. Fig. 11. View largeDownload slide Molar fraction of dissolved Al-species at 230°C and vapor-saturated pressure of 2·80 MPa with pHin situ = 7·0 and aH2O = 1. The thermodynamic data are from Tagirov & Schott (2001), and references therein. Fig. 11. View largeDownload slide Molar fraction of dissolved Al-species at 230°C and vapor-saturated pressure of 2·80 MPa with pHin situ = 7·0 and aH2O = 1. The thermodynamic data are from Tagirov & Schott (2001), and references therein. Linkage of macroscopic Al diffusion and the development of Al zoning Al-Si metasomatism near the Ol-Pl contact caused the macroscopic spatial variation of Al content in the products (Figs 4a and 6a) and the temporal changes in Al concentrations in the solution. To establish the link between macroscopic Al transport during metasomatism and the development of Al zoning in the replacement products in the Ol-hosted region, we consider the one-dimensional reaction–diffusion model within a porous medium (Lasaga, 1998; Philpotts & Auge, 2009): ∂(φCi)∂t=∂∂xφDi∂Ci∂x + φ∑Rchem (9) where φ, Ci, Di and Rchem respectively indicate porosity, the concentration of species i in the solution (mol/cm3 H2O), the diffusivity of species i in porous media (cm2/s), the rate of gain or loss of species i by surface reaction (mol/cm3 H2O/s). For simplicity, we just considered a single overall reaction, and φ and Di were assumed to be constant during the reaction. When we assumed the local equilibrium of Al between minerals and fluid, the equilibrium fluid/solid partition coefficient by volume was defined as KV=Ci/CiSolid ( CiSolid is the concentration of species i in the total mineral (mol/cm3 solid)). Although KV varies with fluid composition, it was introduced as a constant. Under these assumptions, the reaction-diffusion equation in the Ol-hosted region used in the simulations of this study are written as follows: φKV∂CAlSolid∂t=φKVDAl∂2CAlSolid∂x2-(1-φ)∂CAlSolid∂t (10) which could be rearranged to yield ∂CAlSolid∂t=DAl*∂2CAlSolid∂x2 (11) where D*Al (cm2/s) is the apparent diffusion coefficient of Al in the porous medium, which was defined as DAl* =φKVφKV + (1-φ)DAl (12) Equation (11) can be solved analytically as follows: CAlSolid=CAl, 0Solid+(CAl, BSolid – CAl, 0Solid) erfcx2DAl*t (13) where the erfc is the complimentary error function, and the t (second) is the reaction time. The CAl, 0Solid and CAl, BSolid are respectively the concentration of Al in the solid phase at t = 0 (initial conditions), and the concentration of Al in the solid phase at the Ol–Pl boundary (boundary conditions). The values of CAl, 0Solid and CAl, BSolid were set as 0·00 (mol/cm3 solid) and 9·00 × 10-3 (mol/cm3 solid), respectively, based on the observed values of CAlSolid in our experiments. The values of CAlSolid in our experiments were calculated by mass balance using the Al content of the serpentine (β; Fig. 6a) and the modal abundance of Al-serpentine (VAl-srp), and by CAlSolid = β × (1/(1 − φ)) × (VAl-srp/ V-Al-srp). Since KV and DAl in equation (12) could not be obtained independently, the values of D*Al that fit the observed values of CAlSolid at both 3184 h and 7980 h were optimized by the least squares method. The result of simple reaction–diffusion modeling at 3184 and 7980 h are shown in Fig. 12a and b, respectively. The value of D*Al was estimated to be 2·63 × 10-7 (cm2/s), although the estimated diffusivity was on the order of that at ∼10% porosity. The model largely reproduced the Al concentration profile as a function of distance from the Ol–Pl boundary in our experiments (Fig. 12a, b). The model also reproduced the Al front, defined as the distance that satisfies CAlSolid/CAl, BSolid = 0·01, observed in our experiments (Fig. 12c). Figure 12d shows the time- evolution of the calculated CAlSolid at x = 0·5, 0·8, and 1·0 mm. With the progress of the metasomatic front, CAlSolidin the metasomatic zone at a fixed distance increased with time (Fig. 12d). For example, at x = 0·8 mm the value of CAlSolid was <0·1 (mmol/cm3 solid) until 2000 h, and it then increased from 0·1 (mmol/cm3 solid) at 2000 h to 2·0 (mmol/cm3 solid) at 8000 h (Fig. 12d). This increase in the Al content of the Al-serpentine was probably recorded as the growth zoning of domains I and III (Fig. 7c), although the directions of growth are opposite to each other (inwards and outwards, respectively; Fig. 10b). Fig. 12. View largeDownload slide Results of numerical simulations using a one-dimensional reaction–diffusion model. The numerical simulations were carried out from x = 0 mm to 2·0 mm in the Ol-hosted region. (a,b) Al concentrations in bulk solid (CSolidAl) after (a) 3184 h and (b) 7980 h reactions. The values of CSolidAl obtained from experiments (symbols), presented as the mean value of the Al concentration with one standard deviation (1σ), are compared with the calculated values from the numerical simulations (lines). (c) The Al fronts (symbols) obtained from experiments (Fig· 6e) and calculations (lines). (d) Calculated time evolution of CSolidAl at fixed distances of x = 0·5, 0·8, and 1·0 mm. Fig. 12. View largeDownload slide Results of numerical simulations using a one-dimensional reaction–diffusion model. The numerical simulations were carried out from x = 0 mm to 2·0 mm in the Ol-hosted region. (a,b) Al concentrations in bulk solid (CSolidAl) after (a) 3184 h and (b) 7980 h reactions. The values of CSolidAl obtained from experiments (symbols), presented as the mean value of the Al concentration with one standard deviation (1σ), are compared with the calculated values from the numerical simulations (lines). (c) The Al fronts (symbols) obtained from experiments (Fig· 6e) and calculations (lines). (d) Calculated time evolution of CSolidAl at fixed distances of x = 0·5, 0·8, and 1·0 mm. There are some discrepancies between the observed and calculated Al profile around x = 0–0·5 mm at 7980 h (Fig. 12b). Such discrepancies probably result from the effects of the porosity reduction and grain-scale compositional heterogeneity (Al zoning; Fig. 7c) in the experimental products, which were not included in the model. Further detailed modelling including porosity changes and surface reaction rate is needed to reproduce fully the CAlSolid observed in our experiments. Implications for natural hydrothermal alteration Millimetre-scale Al–Si metasomatism The results of our experiments are comparable to the natural alteration of Pl-bearing mafic or ultramafic rocks (including troctolites) in the oceanic lithosphere. Frost et al. (2008) reported that prehnite/grossular replaces plagioclase and chlorite replaces serpentine in olivine-rich troctolite. The estimated distances of movement of Si and Al from plagioclase to the olivine grains were ∼5 mm and ∼1 mm, respectively. In the plagioclase wehrlite from the Izu–Ogasawara forearc region (Fig. 2), the Si front inferred from the first appearance of brucite (>1·0 mm from the Ol–Pl boundary) is more advanced than the Al-front (a zone of Al-rich serpentine; ∼0·9 mm from the Ol–Pl boundary). This relative elemental mobility is consistent with that observed in our experiments (x = 3·2 mm for Si and 1·6 mm for Al after 7980 h; Fig. 6e). Moreover, the lack of magnetite around plagioclase grains in the plagioclase wehrlite is also consistent with that observed in our experiments. In the plagioclase wehrlite from the Izu–Ogasawara fore-arc region, olivine reacted at a faster rate in the Al–Si metasomatic zone (<1·0 mm from the Ol–Pl boundary; Fig. 2b) than in the isochemical zone (>1·0 mm from the Ol–Pl boundary; Fig. 2b). Andreani et al. (2013) and Pens et al. (2016) suggested that a supply of Al enhances hydration, whereas Ogasawara et al. (2013) showed that a supply of silica enhances the hydration rate of olivine with absence of brucite. Since plagioclase supplied Al and Si simultaneously to the olivine region (Fig. 2a), it is difficult to distinguish the effects of Si and Al on the olivine hydration rate. Al-zoning of serpentine aggregate in the mesh texture The Al-zoning in the natural mesh texture, with Al-rich cores, Al-poor rims, and Al-rich serpentine veins (Figs 2g and 3c), is essentially the same as that produced in our experiments (Fig. 7c, e). It is reasonable, therefore, to consider that the mesh rims were formed at the earlier stage under Al-free conditions and that the mesh cores and veins were formed at a later stage when Al was supplied from Al-bearing minerals such as plagioclase and clinopyroxene. This scenario is consistent with the proposed two-stage process of mesh formation, which involves a transition from a closed to an open system during serpentinization (e.g. Viti & Mellini, 1998; Bach et al., 2006). Natural rocks have low porosities compared with the reactants in our experiments and the elements being removed from the mesh core during pseudomorphic replacement would migrate into fractures to form Al-rich veins (Figs 2g, h and 3c, d). In our experiments, the Al contents in domain I (Fig. 7c) are similar to those in the mesh core of the Pl-wehrlite (Fig. 1h), but are an order of magnitude lower than in the mesh core of the harzburgite (Fig. 2d). As suggested by the reaction–diffusion modeling (Fig. 12), the differences in the Al contents of serpentine would have been controlled by the difference in the Al content between Al-source minerals ( CAlSolid; Cpx vs Pl). The preservation of Al-poor domains (domain II) or mesh rims in the zoning (Figs 2g and 3c) indicates that once serpentine was formed from Al-free solutions, the kinetics of re-equilibration was very slow, probably because such re-equilibration involved a two-site coupled substitution (i.e. Tschermak substitution; Anovitz, 1991). However, in our experiments, Al contents in domain II did increase with time to some extent (up to ∼0·2 pfu at 7980h; Fig. 7c), whereas the Al content of the mesh rim in the Pl-bearing wehrlite remained very low (∼0·05 pfu; Fig. 2h). We cannot explain this difference, but it is possibly due to the poor crystallinity of our experimental serpentine. The poor crystallinity of serpentine minerals that were formed in hydrothermal experiments has been reported previously, as judged from relatively high water contents rather than structural formulae (e.g. Kühnel et al., 1975; McCollom et al., 2016) and the poor crystallinity makes the mineral susceptible to chemical alteration via local dissolution and precipitation processes. A potential reference frame of volume changes during serpentinization Serpentinization often proceeds without the formation of brucite, but it is difficult to decide whether brucite-absent serpentinization is induced by a supply of Si, the removal of Mg, or the removal of both Si and Mg (e.g. Putnis, 2009; Putnis & Austrheim, 2010; Malvoisin, 2015) and this difficulty reflects the lack of a reliable frame of reference for mass balance analyses in natural systems. Our results presented here suggest that (1) mesh rims were formed under Al-free conditions, and (2) the mesh cores and veins were formed at a later stage when Al was supplied. During the formation of the mesh rims (the first stage of mesh texture formation; t = t1 in Fig. 10b), volume expansion would have occurred, but it is difficult to estimate accurately the extent of expansion because the initial olivine–fluid boundary is generally unclear. In this study we have shown that the shape of olivine was preserved during the formation of the Al-rich mesh cores (i.e. in the second stage of mesh texture formation; t = t2 in Fig. 10c), which means that a large loss of material occurred during serpentinization of the mesh cores. This provides a reliable physical reference frame for deciphering volumetric changes after the arrival of Al in natural serpentinized rocks, which involves reactions 5–8. With an assumption of isovolumetric replacement of the mesh core, the volume increase at this stage is equal to the volume of veins in the mesh texture. Accordingly, the volume increase ratio can be obtained easily by (area of mesh core + area of vein) / (area of mesh core). For example, the mesh core and vein formation in the harzburgite from the Mineoka ophiolite resulted in a volume increase of ∼13%, and Pl-bearing wehrlite from the Izu–Ogasawara forearc region shows a volume increase of ∼16%. Aluminum occurs in a number of different minerals, including pyroxenes, spinel and plagioclase, in both crustal and mantle rocks (Mellini et al., 2005; Boschi et al., 2006; Frost et al., 2008; Plümper et al., 2012a). In the Pl-bearing werhlite, radial fractures filled by Al-rich serpentine and chlorite were produced around the Pl grain in the metasomatic zone (Fig. 2a). The breakdown of the Al-bearing minerals could release aluminum as well as silica episodically and locally, which results in the heterogeneous progression of serpentinization. The preservation of Al-zoning observed in the mesh texture in the natural rocks (Figs 2g and 3c) suggests that the effective bulk composition of the peridotite could vary with the progression of serpentinization, since the serpentinized parts of the peridotite would become less reactive once they formed. The local increase in serpentinization rate around Al-bearing minerals could also cause reaction-induced fracturing by local volume expansion (Jamtveit et al., 2008; Kuleci et al., 2017). The reaction-induced fracturing would enhance fluid supply, thus leading to a positive feedback between hydration and fracturing. Such positive chemical–mechanical feedback induced by elemental transport provides an important means of enhancing the overall hydration rates around plagioclase–olivine grain contacts and, on a much bigger scale, the crust–mantle boundary. CONCLUSIONS (1) Prominent Al-zoning is present in the mesh textures of serpentinized olivine in samples of harzburgite and Pl-bearing wehrlite from the oceanic lithosphere. In both rocks, the mesh zoning is characterized by Al-rich mesh cores, Al-poor mesh rims and Al-rich veins. (2) Hydrothermal experiments were conducted in the olivine–plagioclase–H2O system at 230°C and under a vapor-saturated pressure. Metasomatism took place during the experiments so that the mineral assemblages changed systematically with distance from the olivine–plagioclase boundary, with Al-rich serpentine + Ca-saponite in the metasomatic zone changing to serpentine, serpentine + magnetite and serpentine + brucite + magnetite. This change of mineral assemblage is explained by decrease in activity of SiO2(aq). (3) In the metasomatic zone, the mesh texture of serpentine showed a clear zoning with an Al-rich domain I, an Al-poor domain II, and an Al-rich domain III. Microtextural and chemical features indicate that this zoning was produced by the initial formation of domain II in an Al-free solution, and the subsequent pseudomorphic replacement of olivine (domain I) and the development of overgrowths (domain III) as the Al metasomatic front migrated. During the pseudomorphic replacement, the transport of large amounts of various elements was facilitated by the nano-scale porosity of the products. The transition of the replacement texture from a non-pseudomorphic to a pseudomorphic product was probably triggered by the supply of Al. The formation of Al–Si complexes probably slowed the rate of diffusive transport rather than the precipitation of serpentine. (4) The similarities between Al zoning in natural mesh textures and the Al zoning in our experimental products suggest that the textures are initiated by the formation of mesh rims in solutions that are initially Al-free. Al is subsequently supplied to the solutions, and an Al metasomatic front develops and migrates with the simultaneous pseudomorphic replacement of olivine to form mesh cores and the precipitation of Al-rich serpentine veins. Such transition from closed to open system could be enhanced by fracturing induced by heterogeneous distribution of Al-Si-bearing minerals. ACKNOWLEDGMENTS We thank Masaoki Uno, Nobuo Hirano, Takayoshi Nagaya, Fumiko Higashino, and Otgonbayar Dandar for valuable discussions, and the captain and crew of the R/V Hakuo–Maru, Osamu Ishizuka, and the KH07–02 science party for their support and cooperation. We also thank Jeffrey Alt, Katy Evans, an anonymous reviewer, and editor James Beard for their helpful comments the manuscript. FUNDING This work was supported by a Grant-in-Aid for Research Fellows from the Japan Society for the Promotion of Science [JP16J04140 to R.O.], a Program for Leading Graduate Schools [to R.O.], a Grant-in-Aid for Scientific Research [16H06347 to A.O.], a Grant-in-Aid for Young Scientists [23684042 to A.O.], and a Grant-in-Aid for Specially Promoted Research [25000009 to N.T.]. SUPPLEMENTARY DATA Supplementary data are available at Journal of Petrology online. REFERENCES Andreani M. , Daniel I. , Pollet-Villard M. ( 2013 ). Aluminum speeds up the hydrothermal alteration of olivine . American Mineralogist 98 , 1738 – 1744 . Google Scholar CrossRef Search ADS Anovitz L. M. ( 1991 ). Al zoning in pyroxene and plagioclase: window on late prograde to early retrograde P–T paths in granulite terranes . American Mineralogist 76 , 1328 – 1343 . Applin K. ( 1987 ). The diffusion of dissolved silica in dilute aqueous solution . Geochimica et Cosmochimica Acta 51 , 2147 – 2151 . Google Scholar CrossRef Search ADS Bach W. , Paulick H. , Garrido C. J. , Ildefonse B. , Meurer W. P. , Humphris S. E. ( 2006 ). Unraveling the sequence of serpentinization reactions: petrography, mineral chemistry, and petrophysics of serpentinites from MAR 15°N (ODP Leg 209, Site 1274) . Geophysical Research Letters 33 , L13306 . Google Scholar CrossRef Search ADS Beard J. S. , Frost B. R. ( 2016 ). The stoichiometric effects of ferric iron substitutions in serpentine from microprobe data . International Geology Review 6814 , 1 – 7 . Google Scholar CrossRef Search ADS Beard J. S. , Frost B. R. , Fryer P. , McCaig A. , Searle R. , Ildefonse B. , Zinin P. , Sharma S. K. ( 2009 ). Onset and progression of serpentinization and magnetite formation in olivine-rich troctolite from IODP hole U1309D . Journal of Petrology 50 , 387 – 403 . Google Scholar CrossRef Search ADS Bethke C. M. ( 2007 ). Geochemical and Biogeochemical Reaction Modeling . Cambridge, UK : Cambridge University Press . Google Scholar CrossRef Search ADS Boschi C. , Früh–Green G. L. , Delacour A. , Karson J. A. , Kelley D. S. ( 2006 ). Mass transfer and fluid flow during detachment faulting and development of an oceanic core complex, Atlantis Massif (MAR 30°N) . Geochemistry, Geophysics, Geosystems 7 , Q01004 . Google Scholar CrossRef Search ADS Boudier F. , Baronnet A. , Mainprice D. ( 2010 ). Serpentine mineral replacements of natural olivine and their seismic implications: oceanic lizardite versus subduction-related antigorite . Journal of Petrology 51 , 495 – 512 . Google Scholar CrossRef Search ADS Connolly J. A. D. ( 2009 ). The geodynamic equation of state: what and how . Geochemistry, Geophysics, Geosystems 10 , Q10014 . Google Scholar CrossRef Search ADS Enami M. ( 2006 ). Mineralogical methods for identification of asbestos and their limitations . Japanese Magazine of Mineralogical and Petrological Sciences 35 , 11 – 21 . (Japanese with English abstract). Google Scholar CrossRef Search ADS Escartín J. , Mével C. , MacLeod C. J. , McCaig A. M. ( 2003 ). Constraints on deformation conditions and the origin of oceanic detachments: the Mid-Atlantic Ridge core complex at 15°45′N . Geochemistry, Geophysics, Geosystems 4 , 1067 – 1103 . Google Scholar CrossRef Search ADS Evans K. A. , Powell R. , Frost B. R. ( 2013 ). Using equilibrium thermodynamics in the study of metasomatic alteration, illustrated by an application to serpentinites . Lithos 168–169 , 67 – 84 . Google Scholar CrossRef Search ADS Frost B. R. , Beard J. S. , McCaig A. , Condliffe E. ( 2008 ). The formation of micro-rodingites from IODP hole U1309D: key to understanding the process of serpentinization . Journal of Petrology 49 , 1579 – 1588 . Google Scholar CrossRef Search ADS Frost B. R. , Evans K. A. , Swapp S. M. , Beard J. S. , Mothersole F. E. ( 2013 ). The process of serpentinization in dunite from New Caledonia . Lithos 178 , 24 – 39 . Google Scholar CrossRef Search ADS Harigane Y. , Michibayashi K. , Morishita T. , Tani K. , Dick H. J. B. , Ishizuka O. ( 2013 ). The earliest mantle fabrics formed during subduction zone infancy . Earth and Planetary Science Letters 377–378 , 106 – 113 . Google Scholar CrossRef Search ADS Hirauchi K. , Katayama I. , Uehara S. , Miyahara M. , Takai Y. ( 2010 ). Inhibition of subduction thrust earthquakes by low-temperature plastic flow in serpentine . Earth and Planetary Science Letters 295 , 349 – 357 . Google Scholar CrossRef Search ADS Holland T. J. B. , Powell R. ( 2011 ). An improved and extended internally consistent thermodynamic dataset for phases of petrological interest, involving a new equation of state for solids . Journal of Metamorphic Geology 29 , 333 – 383 . Google Scholar CrossRef Search ADS Hyndman R. D. , Peacock S. M. ( 2003 ). Serpentinization of the forearc mantle . Earth and Planetary Science Letters 212 , 417 – 432 . Google Scholar CrossRef Search ADS Ishizuka O. , Tani K. , Reagan M. K. , Kanayama K. , Umino S. , Harigane Y. , Sakamoto I. , Miyajima Y. , Yuasa M. , Dunkley D. J. ( 2011 ). The timescales of subduction initiation and subsequent evolution of an oceanic island arc . Earth and Planetary Science Letters 306 , 229 – 240 . Google Scholar CrossRef Search ADS Iyer K. , Rüpke L. H. , Morgan J. P. ( 2010 ). Feedbacks between mantle hydration and hydrothermal convection at ocean spreading centers . Earth and Planetary Science Letters 296 , 34 – 44 . Google Scholar CrossRef Search ADS Jamtveit B. , Malthe-Sørenssen A. , Kostenko O. ( 2008 ). Reaction enhanced permeability during retrogressive metamorphism . Earth and Planetary Science Letters 267 , 620 – 627 . Google Scholar CrossRef Search ADS Johnson J. W. , Oelkers E. H. , Helgeson H. C. ( 1992 ). SUPCRT92: a software package for calculating the standard molal thermodynamic properties of minerals, gases, aqueous species, and reactions from 1 to 5000 bar and 0 to 1000°C . Computers & Geosciences 18 , 899 – 947 . Google Scholar CrossRef Search ADS Katayama I. , Kurosaki I. , Hirauchi K. ( 2010 ). Low silica activity for hydrogen generation during serpentinization: an example of natural serpentinites in the Mineoka ophiolite complex, central Japan . Earth and Planetary Science Letters 298 , 199 – 204 . Google Scholar CrossRef Search ADS Kelley D. S. , Karson J. A. , Blackman D. K. , Früh-Green G. L. , Butterfield D. A. , Lilley M. D. , Olson E. J. , Schrenk M. O. , Roe K. K. , Lebon G. T. , Rivizzigno P. ( 2001 ). An off-axis hydrothermal vent field near the Mid-Atlantic Ridge at 30°N . Nature 412 , 145 – 149 . Google Scholar CrossRef Search ADS PubMed Klein F. , Bach W. , Jöns N. , McCollom T. , Moskowitz B. , Berquó T. ( 2009 ). Iron partitioning and hydrogen generation during serpentinization of abyssal peridotites from 15°N on the Mid-Atlantic Ridge . Geochimica et Cosmochimica Acta 73 , 6868 – 6893 . Google Scholar CrossRef Search ADS Kleppe A. K. , Jephcoat A. P. , Welch M. D. ( 2003 ). The effect of pressure upon hydrogen bonding in chlorite: a Raman spectroscopic study of clinochlore to 26.5 GPa . American Mineralogist 88 , 567 – 573 . Google Scholar CrossRef Search ADS Kühnel R. A. , Roorda H. J. , Steensma J. J. ( 1975 ). The crystallinity of minerals—a new variable in pedogenetic processes: a study of goethite and associated silicates in laterites . Clays and Clay Minerals 23 , 349 – 354 . Google Scholar CrossRef Search ADS Kuleci H. , Ulven O. I. , Rybacki E. , Wunder B. , Abart R. ( 2017 ). Reaction-induced fracturing in a hot pressed calcite-periclase aggregate . Journal of Structural Geology 94 , 116 – 135 . Google Scholar CrossRef Search ADS Lasaga A. C. ( 1998 ). Kinetic Theory in the Earth Sciences . Princeton, Chichester : Princeton University Press . Google Scholar CrossRef Search ADS Malvoisin B. ( 2015 ). Mass transfer in the oceanic lithosphere: serpentinization is not isochemical . Earth and Planetary Science Letters 430 , 75 – 85 . Google Scholar CrossRef Search ADS Malvoisin B. , Brunet F. , Carlut J. , Rouméjon S. , Cannat M. ( 2012 ). Serpentinization of oceanic peridotites: 2. Kinetics and processes of San Carlos olivine hydrothermal alteration . Journal of Geophysical Research: Solid Earth 117 , B04102 . McCollom T. M. , Klein F. , Robbins M. , Moskowitz B. , Berquó T. S. , Jöns N. , Bach W. , Templeton A. ( 2016 ). Temperature trends for reaction rates, hydrogen generation, and partitioning of iron during experimental serpentinization of olivine . Geochimica et Cosmochimica Acta 181 , 175 – 200 . Google Scholar CrossRef Search ADS Mellini M. , Rumori C. , Viti C. ( 2005 ). Hydrothermally reset magmatic spinels in retrograde serpentinites: formation of “ferritchromit” rims and chlorite aureoles . Contributions to Mineralogy and Petrology 149 , 266 – 275 . Google Scholar CrossRef Search ADS Morishita T. , Tani K. , Shukuno H. , Harigane Y. , Tamura A. , Kumagai H. , Hellebrand E. ( 2011 ). Diversity of melt conduits in the Izu–Bonin–Mariana forearc mantle: implications for the earliest stage of arc magmatism . Geology 39 , 411 – 414 . Google Scholar CrossRef Search ADS Nakatani T. , Nakamura M. ( 2016 ). Experimental constraints on the serpentinization rate of fore-arc peridotites: implications for the upwelling condition of the slab-derived fluid . Geochemistry, Geophysics, Geosystems 17 , 3393 – 3419 . Google Scholar CrossRef Search ADS Nishiyama T. ( 1998 ). Uphill diffusion and a new nonlinear diffusion equation in ternary non-electrolyte system . Physics of the Earth and Planetary Interiors 107 , 33 – 51 . Google Scholar CrossRef Search ADS Nozaka T. , Wintsch R. P. , Meyer R. ( 2017 ). Serpentinization of olivine in troctolites and olivine gabbros from the Hess Deep Rift . Lithos 282–283 , 201 – 214 . Google Scholar CrossRef Search ADS Ogasawara Y. , Okamoto A. , Hirano N. , Tsuchiya N. ( 2013 ). Coupled reactions and silica diffusion during serpentinization . Geochimica et Cosmochimica Acta 119 , 212 – 230 . Google Scholar CrossRef Search ADS O’Hanley D. S. ( 1996 ). Serpentinites. Records of Tectonic and Petrological History. Oxford Monographs of Geology and Geophysics . New York, Oxford : Oxford University Press . Okamoto A. , Shimizu H. ( 2015 ). Contrasting fracture patterns induced by volume-increasing and -decreasing reactions: implications for the progress of metamorphic reactions . Earth and Planetary Science Letters 417 , 9 – 18 . Google Scholar CrossRef Search ADS Okamoto A. , Ogasawara Y. , Ogawa Y. , Tsuchiya N. ( 2011 ). Progress of hydration reactions in olivine–H2O and orthopyroxene–H2O systems at 250°C and vapor-saturated pressure . Chemical Geology 289 , 245 – 255 . Google Scholar CrossRef Search ADS Oyanagi R. , Okamoto A. , Hirano N. , Tsuchiya N. ( 2015 ). Competitive hydration and dehydration at olivine–quartz boundary revealed by hydrothermal experiments: implications for silica metasomatism at the crust–mantle boundary . Earth and Planetary Science Letters 425 , 44 – 54 . Google Scholar CrossRef Search ADS Pens M. , Andreani M. , Daniel I. , Perrillat J. P. , Cardon H. ( 2016 ). Contrasted effect of aluminum on the serpentinization rate of olivine and orthopyroxene under hydrothermal conditions . Chemical Geology 441 , 256 – 264 . Google Scholar CrossRef Search ADS Philpotts A. , Auge J. ( 2009 ). Principles of Igneous and Metamorphic Petrology . Cambridge : Cambridge University Press . Google Scholar CrossRef Search ADS Plümper O. , Piazolo S. , Austrheim H. ( 2012a ). Olivine pseudomorphs after serpentinized orthopyroxene record transient oceanic lithospheric mantle dehydration (Leka ophiolite complex, Norway) . Journal of Petrology 53 , 1943 – 1968 . Google Scholar CrossRef Search ADS Plümper O. , Royne A. , Magraso A. , Jamtveit B. ( 2012b ). The interface-scale mechanism of reaction-induced fracturing during serpentinization . Geology 40 , 1103 – 1106 . Google Scholar CrossRef Search ADS Plümper O. , John T. , Podladchikov Y. Y. , Vrijmoed J. C. , Scambelluri M. ( 2017 ). Fluid escape from subduction zones controlled by channel-forming reactive porosity . Nature Geoscience 10 , 150 – 156 . Google Scholar CrossRef Search ADS Powell R. , Holland T. ( 1993 ). On the formulation of simple mixing models for complex phases . American Mineralogist 78 , 1174 – 1180 . Putnis A. ( 2009 ). Mineral replacement reactions . Reviews in Mineralogy and Geochemistry 70 , 87 – 124 . Google Scholar CrossRef Search ADS Putnis A. , Austrheim H. ( 2010 ). Fluid-induced processes: metasomatism and metamorphism . Geofluids 10 , 254 – 269 . Qian G. , Brugger J. , Skinner W. M. , Chen G. , Pring A. ( 2010 ). An experimental study of the mechanism of the replacement of magnetite by pyrite up to 300°C . Geochimica et Cosmochimica Acta 74 , 5610 – 5630 . Google Scholar CrossRef Search ADS Rouméjon S. , Cannat M. ( 2014 ). Serpentinization of mantle-derived peridotites at mid-ocean ridges: mesh texture development in the context of tectonic exhumation . Geochemistry, Geophysics, Geosystems 15 , 2354 – 2379 . Google Scholar CrossRef Search ADS Salvi S. , Pokrovski G. S. , Schott J. ( 1998 ). Experimental investigation of aluminum–silica aqueous complexing at 300°C . Chemical Geology 151 , 51 – 67 . Google Scholar CrossRef Search ADS Sato H. , Taniguchi H. , Takahashi N. , Mohiuddin M. M. , Hirano N. , Ogawa Y. ( 1999 ). The origin of the Mineoka ophiolite . Journal of Geography 108 , 203 – 189 . (Japanese with English abstract). Google Scholar CrossRef Search ADS Schwarzenbach E. M. , Caddick M. J. , Beard J. S. , Bodnar R. J. ( 2016 ). Serpentinization, element transfer, and the progressive development of zoning in veins: evidence from a partially serpentinized harzburgite . Contributions to Mineralogy and Petrology 171 , 1079 . Google Scholar CrossRef Search ADS Shimizu H. , Okamoto A. ( 2016 ). The roles of fluid transport and surface reaction in reaction-induced fracturing, with implications for the development of mesh textures in serpentinites . Contributions to Mineralogy and Petrology 171 , 73 . Google Scholar CrossRef Search ADS Sleep N. H. , Meibom A. , Fridriksson T. , Coleman R. G. , Bird D. K. ( 2004 ). H2-rich fluids from serpentinization: geochemical and biotic implications . Proceedings of the National Academy of Sciences of the United States of America 101 , 12818 – 12823 . Google Scholar CrossRef Search ADS PubMed Stefánsson A. ( 2001 ). Dissolution of primary minerals of basalt in natural waters: I. Calculation of mineral solubilities from 0°C to 350°C . Chemical Geology 172 , 225 – 250 . Google Scholar CrossRef Search ADS Tagirov B. , Schott J. ( 2001 ). Aluminum speciation in crustal fluids revisited . Geochimica et Cosmochimica Acta 65 , 3965 – 3992 . Google Scholar CrossRef Search ADS Viti C. , Mellini M. ( 1998 ). Mesh textures and bastites in the Elba retrograde serpentinites . European Journal of Mineralogy 10 , 1341 – 1360 . Google Scholar CrossRef Search ADS Wada I. , Wang K. , He J. , Hyndman R. D. ( 2008 ). Weakening of the subduction interface and its effects on surface heat flow, slab dehydration, and mantle wedge serpentinization . Journal of Geophysical Research 113 , B04402 . Google Scholar CrossRef Search ADS Wegner W. W. , Ernst W. G. ( 1983 ). Experimentally determined hydration and dehydration reaction rates in the system MgO–SiO2–H2O . American Journal of Science 283A , 151 – 180 . Xia F. , Brugger J. , Chen G. , Ngothai Y. , O’Neill B. , Putnis A. , Pring A. ( 2009 ). Mechanism and kinetics of pseudomorphic mineral replacement reactions: a case study of the replacement of pentlandite by violarite . Geochimica et Cosmochimica Acta 73 , 1945 – 1969 . Google Scholar CrossRef Search ADS Zhao J. , Brugger J. , Chen G. , Ngothai Y. , Pring A. ( 2014 ). Experimental study of the formation of chalcopyrite and bornite via the sulfidation of hematite: mineral replacements with a large volume increase . American Mineralogist 99 , 343 – 354 . Google Scholar CrossRef Search ADS Zimmer K. , Zhang Y. , Lu P. , Chen Y. , Zhang G. , Dalkilic M. , Zhu C. ( 2016 ). SUPCRTBL: a revised and extended thermodynamic dataset and software package of SUPCRT92 . Computers & Geosciences 90 , 97 – 111 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Journal of PetrologyOxford University Press

Published: Apr 14, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off