Seismic faults triggered early stage serpentinization of peridotites from the Samail Ophiolite, Oman

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.


Introduction
Hydrothermal circulation within the oceanic lithosphere has first order effects on its thermal evolution (e.g. Schuiling, 1964;Stein and Stein, 1994), and on the mass-transfer between the oceans and the solid Earth (e.g. Paulick et al., 2006). Serpentinization involves hydration of olivine and pyroxene and their replacement by serpentine and other hydrous minerals, and leads to major changes in the physical and chemical properties of the lithosphere (e.g. Escartin et al., 2001;Malvoisin, 2015). It also affects the biological activity at the seafloor (e.g. Holm and Charlou, 2001). Whereas fluid circulation in the shallow oceanic crust is exten-  Nicolas et al., 2000) and the approximate inferred position of the fossil ridge axis and mantle diapirs based on structural observations (after Boudier et al., 1997). b) Map of the Batin dunite area showing the location of the three drill holes where cores were recovered during the Oman Drilling Project (after Noël, 2018). The legend is the same as in a). Lineation isodips (after Ildefonse et al., 1993) indicate the approximate location of the diapir at the North-West of the Oman DP drilling sites. c) Profile showing the relative location of the three recovered cores and location of the two main samples presented in this study. (For interpretation of the colors in the figure(s), the reader is referred to the web version of this article.) associated with fast-spreading ridges are harder to constrain as mantle rocks are covered by ca. 6 km of basaltic crust and direct sampling by Ocean Drilling at these depths has not yet been performed. An alternative approach is to study fragments of oceanic crust emplaced onto continental crust during plate convergence, i.e. ophiolites.
Here we describe highly localized brittle faults through dunite and harzburgite in the Samail ophiolite based on a study of drill cores obtained through the Oman Drilling Project. We focus on the earliest stages of faulting, which we interpret to have occurred during the incipient stages of serpentinization and hydrothermal activity.

The Batin dunite area
The Samail ophiolite is located on the eastern corner of the Arabic peninsula (Fig. 1a) and covers an area ca. 500 km long and 60 km wide along the coast. The lithospheric section is up to approximately 20 km thick with the typical structure and composition associated with fast spreading ridges. Remarkably, the paleoridge axis seems to have been preserved (Boudier et al., 1997;Fig. 1a). Former oceanic lithosphere overlays an amphibolite grade metamorphic sole with an inverted temperature gradient. The ages of the metamorphic sole and the crust indicate that the rocks composing the ophiolite were formed and obducted between 96-95 Ma (Rioux et al., 2016) during the closure of the Neo-Tethys ocean.
The Batin dunite is a 10 km long and 2.5 km wide tabular dunitic lens located in the southernmost massif of the Oman ophiolite (Fig. 1a) in the upper part of the mantle section, close to the mantle-crust transition. Dunite is believed to form by dissolution of orthopyroxene from harzburgite during interaction with basaltic melt (Kelemen et al., 1997). The Batin dunite is thus considered to be a zone of melt accumulation. This observation, in addition to plunging mantle lineations (Fig. 1b), led Nicolas et al. (1988) to interpret the area as a former mantle diapir feeding the former ridge axis with magma. However, while normal diapirs are normally located below the ridge axis, like the ones observed along the East Pacific Rise (Toomey et al., 2007) and most of the other mantle diapirs identified in Oman (Fig. 1a), the Batin diapir is an off-axis diapir located ca. 50 km from the ridge axis.
The Batin region is tectonically complex. Many generations of faults crosscut the area. Some of them are associated with the emplacement of the off-axis diapir into the cold lithosphere as is observed in the Mansah diapir, also in Oman (Jousselin and Nicolas, 2000). Other faults likely formed during the emplacement of the ophiolite on the continental lithosphere, but the largest ones seem to have formed during the formation of the oceanic lithosphere. Indeed, similar faults in the neighboring Samail massif were interpreted as syn-magmatic by Rospabé et al. (2019) and hence have formed while the ridge axis was still active. Another study by Zihlmann et al. (2018), conducted in the same massif as the BA site on a fault parallel to the set of major faults observed in the Batin region, recognized these lithospheric faults as major fluid pathways controlling the hydrothermal alteration of the lower crust and eventually uppermost mantle.

The Oman Drilling Project -active alteration cores
Samples in this study come from drill cores produced during the second drilling phase of the Oman Drilling Project (Oman DP; https://www.omandrilling .ac .uk/) during the winter 2017-2018. The Active Alteration (BA) site is located at the southeast end of the Batin dunite lens area (Fig. 1b). Seven holes were drilled including three recovered cores at Holes BA1B, BA3A and BA4A. Thorough descriptions of these cores (a total of 1000 m) made onboard of the research vessel Chikyu during summer 2018 are available in an IODP report (Kelemen et al., 2020). Summary logs of the cores are provided in Supplementary Fig. S1.
The cores are ca. 300 m (BA3A and BA4A) and 400 m long (BA1B) and located within or near the Batin dunite body (Fig. 1c). BA4A is mostly composed of dunite and located at the tip of the Batin dunite. The upper 160 m of BA1B are composed of dunite, likely belonging to the same body as BA4A. A 3-meters wide fault zone separates the dunitic part from the underlying 240 m, mostly composed of harzburgite. This major fault may be of the same type as the syn-magmatic hydrothermal faults described by Rospabé et al. (2019) and Zihlmann et al. (2018). However, a thorough study would be required to verify this assertion. BA3A is located outside the Batin dunite lens and is mostly composed of harzburgite. The three cores display very high degrees of serpentinization (above 80%) both as pervasive alteration and in veins. All cores are crosscut by a number of dykes, most of them with a gabbroic or clinopyroxenitic composition. These dykes have highly variable degrees of alteration but are never completely fresh. Clinopyroxene is locally preserved. Deformation within the BA cores is relatively abundant and mainly brittle. Apart from a decrease in the number of open cracks and cataclastic zones within the weathering profile, deformation structures do not have a specific repartition pattern. The fault zone separating the dunitic and harzburgitic parts in BA1B is the only large scale deformation feature that has been reported in the cores.
In the following, we focus on one specific type of fault that has been observed during the description of the BA cores by the Oman DP scientific team. Even though the samples we describe are strongly serpentinized, we refer to their original lithologies as their rock type. However, any mineral description refers to the current mineralogy of the samples.

Fault zone characteristics
The oldest generation of faults identified in the Samail peridotites are dark sealed faults and associated cataclastic bands. They are present in all the cores and do not appear to be restricted to a specific depth. However, due to their discreet appearance in the cores, the exact repartition of these faults is unknown. They seem to be an early feature as they crosscut dykes but are crosscut by most of the other structures (e.g. open cracks, veins). They represent single events with no trace of reactivation. Faults are generally relatively thin (several millimeters at most) compared to the displacement they induced (up to several centimeters). However, not all the faults are associated with centimeter-scale displacement, it can be much smaller (sometimes less than one millimeter). We refer to these faults as "dark faults" in the following.
In this study, we compare two samples representing two different stages of the dark faults development. Sample DF_1 is crosscut by a series of thin dark fractures or faults with very small displacements (less than a millimeter; Fig. 2a-b). It comes from one of the freshest dunitic parts of the BA cores. The sample is still associated with a high degree of serpentinization (>90%), but olivine relics are locally common and particularly concentrated in the central parts of unfractured domains (Fig. 2b). The internal part of the fractures is characterized by intense fragmentation (Fig. 2c). Sample DF-2 is crosscut by a dark fault associated with ca. 6 cm displacement (Fig. 2d). The dunitic sample is completely serpentinized. The internal part of the fault zone is developed into a cataclasite with clasts embedded in a fine-grained serpentine matrix (Fig. 2e). The deformation is mostly contained within the cataclastic zone, but some strain is accommodated by microfracturing and small-scale faults in the wall-rock. The wall-rock damage is asymmetrically distributed with one side (thin sections' top sides in Fig. 2e) more damaged than the other. While the wall-rock contact with the main cataclastic zone is sharp on the least damaged side, it is more complex and gradual on the most damaged side. In both samples, the main dark fault is sub-vertical and can be followed over more than one meter within the cores.
The dark fault in sample DF-2 crosscuts a gabbroic dyke. This dyke contains clinopyroxene that has been partially preserved from alteration and provides an opportunity to characterize fault deformation. Two electron backscatter diffraction (EBSD) maps were obtained from a zone of the DF-2 cataclasite where remnant clinopyroxene clasts from the dyke are abundant. Details on the sample preparation and acquisition method are provided in Appendix A. The first orientation map (DF-2_map_1) was carried out in a domain containing rather large clasts of clinopyroxene, whereas the second map (DF-2_map_2) was carried out in a smaller area containing a large number of fine-grained clasts (orientation maps location is given in Fig. 3). Data was processed using the MatLab toolbox MTEX (version 5.2.8; http://mtex -toolbox .github .io; Bachmann et al., 2010Bachmann et al., , 2011Hielscher and Schaeben, 2008). A misorientation threshold of 10 • was applied during grain modeling and detected grains with less than 10 pixels were not considered in our calculations. Clinopyroxene orientation maps are provided in Supplementary Fig. S2.
The largest clasts in the large-grained domain show a relatively strong crystal preferred orientation (CPO; upper pole figures in Fig. 3c) with two maxima, which are likely to be inherited from the original grains' orientations before faulting. However, when we consider all the grains in the map, the orientation of (100), (010) and [001] have a quasi-uniform distribution ( Fig. 3c) even though traces of the pre-faulting CPO are still visible in the intermediatesized clasts (two central rows of pole figures in Fig. 3c). The grain size area (2-D) distribution from domain DF-2_map_1 follows a well-defined power law with a slope of 1.77 (Fig. 4a). In contrast, the grain size distribution from DF-2_map_2 is best fitted by a lognormal distribution (Fig. 4b). The difference in distribution between the two EBSD maps is likely due to a higher alteration degree of the small grains from DF-2_map_2. Indeed, because their larger surface area to volume ratio, small grains are more altered than large ones. Finally, the Shape Preferred Orientation (SPO) of the grains was extracted from the EBSD maps (Fig. 4c). Both maps show a SPO oriented around 110-120 • from the orientation of the dark fault. Interestingly, a secondary SPO parallel to the dark fault orientation can be observed in the results from DF-2_map_2.

Geochemistry of mesh textured serpentinite
The peridotites from the BA cores show extensive and pervasive serpentinization. Whole rock trace element compositions are within the range of results previously obtained for Oman peri-  has been replaced during serpentinization by a well-developed mesh texture composed of mesh veins that form a network of serpentine veins, and mesh cores. We have not analyzed the serpentine polymorph composing the mesh veins in DF-1 and DF-2, but X-ray analyses made during Oman DP on similar samples (Kelemen et al., 2020) and optical properties strongly suggest it is lizardite. Mesh cores are composed of olivine remnants or alteration phases dominated by serpentine. The composition and structure of the mesh cores are highly variable. Magnetite is rare within the mesh texture but occurs as numerous micrometric grains within the cataclastic matrix.
In samples DF-1 and DF-2 the mesh veins form a network of two almost perpendicular sets. In both cases the most developed set of veins is sub-perpendicular to the dark faults. In sample DF-2, where the dark fault is well developed as a cataclastic zone, the mesh texture is present both in the wall-rocks and in the largest clasts in the cataclasite. The orientation of the veins as well as the crystallographic orientation of serpentine inside the veins appear continuous from the wall-rock through the cataclastic zone clasts (Fig. 5). This reflects limited clast rotation after mesh formation.
Oxygen isotope compositions and boron concentration were measured in-situ (in thin section DF-2C) by SIMS within the mesh veins (in the wall-rock and in the clasts), as well as in the cataclastic matrix in the fault zone (Table 1;  There is in consequence no major difference between the boron content in the mesh veins from the wall-rock and from the fault zone clasts. There is no significant correlation between oxygen isotopes and boron concentrations. A striking feature of the mesh cores observed in this study is a pronounced oscillatory zoning pattern. In sample DF-1, serpentinized mesh cores are most common close to the early stage dark fractures (Fig. 2b). These mesh cores appear colorless under optical light, but chemical maps reveal that they are characterized by alternating silica and iron rich layers (Fig. 6a). Magnesium is evenly distributed inside the mesh cores and sulfide grains are rare. In sample DF-2, mesh cores located around and within the main dark fault share some similarities with the patterns observed in sample DF-1, indicative of a similar formation process, but also some significant differences. In plane-polarized light, their color varies from brown, associated with well-defined zoning patterns, to black, showing no obvious structure. This evolution from brown to black is mostly visible in the wall-rock close to the dark fault, but also in some of the large clasts, where the clast cores show lighter and better-defined patterns than their rim. The brown patterns have a structure similar to the patterns observed in the sample DF-1, but their chemical composition is different (Fig. 6b). The silica-iron layering is still very prominent, but magnesium and sulfur (as sulfide grains) are enriched in the iron-rich, silica-poor, layers. These changes are increasingly striking as the mesh cores become darker and the layering less defined as in Fig. 6c.
Major elements measurements by electron microprobe were made in the mesh veins, in the mesh cores, both in the wall-rock and in the fault zone clasts, and in the cataclasite matrix (Supplementary Tables S4-S8; methods in Appendix D). These analyses were made in thin section DF-2C. Mesh veins (both in the wall-rock and in the clasts) and the cataclasite matrix have rather homogeneous serpentine compositions. Mesh cores in the wallrock as well as in the clasts show much more variation (Fig. 7a). Two trends can be distinguished within the mesh cores composition. The first trend (blue arrows in Figs. 7) is characterized by a decrease in silica and is mainly represented by the mesh cores in the wall-rock. It is associated with a decrease in the totals, reaching values as low as 75% for samples with less than 5wt% SiO 2 (Fig. 7b). This is consistent with a composition dominated by brucite. The second trend (orange arrows in Figs. 7) is associated with an increase of iron content and is mainly represented by mesh cores in the clasts. It is associated with sulfur enrichment and totals up to 100% (Fig. 7b). This trend is consistent with a high Fe-sulfide content. We observe a positive correlation between sulfur and alumina contents in all the measurements. The wall-rock and clasts cores have the highest Al and S concentrations, but a slight increase in Al and S concentrations in the clasts mesh veins and the cataclasite matrix is also observed compared to the wallrock mesh veins ( Supplementary Fig. S5).

Early stage faulting
Our observations indicate an intimate link between deformation and serpentinization. The orientation of the mesh veins in sample DF-2 is the same in the wall-rock and in the dark fault's clasts independent of their size (Fig. 5). This is consistent with the main hydration event occurring after the faulting. Moreover, we remark that in both samples DF-1 and DF-2 the orientation of the most developed set of mesh veins is approximately perpendicular to the dark faults. This highlights a possible link between faulting and mesh vein formation. Finally, in sample DF-1, the olivine relics in mesh cores are located exclusively at some distance from the early stage dark fractures (Fig. 2b), indicating that mesh core replacement is also linked to the occurrence of these early stage brittle features.
The dark faults are associated with intense fragmentation, even when displacement is limited as it is the case with sample DF-1 Fig. 4. Grain size distributions and shape orientations of clinopyroxene clasts obtained from the EBSD maps. Grain size distributions are defined by the probability density functions (pdf) of the grain cross-section areas. The pdf returns the probability to encounter a grain of given area. Grain size distributions are fitted following the same method as used in Aupart et al. (2018). Only grains larger than 10 pixels are used. a) Grain size distribution from the large-scale map (DF-2_map_1) presented in Fig. 3a. The distribution is well fitted by a power law, described by two parameters α and C. α is the slope of the power law, describing the proportion of large grains to small grains, and C a scaling coefficient, depending on the grain spatial density. b) Grain size distribution from the small-scale map (DF-2_map_2) presented in Fig. 3b. The distribution is best fitted by a lognormal distribution that is described by two parameters μ and σ . μ is linked to the median of the distribution while σ is related to the spread of the distribution. c) Rose diagram showing the distribution of the shape orientation of the grains in DF-2_map_1 and DF-2_map_2. In this diagram the dark fault is oriented sub-parallel to a 0 • orientation. In a) and b) N is the total number of grains used in the grain size and shape orientation distributions. (Fig. 2c). This observation suggests an energetic, possibly seismic, event. The sharp fault edges and asymmetric wall-rock damage (Fig. 2e) support this hypothesis. The power law scaling exponent of ca. 1.8 in the grain size area distribution (= 2.6 for the radii distribution; Fig. 4a) of the dark fault clasts from sample DF-2 is similar to the value obtained by Sammis et al. (1987) and Steacy and Sammis (1991) for fragmentation during grinding in seismic fault zones. The size dependent rotation of the clasts indicated by the weakening of the CPO with decreasing grain size (Fig. 3c) and the development of a SPO ( Fig. 4c; Cladouhos, 1999) also point towards intense brittle deformation.
Our observations stress that early faulting plays a major role in the initiation of serpentinization by initiating the supply of fluids to initially dry mantle peridotites. The systematic orientation of the main mesh veins subperpendicular to the dark faults may indicate that the tectonic stress responsible for the faulting also affects the mesh orientation.

Conditions of serpentinization
We observe moderate enrichments in U, Rb, Li and B in our samples ( Supplementary Fig. S3a) typical of serpentinization in an oceanic environment (Deschamps et al., 2011). The FME concen-trations we observe are lower than typical abyssal peridotite signature (that have undergone extensive interaction with seawater at the seafloor; Peters et al., 2017) but consistent with depleted peridotites drilled along the Mid-Atlantic Ridge (that have been less exposed to seafloor alteration) (Supplementary Fig. S6; Godard et al., 2008;Paulick et al., 2006). These results are thus consistent with fluid-rock interactions occurring at a certain depth below the seafloor and relatively low water-rock ratio (Agranier et al., 2007).
The mesh veins and the matrix cataclasite have relatively homogeneous compositions, coherent with pure serpentine (Fig. 7a). The oxygen isotope analyses, made in these phases, can be interpreted in terms of temperature of serpentinization (Saccocia et al., 2009). The oxygen isotope composition of serpentine formed by hydration of a peridotite protolith depends on the temperature and fluid composition. The fluid composition may range from seawater (δ 18 O = 0 ) to positive values for hydrothermally altered water. If a thick oceanic crust was covering the peridotite at the time of serpentinization it is likely that the fluid composition was modified by fluid-rock interactions when reaching the cooling peridotites. Temperatures of serpentinization obtained by assuming that seawater was the serpentinizing fluid are in the range 200-250 • C (  Mesh veins have either a cyan or orange color, which indicate two predominant crystal orientations approximately perpendicular to each other, common to the texture of serpentine derived from olivine. These two orientations correspond to the two mesh vein groups identified by oxygen isotopes analyses (Table 1 and Supplementary  Fig. S4). The magenta color indicates isotropic or very fine-grained zones (here, mostly the cataclastic matrix). The dashed red line indicates the limit between the dark fault cataclastic zone (above) and the wall-rock (below).

Table 1
Oxygen isotope compositions and boron concentrations (ion probe analyses) made in thin section DF-2C. The location number refers to the different points analyzed. A detailed map of these is given in Supplementary Fig. S4. Temperatures are calculated based on Saccocia et al. (2009) and considering either seawater or a hydrothermally evolved fluid as the reactive fluid. In the calculation, a δ 18 O-value of 0 is used for seawater and of 2.4 for hydrothermally evolved water (Campbell et al., 1988). WR = wall-rock.  (1988) from the Snake Pit vent field), the estimated temperature of serpentinization is between ca. 250 and 300 • C ( Table 1).
The range of observed δ 18 O-composition for Group 1 and 2 mesh veins (Table 1)  respectively), implies that mesh vein formation occurred at 210-260 • C (Fig. 8). Since the B-content of the fluid is expected to be less sensitive to fluidrock interactions, this is supported by the relatively constant Bconcentration for all mesh veins (Table 1), irrespective of orientation. The obtained serpentinization temperature range is slightly higher than temperatures expected in the absence of magnetite in the mesh veins (Klein et al., 2014) but lower than the magnetite formation peak (around 300 • C; McCollom and Bach, 2009). It is also commonly observed that magnetite preferentially forms within the higher fluid flux pathways (e.g. Bach et al., 2006) this would be consistent with the preferential occurrence of magnetite in the matrix of the cataclastic zone that likely focused the highest fluid flux. The temperature formation of the mesh cores, usually considered to form later than the mesh veins as the system becomes more open (Viti and Mellini, 1998), is likely better constrained by the cataclastic matrix as, as representative of the main fluid pathway. Considering seawater as the reactive fluid suggests temperature estimates around 210 • C. This is broadly consistent with the brucite Mg# (cationic Mg/(Mg+Fe)) near 0.7-0.8 in the mesh cores as inferred from Fig. 7a (McCollom and Bach, 2009).
The BA site is located ca. 50 km from the fossil ridge axis. The temperatures estimated in this study are surprisingly low considering the position of the samples within the lithospheric section. Such temperatures could not be reached purely through conductive cooling of the lithosphere. However, numerous studies highlight the important role of hydrothermal cooling especially close to the ridge axis (e.g. Stein and Stein, 1994). Fast spreading ridges, such as the fossil Oman ridge, are usually not considered to be sites of extensive serpentinization due to their high temperatures. Nonetheless, it has been shown that formation of lithospheric scale hydrothermal cells can provide enough cooling for local serpen-  tinization to occur at distances exceeding ca. 20 km from the ridge axis (Iyer et al., 2010). This is consistent with the BA drilling site being located approximately 50 km from the fossil ridge axis. It also means that the lithosphere sampled at the BA site could have been serpentinized at a shorter distance to the ridge axis than its current position relative to the ridge axis if conditions were favorable. The location and size of the large scale hydrothermal cells is likely controlled by the occurrence of major lithospheric faults such as those described by Zihlmann et al. (2018) and Rospabé et al. (2019) in Oman. The timing of the hydrothermal alteration associated with the lithospheric faults is constrained by the occurrence of hydrated melts reported by Rospabé et al. (2019). This demonstrates that the ridge was magmatically active at the onset of hydrothermal alteration. Additional support for this model is the occurrence of anhydrite, typical of alteration in an oceanic environment, reported by Mariani et al. (2019) within a core (also drilled during Oman DP) crosscutting the fault studied by Zihlmann et al. (2018). Some of these lithospheric faults are present in the Batin region and located close to the drilling site. As mentioned in Section 2.2, the meter-scale fault sampled by the core BA1B may also be one of these faults. Moreover, it is likely that the location of the sampling site at the periphery of an off-axis diapir enhanced hydrothermal activity in the region (Jousselin and Nicolas, 2000). The Batin region was thus a predisposed zone for extensive hydrothermal cooling to occur.

Stages of serpentinization
The formation of mesh veins and replacement of mesh cores is often considered to happen in two stages. Mesh veins form first, in a rock-dominated environment and under relatively closed system conditions (e.g. Viti and Mellini, 1998). Yet, pervasive formation of mesh textures requires "pervasive water supply". Microcracking due to thermal contraction has been suggested as one possible contribution to pervasive permeability generation (Boudier et al., 2005). Considering that fracturing occurred first (step 1 in Fig. 9a), the circulation of cold seawater through the cracks could be the driving force for thermal cracking (step 2 in Fig. 9a). During this step, the volume increase associated with serpentinization is not compensated by transport of elements and reaction-induced fracturing (Plümper et al., 2012) is likely to occur (step 3 in Fig. 9a).
Replacement of olivine relics remaining in the mesh cores are believed to take place in a second step as the system progressively opens (Viti and Mellini, 1998). Preferred replacement of mesh cores in and around the early stage dark faults and fractures described here may reflect that they represent permeable pathways over a significant time interval also after the onset of initial mesh formation (step 4 in Fig. 9a). The formation of Liesegang-like oscillatory zoning pattern in the mesh cores is indicative of serpentinization under far from equilibrium conditions (Jamtveit and Hammer, 2012) and also reflect open-system behavior at this stage as indicated by the observed S and Al enrichments. A model emerges which includes an early stage of serpentinization with associated mesh texture formation. During this stage, the progress of serpentinization leads to clogging of the pore space (Hövelmann et al., 2012) and prevents water supply before the mesh cores are completely hydrated. The result of this stage is observed in sample DF-1 where relict of olivine can still be observed. Following this stage, the porosity is still connected to external fluid sources along the main faults and fractures. Sustained fluid flux then becomes more focused as less favorable pathways are abandoned, leading to metasomatic effects as observed in sample DF-2 where mesh cores in the fault zone are enriched in S and Al and depleted in Si compared to the mesh cores in the wall-rock (step 5 in Fig. 9a and Fig. 9b).

Conclusions
Our observations from the Samail ophiolite peridotites emphasize the close link between serpentinization, tectonic stress, and brittle deformation. We describe a family of highly localized and early faults that predates mesh texture formation. Macroscopic and microscopic structures, as well as EBSD observations are consistent with high energy brittle faulting. The faults act as major pathways for water infiltration in the rock even after pervasive fluid infiltration has ceased (i.e., after mesh formation). The formation of oscillatory zoning in replacement for olivine is symptomatic of far from equilibrium reaction conditions, possibly indicative of an irregular fluid supply to the system.
Bulk rock and SIMS analyses show low to moderate enrichment on fluid mobile elements including B, Li, Rb, and U, indicative of fluid-rock interaction in an oceanic environment characterized by relatively low fluid-rock ratios. This is consistent with a model where fluid-rock interactions took place below a thick magmatic crust. The oxygen isotope compositions of mesh serpentine are consistent with medium temperature serpentinization (200-250 • C) at ca. 50 km distance from the active fast spreading ridge. The Oman DP science team: Resources, Data Curation.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Früh-Green PI), JAMSTEC, the TAMU-JR Science Operator, and contributions from the Sultanate of Oman Ministry of Regional Municipalities and Water Resources, the Oman Public Authority of Mining, Sultan Qaboos University, CNRS-Univ. Montpellier, Columbia University of New York, and the University of Southampton. We thank Siri Simonsen and Muriel Erambert for their help with SEM and electronic microprobe data, Thanusha Naidoo, Ibrahim Khaled and Gunborg Bye Fjeld for their help with rock crushing and milling, and Martin Whitehouse and Heejin Jeon for their help and warm welcome into the NordSIM facilities. We thank C. Martin (Géosciences Montpellier) and L. Causse (AETE-ISO facility, OSU-OREME/Université de Montpellier) for their assistance during whole-rock trace element chemical preparation and analysis by quadrupole ICP-MS. We thank three anonymous reviewers for helpful reviews, and A. Webb for editorial handling.

Appendix A. Method for SEM and EBSD analyses
SEM pictures and EDS maps provided in this study have been obtained at the University of Oslo on a dual Bruker Quantax XFlash 30 EDS system with an accelerating voltage of 15 kV. Samples were carbon coated beforehand.
Electron backscatter diffraction (EBSD) orientation maps were acquired in the Scientific Center for Optical and Electron Microscopy (ScopeM), ETH Zürich. The thin section DF-2C was mechanically polished with diamond solutions up to a 0.25 microns grain size and final polishing was performed with an alkaline solution of colloidal silica for 3 minutes on a neoprene substrate. For the EBSD mapping, the sample was coated with 3 nm of C. The maps were acquired in a FEI Quanta 200F with EDAX EBSD/EDS system, using an accelerating voltage of 20 kV, beam current of 8 nA, working distance of 15 mm and step size of 1 micron. After the acquisition, the raw data was cleaned with the software OIM 85, by first performing confidence index (CI) standardization followed by neighbor CI correlation, assuming a minimum CI of 0.1. Afterwards, all the points with CI smaller than 0.1 and detected grain sizes smaller than 10 pixels were removed and not considered in the texture calculations. The crystallographic preferred orientation data was plotted in pole figures using the sample reference frame via the MatLab Toolbox MTEX (version 5.2.8; http:// mtex -toolbox .github .io; Bachmann et al., 2010Bachmann et al., , 2011Hielscher and Schaeben, 2008).

Appendix B. Methods for whole rock analyses
Samples were crushed and reduced to powders using the facilities of the Geosciences Department of the University of Oslo using an iron plate and a steel mill. Before the crushing, the pyroxene dyke in DF-1 was cut out of the sample.
Trace element concentrations (Li, Sc, Ti, V, Mn, Co, Ni, Cu, Zn, Ga, As, Rb, Sr, Y, Zr, Nb, Cd, Sn, Sb, Cs, Ba, Rare Earth Elements (REE), Hf, Ta, Pb, Th and U) were determined at Géosciences Montpellier (AETE-ISO, OSU OREME, Université de Montpellier, France) using an Agilent 7700X quadrupole ICP-MS. Powdered samples were prepared following the HF/HClO 4 procedure described in Ionov et al. (1992) and Godard et al. (2000). The samples were analyzed after a dilution of 1000. Element concentrations were measured by external calibration, except for Nb and Ta that were calibrated by using Zr and Hf, respectively, as internal standards. This technique is an adaptation to ICP-MS analysis of the surrogate calibration method described by Jochum et al. (1990); it aims at avoiding memory effects due to the introduction of concentrated Nb-Ta solutions in the instrument. The Helium cell gas mode of the Agilent 7700X was used to measure Sc, Ti, V, Mn, Co, Ni, Cu, Zn, Ga, As, Sn and Sb while removing polyatomic interferences. Each ICPMS measurement is an average of three runs and its precision is determined by its standard deviation. The uncertainty of analysis was estimated for each sample using the error propagation method described in Godard et al. (2000), which takes into account the precision of the measurements of (i) the instrumental blank, (ii) the procedural blanks and (iii) the sample analysis. Analyses (i) below the instrument detection limit, (ii) for which the contribution of the procedural blank is > 70% or (iii) having uncertainties >50 % were eliminated (noted "not determined"). We also analyzed rock reference materials DTS-2b and UB-N to assess the external precision and accuracy of our analyses. Our results show good agreement between measured values and expected values for the international standards, and reproducibility is generally better than 3% at concentrations > 10 ng/g, and 3-10 % for concentrations less than 10 ng/g. The limits of detection, the procedural blank contributions and the values obtained for rock standards during this study are reported in Supplementary Table S3.

Appendix C. Methods for ion probe analyses
Oxygen isotope composition ( 18 O/ 16 O, expressed as δ 18 O V-SMOW ) and B-concentrations of serpentine minerals were performed in situ on gold-coated samples from thin section DF-2C using a large geometry CAMECA IMS 1280 ion microprobe at the NordSIMS ion microprobe laboratory at the Swedish Natural History Museum, Stockholm. The spatial resolution for both measurements was similar at ca. 10-15 μm. For oxygen isotopes, the instrument was operated with a critically focused, 10 μm-rastered, ca. 2nA, Cs+ beam together with low energy electron flooding gun to counteract change build up, and simultaneous detection in two low-noise Faraday detectors, closely following the method reported for analyses of Zircon from the same laboratory (Whitehouse et al., 2017). Polished pieces of lizardite L3431 and antigorite Al06-44A reference materials (kindly provided by D. Rubatto) were co-mounted in the sample holder along with the thin section and used to correct for instrumental mass bias using the δ 18 O-values of 5.26 and 8.22 , respectively (Scicchitano et al., 2018). Prior to secondary ion mass spectrometry (SIMS) analysis, both materials were imaged unsing high contrast back-scattered electron (BSE) to reveal areas of alteration that were avoided during δ 18 O measurements. Matrix bias effects among different serpentine minerals (notably antigorite and lizardite) are relatively small during the same analytical session as the unknown analyses in this study, Al06-44A calibrated against L3441 yielded a 8.38 ± 0.18 (1s, n=14), within uncertainty of the expected value. The lizardite reference material was analyzed once every 3 to 5 samples unknowns and yielded a δ 18 O reproducibility of ± 0.20 (1s, n=32/34).
Boron analyses were performed using a critically focused, 5 μmrastered, ca. 5nA, O − 2 beam, with 11 B + / 28 Si 2+ ratios determined by peak jumping 11 B + and 28 Si 2+ with a mass resolution (M/ M) of ca. 2500 into a pulse counting electron multiplier. Two silicate glass reference materials were used for boron analyses. The primary reference, B6 obsidian (203.8 μg/g B, 75.5 wt% SiO 2 ; Gonfiantini et al., 2003), was analyzed at the beginning and at the end of the measurement session, yielding an uncertainty of ca. 2% (1s) in the 11 B/ 28 Si ratio. The secondary reference, GOR132-G (17.2 μg/g B, 45.5 wt% SiO 2 ; Jochum et al., 2006) was analyzed only at the end of the session and yielded a 11 B/ 28 Si ratio uncertainty of ca. 4% (1s). The estimate of boron concentration uses the equation below assuming ∼35.5 wt% SiO 2 in the targeted serpentine minerals based on the electronic microprobe data (Supplementary Table S1 Calculations using the B6 reference material yielded slightly too high values for B concentration in the GOR132-G reference (∼20.9 ppm). This is likely an effect of the difference in SiO 2 content between the two glasses as well as a crystallographic difference. Thus, since the GOR132-G glass has silica content closer to the serpentine sample, we renormalized to this material for final B concentration.