Abstract

Fossil structures, such as exhumed accretionary prisms, are the only direct recorders of the fluids wetting the plate interface near the base of the seismogenic zone. By studying exhumed accretionary prisms, it is thus possible to determine the physicochemical properties of fluids and the geometry and dynamics of their circulation. We considered here two transects encompassing the brittle-plastic transition (BPT) zone, in the Franco-Italian Alps and the Shimanto Belt in Japan, and compared our data with a broader set of examples from the literature. On quartz that grew synkinematically at peak burial conditions, we inferred fluid properties indirectly from quartz trace-element concentrations (using cathodoluminescence [CL] imaging) and directly from fluid-inclusion composition and P-ρ-T properties (using Raman and microthermometry). At ∼250 °C, quartz grew principally through fracturing and two types of quartz, a CL-brown and a CL-blue, precipitated alternately. At ∼350 °C, where plastic deformation and recrystallization is pervasive, only a single, homogeneously CL-brown quartz is present. The salinity of the fluid in the inclusions shallower than the BPT is consistently of the order or lower than seawater, while salinities are very scattered deeper than the BPT and often exceed seawater salinity. The gas dissolved in the fluid is predominantly CH4 shallower than the BPT, and either CH4 or CO2 deeper than the BPT, depending on the nature of the host rock and in particular on the proportion of carbonates. Cathodoluminescence properties, salinity, and nature of the gas all point to a closed-system behavior in rocks deeper than the BPT. In contrast, shallower than the BPT (i.e., at seismogenic depths), textures revealed by CL-imaging evidence the episodic influx of an external fluid, leading to the crystallization of CL-blue quartz. The scale of the circulation leading to the generation of the CL-blue quartz, or its relationship with the seismic cycle, is still unclear. Besides, the fluid pressure recorded in the abundant water-rich fluid inclusions is systematically much lower than the corresponding lithostatic pressure, irrespective of the depth domain considered. For inclusions trapped at large depth, the low fluid pressure recorded in the inclusions reflects post-entrapment reequilibration. For inclusions trapped at shallower conditions, typically at seismogenic depths, the low fluid pressure may as well be the result of large fluid pressure drop after earthquakes.

INTRODUCTION

Fluid Fluxes along the Plate Interface

The flux of fluid along the plate interface and in overlying accretionary prism has been extensively studied in the shallow domain (i.e., less than a few km depth) of subduction zones in the framework of deep-sea drilling campaigns. Based on direct observations or measurements of venting (e.g., Henry et al., 1992) or hydrological models (e.g., Bekins and Dreiss, 1992), the general scheme of shallow fluid circulation is that water dragged passively down in pores of buried sediments is expelled by compaction and flows upward along permeable horizons such as faults and a master décollement (Kastner et al., 1991; Le Pichon et al., 1993; Saffer and Tobin, 2011). This source of water is restricted to a few kilometers of depth, as the pore volume fraction diminishes along with compaction, and is replaced by water produced by dehydration reactions, involving for example clay minerals, abundant in buried sediments (Saffer and Tobin, 2011).

Models of “deep” fluid flow (i.e., deeper than ∼10 km depth) are much more unclear, because the relevant physical properties (permeability, porosity and fluid production) are not well constrained in modern margins. For example, physical properties are to a large extent derived from measurements of P-wave velocities (Vp), using relationships that are well calibrated for a large porosity (Erickson and Jarrard, 1998; Raimbourg et al., 2011; Kitajima and Saffer, 2012). Below 20%–30% porosity, the empirical relationships to convert Vp into porosity and then to relate porosity to permeability (Neuzil, 1994) are much more uncertain. Furthermore, as temperature increases, rock physical properties depend not only on porosity but also on chemical processes such as cementation or mineral reactions, in relationship with solid deformation (e.g., Tsuji et al., 2006).

Models of deep fluid flow rely therefore mostly on geochemical studies in fossil subduction zones, focusing on isotopic reequilibration or elemental transfers in metamorphic rocks. These studies help discriminate whether the chemical exchanges are restricted to compositionally homogeneous domains (e.g., a tectonic slice of metapelites or serpentines) or whether chemical exchanges occur over larger scales, across different domains. The different units from the Franciscan Complex span the T range 100–350 °C, corresponding to the seismogenic zone (Sadofsky and Bebout, 2001). Most carbonate veins formed in these units are host-rock buffered in terms of oxygen and carbon isotopes, with in some cases the contribution of the advection of an external fluid with a higher δ18O (Sadofsky and Bebout, 2004). Such local control is in agreement with the δ18O in the quartz veins from the Shimanto Belt (Lewis and Byrne, 2003; Raimbourg et al., 2015), irrespective of the CL color of the quartz, or from the Alps (Mullis et al., 1994). In Kodiak, similar host-rock buffering of quartz in veins is described in Vrolijk (1987a, 1987b) but for quartz veins in tectonic mélanges, where δ18O is higher than in veins from coherent units. Another contrasting example is provided by the Otago Schists of New Zealand, where the δ18O signature of quartz veins points to the reequilibration with the pore-fluid reservoir rather than host-rock buffering (Fagereng and Harris, 2014). Therefore, in terms of isotopes, the general regime is a buffering of fluid and vein isotopic composition by host rock, even if several cases point to a mixing with a fluid external to the system, possibly flowing from the depth and enriched in heavier 18O.

At larger depths, metasediments in the Franciscan Complex or the Schistes Lustrés in the Alps, buried down respectively to ∼30 km and ∼45 km, mostly retain their concentrations in fluid-mobile elements, attesting to closed-system behavior (Busigny et al., 2003; Sadofsky and Bebout, 2003; Scambelluri and Philippot, 2004; Bebout, 2007; Bebout et al., 2013). Similarly, serpentines buried in the Alps retain their mobile elements and δ18O for T lower than the antigorite break down (∼600–700 °C) (Scambelluri and Philippot, 2004; Scambelluri et al., 2004). These two cases correspond to low-T gradients prevailing in most subduction zones, so that limited mobility of elements seems to be the general case in material subducted to high-pressure conditions, even if more open-system behavior may occur under higher geothermal gradients (e.g., the Catalina Schists in California; Bebout, 2007).

In contrast to this general picture of closed-system behavior given by geochemical analyses, there is a systematic decrease in the water content of subducted rocks as a result of progressive metamorphism. Because we consider the release of water at depths larger than a few km, we focus hereafter on water bound to minerals rather than water in rock pores. Incoming material, composed of sediments, extrusives forming the uppermost horizon of the oceanic crust, dikes, and gabbros carry on average 7 wt% (Plank and Langmuir, 1998), ∼3–6 wt%, ∼2 wt%, and 1 wt%, respectively (for a 20–80 Ma oceanic lithosphere; Jarrard, 2003) of mineral-bound H2O, although these quantities may vary significantly depending, for example, on the nature of subducting sediments or the slab hydration through faulting and influx of water at the outer rise (Ranero et al., 2003). The oceanic crust is the largest H2O reservoir to be subducted (Bebout, 1995; van Keken et al., 2011). Dehydration reactions occur as a result of prograde metamorphism, at different depths depending on the layer considered. In sediments, dehydration reactions start at low temperature, of the order of 50 °C, with reactions such as opal → quartz or smectite → illite (Moore and Saffer, 2001; Saffer and Tobin, 2011). The latter reaction is the principal source of water, and it is completed at T of ∼150 °C. As a result, the water left in metasediments of the Santa Catalina Schists and Franciscan Complex, buried to conditions below 300 °C and 1 GPa, has already decreased to ∼3 wt% (Sadofsky and Bebout, 2003). In the crust, dehydration reactions are deeper and mostly localized at ∼80 km depth, except in the warmest slabs, where they can occur at shallower depths (Van Keken et al., 2011). As a result, by ∼80 km depth, a large fraction of the water initially present has escaped the crust and the sediments of the slab.

There is therefore a decoupling between the large volumes of fluid flowing from and through the subducting slab and the limited transport of elements with the fluid. For this reason, many geochemical tracers seem unable to track the fluid circulation and its geometry, as for example oxygen isotopes that are host-rock buffered while microstructures point to a complex interplay of two fluids during vein precipitation (Raimbourg et al., 2015). Only domains of focused fluid flow, such as the highly sheared rocks of the subduction channel (e.g., Vannucchi et al., 2010), may be able to record isotopically the flow of exotic fluid without complete host-rock buffer.

Additional information on fluid flow is provided by microstructures, in particular veins, which are witnesses of fluid circulation at depth. In fossil subduction zones, the large concentration of veins in rocks interpreted as belonging to a paleoplate interface provides evidence of vigorous and localized fluid circulations (Byrne and Fisher, 1990; Vannucchi et al., 2010; Yamaguchi et al., 2012). The interpretation of such geological structures in terms of fluid circulation is nevertheless difficult. Certain models plead in favor of large-scale advective flow (Ferry and Dipple, 1991; Breeding and Ague, 2002), but other evidence, such as the ubiquitous and pervasive occurrence of dissolution microstructures in accretionary prisms (Kawabata et al., 2007; Raimbourg et al., 2009), or mass-balance calculations on the required amount of fluid flow to fill the veins with quartz (Fisher and Brantley, 1992; Fisher et al., 1995), point rather to local sources for vein-forming elements. Furthermore, the possible link between fluid flow and deformation, in particular earthquakes, suggested by crack-seal textures in veins (Fisher and Brantley, 1992, 2014; Fagereng et al., 2011), remains hypothetical.

Fluid Pressure along the Plate Interface

Associated with the question of fluid circulation, fluid pressure is another key issue in subduction zones. Because it is potentially a major control on plate boundary strength and seismic and/or aseismic character, many studies have focused on the amplitude of the overpressures of fluid in and around the plate interface (defined as the excess with respect to the hydrostatic fluid pressure gradient). At depths of a few km, seismic reflection profiles (Shipley et al., 1990; Tobin and Saffer, 2009) or hydrological models (Shi and Wang, 1985; Saffer and Bekins, 1998) suggest, in many subduction margins, the existence of near-lithostatic fluid pressure in underthrusted sediments and in the overlying décollement (see review by Saffer and Tobin, 2011), resulting from the compaction of high-porosity sediments. At larger depths, down to typically ∼30–40 km, overall large overpressures of fluid within the upper part of the subducting slab are inferred from high Vp/Vs ratio in several subduction zones such as New Zealand (Eberhart-Phillips et al., 2005), Costa Rica (Audet and Schwartz, 2013), Chile (Moreno et al., 2014), or southwestern Japan (Kato et al., 2010). In addition, some of these seismological studies (Audet and Schwartz, 2013; Moreno et al., 2014) indicate a correspondence between the amplitude of overpressure and the seismic-aseismic character of the interface, suggesting some causal relationship.

The geological record corresponding to these geophysical observations of high fluid pressure in modern margins, consists first of the abundance of extension veins in exhumed mélanges interpreted as paleoplate interface zones, such as in the Apennines (Meneghini et al., 2007; Vannucchi et al., 2010), Alaska (Fisher and Brantley, 2014), or the Franciscan Complex (Meneghini and Moore, 2007). Mode I fracture requires fluid pressure to be larger than σ3, hence close to lithostatic pressure, because differential stress is presumably small in subduction zones (Lamb, 2006; Hasegawa et al., 2012; Duarte et al., 2015). These extension veins confirm therefore the average high pore-fluid pressure along the plate interface but provide limited insights into the space and time distribution of such a pressure field.

The study of fluid inclusions is another potentially powerful tool to estimate the fluid pressure at depth and to derive the amplitude of fluid overpressure from its comparison with rock pressure. This objective faces two distinct obstacles in the shallow and large depth domain. At shallow depth, where low temperature (T) prevails (T<∼300 °C), corresponding to the seismogenic zone, there is hardly any reliable geobarometer to derive rock pressure, hence the record of fluid inclusion is in most cases the only available information about pressure. A good example is the Shimanto Belt, where a wealth of fluid-inclusion studies is available to derive the fluid pressure (Sakaguchi, 1996; Sakaguchi, 1999b; Lewis and Byrne, 2003; Matsumura et al., 2003; Okamoto et al., 2014) but where the corresponding rock peak pressure conditions are not well constrained (Underwood et al., 1993), preventing an estimate of the degree of fluid overpressuring. At large pressure, for example in the blueschist and eclogite-facies fields, the fluid inclusions record systematically low fluid pressure with respect to the pressure conditions derived from mineral assemblages (Küster and Stöckert, 1997; Agard et al., 2000). The pressure record in the fluid inclusions is interpreted as reflecting not the actual pressure during entrapment, but posterior reequilibration at lower pressure during uplift. In summary, both at low and high pressure, there is a limited geological record of concomitant fluid and rock pressure, to be compared with geophysical observations.

Still, fluid inclusions may provide useful insights into the time variations in fluid pressure, despite limitation with regard to the assessment of the absolute value of fluid overpressure. Variations in the density of methane trapped in fluid inclusions from synkinematic quartz veins were reported in the Kodiak accretionary complex in Alaska (Vrolijk, 1987a, 1987b) and in the Shimanto Belt (Raimbourg et al., 2017) and were interpreted as reflecting cycles of fluid pressure. In this interpretation, the cyclic drop in fluid pressure results from episodes of fracturing and connection of a fault network, enabling the fluid to escape to shallower domains. It is nevertheless not clear whether the triggering of fracturing is due to the fluid itself (through a fluid pressure rise resulting from tectonic loading or chemical reactions) or results from damage during large earthquakes.

The objective of this work is to provide examples of the geological record of the composition and pressure of the deep fluid, as well as its flow patterns, using quartz veins and the fluid inclusions they contain. The two case studies, in the Alps and in the Shimanto Belt in Japan, encompass the depth range that includes the transition from macroscopically brittle behavior to deformation controlled by crystal plasticity (BPT), which is often correlated to the transition from seismic to aseismic behavior of the plate interface (Hyndman et al., 1997; Oleskevich et al., 1999). We have focused on synkinematic quartz veins and studied the evolution with burial depth and temperature of (1) microstructures and vein generations (with optical microscopy and cathodoluminescence) and (2) nature and pressure of the fluid trapped in inclusions (with Raman spectroscopy and microthermometry). These observations reveal a sharp contrast in fluid circulation between the seismogenic plate interface and the deeper aseismic domain. In addition, a compilation of our data with the literature shows that the record of fluid pressure in fluid inclusions is systematically lower than the corresponding conditions of lithostatic pressure along the subduction plane.

CHOICE OF THE SAMPLES AND METHODS OF ANALYSIS

Sampling Areas

To study the evolution with depth of the fluid along the plate interface, we have already studied in detail the example of the Shimanto Belt in Japan (Raimbourg et al., 2014b; Raimbourg et al., 2015; Palazzin et al., 2016), and we describe hereafter the principal results (Fig. 1). The Western Alps provide another well-studied example of subducted sediments (Fig. 2). In both settings, we have selected quartz veins belonging to two units with a contrast in metamorphic temperature, one “high-T” (∼350 °C) unit and one “low-T” unit (∼250 °C) (as defined by Raman spectrum of organic matter, see below), so as to study the downdip limit of the seismogenic zone and the transition from brittle to plastic deformation of quartz. The position of all samples and outcrops described hereafter can be found in Supplemental Table S11.

Shimanto Samples

The Shimanto accretionary complex, exposed on-land along the Honshu, Kyushu, and Shikoku islands in Japan (Fig. 1), is interpreted as an ancient accretionary prism (Taira et al., 1980; Taira et al., 1988). The whole complex, trending parallel to the modern trench axis of the Nankai Trough, is composed of a stack of units with a general dip toward the N or NW, with a younging trend toward the SE. These units are constituted of either coherent sedimentary units or tectonic mélanges.

Our study focuses on the northern portion of the Belt on Kyushu, namely the basal layer of the Morotsuka Group, i.e., the Foliated Morotsuka (FM) (Raimbourg et al., 2014a) and the upper part of the Hyuga Group, known as the Hyuga Tectonic Mélange (HTM). The two units are juxtaposed by the Nobeoka Tectonic Line (NTL), a large-scale, low-dipping thrust fault (Saito et al., 1996; Murata, 1997).

The HTM constitutes the best example of plate interface deformation: it consists of strongly sheared sandstone and quartz boudins in a pelitic matrix, with a direction of movement to the SE consistent with the coeval subduction of the Pacific plate below Japan (Faccenna et al., 2012). Synkinematic quartz veins formed within rigid blocks, such as sandstones, as extension veins perpendicular to lineation, or as shear veins at low angle to the foliation (Fig. 3A; see also Palazzin et al., 2016). The temperature of deformation is ∼245 ± 30 °C, as can be determined from Raman spectrum of organic matter (Palazzin et al., 2016), and the quartz microstructures recorded mostly pressure solution, with a minor contribution of plastic deformation.

The FM formation, of pelitic composition, constitutes the “high-T” end member of this study, as metamorphic T is ∼345 ± 30 °C (Palazzin et al., 2016). Kinematics of the deformation are mostly subhorizontal flattening, resulting in folding and transposing of earlier-stage quartz veins (Fig. 3C). Quartz in veins is mostly recrystallized and develops accordingly a crystallographic preferred orientation, with a concentration of quartz c-axes along Z kinematic direction (Palazzin et al., 2016). In tectonic reconstructions of the margin evolution, this deformation event occurred in the vicinity of the plate interface as a result of the collapse of the accretionary prism in the Paleogene (Raimbourg et al., 2014a).

Alps Samples

The nappe du Parpaillon, within the Flysch à Helminthoïdes unit (FH), is constituted of ∼1000-m-thick series of calcareous turbidites overlying a level of dark pelites containing thin sandstone beds (“Complexe de Base” [CB]) (Kerckhove, 1963; Kerckhove et al., 2005). Rare fossils give an upper Cretaceous age (Kerckhove, 1963), contemporaneous with the upper levels of the Schistes Lustrés (SL). Similarly to the SL, the FH nappe is interpreted as associated to the Liguro-Piemontese oceanic domain (Kerckhove, 1963; Caron et al., 1981). This paleogeographic position involves large displacement of the Parpaillon Nappe upon emplacement, and the CB constitutes the associated niveau de décollement (Kerckhove et al., 2005). The deformation of the Nappe comprises two phases of folding (Merle and Brun, 1981; Merle, 1982) achieved at low-grade conditions, at most greenschist-facies conditions (Bousquet et al., 2012). The early stage of deformation is recorded by isoclinal folding, with an axis orientated NE-SW and transport direction toward the NW (Merle and Brun, 1981; Merle, 1982). In the CB, this deformation is apparent through the isoclinal folding of thin turbidites beds, which develop a schistosity as axial planes of folds. These folded beds are in addition cut by numerous quartz veins, oriented approximately perpendicular to bedding. These veins constitute an early-stage deformation feature as they are refolded by isoclinal folds (Fig. 3B). All the early-stage deformation features recorded in the CB (isoclinal folding, extension veins, and schistosity) may be interpreted as associated with the displacement along the basal décollement and the layer-parallel shearing, the régime that prevails in underplated units in accretionary prisms (Byrne and Fisher, 1990; Vannucchi et al., 2010; Fisher and Brantley, 2014). So even if the evolution of the FH during the subduction of the Liguro-Piemontese Ocean is relatively ill constrained, we tentatively consider its basal unit and the early deformation it records as an analogue of underplated units in subduction complexes.

The Schistes Lustrés Complex is a large portion of the cover of the Ligurian oceanic basement, including ophiolitic slivers of various sizes (Lemoine et al., 1986; Michard et al., 1996; Stampfli and Marchant, 1997) and subducted down to blueschist- and eclogite-facies conditions during the Alpine orogenesis (Caby et al., 1978; Saliot et al., 1980; Goffé and Chopin, 1986; Agard et al., 2000). In the studied area (Fig. 2), the SL are composed of metapelites affected by increasing metamorphic conditions from west to east (Chopin and Schreyer, 1983; Beyssac et al., 2002). In the westernmost area, the metapelites contain the HP index mineral magnesio-carpholite, principally present with quartz in veins (Fig. 3D), formed at peak metamorphic conditions of ∼350–380 °C and ∼15 kbars (Agard et al., 2000). The close association, with quartz veins, of tiny needles of carpholite and primary fluid inclusions, was taken as evidence of the formation of these fluid inclusions at peak conditions, although they suffered some reequilibration during retrograde evolution (Agard et al., 2000). The kinematics of the tectonic stage associated with the carpholite-bearing veins (D1 stage by Agard et al., 2001) is unclear, as it is in most places obscured by top-to-the-east D2 stage related to exhumation.

Comparison of the Selected Samples

The four cases studied are meant as representative of deformation of sediments along or near the subduction plane at different depths and temperatures. There are nevertheless large contrasts in two different respects between the Alpine and the Japanese example.

The first difference is the metamorphic gradient during burial. Conditions for the Shimanto Belt, derived in the highest-grade units, are of the order of 3–5 kbars and 300–350 °C (Toriumi and Teruya, 1988; Palazzin et al., 2016), i.e., 15–30 °C/km, and hence belong to a relatively high metamorphic gradient. In contrast, the SL metamorphism is associated with a very cold gradient, because roughly similar temperatures but much higher pressures are reported for the western domain of the SL (∼350–380 °C and ∼15 kbars [Agard et al., 2000]; i.e., ∼6 °C/km).

The second difference is the abundance of carbonates. The Alpine units (SL and FH) contain a large proportion of carbonates (Kerckhove, 1969; Caron, 1977). In contrast, the sedimentary units constituting the Shimanto Belt contain a relatively low abundance of carbonates (Taira et al., 1988). This general feature applies to the Hyuga mélange, where recent X-ray diffraction (XRD) semiquantitative analyses of lithology show proportion of calcite less than 10% and in most cases less than a few % (Fukuchi et al., 2014).

Analytical Techniques

Samples similar to Figure 3 were processed into thin sections (for optical microscopy, Raman spectroscopy, cathodoluminescence, and electronic and ion microprobe) and thick sections (for microthermometry on fluid inclusions). Thin sections are ∼30 µm thick and diamond polished on their upper surface. Thick sections are ∼100–200 µm thick and diamond polished on both surfaces. Thin and thick sections were all cut perpendicular to the foliation and, whenever observed, parallel to the stretching lineation.

Raman Spectra of Organic Matter

Raman spectroscopy of carbonaceous material (RSCM) thermometry is based on the quantitative estimate of the degree of structural transformation of carbonaceous material (CM) (Beyssac et al., 2002; Lahfid et al., 2010). Because of the irreversible character of graphitization, Raman spectrum of CM can be correlated with peak-T conditions (noted Tmax hereafter).

Raman spectra were obtained using a Renishaw InVIA Reflex microspectrometer (Institut des Sciences de la Terre d’Orléans [ISTO]–Bureau de Recherches Géologiques et Minières [BRGM]). A laser (514 nm) was focused on the sample by a DM2500 Leica microscope equipped with a ×100 objective. Instrument control and Raman measurements were performed with the software package Renishaw Wire 4.0. Acquisition time was generally over 30 s, and spectra were accumulated once or twice. To avoid defects on the CM related to thin-section preparation, analyses were all performed below the surface of the section by focusing the laser beam beneath a transparent crystal (i.e., dominantly quartz and occasionally calcite or albite). The laser beam power at sample surface was set to ∼0.5 mW.

To interpret Raman spectra, we used the procedure of Lahfid et al. (2010) adapted to the low-grade metamorphic rocks considered here. This decomposition of spectra is based on five peaks, forming a defect band centered ∼1350 cm–1 (peaks D1, D3, and D4) and a graphite band centered ∼1580–1600 cm–1 (peaks D2 and G). Tmax can then be estimated, with an uncertainty of ±25 °C, as RA1 = 0.0008 Tmax + 0.3758, where RA1 is the ratio of the peak areas defined as RA1 = (D1 + D4) ⁄(D1 + D2 + D3 + D4 + G). Note that the thermometer is calibrated for the range 200–320 °C, but for the sake of discussion, we used the correlation RA1 –Tmax for all samples where the decomposition in five peaks appropriately described the whole Raman spectrum, i.e., in the range 170–370 °C (see Table 1). To allow the comparison of RSCM analyses between the high-grade units of the Shimanto Belt and the SL, which were already analyzed by Beyssac et al. (2002), we also carried out the decomposition of Raman spectra in three peaks, calibrated for the temperature range 330–650 °C by the authors.

Cathodoluminescence on Quartz

Samples were placed in a vacuum chamber at an argon pressure of ∼60 × 10–3 mbar. They were irradiated by a cold cathode source electron gun provided by Optique Electronique OPEA (France). Standard tension-current beam conditions were maintained at ∼13–15 kV and 90–110 µA, respectively. The angle between the gun axis and horizontal was 18°. The chamber was placed on a simplified optical mount designed to minimize light absorption on the beam pathway. The optical stage was equipped with a Retiga 2000R 1394a cooled camera (captor size 1 inch, resolution 1600 × 1200 pixel2, square pixel size 7.4 µm). The luminescence of quartz being quite low, acquisition time was fixed to ∼10 s. Whenever present, the strong luminescence of carbonate veins prevented the analysis of quartz.

Raman Spectra of Fluid Inclusions

Raman spectroscopy was carried out using a WITec Alpha500 RA system with a green laser (λ = 532 nm, Nd:YAG frequency doubled laser) at the Centre de Biophysique Moléculaire, Orléans, France. The laser is connected to the system via an optical fiber and is focused on the surface using an optical microscope ×100 objective Nikon E Plan. In this configuration, the laser spot diameter is ∼720 nm. The laser power was set at 8 mW at the sample surface. Finally, two different gratings were used in order to obtain a spectral range of ∼4000 cm–1 with a resolution of ∼3 cm–1 or a spectral range of ∼1300 cm–1 with a resolution of ∼1 cm–1.

Microthermometry

Apparatus and uncertainties. Fluid-inclusion microthermometric measurements were carried out using the THMS-600 Linkam heating-cooling stage at ISTO (Orléans, France). Further details on apparatus and measurement conditions can be found in Raimbourg et al. (2014b). The temperature accuracy for these measurements is ∼±1 °C, over the whole temperature range of investigation, from –120 to +300 °C. The range of temperatures ∼0 °C, relevant for ice-melting temperature of aqueous inclusions, was calibrated by using synthetic inclusions of pure water (freezing point at 0 °C) and water + CO2 inclusions (melting point of CO2 at –56.6 °C), according to the procedure detailed in El Mekki-Azouzi (2010). Measurements of ice-melting temperature, repeated over several sessions, on the same reference sample containing synthetic, pure-water inclusions, yielded differences of the order of 0.2 °C. In addition, measurements on synthetic, pure-water inclusions, over different reference quartz samples of various thicknesses (200–600 µm) and shapes, yielded differences as large as 0.6 °C between the samples. Because the rock slabs containing our natural fluid inclusions have various shapes and thicknesses, we estimate therefore the uncertainty of freezing stage measurement of ice-melting temperatures as ∼0.8 °C.

Choice of the fluid inclusions. Fluid inclusions have a variable abundance, from rare and scattered in optically clear quartz grains to a dense network of fluid inclusions in cloudy quartz. Fluid inclusions typically range from a few up to 5 µm in diameter and they are two-phased at ambient temperature. Shapes range from elliptical to very irregular. We focused here on primary fluid inclusions, i.e., considering fluid inclusions with (1) a regular shape and (2) either isolated in the quartz crystal or at least not connected, spatially or in terms of elongation axis, with trails of secondary fluid inclusions. In addition, we focused on inclusions distributed only in domains of homogeneous CL color, to enable a clear correspondence between fluid properties and CL colors.

For the chosen fluid inclusions, we could observe and measure ice-melting temperature Tmi (where the last crystal of ice melts into the liquid), Tmcl (where the last crystal of gas hydrate melts into the liquid), and homogenization temperature Th (where the bubble of gas vanishes by dissolution into the liquid). Temperatures Th were easily assessed, because gas bubbles are agitated by Brownian motion. In contrast, the crystals of ice or gas hydrate could not be directly observed. Final melting of these crystals can only be inferred from a movement of the gas bubbles. In most fluid inclusions, when the last crystal melted, the gas bubble either (1) moved to a new position of equilibrium in the inclusion (and was then immobile upon temperature increase) or (2) appeared in the inclusion, since it was not visible at lower T.

RESULTS

Quartz Deformation Microstructures

The comparison of HTM and FM provides a good illustration of the onset of quartz plastic deformation as a result of T increase (Fig. 4). In HTM, quartz veins parallel to schistosity are mostly composed of large quartz grains slightly (Figs. 4A and 4B) to strongly (Fig. 4C) elongated parallel to the lineation. These megacrystals of quartz contain systematically a very large number of fluid inclusions. Some of these fluid inclusions align in planes perpendicular to the stretching direction, corresponding to healed extension microfractures (black triangle in Fig. 4C). Incipient recrystallization is apparent in fine-grained domains on the rims of megacrystals, with a large decrease in fluid-inclusion abundance (white triangle in Figs. 4B and 4C). In FM, recrystallization is pervasive: large inherited megacrystals, composed of patches of misoriented subgrains (Fig. 4D; D′ and D″) alternate with fully recrystallized domains. Along with extensive recrystallization, the average fluid-inclusion abundance is much decreased. Additional information about deformation process is provided by electron backscatter diffraction (EBSD) analysis in Palazzin et al. (2016). Quartz crystallographic preferred orientation (CPO) in HTM, defined by a X-maximum, changes to a Z-maximum in FM. We interpreted this difference in CPO as the transition of fracturing and pressure-solution to plastic deformation–dominated deformation. The evolution from HTM to FM microstructures corresponds therefore to the BPT.

Maximum Temperature of Deformation (by RSCM)

Paleotemperatures of the low-grade units HTM and FH, inferred from RSCM using Lahfid et al. (2010) procedure, are ∼240 and ∼270 °C ± 30 °C, respectively (Table 1). The estimated temperature of HTM is consistent with previous results based on vitrinite reflectance (Kondo et al., 2005) and illite crystallinity (Fukuchi et al., 2014). The analysis of the high-grade units, which stand on the edge of both procedures for low- (Lahfid et al., 2010) and high-grade (Beyssac et al., 2002) material, can be carried out using both methods, yielding temperatures that differ only by ∼20 °C. Using the Beyssac et al. (2002) procedure gives paleotemperatures of ∼330 and ∼350 °C for the FM and the SL, respectively. There is therefore in both Japanese and Alpine cases a temperature gap of ∼80 °C between high-grade and low-grade units, which is readily apparent on the general shape of organic matter spectra (Fig. 5).

Quartz Generations (by Quartz Cathodoluminescence)

In low-grade units, quartz domains are composed of two distinct materials (Figs. 6A–6C): (1) CL-brown quartz, similar to the one present in high-grade domains (Figs. 6D and 6E) and (2) quartz characterized by a bright blue luminescence that vanishes upon exposition to the electron beam. The distribution of the two types of quartz, revealed by CL imaging, can be interpreted as fracture filling (Figs. 6A–6C) and growth rims (Fig. 6B). As the two reciprocal crosscutting (CL brown and blue crosscutting CL blue and brown) or growth (CL brown and blue overgrown on CL blue and brown) relationships can be observed, the two types of quartz precipitated alternately (see also Raimbourg et al., 2015). In general, CL-blue quartz forms larger veins (Fig. 6A), while CL-brown quartz fills much thinner veins (black triangles Fig. 6B), giving a “brecciated” aspect to domains initially formed of homogeneously CL-blue quartz (Fig. 6B). In contrast, in high-grade units, quartz is homogeneously constituted of a CL-brown material (Figs. 6D and 6E).

Composition of the Gaseous Phase of Fluid Inclusions (by Raman)

Fluid inclusions present in the four units studied are two-phased at ambient T, with a bubble of gas embedded in liquid water. In HTM, prior Raman analyses have shown that the gas bubble is composed of methane (Raimbourg et al., 2014b). In the higher-grade FM, most of the quartz veins are plastically deformed and recrystallized (Palazzin et al., 2016) and contain a contrasted distribution of fluid inclusions. Inherited grains contain a relatively large amount of fluid inclusions, while recrystallized and fine-grained domains contain a much lower abundance of fluid inclusions, which are in addition of a very small size, hence difficult to analyze. Irrespective of these microstructures, the gaseous bubble in all fluid inclusions is composed of methane (Fig. 7).

In the Alps, the fluid inclusions present with an extremely high abundance in the FH are composed of gaseous methane and water (Fig. 7). In contrast, the large quartz veins from the SL incorporating carpholite needles contain much fewer fluid inclusions. The gaseous bubble is composed there of CO2, with no trace of methane.

In summary, the two low-grade units contain fluid inclusions made of a mixture of liquid water and gaseous methane. The high-grade units have a different gaseous phase in fluid inclusions. If methane fills the bubble in the FM in the Shimanto Belt, the gaseous phase in fluid inclusions from the SL is composed of CO2.

Fluid Physicochemical Properties (by Microthermometry on Fluid Inclusions)

In the Flysch à Helminthoïdes, the three microthermometric properties we could systematically observe—namely, last melting of ice, last melting of clathrate, and homogenization—are homogeneous across the four samples we analyzed (Fig. 8): Tmi ranges principally between –0.8 and –2 °C, Tmcl ranges between 1 and 6 °C, and Th clusters between 180 and 210 °C (Table 2).

There is no systematic variation related to the CL color of the quartz hosting the fluid inclusions but for Tmi. In one sample (FH2), fluid inclusions in CL-blue quartz have a lower Tmi, hence a larger salinity, than those in CL-brown quartz; while in two other samples (FH22 and FH114A), there is no difference between the two types of quartz. In the fourth sample (FH21b), the number of data in the CL-brown quartz is too small for any conclusion to be drawn.

A synthetic, upper P-T isochore of these inclusions can be calculated using –2 °C, 5 °C, and 180 °C, as Tmi, Tmcl, and Th, respectively, in the procedure in Raimbourg et al. (2014b). The associated concentration of methane is 0.34 mol CH4/kg H2O (Fig. 9). The intersection of this isochore with the rock paleotemperature range by Raman occurs at ∼2 kbars.

DISCUSSION

Timing of Formation of the Fluid Inclusions

The timing of entrapment of the fluid in inclusions, with respect to host-rock evolution, is difficult to assess. In FH, we analyzed primary fluid inclusions whose entrapment is contemporaneous with host quartz growth, while in HTM (Raimbourg et al., 2015), we considered also secondary fluid inclusions when they could be unambiguously associated with fracture-filling quartz of a homogeneous CL color.

In the HTM case, both veins and microfractures filled by trails of fluid inclusions are systematically oriented perpendicular to the extension direction (Palazzin et al., 2016, their figures 6–8), and the quartz crystallographic fabrics itself is interpreted as reflecting growth by repeated events of fracturing and/or sealing. Fluid inclusions formed therefore principally during the main deformation event, which corresponds to peak burial conditions (Raimbourg et al., 2014a).

In the FH case, deformation kinematics during burial are obscured by later-stage deformation related to Alpine collision (Merle and Brun, 1981; Merle, 1982). One constraint on quartz veins formation is that they are an early-stage deformation feature, which predates folding, possibly during nappe transport. In addition, the geometry of the veins, i.e., perpendicular to bedding and concentrated in sandstone beds, is very similar to vein geometry in the Shimanto Belt (Raimbourg et al., 2015) or in Kodiak accretionary prism (Fisher and Byrne, 1987; Byrne and Fisher, 1990). In this last example, penetrative extension fracturing of sandstone beds in turbidite units is operative throughout burial. On the basis of deformation analysis, the record of fluid inclusions in both HTM and FH can be associated with burial or peak metamorphic conditions.

Another constraint is provided by the microthermometric record of the fluid inclusions themselves. In FH, homogenization temperatures are concentrated ∼200 °C, which is slightly lower than the host rock peak temperature of 270 ± 30 °C given by RSCM. In HTM, homogenization temperatures range from 170 to 280 °C, which is also slightly lower or of the order of the host-rock peak temperature of 245 ± 30 °C given by RSCM (Palazzin et al., 2016). In both cases, homogenization temperatures are slightly lower or overlap peak rock temperatures. Trapping temperatures of the fluid are equal to or higher than homogenization temperatures, which even reduces the difference between fluid trapping temperature conditions and rock peak temperature.

From structural and microthermometric results, we consider therefore that fluid inclusions in quartz veins from both FH and HTM units formed during the latest stages of burial or even at peak burial and not after. Their record, in terms of fluid P-T conditions, corresponds therefore to the deepest stage of the rock evolution, provided that inclusions were not affected by reequilibration processes during exhumation (see next section).

Fluid Pressure along the Plate Interface—The Record of Fluid Inclusions

The geological record of near-lithostatic fluid pressure is potentially preserved in fluid inclusions trapped at depth and exhumed along with rocks belonging to paleoplate interfaces. Such a record relies on the assumption that inclusions follow an isochoric path since trapping at high pressure, hence do not reequilibrate. Driving forces for reequilibration (meant as a change in fluid density in the inclusion) are the difference in (1) pressure and (2) water chemical potential between the inclusion and the surrounding crystal (Bakker and Hansen, 1994; Sterner et al., 1995; Bodnar, 2003).

Reequilibration of Deep Fluid Inclusions

Experimental studies have shown that the assumption of isochoric path is in many cases not valid and that fluid inclusions reequilibrate under both conditions of internal over- and under-pressure (Sterner and Bodnar, 1989; Hall and Sterner, 1993; Vityk and Bodnar, 1995a, 1995b). Several processes account for reequilibration, such as plastic deformation of the inclusion wall (Vityk et al., 2000), leakage through dislocations (Bakker and Hansen, 1994), or fracturing ( = decrepitation) of surrounding crystal (Vityk and Bodnar, 1995b).

Indeed, in high-pressure rocks, reequilibration of fluid inclusions during uplift accounts well for the large discrepancy between rock peak P-T conditions in blueschist- or eclogite-facies and the much lower pressure in fluid inclusions trapped concomitantly (Küster and Stöckert, 1997; Agard et al., 2000). The primary reason for such reequilibration can be illustrated in the P-T diagram showing the evolution of fluid-inclusion pressure and rock pressure (Fig. 10), for primary fluid inclusions from the SL trapped concomitantly with high-pressure carpholite (Agard et al., 2000; this study).

Fluid-inclusion P-T evolution is given by the isochores, which we calculated using equation of state of pure water (Wagner and Pruss, 2002). The actual fluid in the inclusion contains not only water but also a small proportion of gas (CO2; see Fig. 7), which should be accounted for in calculating the isochores. Nevertheless, this gaseous phase is much more compressible than water (Duan et al., 1992), which decreases the fluid P-T slope in Figure 10, so that pure-water isochore is an upper bound on actual fluid isochore.

Considering a fluid trapped at peak P-T conditions of the SL (Agard et al., 2000), its isochoric evolution during exhumation follows a slope of ∼3 × 10–2 kbar/°C (or equivalently ∼12 °C/km; red line in Fig. 10). This contrasts with typical upward-concave rock exhumation P-T path, whose first stage follows a much larger P-T slope (blue line in Fig. 10). As a result, during exhumation, there is a divergence between rock and fluid pressure, and internal overpressure develops in the inclusion. The amplitude of overpressure is of a few kbars, that is of the order or larger than what was necessary to trigger reequilibration, in high P and T experiments on Brazilian quartz containing synthetic fluid inclusions (Hall and Sterner, 1993; Vityk and Bodnar, 1995b). In general, it is therefore unlikely that fluid conditions at high-P are preserved.

Reequilibration of Shallow Fluid Inclusions

The difference in pressure between the fluid inclusions and the surrounding solid medium is the main parameter controlling fluid-inclusion reequilibration. Following this idea, the best preservation of fluid inclusions (hence of the record of their entrapment conditions) is to be expected for units buried to shallower depth (i.e., of the order of seismogenic depths), because the possible divergence between rock and fluid P-T path is then reduced. Furthermore, the T gradient of isochoric lines of pure water (∼10–15 °C/km; see Fig. 10) is compatible with the T gradient along some subduction interface (Agard et al., 2009). In some cases, fluid trapped at peak conditions could in principle follow the same P-T exhumation path as rocks without being affected by reequilibration.

The P-T conditions recorded in the fluid inclusions from the present work, as well as a compilation of similar examples from various worldwide accretionary prisms, show nevertheless a discrepancy with respect to the thermal regime in modern subduction zones (Fig. 11). Fluid inclusions record a lower pressure than all modern zones, including the “hot” subduction zone of SW Japan where the young crust of the Philippines Sea Plate is subducting (Chamot-Rooke et al., 1987; Marcaillou et al., 2012).

Taken at face value, the fluid-inclusion record implies that the fluid pressure is in general significantly lower than the lithostatic pressure near or along the subduction plate interface at depths corresponding to ∼200–300 °C, i.e., near the base of the seismogenic zone (Fig. 11). This is possible but would be at variance with geophysical measurements showing large Vp/Vs ratio in this depth domain (Audet et al., 2009; Kato et al., 2010; Audet and Schwartz, 2013; Moreno et al., 2014).

A way out of this discrepancy is to use the fault-valve model (Sibson et al., 1988; Sibson, 1994), where large variations in the fluid pressure occur in relation with the seismic cycle. In such a model, fluid reservoirs, pressurized during the interseismic period, connect in the coseismic stage to shallower aquifers due to earthquake-related damaging. The resulting large-scale fluid flow leads to transient fluid pressure decrease, before fractures seal, permeability decreases, and fluid pressure increases back to lithostatic values. In this framework, it would be possible that fluid inclusions record principally the low pressure stage of the fluid pressure cycle, i.e., lower values than coeval lithostatic pressure. Experimental studies indicate a significant decrease in quartz solubility when pressure decreases (Anderson and Burnham, 1965), in agreement with the idea of closing cracks and trapping fluid inclusions after pressure drop (Fisher and Brantley, 1992). In addition, in Kodiak accretionary prism (Vrolijk, 1987a, 1987b) and in the Shimanto Belt (Raimbourg et al., 2017), fluid inclusions actually recorded variations in fluid pressure, supporting the model above. Nevertheless, the range of variations is of the order of 1–1.5 kbars, too small to account for the difference between the fluid pressure in the inclusions and the lithostatic pressure along the subduction plane (Fig. 11).

Alternatively, fluid inclusions from the shallower domains of subduction zones could be modified by post entrapment processes, similarly to the deeper domains (4.2.1). This hypothesis is somehow substantiated by the observation that the water-rich fluid inclusions, which are the most common type of fluid inclusions, record the lowest pressure, when compared to coexisting methane-rich fluid inclusions (Sakaguchi, 1999a; Raimbourg et al., 2014b). Because water dissolves the host mineral whereas methane does not, water-rich fluid inclusions are indeed more likely reequilibrated than methane-rich inclusions, which potentially record an internal pressure closer to the pressure of entrapment. One can also note that taking into account the small fraction of methane in the water-rich fluid inclusions (i.e., in the system H2O-CH4-NaCl), instead of neglecting it as is often done, offsets the isochore toward higher pressure, but to a very limited extent. In the FH (this study) or the Shimanto Belt (Raimbourg et al., 2014b), methane is always present with a low concentration (<1–1.5 mol/L H2O), so that this effect becomes negligible, and isochores are similar to pure-water ones and run at very low pressure.

In summary, in the deep domains of subduction zones (blueschists and eclogite facies), the record by fluid inclusions of the fluid pressure is dubious, as postentrapment processes are likely to modify the P-ρ-T properties of the inclusions in the general case. At shallower depths, corresponding to the seismogenic zone, the record of fluid inclusions is ambiguous. The very low fluid pressure at the time of trapping, with respect to corresponding rock pressure, may reflect the action of reequilibration processes, similarly to the deeper domains. Alternately, fluid inclusions may be systematically trapped after a fluid pressure drop that would be the consequence of the large earthquakes that rupture episodically the plate interface.

Compositions of the Fluid across the BPT

The low salinity observed in fluid inclusions from the Shimanto Belt or the western Alps is in agreement with other examples of low-grade terranes (Fig. 12), such as the Franciscan Complex (Sadofsky and Bebout, 2001) or the external domain in the Central Alps (Mullis et al., 1994). In Kodiak and nearby Afognak islands, the salinity spans a larger range, which encompasses seawater salinity (Brantley et al., 1998), while the fluid is made of H2O ± CO2. The transect in the Central Alps by Mullis et al. (1994) shows an increase in salinity from the lower-grade CH4-H2O domain to the higher-grade CO2-H2O domain, with a salinity respectively lower and higher than seawater.

At the much higher conditions of the eclogite facies, the salinity is much more variable, between highly saline brines in serpentine (Scambelluri et al., 1997) and metagabbros (Philippot and Selverstone, 1991) from the Alps and fluid with seawater-like salinity in eclogite blocks from the Franciscan Complex and the Dominican Republic (Giaramita and Sorensen, 1994). This salinity contrast between HP inclusions was interpreted as reflecting initial ( = pre-subduction) salinity contrasts by Philippot et al. (1998), but this hypothesis is at variance with the correlation between salinity and metamorphic grade shown by Mullis et al. (1994) and the relatively homogeneous salinity observed in metasediments shallower than the BPT. The extremely large variability in salinity reported in blueschist-facies metasediments on the island of Syros in the Cyclades (Barr, 1990) is therefore more likely to be connected with local-scale mineral-fluid interactions than with salinity contrasts inherited from the pre-subduction stage.

The nature of the gas dissolved in water shows also a contrasting behavior across the BPT. Shallower than the BPT, the gas phase is predominantly methane, as for example in the Alps (this study), the Shimanto Belt (Sakaguchi, 1999a; this study), or the Franciscan Complex (Sadofsky and Bebout, 2004). The nature of the dissolved gas phase becomes more variable with increasing depth. In the Alps, where buried sediments are rich in carbonates, the gas phase changes from CH4 to CO2 in the T range from 250 to 350 °C (Fig. 7). In contrast, in the Shimanto Belt metasediments, poor in carbonates, the gas phase is still methane at ∼350 °C (Fig. 7), except for CO2-bearing fluid inclusions in the Muroto Peninsula, associated with Miocene near-trench magmatism (Okamoto et al., 2014). Higher-grade metamorphic rocks are absent from the Shimanto Belt, but in the Jurassic subduction belt of Sambagawa (Japan), also poor in carbonates, all fluid inclusions down to eclogite-facies conditions contain a mixture of water, N2, and CH4, without any CO2 (Yoshida et al., 2015). The comparison of the HP domains in the Alps and in the Japanese subduction belts shows how the nature of the host-rock sediment, in particular its concentration in carbonates, controls the nature of the gas dissolved in the water, even if it is unclear which reactions precisely control the release of CO2 from carbonates (Kerrick and Connolly, 2001; Frezzotti et al., 2011; Cook-Kollars et al., 2014).

In summary, both in terms of salinity and composition of dissolved gas, there is a contrast across the BPT between a relatively homogeneous shallow domain and a much more heterogeneous deep domain. This implies that these two domains are, in terms of fluid, disconnected. It also reflects a relatively open-system behavior of the subducting material shallower than the BPT, contrasting with a closed-system behavior deeper. Such a contrast, in subduction context, is similar to what was observed in context of postorogenic collapse, during the exhumation of terranes along a detachment in the Cycladic islands of Greece (Famin et al., 2004; Famin et al., 2005).

Model of Fluid Flow along the Plate Interface

In agreement with the geochemical signatures described above, the cathodoluminescence study of quartz veins from the Alps and the Shimanto Belt described here (Fig. 6) reveals a major difference between low- and high-grade domains. While high-grade quartz veins are constituted of a homogeneous, CL-brown material, low-grade veins incorporate, in addition to the CL-brown quartz, a second type of quartz with a bright-blue luminescence. The respective occurrences of CL-blue and CL-brown quartz provide insight into the conditions for their formation. Because CL-blue and CL-brown quartz precipitate in “macro” and “micro” veins (e.g., Figs. 6A and 6B), we interpreted in Raimbourg et al. (2015) the CL-blue and CL-brown quartz as reflecting equilibrium with an external and local fluid, respectively. This interpretation was further supported by the CL-brown color of recrystallized domains in low-grade units, reflecting equilibration with the local fluid wetting the rock (Raimbourg et al., 2015). At high grade, quartz is pervasively recrystallized and homogenized with the local fluid, resulting in a homogeneous CL-brown color.

There is therefore a strong contrast between low-grade units, where episodic influx of external fluid is recorded in fracturing and precipitation of CL-blue quartz, and high-grade units, where the rock evolves as a closed system. In terms of mechanical behavior, the transition corresponds to the onset of plastic deformation of quartz (Fig. 4) or, at larger scale, to the seismic-aseismic transition of the plate interface behavior (Byrne et al., 1988; Hyndman et al., 1997; Oleskevich et al., 1999), for temperatures in the range 250–350 °C (Fig. 5 and Table 1).

Similarly to our observations of CL textures in low-grade material, in other examples of subduction thrusts (Vannucchi et al., 2010; Fagereng et al., 2011; Fisher and Brantley, 2014), crack-seal textures have been reported as evidences of repeated events of fracturing, fluid infiltration, and mineral precipitation (Ramsay, 1980). In the mélange zone of the Chrystalls Beach Complex, shear veins with crack-seal textures indicative of displacements of ∼10–100 µm were interpreted as the result of repeated micro-earthquakes occurring at ∼300 °C, i.e., near the down-dip limit of the subduction thrust (Fagereng et al., 2011). In the Kodiak accretionary complex, in Alaska, veins concentrated near the paleodécollement record hydraulic events with two distinct timescales: thin veins record repeated crack-seal events (with typical thickness of precipitated quartz of ∼8 µm), while a set of thicker veins record less frequent events of fracture porosity collapse (Fisher and Byrne, 1990; Fisher and Brantley, 1992, 2014).

This set of observations leads to a model of circulations with two compartments, limited by the BPT (Fig. 13). Deeper than the BPT, fluid circulation is only local, and the fluid composition (salinity and dissolved gas) is controlled by chemical exchanges with surrounding host rock. Shallower than the BPT, i.e., along the seismogenic zone, stages of closed-system evolution alternate with events of system opening and infiltration of external fluid. The associated fluid has seawater-like or lower salinity and contains dissolved CH4 with a low concentration.

The connection between these hydrological cycles and the earthquake cycle is unclear, as discussed in Fisher and Brantley (2014). Crack-seal textures may as well result from cyclic variations in pore-fluid pressure, without direct correlation with the seismic slip. The information provided by isotopic studies is also ambiguous. In veins from the Shimanto Belt, Yamaguchi et al. (2011) show a diminution in the fluid redox state during faulting and associated vein precipitation, but their observations cannot distinguish between a transient influx of exotic, deep fluid or the in situ production of hydrogen during seismic faulting. Therefore, further work is necessary to provide clear geological observations connecting the fluid flow to the seismic cycle.

CONCLUSIONS

Our comparative study of the Alps and the Shimanto Belt, which spans the lower portion of the seismogenic zone and the aseismic, plastic domain at larger depths, shows a relatively contrasted pattern, in terms of fluid circulation, between these two domains. Deeper than the BPT, the system is hydrologically closed, and the quartz, pervasively recrystallized, has a homogeneous, CL-brown color. The fluid salinity is variable and often larger than seawater, and the dissolved gas depends on the host rock. In contrast, shallower than the BPT, CL colors reveal the presence of two types of quartz, which we interpret as reflecting hydrological cycles of transient opening and closure of the system. The salinity of the fluid is of the order or lower than seawater, and the dissolved gas is consistently composed of methane, with a low concentration. The seismic cycle constitutes a good candidate for the transient hydrological opening of the system, but further work is required to establish such a connection. In addition, the fluid pressure recorded in fluid inclusions is much lower than lithostatic pressure along the subduction plane, which may reflect either (1) postentrapment, reequilibration processes or (2) trapping of fluid inclusions after a fluid pressure drop that would follow earthquake occurrence.

ACKNOWLEDGMENTS

We thank Frédéric Foucher at Centre de Biophysique Moléculaire (CBM), University of Orléans, for Raman analyses and J.G. Badin and S. Janiec for the preparation of thin sections. We also thank Donald Fisher and Åke Fagereng (reviewers) and Philippe Agard (guest associate editor) for their constructive comments, which helped to improve the manuscript. This work has received funding from (1) the European Research Council (ERC) under the seventh Framework Programme of the European Union (ERC Advanced Grant, grant agreement no. 290864, RHEOLITH), (2) the Labex VOLTAIRE (ANR-10-LABX-100-01), and (3) the program “Sakura” by the French Ministry of Foreign Affairs.

1Supplemental Table S1. Location of all samples and outcrops used in this work. Please visit http://doi.org/10.1130/GES01504.S1 or the full-text article on www.gsapubs.org to view Supplemental Table S1.
Science Editor: Shanaka de Silva
Guest Associate Editor: Philippe Agard
Gold Open Access: This paper is published under the terms of the CC-BY-NC license.

Supplementary data