Fluid evolution in an Oceanic Core Complex: A fluid inclusion study from IODP hole U1309 D—Atlantis Massif, 30°N, Mid‐Atlantic Ridge

In the detachment mode of slow seafloor spreading, convex‐upward detachment faults take up a high proportion of the plate separation velocity exposing gabbro and serpentinized peridotite on the seafloor. Large, long‐lived hydrothermal systems such as TAG are situated off axis and may be controlled by fluid flow up a detachment fault, with the source of magmatic heat being as deep as 7 kmbsf. The consequences of such deep circulation for the evolution of fluid temperature and salinity have not previously been investigated. Microthermometry on fluid inclusions trapped in diabase, gabbro, and trondjhemite, recovered at the Atlantis Massif Oceanic Core Complex (30°N, Mid‐Atlantic Ridge), reveals evidence for magmatic exsolution, phase separation, and mixing between hydrothermal fluids and previously phase‐separated fluids. Four types of fluid inclusions were identified, ranging in salinity from 1.4 to 35 wt % NaCl, although the most common inclusions have salinities close to seawater (3.4 wt % NaCl). Homogenization temperatures range from 160 to >400°C, with the highest temperatures in hypersaline inclusions trapped in trondjhemite and the lowest temperatures in low‐salinity inclusions trapped in quartz veins. The fluid history of the Atlantis Massif is interpreted in the context of published thermochronometric data from the Massif, and a comparison with the inferred circulation pattern beneath the TAG hydrothermal field, to better constrain the pressure temperature conditions of trapping and when in the history of exhumation of the rocks sampled by IODP Hole U1309D fluids have been trapped.


Introduction
In previous studies, oceanic and ophiolitic fluid inclusions have been interpreted in the context of purely magmatic ocean floor spreading, with the principle driving force for hydrothermal circulation being a shallow axial magma chamber [Kelley and Delaney, 1987;Kelley et al., 1992Kelley et al., , 1993. The consequences for hydrothermal fluid circulation in Oceanic Core Complexes (OCCs) of the recently established ''detachment mode' ' [Escartin and Canales, 2011;McCaig and Harris, 2012] of seafloor spreading have not been investigated. In particular, this mode of spreading may involve deeper magma chambers (up to 7 km below seafloor) , focussed fluid discharge along detachment faults [McCaig et al., 2007[McCaig et al., , 2010, and exhumation of gabbros and peridotites onto the seafloor [Cannat, 1993;Ildefonse et al., 2007;McCaig and Harris, 2012].
Fluids of variable salinities have been found in ocean crust. Processes responsible for such variation have been widely investigated in previous studies. Two-phase separation has been suggested as an important process to explain the observed salinity variation of hydrothermal fluids (10-200% of the seawater value) at different localities in the Mid-Atlantic Ridge such as the MARK area (Mid-Atlantic Ridge at Kane) [Kelley et al., 1993], and in the South West Indian Ridge [Kelley and Fr€ uh-Green, 2001]. Nevertheless, other processes such as exsolution from a melt [Kelley et al., 1992;Kelley and Malpas, 1996] and hydration/dehydration reactions with precipitation/dissolution of associated chloride-bearing minerals [Kelley and Robinson, 1990] need to be considered in certain cases.
The Atlantis Massif (AM) is an Oceanic Core Complex (OCC) located at 30 N at the inside corner of the Mid-Atlantic Ridge (MAR) and the Atlantis Transform Fault (ATF). Gabbros and mantle peridotite were exhumed to the seafloor by the movement of a detachment fault that is the locus of hydrothermal fluid discharge [Boshi et al., 2006;McCaig et al., 2007McCaig et al., , 2010. IODP Hole U1309D has been drilled in the central dome of the AM which is believed to be the footwall of this major detachment fault. The drillcore shows evidence of fluid

Geological Setting
The Atlantis Massif is an OCC, located at the inside corner of the MAR and the ATF (30 N) ( Figure 1). An OCC is a dome-like exposure of variably deformed and metamorphosed lower crustal and upper mantle rocks that has been unroofed by movement on a major detachment fault ( Figure 2) [Blackman et al., 2002[Blackman et al., , 2011Ildefonse et al., 2007;Tucholke et al., 1998;McCaig and Harris, 2012]. Gabbros from the massif have been dated between 1.08 6 0.07 and 1.28 6 0.05 Myr [Grimes et al., 2008].
The AM is composed of three different parts: (i) the central dome in which two deep holes (U1309B and U1309D) and five shallow-penetration holes (U1309A and U1309E-H), recovering upper sediment cover and fragments of detachment fault schist have been drilled (Figure 1), (ii) the southern wall which is dominated by serpentinized peridotite capped by a 100 m thick detachment shear zone rich in talc and tremolite [Boschi et al., 2006]; it is the host of the Lost City hydrothermal field, and (iii) the eastern block, interpreted as a fault-bounded block of basaltic material lying structurally above the central dome. The central dome is characterized by a corrugated surface believed to be an exposure of the major detachment fault responsible for the uplift of the massif. The corrugations are parallel to the spreading directions and have a wavelength of approximately 1000 m, amplitude of tens of meters, and length of several kilometers [Cann et al., 1997].
Hole U1309B (30 10.11 0 N, 42 07.11 0 W; 1642 mbsl) was drilled up to 101.8 mbsf with an average recovery of about 50%. Hole U1309D, located at 30 10.12 0 N, 40 07.11 0 W, 1645 mbsl, (20 m from Hole U1309B) penetrated 1415.5 mbsf, with a recovery of 75% comprising intrusive basalt and diabase (3%), gabbroic (91%), and olivine-rich rock (5%) consisting of dunites, wehrlites, troctolites, as well as a few mantle peridotites (harzburgite) in the upper 200 m. Seismic tomography suggests that a gabbro body several kilometers across forms the core of the massif, with relatively steep contacts against serpentinized peridotite to the south and west [Canales et al., 2008;Henig et al., 2012]. The drillcore recovered at Hole U1309D records strong evidence of penetration of altering fluids. Alteration occurred over a range of temperatures ranging from granulite facies to zeolite facies, but was dominantly in the greenschist and lower amphibolite facies [Blackman et al., 2011]. A few samples of detachment-related talc-tremolite schists were sampled in the uppermost 20 m of IODP Holes 1309B and 1309D, and fault breccias derived from diabase and gabbro are common in the upper 120 m of the holes [Blackman et al., 2006[Blackman et al., , 2011; there is no doubt that the holes were drilled into the footwall of the detachment fault observed in outcrop at the summit of the south wall.
Gabbros in U1309B and U1309D vary in grain size (microgabbro to coarse grained gabbro) and deformation type, and can also be divided in several groups: microgabbro, oxide gabbro, gabbronorite, gabbro, olivine gabbro, troctolitic gabbro, and troctolite [Blackman et al., 2006]. Overall, the gabbros are equigranular, but can exhibit different types of deformation ranging from (rare) mylonitic to absolutely undeformed. Plagioclase is generally unaltered, but can show evidence of albitization in the vicinity of veins and magmatic intrusions and can also be altered to chlorite along fractures and in a corona reaction with olivine. Clinopyroxene rarely survives alteration in the upper 350 m of the Hole and is replaced by amphiboles (hornblende, actinolite, and tremolite). When olivine is present, it is more or less replaced by amphibole to form coronas in the upper part of the core, and to serpentine often accompanied by rodingitization of plagioclase [Frost et al., 2008] at greater depths. Schoolmeesters et al. [2012] have used U-Pb zircon crystallization ages [Grimes et al., 2008[Grimes et al., , 2011, U-Th/He zircon thermochronometry, and multicomponent magnetic remanence data [Morris et al., 2009] to constrain the cooling rates of the AM. It has been shown that the upper 800 m of the central dome at the AM cooled from 780 C to 250 C at a rate of 2895 (11276/21162) C/Myr, whereas the lower 600 m cooled at a slower rate of 500 (1125/2102) C/Myr, from 780 C to present day temperatures. Rocks from the uppermost part of the hole appear to have cooled more slowly from 250 C to 190 C at a rate of 300 C/Myr due to the hydrothermal circulation along the detachment fault. These results imply a thermal structure of the AM such that the depth of the root of the detachment fault is 7 km. The depth of the 190 C isotherm resides around 1.5 kmbsf while the temperature at Moho depth of 4.5-5 km [Blackman and Collins, 2010] is >500 C. According to the exhumation model of Schoolmeesters et al. [2012], the depth of the 250 C, and 580 C isotherms along the detachment fault, respectively, are 3.75 kmbsf and 6 kmbsf.  On the basis of these criteria six samples were selected for detailed study.

Electron Microprobe Analyses
Electron microprobe analyses were conducted using a Cameca V R SX-50 fitted with three wavelengthdispersive spectrometers for full quantitative analyses and with an Oxford MicroAnalysis Division Link 10/ 55S Energy Dispersive System for reconnaissance of phases and qualitative analyses. Polished thin sections were carbon coated (10-15 nm) before analyses. The microprobe is calibrated with a certified jadeite (Na), a pure synthetic MgO (Mg), a pure synthetic Al 2 O 3 (Al), wollastonite (Si, Ca), halite (Cl), a certified orthoclase (K), a pure synthetic rutile (Ti), chromite (Cr), rhodonite (Mn), and hematite (Fe). This calibration is cross checked against silicate standards such as diopside (Si, Ca, Na), almandine (Si, Al, Fe), K-feldspar (Si, Al, K), and albite (Si, Al, Na) before every silicate analyses sessions.
Chlorite analyses were used to estimate temperatures of formation based on Cathelineau and Nieva [1985]. The calculation uses the relationship between the Al [iv] of chlorite and the temperature which is given by the following equation:  [Roedder, 1984], were not included in the statistics.
Repeated homogenization and freezing measurements were undertaken on individual inclusions in order to observe phase changes (such as ice melting point, halite dissolution temperature, liquid-vapor homogenization temperature) and to obtain homogenization temperatures and fluid salinities (in wt % NaCl eq). Salinities were calculated using the temperature of melting of ice for low-salinity fluids [Bodnar, 1993]. Salinities for saturated fluids were calculated using the temperature of dissolution of solid halite [Sterner Geochemistry, Geophysics, Geosystems

. Quartz Vein in Diabase
Sample U1309D 1R-1 41-44 (depth 5 20.9 mbsf) is a subophitic medium grained diabase composed of laths of unaltered and fractured plagioclase 0.1-2.5 mm in length, with poikilitic augite generally partially replaced by green hornblende (Figure 3a). The chemistry of several laths of plagioclase is presented in Table  1. Close to the quartz-chlorite vein, laths of plagioclase tend to have an albitic core and intermediate edges (labradorite), whereas away from the vein (laths 3, 4, and 5), the opposite is observed. The amphibole in the diabase is green magnesiohornblende (Table 1). Ilmenite partially replaces magnetite.
A 3 mm wide quartz-chlorite vein crosscuts the general fabric of the matrix. The quartz vein is equigranular and quartz grains are commonly of irregular shape. They exhibit a radial extinction (Figure 3b). Chlorite   forms vermiform and radiating aggregates (Figure 3c), primarily on one side of the vein, and shows three distinct growth phases; an early growth phase with Mg numbers in the range 32-40 (Table 2, Chl 1-4) at a temperature of 283 6 5 C [Cathelineau and Nieva, 1985], a second phase with Mg numbers from 54 to 60 (Table 2, Chl 5-8) at a temperature of 259 6 5 C [Cathelineau and Nieva, 1985], and a final phase with Mg numbers from 62 to 71 (Table 2, Chl 9-12) at a temperature of 242 6 17 C [Cathelineau and Nieva, 1985]. Chlorite appears to largely predate quartz, which is often intergrown with amphibole needles (magnesiohornblende) at the edge of the vein (Table 2, Am 1-3; Figure 3c). Fluid inclusions occur as irregular shaped primary (?) inclusions clustering in the clear central part of the quartz grains of the vein. They are two phase, liquid-dominated inclusions ranging in size from 5 to 10 mm.

Gabbro
Sample U1309D 5R-3 107-110 (depth 5 39.9 mbsf) is a medium grained troctolitic gabbro exhibiting plagioclase grains of intermediate composition (Table 3); it contains corona textures in which olivine is replaced by tremolite and plagioclase is partially replaced by chlorite and cut by chlorite veins (Figure 3d). Actinolite and magnesiohornblende replace clinopyroxene ( Figure 3e and Table 3). Plagioclase is slightly deformed and exhibits subgrain boundaries and deformation twins. Sample U1309D 10R-1 127-129 (depth 5 61.5 mbsf) is a mylonitized coarse gabbro composed of roughly 60% deformed, partially recrystallized plagioclase and 40% green-brown hornblende and actinolite, replacing clinopyroxene; the plagioclase is partially altered to chlorite. Granoblastic recrystallization affects the boundaries of amphibole grains suggesting that the shear zone was active at amphibolite facies conditions. In these two last samples, fluid inclusions occur as both regular and irregular shaped secondary inclusions in plagioclase that occur in trails decorating fractures and as larger cigar-shaped inclusions perpendicular to chlorite veins. They are two phase and liquid dominated, ranging in size from 10 to 20 mm.
Samples U1309D 40R-1 6-12 (depth 5 214.9 mbsf) and U1309D 40R-1 17-19 (depth 5 215.0 mbsf) are troctolites showing the same textural characteristics as samples U1309D 5R-3 107-110 with more abundant olivine and crosscutting quartz veins. Fluid inclusions in these samples occur as irregular shaped primary inclusions. They generally cluster in the central part of the quartz grains of the vein. These fluid inclusions are two-phase liquid-dominated ranging in size from 5 to 20 mm.

Trondjhemite and Crosscutting Quartz Vein
Sample U1309D 40R-1 21-24 is a fine to medium grained trondjhemite composed of albitic plagioclase and quartz. Both are usually anhedral and form graphic intergrowths ( Figure 3f). Alteration minerals (traces of tremolite and titanite along the edges of the graphic intergrowths) are not common. A large number of apparently primary fluid inclusions have been studied in a single 5 mm quartz grain of the trondjhemite. Two populations are observed; one with irregular shaped liquid-dominated inclusions and the other one with halite daughter crystals. Halite-bearing inclusions are generally bigger than those without halite and are irregular in shape (20-50 mm).
A late crack-seal quartz vein 4-5 mm in width crosscuts this late leucocratic magmatic intrusion. Quartz grains are elongated and also show radial extinction. Fluid inclusions in the vein are generally distributed as clusters in the clear central part of the grains, and are interpreted as being primary in origin. They are of irregular shape ranging in size from 10 to 20 mm, with some reaching several 10s of mm. They do not contain halite crystals.

Fluid Inclusions Typology and Results
Four types of fluid inclusions have been identified in the samples studied ( Figure 4).  Geochemistry, Geophysics, Geosystems Liquid-dominated, low-salinity inclusions have been found in all the samples in both quartz and plagioclase, and they generally occur as irregular inclusions ranging in size from 5 to 30 mm with the exception of sample U1309D 40R-1 17-19, where inclusions up to 100 mm in size have been found. Rare regular shaped inclusions are found in plagioclase grains. Irregular shaped inclusions might be the result of necking down, stretching, and/or leaking. Therefore, particular care has been taken to verify the similarity in behavior of irregular shaped inclusions relative to regular shaped ones. Where fluid inclusions exhibited salinities and/ or homogenization temperatures significantly different from the other inclusions in the same population, those were not included in the statistics (11 out of 329). Results from sample U1309D 40R-1 6-12 are not included in the statistics either as only six measurements were undertaken in this sample. Unlike the other samples, the distribution of homogenization temperatures is not unimodal ( Figure 6) and the wide range of homogenization temperature is possibly the result of necking down due to fracturing and chlorite filling veins in the vicinity of the inclusions. The mode in salinities is 0.5 wt % NaCl eq (depleted relative to the seawater value); N 5 163. In sample U1309D 10R-1 127-129, type 1a inclusions are also present and homogenize in the same range of temperatures at equivalent salinities of 2.0-4.3 wt % NaCl (Figures 5 and 6).

Type 3a: Liquid-Dominated High Salinity [L 1 V (L)]
Liquid-dominated high-salinity inclusions lacking daughter minerals (type 3a) are found in quartz grains of the trondjhemite sample (U1309D 40R-1 21-24), and are associated with type 3b inclusions. Their primary or secondary origin is not really clear. They are irregular in shape with a range in size of 5-20 mm. Homogenization occurred in the liquid phase at temperatures of 314.5 to >400 C. Melting of ice occurred at 26.2 to 217.6 C, indicating salinities of 9.5-20.7 wt % NaCl ( Figure 5); N 5 12.

Type 3b: Daughter Mineral-Bearing Inclusions [L 1 V 1 H (L)]
Halite-bearing fluid (type 3b) inclusions have been found in quartz grains of the trondjhemite and are associated with type 3a, although their temporal relationship is not clear. They are irregular in shape with variable size of 10-50 mm, liquid dominated, and rarely contain other daughter minerals (Figure 4e). Dissolution of the halite cube (184-294 C) always occurred at lower temperatures than the homogenization temperature, marked by the disappearance of the vapor bubble (336.4 to >400 C). Some inclusions remained unhomogenized at a temperature of 400 C (the limit of the stage used). Since halite dissolution had already been observed, the behavior of the vapor bubble suggested that homogenization would occur in the next 20-30 C. The vapor bubble had considerably decreased in volume and was intensely moving around the inclusion. This behavior, observed in all other inclusions, characterizes approach homogenization. These particular inclusions are indicated by arrows on Figure 5. Halite dissolution temperatures indicate equivalent fluid salinities of 31.1-37.7 wt % NaCl eq; N 5 7.
Cooling experiments were undertaken in order to test for the presence of additional gas species in the vapor phase, but no clear phase changes were observed.

Liquid-Dominated Related to Chlorite Veins
The elongated cigar shape inclusions that have been observed in plagioclase (Figure 4f) could not be studied microthermometrically because the thickness of the wafer made observation of a single inclusion impossible.

Fluid Evolution in Oceanic Core Complexes
The

Hydrostatic and Lithostatic Pressure Gradients
Interpretations of the pressure-temperature conditions of the two-phase curve between liquid and vapor are strongly dependent on the type of fluid pressure which occurs in the crust. Hydrostatic pressure applies to fluids in cracks under brittle conditions. Lithostatic pressure applies to fluids exsolving from a melt under ductile conditions and to fluids isolated from the convective circulation. The switch between those conditions is likely linked to the depth interval of the brittle-ductile transition. Calculations estimate this transition at a temperature of 700-800 C in moderately shallow gabbroic rocks [Hirth et al., 1998]. Hydrostatic pressure will be favored in a model where seawater is the fluid source whereas if magmatic fluid is the parent fluid the pressure may be hydrostatic or lithostatic, depending on depth of fluid circulation.
The PT gradient under hydrostatic pressure will vary depending on water temperature. A cold hydrostatic pressure gradient is 100 bars/km at normal seawater temperature, whereas a hot hydrostatic pressure gradient is 30 bars/km at black smoker temperature [Coumou et al., 2009] (see below and Figure 7). The vertical pressure gradient must be small enough for cold water to flow down in the recharge zone, and large enough for hot water to flow up in the discharge zone, implying the pressure gradient generally lies between cold and hot hydrostatic pressure [Jupp and Schultz, 2000]. Nonetheless, several studies [Jupp and Schultz, 2004;Wilcock and McNabb, 1996] assume that a pressure gradient very close to cold hydrostatic pressure defines the properties of the circulating fluids such as viscosity and flow resistance.

Pressure and Temperature Conditions of Fluid Entrapment
In OCCs, where the detachment fault is the locus of large fluid flux and the magma chamber is supposed to be approximately at a depth of 7 kmbsf, fluid may circulate from the depth of the seafloor to as deep as 7 kmbsf. We assume a water depth of 3.5 km, as seen at the TAG hydrothermal field. Over this depth range, Geochemistry, Geophysics, Geosystems A pressure correction is necessary to estimate the temperature of fluid entrapment from the microthermometry. This correction is valid for inclusions that contain pure NaCl solution, for which the salinity of the fluid has been correctly determined, and for which homogenization occurs in the liquid phase, and when pressure of formation can be estimated [Roedder, 1984]. The extreme pressures given above correspond to the minimum and maximum hydrostatic and lithostatic pressures that the inclusions experienced on trapping, depending on when they formed during the exhumation of the Atlantis Massif. These pressures allow calculation of isochores ( Figure 8) which the fluid inclusions followed from trapping to homogenization while cooling. Pressure-corrected trapping temperatures for fluids with less than 20.8 wt % NaCl were calculated from the software Loner38V C from http://fluids.unileoben.ac.at that computes the equations from Zhang and Frantz [1987] for the system H 2 O-NaCl. Pressure-corrected trapping temperatures for supersaturated fluid were calculated using equation 4 of Bodnar and Vityk [1994] that gives dP/dT (bar/ C) as a function of the salinity and the homogenization temperature.
The pressure-temperature conditions described in Schoolmeesters et al. [2012] are used here to better constrain the trapping conditions of fluid inclusions. In Figure 8, the intersection of the isochores and the P-T curves of Schoolmeesters et al. [2012] gives a range of possible trapping conditions for each individual fluid type. Pressures for the PT curves of Schoolmeesters et al. [2012] are calculated assuming either hydrostatic pressure or lithostatic pressure from the depth of the top core pressure/depth-temperature curve in their Figure 6 using a water column of 3500 m and a rock density of 3000 kg m 23 . Seawater-like fluid (type 1a) could have been trapped at depths between 1.8 and 3.8 kmbsf and temperatures between 190 and 250 C if in an active buoyancy-driven convective system under hydrostatic conditions. If they were isolated from any active system they could have been trapped deeper and at higher temperatures (2.8 to 4.2 kmbsf and up to 300 C) ( Table 5). Low-salinity fluid with respect to seawater (type 1b) could have been trapped at hydrostatic pressure of 780-800 bars and temperatures up to 290 C. They would have been trapped at greater pressures (up to 1.9 kbars at 460 C) under lithostatic conditions it they were isolated from any active  (Figure 7a) Seafloor is at an approximate depth of 3500 m corresponding to the TAG model. The two-phase curve is shown for hydrostatic conditions and separates the single phase liquid field from the two-phase vapor 1 liquid field [Sourirajan and Kennedy, 1962]. The critical point (Cp) of seawater is also shown (407 C, 298 bars) as well as the three phase curve that separates the stability field of liquid and vapor from that of vapor and halite. (b) Relationship between fluid pressure and depth for hydrostatic pressure under cold (100 bars/km) and hot (30 bars/km) gradients [Coumou et al., 2009] and a lithostatic pressure gradient. (Figure 7c) The two-phase curve is shown here in a temperature against depth diagram for cold and hot hydrostatic conditions using both diagrams of Figures 7a and 7b, and for lithostatic conditions. In this case, at a given depth, phase separation will occur at higher temperature under lithostatic conditions than under cold hydrostatic conditions, and in turn at higher temperature under cold hydrostatic conditions than under hot hydrostatic conditions. (d) The solidus of a water-saturated tonalite [Wyllie, 1977], whose intersection with the two-phase curve under lithostatic pressure separates the crystal 1 melt 1 liquid 1 vapor field from the crystal 1 melt 1 liquid field, is plotted. At depth <8 km and temperature > the solidus, an exsolved fluid following path 1 will consist of supercritical droplets of brines in a vapor phase; whereas a fluid exsolving along a path similar to 2 will exsolve as a single phase.
Geochemistry, Geophysics, Geosystems conductive system (Table 5). High-salinity fluid with respect to seawater salinity (type 3a) could have been trapped at hydrostatic pressure of 870 bars and at temperature of 470 C if connected to a convective system whereas it would have been trapped at higher pressure of 2.2 kbars and higher temperature of 635 C if not connected to such system (Table 5). If hypersaline fluids (type 3b) were connected to a conductive system, they could have been trapped at depths of 5 kmbsf and at temperatures of 430 C. Under lithostatic conditions, these same fluids would have been trapped at greater depths (5.5 kmbsf) and greater temperatures (500 C; Table 5).

Processes Modifying the Salinity
Several processes have been suggested to explain the variations in salinity observed in fluids circulating in the oceanic crust. They include subcritical phase separation (boiling) or supercritical phase separation (condensation) of a seawater-like fluid [Kelley et al., 1993;Kelley and Malpas, 1996] or magmatic fluid [Kelley and Figure 8. Range of possible trapping conditions for each type of fluid. A range of possible trapping conditions can be read at the intersection between the isochore and the P-T curves from Schoolmeesters et al. [2012]. These P-T curves were calculated for hydrostatic and lithostatic gradients from depths in their Figure 6 using a water column of 3500 m (assumed to be the depth at the end of detachment fault movement) and a rock density of 3000 kg m 23 . The liquid-vapor curve and the liquid-vapor-halite curve are from Khaibullin and Borisov [1966] and Sourirajan and Kennedy [1962]. The isochores were calculated with the software Loner38V C from http://fluids.unileoben.ac.at and from equation 4 of Bodnar and Vityk [1994]. Values are calculated with the averages of every type of inclusion, using the P-T curves of Schoolmeesters et al. [2012] for both hydrostatic and lithostatic pressure gradients and an assumed water depth of 3.5 km. Th 5 temperature of homogenization; Tt 5 temperature of trapping in degree Celsius; Pt 5 Pressure of trapping in bars; Depth t 5 Depth of trapping in kmbsf.
Fr€ uh-Green, 2001], magmatic fluids exsolving from melts [Kelley et al., 1992[Kelley et al., , 1993Kelley and Malpas, 1996], hydration/dehydration reactions with precipitation/dissolution of associated chloride-bearing minerals [Kelley and Robinson, 1990;Kelley et al., 1992], and variable mixing of hydrothermal fluid with a phase-separated brine or vapor [Kelley and Robinson, 1990]. The most common explanation for the generation of low-salinity fluids is phase separation of seawater-like fluids, and the most usual explanation for generation of high fluid salinities is phase separation of either magmatic or seawater-like fluids. Hydration reactions may also play a role. Variable mixing of hydrothermal seawater with phase-separated brines and vapor can also change the salinity of fluids as a late process.

Hydration/Dehydration
Under rock dominated conditions, hydration reactions or retrograde dissolution of chloride-bearing mineral phases have the potential to modify the ionic strength of hydrothermal fluids by consuming or liberating chloride ions [Kelley and Robinson, 1990;Kelley et al., 1992]. Formation of secondary amphibole containing up to 4 wt % chlorine [Vanko, 1986] can then result in decrease of fluid salinities, and dissolution of such phases might increase fluid salinities [Seyfried et al., 1986]. These processes could then account for slight changes in fluid salinity (low-temperature, low-salinity fluid generation) at relatively low water-rock ratio conditions, preferentially in a near axis environment recharge zone [Kelley et al., 1995], in contrast to an outflow zone where fluids rapidly pass through the oceanic crust .
Electron microprobe data (Tables (1-3)) of amphiboles in samples from which microthermometry was undertaken, show Cl/H 2 O of 0.000-0.049 with a mean of 0.01 6 0.014, whereas Cl/H 2 O in seawater is 0.0195. Formation of such amphiboles is unlikely to have influenced salinity of the residual fluid as Cl content is too low compared to that of seawater. The complete collection of electron microprobe data from holes U1309B and U1309D show a Cl/H 2 O ratio of 0.000-0.148 with a mean at 0.024 6 0.028 (unpublished data of our collection). In general, formation of this type of amphibole cannot account for the great variety of salinity observed in U1309D, but some amphiboles show such high Cl/H 2 O (0.148 for the maximum in this study) that precipitation of such minerals can lead to a slight salinity decrease in the residual fluid.

Phase Separation 5.4.2.1 Generation of Brines
Fluid sources and pressure conditions for brine-bearing inclusions are difficult to determine. Brine inclusions homogenize by disappearance of the vapor bubble at temperature >400 C. Two models for the generation of brine are as follow: 1. Brine and vapor are generated during supercritical phase separation (condensation) of either magmatic or seawater-derived fluids with segregation of the phases driven by density differences and entrapment of the brine at depth.
2. Direct exsolution of magmatic brine from late stage melts with significant cooling during the migration of the brines along microfractures (Figure 7).
The system H 2 O-NaCl will be used in the discussion below as an analog for fluid circulating in the oceanic crust. In a temperature-pressure diagram, the two-phase curve separates the one-phase field (liquid) from the two-phase field (liquid 1 vapor) at pressure-temperature conditions greater than the critical point of seawater (Cp: 407 C; 298 bars). Fluids of seawater-like salinity or magmatic fluids that circulate at deep levels of the oceanic crust and intersect the two-phase curve will undergo supercritical phase separation (commonly described as condensation) where droplets of brines will separate out of a vapor-rich phase ( Figure  7a). Fluids circulating under low-pressure conditions will boil and separate a vapor from a low-salinity fluid. However, since a water depth of approximately 3500 mbsf is assumed in our TAG-based model for the Atlantis Massif, boiling cannot happen in this system. In addition, seawater-like fluids in hydrothermal systems mainly circulate at temperatures of 400 C or less [Coumou et al., 2009]. According to Figure 7c, fluids which circulate at this temperature and at any depth or pressure will stay in the single phase region, and phase separation is therefore impossible. Fluids need to be heated up by magmatic intrusion such as diabase dikes [McCaig and Harris, 2012] to undergo phase separation in this system.
As seawater-like fluids circulate down to depth, fluids will traverse several condensation curves depending on the composition while approaching the heat source [Kelley et al., 1993]. For instance, a fluid of seawater composition (3.2 wt % NaCl), which circulates at crustal depth of approximately 2 km and under 550 bars of pressure assuming a water column of 3.5 km and cold hydrostatic conditions, will encounter the two-phase curve if heated to 500 C by intrusions, and will condense a fluid containing 25.3 wt % NaCl and a vapor with 2.2 wt % NaCl (Figure 9a). Under hot hydrostatic conditions, the fluid would need to circulate at a total depth of 10 km-crustal depth of 6.5 km (Figures 7b and 7c) in order to generate the same result by phase separation at 550 bars and 500 C. This would imply that fluids were trapped almost immediately after being generated since the maximum depth of circulation in the TAG model is 6-7 kmbsf deMartin et al., 2007;McCaig et al., 2010]. Higher-salinity fluids (in comparison to seawater) circulating under the same pressure temperature conditions would give the same compositions for brine and vapor, but would separate a bigger proportion of brine. If fluids circulate at shallower levels, they will separate out a greater volume of vapor given the pressure dependence on the shape of the two-phase curve [Kelley et al., 1993]. Under lithostatic conditions, the same fluids must be at higher temperature (600 C) in order to undergo phase separation (Figure 7c).
Generation of hypersaline magmatic brines can be explained by two different processes. Figure 7d shows the two-phase curve for a fluid of seawater salinity (3.2 wt % NaCl eq) for lithostatic conditions and illustrates two scenarios: 1. Exsolution of magmatic fluids under supercritical conditions and condensation of droplets of brines in a vapor phase (Figure 7, path D1) [Kelley and Delaney, 1987;Kelley and Fr€ uh-Green, 2001], Fluids in the melt are under lithostatic pressure, and at depth <8 km and at temperature above the solidus, crystal, melt, liquid, and vapor coexist. A fluid exsolving under conditions of the two-phase field will undergo exsolution under supercritical conditions and condensation of immiscible droplets of brine in a vapor phase.
2. Direct exsolution of brines in the absence of a vapor phase (Figure 7, path D2). At depths greater than 8 km, the solidus separates a field where solid, melt, and liquid coexist from a field with a single liquid plus Figure 9. Processes generating variations in salinity observed in the Atlantis Massif. (a) P-X projection of the system H 2 O-NaCl contoured for T under hydrostatic conditions. The critical curve (dashed line), the isotherms (dotted lines), and the three phases curve (plain line) are from Sourirajan and Kennedy [1962]. A seawater-like fluid which intersects the two-phase curve, at a temperature of 500 C and pressure of 550 bars will undergo supercritical phase separation (condensation) and separate droplets of brines with salinities of 25.3 wt % NaCl eq from a vapor of salinity close to 2.2 wt % NaCl equivalent. (b) In this model, it is assumed that the vapor-like fluid generated by phase separation has a salinity of 0.2 wt % NaCl. That salinity can be obtained under various conditions that will generate different brine salinities. The minimum temperature, at which supercritical phase separation generates a vapor phase of 0.2 wt % NaCl under hydrostatic conditions, is 450 C at a pressure of 340 bars and the maximum temperature is 600 C at 550 bars. (c) Temperature-depth diagram under hydrostatic conditions above seafloor (dashed line) [Sourirajan and Kennedy, 1962]  solid, such that any fluid under those conditions would be exsolved as one single phase of uncertain salinity.
Halite-bearing inclusions have been found only in trondjhemite. Trondjhemite being a late magmatic intrusion, a seawater-derived parent fluid is not likely. A magmatic fluid source for generation of hypersaline inclusions seems then to be the most probable. The condensation model is preferably applicable for brines and associated low-salinity vapor-rich inclusions, whereas the direct exsolution of brines model is more applicable to inclusions that homogenize by halite dissolution [Kelley and Fr€ uh-Green, 2001]. Since hypersaline inclusions in IODP hole U1309D homogenize by vapor disappearance and not by halite dissolution, it is suggested that brines have been formed by exsolution of a magmatic fluid under condensation (Figure 9c). The maximum pressure-temperature conditions are 770 C and 7 km depth, which is equivalent to 1.5 kbars according to the lithostatic gradient of Figure 7b. The minimum conditions in terms of pressure depth are 790 C and 4 km depth, which is equivalent to 515 bars according to Figure 7b. Note that since it concerns a magmatic intrusion (trondjhemitic intrusion), the maximum conditions seem to be the most probable.

Generation of Low-Salinity Fluid
Fluid of low salinity relative to seawater can be generated by phase separation as described above. Trapping of a low-salinity vapor fluid has not been observed. What is in fact observed, are liquid-dominated low-salinity fluids in plagioclase at lower temperature than the brine inclusions. As brine and vapor are segregated after separation by density effects, late mixing between the initial vapor-like fluid with seawaterlike fluid can occur to reach the salinity and temperatures of homogenization observed. Note that the phase separation event does not have to be the same event described in the previous paragraph for generation of brine observed in trondjhemite. In the model of Figure 9b, we discuss the conditions for the generation of low-salinity fluid (fluid type 1b) with respect to seawater salinity. It is assumed that the vapor-like fluid generated by phase separation has a salinity of 0.2 wt % NaCl because that is the minimum salinity of low-salinity inclusions (type 1b) found in the system. That salinity can be obtained at various conditions that will generate different brine salinities. The minimum temperature, at which supercritical phase separation generates a vapor phase of 0.2 wt % NaCl under hydrostatic conditions, is 450 C at a pressure of 340 bars (Figure 9b). However, the minimum hydrostatic pressure possible in the system constrained by our TAG-based AM model is the seafloor pressure (P 5 350 bars). The maximum temperature is 600 C at a hydrostatic pressure of 550 bars (Figure 9b). Such temperatures can most easily be reached if fluids are heated up by dikes or other intrusions [McCaig and Harris, 2012] that are formed deep and that are unroofed during exhumation of the massif. However, in both cases, significant cooling must occur in order to reach the trapping temperatures of such inclusions that do not exceed 345 C on average under hydrostatic conditions and 440 C on average under the maximum pressure correction at lithostatic conditions. Estimated pressures of trapping based on the Schoolmeesters et al. [2012] exhumation curves (790-1860 bars; Figure 8 and Table 5) are much higher than those suggested by our H 2 O-NaCl model. At 790 bars under hydrostatic conditions, fluid that intersects the critical curve cannot produce such a low-salinity fluid. It may be that the assumed water depth of 3.5 km in the model is too great, or that some of the assumptions regarding exhumation rate in the Schoolmeesters et al. [2012] exhumation paths need adjusting. Tivey et al. [2003] first hypothesized that the TAG hydrothermal field [Alt and Teagle, 1998;Rona et al., 1993;Teagle et al., 1998aTeagle et al., , 1998bTeagle et al., , 1998c is located on the hanging wall of an active detachment fault. deMartin et al. [2007] studied seismic refraction and microearthquake from the TAG segment and suggested that the TAG hydrothermal field is sited on the hanging wall of a dome-shaped detachment fault that penetrates to depths of >7 km below the seafloor. Their results suggest that high-temperature fluid discharge at TAG is controlled by the detachment fault, a hypothesis supported by evidence that exhumed detachment faults in the Atlantic were the locus of significant fluid flow at black smoker temperatures [McCaig et al., 2007[McCaig et al., , 2010. A seismic refraction study from Canales et al. [2007] also supports this hypothesis and adds that magmatic intrusions at depth (>4 km and perhaps as deep as 7 km) must be the heat source for sustaining the long term, high-temperature hydrothermal circulation at TAG. McCaig et al. [2010] interpreted the cross section of deMartin et al. [2007] and added isotherms to the model in order to predict the thermal evolution of the TAG detachment fault and its footwall. It is assumed that the configuration of the TAG hydrothermal mound fits the possible early configuration of Atlantis Massif. Schoolmeesters et al. [2012] propose a similar model with hydrothermal cooling of the detachment fault and its footwall, but with lower fluid temperatures as suggested by thermochronometric data. Thermal balance arguments suggest that large black smoker systems like TAG cannot be continuously active for long periods [Cannat et al., 2004], yet TAG has a long history of activity [Lalou et al., 1995]. It may be that black smoker circulation was episodic, punctuated by periods of lower temperature flow, with an average temperature consistent with the Schoolmeesters et al. [2004] cooling curves. The importance of the TAG model is that fluid is predicted to circulate much deeper than in previous models used to understand fluid inclusion microthermometry in seafloor hydrothermal systems [Kelley and Delaney, 1987;Kelley et al., 1993Kelley et al., , 1995Kelley and Malpas, 1996].

Detachment-Controlled Fluid Discharge-The TAG Model
Seawater fluids circulating at 400 C or less should remain in the single phase region throughout ( Figure  7c). Figure 8 and Table 5 indicate that these type 1a fluids could have been trapped between 1.5 and 3.8 kmbsf in the Schoolmeesters et al. [2012] model at hydrostatic pressures. In Figure 10, this would be between locations A and B. These fluids may have circulated much deeper in the system before being trapped, or they could be recharge fluids. In an active hydrothermal system such fluids would be expected to remain at hydrostatic pressures. However, a trapping pressure of around 900 bars would be required to reconcile the fluid inclusion temperatures with those based on chlorite geothermometry (283-242 C, see section 4) in sample 1R1 41-44. This suggests a trapping pressure between hydrostatic and lithostatic, although the depth of trapping may be little different. These data suggest some cycling of pressure between hydrostatic and higher values, perhaps during quiescent phases of circulation. Small variations in salinity in these fluids may be due to hydration reactions or mixing with type 1b or type 3 fluids. Type 1b fluids require higher temperatures in their history in order to allow phase separation (Figures 7 and  9). The detachment fault was intruded by diabase dykes, including the host of the vein in U1309D 1R-1 41-44. These contain high-temperature amphiboles [McCaig and Harris, 2012], indicating intrusion into a wet zone of hydrothermal flow. In the footwall of the detachment, intrusion of gabbroic magma creating a conductive thermal boundary layer to the hydrothermal system has been suggested [McCaig and Harris, 2012]. In either of these circumstances transient temperature increase could lead to phase separation if the pressure is not too high. The low-salinity fluids cooled and mixed with seawater before trapping. Dyke injection could have happened between locations A and B.
Finally, we suggest that the hypersaline type 3 inclusions formed by magmatic exsolution of fluid, which could occur either in the single-or two-phase region depending on pressure. If the melt lens was at a depth of 7 kmbsf as suggested by de Martin et al. [2007], this would be too deep to allow exsolution in the twophase region according to Figure 9c, but this diagram assumes seawater salinity and it is not clear what the salinity of a single phase fluid exsolving from a melt would be. Melt could also have intruded shallower as suggested by the range of U-Pb zircon ages reported by Grimes et al. [2008]. These fluids were not trapped at magmatic temperatures, but at temperatures of 505-635 C assuming lithostatic conditions (Table 5). The fact that the inclusions occur as secondary inclusion within quartz grains in a trondjhemite suggests that the brine may have been carried upward within the host intrusion before being remobilized into the observed fluid inclusions. This may have occurred as result of hydrofracture as the lithostatic pressure fluid was decompressed by exhumation. Based on Figure 8, we have suggested a depth of around 5 kmbsf for the trapping of these inclusions, with exsolution from the melt occurring somewhat deeper. The trondjhemite (sample 40R1 21-24) was then cut by a crack-seal quartz vein containing type 1a fluid inclusions. This likely happened at higher levels, between A and B, and in a hydrostatic regime, although lithostatic fluids cannot be ruled out.
The pattern of circulation shown in Figure 10 is highly schematic, and in reality both up and downflow zones are likely to be focused within the fault zone in the third dimension. It is also likely that fluid flow was episodic, with strong upflow events at black smoker temperatures of 350-400 C triggered by magmatic intrusion and intervening periods of less intense, lower temperature circulation [McCaig et al., 2010].

Conclusions
Fluid inclusion microthermometry demonstrates the occurrence of four different types of fluids in IODP Hole U1309D with different salinities (ranging from 1.4 to 35 wt % NaCl eq) and homogenization temperatures (160 to >400 C). The lowest homogenization temperature is exhibited by fluid trapped in quartz veins (type 1a), and the highest by fluid type 3.
High-salinity (Type 3b) fluid, only found in the evolved trondjhemite intrusion, is proposed to have been generated by condensation of a magmatic fluid at maximum temperature of 770 C at depth of 5-6 kmbsf. These fluids were trapped as fluid inclusions at somewhat lower temperatures during exhumation of the trondjhemite.
Low-salinity fluids (Type 1b), only found in plagioclase, are believed to have been generated by mixing between seawater-derived fluid and supercritical phase-separated seawater at temperatures of 450-600 C and pressures of 340-550 bars and are assumed to have been trapped at a depth of 3 kmbsf. Black smoker fluids do not normally circulate at such temperatures and it is suggested that phase separation occurred due to injection of dikes.
A late stage fracturing event has provoked precipitation of quartz veins at low pressure (450 bars) and temperature (210 C) that have trapped seawater-like salinity fluid after the footwall had been excavated to shallower depths. Comparison with chlorite geothermometry results suggests that these inclusions may have been trapped at suprahydrostatic pressures, possibly during relatively quiescent phases of fluid circulation.
Results show that the TAG model of McCaig et al. [2010], with temperature data refined by Schoolmeesters et al. [2012] provides a good framework for explaining the fluid evolution at the Atlantis Massif.