Diagenetic evolution of a fractured evaporite deposit (Vilob ı Gypsum Unit, Miocene, NE Spain)

The upper Burdigalian Vilob ı Gypsum Unit, located in the Vall es Pened es half-grabben (NE Spain), consists of a 60-m thick succession of laminated-to-banded secondary and primary gypsum affected by Neogene extension in the western part of Mediterranean Sea. This Tertiary extensional event is recorded in the evaporitic unit as six fracture sets (faults and joints), which can be linked with basin-scale deformation stages. All fractures are totally or partially infilled by four types of new gypsum precipitates showing a large variety of textures and microstructures. A structural and petrological study of the unit allows us to establish the following chronology of the fracture formation and infilling processes, from oldest to youngest: (i) S1 and S2 normal faults sets formation and precipitation of sigmoidal gypsum fibres; (ii) S3 joint set formation and growing of perpendicular fibres; (iii) S4 inverse fault development, infilled by oblique gypsum fibres and deformation of the previous fillings; and (iv) S5 and S6 joint formations and later dissolution processes, infilled by macrocrystalline gypsum cements. The fractures provided the pathway for fluid circulation in the Vilob ı Unit. The oxygen, sulphur and strontium isotope compositions of the host rocks and the new precipitates in the fractures suggest clear convective recycling processes from the host-sulphates to the fracture infillings, recorded by a general enrichment trend to heavier S–O isotopes, from the oldest precipitates (sigmoidal fibres) to the youngest (macrocrystalline cements) and to the preservation of the strontium signals in the infillings.


INTRODUCTION
Evaporite rocks are usually considered as ductile rocks under tectonic stresses, as they tend to deform and flow without rupture, due to their general low density and viscosity. Evaporites expulse their pore waters in the first tens of metres when buried; thus, they become nonporous and noncompressible rocks under burial conditions (Casas & Lowenstein 1989;De la Cueva 1992;Warren 1999). Salt and sulphate rocks have low permeability, very low solubility in hydrocarbons and no or very little porosity. Such rocks frequently develop large scale halokinetic structures (pillows, diapirs) associated with large oil and gas accumulations. Furthermore, they also serve as d ecollement horizons in foreland fold and thrust belts (De Meer et al. 2000; and references therein). Gypsum and anhydrite rocks show a different mechanical behaviour than carbonates. Even under moderate stresses, anhydrite can develop structures generally observed in metamorphic rocks, such as microfolding, boudinage, foliation, stretching lineation, shear bands and recrystallization (Mossop 1973;Marcoux et al. 1987;Lugli 2001;Rouchy & Blanc-Valleron 2006;Torres et al. 2012). Anhydrite is believed to show a more rigid behaviour than gypsum; anhydrite fragments have been described as rigid objects that develop spectacular gypsum pressure shadows in d ecollement zones (Triassic evaporites) from the Southern Alps (Malavieille & Ritz 1989).
Despite the very ductile behaviour of gypsum and anhydrite rocks, the Vilob ı Gypsum Unit shows several totally or partially filled fracture sets, unrelated to folding or other ductile structures; thus, the unit has suffered brittle deformation, and fluids have flowed along the fracture planes. Fracture processes and fluid flow evolution in evaporites are poorly understood in the literature. As stated above, most of the references are dedicated to ductile deformation and microstructures. For example, Sadooni (1995) describes fracturated sulphates, but he associates such brittle mechanisms to folding.
Most of the literature dealing with gypsum precipitates in fracture pays attention to the satin spar (fibrous) gypsum veins, as the origin of these infillings has been under great debate (Shearman et al. 1972;Machel 1985;Marcoux et al. 1987;Gustavson et al. 1994;El Tabakh et al. 1998;Philipp 2008). All these authors agree that fibrous fillings are cements in opened fracture planes and mainly assume that rehydration of anhydrite rocks and expulsion of CaSO 4 À rich fluids are the origin of the parental fluids. But in general, satin spar veins are horizontal and exclusively display fibrous gypsum; combination of several cements in fractures and/or gypsum cements in nonstrata-bound fractures have not been described in the literature, but they have been recognized in the studied evaporite unit.
The main objectives of this study are: (i) to determine the parental fluids of the fracture infillings and the evolution of the fluids during the different precipitation events, according to the whole diagenetic history of the unit; and (ii) to establish direct relationships between the local deformation stages and the regional deformation events, paying special attention to the fragile deformation structures.

GEOLOGICAL SETTING
The Vilob ı Gypsum Unit is located in the Pened es halfgraben, which is located in the Catalan Coastal Ranges, at the NE part of Iberian Peninsula (Fig. 1). The Catalan Coastal Ranges represent the continental northern margin of the Valencia Trough, which is a large NE-SW orientated basin developed as a result of the regional extensive Neogene event, which affected the western Mediterranean. The Valencia Trough is an oil-producing basin where accumulations are in Mesozoic and Neogene rocks; over a hundred wells have been drilled in the Valencia Trough since initial discovery in 1970. The most important fields are the Amposta and Casablanca oil exploitations (Clavell & Berastegui 1991;Varela et al. 2005;Play a et al. 2010;Rodr ıguez-Morillas et al. 2013).
The Catalan Coastal Ranges are mainly constituted by a NE-SW oriented horst and graben system developed during Neogene extension, resulting from the reactivation of previous compressive fractures developed during the regional Paleogene contraction (late Eocene-early Oligocene) (Roca 1994;Cabrera et al. 2004;Gaspar-Escribano et al. 2004). Part of the Valencia Trough and the present Cata-lan Coastal Ranges were uplifted during this compressive event and during the late Oligocene-early Miocene, the compressional structures were overprinted by extensional faulting related to a regional rifting stage (giving rise to the NE-SW trending). That regional Neogene extension was characterized by two main stages: a syn-rift episode developed between late Oligocene and early Burdigalian, and a thermal subsidence stage from the latest Burdigalian to Quaternary. The thermal subsidence stage was not a continuous stage because minor steps can be differentiated (Bartrina et al. 1992;Roca & Guimer a 1992;Roca et al. 1999): (i) Quiescent stage with little tectonic activity stage (late Burdigalian-early Serravallian); (ii) brief reactivation stage (late Serravallian-Messinian) including minor strikeslip and compressive events, Serravallian and Messinian in age; and (iii) attenuation of extensional activity (Pliocenepresent) (Bartrina et al. 1992;Roca 1994;Trav e & Calvet 2001).
The thickness of the Neogene sediment filling the Pened es basin reaches up to 4000 m. As a consequence of the rifting stage, followed by thermal subsidence, several transgressive pulses in the upper Burdigalian to Langhian led to a partial flooding of the western part of the basin by sea water changing to nonmarine conditions, which have prevailed up to present (Cabrera 1981). The basin exhibits a heterogeneous sediment filled with continental clastics, marine deposits including reefs build-ups and sabkha-salina evaporite sediments.

METHODS
This structural study was carried out in the region where evaporites crops out. We collected approximately 80 fracture plane measurements, which were chronologically arranged according to their crosscutting relationships. Sixty-one samples of host gypsum rocks and fracture infillings (gypsum) were collected, and twenty standard thin sections were prepared. Petrological and microstructural studies were performed using optical and scanning electron microscope (ESEM QUANTA 200 with a X-ray microanalysis device-EDAX Genesi). Twenty-two powdered samples were analysed by X-ray diffraction for a mineralogical characterization in a Bragg-Brentano PANalytical X'Pert PRO MPD alpha-1 diffractometer.
All the powdered gypsum samples were first reacted with HCl 2 N before carrying out the geochemical analyses, to remove carbonate fraction and part of celestite content (Lucchesi & Whitney 1962;Play a & Rosell 2005).
The d 18 O VSMOW and d 34 S VCDT values of the sulphates, in, were analysed in fifty-one gypsum samples all converted to BaSO 4 . The respective CO and SO 2 gases produced from the sulphates were analysed on a continuous-flow elemental analyser Finnigan DELTA plus XP mass spectrometer, with TC/EA pyrolyser for oxygen and Finnigan MAT CHNS 1108 analyser for sulphur. The results are precise to 0.4 for d 34 S VCDT and 0.5 for d 18 O VSMOW .
To determine 87 Sr/ 86 Sr ratios, ten samples of both host gypsum and fracture infillings were selected. The strontium and rubidium contents were first analysed to control possible contamination from celestite (Sr), while contribution of radiogenic 87 Sr coming from the 87 Rb decay (Rb). Sample were dissolved in 18 mΩ cmÀ1 Milli-Q water; to avoid gypsum saturation in the final solution and to ensure a similar range of concentrations, a constant dilution ratio of 1:500 was maintained for all samples. The samples were then filtered through a 0.45-lm cellulose acetate and cellulose nitrate membrane filter to remove any quartz or other insoluble mineral particles. Sr contents were analysed in a ICP-OES with a Perkin Elmer, model Optima 3200RL, with wavelength of 421.552 nm, while Rb was charged in a Perkin Elmer ICP-MS, Elan 6000 model. The ten powdered (and HCl 2N leached) samples were dissolved in Teflon PFA (Savillex ® ) vessels using HNO 3 2N. The solutions were subsequently centrifuged to separate the fraction insoluble in acid. Chemical procedures for Sr extraction using the Sr-Spec resin of Eichrom. All the procedures were carried out in a clean room. The Sr fractions of the rock samples were measured for isotope ratios by plasma ionization multicollector mass spectrometry (MC-ICP-MS) using a high-resolution Thermo Fisher Scientific Neptune instrument at the SGIker facility. The accuracy and reproducibility of the method were verified by periodic determinations under the same conditions using the NBS-987 strontium standard, averaging (n = 4) 0.710272 AE 0.000022 (2r). The precision of the measurement is 0.0031% (2r).

GYPSUM UNIT
The Vilob ı Gypsum Unit is upper Burdigalian in age. The unit outcrops at the northwest of Vilob ı village and is orientated NW-SE and dips 30°to the NE. This evaporite succession (60 m thick) is laying on nonmarine grey shales and limestones and is overlain by nonmarine red and grey shales and marine calcarenites (Langhian) (Ort ı & Pueyo 1976;Bitzer 2004). These Tertiary deposits are separated by an erosional unconformity from the underlying Cretaceous carbonate rocks.
The unit consist of two subunits. The Lower Subunit is characterized by three different textural varieties that progressively evolve into the other with no defined boundaries, preserving the depositional lithofacies (laminated-to- banded gypsum alternating with thin carbonate laminae) (Ort ı & Pueyo 1976; Fig. 2). Several lutite levels, up to 20 cm thick, are interlayered in the evaporite unit. Part A of this subunit (Fig. 2) is up to 30 m thick and is texturally formed by micronodular-nodular fine-grained alabastrine (microcrystalline) gypsum. The following 20 m (part B) are formed by radial aggregates with diameters up to 10 cm. In the upper part (part C), the radial aggregates become larger enough (several tens of cm in length) allowing the identification of individual elongated and lenticular crystals crosscutting the bedding. Some enterolithic beds are intercalated throughout this gypsum interval. Under the microscope, all the textural varieties show anhydrite relics within the gypsum crystals (Fig. 2). Celestite is also a minor accesory mineral widely present in the whole unit.
The Upper Subunit (4 m thick) is characterized by laminated-to-banded gypsarenite and microselenite (grass-like) layers; crystals are sized up to 1 cm. Carbonate laminae and celestite are also present, and anhydrite is lacking. The contact with the secondary gypsum subunit is sharp.

FRACTURE SET ANALYSIS
Up to 6 fracture sets (faults and joints) have been identified and related to different deformation events ( Fig. 3): (1) The first two fracture sets (S1 and S2) consist of coeval conjugate normal faults structures. S1 consists of NE-SW trending extensional faults dipping between 33 and 70°SE (mean of 50°). S2 is composed of a normal fault system trending NE-SW with a mean dip value of 50°NW (from 38 to 80°) (Fig. 3). Both sets of fractures are nonstrata-bound, showing scale spacing up 10 m, lengths of up to 10 m and displacement in the order of cm-dm ( Fig. 4). In most cases, the facture porosity has been enlarged by dissolution along the   fracture plane, being more evident in the crosscutting areas. All fracture and cavern porosities (openings up to 10 cm) are totally filled by sigmoidal fibrous and/or macrocrystalline gypsum cements.
(2) Set 3 (S3) is formed by NE-SW trending, subvertical, NW-dipping joints (Fig. 3) that crosscut S1 and S2, indicating a younger formation age. This set of fractures occurs very locally, in groups of three or four joints spaced cm to dm and up to 1 cm in width. These nonstrata-bound fractures have not been enlarged by dissolution, and they are totally filled by perpendicular fibrous gypsum (satin spar).
(3) Set 4 (S4) is made up of NE-SW trending thrust faults with both NW and SE dipping (Fig. 3). S4 fractures have a scattered presence in the Vilob ı Gypsum Unit, corresponding only to 6% of the total recorded measurements. Fracture lengths range from 15 cm to 3 m and are randomly distributed throughout the unit (Fig. 4). These structures are totally filled by oblique gypsum fibres. The displacement of these structures is unknown. (4) S5 and S6 comprise two orthogonal sets of nonstratabound joints. S5, composed by SE-dipping joints with cm to dm spacing, is the most penetrative set, whereas S6 (SW dipping) shows a scattered distribution throughout. The most penetrative surfaces (S5) are usually enlarged by dissolution, resulting in irregular boundaries. Openings by dissolution in some joints can reach 80 cm. Fracture porosity can be totally or partially filled by macrocrystalline gypsum cements or uncemented in the least penetrative planes.

GYPSUM VEINS AND CEMENTS
All fractures are filled by very pure gypsum crystals with 4 different microstructures and arrangements (Figs. 5 and 6): Sigmoidal fibrous veins, perpendicular fibrous veins, oblique gypsum veins and macrocrystalline cements. Macrocrystalline fillings are considered as cement, understanding 'cements' as new minerals precipitated in pore spaces (Tucker 1981;Boggs 2003). In the studied outcrops, the pore space is fracture and cavern porosity types (Choquette & Pray 1970;Tucker & Wright 1991;Moore 2001). The remainder gypsum precipitates in fractures are called gypsum veins, according to the vein definition as 'mineral aggregates that precipitated from a fluid in dilatational sites' (Bons et al. 2012). (1) Sigmoidal fibrous veins. These veins consist of fibres from 1 to 5 cm long filling S1 and S2 faults. Fibres show curved morphology and oblique orientation to the fault walls (Figs. 5A and B), with no visible median line in the veins. The fibrous crystals show constant thickness from fracture centre to fracture wall, precipitating in optical continuity and constant crystallographic orientation, and lack of wavy extinction and internal breakage (Fig. 5B). The fibres are commonly limited by microfractures and Riedel Y-shears within the fracture planes, indicating several growing fracturation events and a complex infilling history of the veins. Small elongated gypsum crystals are observed near the wall rock; these crystals, which are aligned parallel to the fault plane, are interpreted as broken sigmoidal fibres (Fig. 5B). A few host gypsum fragments, up to 1 mm in size, are randomly embedded within the fibres, but always close to the fracture walls, indicating short displacement. Chips are subangular and isolated, uncommon in secondary micro-to macrocrystalline gypsum with carbonate matrix.
(2) Perpendicular fibrous veins (satin spar). These veins are characterized by fibrous gypsum crystals from 250 lm to 0.5 cm in size, growing perpendicular to the S3 fracture walls (Figs. 5C and D). The crystal contacts are straight and display internal fracturation and wavy extinction. The S3 veins sometimes display a median zone, parallel to the vein walls, dividing the gypsum fillings into two isopachous seams, equal in thickness and symmetrical (Fig. 5D). On a microscopic scale, the parting is composed of microcrystalline secondary gypsum.
(3) Oblique gypsum veins. These veins are mainly composed of fibres from <1 mm to 2 cm in length distributed parallel or slightly oblique to the walls of the S4 fracture planes. Several morphologies can be differentiated under microscope: short aligned crystals, large fibres and subeuhedral crystals. The aligned crystals (Fig. 6A) are short elongated gypsum crystals with wavy extinction close to the fracture walls, always in contact with the host gypsum. The large fibres are slightly oblique to the fault walls with wavy extinction, evolving gradually to subeuhedral elongated morphologies with irregular boundaries and wavy extinction. (4) Macrocrystalline cement. Euhedral (lenticular) to subeuhedral, but not twinned, gypsum, up to 20 cm, is found isotropically distributed mainly within S5 (and occasionally in S6) joints, displaying no changes in size from the wall to the centre of the cavities (Fig. 6B).
Most of the joints have not been enlarged by dissolution and remain noncemented, but the most penetrative ones (S5) are usually enlarged by dissolution (creating cavern porosity) and are partially or totally cemented by large lenticular gypsum. The lenticules mainly precipitate along the c-axis parallel to the walls of the fractures, but are not developed completely where the fractures display little opening. These same macrocrystalline cements also infill some S1 and S2 fractures, accompanying the fibrous gypsum. In such cases, the macrocrystals mainly occur in the S1-S2 cross-points, in which fracture porosity has been enlarged by dissolution. Macrocrystals show only local wavy extinction. Small euhedral gypsum crystals usually appear in the contact between the macrocrystals (suggesting crystal boundary recrystallization), and such recrystallized particles show oblique alignment (suggesting coeval movement during recrystallization) (Fig. 6C).

GEOCHEMISTRY Sulphur and oxygen isotopes
Sulphur (d 34 S) isotope analyses of the host gypsum rocks display a range of compositions from +20.9 to +22.3, in the basal secondary gypsum subunit, and between +17.8 and +21.8 in the upper gypsarenite-microselenite subunit. For oxygen (d 18 O), secondary gypsum rocks show a range between +12.9 and +16.2, while the primary lithofacies are in a range from +15.6 to +18.6. These values are mostly within the range of the preexisting data (Agust ı et al. 1990), but some samples of the primary subunit offer lighter sulphur compositions (Fig. 7, Table 1).
The oxygen and sulphur isotope compositions of gypsum infillings vary between +14.3 and +19.8, and from +21.8 to +25.1, respectively. Contrary to the host gypsum, which values plot in a dispersed area, the compositions of the infillings display a relatively good positive correlation. The d 34 S values clearly lighter than those of the host gypsum (Fig. 7, Table 1). Additionally, two groups of values are recorded within the infillings; the sigmoidal fibrous gypsum show lighter values, and the macrocrystalline cements display the heavier compositions (and also displaying a wider range of compositions).

Sr/ 86 Sr analysis
Rb contents range from 32 to 285 ppb, while a wider range of Sr contents is defined (between 246 and 3498 ppm). The strontium isotope ratios are in a narrow range between 0.70863 and 0.70874. The most radiogenic values are in general displayed by the host gypsum rocks, while the sigmoidal fibrous gypsum offers the lowest one; the macrocrystalline cements show intermediate values (Fig. 8, Table 1).

DISCUSSION Evaporite sedimentation and early diagenesis
The Vilob ı Gypsum Unit succession is composed by two subunits (Fig. 2). According to Ort ı & Pueyo (1976), the lower subunit is interpreted as precipitated in a coastal salina (laminated-to-banded gypsum lithofacies), which continuously evolved to a sabkha environment (enterolithic and nodular-micronodular anhydrite lithofacies). The growth of these anhydrite nodules is related to synsedimentary-early diagenetic processes in the sabkha environment. Nevertheless, this Lower Subunit was formed by secondary gypsum, after hydration of precursor anhydrite, evidenced by the gypsum textures (alabastrine and radial aggregates) and the presence of anhydrite relics. The mean isotopic signals for the lower gypsum are +21.5 and +14.3 for sulphur and oxygen, respectively, and the strontium isotope ratios are between 0.70863 and 0.70873 (Table 1). The S-O and Sr isotopic compositions of the Vilob ı Gypsum suggest predominantly marine origin of the brines (Fig. 7). The Sr contents of these samples oscillate from 2240 to 3500 ppm, which is clearly due to celestite contamination.
The Upper Subunit in the Vilob ı Gypsum is composed of a thin (few metres thick) interval of primary (depositional) gypsum, including laminated-to-banded microselenitic and gypsarenitic lithofacies, revealing a sedimentological change in the basin. Ort ı & Pueyo (1976) interpreted this upper subunit as deposited in a coastal salina environment. Moreover, the sharp contact at its base can be interpreted as an

Deformation and infilling-cementation stages
The deformation history affecting the Vilob ı Gypsum Unit can be divided in five events, all related to the major Neogene extensional event that occurred in the NW margin of the Mediterranean Sea (Fig. 9).
(1) S1-S2 faults: These are the main normal faults affecting the evaporite unit and the oldest fragile structure recorded. They are totally obliterated by sigmoidal gypsum fibres (stacking of several growing events limited by fractures), coexisting with macrocrystalline cements. Fibrous crystals growing oblique to the fracture walls and lacking wavy extinction and internal breakage suggest coeval movement of fractures during crystallization (El Tabakh et al. 1998). Fibrous morphologies are usually linked to a crystal growth rate occurring at the same rate as fracture opening (Machel 1985;El Tabakh et al. 1998). The studied gypsum fillings display a clear asymptotic trend of the fibres towards the walls supporting syntectonic precipitation. The small elongated crystals distributed parallel to the fracture wall with undulant extinction (Fig. 5), observed in some S1 and S2 fractures, could correspond to a fragmentation of previous crystals during this extensive event or later during the compressive event.
The main trend of the S1-S2 fractures and the direction of their movement are well correlated with the main basin structures (normal faults with NW-SE trend) produced during the Neogene rifting that affected the Vall es-Pened es basin (Banda & Santanach 1992;Bartrina et al. 1992). Taking into account that the evaporite unit was deposited during the upper Burdigalian (Cabrera 1981), these structures could not have developed during the syn-rift event (late Oligocene to early Burdigalian), but the evaporites probably fractured during the syn-rift to postrift transition (late Burdigalian to early Langhian), which is in accordance with a similar deformation event defined by (Baqu es et al. 2012b).
(2) S3 joints: A gentle tectonic dilation was recorded by the subvertical S3 joints, in the latest early postrift stage, characterized by extension and a general decrease of tectonic activity. These small scale joints (S3) are filled by perpendicular fibrous crystals. Fibrous gypsum grows perpendicular to wall fractures symmetrically from a median zone. This vein type, generally called satin spar in the literature, commonly referred  to subhorizontal to slightly inclined planes (Machel 1985;Gustavson et al. 1994;El Tabakh et al. 1998;among others). In the S3 joints, however, they display a subvertical distribution. (3) S4 faults: These thrust faults show the same trend as the S1-S2 faults (Fig. 3) and developed taking advantage of previous normal faults (S1 and S2). Minor compressive events have been described in the Vall es-Pened es basin during both the upper Langhian-lower Serravallian and the Messinian (Bartrina et al. 1992;Roca 1994;Trav e & Calvet 2001). Taking into account the crosscutting relationship between the six fracture sets defined in the Vilob ı Gypsum Unit, the upper Langhian-lower Serravallian compressive event is the more reliable as the reactivation episode of the S1-S2 normal faults that promoted the formation of S4 thrust. This chronology is in agreement with the chronology established at basin scale by Baqu es et al. .
At a microscope scale, compressive deformation is also observed in the fillings of the S4 fractures, which are mainly filled by oblique crystals, mainly composed of fibrous gypsum crystals orientated in a very low angle to the fracture walls and subeuhedral crystals with wavy extinction. These fibrous crystals developed synchronously to the evolution of S4 thrust. Their distribution (low angle to the wall vein) and morphology are linked to field stresses and low dilation rate linked to these structures, where fibrous growth direction is almost parallel to the vein offset direction (Bons & Montenari 2005).
The subeuhedral crystals are interpreted as previously formed crystals, which were deformed during inverse fault activity, thus modifying their original distribution. The wavy extinction supports this interpretation, as is also indicated by a posterior deformation of the crystal (Hilgers and Urai 2002). However, it is impossible to determine whether the deformed crystals were formed during the same tectonic event (early formed fibres) or whether they come from a previous stage (S1 and S2 fillings). (4) S5-S6 joints: After the minor compressive event, a general decrease in tectonic activity occurred in the studied area (Bartrina et al. 1992;Roca 1994), as recorded by S5 and S6 joints. Likewise, a general dissolution of the host rock in the fracture sectors is observed around outcrops, this dissolution process being more intense in intersection of several fractures and in the most penetrative S5 joints. In this context, large fracture opening allowed the precipitation of the macrocrystalline cements. Moreover, the crosscutting relationships between S5-S6 joints and S4 faults reveal that the joints developed after the lower Serravallian. The age of the dissolution is unknown, but could be tentatively attributed to the Messinian, taking into account the generalized sea level decrease in the Mediterranean area and the large dissolution processes occurring in the Vall es-Pened es basin due to the phreatic level fall (Baqu es et al. 2012a).

Fluid flow evolution
Both the micro-to macrocrystalline gypsum of the Lower Subunit (Fig. 2) are diagenetic gypsum textures overprinted on the original laminated and nodular lithofacies, as indicated by the textural varieties of gypsum and the extensive presence of anhydrite relicts. Nevertheless, such laminated to nodular-enterolithic relict lithofacies suggest that the unit was precipitated in a marine sabkha-salina environment that evolved to salina deposits (gypsarenite and microselenite lithofacies) towards the top, probably coupled with the input of nonmarine fluids. The d 34 S and d 18 O values of cements and vein fillings show a general isotopic enrichment trend, going from similar host rock isotopic values to heavier ones. The oldest fibrous gypsum, linked to the S1-S2 fractures (mean of +22.5 and +15.6 for sulphur and oxygen), shows the most similar isotopic values to those of the host gypsum rocks, which are close to Tertiary marine evaporite signatures. The macrocrystalline cements, which were precipitated in the latest deformation stages, show the most enriched isotopic values (mean of +24.6 and +18.3 for d 34 S and d 18 O) (Fig. 7).
The observed tendency is easily explained by in situ chemical recycling, not by external basinal brine input. External brines could come into the area from the dissolution of Triassic evaporates, in which case the d 34 S would be more depleted than the obtained values. Fluids coming from interaction with Triassic evaporites in Spanish basins show d 34 S values between 14 and 16 according to Utrilla et al. (1992). The evolution of the latest cements to heavier values points to a continuous process of host dissolu-tion and reprecipitation, taking into account that gypsum precipitation processes produce a sulphur fractionation of +1.65‰ (Thode & Monster 1965), whereas d 18 O enrichment is around +3.5‰ (Claypool et al. 1980). The higher oxygen enrichment, with respect to the sulphur enrichment, suggests successive chemical recycling processes from the host gypsum, the vein precipitates and finally the macrocrystalline cements.
The main source of the calcium sulphate ions is dissolution of the host Vilob ı Gypsum Unit. The calcium sulphate-rich parental fluids flowed along the fractures, some of them enlarged by dissolution, recording changes in the isotopic signatures of the new gypsum precipitates. The source of Ca-SO 4 saturated fluids could have been first related to the early diagenetic anhydrite hydration. Such a process has been invoked in the literature as being responsible for some fibrous gypsum precipitates (Shearman et al. 1972). Nevertheless, the Vilob ı Unit was tight, lithified and transformed to secondary gypsum (chips of host secondary gypsum are embedded within the veins fillings) when it started to fracture during the early postrift stage. This suggests that the anhydrite-gypsum transformation occurred before the fracture processes and that the Ca-SO 4 enriched fluids had already escaped. The absence of satin spar gypsum veins in the underlying and overlying lutite units supports this hypothesis. Initial sea water-related input in the first fibrous crystals cannot be ruled out, considering their marine-like sulphate isotopic signatures and the presence of Langhian marine facies close to the studied area, which coincide in age with the development of the S1-S2 faults and their coeval fibrous fillings (Baqu es et al. 2012a). Sulphate chemical recycling and the onset of gypsum precipitation in fractures could have occurred in meteoric fluids and/or marine-related fluids.
The Sr contents in gypsum could come from direct incorporation from brine to crystals during crystallization or from external Sr inputs by means of celestite contamination or rubidium decay (Rosell et al. 1998;Play a & Rosell 2005). Rb and Sr contents were analysed to guarantee the Sr origin. Taking into account that no contamination by clay or celestite, evidenced by lower Sr content, occurs in the analysed strontium rations of the infillings (from 200 to 400 ppm), nevertheless, slight differences exist between the fibrous crystals and the macrocrystalline cements (Fig. 8). For example, the 87 Sr/ 86 Sr values in the fibrous precipitates (0.70863) are less radiogenic than those of the macrocrystalline cements (values between 0.70869 and 0.70872). These isotopic signals fit with those of the marine Vilob ı Gypsum Unit, thus reinforcing the chemical recycling of these gypsum host rocks.
Macrocrystalline gypsum grew in highly saturated calcium sulphate brines, promoting a reduced number of nuclei, slow crystallization rates and well-developed (euhedral) gypsum crystals. The cavern porosity (enlarged fractures) is filled by randomly oriented lenticular crystals indicating no stratification of the brines and continuous flowing through the fractures. Because achievement of Casulphate saturation in the brines is the key factor in the precipitation of these cements, this must be related to dissolution processes in gypsum rocks, not to evaporation processes. Slow fluid circulation rates along fractures and high fluid-host gypsum rock interaction are evidenced, promoting host gypsum dissolution. As the residence time of the brines is long suggesting a closed palaeohydrological system with a prolonged fluid-rock interactions with the host gypsum rocks. The more radiogenic strontium ratios in the macrocrystalline cements can be related to the presence of more radiogenic detrital components in the host gypsum rock, thus increasing the 87 Sr/ 86 Sr signatures of the cements.

CONCLUSIONS
The purpose of this study was to establish the parental fluids of the fracture infillings of the Vilob ı Gypsum Unit and their evolution during the different precipitation events. Our results suggest that Vilob ı Gypsum Unit was affected by a series of postrift deformation events of Neogene extension. As a result, six fracture sets formed at four different deformation stages linked to basin-scale events. In the early postrift (Langhian), S1 and S2 conjugated normal faults, and S3 joints were developed. Whereas S1 and S2 were totally filled by sigmoidal fibrous gypsum, which grew coeval to fault development, S3 was totally filled by perpendicular fibrous gypsum (satin spar gypsum veins).
Later, a minor compression occurred (lower Serravallian) originating the S4 inverse faults with the same orientation as S1 and S2 faults, most of which filled with oblique fibrous gypsum, which developed synchronously to the evolution of S4 fractures. Subeuhedral crystals, interpreted as previously formed crystals deformed during inverse fault activity, can be also related to these fracture infillings.
Finally, during the late postrift (upper Serravallian-Pliocene), the orthogonal system of the S5 and S6 joints were formed, followed by a general dissolution and a decrease in the tectonic activity in the basin. The most penetrative of these joints (S5) are totally or partially cemented by macrocrytalline gypsum, whereas the least penetrative (S6) remained uncemented.
Isotopic analysis provide compelling evidence for a closed (convective) hydrologic system associated with the formation of the fracture-filling cements. The d 34 S and d 18 O values of cements and vein fillings show a general isotopic enrichment trend, going from similar host rock isotopic values to heavier ones. The oldest fibrous gypsum, linked to the S1-S2 fractures, shows the most similar isotopic values to the values of the host gypsum rocks, which are close to Tertiary marine evaporite signatures. The macrocrystalline cements, which were precipitated in the latest deformation stages, show the most enriched isotopic values. This trend reveals chemical recycling processes linked to fluid flow along the fractures dissolving the closest sulphates. Each successive infilling event resulted from recycling of the preexisting gypsum. Strontium isotopic signals of fracture gypsum infillings fit with those of the marine Vilob ı Gypsum host rocks, thus reinforcing the chemical recycling of these gypsum host rocks.