Magmatic volatiles episodically flush oceanic hydrothermal systems as recorded by zoned epidote

Circulation of seawater at oceanic spreading centers extracts heat, drives rock alteration, and transports leached metals to shallower levels of the crust, where they may precipitate and form ore deposits. Crystallization of the lower crust, may exsolve and introduce magmatic volatiles into the seawater-dominant system. However, the role of magmatic volatiles added to the hydrothermal system, including pathways of these fluids are lesser known. Here we present coupled in-situ strontium isotope and rare earth element data of distinct domains in epidote, a common hydrothermal mineral throughout the Troodos ophiolite, to track magmatic fluid input and flow. Epidote crystal domains characterize three distinct strontium isotope-rare earth element signatures—suggesting sequential growth from magma-derived fluids (0.704, negative europium anomalies), rock-buffered fluids (0.7055, positive europium anomalies) and seawater-derived fluids (0.7065, negative cerium anomalies). Epidote records episodic fluxing of magmatic fluids from plagiogranites, through epidosites in the upflow zone and into metal ore deposits. Deep-sea hydrothermal systems are episodically flushed by fluids of magmatic origin, as revealed by the chemical signatures of zoned epidote crystals from the Troodos Ophiolite, Cyprus

S ince the first discovery of submarine hydrothermal vents (black-smokers), at the East Pacific Rise in the late 1970's 1 , hydrothermal mineralization has been found abundantly along earth oceanic spreading centers 2 . The modern fluid venting from seafloor hydrothermal fields is predominantly seawater, heated by crystallizing magma emplaced in the lower crust 3 . Metals in ocean floor hydrothermal vents are thought to originate mostly from leaching of crustal rocks by heated seawater and subsequently precipitated as sulfides, when the discharging hydrothermal fluids mix with cold, metal-depleted seawater 4,5 .
Magma-derived fluids added to the hydrothermal cell may contribute to the formation of ocean-floor sulfide deposits 6 . Favorable settings to test this hypothesis are Back Arc Basins (BAB) because volatile-rich felsic magmas, which concentrate many metals as they evolve, are much more common in back-arc settings than along mid-ocean ridges 7 . A particularly useful tool to identify the contribution of volatiles exsolved from silicic magmas to black smoker fluids are the set of Rare Earth Elements (REE). In Mid-Ocean Ridge (MOR) settings, the chondritenormalized REE patterns of hydrothermal fluids are universally characterized by a strong positive Eu anomaly and LREE (light REE, La-Sm) enrichment 8,9 . This pattern stems from preferential complexing of the LREE and Eu 2+ with chlorine, a major constituent of seawater 10 . However, recent discoveries of HREE (heavy REE, Gd-Lu) enrichment and weaker Eu anomalies in back-arc basin (BAB) fluids are explained by REE complexing with other ligands, such as fluoride and sulfate, under varying pH conditions 10,11 . The source of these fluids was inferred to be degassing of silicic magmas 12 , but the exact identity of these silicic magmas has yet to be described. Furthermore, whether BABhosted complexing ligands only affect REE leaching from solid rocks or in fact transport the REE from the silicic magmas directly to the hydrothermal system is still unresolved.
The lower oceanic crust occurs at depths of ≥2 km below the seafloor, and hence the rootzones of the present-day oceanic hydrothermal systems are rarely accessible. Ancient oceanic crust exposed on-land in ophiolites is thus invaluable for assessing the role of magmatic fluids in oceanic hydrothermal systems and seafloor mineralization. In this respect, the Troodos ophiolite of Cyprus is a key exposure: (1) a well-preserved, intact crustal section is exposed, allowing the inspection of the ancient hydrothermal fingerprint in its full extent from rootzone through upflow zones to vents; (2) the most significant evidence for Volcanogenic Massive Sulfide (VMS) ores in ophiolites being ancient analogs of black smokers comes from Troodos 13,14 . Moreover, negative δ 34 S values measured in sulfides from a few VMS deposits in Troodos suggest magmatic sulfur influx; 15 (3) Troodos is a supra-subduction zone (SSZ) ophiolite, representing BAB-type crust 16,17 and predictably contains conspicuous silicic intrusions, commonly known as plagiogranites.
The Troodos ophiolite (~92 Ma) 18,19 comprises mantle peridotites, overlain by a crustal section made of, from bottom to top (i) gabbros and minor plagiogranites, (ii) sheeted diabase dykes and (iii) lavas, mostly pillowed (Fig. 1). The VMS deposits of Cyprus, the type location for such worldwide Cyprus-style copper ore deposits, mainly occur near the top of the volcanic section of the ophiolite. Their geometry and alteration zonation are similar to hydrothermal vents at modern spreading centers 4,20 . The Cu, Zn, Pb and minor amounts of precious metals found in the VMS deposits are thought to derive from diabase dykes below the deposits by leaching along intensive hydrothermal upflow zones 21,22 . Mass balance calculations show that the base metal quota hosted in diabase is sufficient to supply the overlying VMS deposits, but requires fluid/rock interactions at substantially higher ratios than deduced for modern black smoker environments 5,23 . The leached zones are often seen as lenticular bodies rich in epidote-quartz ± chlorite assemblage in the center of dykes. In areas where the alteration is most destructive, pure epidote and quartz encompass the entire rock, known as epidosite 24 . These rocks were hypothesized to form where seawaterderived hydrothermal fluids are focused into narrow upflow pipes and epidotize the dike rocks at high water-to-rock ratios of 20 to 1000 23,25 . This model is petrologically implausible, because fluids that are multiply saturated with 5-6 greenschist facies minerals in the root zones are unlikely to dissolve all but two of those phases upon initial upwelling 26 . Thermodynamic modeling showed that epidosite formation requires some influx of hard mineral acids, like that seen in exsolved magmatic volatiles from siliceous magmas 26 .
Fluid inclusions trapped in gabbros and plagiogranites of the Troodos ophiolite include dense brines of up to 60 wt.% NaCl, interpreted as derived from magmatic fluid sources 27 . The highly saline inclusion trails are not associated with low-salinity vapor phase inclusions, indicating derivation by direct exsolution from magma and not by phase separation of a seawater-derived fluid. Secondary epidote, especially abundant in Troodos plagiogranites, turns out to be excellent tracer of the evolution of fluids exsolved from magmas 28 . Hydrothermal epidote, precipitated in miarolites in plagiogranites, records REE-rich and Eu-depleted fluids exsolved from silicic magma early on, which gradually transform into the REE-depleted and Eu-enriched pattern prevalent throughout conventional seawater-derived sub-seafloor fluids. Thus, epidotization of plagiogranites is initially an autometasomatic process, whereby the rock is altered by supercritical fluids exsolving from the crystallizing magma rather than by seawater-derived fluids.
The question posed here is whether exsolved magmatic fluids are able to migrate through the plagiogranites into nearby and overlying rock suites, the sheeted diabase dykes, which are considered the root zone of the oceanic hydrothermal system [29][30][31][32] . Furthermore, do magmatic fluids enhance alteration of the middle and upper oceanic crust and contribute to mineralization on the ocean floor?

Results
Using Epidote to trace hydrothermal fluids. Epidote in the ocean crust is stable over a wide range of conditions 33 and potentially an excellent tracer mineral for the evolution of the sub-seafloor oceanic hydrothermal system. Epidote is also a strontium (Sr) and REE-bearing mineral, and since the possible end-members of circulating fluids, e.g. seawater and magmatic water, strongly differ in their REE contents and patterns as well as Sr isotope ratios, it makes a perfect recorder of relative contributions of these two fluid sources.
Textural types of epidote represent various stages of hydrothermal alteration of the Troodos crustal section: (a) epidote formed at the expense of plagioclase in the sheeted dykes as part of greenschist facies alteration; (b) hydrothermal epidote overgrowing and, in some cases, totally replacing type a; (c) precipitates in veins, miarolites and amygdules. Often all three textural types of epidote can occur in a single thin section. In-situ REE contents and Sr isotope ratios were measured by laser ablation inductively-coupled plasma mass spectrometry (LA-ICP-MS) techniques (quadrupole and multiple-collection sector MS, respectively) in epidote crystals from various crustal levels of the Troodos ophiolite. Sr isotope ratios of bulk epidote separates from rock matrices, veins and amygdules of the same samples chosen for in-situ analysis were measured by thermal ionization mass spectrometry ( Sr isotope-REE coupling in epidotes from miarolites and epidotized regions in plagiogranites are similar to that observed in epidosite matrix and amygdules in shallower rocks, however, in the plagiogranites (M)-like lower Sr isotope ratios and negative Eu anomalies are more abundant (Fig. 3a-f). Epidote-quartz veins ( Fig. 3g), sharply crosscutting plagiogranites and gabbros, are accompanied by a metasomatic front that emanates from the vein into the host rock causing a secondary epidotization alteration. Massive epidote from the host rock matrix is characterized by (M)-like Sr isotope ratios of 0.7039-0.7044; ratios in the vein are also mostly (M)-like, but some crystals yielded higher (H)-type and even (MS)-type values (Fig. 3h). REE patterns of the vein epidote corresponds to (M)-and (MS)-like patterns of epidote found in basalts, epidosites and plagiogranites (Fig. 3i).

Discussion
The large range in whole-rock Sr isotope ratios previously measured in the extrusive suite of Troodos 29,34 , 0.703 to 0.707, was interpreted to represent variable degrees of low-temperature alteration resulting in mixing between fresh MORB glass, 87 Sr/ 86 Sr = 0.7030, and Sr derived from Cretaceous seawater 35 , 0.7073 (Fig. 4). Nonetheless, going deeper in the crust into epidosites of the sheeted diabase dyke layer, this range significantly narrows down to 0.7050-0.7055. These relatively uniform whole rock values were attributed to interaction with an average Troodos hydrothermal fluid-a seawater-derived fluid, buffered by the mafic crust at the recharge zones of the hydrothermal cell 29 Fig. 4) and indicate that no single fluid could have formed the Troodos epidosites. Whole-rock and bulk epidote measurements tend to average out chemical and isotope zoning and thus cannot detect the origin or temporal variation of epidosite-forming fluids. The draw-back of bulk-rock analyses averaging grain-scale information is also known from other isotope studies of water-rock interaction. For example, in-situ measurements of oxygen isotope heterogeneities between grain centers and boundaries of highgrade metamorphic calcite from Naxos, Greece, have been shown to be averaged out by whole-rock powder analysis 36 , thus concealing distinctive late meteoric fluids and their pathways along cracks and grain boundaries. The chromatographic continuum model previously suggested to account for gradual whole rock isotope evolution towards that of an incoming fluid 37 may be useful for a single episode of infiltration of a homogenous fluid. However, in complex cases where several generations of fluid infiltration are indicated by in-situ analysis, for example in Naxos marbles 37 and the Troodos ophiolite 29 , the model is not compatible, and instead higher resolution grain-scale profiles should be used.
In general, REE patterns of epidotes from worldwide magmatic and metamorphic rocks display flat to LREE-enriched patterns 38 . However, REE patterns of epidote from metasomatic environments, like amygdules, miarolites and veins, are extremely variable 39 . Diverse REE patterns have been previously recorded in hydrothermal epidote from Troodos ( Fig. 5a): (1) whole-grain groundmass-replacing epidote from greenschist facies diabase has flat patterns that are parallel to fresh-rock patterns, albeit with enriched absolute-REE contents 40 ; (2) epidote vug-precipitates show LREE-enrichment, a positive Eu anomaly and flat HREE 40 (Fig. 5a); REE patterns (and 87 Sr/ 86 Sr) of vug-epidote thus resemble those of modern hydrothermal vent fluids (Fig. 5b); (3) in situ analysis of miarolite-epidote cores display HREEenrichment with negative Eu-anomalies resembling the REE pattern of the whole-rock host plagiogranite 28 . These epidote cores may be interpreted as precipitates of an exsolved magmatic fluid. The high temperatures and carrier ligand concentrations characteristic of magmatic fluids are expected to enable transportation of at least ppm-level REE contents 41 and up to 1300ppm in some cases 42 , thus providing a source for REE-bearing precipitates.
REE patterns of several source fluids that may have been involved in epidote-precipitation are given in Fig. 5b. The granophyre and aplite REE patterns, determined by crush-leach extraction analysis of quartz-hosted fluid inclusions in felsic rocks, are LREE enriched and have negative Eu anomalies 42 . These may be representative of early, Eu-depleted, magmatic-derived fluids and resemble (M)-type patterns in epidote. The Pacmanus and Vienna Woods vent fluids from modern BAB hydrothermal systems 9 , used as analogs for ancient vent fluids, resemble the patterns of (H) domains albeit with slightly increased LREE ≥ HREE. Finally, the pattern of modern seawater 9 , used as an analog for Cretaceous seawater, closely resembles the (MS)-type patterns with low REE contents and negative Ce anomalies; although in seawater positive Eu anomaly is absent. Our in-situ analyses show that regardless of crustal depth, host rock type, and textural context, epidote growth domains can be classified by their coupled REE-Sr isotope characteristics into three major types: M-H-and MS-type fluid sources (Figs. 4, 5). Miarolites are cavities forming during the late stages of granite magma crystallization, often trapping exsolved magmatic fluids. Expectedly, the cores of miarolitic epidote from our study are characterized by REE patterns, which are similar to those measured in plagiogranites (Figs. 3f, 5a) and their Sr isotope ratios, 0.703-0.704, coincide with the whole rock values of fresh gabbros in Troodos 29 (Fig. 4). On the other edge of the fluid spectrum, the MS-type domains, which mostly occur at epidote rims, are most likely precipitated from seawater, as indicated by their negative Ce anomaly. However, the positive Eu anomaly coupled with Sr isotope ratios of~0.706 suggest some modification of seawater by low water/rock ratio interaction with the mafic crust prior to epidote precipitation.
The most dominant fluid recorded by epidote is H-type, which is intermediate in REE contents and Sr isotope ratios with respect to the M-and MS-type fluids, and unanimously has a positive Eu anomaly. There are three possible scenarios that may account for the origin of H-type fluids. (1) The zoning of miarolite epidote from M-to H-type in REE patterns may record the evolution of magmatic fluids in a closed system 28 , however since the Sr isotope ratio is significantly higher at the epidote rim, reaching 0.7052 (Fig. 3b), mixing of the magmatic fluid with an external one is more probable. (2) Mixing of magmatic water (M-type) and seawater (MS-type) at approximately equal amounts will produce the measured H-type REE-Sr values. (3) Seawater infiltration through the recharge zones involving prolonged interaction with diabase may result in averaging the Sr isotope ratio between Cretaceous seawater (0.7073) and mafic crust (0.703), forming an average Troodos hydrothermal fluid 29 . It would also produce fluids with positive Eu anomalies due to albitization of plagioclase. H-type domains have this characteristic REE-Sr coupling (Fig. 2, 3)), and may thus represent this large-scale rock-buffered seawater circulation. It follows that three distinct types of fluids sequentially flowed through the Troodos hydrothermal cell.
While epidosites are rarely found in modern oceanic crust 43 , a bevy of epidosite alteration is recorded in Cretaceous ophiolites 44 (e.g. Troodos and Oman). It has been suggested that the higher Ca/Mg ratios 45   suggest that epidosites form by overprinting of diabase that was previously altered to greenschist facies mineral assemblage 23 . More recently, some epidosites were shown to precipitate into a newly-formed porosity, interpreted to form by dissolution of primary magmatic diabase by black-smoker hydrothermal fluid 44 .
Our study of the epidosite paragenesis show that M-type epidote cores, precipitated by magmatic fluids, are truncated and later overgrown by euhedral rims precipitated by H-type fluids (Fig. 2g-i). Thus, irrespective of whether the initial rock was a primary mafic diabase or greenschist-facies assemblage, the introduction of acidic magmatic fluids (M-type), like those found venting at modern supra-subduction spreading environments 9 , was necessarily part of the epidotization. Eventually, a dissipation of the magmatic fluid led to the influx of H-type fluid, partial resorption of the core due to disequilibrium and precipitation of euhedral H-type rims into the newly-formed porosity. While this may be the major mode of epidotization, some epidote crystals bear M-type characteristics throughout their growth. These may indicate regions in the dyke where permeability was low. The consensus model for hydrothermal circulation in oceanic spreading centers involves deeply penetrating seawater interacting with diabase to obtain near magmatic Sr isotope ratios 29 . Upon fluid ejection through the upflow zones, the forming epidosites are supposed to inherit the near magmatic-like leached diabase isotope signature (Fig. 6, point 1). This is not likely the case as seen in Troodos, also applicable to other worldwide epidositebearing ocean crust sections, because (a) seawater-derived fluids cannot drive the immense mass transfer that is necessary to form intensive epidosite alteration. This may be the reason for the lack of documented epidosites from modern mid-ocean ridges. However, ingassing of hard mineral acids, such as HCL and H 2 SO 4 , found venting at modern supra-subduction environments, may destabilize chlorite, actinolite, and albite and leach Mg from the rocks 26 . (b) domains of zoned epidote crystals in epidosites, but also at shallower and deeper levels of the crust, are characterized by coupled REE-Sr isotope signatures indicative of sequential precipitation from magmatic, hydrothermal and modified seawater fluids. Moreover, the abundant domains with H-type intermediate Sr isotope ratio of~0.705 are not necessarily derived from diabase-buffered fluid, but may have precipitated from a mixed seawater-magmatic fluid (Fig. 6, point 2, 3). The occurrence of M-type domains in epidote from the VMS deposits suggest mixing of the magmatic fluids with the H-and MS-type fluids may have been limited at times (Fig. 6, point 4). This process involves magmatic fluids episodically flooding the seawater-derived hydrothermal cell circulating through the upper crust. A possible mechanism for the pulsating magmatic fluid supply is volatile pressure build-up eventually leading to breach of the magma chamber, as commonly invoked to explain massive volatile degassing in oceanic dyke injection 46 and terrestrial volcanoes prior to eruption (Fig. 6, point 3). While episodic pulses of magmatic volatiles prevail, some epidote domains, even in deepseated rocks, i.e. epidosites and plagiogranite, were precipitated from slightly modified seawater. This requires minimum interaction with rocks and thus rapid draw-down of seawater to~2 km depth in the crust. Such intense downward seawater flow may be localized along steep normal faults during episodes of waning magmatic activity (Fig. 6, point 2).
With this model we suggest there were at least three types of fluids in circulation feeding the on-axis hydrothermal system of the Troodos oceanic crust-a modified seawater-derived fluid (MS-type), a magmatic fluid (M-type) and, a hydrothermal fluid (H-type)-possibly composed of the M-and MS-types, but most likely a seawater-derived rock-buffered fluid. The sequential Mto H-type zonation in epidote from the upflow zones (epidosites) suggests, magmatic fluids episodically flush the hydrothermal system, which is later dominated by a H-type fluid.

Methods
TIMS-whole grain strontium isotope ratio. For strontium isotope measurements, epidote separates (1.1 to 3.5 mg) were brought into solution by two consecutive steps. At first, the separates were digested in 0.5 ml of a mixture of concentrated HF and HNO 3 (5:1) and were placed on a hot plate for 42 h at 140°C. Secondly, the solutions were dried at 95°C and the residuum were re-dissolved in 1 ml 7 M HNO 3 , placed on the hot plate for 24 hours, and afterwards dried at 90°C. Strontium was purified by using 70 µl Sr. spec™ resin in miniaturized columns 47 . The samples were digested in 0.5 ml 2 M HNO 3 and loaded on the columns in 100 µl steps. Following, unwanted elements were removed by adding 1.2 ml of 2 M HNO 3 in 100 µl steps, 1 ml 7 M HNO 3 in 500 µl steps and 0.3 ml 2 M    LA-ICP-MS-in-situ trace element contents. The determination of trace element concentrations from thin sections was performed by using a Thermo-Scientific Element 2 inductively coupled plasma-mass spectrometer (ICP-MS) coupled to a NewWave UP193 laser ablation system at the Faculty of Geosciences, University of Bremen. The irradiance for ablation was set to 1 GW/cm 2 and for spot measurements the beam diameter was set to between 75 µm and 50 µm depending on sample size. Pulse rates of 5 Hz were used. The carrier gas was helium and argon was added as make-up gas, both with 0.8 l/min. NIST 610 was used as the external calibration standard and Ca as the internal standard previously measured by EMPA. The computation of the trace element concentrations was performed with the Cetac GeoPro TM software. The reference materials BCR-2G and BHVO-2G were measured at least every 20 measurements to ensure the analytical precision and accuracy. The relative standard deviation (RSD) of the reference material was used as an indicator for precision.
LA-MC-ICP-MS-in-situ strontium isotope ratio. Epidote crystals were polished, imaged and analyzed either in-situ in petrographic thin sections or as separated grains mounted in 1-inch epoxy rounds. 87 [48][49][50] . MPI DING iRM StHs6/80-G (St. Helens Andesitic Ash Glass) and in-house White Sturgeon (Acipenser transmontanus) pectoral fin ray named 'SSFR' (from Sterling Caviar Aqua Farm, CA, USA) were also measured in this study. StHs6/80-G was used to evaluate potential polyatomic spectral interferences (predominately CaCa+ and FeSi+) and doubly-charged REE spectral interferences (Er 2+ , Yb 2+ , Lu 2+ and Hf +2 ). SSFR was used to evaluate the accuracy threshold of the Rb correction because of its similar Rb/Sr ratio to the epidote grains (raw 85 Rb/ 88 Sr~0.003). These iRMs yielded 87 Sr/ 86 Sr means of 0.70471 ± 0.00144 (n = 16, ±2σ) and 0.70570 ± 0.00019 (n = 9, ±2σ) respectively, both within measurement error of accepted values (0.703497 ± 0.000035 51 , and 0.70575 ± 0.0001, in-house average). StHs6/80-G yielded a heavy 87 Sr/ 86 Sr value and greater measurement error due to its abundant rubidium content (Rb/Sr = 0.06 51 ), which is significantly greater than Rb/Sr in the epidote samples. We carried out our Sr isotope analyses using enhanced mass resolving power >7000 (source slit set to 0.05 mm and alfa slits engaged, each with 25% signal reduction) to attenuate potential spectral interferences from REE doubly-charged species (e.g., 174 Yb 2+ ) and molecular species (i.e., CaCa+ and FeSi+) endemic in epidote sample matrices. 56 Fe 28 Si + was monitored with 84/88 ratios in StHs6/80-G and epidote grains with no significant effect. This was measured directly in StHs6/ 80-G average raw 84/88 = 0.0063 ± 0.0001 (n = 16, ±2σ) compared to White Seabass average raw 84/88 = 0.0065 ± 0.0001 (n = 70, ±2σ). Note that it is not possible to spectrally resolve FeSi + species using the mass resolution capabilities of the Nu Plasma HR instrument. However, increasing resolving power does effectively reduce the transmission of the FeSi+ species to below the threshold of the faraday detectors. These signals were not observed. Alternatively, it may be possible that these molecular species do not abundantly form and transmit with laser ablation coupled with hot-plasma sample introduction (RF power = 1300). Increasing the mass resolving power was determined the most accurate and robust method for epidote samples considering their range in REE, Sr and Ca composition. LA-MC-ICP-MS data were initially reduced offline using a custom Microsoft Excel spreadsheet. A five-point moving average was applied to reduce the noise typically associated with LA data. A data time window, visualized as a peak in the Sr signal above background measurements, was selected and the average of all the 87 Sr/ 86 Sr ratios within that time window were averaged with 2SE outlier removal. Standard error of the mean is used to report the precision of individual 87 Sr/ 86 Sr analyses.

Data availability
All data generated during the study, including in-situ REE contents and both in-situ and whole grain Sr isotope ratios, are available in supplementary information file (Supplementary Data 1). While, Fig. 2, 3 shows spatially correlated REE-Sr data in epidote, there are cases where this effort was not possible, e.g., fine-grained epidote. As such, there are instances of either REE patterns or Sr isotope data for an individual rock/ crystal. However, in these cases, the distinctive REE patterns or Sr isotope fluid signatures, discussed in the main text, are still represented. Data can be found on Pangaea.de repository under the title 'In-situ REE patterns and Sr isotope ratios of zoned Epidote minerals'.