Constraints on barium isotope fractionation during aragonite precipitation by corals

We present a barium (Ba) isotope fractionation study of marine biogenic carbonates (aragonitic corals). The major aim is to provide first constraints on the Ba isotope fractionation between modern surface seawater and coral skeleton. Mediterranean surface seawater was found to be enriched in the heavy Ba isotopes compared to previously reported values for marine open ocean authigenic and terrestrial minerals. In aquarium experiments with a continuous supply of Mediterranean surface water, the Ba isotopic composition of the bulk sample originating from cultured, aragonitic scleractinian corals (δ137/134Ba between +0·16 ± 0·12‰ and +0·41 ± 0·12‰) were isotopically identical or lighter than that of the ambient Mediterranean surface seawater (δ137/134Ba = +0·42 ± 0·07‰, 2SD), which corresponds to an empirical maximum value of Ba isotope fractionation of Δ137/134Bacoral‐seawater = −0·26 ± 0·14‰ at 25°C. This maximum Ba isotope fractionation is close and identical in direction to previous results from inorganic precipitation experiments with aragonite‐structured pure BaCO3 (witherite). The variability in measured Ba concentrations of the cultured corals is at odds with a uniform distribution coefficient, D(Ba/Ca), thus indicating stronger vital effects on isotope than element discrimination. This observation supports the hypothesis that the Ba isotopic compositions of these corals do not result from simple equilibrium between the skeleton and the bulk seawater. Complementary coral samples from natural settings (tropical shallow‐water corals from the Bahamas and Florida and cold‐water corals from the Norwegian continental shelf) show an even wider range in δ137/134Ba values (+0·14 ± 0·08 to +0·77 ± 0·11‰), most probably due to additional spatial and/or temporal seawater heterogeneity, as indicated by recent publications.


INTRODUCTION
The earth alkaline element barium (Ba) has attracted significant attention in the biogeochemistry community. In the photic zone of marine surface waters, Ba is incorporated in or adsorbed on organic matter (Sternberg et al., 2005) and planktonic carbonate tests, causing depletion of dissolved Ba (Cardinal et al., 2005) and resulting in a nutrient-like distribution in the water column (Wolgemuth & Broecker, 1970;Chan et al., 1976;Church, 1979;Monnin et al., 1999;Chester, 2009). Therefore, variation in the accumulation rates of Ba in marine sediments is thought to be indicative of variations in marine biological productivity through time (Schmitz, 1987;Franc ßois et al., 1995;McManus et al., 1999;Martinez-Ruiz et al., 2000;Babu et al., 2002;Paytan & Griffith, 2007;Hull & Norris, 2011). Furthermore, Ba/Ca ratios in the shells of foraminifers are used to reconstruct past ocean alkalinity and circulation patterns (Lea et al., 1989;Hall & Chan, 2004;Weldeab et al., 2007a) or act as a continental run-off proxy in coastal waters (Weldeab et al., 2007b;Sprovieri et al., 2008).
The marine Ba cycle, however, is not fully understood (Dymond et al., 1992;Dickens et al., 2003). Different Ba sources to the ocean and their relative contributions in the past are still difficult to quantify. If different Ba sources could be distinguished based on their respective isotopic composition, the relative contribution of these sources to the marine Ba pool could be approached by mass balance models. Further, transport, dissolution-reprecipitation or vital processes may induce resolvable Ba isotope fractionation in sediments and biogenic carbonate. Thus, the analysis of the Ba isotopic composition of different Ba sources in seawater and marine sediments bears the potential to improve the understanding of the Ba cycle. Eugster et al. (1969) measured various meteorites, reagents Ba and one terrestrial standard and concluded that there is no anomaly present in Ba isotope abundances in excess of 1&. Subsequently, only a very limited number of studies have dealt with Ba isotopes in natural terrestrial samples (von Allmen et al., 2010;Miyazaki et al., 2014;Horner et al., 2015;Nan et al., 2015;Cao et al., 2016), or experimental carbonate and sulphate precipitates (von Allmen et al., 2010;B€ ottcher et al., 2012). Von Allmen et al. (2010) found that synthetic witherite (BaCO 3 ) and barite (BaSO 4 ) are generally depleted in the heavy Ba isotopes down to À0Á3& in D 137/134 Ba solid-liquid . The calculated total amounts of Ba precipitated in those experiments were below 5%, that is, the results were not significantly influenced by Rayleigh effects. Terrestrial Ba gangue minerals (four barites, one norsethite [BaMg(CO 3 ) 2 ]) were all close to standard solutions. In contrast, a d 137/134 Ba value of À0Á5& was found in a diagenetic barite sample from ODP Leg 207 that formed under reducing diagenetic conditions. It is only recently that modern seawater data have been published for Ba. Open ocean water profiles in the South Atlantic (Horner et al., 2015) and the East China Sea (Cao et al., 2016) showed consistently a general decrease in d 137/134 Ba of about 0Á2 to 0Á3& with water depth. The actual depth in which the transition to lower d 137/134 Ba values starts and the Ba isotopic composition of (sub)surface waters, however, are variable. The latter varies between 0Á45& in d 137/134 Ba in the South Atlantic (Horner et al., 2015) and 0Á8& in the East China Sea (Cao et al., 2016), recalculated relative to the Fluka Ba(NO 3 ) 2 standard used here (see 'MC-ICP-MS analysis' below).
In this study, the Ba isotope composition of cultured coral skeletons is investigated and compared to the composition of the ambient seawater. Corals were chosen because they are abundant in the photic zone where Ba is depleted. Given the observed fractionated Ba isotope values in (sub)surface waters (Horner et al., 2015;Cao et al., 2016) and Ba depletion related to microorganism activity, corals may record changes in productivity and seawater heterogeneities with respect to their Ba isotopic compositions. In order to use coral skeletons as archives, however, the fractionation factor between carbonate and ambient seawater must be known. Next to isotope effects due to carbonate precipitation (von Allmen et al., 2010), vital effects accompanying coral growth may contribute to the overall Ba isotope fractionation in an analogous manner to that observed in the Ca-isotope system. For example, Inoue et al. (2015) found significant difference in D 44/ 40 Ca (solid-solution) in cultured Porites australiensis compared to published data from inorganic aragonite precipitates, what these authors attributed to coral-specific bio-mineralization processes. In contrast, vital effects on stable Sr isotope ratios are ambiguous. Constant D 88/86 Sr values of À0Á20& were reported for various samples of Lophelia pertusa derived from different locations along the European continental margins (Raddatz et al., 2013). Fietzke & Eisenhauer (2006), on the other hand, reported temperature-dependent strontium isotope fractionation during calcium carbonate precipitation. Natural coral samples from Pavona clavus yielded a steeper temperature gradient than inorganically precipitated aragonite.
Here, the experimental setup with different coral species ( Fig. 1) and constant ambient conditions allowed the number of factors involved to be reduced and, in particular, the Ba isotopic composition of the ambient water to be measured. Thus, the magnitude and direction of Ba isotope fractionation during coral growth can be investigated. In addition, corals from different natural settings (tropical shallow-water corals: Bahamas/Florida; deep, cold-water corals: Norwegian continental shelf, Fig. 2) were analysed to get a first indication of Ba isotope fractionation in natural habitats. Further, a set of five Mediterranean surface water samples taken over a distance of >2200 km east to west provides a first glimpse of surface water homogeneity at this scale (Fig. 3). The purpose of this study is to evaluate (i) if Ba isotope fractionation in corals is on the same order as that of abiogenic precipitation or if significant vital effects are observable; and (ii) if corals from natural settings show larger variations than explainable by precipitation, that is, are there indications for isotopically heterogeneous seawater.

Coral skeleton and seawater samples, experimental approaches and sampling environments
Characteristics of the coral samples are summarized in Tables 1 and 2. Two sets of samples consisting of tropical scleractinian corals (cultured and from natural settings), a standardized coral powder and a cold-water scleractinian coral were analysed.
The first sample set is composed of Porites sp., Acropora sp., Stylophora sp. and Montipora sp. cultured under constantly monitored conditions at the Centre Scientifique de Monaco. The aquaria were supplied with Mediterranean seawater pumped from a site situated at 50 m depth and 30 m off the coastline (Centre Scientifique de Monaco, Fig. 3). The salinity was monitored using a conductivity meter (Mettler LF 196) and was found to stay almost constant at 38. Seawater was maintained at a temperature of 25°C using a temperature controller (EW, PC 902/T). Metal halide lamps (Philips HPIT, 400 W) provided irradiance of 204 AE 3 lmol m À2 s À1 on a 12 : 12 photope-riod. The seawater was continuously aerated with ambient air. A water sample was collected in the aquarium, representing the ambient seawater in which the cultured corals grew ('Monaco seawater'). Prior to grinding of the coral tips in an agate mortar, coral skeletons were bleached for 24 h with NaClO (10%), ultrasonic-cleaned and rinsed with Milli-Q water in order to remove organic matter.
The second sample set comprises sub-Recent tropical corals Montastraea cavernosa, collected off Mayaguana Island (Bahamas, Fig. 3; 22°25 0 2Á47″N, 73°8 0 8Á99″W) in 2008, and Favia fragum and Agaricia agaricites from the Florida Keys (Florida, USA; 24°44 0 23Á94″N, 80°58 0 51Á54″ W) collected in 2004. The samples were ultrasoniccleaned in Milli-Q water and powdered in an agate mortar prior to analysis. Unlike the other samples, the coral skeletons from the Bahamas and Florida are not pristine as their aragonite content ranges from 79Á6 to 89Á4% (Table 2), that is, up to 20% of aragonite is recrystallized into secondary calcite. In addition, a standardized coral powder (JCp-1, Geological Survey of Japan) was analysed. JCp-1 is used as an international reference material in coral skeleton studies (Okai et al., 2002;Inoue et al., 2004). This sample comes from a colony of Porites sp. off Japan and consists of 100% aragonite.
A sub-Recent Lophelia pertusa from the Norwegian continental shelf collected at a depth of around 200 m (Freiwald et al., 2002) represents a cold-water scleractinian coral specimen. Prior to analysis, the coral sample was ultrasonic-cleaned in Milli-Q to remove attached sedimentary particles. The aragonitic skeleton of coldwater corals like L. pertusa may be coated post-mortem by a ferromanganese oxyhydroxide crust (Lomitschka & Mangini, 1999) as is the case for our sample (Fig. 2E). To test the potential influence of this coating on Ba isotope data, a portion of the sample was treated with 0Á1M HCl to remove the coating (Copard et al., 2010). Barium contents were very similar prior to (15 AE 0Á7 mg kg À1 , Loph-1) and after this procedure (13 AE 0Á6 mg kg À1 ,   Veron et al. (1996) with cultured species in bold.
Loph-2). Moreover, the two aliquots have identical d 137/134 Ba values (Table 2). We therefore conclude that coating by ferromanganese oxyhydroxides has a negligible influence on the Ba isotope composition of the coral skeleton analysed in this study. A previous study has shown that dissolved Mn does not affect the measured Ba isotope signal (B€ ottcher et al., 2012).
A set of four surface seawater samples was taken on a transect through the Mediterranean Sea (Fig. 3) on-board R/V Maria S. Merian (Cruise MSM33) in November, 2013. Continuously supplied surface water from about 5 m water depth (MSM on-board sampling system) was sampled with 60 cm³ plastic syringes, filtered through 0Á2 lm syringe filters (Minisart, Sartorius) into pre-conditioned polyethylene bottles, and acidified to final 1 vol % HNO 3 (Winde et al., 2014) according to established procedures of filtering alkaline seawater through the membrane filters for earth alkaline element analyses. Seawater salinity and temperature data as well as geographical site coordinates were obtained from the standard MSM board equipment.

Sample preparation and analytical procedures
The analytical procedure of Ba extraction for isotope ratio measurements was conducted using PFA labware in a class 100 laminar flow hood. All acids were distilled before application. A detailed description of the chromatographic purification of Ba from carbonate matrices can be found in Pretet (2013). In brief, several grams of bulk coral material (see Figs 1 and 2) were powdered and homogenized. Aliquots comprising 2 lg of Bacorresponding to 40 to 400 mg per analysiswere first mixed with an appropriate amount of 130 Ba-135 Ba double-spike, following the procedure reported in von Allmen et al. (2010). The spiked carbonate samples were then digested in 5 ml of 6Á4 M HCl at 90°C for 12 h. After digestion, the samples were dried down and then redissolved in 0Á6 ml of 2Á5 M HCl to be loaded onto the column. The column chemistry was performed in PFA micro-columns (4Á0 mm ID; Savillex, Eden Prairie, MN, USA) equipped with 4 mm PFA frits (30 lm pore size) and filled with 1 cm 3 of Dowex â 50WX8 cation-exchange resin (200 to 400 mesh). First, the resin was cleaned with 10 ml of 6Á4 M HCl and conditioned with 4 ml of 2Á5 M HCl. Sample loading and removal of the matrix were conducted with 8Á9 ml of 2Á5 M HCl in total. Finally, Ba was eluted in 6 ml of 6Á4 M HCl and taken to dryness.
A modified chromatographic procedure was applied for Ba separation from seawater, consisting of two cationexchange columns. Before purification, 15 ml of Mediterranean seawater, corresponding to ca 100 ng Ba, were double-spiked and evaporated at 100°C. This aliquot size provides enough Ba for one isotope measurement, while minimizing the amount of matrix elements. The precipitated salt was then equilibrated with 0Á5 ml of 2Á5 M HCl for about 10 h prior to ion-exchange purification.
Remaining salt that did not re-dissolve was removed by centrifugation and discarded. Barium recovery from the evaporates was found to be quantitative, that is, less than 1% of total Ba was lost. In a first column, we removed the largest fraction of major elements. For this we used Spectrum â PP Columns 104704 (Spectrum Chromatography, Houston, TX, USA) filled with 2 cm 3 of Dowex â 50WX8 cation-exchange resin. Cleaning and conditioning of the resin was done with 20 ml of 6Á4 M HCl and 7 ml of 2Á5 M HCl, respectively. The samples were loaded in 0Á5 ml of 2Á5 M HCl, followed by matrix removal with 5Á5 ml of 2Á5 M HCl and 2 ml of 6Á4 M HCl. Barium was collected in 15 ml of 6Á4 M HCl. The Ba yield of this first column was ca 90%. Removal of Na, Mg and K was close to 100%, whereas ca 15% of Ca and ca 55% of Sr were eluted together with Ba. Therefore, Ba was further purified from the remaining matrix in a second step, following the procedure outlined above for carbonate samples. The total Ba yield of our chromatographic separation from seawater was ca 80%. To test for potential analytical isotope fractionation, four aliquots of one seawater sample were spiked at different, successive analytical steps. They gave identical d 137/134 Ba. Thus, despite the non-quantitative recovery of Ba, no impact on the Ba isotope signatures was recognized. Furthermore, the applied double-spike technique in addition corrects for analytical mass-dependent isotope fractionation when spiking is done before purification. Following the chromatographic separation, remaining organic matter from the sample or from the ion-exchange resin was effectively oxidized using ca 150 ll of suprapure 30% H 2 O 2 and 650 ll of 7M HNO 3 , heated to 90°C for about 7 h. Prior to measurements, samples were evaporated and re-dissolved in 0Á5 M HNO 3 . The total blank for the chemical procedure was below 3 ng.

MC-ICP-MS analysis
Barium isotope analyses were performed using a double focusing Nu Instrument â 'Nu plasma' multi-collector ICP-MS at the Institute of Geological Sciences, Bern. Samples were introduced via a PFA nebulizer connected to an ESI â Apex-Q. The measurements were controlled for isobaric interferences of tellurium (Te), affecting mass 130 Ba, and of xenon (Xe) on 130 Ba, 132 Ba and 134 Ba. No Te was detected during the measurements. Signal intensities of 130 Xe, 132 Xe and 134 Xe were calculated based on the 131 Xe beam and Xe isotope ratios taken from IUPAC (de Laeter et al., 2003), which were exponentially corrected for instrumental isotope fractionation.
Data acquisition included three to five blocks of 10 cycles of 10 seconds, depending on the available quantity of sample Ba. Peak centring was performed prior to each block. The isotopic values were expressed relative to a Ba nitrate ICP-OES standard solution from Fluka-Aldrich â as (expressed in &): d 137=134 Ba ¼ ðð 137 Ba= 134 BaÞ sample =ð 137 Ba= 134 BaÞ standard À 1Þ Ã 1000 The Fluka Ba(NO 3 ) 2 standard used as reference solution was previously introduced and calibrated against IAEA standards (IAEA-SO-5 #34, IAEA-SO-6 #34, IAEA-CO-9 # 56) by von Allmen et al. (2010). A minimum of five standards (one after each three samples) was measured per analytical session. The d 137/134 Ba values of the samples reduced by the double-spike routine were further corrected with the average value of the standard solution measurements during each session.

Barium concentrations
Barium concentrations in skeletal material are commonly normalized to calcium (Ca) and reported as Ba/Ca ratios. A distribution coefficient D (Ba/Ca) is then applied to derive the respective composition of seawater from the analysed skeleton (Lea & Spero, 1994;B€ ottcher & Dietzel, 2010), which is defined according to the following equation: However, as (Ba/Ca) solution is constant for the cultured corals (Ba/Ca) carbonate ratios bear the same information on relative variations as D (Ba/Ca) . Furthermore, Ba concentrations vary by one order of magnitude in the corals, while the range of Ca concentrations is smaller than 2% (Table 2). Thus, in the data set of cultured corals, Ba concentrations basically carry the same information as Ba/Ca ratios or D (Ba/Ca) .

Mediterranean seawater barium isotope composition
Sampling of the seawater supplied for the aquarium experiments showed a Ba isotopic composition of 0Á42 AE 0Á18& in d 137/134 Ba (2sd, Table 2). In order to estimate potential variations in modern surface seawater, a further set of four samples from Mediterranean surface waters were analysed for their Ba isotope composition (Table 3). Results range from 0Á38 to 0Á46& with individual uncertainties of ≤0Á07 (2sd; Fig. 4). These data indicate an essentially homogeneous d 137/134 Ba for Mediterranean surface waters. The Ba isotope composition of deeper water masses in the Mediterranean may differ from those of the surface waters and are the subject of further investigations.

Barium isotope composition of coral skeleton
The d 137/134 Ba values of scleractinian coral skeletons, comprising different coral families and genera, range   Table 2). between 0Á14 AE 0Á11 and 0Á77 AE 0Á11& (2s p ,  (Fig. 6). Also, the values are independent from the systematic position of the respective genera and species (cf. Table 1). The range in d 137/134 Ba of the analysed natural corals is much larger (0Á14 to 0Á56, with one specimen yielding 0Á77 AE 0Á11&, Table 2) than that of the cultured species. Here, significant differences exist between the cold-water coral Lophelia pertusa (d 137/134 Ba = 0Á18 AE 0Á08&) and the tropical corals (Florida and the Bahamas; A. agaricites, F. fragum and M. cavernosa).

Barium concentrations and empirical distribution coefficients
The D (Ba/Ca) values of cultured corals vary between 1Á4 and 9Á8 (Table 2). Most values are in agreement with the pioneering work of Lea et al. (1989), who found average D (Ba/Ca) values of 1Á41 AE 0Á14 and 1Á27 AE 0Á03 for two different coral species and locations. The D (Ba/Ca) values for aragonite that was abiogenetically precipitated under laboratory conditions were found to be in the range of about 1Á0 to 2Á6, and to be inversely correlated with temperature (Dietzel et al., 2004;Gaetani & Cohen, 2006). Thus, estimated D (Ba/Ca) values of all but one cultured coral are close to the range of experimental results for abiogenic Ba incorporation into aragonite at high precipitation rates. However, the pronounced D (Ba/Ca) of one aquarium coral of 9Á8 is too high to be explained as dominated by abiogenic Ba incorporation and cannot be reconciled by a unique D (Ba/Ca) value. Thus, it appears reasonable to consider additional vital effects, as suggested previously by Gaetani & Cohen (2006), impacting internal pool sizes or crystallization kinetics beyond the scale considered in experiments so fare.
The Ba concentrations of the corals from natural settings presented here (5 to 24 mg kg À1 ) largely overlap with the cultured corals (8 to 57 mg kg).

Coral skeleton Ba isotope fractionation
Barium isotope fractionation between seawater and coral skeleton is derived from the data of cultured corals ( Table 2). As d 137/134 Ba skeleton is close to or lower than d 137/134 Ba seawater , Ba stable isotope fractionation during biogenic Ba uptake seems to be similar to isotope effects during precipitation of aragonite-structured BaCO 3 (witherite) (von Allmen et al., 2010). It is further comparable to isotope fractionation of Sr (Krabbenh€ oft et al., 2009), Mg (Yoshimura et al., 2011), andCa (B€ ohm et al., 2006) in coral skeletons (see introduction). The experiments performed by von Allmen et al. (2010) suggest that different supersaturation degrees, and associated precipitation rates, influenced the reaction and the enrichment of light isotopes in the precipitates. The maximal observed Ba isotope fractionation between seawater and cultured corals (À0Á26& in D 137/134 Ba coral-seawater ) is still consistent both in direction and magnitude with previous precipitation experiments (von Allmen et al., 2010;B€ ottcher et al., 2012). A similar fractionation factor was approximated by Horner et al. (2015), whose minimum estimate of a BaSO4-seawater corresponds to a D 137/134 Ba BaSO4-seawater of À0Á21&.
Substantial differences in d 137/134 Ba values for corals from the same environment, that is, the aquarium, cannot be attributed to simple, quasi abiogenic precipitation processes in isotope exchange equilibrium with bulk (aquarium) water, as already indicated by the variable D (Ba/Ca) values. A simple first order model for this observation is that variable fractions of the given amount of seawater were precipitated during the bio-mineralization process. The question whether skeletogenesis in corals includes an open or semi-closed calcification compartment, is still a matter of intense debate (B€ ohm et al., 2006; , 2009;Cohen & Gaetani, 2010;Allemand et al., 2011;Gaetani et al., 2011;Tambutt e et al., 2011;Gagnon et al., 2012). A semi-closed calcification compartment might provide the restricted reservoir needed for a Rayleigh-type Ba isotope fractionation. Taken at face value, the variations in d 137/134 Ba of the cultured corals could reflect a range of 25 to 99% of Ba from a given amount of seawater precipitated in a Rayleigh-type fractionation process, using a model with a Mediterranean d 137/134 Ba seawater value as initial seawater value (0Á42 AE 0Á07&, 2sd) and a theoretical D 137/134 Ba precipitate-fluid of À0Á3& (von Allmen et al., 2010). However, a semi-closed calcification compartment is just one hypothesis to explain isotope fractionation in corals. For example, Inoue et al. (2015) reported combined d 44/40 Ca and Sr/Ca of Porites australiensis samples from multiple culture experiments. They conclude that their results are not compatible with a Rayleigh-type fractionation directly from a fluid, which is seawater-like in terms of d 44/40 Ca and Sr/Ca. Therefore, Inoue et al. (2015) favoured Ca-isotope fractionation in corals to be the result of Ca transmembrane transport (Tambutt e et al., 1996). Although these data give indications of isotope fractionation in corals also for Ba, the fractionation is comparatively small given the associated uncertainties and, thus, do not allow us to make conclusions on the competing hypotheses on bio-mineralization.

Diagenetic recrystallization
Following the experimentally derived Ba isotope fractionation factor during carbonate precipitation and the results of our coral culture experiments, the d 137/134 Ba value of a bulk coral is not higher than that of the ambient seawater, unless secondary processes contribute to the overall isotope fractionation. This finding is consistent with results for Sr or Ca (see introduction). In the coral samples from Florida and the Bahamas, up to 20% of the aragonite is recrystallized into calcite (Table 2), a process that may change the Ba isotopic compositions significantly. However, only very low amounts of Ba can be incorporated into the lattice of secondary calcite during recrystallization (B€ ottcher, 1997;B€ ottcher & Dietzel, 2010) or nano-scale dissolution-reprecipitation. Accordingly, calcitic fossils show D (Ba/Ca) values consistently <1 [e.g. foraminifers (Lea & Spero, 1994;H€ onisch et al., 2011) and calcitic bivalve Mytilus eludes (Gillikin et al. 2006)]. Pingitore & Eastman (1985) investigated the Ba partitioning during recrystallization of coral aragonite to calcite. They compared altered and unaltered corals from the Pleistocene reef terraces exposed on Barbados, West Indies, and reported a Ba decrease from typical ranges of 8 to 15 mg kg À1 in aragonite to 1 to 3 mg kg À1 in secondary calcite. Therefore, the bulk carbonate d 137/134 Ba should remain dominated by the composition of the original aragonite fraction. This phenomenon can be illustrated by a simple mass balance model. To increase an aragonitic skeleton d 137/134 Ba value of 0Á4 (Mediterraneanlike seawater) to 0Á8 (like A. agaricites from the Florida setting), considering a 12% fraction of recrystallized calcite (as observed), and a Ba concentration five times lower in calcite than in primary aragonite (based on Pingitore & Eastman, 1985), the calcite fraction would have to carry a d 137/134 Ba value as high as +16&. Therefore, partial diagenetic carbonate recrystallization is deemed unlikely to alter the primary Ba isotopic composition of a coral significantly.

Heterogeneity of the Ba isotopic composition of seawater
Precipitation of carbonate causes Ba isotope fractionation towards lower d 137/134 Ba values in a solid. Currently no data are available that would indicate a process that drives the Ba isotopic composition of a precipitate to d 137/134 Ba values above the corresponding dissolved Ba. This holds generally also true for isotope systems of other Earth alkaline elements (Nielsen et al., 2012). Observed exceptions are Ca-isotope fractionation between aqueous solutions and inorganic solids that trap the hydrated ion in the lattice (Colla et al., 2013) and the equilibrium Mg isotope fractionation factor between inorganic hydrous magnesium sulphate and aqueous MgSO 4 solutions (e.g. epsomite MgSO 4 *7H 2 O, Li et al., 2011).
The d 137/134 Ba value of A. agaricites from Florida (0Á77 AE 0Á11&) is significantly higher than that of Mediterranean seawater (0Á42 AE 0Á07&). Given the above outlined conditions, carbonates of A. agaricites cannot have been precipitated from seawater with a Ba isotopic composition identical to that of Mediterranean surface waters. Indeed, recent studies indicate heterogeneous Ba isotopic compositions of surface ocean waters, ranging in d 137/134 Ba from 0Á45 to 0Á8& (Horner et al., 2015;Cao et al., 2016). Similarly, the vertical seawater heterogeneity, with shifts in d 137/134 Ba of about 0Á2 to 0Á3& (Horner et al., 2015;Cao et al., 2016), could be an explanation for the lower d 137/134 Ba values of the cold-water coral L. pertusa of 0Á18 AE 0Á08&. We thus tentatively interpret the variations in the Ba isotopic composition of the investigated natural corals as indicators of spatial and/or temporal differences in the barium isotope composition of seawater. isotope fractionation between seawater and coral skeletons under controlled standard state conditions (25°C, 1 atm total pressure).

CONCLUSIONS
The seawater sample of the supply for the aquarium experiments and a set of four additional surface water samples from the Mediterranean Sea, collected over a distance of >2200 km from east to west, support a homogeneous Ba isotopic composition for Mediterranean surface waters with an average d 137/134 Ba value of 0Á42 AE 0Á07&.
The cultured corals show either no Ba isotope fractionation relative to ambient seawater or slight enrichment in the lighter Ba isotopes (D 137/134 Ba coral-seawater between À0Á01& and À0Á26&). The direction and magnitude of Ba isotope fractionation during coral growth is comparable with fractionation during precipitation of inorganic carbonates observed by a previous experimental study. The observed range of D (Ba/Ca) values in the cultured corals in part exceed previously reported Ba distribution in abiotic experiments, and indicate additional vital effects leading to reservoir effects or crystallization velocities beyond the conditions of previous experimental calibrations. However, no correlation between D (Ba/Ca) and the Ba isotopic composition of the corals is found in the growth experiments of this study. Instead, isotope fractionation might be governed by the efficiency of Ba incorporation into the coral skeleton. The number of investigated samples and the analytical resolution, however, do not allow further conclusions to be drawn.
One naturally grown tropical coral from Florida (A. agaricites) exhibits a d 137/134 Ba value significantly higher than that of Mediterranean surface water, suggesting heterogeneous Ba isotope compositions of surface seawaters, which is consistent with recently published data by Horner et al. (2015) and Cao et al. (2016). While the Ba isotope system is still in its infancy, these findings provide support to the use of coral Ba isotopes as a palaeo-proxy for water masses and the productivity related isotope composition they carried.