Diagenetic incorporation of Sr into aragonitic bivalve shells: implications for chronostratigraphic and palaeoenvironmental interpretations

Aragonite is easily altered during diagenesis, therefore presumed pristine when present. In effect, beyond polymorphic transformation to calcite, alteration paths of aragonite remain poorly understood despite heavy reliance on such material to produce palaeoenvironmental and chronostratigraphic interpretations. Previous work on core material from Southern McMurdo Sound, Antarctica, showed that unlike their calcitic counterparts, seemingly unaltered aragonite shell fragments invariably produced older than expected 87Sr/86Sr ages. In this study, we pursued additional analyses of these aragonite shells and of the porewater of the core to understand this discrepancy. Aragonite mineralogy was reconfirmed and elemental mapping of shell fragments revealed growth lines within the middle layer suggestive of good preservation. The outer layer, however, showed anomalously high Sr concentrations (average 4·5 ± 0·6 mole% SrCO3; ca 25 mmol mol−1 Sr/Ca) and was depleted in 18O and 13C compared to the middle layer, both features inconsistent with pristine material. The δ18O values and Sr concentrations of the porewater were used to model outer layer compositions reasonably well. Coincidentally, porewater Sr isotope composition was in general agreement with the age model of the core only at the aragonite‐bearing interval suggesting that Sr‐isotopic disequilibrium between porewater and the carbonates was the rule rather than the exception in the core. The Sr isotope compositions of the aragonite shells are most likely the result of early diagenesis as suggested by the inconsistent O and C isotope compositions between shell layers and the anomalously high Sr concentrations. We conclude that knowledge of Sr concentration and distribution in shells is critical to determine the viability of Sr stratigraphy and the scale at which it may be applied. Reliance on traditional indicators of lack of alteration, such as cathodoluminescence, Mn‐Fe concentration, and the presence of labile mineralogies to assert chronostratigraphic and palaeoenvironmental questions may produce erroneous conclusions due to obscurely altered material.


INTRODUCTION
The 87 Sr/ 86 Sr ratios of unaltered marine biogenic carbonates are routinely compared to the well-established secular variation of Sr isotopic composition of sea water (Hess et al., 1986;Howarth & McArthur, 1997;McArthur et al., 2001) to provide chronostratigraphic control. Because the residence time of Sr in sea water is considerably longer This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
The journal of the International Association of Sedimentologists than the mixing time of the oceans and because Sr is not fractionated by near-surface physicochemical processes, Sr isotopic composition of sea water at any one time is homogeneous and thus reflects the balance between global rock distribution, rock-type proportion and weathering intensity (Graustein, 1989;Faure & Mensing, 2005). Given that carbonate secreting organisms do not fractionate Sr when they incorporate it in their skeletons, the 87 Sr/ 86 Sr ratio of the open oceanic sea water can be reconstructed through geological time using unaltered biogenic carbonate material. In this context, well-preserved samples can also be used to estimate Sr sources and to model past mixing and weathering rates, although these relationships are not straightforward and numerous components of the climate system appear to play different roles of variable importance at different points in time.
For example, modelling suggests that most of the increase in sea water 87 Sr/ 86 Sr since the mid Pliocene can be explained solely by the relative rise in phyllosilicates weathering prompted by global cooling (Li et al., 2007), while Kashiwagi et al. (2008) conclude that sea water 87 Sr/ 86 Sr cannot be used as a direct proxy for silicate weathering and atmospheric CO 2 decrease during the Cenozoic. Strontium stratigraphy is not applicable to intervals where plateaus in the secular variation of radiogenic Sr prevent time discrimination, or to areas where large freshwater inputs can deviate the biogenic carbonate signal from the global averages (Ingram & Depaolo, 1993;Israelson & Buchardt, 1999;Sessa et al., 2012). However, Sr isotope composition from well-preserved biogenic carbonates of neritic and bathyal marine environments, where conventional chronostratigraphic methods are problematic due to frequent hiatuses and reworking, can be helpful in establishing time constraints. That was an immediate analytical target offered by the several macrofossil-bearing intervals of core AND-2A recovered by the ANDRILL (ANtarctic Geological DRILLing) program during its second field campaign (i.e., the Southern McMurdo Sound (SMS) project) during the austral summer of 2007 (Fig. 1).
The multinational ANDRILL program is a joint effort to drill the margin of Antarctica in search of Cenozoic stratigraphic records to study variation in ice sheets and obtain a better understanding of polar climate evolution (Harwood et al., 2005(Harwood et al., , 2006. Specific goals for the SMS project included: (i) improve chronostratigraphic control; (ii) document melt-water discharge events from the Dry Valleys of the TransAntarctic Mountains and (iii) evaluate the persistence of polar conditions in Antarctica over the past 15 Myr. Extracting this critical information on continental evolution from marginal sediments, however, is not without complications. The biogenic-rich intervals of the core were sampled in part to assess the aforementioned objectives using Sr and O isotopes.
The Sr isotope compositions of pristine marine samples are considered to accurately reflect precipitation age. Previous analyses of venerid and pectinid shells from the macrofossil-rich intervals of core AND-2A  were suggestive of minimal to no diagenetic alteration. The Sr isotope compositions of unaltered calcitic pectinid fragments provided additional chronostratigraphic control to the core (Acton et al., 2008). In contrast with calcite fragments, Sr isotope compositions of the aragonitic venerid shells were lower than expected in all samples resulting in older than reasonable Sr-isotopic ages, while Sr concentrations of the aragonite shells were higher than those observed in modern bivalves (Marcano et al., 2009).
Given the significance of chronostratigraphic and palaeoenvironmental interpretations reliant on the Sr content of carbonate samples, and considering that the aragonite anomalous Sr isotope compositions of samples from core AND-2A were not unequivocally explained using basic isotopic mass balance calculations, these samples were further explored here using more detailed techniques and evaluated against new Sr data from the porewater of the core. Unlike biogenic carbonates, interstitial fluids in AND-2A appeared to be highly modified as suggested by initial chemical analyses of porewater and fracture filling cements (Gui, 2009;Frank & Gui, 2010). Here, the Sr concentrations and isotope compositions of AND-2A porewater were explored to better understand the results from shell studies and more broadly assess the use of 87 Sr/ 86 Sr as a definite chronostratigraphic indicator from carbonates whose alteration has been ruled out using limited criteria.

Previous work: Sr compositions of biogenic material from other Antarctic cores
Several studies from previous Antarctic coring efforts produced Sr isotope ages consistent with independently determined age estimates. The Sr isotope ratios measured in aragonitic bivalves recovered from DSDP Site 270 in the southeastern portion of the Ross Sea and from CIROS-1, off Ferrar Glacier in McMurdo Sound (Barrera, 1989), produced ages consistent with biostratigraphic estimates. Similarly, several aragonitic bivalve samples from the Cape Roberts Project (CRP) cores, drilled about 65 km north of CIROS-1, produced Sr ages consistent with Ar isotope ages and diatom-based biostratigraphy (Lavelle, 1998(Lavelle, , 2000(Lavelle, , 2001. Most of these ages (8 out of 11) remain valid even after the age model of CRP-2/2A was astrochronologically adjusted across the Oligocene-Miocene boundary (Naish et al., 2008). Other studies have produced mixed results, with Sr isotope compositions inconsistent with stratigraphic age. For example, shells from the first ANDRILL core recovered in McMurdo Sound (AND-1B, MIS -McMurdo Ice Sheet Project, Naish et al., 2007) yielded somewhat ambiguous Sr isotope ages . Within the uncertainty of shell fragment identification, the bivalves appeared to be consistently associated with lower than expected Sr isotope values, while other macrofossils (mainly foraminifera and cirripeda) produced ages closer to the age model of the core. Mineralogy of bivalve fragments was not specified, and therefore it is uncertain whether the older ages were associated with aragonitic bivalves only, as was the case in the AND-2A core. Wilson et al. (2007) concluded that the most likely cause of the age discrepancy was contamination by matrix sediments. However, results continued to be equivocal after several cycles of HCl etching of the shell fragments, suggesting that ambiguous Sr isotope compositions were probably intrinsic to the shell and not a product of contamination. Partial equilibration with porewaters was proposed by Wilson et al. (2007) as an alternate cause of the anomalous isotopic compositions, but this hypothesis was not tested.
The Sr isotope compositions of bivalve shells positively identified as aragonite reported from ODP Site 739 in Prydz Bay, Antarctica, also produced unexpected ages (Thierstein et al., 1991). In this case, the Sr isotope ages were younger (latest Early Oligocene-earliest Miocene) than the aragonite shell-bearing stratigraphic units (uppermost Eocene-lowermost Oligocene). Porewaters from the same depths as the shell-bearing diamictites were analysed for Sr concentration and isotopic composition. The Sr concentrations were between 1Á5 and 2 times higher than that of modern sea water, and Sr isotope ratios were among the most radiogenic ever recorded by the ODP and DSDP. Thierstein et al. (1991) calculated that measured aragonite Sr isotope compositions would require either a 90% contribution from freshwater to the depositional fluid or 10% to 17% post-depositional incorporation of Sr from the extant porewater. They also calculated that porewater compositions could be reasonably derived from altered continental detritus.
Approaches used in the studies mentioned above were not consistent. For example, identification, description and criteria to rule out alteration or transport varied widely. This may in part explain the disparity of results. In any case, the apparent pristine aragonite of the macrofossil shells remained puzzling and required further investigation.

MATERIALS AND METHODS
Previous analyses to identify and discriminate alteration effects on these and other selected macrofossil fragments are described in Marcano et al. (2009). Here, we focus on AND-2A venerid aragonitic fragments recovered between 429Á28 m and 430Á51 m below seafloor (mbsf) identified as in-house samples number 9, 10, 11-1, 15-2 and 15-3. These fragments belong to a venerid bivalve recently shown to be a species new to Science and described by Beu & Taviani (2014) as Retrotapes andrillorum. The outer and middle layer could be clearly recognized in most fragments. Marcano et al. (2009) sampled the outer and middle layers separately for chemical analyses, but relatively large sample sizes were required for conventional powder X-ray diffraction technique. Results could not account for the presence of small, localzed alteration products or resolve differences among layers. Powder X-ray microdiffraction (PXRD) was used in this study to reconfirm mineralogy without sample homogenization. PXRD patterns were collected on a Rigaku R-Axis Spider diffractometer with an imaging plate detector using graphite monochromated Cu-Ka radiation (1Á5406 A) at ambient temperature. Whole fragments of venerid shell exhibiting both outer and middle layers were used. The sample was mounted on a cryoloop with paratone N oil for analysis, and the diffractogram obtained from sections of samples where both layers occupied approximate equal volumes. To avoid preferred orientation, images were collected for 5 min while rotating the sample about the φ-axis at 10°s À1 and oscillating x between 120°and 180°at 1°s À1 , with w set at 45°(see additional online material for an illustration of the sample orientation). These were integrated with a 0Á05°step size with the AreaMax2 software package.
The aragonite shell fragments were further sampled from the outer and middle layer to measure oxygen and carbon isotope composition and supplement previous published values (Marcano et al., 2009). Separates were roasted in vacuo, at 200°C, to eliminate volatile contaminants. Oxygen and carbon isotope ratios were determined using an automated Kiel IV device coupled to a triplecollector gas source Finnigan MAT 253 isotope-ratio mass spectrometer and reported against the VPDB (Vienna PeeDee Belemnite) standard. Standard deviations for both carbon and oxygen are equal to or better than 0Á1&.
Electron microprobe elemental mapping was done to localize areas of anomalous Sr concentration in the shell. The X-ray mapping was carried out on a cross section of a shell fragment (sample 10) using wavelength-dispersive (WLD) spectrometry in a Cameca SX-100 electron probe microanalyser (EPMA). Data were collected from three spectrometers with a detection limit of 270 p.p.m. To avoid precision limitations imposed by the analytical time necessary to quantify the entire Sr map, concentrations were measured along three transects, each 270 lm in length along the cross-section of the fragment. Beam diameter was adjusted to 5, 2 and <0Á5 lm with each step size varying from 5 to 2 lm, producing a total of 245 measurements. Secondary electron images were also acquired from the mapped areas and for the rest of the shell fragments, which showed cleaned, unaltered prismatic material.
Porewater chemistry was measured on-ice and discussed elsewhere . Here, Sr concentrations of porewaters were determined on a Perkin Elmer Optima 3300 DV inductively coupled plasma-optical emission spectrometer (ICP-OES) using 10-point calibration curves. One High-Purity â standard solution (Trace Metals in Drinking Water) and one in-house standard indicated that accuracy of the chemical analyses was AE5% RSD or better for Sr and Ca.
The Sr concentrations were used to determine initial porewater sample size for Sr isotope determinations. After complete porewater evaporation 2Á5N HCl was added to the residues. Strontium was then separated using column chromatography (Mukasa et al., 1991). Samples were dried to a solid, treated with a drop of 14N HNO 3 , redried, and loaded on a single Re filament. The Sr was loaded with 0Á1 vol% H 3 PO 4 and TaClO 4 solution. The 87 Sr/ 86 Sr measurements were done on a VG Sector multicollector thermal ionization mass spectrometer (TIMS). The Sr isotope composition was corrected for mass fractionation using 86 Sr/ 88 Sr = 0Á1194. The repeated analyses of NBS-987 standard (n = 3) gave an average 87 Sr/ 86 Sr = 0Á710252 AE 10 9 10 À6 . Total blanks averaged 0Á35 ng for Sr, which are negligible (Table 1).
All analyses described above were carried out in the Chemistry Department and the Earth and Environmental Sciences Department of the University of Michigan in Ann Arbor, Michigan.

RESULTS
Powder patterns from the PXRD patterns were processed in Jade Plus3 to calculate peak positions and intensities. The Jade software package developed by Materials Data Inc. (MDI) supports and provides access to comprehensive XRD databases such as the Powder Diffraction Files (PDF 2 and 4), produced by the International Center for Diffraction Data (ICDD). The suggested match was exclusively aragonite, with both layers contributing equally to the result.
Electron microprobe elemental mapping identified areas of anomalous Sr concentration in the shell ( Fig. 2; see additional online material for a complete list of the results). The backscattered electron (BSE) image suggested that the mineralogy of the middle and outer layers of the venerid fragment was uniform (i.e., no high-Sr alteration phase). Instead of occurring in isolated phases, Sr appeared highly and evenly concentrated in the prismatic outer layer of the shell and along well-defined lines in the crystallographic homogeneous middle layer. Samples strictly within the outer prismatic layer averaged 2Á3 AE 0Á4 wt% Sr (4Á5 AE 0Á6 mole% SrCO 3 ) (n = 15, error is AE 2r), a concentration about five times that of the middle layer lines (0Á5 AE 0Á2 wt% Sr or 1Á0 AE 0Á4 mole% SrCO 3 , n = 43, error is AE2r). These high-Sr lines in turn had about twice the background Sr concentration of the middle layer (0Á3 AE 0Á1wt%, 0Á5 AE 0Á2 mole% SrCO 3 , n = 184, error is AE2r).
The Sr 2+ concentration of porewater varied from a minimum of 1Á5 to 84Á4 p.p.m.; while Ca 2+ varied from 175 to 4,137 p.p.m. (Table 1). Behaviour of Sr 2+ and Ca 2+ downcore was very similar decreasing in the upper 60 m to their respective minima and increasing thereafter. The Sr 2+ and Ca 2+ concentrations diverged between ca 336 mbsf and 545 mbsf, where Ca 2+ decreased while Sr continued to increase. In the uppermost 60 m of the core, Sr 2+ /Ca 2+ varied from 7Á75 mmol mol À1 at ca 10 mbsf to its minimum (ca 4 mmol mol À1 ) and back to ca 8 mmol mol À1 . From about 93 mbsf to the bottom of the core, porewater Sr 2+ /Ca 2+ steadily increased to a maximum of ca 10 mmol mol À1 , with only one sample off the trend at ca 545 mbsf (Fig. 3).
Porewater Sr isotope compositions in the upper ca 200 m of the core showed large deviations from the age model of the core ( Fig. 4; Table 1). Close to the top of the core, porewater 87 Sr/ 86 Sr was well below modern sea water values (0Á70807 AE 15). From this depth to ca 235 mbsf, Sr isotope compositions increased to their maximum (0Á71022 AE 15 at 73Á18 mbsf) and then decreased to values close to those predicted by the age model of the core (0Á70889 AE 15 at 235Á71 mbsf) where they remained relatively stable to a depth of ca 620 mbsf. Porewater 87 Sr/ 86 Sr dropped to its lowest values at 779Á74 mbsf (0Á70692 AE 15) and finally increased to the deepest porewater measured at ca 960 mbsf to values still below those corresponding to the age model (0Á70802 AE 15).
Supplementary stable isotopes samples produced average d 18 O of À5Á7& from the outer layer and À2Á1& from the middle layer, while d 13 C averages are À4Á5& and À0Á2& from outer and middle layer samples, respectively (Fig. 5). Aragonite d 18 O and d 13 C are highly correlated.

Strontium in aragonite bivalve shells
Thin bivalve shells are considered by some to be among the least reliable materials to preserve sea water Sr isotope ratios (Smalley et al., 1994). However, because aragonite is easily altered during early diagenesis (Bathurst, 1975;Brand & Veizer, 1980), bivalve shells that retain their original aragonite mineralogy are presumed pristine and thus expected to preserve their original chemical composition. Lack of pervasive alteration of AND-2A venerid 1Á0 9 10 À5 * Accuracy better than AE5% for Sr and Ca concentrations. shells was previously inferred based on the absence of cathodoluminescence, high Sr and low Fe-Mn concentrations (Marcano et al., 2009). In the present study, the mineralogy of both layers of the same bivalve samples was reconfirmed by microdiffraction. The Sr concentration contrast observed within the middle layer of the venerid fragments from AND-2A (Fig. 2) may be reasonably explained as the result of Sr concentration fluctuating in response to seasonal changes, which either influence Sr availability or the physiology of the bivalve that in turns controls Sr distribution. For example, in a number of previous studies, Sr incorporation into bivalves has been variably correlated to temperature, and also to growth rate and maturity (Dodd, 1965;Stecher et al., 1996;Hart & Blusztajn, 1998;Dutton et al., 2002;Gillikin et al., 2005;Freitas et al., 2006;Surge & Walker, 2006;Elliot et al., 2009). However, no unique relationship has been established, and as suggested by some, controls on Sr incorporation into bivalve shells may even be species specific (Gillikin et al., 2005;Bailey & Lear, 2006). Regardless of the cause, controlling factors of Sr incorporation into skeletal aragonite vary as the carbonate is incrementally added to the shell, thereby producing a record of growth such as that observed in Figure 2.
However, suggestive of primary precipitation the observed growth lines are, their Sr concentrations are anomalously high compared to modern bivalves, in particular those of the outer layer. Modern molluscs in general discriminate against skeletal Sr incorporation, and their Sr content is usually less than 0Á77 mole% SrCO 3 (Kinsman, 1969;Veizer, 1983;Morse & MacKenzie, 1990). Modern aragonite cements, on the other hand, contain between 1Á2 mole% SrCO 3 and 2Á3 mole% SrCO 3 , although inorganic aragonite with up to 14 mole % SrCO 3 has been reported from hot spring deposits (Morse & MacKenzie, 1990). Unusually high Sr contents in AND-2A bivalves suggest that recrystallization of the outer layer, a neomorphic process in which the mineral persists after reaction, probably occurred (Folk, 1965;Bathurst, 1975). Therefore, the prismatic structure observed in the Sr-rich outer layer of AND-2A venerid samples, observed also in modern aragonite bivalves, is not sufficient evidence to reject the potential presence of a secondary carbonate because the crystal form can persist through mineralogical stabilization. Crystal structure preservation occurs when diagenetic reactions take place through migrating solution films, which allows chemical changes to occur at the crystal boundaries without developing porosity (Kinsman, 1969;Maliva et al., 2000).
Authigenic aragonite has not been described in AND-2A, but general conditions favourable for carbonate precipitation from the porewater (i.e., very high alkalinity) exist throughout the length of the core. Also, Wada & Okada (1989) described aragonite cements at a variety of depths in the CIROS-1 core, which was drilled less than 30 km away from the locality of this study (Fig. 1). Detailed description of the porewater chemistry and the diagenetic conditions are published elsewhere Frank & Gui, 2010). Knowledge of the specific saturation state of the pore fluids, however, is not useful given the disequilibrium between porewater and shell aragonite shown by Sr isotopes and the localized nature of the recrystallization process.
As pointed out by Marcano et al. (2009), calculated non-thermodynamic Sr partition coefficients (K Sr ) for aragonite between 0Á4 and 2 are necessary to obtain the measured Sr concentrations in AND-2A shells if these were precipitating from normal modern sea water. A simple non-thermodynamic calculation using a Sr partition   coefficient of 1Á13 (calculated by interpolating to 24°C the values of the distribution coefficient calculated by Kinsman & Holland (1969) at 16°C and 80°C), suggests that the Sr concentrations of aragonite cement precipitating from modern porewater at the level of interest, where Sr 2+ /Ca 2+ could potentially vary between 8 mmol mol À1 and 14 mmol mol À1 (Fig. 3), should be between 9 mmol mol À1 and 16 mmol mol À1 . Although narrower, this range is in good agreement with the 4 to 18 mmol mol À1 Sr/Ca measured in bulk samples of the AND-2A aragonite shell fragments (Marcano et al., 2009). There are limitations associated with this calculation, in particular when considering the high ionic strength of AND-2A porewater. Nevertheless, higher aragonite Sr concentrations can certainly be modelled using the Sr in porewater instead of sea water.
Inorganic precipitation of aragonite has been used to calculate the temperature-dependent partition of Sr in aragonite. These experiments have produced relatively consistent K Sr values close to 1 for temperatures <100°C (Kinsman & Holland, 1969;Dietzel et al., 2003;Gaetani & Cohen, 2006). All of these calculations imply that the Sr/Ca in aragonite should be very close to the Sr 2+ /Ca 2+ in the fluid from which it precipitates. However, Bathurst (1975) in analysing the calculations of Kinsman & Holland (1969), noted the large uncertainties involved (>3500 p.p.m. variation in Sr concentration of inorganic aragonite precipitating from a known fluid at a fixed temperature). Moreover, Gaetani & Cohen (2006) pointed out the large discrepancy between the values based on experimental precipitation and their theoretical calculations. Their theoretical approaches return considerably lower distribution coefficients, which are directly instead of inversely correlated to temperature. The authors suggest that aragonites that conform to a partition coefficient close to 1 are actually strongly enriched in Sr relative to the expected equilibrium concentrations, which should be between 0Á5 mmol mol À1 and 1 mmol mol À1 for temperatures from 15 to 75°C. This disagreement suggests that the incorporation of Sr in aragonite is a complex process not yet fully understood and probably more so in biogenic aragonite.

Alteration scenarios
Although recrystallization without mineralogical change probably occurred, a satisfactory diagenetic process to explain it and further physical evidence of its occurrence are both difficult to produce. The original aragonitic outer shell material must have been dissolved to once more precipitate as aragonite. Carbonate cement compositions are in part controlled by the mineralogy of the particles present in the sediment (Walter, 1986), and although in this case cementation sensu stricto did not occur, it illustrates the plausibility of primary mineralogy controlling diagenetic phases. Unless conditions that inhibit calcite precipitation exist, the more thermodynamically stable calcite crystals will tend to precipitate from fluids saturated with respect to aragonite. Aragonite and high-magnesian calcite, the common skeletal carbonate materials to precipitate in shallow environments, are in metastable equilibrium with sea water and will tend to recrystallize to calcite in the early diagenetic environment (inversion of Folk, 1965; polymorphic transformation of Bathurst, 1975). Although the presence of ions in the pore fluids influences mineral equilibria, and the free energies of formation of pure aragonite and pure calcite are close enough that small changes in the fluid can impact their equilibria (Bathurst, 1975;Morse & MacKenzie, 1990), it is difficult to explain fluid changes across migrating films so that aragonite is dissolved and reprecipitated as the film advances. Even if kinetic processes supersede thermodynamic controls of solution-precipitation (as is common in natural carbonates), it is unclear why diagenetic stabilization should increase Sr concentration. Diagnostic physical products of recrystallization without mineralogical change were absent in petrographic and SEM preliminary observations of AND-2A samples.
Studies on solid-state diffusion of Cd 2+ on calcite demonstrated that this mechanism is capable of incorporating cations into the bulk of the solid and likely to play a role in the uptake of trace metals by other mineral groups as well (Stipp et al., 1992). Although diffusion in carbonate minerals has been regarded relevant only at high temperature and pressure, it may be an important factor in standard conditions, highlighting the potential uncertainties associated to techniques based on the assumption of minerals functioning as perfectly closed system (Stipp et al., 1998).
AND-2A aragonite shells are characterized by low Mn and Fe concentrations, an otherwise good indicator for lack of extensive diagenetic alteration. However, the low Mn and Fe concentrations may be a reflection of their absence in the diagenetic environment, probably an indication of conditions sufficiently oxidizing to prevent reduction of oxides and hydroxides. Abundant iron-rich oxidizing phases and the near absence of organic matter support this possibility. In addition, Mn and Fe share structural affinity to the trigonal calcite, not the orthogonal aragonite. This alone may have prevented their incorporation into the secondary carbonate, even if present in the sediment after the death of the bivalves. Slow recrystallization may have also prevented inclusion of Mn and Fe in aragonite (Morse & MacKenzie, 1990).
If Sr concentrations of the aragonite shell can be achieved as described, and recrystallization occurred and was limited largely to the outer layer, crystal structure may have played a role in promoting chemical changes. Solubility can be influenced by crystal properties and mineral structure differences typically exist between bivalve layers and also between growth lines within a sin-gle layer. According to Walter & Morse (1985), thermodynamically more stable carbonate phases can in some cases dissolve faster than less stable ones due to differences in microstructural complexity. Consequently, the original contrast in crystalline structure is likely to translate into a stability gradient between the bivalve layers, which will result in selective or localized diagenetic stabilization.
Because the alteration scenarios discussed above are based mainly on the excessive Sr concentration of the outer layer and its contrast with the middle layers, which are not by themselves diagenetic indicators, it is necessary to examine them as potential primary features. While some modern aragonite bivalves show Sr enrichment in the outer layer (Shirai et al., 2014), not all do. Elliot et al. (2009) obtained similar average Sr/Ca from different layers of modern Tridacna gigas specimens, although the middle layers showed higher variability. Neither average nor variability differences could be distinguished between the Sr/Ca ratios from the prismatic outer layer and the crossed-lamellar middle layer of Mercenaria campechiensis from the Gulf Coast of Florida (salinity between 19 and 36), although Sr concentration behaviour with respect to d 18 O differed between layers (Surge & Walker, 2006). Takesue et al. (2008) found significant differences in several cation-to-Ca ratios as a function of the aragonite crystal structure in the estuarine bivalve Corbula amurensis, with the notable exceptions of Sr and Na. The shell used to perform the latter analysis grew in waters with a maximum salinity of 28Á5. Interestingly, Foster et al. (2009) studying the marine cold water bivalve Arctica islandica found significant Sr/Ca differences between samples from the axis of maximum growth and those from transects parallel to it, as well as between and within aragonite from the umbo and the outer shell layer. They concluded that changes in shell architecture were a likely candidate to control Sr incorporation through Sr distribution coefficient changes associated with differences in crystal growth. Their analyses were not designed to sample different crystal structures, but the differences they observed keep open the possibility of structure mediated Sr incorporation in some aragonitic bivalves, perhaps an indirect consequence of contrasting growth rates. Still, in the studies mentioned above, Sr/Ca in the aragonite shells remains below ca 3 mmol mol À1 , while the outer layer of AND-2A sample 10, for example, averages 25 mmol mol À1 . This mechanism is most likely unable to concentrate Sr at the levels found in this outer layer.
In summary, a primary origin of aragonitic shells from AND-2A may be suggested by the presence of primary growth lines in the shells and the lack of Sr-isotopic equilibrium between the shells and the ambient porewater. The difficulties associated with physically documenting and chemically explaining potential diagenetic changes in the outer layer of AND-2A Retrotapes and the uncertainties in the partition coefficient of Sr in aragonite could be used to further the argument of lack of alteration. Nevertheless, the extreme Sr concentration of the outer layer is difficult to reconcile with precipitation by a living bivalve and also with the fresh water input that would be required to explain the more negative d 18 O measured in the outer layer, which will be discussed later. A primary origin is therefore unlikely.

Porewater Sr isotope compositions
Excluding a relatively small deviation at 545Á06 mbsf, porewater Sr isotope compositions are within the age model of the core and range between ca 336 and 620 mbsf (Fig. 4). This section includes the aragonitebearing interval at ca 430 mbsf. At all other depths porewater Sr isotope compositions diverge substantially from the age model of the core. Unlike porewater, Sr-isotopic ratios of unaltered calcite shell fragments were in agreement with the age model in the aragonite-bearing interval (366Á83 mbsf) and also above and below (144Á05 mbsf and 1063Á72 mbsf; Marcano et al., 2009). Although porewater may appear to have maintained its original Sr isotope composition at the aragonite-bearing level, this is probably not the case.
Carbonate Sr isotopes require high water-to-rock ratios (on the order of 10 3 ) to equilibrate with fluids unless these are brines. In that case equilibration can occur at considerably lower water-to-rock ratios (ca 10), similar to those required to equilibrate oxygen during freshwater diagenesis (Banner & Hanson, 1990). AND-2A porewater is highly saline, increasing linearly at a rate of about 30 per 100 m depth and stabilizing below ca 500 mbsf to salinities between 150 and 200 in the practical salinity scale (Frank & Gui, 2010). Alkalinity is also high (maximum ca 55 meq kg À1 ).
For brines in particular, appropriate chemical models to describe mineral solubility in subsurface conditions are lacking, and non-equilibrium processes are probably the norm (Morse & MacKenzie, 1990). Nevertheless, given the observed carbonate-porewater isotopic contrasts, the highly modified chemistry of AND-2A porewater and the age of the stratigraphic column, late modification or substitution of the porewater probably occurred in this area, which is located only a few tens of kilometres from the coast. Porewater compositions of the DSDP Sites 270 to 273 in the continental shelf of the Ross Sea, for example, are very different from AND-2A fluids. Maximum alkalinity reported is 25 meq kg À1 , although it remains mostly below 10 meq kg À1 , while salinity drops with depth from normal marine to ca 27 (Mann & Gieskes, 1975). They also reported a slight Sr concentration increase downcore possibly associated with aragonite dissolution. Sr concentration in AND-2A decreases in the uppermost part of the core and begins to increase steadily below ca 120 mbsf. The unusual Sr isotopic composition of AND-2A porewater is highlighted by the Elderfield & Gieskes (1982) study, which summarized Sr concentration and isotopic composition trends from 37 DSDP holes with latitudinal distribution between 0Á5°and 69Á9°( Median = 31Á53°lat). In contrast to AND-2A, overall porewater isotopic compositions in all analysed drill-holes decrease with depth from normal or close to normal modern marine values in the upper 10 mbsf, to a minimum of 0Á70490. Strontium isotope compositions significantly greater than modern sea water were not measured in any of the more than 160 porewater data sets included in this study.
The extant porewater Sr isotope compositions appear not to be just the direct result of reactions with volcanic glass. Throughout AND-2A volcanic alteration products are found in close proximity to unaltered glass . Given the reactive nature of volcanic material, the presence of unaltered grains may in part indicate that some of the alteration products originated at the sedimentary source instead of the subsurface. The Sr isotope compositions of porewater are inadequate to explain the anomalous Sr isotope compositions in AND-2A aragonites and give support to a parautochtonous origin of the porewater. What is more, these observations highlight the potential for invalid interpretations if stabilization with the current porewater had occurred. In such a case, the Sr isotope composition of the water fortuitously matching the age model of the core would have not raised further suspicions about the integrity of the aragonite shells. Consequently, ruling out alteration using only basic criteria such as mineralogy, crystal structure, and Mn and Fe content may be misleading.

Oxygen and Carbon isotopic compositions of aragonite bivalve shells
Calcite pectinids and aragonite venerids at ca 430 mbsf have contrasting oxygen and carbon isotope compositions (Fig. 5). While all measured shell calcites at ca 430 mbsf (n = 3) produced very similar d 18 O and d 13 C values, aragonites did not. Calculated equilibrium d 18 O values for aragonite (between À10Á3& and À11Á0&) using the oxygen isotope composition of the porewater (about À10 &; Frank & Gui, 2010) and the borehole temperatures at the level of the aragonite-bearing interval (between 21°C and 24°C; Wonik et al., 2008) are at least 2& lower than the lowest d 18 O measured in the venerid shell outer layer (À8Á4&) and up to 9Á5& lower than the most positive shell d 18 O value (À1Á5&). These estimates were calculated using the Grossman & Ku (1986) corrected palaeotemperature equation, which is in good agreement with relationships based on theoretical estimates of water-aragonite oxygen fractionation at equilibrium (Kim et al., 2007).
The modelled d 18 O values show variable levels of disequilibrium between the aragonitic shells and the porewater for the conditions of the core today. However, using lower temperatures, it is possible to replicate the measured aragonite d 18 O values. This scenario requires alteration to occur earlier in the burial history of the samples without further re-equilibration. In the upper 50 m of the core, temperatures increase from ca 0 to 5°C, while water d 18 O decreases from modern sea water values of ca À1& to about À10& (Wonik et al., 2008;Frank & Gui, 2010). Aragonite d 18 O precipitating in this interval would have compositions between ca +3Á5& and À6&. Under stable conditions with respect to temperature and porewater d 18 O, maximum alteration could have occurred when the sample was ca 200 mbsf. This is a good indication that parts of the shell could have been subjected to early alteration. The moderate Sr 2+ /Ca 2+ increase at depth (Fig. 3) suggests that dissolution of Sr-rich phases is occurring, although this normal diagenetic trend may not be locally dominant as appears to be the case at the aragonite-bearing interval.
The carbon isotope composition of the shell is a function of the isotopic composition of the bicarbonate from which it forms, which in turn reflects the incorporation of organic and inorganic carbon, as well as metabolic fractionation. Variations in carbon isotopic composition indicate, in part, changes in productivity that are a function of depth or microhabitat conditions (Grossman & Ku, 1986;Hoefs, 1997), and these are difficult to predict or estimate. Although metabolic CO 2 appears to have a limited influence on molluscan carbonate (McConnaughey & Gillikin, 2008;Beirne et al., 2012), it is unlikely that C can be incorporated into biogenic carbonate exclusively reflecting hydrological conditions (i.e., without vital effects). The d 13 C difference between calcite and aragonite was argued to reflect in part contrasting life styles between the pectinid and venerid bivalves (Marcano et al., 2009).
Considerable differences exist in both oxygen and carbon isotope values between outer and middle layers. Values in outer layers are more negative than those from corresponding middle layers. Bivalve shell layers are formed from different pallial fluid sources along different areas of the mantle (Moore, 1969;McConnaughey & Gillikin, 2008), and adjacent points from different layers do not necessarily form at the same time. This discrepancy may be exacerbated by rapid growth. Primary intra-shell contrast in d 13 C is common, but intra-shell contrasts in d 18 O, although reported (Elliot et al., 2009), are normally absent (Surge & Walker, 2006). While d 18 O comparisons between layers in most studies are based on long, usually multiyear data from complete valves, the millimetre-sized fragments analysed here provide data from a single point and this may help explain the contrast. However, the d 18 O difference between the layers can be better explained as the product of localized alteration.
The high correlation between d 18 O and d 13 C values and its similarity with such relationships observed in corals (Gonzalez & Lohmann, 1985;McConnaughey, 1989a, b) could be argued to represent disequilibrium precipitation in contrast with the equilibrium fractionation that characterizes most modern molluscs (Epstein et al., 1953;Grossman & Ku, 1986). A fairly good correlation between d 18 O and d 13 C values has been reported for the modern venerid M. mercenaria (Elliot et al., 2003), and departures from established oxygen isotope equilibrium fractionation of aragonite exist in other clams (Carr e et al., 2005). These observations suggest that disequilibrium precipitation in aragonite bivalves is possible. Kinetic effects probably associated with periods of rapid growth are the main cause of disequilibrium and likely a consequence of the seasonal growth in bivalves. Although the similarity between the slopes of linear fits to the venerid data and to coral data is intriguing (Fig. 5), the possibility of primary precipitation out of equilibrium cannot be tested, and as was mentioned above, the more negative values of the outer layer are not explained using the primary precipitation argument. Thus, the high d 18 Od 13 C correlation remains unexplained.

CONCLUSION
Previous work had shown that aragonite bivalve fragments recovered from ANDRILL AND-2A core in Southern McMurdo Sound, Antarctica, passed generally accepted criteria for unaltered aragonite and showed Sr concentration contrast along growth lines of the middle layer. The Sr concentrations in the outer layer, however, were anomalously high, and Sr isotope compositions of all subsamples were less radiogenic than expected (Marcano et al., 2009). Within the uncertainties associated with the incorporation of Sr into aragonite, the observed high-Sr concentration of the shells was modelled here using AND-2A porewater. Alteration was also suggested by the persistent depletion in 18 O of the outer layers compared to the middle layers, and the successful calculation of shell isotopic values using core temperatures and porewater d 18 O. The Sr isotope compositions of the porewater and the aragonite shells differed. AND-2A porewater was highly modified and in general disequilibrium with the core carbonates. In addition, Sr isotope compositions of the porewater deviated from that of the Ross Sea shelf and from ca 160 other porewater compositions reported by Elderfield & Gieskes (1982). The Sr concentration at the levels reported here (i.e., Sr/Ca as high as 25 mmol mol À1 ), although not easily explained chemically or supported petrographically as a product of alteration, is in strong disagreement with modern primary shell precipitation. The presence of aragonite with no visual indications and only partial evidence of chemical alteration have been also reported for other molluscs, mostly on the nacreous layer of inoceramids, ammonoids and gastropods (Dauphin et al., 2007;Cochran et al., 2010;Wierzbowski et al., 2012). Together, these observations confirm that the anomalous Sr concentrations and isotope compositions observed in the aragonite bivalves of AND-2A, equally expressed along diachronous layers of the shell, are likely the result of early diagenesis. Changes may have been selective, affecting differently the distinct crystal structures of the shell, which are variably susceptible to alteration. Evaluating mineralogy, cathodoluminescence and abundance of minor elements typically associated with diagenesis (i.e., Mn, Fe) is not enough to rule out alteration in Cenozoic Antarctica biogenic aragonites. Knowledge of Sr concentration and distribution in the shell is critical to determining the viability of Sr stratigraphy and the scale at which it may be applicable.