University of Birmingham Assessing the impact of diagenesis on 11B, 13C, 18O, Sr/Ca and B/Ca values in fossil planktic foraminiferal calcite

The geochemical composition of foraminiferal tests is a valuable archive for the reconstruction of paleo-climatic, -oceanographic and -ecological changes. However, dissolution of biogenic calcite and precipitation of inorganic calcite (over-growth and recrystallization) at the seaﬂoor and in the sediment column can potentially alter the original geochemical composition of the foraminiferal test, biasing any resulting paleoenvironmental reconstruction. The d 11 B of planktic foraminiferal calcite is a promising ocean pH-proxy but the eﬀect of diagenesis is still poorly known. Here we present new d 11 B, d 13 C, d 18 O, Sr/Ca and B/Ca data from multiple species of planktic foraminifera from time-equivalent samples for two low latitude sites: clay-rich Tanzanian Drilling Project (TDP) Site 18 from the Indian Ocean containing well-preserved (‘glassy’) foraminifera and carbonate-rich Ocean Drilling Program (ODP) Site 865 from the central Paciﬁc Ocean hosting recrystallized (‘frosty’) foraminifera. Our approach makes the assumption that environmental conditions were initially similar at both sites so most chemical diﬀerences are attributable to diagenesis. Planktic foraminiferal d 18 O and d 13 C records show oﬀsets in both relative and absolute values between the two sites consistent with earlier ﬁndings that these isotopic ratios are strongly inﬂuenced by diagenetic alteration. Sr/Ca and B/Ca ratios in planktic foraminiferal calcite are also oﬀset between the two sites but there is little change in the relative diﬀerence between surface and deep dwelling taxa. In contrast, d 11 B values indicate no large dif-ferences between well-preserved and recrystallized foraminifera suggesting that despite extensive diagenetic alteration the d 11 B of biogenic calcite appears robust, potentially indicative of a lack of free exchange of boron between pore ﬂuids and the recrystallizing CaCO 3 . Our ﬁnding may remove one potential source of uncertainty in d 11 B based pH reconstructions and provide us with greater conﬁdence in our ability to reconstruct pH in the ancient oceans from at least some recrystallized foraminiferal calcite. However, further investigations should extend this approach to test the robustness of our ﬁndings across a range of taphonomies, ages and burial settings.


INTRODUCTION
Foraminifera precipitate their calcium carbonate tests from the seawater in which they live, providing a record of key environmental conditions at the time and depth of calcification (e.g., Lea, 2014). There are many chemical proxies (elemental and isotopic) that can be measured in foraminiferal tests to reconstruct both past environments and organism ecology. To highlight a few, the classic d 18 O and d 13 C proxies provide insights into changes in past ocean temperatures, salinity and/or global ice volume, and ocean dissolved inorganic carbon, respectively (Lea, 2014;. Sr/Ca ratios are more difficult to interpret reflecting some combination of calcification temperature, calcification rate, carbonate chemistry, seawater Sr/Ca, salinity and/or dissolution (e.g., Brown and Elderfield, 1996;Lea et al., 1999;Stoll et al., 1999) but may be a good qualitative indicator of post-mortem alteration of foraminiferal calcite (Baker et al., 1982;Bralower et al., 1997;Regenberg et al., 2007).
The most powerful tool that we have to reconstruct paleo-ocean pH and ultimately past concentrations of atmospheric CO 2 is the boron isotopic composition (expressed as d 11 B) of planktic foraminifera (Foster, 2008;Foster et al., 2012;Hemming and Hö nisch, 2007;Hö nisch et al., 2009;Palmer et al., 1998;Pearson andPalmer, 1999, 2000;Penman et al., 2014;Spivack et al., 1993). The B/Ca of planktic foraminiferal tests also looked to be a promising candidate for reconstructing surface ocean pH (Yu et al., 2007b) or [CO 3 ] 2À (Foster, 2008). However, subsequent investigations indicate that the controls on B/Ca are not clear and thus, interpretation of B/Ca records (particularly in the geological record) may not be as straightforward as originally anticipated Babila et al., 2014;Henehan et al., 2015).
There is increasing awareness of how both the microstructure and geochemistry (elemental and isotopic) of fossil foraminifera tests from deep-sea sediments are modified by post-mortem diagenetic alteration (Norris and Wilson, 1998;Pearson and Burgess, 2008;Pearson et al., 2001Pearson et al., , 2007Schrag et al., 1995;Sexton et al., 2006a;Wilson et al., 2002). Yet, the impact of diagenesis on foraminiferal calcite, in particular on new proxies such as d 11 B and B/Ca, is not well known and thus represents a potentially large source of uncertainty in paleo-reconstructions of ocean carbonate chemistry.

Impact of diagenesis on the microstructure and geochemistry of foraminiferal tests
There are at least three distinct diagenetic processes by which foraminiferal test wall structure and morphology can be modified: partial dissolution, overgrowth and recrystallization. All three can potentially act to offset the original geochemistry of foraminiferal tests post-mortem either in the water column or at and beneath the seafloor . Firstly, calcite dissolution can modify the elemental and isotopic composition of foraminiferal calcite (e.g., Brown and Elderfield, 1996;Coadic et al., 2013;Lohmann, 1995;Pearson, 2012). For instance, one hypothesis suggests that ontogenetic calcite is more susceptible to dissolution than gametogenic calcite precipitated at greater depths and as such with a potentially different chemical composition (Bé et al., 1975;Hemleben et al., 1989). Secondly, precipitation of inorganic calcite from sediment pore fluids onto the internal and/or external walls of foraminiferal tests (overgrowths) can add significant amounts of secondary calcite to the test which may be geochemically very different to the primary test composition. These inorganic calcite crystals are typically much larger than their biogenic counterparts making them easy to identify by scanning electron microscopy and can, in some cases, completely infill specimens . Finally, the in-situ replacement of the original microgranular calcite test wall structure by larger, blockier calcite crystals Pearson et al., 2001;Sexton et al., 2006a) is frequently termed either neomorphism (as per Folk, 1965) or recrystallization (Sorby, 1879). These two terms are sometimes used interchangeably because of their overlapping definitions. However, neomorphism encompasses all in-situ transformations of older crystals to new crystals of the same mineral or its polymorph (recrystallization and inversion, respectively) (Folk, 1965). This term is most commonly applied to the aragonite-calcite transformation or when the original mineral phase is unknown. Here we use the term recrystallization which explicitly refers to the replacement of primary crystals by new crystals of the same mineral species, in this case calcite (see Folk, 1965 for discussion). Recrystallization is potentially gradual and very localized such that the new crystal lattice can be constructed wholly or in part from ions in the pre-existing crystal phases. There may be many rounds of dissolution and re-precipitation as a new crystal is formed  leading to foraminifera appearing opaque or 'frosty' under reflected light (Sexton et al., 2006a).
The potential for diagenetic effects on the isotopic composition of marine carbonates to complicate paleoclimatic reconstructions is best illustrated by the so-called 'cool-tropic paradox' (D'Hondt and Arthur, 1996). This is the mismatch between climate simulations indicating warming at all latitudes and Cretaceous and Early Paleogene paleo-data that reflect similar to modern or cooler than modern sea surface temperatures (SSTs) (D'Hondt and Arthur, 1996). This discrepancy arises from analysis of foraminifera that yield artificially high d 18 O values and thus, underestimates of tropical SSTs because of diagenetic alteration of tropical planktic foraminiferal calcite at the seafloor. One solution to the 'cool-tropic' paradox has been to target sites containing well-preserved foraminifera obtained from clay-rich sediments (e.g., Pearson et al., 2001). These foraminifera typically show little evidence of micron-scale diagenetic alteration appearing translucent under the binocular light microscope or 'glassy' and contain near-pristine biogenic calcite (e.g., Burgess et al., 2008;Norris and Wilson, 1998;Pearson et al., 2001Pearson et al., , 2007Sexton et al., 2006a;Wilson and Norris, 2001;Wilson et al., 2002). The relatively impermeable nature of these clay-rich sediments may prevent significant interaction of foraminiferal calcite with surrounding pore fluids leading to excellent carbonate preservation (Bown et al., 2008). However, sites containing glassy foraminifera are relatively rare (and largely undersampled) in the geological record and are usually limited to continental shelves and slopes. The majority of available deep-sea sites for paleoceanographic reconstructions are carbonate-rich and predominantly contain recrystallized foraminiferal calcite (see Schrag et al., 1995). Despite an obvious diagenetic impact on several well-known geochemical proxies including d 18 O, some component of the original chemical composition is often retained (e.g., Kozdon et al., 2011;Pearson et al., 2001;Sexton et al., 2006a). Hence, the identification and quantification of diagenetic alteration and its impact on the chemical composition of fossil foraminiferal calcite remains a major challenge for paleoenvironmental reconstructions.
The use of d 11 B in marine carbonates as a paleo-pH proxy has undergone a revival in recent years in-line with the increased focus on understanding past changes in ocean acidification and the link between atmospheric CO 2 and climate (Badger et al., 2013;Foster et al., 2012;Hö nisch et al., 2009;Pearson et al., 2009;Penman et al., 2014). Early studies suggest that diagenesis could modify the d 11 B of bulk carbonate sediments (Spivack and You, 1997), but more recent work suggests that the d 11 B of foraminiferal tests may be more robust. Specifically, as with d 18 O and d 13 C records (e.g., Pearson et al., 2001), diagenesis does not apparently overprint inter-specific differences in the d 11 B of fossil foraminiferal calcite (Foster et al., 2012;Palmer et al., 1998;Pearson and Palmer, 1999). Furthermore, there is close agreement between time-equivalent planktic foraminiferal d 11 B values measured in Mid-Miocene aged Globigerinoides sacculifer at ODP Sites 926 and 761, and in the clay-rich Ras-il Pellegrin section in Malta, all three of which have different burial histories and carbonate preservation (Badger et al., 2013;Foster et al., 2012). Similarly, recently published planktic foraminiferal d 11 B data from a suite of deep-sea sites spanning the Paleocene-Eocene Thermal Maximum record similar patterns of change across the event implying minimal diagenetic bias of the d 11 B values given the different preservation histories of the analysed sites (Penman et al., 2014). In contrast, partial dissolution of planktic foraminifera has been hypothesised to explain lower than expected d 11 B values for some species reported from sites situated at greater water depths (Hö nisch and Hemming, 2004;Ni et al., 2007;Seki et al., 2010).
To assess the impact of diagenetic alteration on planktic foraminiferal calcite we present a suite of new d 11 B, d 13 C, d 18 O, B/Ca and Sr/Ca data in planktic foraminiferal species from two time-equivalent sedimentary settings with different taphonomies. Foraminiferal tests from ODP Site 865 are recrystallized whereas those at Tanzanian Drilling Project (TDP) Site 18 have experienced little or no recrystallization. Nearly identical middle Eocene planktic foraminiferal assemblages are found at these two tropical sites implying very similar ages and paleoenvironmental conditions. Further similar oceanographic settings (and temperatures) are inferred for the two sites from Eocene General Circulation Model (GCM) simulations (Huber and Caballero, 2011;Tindall et al., 2010). Thus, our basic assumption is that the elemental and isotopic composition of the various species and their size fractions were originally similar at both sites, hence any large and consistent differences that we observe are likely to be the result of differential diagenesis. At each of the sites we analysed morphospecies that calcified across a substantial environmental gradient in temperature, d 13 C, pH etc., from the surface mixed layer through to the deep-thermocline. More specifically, we assess: (1) the consistency of relative depth rankings of planktic foraminifera in the water column at each site; (2) any inter-specific offsets; (3) differences in the absolute values measured at the two sites; (4) the susceptibility of different taxa to diagenetic alteration; and (5) potential explanations for the chemical trends observed.

Site location
This study utilises sediments from TDP Site 18, a short core (28.4 m long) drilled in the Eocene Masoko Formation (9°261.982 0 S, 39°53.389 0 E; Fig. 1) which comprises greenish-grey silty clays inter-bedded with occasional limestone debris flows (Nicholas et al., 2006). The entire core falls within short-lived planktic foraminiferal Zone E12 defined by the total range of the planktic foraminifera Orbulinoides beckmanni ($40.0-40.5 Ma;Nicholas et al., 2006;Wade et al., 2011). Samples utilised here are from dark greenish-grey silty clays situated in the lowermost part of the core below the weathering oxidation front. Foraminifer assemblages confirm that the studied sediments belong to the lower part of Biozone E12, prior to the warming associated with the Middle Eocene Climate Optimum (MECO). Sediments were deposited on the mid or outer shelf-upper continental slope under estimated paleowater depths of >300 m based on in-situ benthic foraminiferal assemblages (Nicholas et al., 2006). Seismic, sedimentary facies, paleogeography, and nannofossil and planktic foraminiferal assemblages indicate that these sediments were deposited under open-ocean conditions with waters derived from the Indian Ocean subtropical gyre (Bown et al., 2008;Pearson et al., 2007).
Ocean Drilling Program Hole 865C (18°26.425 0 N, 179°33.339 0 W; 1300-1500 m paleowater depth) is located on the summit of Allison Guyot under the North Pacific oligotrophic gyre in the western Pacific Ocean ( Fig. 1; Shipboard Scientific Party, 1993). Eocene sediments are predominantly foraminiferal nannofossil ooze and foraminiferal sands and are shallowly buried (<100 m). Samples here are from uppermost planktic foraminiferal Zone E11/lowermost Zone E12 as rare transitional specimens between Globigerinatheka euganea and O. beckmanni are present but also Acarinina bullbrooki, $40.5 Ma (age assignment based on Wade et al., 2011). Samples were taken from this transitional interval because no stable   and TDP Site 18 (black stars). Base map generated from www.odsn.de. isotope stratigraphy is currently available for ODP Site 865 and we wanted to ensure that we did not take samples from the MECO, which could bias our comparison.
To constrain bottom water conditions at each of the sites the epifaunal benthic foraminifera Cibicidoides was also measured (0.7-1.0 mg or $6 individuals). It is important to note that benthic foraminiferal chemical values at the two sites are not necessarily comparable with one another because our TDP and ODP samples reflect bottom water conditions on the continental slope and from mid-bathyal water depths, respectively.

Scanning electron and light microscopy
To document the extent of diagenetic alteration of fossil planktic foraminiferal tests, representative specimens from each site were selected for imaging. Reflected light microscope images were captured using a Leica DFC 480 camera and Qimaging software -samples were immersed in water before imaging. Prior to mounting, all specimens were ultrasonicated in de-ionised water for approximately two seconds to remove any loosely adhered material from the test surface. Individual tests and broken specimens were then mounted on a black adhesive tab on top of a metal SEM stub. Samples were gold sputter-coated for SEM analysis on a Veeco FEI (Phillips) XL30 environmental SEM at the School of Earth and Ocean Sciences at Cardiff University.

Stable isotope and elemental analyses
Paired isotopic (d 18 O, d 13 C and d 11 B) and elemental analyses were conducted on each mono-specific foraminiferal separate. Specimens in each separate were crushed, homogenised and then divided into two parts with the majority analysed for d 11 B and trace metals ($97%) and $100 lg reserved for d 18 O and d 13 C analyses.
All d 13 C and d 18 O measurements were made on a Thermo Scientific Delta V Advantage mass spectrometer coupled to a GasBench II in the School of Earth and Ocean Sciences at Cardiff University and are reported relative to the Vienna Pee Dee Belemnite (VPDB) standard. Stable isotope values have a standard external analytical precision of 0.06& for d 13 C and 0.07& for d 18 O (at 68% confidence). Mono-specific separates were cleaned for elemental and d 11 B analysis following the oxidative cleaning methodology of Barker et al. (2003). The reductive step designed to remove metal oxide coatings was omitted because these coatings are not a major source of boron contamination (the main focus of this study; Yu et al., 2007a) and the reducing reagent is corrosive causing partial dissolution of foraminiferal carbonate and artificial lowering key element/Ca ratios (Barker et al., 2003). Thus, we do not present any new Mg/Ca data here because Fe-Mn oxide coatings (with high [Mg]) are present on samples from TDP Site 18 but not on those from ODP Site 865 rendering a direct comparison of Mg/Ca ratios from these two sites invalid. Consistent with Yu et al. (2007a) we find no correlation between B/Ca and indicators of Fe-Mn oxide coatings indicating that [B] is low or absent in Fe-Mn oxide coatings. Following cleaning but prior to d 11 B analysis a small aliquot ($7%) of each dissolved foraminiferal separate was analysed for minor and trace elements on a Thermo Element 2 Inductively Coupled Plasma Mass Spectrometer (ICP-MS) at the University of Southampton to test the efficiency of the oxidative cleaning protocol and determine the concentration of boron and strontium in each foraminiferal separate. Subsequently boron was separated from dissolved foraminiferal separates using amberlite IRA-743, a boron-specific anion exchange resin, following the method of Foster (2008). d 11 B analyses were then conducted on a Thermo Scientific Neptune Multi-Collector ICP-MS at the University of Southampton and are reported relative to the NIST-SRM 951 boric acid standard (Catanzaro et al., 1970). External precision is described by the reproducibility of repeat analyses of Japanese Geological Survey Porites coral standard (JCP; d 11 B = 24.3&) and is dependent on the amount of boron analysed . In a slight departure from Henehan et al. (2013) we use two equations, one for single analyses and one for duplicate analyses (Eqs. (1) and (2), respectively) to determine external precision on each species separate.

Constraining the composition and fraction of inorganic calcite contributing to foraminiferal test chemistry
To quantify the amount of chemical alteration that has occurred in frosty specimens, which may be independent of how texturally recrystallized specimens are, we define an 'Index of d 18 O diagenetic overprint'. We developed species-specific linear regressions between the d 18 O values of 'primary' foraminiferal calcite and inorganic calcite precipitated at the seafloor assuming contributions of 0 and 100%, respectively to foraminiferal test chemistry. The d 18 O of inorganic calcite is based on the d 18 O of epifaunal benthic Cibicidoides sp. measured at ODP Site 865 (0.85&) which is thought to precipitate its test in isotopic equilibrium with bottom waters (Bemis et al., 1998). We assume that the d 18 O of primary calcite is equivalent to glassy species-specific d 18 O values (Table 1).
Using the calculated Index of d 18 O diagenetic overprint, mass balance equations (Eqs. (3) and (4)) were solved to estimate the B/Ca and d 11 B of inorganic calcite.
Subscript letters in Eqs. (3) and (4) refer to the isotopic value/elemental ratio in 'glassy' calcite ( G ), inorganic calcite ( I ) and in 'frosty' calcite ( F ). F I is the fraction of the total test chemistry that is attributed to inorganic calcite approximated by the Index of d 18 O diagenetic overprint and (F G ) is the remaining fraction of test chemistry attributed to primary calcite calculated by: 2.6. Hydrographic and pore fluid data To investigate the d 11 B value of inorganic calcite precipitated from the water column and sediment pore fluids we calculate the d 11 B of B(OH) 4 À at ODP Site 865 (no pore fluid data are available for TDP Site 18). Hydrographic data is taken from World Ocean Circulation Experiment section P14N (Sites 88 and 91) located close to ODP Site 865 today (Roden et al., 1995). The d 11 B of B(OH) 4 À in sediment pore fluids (d 11 B BOH 4 -pore fluid ) was determined using chemical data from ODP Site 865 collected by the Shipboard Scientific Party (1993). Profiles were calculated in CO2sys (van Heuven et al., 2011) using salinity, temperature, depth, alkalinity, pH, and the concentrations of silicate and phosphate from each site. Total B (B T ) was estimated from salinity to give the K B . B T and K B were then used to calculate the amount of BOH 3 and B(OH) 4 À in the water column and pore fluids assuming a modern d 11 B sw = 39.61& . The boron isotopic fractionation factor of 1.0272 was applied throughout (Klochko et al., 2006) and phosphate was set at 2 lmol/kg. Sparse in-situ down-hole temperature data was available for ODP Site 865 thus, a geothermal gradient of 46°C/km from nearby sites was used to calculate the temperature for each of the sample depths (Sager et al., 1993). The temperature gradient has little impact on calculated d 11 B values.

Middle Eocene planktic foraminiferal taphonomy
Planktic foraminifera from TDP Site 18 are glassy sensu Sexton et al. (2006a) (Fig. 2a-c) akin to non-encrusting foraminiferal taxa collected live from the modern ocean. SEM images reveals that the test walls of these planktic foraminiferal are constructed from aggregates of sub-micron scale crystallites (microgranules sensu Blow, 1979) creating smooth interior and exterior surfaces with fine scale surface features such as spines, pores and carbonate pustules (also known as 'muricae') preserved ( Fig. 2d-i). In cross section, the test wall is cohesive with biogenic layers and no apparent gaps. In contrast, planktic foraminifera from ODP Site 865 are opaque ('frosty') under reflected light indicating that the walls have been recrystallized ( Fig. 2j-l). SEM analysis shows that external test surfaces are overgrown by relatively large (>1 lm), loosely packed calcite crystals that act to obscure the pores and other surface ornamentation. The delicate biogenic muricae that are characteristic of the dominant Eocene surface dwelling groups Acarinina and Morozovelloides have been overgrown by large, euhedral, blade-like crystals presumably having acted as a locus for crystallisation ( Fig. 2g vs. o). In cross-section it is evident that the structural integrity of the test wall has been compromised with dissolution exploiting existing lines of weakness such as the position of the former primary organic membrane leading to delamination of the inner test wall and there are pervasive small inorganic crystals projecting from the internal test surface (see also Pearson et al., 2007Pearson et al., , 2015Sexton et al., 2006a for similar comparisons). However, specimens at both sites are free of significant infilling. Notably the average test weight of frosty specimens is $30% lower than in their glassy counterparts reflecting the loss of material by dissolution and subsequent looser packing of inorganically precipitated crystals within the recrystallized test walls (Table 2).

Foraminiferal d 18 O and d 13 C and species ecology
The main geochemical tools used to reconstruct the paleoecology of fossil foraminifera, e.g., water depth habitat and presence/absence of algal photosymbionts, are the carbon and oxygen isotopic (d 13 C and d 18 O) compositions of foraminiferal tests (Berger et al., 1978;Birch et al., 2013;Fairbanks et al., 1980;Pearson, 2012;Spero and Williams, 1988). Their utility arises because 12 C is preferentially utilised by photosymbionts and phytoplankton, leaving the foraminiferal microenvironment and ambient seawater enriched in 13 C. Below the photic zone, the 12 C of the water column increases relative to surface waters as a function of reduced photosynthetic activity and remineralisation of 12 C enriched-organic matter leading to lower test d 13 C values (many foraminifera also lack photosymbionts). In contrast, foraminiferal test d 18 O values increase with depth because of the decrease in temperature and the strong temperature-dependence of oxygen isotope fractionation between ambient seawater and foraminiferal calcite during calcification (Bemis et al., 1998;Emiliani, 1954;Pearson, 2012). Thus, taxa that precipitate their tests in warm surface waters have relatively low d 18 O and high d 13 C values relative to those that occupy a deeper position in the water column (Berger et al., 1978;Birch et al., 2013;Fairbanks et al., 1980;Pearson, 2012;Spero and Williams, 1988).   Fig. 3a and b), for all of the species analysed at each of the study sites. Well-preserved specimens from TDP Site 18 indicate species ecologies consistent with previous findings (e.g., Edgar et al., 2013a;John et al., 2013;Pearson et al., 1993Pearson et al., , 2001Sexton et al., 2006b;Wade et al., 2008). Specifically, the muricate taxa Morozovelloides and Acarinina have among the lowest d 18 O and highest d 13 C values in the assemblage suggesting that they lived in the uppermost water column and were likely host to dinoflagellate photosymbionts. G. euganea and Subbotina senni have a similar surface dwelling habitat to the muricate taxa but develop a thick calcite crust in deeper waters towards the end of their life. The lower d 13 C values of S. senni than G. euganea (Fig. 3a and b) is likely attributable to ontogenetic effects resulting from the large difference in the size fractions measured between the two species (250-300 lm versus >500 lm, respectively; Edgar et al., 2013a) and/or the absence of photosymbionts. Turborotalia and Subbotina (other than S. senni) have isotopic compositions suggesting calcification at greater depths in the upper and lower thermocline, respectively and asymbiotic ecology. Finally, the epifaunal benthic Cibicidoides spp. record the highest d 18 O and lowest d 13 C values in the foraminiferal assemblage. Species isotope data show a similar overall pattern at the two study sites with two notable exceptions -in contrast to observations from TDP Site 18, G. euganea and S. senni at ODP Site 865 have the lowest d 18 O values in the assemblage (Fig. 3a vs. b).
d 18 O values in the same species (and size fraction) are consistently higher in recrystallized planktic foraminifera from ODP Site 865 than in glassy foraminifera from TDP  Table 1). However, as previously noted in Pearson (2012) this inter-site difference is least pronounced amongst the globigerinathekids and in S. senni (<2.0&) suggesting that these thick-walled species preserve a signal closer to primary values than other planktic foraminiferal taxa (e.g. they are closest to the 1:1 line in Fig. 3d). In contrast, d 13 C offsets between glassy and recrystallized examples of any given species are typically much smaller (d 13 C values are closer to the 1:1 line in Fig. 3e) and more variable with surface dwelling taxa typically having d 13 C offsets $ ±0.3&, intermediate taxa $ +0.4& and the deepest dwelling planktic foraminifera giving the most pronounced offsets $ +0.8& (Table 1; Fig. 3e). This differential specific response results in a $50% reduction in the surface-thermocline isotopic gradient for d 13 C at ODP Site 865 relative to TDP Site 18 implying compression of inter-species isotopic offsets (Fig. 3c) whereas there is relatively little change in d 18 O gradients between planktic species (Table 1).

Trace element/calcium ratios in foraminiferal calcite
Planktic foraminiferal Sr/Ca ratios generated in the same samples as the stable isotope measurements tend to be marginally higher in the intermediate and deeper dwelling species than in the surface dwellers at both sites ( Fig. 4a and b; Table 2). B/Ca ratios have a more pronounced water column gradient with the highest ratios amongst known surface dwelling taxa and lowest in intermediate and deeper dwelling planktic foraminifera ( Fig. 4d and e; Table 2). Benthic foraminifera at both sites are relatively enriched in [B] relative to their planktic counterparts, consistent with previous work (Table 2 only, not shown in Fig. 4; e.g., Rae et al., 2011). Comparison of trace element/calcium ratios between glassy and frosty planktic foraminiferal samples indicates that recrystallized specimens consistently have Sr/Ca and B/Ca ratios up to 40% lower than their glassy counterparts (Table 2; Fig. 4c and f). However, as with d 18 O and d 13 C values, Sr/Ca offsets are smallest in G. euganea and S. senni (e.g. closest to the 1:1 line in Fig. 4c) between the two sites but this remains true only for G. euganea B/Ca ratios (Table 2). There are also no apparent large changes in the B/Ca or Sr/Ca gradient between surface and thermocline waters at the two sites.

Foraminiferal d 11 B in foraminiferal calcite
Measured d 11 B values at our study sites decrease from $15.5& in the surface dwelling Morozovelloides and Acarinina to $12.5& in thermocline dwelling and benthic species (Fig. 5 and Table 3). This is consistent with the calculated pattern of change in the d 11 B of the B(OH) 4 À in the modern water column (but not in absolute values, compare Figs. 5 and 8a). At each of the sites the relative depth habitat of foraminifera species from d 11 B-depth ranking is similar to those reconstructed using d 18 O and d 13 C values (Fig. 3 vs. 5) and inter-specific offsets are similar at each of the sites. However, G. euganea and S. senni at ODP Site 865 both record lower d 11 B values than predicted from their d 18 O values alone and are more in-keeping with the other surface dwelling taxa (Morozovelloides and Acarinina) (Fig. 3b vs. 5b). Crucially, comparison of d 11 B in the same species (and size fraction wherever possible; Table 3) with frosty and glassy taphonomies reveals little apparent offset outside of the calculated analytical uncertainty on the measurements such that no consistent patterns emerge between taxa at the two sites (Fig. 5c). The one

Quantifying the composition and contribution of inorganic calcite to foraminiferal test chemistry
We calculate a >50% contribution of inorganic calcite to the chemical signature of foraminiferal tests at ODP Site 865 with the smallest estimated values recorded in G. euganea and S. senni (Fig. 6). Utilising the species-specific Index of d 18 O diagenetic overprint and Eq. (3), we estimate the B/Ca ratio of inorganic calcite as <56 lmol/mol, $20 lmol/mol lower than primary (glassy) foraminiferal values (Table 2). From Eq. (4) we calculate the d 11 B of inorganic calcite, species-specific estimates range from 12& to 16& typically falling within the uncertainty of the d 11 B of their glassy counterparts.
We tested the sensitivity of our findings to the d 18 O of inorganic calcite (d 18 O I ) employed and found that it had little impact on our overall conclusions. This is because the relative ordering of the degree of geochemical alteration of taxa remains the same so the impact on absolute values is relatively minor.  (Table 1). This large offset is inconsistent with the comparable oceanographic settings (and temperatures) of the two sites in the modern ocean (Locarnini et al., 2010) and inferred from Eocene GCM simulations (Huber and Caballero, 2011;Tindall et al., 2010). The most reasonable explanation for the large inter-site d 18 O offsets is the diagenetic recrystallization of planktic foraminiferal test calcite in cooler waters at the seafloor as previously predicted by theoretical and numerical modelling, in-situ secondary ion mass spectrometry of tests and diagenetic crystallites, and SEM observations (Kozdon et al., 2011;Pearson et al., 2001;Schrag et al., 1995;Sexton et al., 2006a). Indeed d 18 O values do converge towards the estimated d 18 O of inorganic calcite (see black star in Fig. 3c). Thus, the offset in d 18 O values is likely an artefact of this process and results in an underestimate of $12°C in sea surface temperatures calculated from recrystallized foraminifera (Table 1), consistent with the magnitude previously reported (Kozdon et al., 2011;Pearson et al., 2001;Sexton et al., 2006a). This interpretation is supported by our SEM observations (Fig. 2), which indicate that individual foraminifera tests from ODP Site 865 are completely recrystallized -no evidence of the original microgranular texture remains (Fig. 2j-r). Support is also given by the trace element/Ca ratio (see Sections 4.2 and 4.3) variations observed at the two study sites.
One explanation for the relatively small d 18 O offset between glassy and frosty samples in G. euganea and S. senni (Table 1) is that these taxa have thicker test walls than other taxa. There is no textural evidence that non-recrystallized areas may survive within the test wall (cf. Kozdon et al., 2011), because blocky recrystallization appears pervasive; however internal areas may have chemically exchanged less freely with external areas during the recrystallization process. This may also explain why G. euganea and S. senni record the lowest d 18 O values (warmest temperatures therefore the shallowest inferred water depth) in the whole foraminiferal assemblage at ODP Site 865 ( Fig. 3b; see also Pearson, 2012).
In contrast to d 18 O, d 13 C offsets between the two sites are more modest and within the range of natural surface water variability ($0.5&) between the sites today (Tagliabue and Bopp, 2008). However, here we assume that biogenic carbonate precipitated under similar initial conditions. The d 13 C of diagenetic calcite is more similar to bulk carbonate values (black star in Fig. 3e) than bottom water values (Table 1), in these high carbonate, low organic matter sediments. Indeed planktic foraminiferal d 13 C values do converge towards a bulk carbonate value, which records an upper water column signal because it is predominantly comprised of planktic foraminifera and calcareous nannofossils (Berger et al., 1978). This scenario helps to reconcile the variable inter-specific d 13 C offsets in diagenetically altered foraminifera. For instance, the subbotinids, which have a biogenic d 13 C signature with the largest offset from diagenetic (inorganic) calcite, are most strongly impacted by diagenetic alteration (Table 1; Fig. 3d). In comparison, surface dwelling taxa typically show a much smaller and more variable shift in composition presumably because of the smaller difference between the d 13 C of diagenetic calcite and parent seawater (see also Pearson, 2012;Pearson et al., 2001). The variable response of d 13 C in the surface dwellers may reflect some combination of dissolution (shifting test chemistry to lower d 13 C values), a function of inter-specific dissolution susceptibility, and the spread of primary d 13 C values as a function of ontogeny. Therefore, d 13 C is a less sensitive parameter to quantify diagenetic alteration than d 18 O because of additional uncertainties relating to estimating the d 13 C composition of diagenetic calcite.
In comparison to the obvious diagenetic offset on calculated sea surface temperatures using d 18 O, little change is observed in the surface-thermocline water column temperature gradient in recrystallized calcite compared to that from well-preserved fossil calcite ( Table 1). Compression of the d 13 C gradient is however more pronounced and creates the appearance of a more gradual transition in d 13 C values with increasing water depth (as also noted by John et al., 2013). The impacts of this are difficult to quantify but could bias both spatial and temporal reconstructions of relative changes in export production and/or nutrient utilisation in the photic zone.

Planktic foraminiferal Sr/Ca sensitivity to diagenetic alteration
The effect of dissolution and subsequent recrystallization on foraminiferal test chemistry is elegantly demonstrated by in situ analyses which reveal that inorganic crystallites in diagenetically altered foraminifera have Sr/Ca ratios two to three times lower than in the associated foraminiferal calcite (Kozdon et al., 2013;Regenberg et al., 2007). Our new results confirm this, with the Sr/Ca ratios of frosty planktic foraminiferal tests from ODP Site 865 $33% lower than their glassy counterparts ( Fig. 4c and Table 2). Notably the smallest change in Sr/Ca ratios (<25%) occurs in G. euganea and S. senni, which have a thick calcite crust.
This pattern is easily explained because Sr 2+ is released from marine carbonates into pore fluids during calcite dissolution in the sediment column and then excluded during subsequent inorganic calcite precipitation from the same pore fluids (Baker et al., 1982;Gieskes et al., 1975;Richter and Liang, 1993). This leads to progressively lower [Sr 2+ ] and Sr/Ca ratios in marine carbonates with increasing diagenetic alteration. Thus, diagenesis can result in distinctive down-core profiles of Sr/Ca and [Sr 2+ ] of sediment pore fluids that can constrain the early diagenetic history of shallowly buried (<200 m) Neogene marine carbonates (see open circles in Fig. 7a; Baker et al., 1982;Matter et al., 1975). Many Neogene sediment pore fluid profiles show a pronounced increase in [Sr 2+ ], from seawater values at the sediment-water interface to a maximum at $200 m (see open symbols in Fig. 7a) inferred to represent the zone of maximum carbonate recrystallization (i.e., [Sr 2+ ] exchange by carbonate dissolution and re-precipitation) (Baker et al., 1982). In these profiles, between $200 m and the basement, [Sr 2+ ] is invariant reflecting equilibrium between pore fluids and sediments. Thus, below $200 m these records are not useful for constraining the diagenetic history of carbonate sediments (Baker et al., 1982;Richter and Liang, 1993;Rudnicki et al., 2001).
In contrast to the profiles observed in Neogene deep-sea sediments, the Sr/Ca pore fluid profile at ODP Site 865 shows little change in ratio with increasing sediment burial depth (see solid circles in Fig. 7). This vertical pore fluid profile is characteristic of the shallowly buried Paleogene deep-ocean sediments that are routinely targeted for paleoceanographic analyses, e.g., compare the almost identical downcore Sr/Ca profile of Sites U1334 and 865 in Fig. 7 (Edgar et al., 2013b;Rudnicki et al., 2001). This profile arises because of the absence of a thick pile of 'young' (<20 Ma), reactive sediments overlying Paleogene sediments at ODP Site 865 so there is little exchange of Sr 2+ in the sediment column at the present day (Richter and Liang, 1993). Sr 2+ exchange still occurs between pore fluids and ocean crust but any chemical gradients are quickly lost by upward diffusion through the sediment column. Therefore unfortunately, this flattened profile provides little information on either the timing and/or location of carbonate alteration at this site in the past or present except to highlight that significant carbonate diagenesis is not occurring today consistent with decreasing sediment reactivity with sediment age (Richter and Liang, 1993). However, the large size and well-developed crystal faces on diagenetic carbonate crystallites in Early Paleogene sediments at ODP Site 865 suggest that diagenesis occurred early in the burial history of these sediments (<100 kyrs) before significant burial compaction occurred (Kozdon et al., 2013).
Given the obvious reduction in test Sr/Ca ratios as a function of diagenetic alteration it has been suggested (Bralower et al., 1997;Kozdon et al., 2013;Regenberg et al., 2007) that the Sr/Ca of foraminiferal calcite can be used to distinguish between altered and un-altered specimens. Ancient calcite is considered 'unaltered' if Sr/Ca ratios are >0.9 mmol/mol in the Holocene (Regenberg et al., 2007) or >1.2 mmol/mol in the Cretaceous (Bralower et al., 1997). However, any Sr/Ca cut-off will vary as a function of both the time interval and taxa being investigated. Here we plot our new Sr/Ca ratios against the Index of d 18 O diagenetic overprint for each species to try to place more quantitative constraints on diagenetic alteration (Fig. 7b). We find a correlation between the two variables and note that all altered specimens have Sr/Ca ratios of <1.2 mmol/mol. However, given the spread in the data, it appears unlikely, except in direct comparisons of time-equivalent samples from different sites that Sr/Ca ratios will be a valuable tool for quantitatively assessing the amount of diagenetic alteration a sample has undergone.

How robust are planktic foraminiferal test B/Ca ratios to diagenesis?
B/Ca ratios are consistently lower in recrystallized than in glassy foraminifera (by $6% to 35%; Table 2). However, the controls on planktic foraminiferal B/Ca ratios in biogenic calcite are still poorly constrained and in the modern ocean can vary by $10 lmol/mol between similar sites (Foster, 2008;Ni et al., 2007). If we assume that the B/Ca of biogenic calcite precipitated in surface waters was similar at the two study sites then the positive (albeit weak) correlation between those taxa with the highest Index of d 18 O diagenetic overprint and the largest B/Ca offsets between glassy and frosty taxa may support a diagenetic rather than primary origin for the offsets (Fig. 7c). Certainly published studies show decreasing B/Ca ratios with increasing water depth implying that B/Ca ratios are susceptible to depth dependent dissolution of a similar or lower magnitude to the offsets observed here, albeit over a much larger (and deeper) range of water depths than exists between our two sites (Coadic et al., 2013;Ni et al., 2007;Seki et al., 2010;Yu et al., 2007b). Recrystallized individuals at ODP Site 865 do show evidence of dissolution in the form of etching of the outer test surface, delamination and/or replacement of the test wall, and lower average test weights than in glassy foraminifera. Thus, dissolution and/or subsequent replacement of biogenic calcite with inorganic calcite at the seafloor and in the sediment column may modify test B/Ca ratios with implications for resulting paleoenvironmental reconstructions. However, further work is necessary to confidently discern the full impact of diagenesis on fossil test B/Ca ratios.
An understanding of the B/Ca of inorganic calcite precipitated from sediment pore fluids is essential to constraining the impact of diagenesis on fossil test B/Ca ratios. The B content of calcite is thought to be a function of the partition coefficient (K D ) of B between solid and liquid phases and the B(OH) 4 À to HCO 3 À ratio of the precipitating medium which is pH-dependent (K D = [B/Ca] calcite / [B(OH) 4 À /HCO 3 À ] solution ) (Hemming and Hanson, 1992;Yu et al., 2007b;. There is also an apparent pH dependency to the value of K D itself, in addition to the pH dependency of B(OH) 4 À :HCO 3 À (e.g., . Early empirical studies indicated that at pH values >8.5 K D was typically higher for inorganic than for foraminiferal calcite (Sanyal et al., 1996(Sanyal et al., , 2000. Thus, at pH values above $8.5 inorganic calcite should contain more B than foraminiferal calcite and vice versa at pH values <8.5. However, more recent foraminiferal studies (core-top and culture experiments) from a number of different planktic foraminiferal species yield K D values that are more similar to, or higher (0.61 Â 10 À3 -1.79 Â 10 À3 ) than, inorganic calcite at similar pH values to the original inorganic calcite precipitation experiments (Allen et al., 2011Foster, 2008;Ni et al., 2007;Sanyal et al., 1996Sanyal et al., , 2000Yu et al., 2007b). Thus, these new data suggest that the B/Ca ratio of inorganic calcite may be more similar to, or lower than, foraminiferal calcite, particularly when the lower pH of pore water and deep waters is considered Spivack and You, 1997). This is consistent with calculations in Section 3.2.4, which suggests that the B/Ca ratio of inorganic calcite is likely to be lower than in biogenic calcite (<56 lmol/mol), which if correct, implies that B is preferentially lost and/or subsequently excluded during diagenesis. However it should be noted that considerable uncertainty remains as to the importance of factors other than pH (and the B(OH) 4 À :HCO 3 À ratio) in determining the B content of calcite, including (but not limited to) the rate of precipitation (Gabitov et al., 2014;Mavromatis et al., 2015;Ni et al., 2007), temperature (Foster, 2008;Tripati et al., 2009;Yu et al., 2007b), salinity (Allen et al., 2011) and light intensity (Babila et al., 2014). Indeed the concept of K D as defined here (following Hemming and Hanson, 1992) may itself be at least partly incorrect (see Henehan et al., 2015).

Diagenetic impacts on planktic foraminiferal d 11 B records?
In contrast to d 18 O, d 13 C, Sr/Ca and perhaps B/Ca, the d 11 B of foraminifera analysed here are not obviously impacted by digenesis and the same d 11 B-depth profile is observed at both sites (compare measurements to 1:1 line in Fig. 5c). This raises the following questions: (1) Why does recrystallization/diagenesis impact Sr/Ca, d 18 O, d 13 C and possibly B/Ca but not, apparently, d 11 B? and (2) are there any scenarios under which we might expect to observe a large diagenetic impact on the boron isotope composition of a foraminiferal test?
The d 11 B of G. sacculifer may be modified by partial dissolution (Hö nisch and Hemming, 2004;Ni et al., 2007) with recent estimates showing that the d 11 B of G. sacculifer was lowered by $0.7& when bottom waters become under-saturated with respect to carbonate ion (Seki et al., 2010). In contrast, there is little evidence for a partial dissolution effect on the d 11 B of G. ruber (Henehan et al., 2013;Ni et al., 2007;Seki et al., 2010). One hypothesis to explain this discrepancy is that the preferential dissolution of ontogenetic calcite from the test may shift the d 11 B of G. sacculifer towards the lower d 11 B of gametogenic calcite (precipitated in deeper waters) compared to G. ruber, which possesses little if any gametogenic calcite. However, [B] is heterogeneously distributed in foraminiferal tests and the divide between the different types of calcite is not clear-cut (Allen et al., 2011;Hathorne et al., 2009). Thus, the preferential removal of certain calcite layers with distinct chemical compositions within the test by dissolution is likely an overly simplistic hypothesis as highlighted by studies of other proxy systems (e.g., Hö nisch and Hemming, 2004;Nü rnberg et al., 1996). Regardless, dissolution of samples from ODP Site 865 does not appear to have lead to any significant modification of planktic foraminiferal d 11 B values.
The d 11 B of diagenetic calcite is a function of the pH and isotopic composition of the pore fluids from which it precipitates. The d 11 B (and the [B]) of pore fluids are in turn the product of a number of parameters and processes including temperature, pH, decomposition of organic matter, silica diagenesis, desorption of boron from and onto clays, chemical exchange with seafloor basalt and dissolution of carbonates (Ishikawa and Nakamura, 1993;Palmer et al., 1987;Rae et al., 2011;Spivack and Edmond, 1987;Spivack and You, 1997;Vengosh et al., 1991;. Therefore the relative impact of these processes on pore fluid d 11 B and thus, the d 11 B of B(OH) 4 À incorporated into diagenetic calcite may be strongly site-specific (Brumsack and Zuleger, 1992;Ishikawa and Nakamura, 1993;Spivack and You, 1997;Spivack et al., 1993). The cumulative outcome of calcite and opal dissolution, oxidation of organic matter and desorption of B from clays in the sediment column should lead to lower pH and d 11 B values in pore fluids Spivack and You, 1997;. Thus, inorganic calcite precipitated from these low pH and d 11 B fluids will also have a lower d 11 B than biogenic calcite precipitated in the water column. Indeed this interpretation is supported by Spivack and You (1997) who observed very low d 11 B values ($ À5.5&) in bulk carbonate at carbonate-rich site ODP Site 851. Unfortunately few d 11 B sediment pore-fluid studies exist, with which to constrain the d 11 B of inorganic calcite, for carbonate-rich deep-sea sections that are typical of those selected for paleoceanographic studies.
Seawater is able to mix rapidly (<10 kyrs; Paull et al. (1995)) through the highly permeable calcareous oozes, foraminiferal sands and limestones found at ODP Site 865, hindering the development of geochemical gradients within the pore fluids (as demonstrated by elemental profiles including Sr/Ca in Fig. 4a Shipboard Scientific Party (1993)). This is likely to have also been the case further back in time given the sites persistent position under an oligotrophic gyre and evidence of winnowing indicating high bottom water current flow at this site (Shipboard Scientific Party, 1993). Further, the high calcium carbonate content (>95%) of sediments and minor contributions of clay, silica, organic matter and other components reduces the possible contribution of boron from other sources to pore fluid chemistry (Shipboard Scientific Party, 1993).
Unfortunately [B] or d 11 B sediment pore fluid data were not measured at either of our study sites. Thus, to determine how the composition of inorganic calcite precipitated from pore fluids may vary down-core at ODP Site 865 we estimated the d 11 B of borate in pore fluid (d 11 B BOH 4 -pore fluid ) at the present day ( Fig. 8b; Section 2.6). The resulting d 11 B BOH 4 -pore fluid profile shows little change with increasing burial depth and absolute values are equivalent to bottom waters at this site -in-keeping with the invariant pH profile, strong seawater influence on pore fluid make-up and shallow sample burial ($39 m). Thus all else being equal, the d 11 B of calcite precipitated at the seafloor (i.e. benthic foraminifera) and from seawater-dominated pore fluids, should be similar.

Hypotheses for diagenetic alteration in foraminiferal tests
We propose three potential scenarios for how the diagenetic alteration of foraminiferal tests may occur and discuss the viability of each in light of our new dataset (with particular focus on the insights provided by our new d 11 B data).
(1) 'Open' system -if inorganic calcite precipitates from pore fluids in exchange with bottom waters, and thus with the same chemical composition and physical characteristics, then the d 11 B of inorganic calcite at ODP 865 will be equal to that of biogenic calcite precipitated from bottom waters. However, large pH offsets between the waters in which surface and deep-dwelling taxa precipitate their calcite (Fig. 5) means that surface dwelling taxa have biogenic calcite d 11 B values significantly higher than those at the seafloor (Table 3). Thus, their test calcite d 11 B will become progressively lower with increasing alteration. Given the degree of geochemical overprinting on foraminiferal test chemistry estimated from d 18 O values (Fig. 6), recrystallized surface dwelling taxa should have d 11 B values >1& lower than their glassy counterparts based on simple mixing calculations between two end-members (surface and bottom water d 11 B values; Eq. (4); Fig. 9). Carbonate dissolution (and other processes -see above) in the sediment column should lead to a further reduction in pore fluid d 11 B values exacerbating this offset. Such a large depletion in d 11 B values in surface dwelling frosty taxa is not observed (Fig. 5c). However, this situation could be mitigated if there was much less chemical alteration (<30%) than implied by the Index of d 18 O diagenetic overprint, but this necessitates a very high d 18 O of inorganic calcite (>6&) or if our Index is not a sensitive quantifier of chemical alteration. For now it is difficult to reconcile precipitation of inorganic carbonate directly from bottom-water derived pore fluids with our observations. (2) 'Closed' system -The best-case scenario for paleoceanographic studies is that diagenesis proceeds in a closed system, i.e., the foraminifera test dissolves and re-precipitates with no interaction with surrounding pore fluids. However, the observations that ( Pingitore, 1982) suggesting that foraminiferal recrystallization probably occurs within aqueous films at a very local scale within/immediately surrounding the foraminifer test (i.e., with only limited pore fluid exchange). Thus, foraminiferal calcite itself is likely the major source of the 'raw' chemical components utilised during recrystallization ensuring, in this case at least a near 100% preservation of original d 11 B-signatures. This scenario is consistent with d 13 C and d 18 O values and Sr/Ca (and possibly B/Ca) ratios in foraminiferal calcite that while more sensitive to alteration than d 11 B (as shown by the larger offsets in absolute values between sites) largely preserve inter-specific offsets at each site despite the very different chemical compositions of inorganic calcite (particularly for d 18 O).  (Table 3) based on the assumption that epifaunal Cibicidoides sp. precipitates its test in isotopic equilibrium with bottom waters . Note that if the d 18 O of inorganic calcite is >0.85& the offset between predicted and observed values marginally decreases (offset is $1& if d 18 O is 2&) and if <0.85& it increases. Symbols as defined in Fig. 3.
Other factors that may help to reduce d 11 B offsets between biogenic and inorganic calcite are the reaction rates of recrystallization which are, as highlighted by numerical modelling, sediment pore fluid studies, bulk carbonate and foraminiferal studies, fastest in the first one million years after deposition and rapidly decrease with time (e.g., Fantle and DePaolo, 2006;Richter and Liang, 1993). Thus, early diagenesis typically proceeds rapidly after sediment deposition and thus, at relatively shallow burial depths (Edgar et al., 2013b;Richter and Liang, 1993;Rudnicki et al., 2001;Schrag et al., 1995). Therefore the d 11 B of pore fluids and environmental parameters should remain most similar to those the foraminifera originally precipitated from. Alternatively if inorganic calcite is subsequently demonstrated to have very low B/Ca ratios (<20 lmol/mol) as compared to biogenic foraminiferal calcite then regardless of its d 11 B value its contribution to test geochemistry will be very minor and there is a strong likelihood that some component of the original signal will still be preserved. However, in this situation a d 11 B of B(OH) 4 À of pore fluids of >30& is necessary to ensure similar d 11 B values in glassy and frosty foraminiferal calcite.
Regardless of which scenario is correct, it is clear from the data presented here that the chemical and isotopic composition of recrystallized foraminifera is not simply the result of mixing primary biogenic and secondary pore-water derived diagenetic calcite. We hope that our dataset and these hypotheses will provide the impetus for future work to fully constrain the mechanism by which diagenesis occurs in foraminifera.

CAVEATS AND WAYS FORWARD
While our initial conclusions on the relative stability of d 11 B values in fossil foraminiferal calcite are promising for its use as a proxy, even in recrystallized foraminifera, caution is warranted before our findings are more widely extrapolated to different environments and other depositional settings where there may be additional controls on pore fluid compositions. For instance, if porewater d 11 B (and to a lesser extent pH, temperature and salinity) strongly vary from 'normal' seawater values as a function of processes on-going within the sediment, inorganic and biogenic calcite may have very different d 11 B values. Thus, even a small amount of diagenetic calcite will have a large impact on test d 11 B. This will be particularly problematic if, unlike the situation at ODP 865, pore waters are closed to seawater preventing diffusion of pore fluid and promoting the development of strong geochemical gradients. The problem is also likely to be exacerbated as sediments are increasingly lithified with overgrowths becoming more extensive, eventually infilling specimens (Schlanger and Douglas, 1974). In these situations inorganic calcite typically precipitates deep in the sediment column where environmental parameters and the geochemistry of pore fluids are very different from those that the foraminifera originally precipitated in. However, such poorly preserved foraminifera are typically not utilised for geochemical analysis.
Future work should aim to test the robustness of our findings elsewhere by assessing foraminiferal d 11 B values across a wider range of taphonomies, time scales and sedimentary settings. In particular, the collection of d 11 B and [B] porewater profiles in seafloor sediments coupled with measurements of bulk and foraminiferal calcite from the same burial depths will be a valuable contribution to understanding diagenetic alteration of the boron system. Although perhaps the most direct method to identify the geochemical composition of diagenetic calcite is to utilise in-situ techniques such as Secondary Ionization Mass Spectrometry (SIMS) to compare the d 11 B of inorganic calcite overgrowths and primary foraminiferal calcite akin to work done on other paleoceanographically important proxies (Kasemann et al., 2009;Kozdon et al., 2013).

CONCLUSIONS
Here we utilise new geochemical data from two sites with inferred similar paleoceanographic settings but different carbonate fossil preservation states to assess the different sensitivities of elemental and stable isotope proxies to diagenetic alteration. Despite the influence of pervasive micron-scale diagenetic alteration on foraminiferal calcite observed by SEM we find that the relative depth stratification shown by d 18 O and d 13 C values in planktic foraminifera is generally well preserved. However, taxa with a thick calcite crust specifically the globigerinathekids and S. senni record a geochemical signal closest to original values. Foraminiferal Sr/Ca and B/Ca ratios are consistently lower in recrystallized compared to glassy tests suggesting susceptibility to post-mortem alteration, but further work is needed to more fully understand carbonate B/Ca ratios. In contrast, we show that the d 11 B values recorded in foraminiferal calcite (and inter-specific offsets) are not significantly impacted by extensive diagenetic alteration in shallowly buried samples (<50 m) from ODP Site 865. This may imply that recrystallization of foraminiferal calcite occurs in a relatively localized (and isolated) environment with the major contribution to diagenetic calcite geochemistry deriving from the dissolution of the foraminiferal test itself. Regardless of the exact mechanism our results potentially open the door to utilising the vast number of deep-sea sites hosting frosty foraminifera for reconstructing ocean pH in the past.