Tripati Et Al. 2010.Pdf

Available online at www.sciencedirect.com

Geochimica et Cosmochimica Acta 74 (2010) 5697–5717 www.elsevier.com/locate/gca

13C–18O isotope signatures and ‘clumped isotope’ thermometry in foraminifera and coccoliths

Aradhna K. Tripati a,b,*, Robert A. Eagle b, Nivedita Thiagarajan b, Alexander C. Gagnon b, Henning Bauch c, Paul R. Halloran d, John M. Eiler b

a Departments of Earth and Space Sciences and Atmospheric and Oceanic Sciences, Institute of Geophysics and Planetary Physics, University of California, Los Angeles, USA b Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, USA c Mainz Academy of Sciences, Humanities, and Literature, IFM-GEOMAR, Kiel, Germany d Met Oﬃce, Hadley Centre, Exeter, UK

Received 12 August 2009; accepted in revised form 1 July 2010; available online 15 July 2010

Abstract

Accurate constraints on past ocean temperatures and compositions are critical for documenting climate change and resolv- ing its causes. Most proxies for temperature are not thermodynamically based, appear to be subject to biological processes, require regional calibrations, and/or are influenced by fluid composition. As a result, their interpretation becomes uncertain when they are applied in settings not necessarily resembling those in which they were empirically calibrated. Independent proxies for past temperature could provide an important means of testing and/or expanding on existing reconstructions. Here we report measurements of abundances of stable isotopologues of calcitic and aragonitic benthic and planktic foraminifera and coccoliths, relate those abundances to independently estimated growth temperatures, and discuss the possible scope of equilibrium and kinetic isotope effects. The proportions of 13C–18O bonds in these samples exhibits a temperature dependence that is generally similar to that previously been reported for inorganic calcite and other biologically precipitated carbonate- containing minerals (apatite from fish, reptile, and mammal teeth; calcitic brachiopods and molluscs; aragonitic coral and mollusks). Most species that exhibit non-equilibrium 18O/16O(d18O) and 13C/12C(d13C) ratios are characterized by 13C–18O bond abundances that are similar to inorganic calcite and are generally indistinguishable from apparent equilibrium, with possible exceptions among benthic foraminiferal samples from the Arctic Ocean where temperatures are near-freezing. Observed isotope ratios in biogenic carbonates can be explained if carbonate minerals generally preserve a state of ordering that reflects the extent of isotopic equilibration of the dissolved inorganic carbon species. Ó 2010 Elsevier Ltd. All rights reserved.

1. INTRODUCTION 2000; Lear et al., 2000; Lea, 2003; Sadekov et al., 2008), and the molecular structures of organic biomarkers (Muller Records of past ocean conditions are essential to under- et al., 1996; Volkman, 2000; Schouten et al., 2002) can be standing the evolution of the Earth. The d18O of carbonates used to give quantitative estimates of temperature and (Epstein et al., 1951; Shackleton and Opdyke, 1973; Erez water composition. If accurately interpreted, sedimentary and Luz, 1982), metal/calcium ratios in foraminifera (Boyle records of these and other proxies can be used to examine and Hester, 1982; Delaney et al., 1995; Rosenthal et al., temperatures, global ice volume, water salinity, and carbon- 1997; Lea et al., 1999, 2000; Elderfield and Ganssen, ate chemistry (Emiliani, 1955; Shackleton, 1974; Keigwin, 1980, 1982; Lynch-Steglitz et al., 1999; Lear et al., 2000; Tripati et al., 2001, 2003, 2005, 2006, 2009; Tripati and * Corresponding author. Tel.: +1 626 395 2975. Zachos, 2002; Tripati and Elderfield, 2004, 2005; Elderfield E-mail address: [email protected] (A.K. Tripati). et al., 2010).

However, the sensitivity of inorganic (d18O, Mg/Ca) and Here we present a global calibration dataset that demon- K0 organic (the alkenone unsaturated index U37 and the tetrae- strates the applicability of a new thermodynamically-based ther index of aracheal lipids with 86 carbon atoms TEX86) temperature proxy, the ‘clumped isotope’ thermometer, to temperature proxies to biological, ecological, and non-tem- foraminifera and coccoliths. We also use these new data perature environmental factors can introduce systematic to discuss the origins of vital effects (or non-equilibrium differences, errors, or ambiguities in climate reconstructions isotopic fractionations) in the d18O of foraminifera and (e.g., Sachs et al., 2000; Lipp et al., 2008). This is of impor- coccoliths. Carbonate clumped isotope thermometry in tance because discrepancies between proxies (MARGO foraminifera and coccoliths has the potential to advance Project Members, 2009) have led to uncertainties in our our understanding of past oceanographic conditions be- understanding of past climate, including attempts to assess cause theoretically it could circumvent many of the issues climate sensitivity to greenhouse gas forcing. Almost all associated with other proxies (i.e., sensitivity to water com- temperature proxies are subject to “vital effects” (i.e., differ- position). This approach is based on the principle that ences between the chemical properties of biologically pre- ordering, or ‘clumping’, of heavy isotopes into bonds with cipitated and inorganic minerals) and/or fluid each other in molecules is temperature-dependent due to composition (Shackleton and Opdyke, 1973; Bemis et al., an internal isotope exchange reaction in carbonate minerals 1998; Turich et al., 2007) (e.g., Figs. 1 and 2). For example, (Ghosh et al., 2006a; Schauble et al., 2006; Eiler, 2007). both d18O and element/Ca ratios in foraminifera and many other biogenic carbonates need to be used with a species- 2. BACKGROUND specific fractionation factor (or apparent partition coeffi- cient) in order to estimate temperature (e.g., Bemis et al., 2.1. Carbonate clumped isotope thermometry 1998; Tripati et al., 2003; Elderfield et al., 2006). Most existing temperature proxies must be referenced to fluid compo- In carbonates, rare heavy isotopes (i.e., 13C and 18O) oc- sition: use of d18O or Mg/Ca ratios in foraminifera as a cur in a pool of abundant light isotopes. The proportion of proxy for temperature requires knowledge of seawater 13C and 18O that are bound to each other within the car- d18O or Mg/Ca, respectively. For example, there is a signif- bonate mineral lattice is predicted to be temperature-depen- icant correlation between Mg/Ca and salinity in both cul- dent based on principles of statistical thermodynamics ture and core-top samples (Nuernberg et al., 1996; Lea (Schauble et al., 2006). The equilibrium constant for the fol- et al., 1999; Arbuszewski et al., 2008; Ferguson et al., lowing reaction (Schauble et al., 2006) forms the theoretical 2008; Kısaku¨rek et al., 2008; Mathien-Blard and Bassinot, basis for the carbonate ‘clumped isotope’ thermometer: 2009). The TEX86 molecular temperature proxy is signifi- Ca12C18O16O Ca13C16O 2 þ 3 cantly correlated with seawater nitrate levels (Turich 13 18 16 12 16 ¢ Ca C O O2 Ca C O3 Reaction1 et al., 2007). These and other factors contribute to uncer- þ ð Þ tainties in the interpretation of past conditions, particularly The temperature-dependent ‘clumping’ of heavy isotopes into for extinct taxa, for which one cannot establish robust bonds with each other is driven by subtle vibrational energy empirical calibrations of the proxy. differences between isotopologues. Each isotopologue has

Fig. 1. d18O ratios of diﬀerent species of foraminifera and coccoliths measured in this study. (A) Data for mixed-layer dwelling and thermocline-dwelling planktic foraminifera, and bulk carbonate (assumed to be composed primarily of coccoliths and therefore to record mixed-layer temperatures). The dashed line is the relationship reported for inorganic calcite (Kim and O’Neil, 1997). Carbonate data is reported relative to V-PDB and water data relative to V-SMOW. Temperature data are from Levitus et al. (1994) and are reported in Table 2 of Electronic Appendix 2. Water d18O values are from LeGrande and Schmidt (2006) and are reported in Table 3 of Electronic Appendix 2. (B) Data for benthic foraminifera (assumed to record bottom-water temperatures). The dotted line is the relationship reported for inorganic aragonite (Kim et al., 2007). 13C–18O isotope signatures and ‘clumped isotope’ thermometry 5699

Fig. 2. d18O and d13C ratios of different species of foraminifera and coccoliths measured in this study compared to values predicted from an inorganic calibration. Inorganic values represent carbonate precipitation in apparent equilibrium. (A) Relationship between deviations from inorganic carbonate values of d18O and d13C for mixed-layer dwelling planktic foraminifera and bulk carbonate. (B) Same as (A) but for thermocline-dwelling planktic foraminifera. (C) Same as (A) and (B) but for benthic foraminifera. unique vibrational properties which therefore impact their reflect any combination of analytical and/or experimental thermodynamic stability, with the species on the right-side artifacts in either or both of these studies. Further experi- of Reaction 1 favored at lower temperatures, promoting mental work will be needed to resolve this issue; for the pur- the clumping of heavy isotopes into bonds with each other poses of our discussion, we compare our results to the (Schauble et al., 2006). In an equilibrium precipitate, if the inorganic calcite calibration from Ghosh et al. (2006a) as equilibrium constant for Reaction 1 is known and the their calibration samples were measured on the same instru- abundance of all four isotopic species is measured, the ment as the ones used for this study. temperature can be calculated independent of the isotopic The biogenic carbonates examined in previous studies composition of the water in which the carbonate grew. generally adhere to the inorganic calcite calibration, with In practice, clumped isotope thermometry is based on a sensitivity of 0.004–0.005&/°C. The observed excep- 13 18 16 $ analyses of C O O in CO2 produced by acid digestion tions to this trend (Fig. 3) are measurements of aragonite of carbonates. Fractionation during acid digestion is a from two different types of marine organisms: Red Sea cor- predictable and apparently reproducible function of diges- al (Ghosh et al., 2006a) and fish otoliths (Ghosh et al., tion temperature (Ghosh et al., 2006a; Guo et al., 2009). 2007). The Red Sea coral data exhibit a wide range of var- The abundance of the doubly substituted isotopologue iability in D47 (a range of 0.10–0.12&) for a relatively 13 18 16 $ C O O is reported using the variable D47, defined as small (seasonal) range of calcification temperatures the enrichment, in per mil (&), of 13C18O16O above the ( 6 °C; defined in the original publication by the instru- $ amount expected for a random distribution of isotopes mental record for the region, and also by Sr/Ca-based tem- among all CO2 isotopologues (Eiler and Schauble, 2004). perature estimates). Most of the samples from this 13 18 Measurements have been made to calibrate D47 in inorganic individual coral are enriched in C– O bonds by 0.02– calcite precipitated at known temperatures and biogenic 0.04& relative to inorganic calcite. This data set also possi- carbonates including fish otoliths, mollusks, brachiopods, bly exhibits a higher temperature sensitivity relative to most teeth, micritic carbonates in lake, and coral (Ghosh et al., biogenic carbonates and inorganic calcite. In contrast, the 2006a, 2007; Came et al., 2007; Eagle et al., 2010; Hunting- otolith dataset exhibits a similar temperature sensitivity to ton et al., 2010). These published calibration data are the inorganic calcite calibration, but lower absolute values. shown in Fig. 3. The origin of these two types of systematic deviations is A recently published calibration of synthetic calcite unknown. They may indicate the occurrence of non-equilib- (Dennis and Schrag, 2010; not shown) differs from the trend rium biological fractionations in some calcifying organisms, reported by Ghosh et al. (2006a) at low temperatures, with record analytical artifacts, and/or reflect uncertainties in a shallower slope and higher intercept than previously re- estimating the temperature of calcification for the analyzed ported. The source of this discrepancy is not clear. It might materials. 5700 A.K. Tripati et al. / Geochimica et Cosmochimica Acta 74 (2010) 5697–5717

Fig. 3. Published D47 data (mean and 1 SE) for modern inorganic calcite and biogenic carbonates compared to independent estimates of growth temperatures (mean and range). The line corresponds to the linear regression from Ghosh et al. (2006a) through the inorganic calcite data.

Since this thermometer was developed, it has been ap- calcite (McCrea, 1950; Usdowski and Hoefs, 1993; Beck plied to reconstruct temperature in several contexts. In et al., 2005), foraminiferal calcite (Spero et al., 1997; Bemis addition to constraining temperature, measurements have et al., 1998), and other biogenic carbonates (Rollion-Bard been paired with carbonate d18O to determine ambient et al., 2003) may also vary as a function of carbonate ion water d18O. Published applications thus far include (a) esti- concentration. These factors contribute to uncertainties in mating continental temperatures using paleosol carbonates the interpretation of past conditions from carbonate d18O, to study the tectonic history of the Bolivian Altiplano and particularly for biogenic carbonates precipitated by extinct other regions (Ghosh et al., 2006b; Quade et al., 2007; taxa, for which one cannot establish robust empirical cali- Huntington et al., 2010), (b) constraining variations in ter- brations of the fractionation with respect to water. restrial temperature using speleothems (Affek et al., 2008), (c) reconstructing Pennsylvanian and Silurian temperature 3. METHODS and seawater d18O using well-preserved molluscs and brachiopods, to look at the long-term influence of CO2 and 3.1. Samples cosmic rays on climate (Came et al., 2007), (d) using the precipitation temperature of carbonates in primitive mete- For this study, we calibrated the thermometer with mea- orites to study their alteration history (Guo and Eiler, surements of 17 species of foraminifera and coccoliths that 2007), (e) reconstructing the body temperature of extinct are independently known to have grown at temperatures taxa (Eagle et al., 2010), (f) assessing the susceptibility of ranging from 0.9 to 29.2 °C, including both calcitic and À D47 to diagenetic resetting (Came et al., 2007; Dennis and aragonitic taxa. Of the 41 calibration samples we examined, Schrag, 2010; Eagle et al., 2010). 24 were of planktic foraminifera, 11 of benthic foraminifera, 4 of bulk carbonate in core-top sediments, and 2 of cul- 2.2. Controls on the oxygen isotope composition of tured coccoliths. Counting replicates of 13 samples, 61 foraminifera and coccoliths separate analyses (i.e., extractions of CO2 from carbonate) were made in total. This represents the first detailed survey The d18O values of planktic and benthic foraminifera, of the clumped isotope composition of multiple marine spe- and coccoliths, depend on both temperature and seawater cies from sediment core-tops, and the first measurements of d18O, which itself depends on global ice volume and salin- cultured carbonate-secreting organisms. All samples are ity. Therefore interpretations of carbonate d18O are ambig- monospecific except where noted (i.e., for bulk carbonate uous unless one can call on independent constraints on samples or for mixed species of benthic foraminifera). Sam- either temperature or seawater d18O. In addition, although ples ranged from 4 to 12 mg in size, with 200–500 individu- a few species appear to precipitate their tests in apparent als for each planktic foraminiferal sample and between 100 isotopic equilibrium with the seawater from which they and 200 individuals for each benthic foraminiferal sample. grew, disequilibrium fractionations of unknown origin Planktic foraminifera were picked from the >250 lm size (i.e., vital effects) are common (Shackleton and Opdyke, fraction and benthic foraminifera were picked from the 1973; Bemis et al., 1998). The d18O values of some modern >212 lm size fraction. Samples of Arctic benthic foraminif- foraminiferal taxa deviate from apparent equilibrium (i.e., era were provided by author H.B. Table 1 of Electronic values predicted using inorganic calibrations) by up to 3– Appendix 2 contains a description of sample and locality 4& (e.g., Fig. 2; Fig. 1 of Electronic Appendix 1), and it information. Tables 2 and 3 of Electronic Appendix 2 list is possible that the magnitude of this disequilibrium may constraints on calcification temperature and other hydro- potentially vary over time. A similar range of non-equilib- graphic parameters (water d18O, the d13C of dissolved inor- rium fractionations has been reported in coccoliths (Ziveri ganic carbon), respectively. The latter table also shows et al., 2003). It has been shown that the d18O of inorganic estimates of the saturation state of bottom waters and pH 13C–18O isotope signatures and ‘clumped isotope’ thermometry 5701 for localities where samples of benthic foraminifera were ta- were made to yield a 16-V signal for mass 44, with peak ken, with carbonate system parameters calculated from centering, background measurement, and pressure balanc- hydrographic data in the GLODAP database (Sabine ing before each acquisition. These conditions were selected et al., 2005) using constants from the Best Practices Hand- to ensure stable ion currents and minimize time-dependent book (Dickson et al., 2007). fractionation of sample and reference gas reservoirs. Monospecific cultures of Emiliania huxleyi and Coccoli- Data for this study was collected between April 2007 thus pelagicus were grown by author P.H. in 0.1 lm filtered and January 2009. Each day we analyzed a heated gas com- and aged-seawater, enriched with nutrients, vitamins and posed of CO2 with a stochastic distribution of isotopes be- trace elements. Coccolithophores were grown for at least tween isotopologues. Gases with different bulk d18O and three consecutive batch cultures ( 20 generations for d13C ratios in quartz breakseals were heated to 1000 °C $ E. huxleyi) at the desired temperature before inoculating for 2 h and then quenched at room temperature. These final cultures in three 2 l acid-cleaned bottles. Coccolitho- standard gases (or heated gases) were then purified and phores were maintained under a 14/10 h light/dark cycle, analyzed using the same protocol as sample gases, and used and the growth rates were monitored to ensure that harvest- for the calculation of sample D47 values, where R47 ing was undertaken during the exponential growth phase. sample 47 D47 R47 1 1000 and R refers to the 47/44 ratio Samples were harvested onto 0.2 lm polycarbonate mem- ¼ stochastic À ÂÞ brane filters under vacuum, rinsed with Aristar grade of the analyzed CO2 (Eiler and Schauble, 2004). Several of ethanol, taken into suspension in ethanol and transferred the samples were analyzed blind by authors R.E. and N.T. to 10 ml centrifuge tubes. (i.e., without knowledge of their origin or relationship to the present study). No systematic differences were observed 3.2. Sample pre-treatment between these analyses and the remaining samples (Table 4 of Electronic Appendix 2). These analyses were performed To remove contaminants, we developed a protocol that in the lab of author J.E. uses a cold dilute buffered solution of hydrogen peroxide We used published acid digestion fractionation factors to remove organic matter and adhering particles, which is for determining carbonate d18O ratios. The value used for 18 16 modified from the cleaning protocols designed for measuring calculating calcite O/ O ratios from analyzed CO2 is metal/calcium ratios in foraminifera (Boyle and Keigwin, 1.01025 for samples digested at 25 °C(Das Sharma et al., 1982). All foraminiferal and coccolith samples were dried 2002; Guo et al., 2009) and 1.00821 for samples digested overnight in an oven at 50–70 °C. Most samples were at 90 °C(Swart et al., 1991). Aragonitic samples reported cleaned to remove secondary carbonates and organic mat- in this paper were digested at 25 °C and the acid digestion ter (Table 4 of Electronic Appendix 2). Foraminiferal sam- fractionation factor used for calculations is 1.01034 (i.e., ples were crushed between two glass plates, and then 1000 ln a = 10.44; Sharma and Clayton, 1965). sonicated in distilled water for 1 min and rinsed (repeated For calculations of equilibrium 18O/16O ratios in calcite 5 ), sonicated in methanol and rinsed (repeated 5 ), and (and to calculate d18O-calcification temperatures), we used Â Â then sonicated in distilled water again and rinsed (repeated a published relationship (Kim and O’Neil, 1997) to describe 18:03 103 3 ). Samples with high organic contents were cleaned to re- calcite–water fractionation: 1000 ln acalcite–H O Â Â 2 ¼ T À move organic matter by reaction for 15–60 min (depending 32:42 and another expression for aragonite (Kim et al., 17:88 103 on the organic content) with a cold dilute solution of 1% 2007): 1000 ln a Â 31:14, where T is aragonite–H2O ¼ T À H2O2. Samples were then centrifuged, decanted, and rinsed temperature in K. For these calculations, temperatures four times with distilled water, twice in methanol, and then are determined by applying the published inorganic calibra- twice again in distilled water. Samples were then dried again tion equation of Ghosh et al. (2006a,b), and water d18O is in a 50–70 °C oven overnight. Treatment of carbonate stan- from a published database (LeGrande and Schmidt, dards with this protocol did not affect their value. 2006). Calculations of equilibrium 13C/12C ratios in calcite and aragonite used the published relationships determined 3.3. Sample analyses by Romanek et al. (1992), d13C ratios for dissolved inorganic carbon shown in Table 3 of Electronic Appendix 2. As described elsewhere (Ghosh et al., 2006a; Eiler, 2007; The presence of organics can yield significant isobaric Huntington et al., 2009), between 4 and 12 mg of sample interferences with CO2 isotopologues that can impact both was reacted with phosphoric acid at 25 °C (with the excep- accuracy and precision, resulting in an apparent tempera- tion of some samples from localities MD97-2138, V24-109, ture bias and/or larger internal errors. We monitored and OP 806A which were reacted at 90 °C) and the product masses 48 and 49 to screen for isobaric interferences from CO2 purified using cryogenic separation and gas chroma- the presence of these compounds. Data for samples that tography. The sample gas was analyzed for masses 44–49 were measured but are excluded from the discussion are AMU on a Thermo 253 gas source isotope ratio mass spec- listed in Table 5 of Electronic Appendix 2. trometer, for 8–16 acquisitions of 10 cycles each, with an integration time of 8 s per cycle (Ghosh et al., 2006a). 3.4. Precision and accuracy The total integration time was 640–1280 s (for both sample gas and working gas) and total analysis time (including 3.4.1. Precision peak centering, background measurement, and pressure We report mean sample D47 values and the standard error balancing) ranged from between 4 and 8 h. Measurements of each sample mean in Table 4 of Electronic Appendix 2. 5702 A.K. Tripati et al. / Geochimica et Cosmochimica Acta 74 (2010) 5697–5717

Error propagation factored in uncertainty in measurement occasionally referred to here as ‘Levitus temperatures’. of the D47 of CO2 and uncertainty in heated gas normaliza- Fig. 4A vs. C shows that the D47 of carbon dioxide pro- tion (Huntington et al., 2009). Several of the measurements duced from acid digestion of the calibration samples mea- of foraminifera and coccoliths made in this study required a sured in this study exhibits a negative correlation with the substantial increase in counting times and numbers of independently known Levitus temperatures. An alternative replicated samples, relative to past clumped isotope studies, means of estimating calcification temperatures for modern in order to achieve precisions appropriate for paleoceano- samples is based on combining measurements of carbonate graphic application. Most of these data are relatively pre- and water d18O(Table 2 of Electronic Appendix 2). Water cise (SE of 0.005–0.011&) and have approximately half d18O can be estimated for modern samples using a pub- the uncertainty typical of such data (0.011–0.025&) lished database (LeGrande and Schmidt, 2006). For these (Huntington et al., 2009). calculations, we used the inorganic fractionation factors de- A histogram of the standard errors for each sample is scribed in Section 3.3. shown in Fig. 2 of Electronic Appendix 1, with separate dis- Fig. 4D and F show D47 values of samples measured in tributions shown for samples that were cleaned to remove this study compared to d18O-based calcification tempera- organic matter and those samples that were analyzed un- tures. Systematic errors can be introduced into temperature cleaned. The average standard error of sample means for calculations that use this method simply due to the presence uncleaned samples is 0.019 ± 0.014&. In contrast, the aver- of non-equilibrium isotope fractionation (i.e., vital effects) age standard error of sample means for cleaned samples are for oxygen isotopes in foraminifera and coccoliths (Bemis about 1/2–1/3 this value (0.008 ± 0.003&). In some cases, et al., 1998; Ziveri et al., 2003). As a result, we use the first the standard error for each sample mean was reduced to method for estimating calcification temperatures in the 0.003&, corresponding to temperature uncertainties of remainder of this paper except where noted. However, the $ 18 ±0.5–0.8 °C (1 SE). Low errors similar to these should be use of d O-based calcification temperatures to estimate routinely achievable through the measurement of large the actual growth temperatures of our calibration samples samples, many replicate analyses, and sample pre-treatment does not significantly impact the results for all samples with to remove organic contaminants that can introduce isobars the exception of the thermocline-dwelling planktics (Fig. 4B within the relevant mass range. vs. E), and does not affect our conclusions (for example, results in Table 6 of Electronic Appendix 2 show that the 3.4.2. Accuracy slope and standard error of a regression through the foram- We measured multiple standards repeatedly over this inifera and coccolith data calculated using d18O-calcifica- period, typically every 4–10 analyses, in order to determine tion temperatures is 0.05051 ± 0.00124, compared to accuracy and monitor long-term external precision. These 0.05037 ± 0.00119 calculated using Levitus temperatures). standards included NBS-19 which yielded a mean D47 of All but two of the measured cultured and core-top sam- 18 0.350 ± 0.008& (1 SE), mean d OV-PDB of 2.193 ± ples are within one standard error of the inorganic calcite 13 À 0.031&, and mean d CV-PDB of 1.845 ± 0.099&. These sta- data (Fig. 4). A sample of the thermocline-dwelling planktic ble isotope ratios are statistically indistinguishable at the foraminifera Globorotalia tumida is offset by about 0.04&. 95% confidence level to those previously reported (Coplen, The other three samples of G. tumida that were analyzed 1995; Ghosh et al., 2006a). do also systematically deviate by lesser amounts from the inorganic calcite calibration (brown squares in Fig. 4B 4. RESULTS AND DISCUSSION and E). The second sample is of the epifaunal foraminiferal species Planulina wuellerstorfi with growth temperatures of 4.1. Relationship between temperature and D 0.8 ± 0.2 °C, and is also offset to lower D values (by 47 À 47 0.04&) relative to the inorganic calcite data, although the Theoretical calculations predict that in carbonate miner- other two samples of P. wuellerstorfi that have been ana- als the enrichment in 13C–18O bonds relative to a stochastic lyzed (that calcified at warmer temperatures) do not deviate distribution will decrease with increasing temperature from the inorganic dataset (green triangles in Fig. 4C and (Schauble et al., 2006; Guo et al., 2009). The relationship F). The 95% confidence interval for the D47 values of these between the D47 of CO2 produced during acid digestion of two samples overlap with the 95% confidence interval of the carbonate minerals and temperature has been determined inorganic calcite calibration. using measurements of inorganic calcite (Ghosh et al., These two samples belong to two populations of data 2006a), aragonitic fish otoliths (Ghosh et al., 2007), and that are notable because both populations systematically apatite (Eagle et al., 2010). Ghosh et al. (2006a) reported deviate from the inorganic calcite data (and the other bio- the following D47–temperature relationship for inorganic genic carbonate data shown in Fig. 5). The largest system- 0:0592 106 calcite: D 2Ã 0:02 (with T in K). Similar slopes atic offset between the inorganic calcite line and our data is 47 ¼ T À have been reported for otoliths (Ghosh et al., 2007) and observed in thermocline-dwelling planktic foraminifera, biologically precipitated apatite (Eagle et al., 2010); the which generally have lower D47 values than the remainder intercept for otoliths was reported to be significantly lower. of the samples (Figs. 4C and 5B). Estimated growth temper- For calibration samples measured in this study (Table 2 atures are more uncertain for these samples than for others of Electronic Appendix 2), calcification temperatures were surveyed in this study, due to the presence of strong vertical estimated from cruise reports and the Levitus Ocean Atlas gradients in seawater temperature for the environmental (Levitus et al., 1994). These calcification temperatures are niche inhabited by these organisms. If we assume the 13C–18O isotope signatures and ‘clumped isotope’ thermometry 5703

Fig. 4. D47 data (mean and 1 SE) compared to independent estimates of growth temperature (mean and range) for foraminifera and coccolith samples. Also shown is relationship for inorganic calcite. (A) Data for mixed-layer dwelling planktic foraminifera, coccoliths, and bulk carbonate compared to atlas temperatures (Levitus et al., 1994). (B) Data for thermocline-dwelling planktic foraminifera compared to atlas temperatures (Levitus et al., 1994). (C) Data for benthic foraminifera compared to atlas temperatures (Levitus et al., 1994). (D–F) Same as (A)–(C) except compared to d18O-calciﬁcation temperatures. inorganic calcite calibration applies to these samples, then estimated using the Levitus database (Fig. 4B) and may it suggests that growth temperature may have been under- be more accurately estimated using d18O-calciﬁcation 5704 A.K. Tripati et al. / Geochimica et Cosmochimica Acta 74 (2010) 5697–5717

temperatures (Fig. 4E). We assume that inaccuracies in estimating growth temperature for thermocline-dwelling planktics accounts for the large range of values observed in the D47 residuals calculated using the Levitus database (i.e., the vertical range observed in Fig. 6B and C). All of the benthic foraminiferal samples from the Arctic Ocean also are offset to heavier values relative to the inorganic calcite line (Fig. 4). One sample is of the benthic foraminifera P. wuellerstorfi, as described above. A second sample of the infaunal species Oridorsalis umbonatus is offset by 0.02& in the same direction (purple triangles in Fig. 4E and F) relative to inorganic calcite. Other samples of P. wuellerstorfi and O. umbonatus analyzed that calcified at warmer temperatures do not exhibit a systematic offset from the inorganic dataset. The offset observed in these Arctic benthic foraminifera does not appear to be due to a carbonate ion effect (Fig. 7A). We discuss the possible significance of these observed offsets in Section 4.3.2.

4.1.1. Calibration of the carbonate ‘clumped isotope’ thermometer Table 6 of Electronic Appendix 2 compiles the slopes and intercepts for D47–temperature relationships calculated using different populations of data taken from this study and other publications (Ghosh et al., 2006a,b, 2007; Came et al., 2007; Eagle et al., 2010; Huntington et al., 2010). We applied weighted least-squares linear regression analysis to the data and used the regression model of York (1969) (described in the appendix of Huntington et al., 2009). Each calibration sample has an uncertainty associated with both the determined D47 value and the estimated growth temperature. This regression model assumes that the errors in both variables are uncorrelated, with a symmetrical solution that yields the same regression and errors if the variables are switched. The values for slopes and intercepts, and their 95% confidence bands, are shown in Fig. 8. A new calibration 6 0:05875 0:0133 10 through the inorganic calcite data is D ðÆ 2 ÞÃ 47 ¼ T À 0:014 0:138 , similar to the calibration reported by 0ðÆ 4Þ Ghosh et al. (2006a). Fig. 4 shows that the confidence bands for the inorganic calcite calibration line would overlap with the 95% confidence bands for all the samples (the plotted error bars are 1 SE). A weighted least-squares linear regression of any of the other populations (of data for biogenic carbonates or through all of the data) yields a slope and intercept that is shallower but indistinguishable from the value determined for inorganic calcite (Figs. 5 and 8). Fig. 8 shows that the slopes and intercepts of all the populations are statistically equivalent at the 95% confidence level. The slope of the equilibrium line therefore may be shallower than the Ghosh et al. (2006a) inorganic calibration and be better Fig. 5. D47–temperature data for calibration samples from this approximated by the more precisely determined slope of study and other publications compared to calibration for inorganic the regression through all of the inorganic and biogenic data calcite (yellow lines). See Fig. 8 for a comparison of slopes and (Table 6 of Electronic Appendix 2), which is closer to theo- intercepts for these and other populations. (A) Data with precise retical predictions (Schauble et al., 2006; Guo et al., 2009). constraints on growth temperatures and regression through the 13 18 data. (B) All data (excluding Red Sea coral) and regression through These observations indicate that C– O bond abun- the data. (C) Regression through all data excluding benthic dance in most or all of these biogenic carbonates are statis- foraminifera. (For interpretation of the references to color in this tically indistinguishable from inorganic calcite that is figure legend, the reader is referred to the web version of this assumed to have precipitated in equilibrium. Although article.) the data as a group conform to the slope predicted by the 13C–18O isotope signatures and ‘clumped isotope’ thermometry 5705

18 Fig. 6. Comparison of D47 and d O residuals (measured minus predicted from an inorganic calibration) to each other. To estimate residuals, D47 data were compared to the new inorganic calibration derived in this study listed in Table 6 of Electronic Appendix 2 (which is similar to the Ghosh et al. (2006a) calibration). d18O data for calcite samples were compared to values calculated using the inorganic calibration of Kim and O’Neil (1997); data for aragonite samples were compared to values calculated using the calibration of Kim et al. (2007). Temperature estimates used for calculated inorganic values are from Levitus et al. (1994) with the exception of (C) (estimates use d18O-based calcification temperatures). With the exception of the data shown in (B) and (C), the slopes of linear regressions through all of the data in each figure panel do not significantly differ from zero at the 95% confidence level; no significant species-specific trends are observed. (A) Comparison of D47 residuals to d18O residuals for different species of mixed-layer dwelling planktic foraminifera and bulk carbonate. (B) Same as (A) but for thermocline-dwelling planktics. The slope of a linear regression through these data is 0.026 ± 0.010 (1 SE) and is significantly non-zero (p = 0.05). We attribute this systematic deviation to uncertainties in estimating growth temperatures for thermocline-dwellers. (C) Same as (B) but uses d18O-based calcification temperatures. (D) Same as (A) and (B) but for benthic foraminifera. regression through the inorganic calcite data, a potential thermometer may be simpler to interpret than the d18O, source of uncertainty stems from the small range of growth Mg/Ca and TEX86 thermometers. The relationship between temperatures among samples of any one species. Therefore measured stable isotope compositions and seawater chemis- any attempt to assess whether each species individually con- try is shown in Figs. 7 and 9. The d13C of dissolved inor- forms to the same overall trend as the group is subject to a ganic carbon and the d18O of water (Table 3 of Electronic large uncertainty. We also note that the regression through Appendix 2) are both uncorrelated with the deviations in data for benthic foraminifera (Fig. 8) has a lower slope and D47 from apparent equilibrium (Fig. 9). These observations higher intercept than the values for the inorganic calcite are consistent with predictions that bond order should be 18 13 regression due to the low D47 values observed in the two independent of the d O of water and the d C of dissolved samples from the Arctic Ocean (with temperatures of inorganic carbon (Schauble et al., 2006). 0.8 °C), although these systematic deviations do not ap- Although the dataset is small, no evidence for a carbonate À pear to be taxon specific, as the samples of O. umbonatus ion effect is apparent in the D47 data for benthic foraminifera and P. wuellerstorfi from warmer settings do not exhibit (Fig. 7). A relationship is observed for d13C in these data, and similarly large departures from the inorganic calcite calibra- has been reported for foraminiferal d18O in previous work tion (or the remaining population of biogenic data). (Spero et al., 1997). The observations of a carbonate ion effect on carbonate d18O and d13C in inorganic precipitation 4.2. Relationship between seawater composition and D47 experiments is hypothesized to reflect the difference in 18 16 13 12 2 O/ O and C/ C ratios in HCO3À and CO3 À ions, and Seawater composition appears to have a negligible influ- the pH-dependence of speciation for dissolved inorganic car- 13 18 ence on D47. In this respect, at least, the clumped isotope bon (e.g., Zeebe, 2007). Similar abundances of C– O 5706 A.K. Tripati et al. / Geochimica et Cosmochimica Acta 74 (2010) 5697–5717

18 13 Fig. 7. Comparison of D47, d O, and d C residuals (measured minus predicted from an inorganic calibration) for benthic foraminiferal samples to saturation state of bottom waters. (A) Comparison of D47 residuals for data from benthic foraminifera to estimates of the calcite 2 2 (or aragonite) saturation state of bottom waters, defined as the difference between measured [CO3 À] and the [CO3 À] at calcite (or aragonite) 2 18 saturation (labelled DCO3 À in the figure). (B) Comparison of d O residuals for data from benthic foraminifera to estimates of the saturation state of bottom waters. (C) Comparison of d13C residuals for data from benthic foraminifera to estimates of the saturation state of bottom waters. bonds (relative to a random distribution) have been pre- measured stable isotope values and in the temperature cali- 2 18 dicted for both HCO3À and CO3 À ions (Schauble et al., brations for d O and D47 (Table 7 of Electronic Appendix 2). 2006; Guo, 2008; Guo et al., 2008), although there are no Given the general conformity of the D47 data to the inor- observational constraints available. At 25 °C, the equilib- ganic carbonate calibration (Fig. 4), these deviations could rium constant for the ‘clumping’ reaction in calcite (Reaction potentially be used to assess if d18O ratios in certain taxa 1) is predicted to be 1.000410, while the equilibrium constant are subject to significant non-equilibrium fractionations. for the analogous isotope exchange reaction between Most of the data are within 2& of the 1:1 line shown in HCO3À isotopologues is estimated to be 1.000427. The theo- Fig. 10, consistent with evidence for species-specific offsets retically predicted 0.02& difference in 13C–18O bond abun- of 0–2& which have been reported for several of the taxa (Be- $ 2 dance between dissolved HCO3À and CO3 À ions is subtle mis et al., 1998). If the species-specific offsets (Table 7 of Elec- (Guo, 2008; Guo et al., 2008). The similarity in 13C–18O bond tronic Appendix 2) can be attributed solely to ‘vital effects’ excess for carbonate and bicarbonate ions implies there may and are precisely determined, then the results indicate that be a weak and possibly pH effect on clumping in marine car- the magnitude of disequilibrium fractionation for G. ruber bonates given the range of pH observed within foraminifera and P. obliquloculata < G. sacculifer and G. bulloides < ben- and other organisms (Erez et al., 2008). thic foraminifera, and for G. menardii < G. hirsuta < G. tumida < G. truncatulinoides. The small sample sizes, however, 4.2.1. Calculated seawater 18O values mean that this analysis is not statistically significant for most In Fig. 10 we show a comparison of observed (i.e., mea- of the species, with the possible exceptions of G. ruber (n = 6) sured) seawater d18O(x-axis; LeGrande and Schmidt, and G. sacculifer (n = 7) which appear to have shell d18O val- 2006) to estimated water d18O values for the foraminifera ues that are close to equilibrium with seawater. The same ap- and coccolith samples (y-axis), with average species-specific proach could potentially be applied to larger datasets 18 differences reported in Table 7 of Electronic Appendix 2. comprised of paired d O and D47 measurements of core- The calculated water d18O values use measurements of car- top foraminifera samples as a method of constraining the 18 bonate d O, measurements of D47 converted to temperature magnitude of vital effects in different species. using the new regression through inorganic calcite data, and published relationships for the temperature-dependent oxy- 4.3. The significance of D47 observations for the origin of vital gen isotope fractionation between inorganic calcium carbon- effects in 18O and 13C ate (Kim and O’Neil, 1997; Kim et al., 2007). The difference between these estimated values and measured water d18O All of the samples we have measured with precisely con- (LeGrande and Schmidt, 2006) reflects uncertainties in the strained growth temperatures (i.e., mixed-layer planktic 13C–18O isotope signatures and ‘clumped isotope’ thermometry 5707

Fig. 8. Slope and intercepts (mean and 95% confidence interval) for D47–temperature relationships determined using different populations of calibration samples listed in Table 6 of Electronic Appendix 2. The large horizontal bar corresponds to the 95% confidence interval for the inorganic calcite relationship as determined in this study, and overlaps with all of the other regressions. The thin horizontal bar corresponds to the 95% confidence interval for a regression through all data, and overlaps with almost all of the other regressions. foraminifera, coccoliths, bulk carbonate, benthic foraminif- on the origin of vital effects, and therefore on the mecha- era excepting two samples of benthic foraminifera from the nism of biomineralization, in foraminifera and coccoliths. Arctic Ocean) do not exhibit large deviations from the inor- While additional experimental and theoretical studies are ganic calibration (differences are <0.02&; Fig. 5). Given the needed to fully understand the significance of this result, taxonomic breadth represented by these samples, these re- we discuss some possible explanations of the following sults suggest that foraminifera and coccoliths generally observations: (1) values of D47 that are statistically indistin- grow at equilibrium with respect to their proportions of guishable from inorganic calcite (the Arctic benthic foram- 13C–18O bonds, despite significant ‘vital effects’ of up to inifera that are the exception to this trend is discussed in 1–2& expressed in d18O(Fig. 6) and d13C(Fig. 11). In al- more detail in Section 4.3.2), (2) enriched/depleted d18O most all of the samples, there is no evidence for species-spe- and d13C relative to inorganic calcite, and (3) small depar- cific offsets from inorganic calcite in D47, and deviations of tures in D47 from inorganic calcite that are statistically 18 18 D47 and d O from inorganic values are generally not corre- uncorrelated with variations in d O. lated with each other (Fig. 6). Fig. 11 shows that the same is One model that has been proposed is that oxygen and 13 true for D47 and d C. This result provides a new constraint carbon isotope disequilibrium in carbonate minerals may 5708 A.K. Tripati et al. / Geochimica et Cosmochimica Acta 74 (2010) 5697–5717

18 Fig. 9. Comparison of D47 residuals (measured minus predicted from an inorganic calibration) to independent estimates of water d O and the 13 d C of dissolved inorganic carbon (Table 3 of Electronic Appendix 2). To estimate residuals, D47 data were compared to values calculated using the inorganic calibration derived in this study (which is similar to the Ghosh et al. (2006a) calibration) and temperature data from

Levitus et al. (1994). (A) Comparison of D47 residuals for different species of mixed-layer dwelling planktic foraminifera and bulk carbonate to water d18O. (B) Same as (A) but for thermocline-dwelling planktics. (C) Same as (A) and (B) but for benthic foraminifera. (D–F) Same as (A)– (C) but for the d13C of dissolved inorganic carbon. arise if carbonate precipitation is from a mixture of dis- calcify from a mixture of dissolved inorganic carbon spe- solved inorganic carbon species that are at local isotopic cies, or from an internal dissolved inorganic carbon reser- equilibrium, and not just from carbonate ions (Zeebe, voir that differs in pH and/or isotopic composition from 1999, 2007). The speciation of dissolved inorganic carbon seawater but maintains local equilibrium. Metabolic CO2 is pH-dependent. Several studies have determined that the can also contribute to the dissolved inorganic carbon pool d18O and d13C of the individual species are isotopically dis- of foraminifera and coccoliths (Spero et al., 1997; Adkins tinct at a given temperature (Usdowski and Hoefs, 1993; et al., 2003; Erez, 2003). The pattern of isotopic fraction- Zeebe, 1999, 2007; Beck et al., 2005). As pH varies, the iso- ation we report here is consistent with this “pH hypothe- topic composition of the DIC pool is predicted to change. sis,” given the pH-dependence of dissolved inorganic The different equilibrium fractionations between dissolved carbon speciation, coupled with the relatively large differ- inorganic carbon species and water, for example, results ence in equilibrium 18O/16O ratios (and 13C/12C ratios) 18 2 in a pH-dependence for the bulk d O of the dissolved inor- for HCO3À and CO3 À (Zeebe, 1999, 2007, 2009; Beck ganic carbon pool (e.g., Zeebe, 2007). et al., 2005) and the theoretically predicted pattern of 13 18 It is known that carbonate-precipitating organisms can C– O bond abundance for dissolved HCO3À and 2 maintain differences in pH between internal and external CO3 À (Guo, 2008; Guo et al., 2008). environments (e.g., across membranes) (Spero et al., 1997; Another possibility is that vital effects or isotopic disequi- Adkins et al., 2003; Erez, 2003), and in foraminifera, invag- librium in d18O and d13C reflect kinetic isotope effects during inated seawater vacuoles exhibit a range of pH values from the hydration and hydroxylation of CO2 (which would drive 6 to 9 (Erez et al., 2008). It may be that these organisms 18O depletion and 13C depletion), perhaps combined with 13C–18O isotope signatures and ‘clumped isotope’ thermometry 5709

Fig. 10. Comparison of water d18O(LeGrande and Schmidt, 2006) to calculated water d18O values for individual samples based on combining 18 D47-based temperatures (using new inorganic calibration shown in Table 6 of Electronic Appendix 2), measured carbonate d O, and published carbonate–water oxygen isotope fractionation factor (Kim and O’Neil, 1997; Kim et al., 2007). Diagonal line in all panels is a 1:1 line. Mean species-specific differences are reported in Table 7 of Electronic Appendix 2. (A) Data for different species of mixed-layer dwelling planktic foraminifera and bulk carbonate. (B) Same as (A) but for thermocline-dwelling planktics. (C) Same as (A) and (B) but for benthic foraminifera. metabolic changes in the d13C of the dissolved inorganic equilibrium, as observed in some foraminiferal taxa carbon pool that derive from photosynthesis and/or respira- (Fig. 6) and other biogenic carbonates (e.g., McConnaughey, tion (McConnaughey, 1989). If the rates of reaction for var- 1989). In addition, kinetic isotope effects during the hydra- ious isotopically normal, singly and doubly substituted tion and hydroxylation of CO2 are unlikely to be important isotopologues of CO2 differ, then a kinetic origin for the vital for organisms in which carbonic anhydrase is present, as this effect could produce non-equilibrium clumped isotope values enzyme catalyzes equilibrium between carbonate species and in calcium carbonate, and might results in observations of water. Different forms of carbonic anhydrase have been correlated deviations from inorganic calcite in d18O, d13C, detected in a number of organisms but its direct role in the and D47. No such patterns are systematically observed in calcification process is uncertain (e.g., coccoliths, corals, our data set (Figs. 6 and 11). The ‘kinetic’ hypothesis for molluscs, eggshells; Gay and Mueller, 1973; Miyamoto the origin of vital effects in foraminifera and coccoliths can et al., 1996; Ziveri et al., 2003). An important caveat is that only be consistent with our observations if the kinetically despite its occurrence in many systems, equilibration of the controlled oxygen and carbon isotope fractionations lead dissolved inorganic carbon system and associated isotopic to proportions of 13C–18O bonds that are typically within equilibrium in biologically precipitated minerals is only 0.01& of thermodynamic equilibrium across a range of likely to occur if (1) carbonic anhydrase specifically occurs $ temperatures, which is possible but has not yet been at the site of uptake/production, (2) it is present at the site demonstrated. of calcifying reactions, and (3) is working efficiently. The Several observations argue against the ‘kinetic’ hypothe- observation, however, that D47 residuals for almost all sam- sis for the origin of vital effects in foraminifera and coccoliths ples are apparently normally distributed (see discussion in (excepting the Arctic benthic foraminifera). Counter- Section 4.5) is also consistent with the hypothesis that kinetic examples of kinetically controlled precipitation mechanisms isotope effects are not discernable in most of the biogenic that lead to large departures from equilibrium D47 values data. Future theoretical calculations and experiments are of product carbonate are known (Ghosh et al., 2006a; necessary, however, to quantify isotopic fractionations of Affek et al., 2008; Guo, 2008). It is not clear how kinetic doubly substituted isotopologues associated with different models would account for enrichments in d18O relative to reactions (i.e., hydroxylation and hydration reactions; 5710 A.K. Tripati et al. / Geochimica et Cosmochimica Acta 74 (2010) 5697–5717

13 Fig. 11. Comparison of D47 and d C residuals (measured minus predicted from an inorganic calibration) to each other. To estimate residuals, D47 data were compared to values calculated using the inorganic calibration derived in this study (which is similar to the Ghosh et al. (2006a) calibration) and temperature estimates from Levitus et al. (1994). d13C data for all samples were calculated as discussed in the methods. (A) 13 Comparison of D47 residuals to d C residuals for diﬀerent species of mixed-layer dwelling planktic foraminifera and bulk carbonate. (B) Same as (A) but for thermocline-dwelling planktics. (C) Same as (A) and (B) but for benthic foraminifera.

protonation of carbonate ion and deprotonation of bicar- D47. If precipitation is slow, there may be sufficient time bonate ion) accurately. for newly formed carbonate ions to isotopically re-equilibrate with water and dissolved inorganic carbon, so that 18 13 4.3.1. Mechanisms for recording solution pH in solid phase d O, d C, and D47 signatures are all close to equilibrium. 2 The “pH hypothesis” requires that the HCO3À/CO3 À Observations supporting this explanation are discussed in ratio of ambient dissolved inorganic carbon to be preserved Section 4.3.2. in the solid phase, although a mechanism for recording Alternately, biologically precipitated carbonates may be solution pH in calcium carbonate has not yet been eluci- quantitatively precipitated from a pool of HCO3À and 2 dated (Zeebe, 1999, 2007; Adkins et al., 2003; Watson, CO3 À, and therefore the pH of the original solution (which 2004). Although both bicarbonate and carbonate ions sets the proportions of these dissolved inorganic carbon may be present at the site of calcification, surface complex- species) is recorded in the solid phase (Zeebe, 1999, 2007). ation modelling of the carbonate–water interface indicates The reservoir being quantitatively precipitated need not that the mineral surface preferentially sorbs carbonate ion be an entire seawater vacuole or seawater-like calcifying relative to bicarbonate (Van Cappellen et al., 1993; Kim space, although this is possible. Instead, it should be suffi- 2 et al., 2006). One possible explanation is that the ratio of cient to quantitatively precipitate HCO3À and CO3 À ions chemisorbed carbonate and bicarbonate at the mineral sur- in the solution boundary layer, or at the site of calcification face in the site of calcification is proportional to the solu- itself, as long as there is local isotopic equilibrium. tion carbonate/bicarbonate ratio. If precipitation is A third possibility is that if there is a relationship be- sufficiently rapid (relative to the timescale for equilibration tween calcification rate (which is related to carbonate ion of dissolved inorganic carbon species), when deprotonation concentration) and the occurrence of lattice defects, the of bicarbonate occurs there is not time for the newly formed amount of HCO3À directly incorporated in the calcite struc- carbonate ion to isotopically re-equilibrate with the solu- ture may function to maintain local charge balance (Duffy tion, resulting in apparent vital effects in d18O and d13C, et al., 2005). This idea is consistent with the recent observa- and subtle effects that may or may not be detectable for tion using nuclear magnetic resonance spectroscopy that 13C–18O isotope signatures and ‘clumped isotope’ thermometry 5711

HCO3À is present in synthetic and natural biogenic and bonic anhydrase present at the site of calcification and abiogenic calcite, with significant structural accommoda- working efficiently, or in organisms that do not have car- 2 tion associated with the substitution of HCO3À for CO3 À bonic anhydrase present at the site of mineral formation groups (Feng et al., 2006). but mineralize extremely slowly. The converse should be observed for other calcifying organisms that mineralize rap- 4.3.2. Data for samples from near-freezing temperatures: idly and do not have carbonic anhydrase present at the site possible kinetic effects? of calcification or working efficiently. Data for the two samples of benthic foraminifera from 0.8 °C do exhibit deviations of up to 0.02& to 4.3.3. Uncertainties in the inorganic calcite calibration and À À 0.04& from the calibration for inorganic calcite implications À (Fig. 4C and F). It is possible that (1) these offsets simply It is possible that the observed differences between the reflect measurement error, (2) that CO2 produced during biogenic data and the published inorganic calcite data at acid digestion of these samples partially re-equilibrated dur- low temperatures reflect uncertainties in the inorganic cal- ing sample preparation, or (3) that the samples still retained cite calibration. The D47 value for the sample grown at contaminating phases (e.g., coccoliths). It is also possible near-freezing temperatures is poorly constrained (Ghosh that these departures in D47 reflect non-equilibrium isotope et al., 2006a). The inorganic calcite calibration is based effects due to the slow kinetics associated with the hydra- on replicate measurements of synthetic materials precipi- tion and hydroxylation of aqueous CO2 (McConnaughey, tated at 23, 33, and 50 °C, and a single analysis of a syn- 1989). Another kinetic mechanism proposed for depletion thetic sample that was grown at 1 °C that had a in d18O and d13C in carbonate minerals relative to equilib- measurement uncertainty of 0.016&. The samples at 23, rium is diffusion; however, it is unlikely that the pattern re- 33, and 50 °C have carbonate d18O values that are lower flects isotopic discrimination during diffusion of carbonate than predicted by the Kim and O’Neil (1997) equation by species to the growing crystal surface (Turner, 1982; Gross- 1.07&, 0.39&, and 0.04&, respectively. Uncertainties À À À man, 1984) because it should drive enrichment in D47 (Eiler in the temperature of precipitation contribute an error in and Schauble, 2004). the calculation of equilibrium values of 0.25&, 0.50&, These observations could indicate that it is the relative and 0.25&. This inorganic calcite sample precipitated at balance of the timescale for equilibration relative to the rate cold temperatures also had a measured d18O value that of precipitation that governs whether a kinetic isotope effect was 0.54 ± 0.05& lighter than the equilibrium value. 18 13 18 is expressed in D47 (and also in d O and d C). Given that It is possible that the d O and D47 ratios reported for the timescales for dissolved inorganic carbon equilibration this synthetic sample (and possibly the 23 °C sample) may are longer at lower temperatures (Zeebe and Wolf-Glad- reflect a kinetic isotope effect that drives d18O to low values row, 2001), it is possible that the observed departures for and D47 to higher values, relative to equilibrium. If true, the two low-temperature benthic foraminiferal samples re- then the D47 values of the Arctic Ocean benthic foraminif- flect a kinetic isotope effect. Constraints on growth rate, eral samples may be a closer determination of equilibrium however, are not available to evaluate whether the rate of calcite D47 values at low temperatures, and the D47 values carbonate precipitation by these organisms is sufficiently ra- previously reported for deep-sea coral (Ghosh et al., pid relative to the timescale for equilibration of dissolved 2006a) could in fact reflect a kinetic isotope effect. One ki- inorganic carbon species. netic process that would result such an effect in d18O (deple- However, it is interesting to note that similarly large tion relative to equilibrium) and D47 (enrichment relative to deviations from inorganic calcite are not reported in equilibrium) is diffusion (Eiler and Schauble, 2004). It is deep-sea corals that have grown at low temperatures interesting to note that the Red Sea coral reported in Ghosh (Fig. 3)(Ghosh et al., 2006a). This difference may reflect et al. (2006a) exhibits a similar pattern of apparent isotopic the extremely slow growth rates of deep-sea coral, and/or disequilibrium. Careful studies of kinetic isotope effects the efficient activity of carbonic anhydrase. Growth rates from growth experiments are necessary to fully explore for deep-sea coral of 0.04–2 mm/year have been reported the significance of these observations in biogenic for Desmophyllum cristagalli and Caryophyllia profunda, carbonates. and higher rates of up to 6–8 mm/year reconstructed for Enallopsammia rostrata and Lophelia sp. (Wilson, 1979; 4.4. Mineral structure effects Bell and Smith, 1999; Risk et al., 2002; Adkins et al., 2004). If this explanation for large (0.02–0.04&) deviations We observe no discernable mineralogic differences in D47 from the inorganic calcite calibration is true, then we would between aragonitic and calcitic biogenic carbonates predict that detailed measurements may show that the slow- (Figs. 12 and 13). D47 analyses of other aragonitic biogenic growing deep-sea coral D. cristagalli is closest to equilib- carbonates, such as deep-sea coral and mollusks, also rium, whereas the faster growing taxa Lophelia sp. and E. resemble the inorganic calcite calibration (Ghosh et al., rostrata exhibit larger deviations from equilibrium. Other 2006a; Came et al., 2007; Eiler, 2007; Huntington et al., taxa from cold-water environments with rapid growth rates 2010). These observations are consistent with ab initio cal- (i.e., molluscs) might also show much larger deviations culations that predict D47 differences between carbonate from equilibrium than the data previously reported. Final- minerals with different mineral structures should be small 18 ly, we would predict that D47 and d O deviations from (less than 0.02& between aragonite and calcite at 25 °C) equilibrium should not be observed in organisms with car- (Schauble et al., 2006; Eagle et al., 2010). These similarities 5712 A.K. Tripati et al. / Geochimica et Cosmochimica Acta 74 (2010) 5697–5717

Fig. 12. Comparison of measured D47 values to D47 residuals (measured minus predicted from an inorganic calibration) for data from foraminifera, coccoliths, and bulk carbonate. may arise if the diﬀerent carbonate minerals preserve a state of ordering that reﬂects isotopic equilibration of the dissolved inorganic carbon species, as proposed above, rather than of carbonate ions within the solid itself. The observed slopes through both the inorganic and biogenic data are similar but slightly steeper than the theoretical values (Schauble et al., 2006; Guo et al., 2009)(Fig. 8). The 0.001–0.002&/°C discrepancy is potentially attributable to uncertainties in the theoretical calculations.

4.5. Utility for paleoclimatic and paleoceanographic studies

If the isotope systematics (including D47 values) of foraminifera and coccoliths generally conform to our observations, then the application of carbonate clumped isotope thermometry of the biogenic carbonates could significantly improve our ability to reconstruct past temperatures and seawater composition. The observations reported above suggest that nearly all of the biogenic samples that have been studied are indistinguishable from inorganic calcite and therefore likely precipitated in apparent equilibrium. These observations imply that by using this proxy, we are essentially exchanging unknown sources of error of an empirically calibrated proxy with known errors from clumped isotopes that we can interrogate statistically. The standard error of sample means for cleaned samples (Fig. 2 of Electronic Appendix 1) are indistinguishable from a normal distribution at the 95% confidence level (Shapiro– Wilk W test; p = 0.54). The difference between the measured D47 values from these biogenic carbonates and the values predicted from the inorganic calcite calibration (or the calibration through all of the data) are also indistinguishable from a normal distribution at the 95% confidence level (Fig. 14). This distribution is what we would predict if both the measurement errors and the residuals are simply due to random errors in counting statistics with a Poisson Fig. 13. Comparison of D47–temperature relationships for samples distribution. with different mineralogy to theoretical calculations (Guo et al., On the basis of these observations, we suggest that the 2009). The regressions for different populations of data have a steeper slope and shallower intercept to the modelled values. (A) calibration derived from all inorganic and biogenic carbon- Data for aragonitic and calcitic samples (mean and 1 SE for ate data with precise temperature constraints should be y-values and mean and range for x-values shown) compared to used for paleothermometry (Table 6 of Electronic Appen- 6 regressions through inorganic calcite data, all data, and modelled 0:05208 0:00099 10 dix 2). This calibration is D47 ðÆ 2 ÞÃ ¼ T þ calcite and aragonite relationships. (B) Comparison of modelled 0:06274 0:01157 . The average difference between temper- and observed regression lines for calcite. (C) Comparison of ðÆ Þ atures predicted using this calibration (or the inorganic modelled and observed regression lines for aragonite. 13C–18O isotope signatures and ‘clumped isotope’ thermometry 5713

Fig. 14. Histogram of D47 residuals (measured minus predicted from inorganic calibration). (A) Estimates for surface-dwelling planktic foraminifera, coccoliths, and bulk sediment compared to a normal distribution. Data are indistinguishable from a normal distribution at the 95% conﬁdence level (Shapiro–Wilk W test; p = 0.43). The same is true if residuals are calculated using the calibration through all of the data (p = 0.08). (B) Estimates for all samples with precisely known growth temperature data (data shown in Fig. 5A). Grey bars are for the data shown in (A); white bars are for benthic foraminifera; light blue bars are for published data. The distribution of residuals is also indistinguishable from a normal distribution (p = 0.43). The same is true if residuals are calculated using the calibration through all of the data (p = 0.09). (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.)

calcite calibration) and independently estimated calcifica- long analytical durations, and/or extensive replication in tion temperature is 0.7 °C ± 2.5 °C (1r), similar to what order to reduce the errors of D measurements to the level $ 47 has been reported for calibrations of other temperature necessary for precise temperature determinations. Our anal- proxies including d18O(Bemis et al., 1998) and Mg/Ca yses of such samples suggest that clumped isotope ther- (Anand et al., 2003; Elderfield et al., 2006). mometry can yield accurate and precise determinations of In addition to allowing us to accurately reconstruct tem- past calcification temperatures when applied to foraminif- perature, the development of D47 records for sites where era and coccoliths. The results of this detailed survey of there are existing proxy data would help leverage the multiple taxa implies that ‘vital effects’ in D47 may be unli- interpretation of other proxies. We suggest that clumped kely in extinct species (in contrast to d18O and d13C), isotopes can potentially be applied to assess whether differ- although additional studies on the systematics governing ences between existing proxy reconstructions reflects envi- ‘clumping’ in carbonates are necessary. In this study, we ronmental factors (i.e., fluid composition, seasonality and observe that the D47 of primary carbonate precipitated by depth of production) or other factors. Proxy “carrier” foraminifera and coccoliths and other biogenic carbonates phases (e.g., alkenones vs. planktic foraminifera) may have is generally indistinguishable from inorganic precipitates. different depths and seasons of maximum abundance. As a Clumping of heavy isotopes of carbon and oxygen into result, reconstructions for the same site may not reflect con- bonds with each other is independent of the d18O of the ditions during the same time of year or at the same water water in which the mineral precipitated, and the d13C of dis- depths (e.g., Mix, 2006). Different carrier phases also do solved inorganic carbon. No evidence for a relationship be- not have the same susceptibility to secondary processes tween pH and 13C–18O bond order is discernable in the such as lateral transport and water column dissolution limited dataset presented, and no mineralogic differences (e.g., Benthien and Mueller, 2000). One direct application are observed. In the species we have examined, the mecha- of clumped isotope thermometry could be to assess whether nism(s) responsible for significant non-equilibrium fraction- 18 apparent mismatches in sea surface temperature reconstruc- ation in d O do not bias D47, with a possible exception tions from alkenones and foraminiferal Mg/Ca data for the temperatures a few cases where D47 values deviate by Last Glacial Maximum at specific sites (MARGO Project 0.02–0.04& relative to inorganic calcite and other biogenic Members, 2009) arise due to some of these processes. carbonates. These data constrain the upper magnitude of D47 frac- 5. CONCLUSIONS tionation in models of stable isotope vital effects that rely on equilibrium or alternative kinetic mechanisms. Our A challenge for all forms of carbonate clumped isotope findings suggest that in foraminifera and coccoliths, dis- thermometry is the need for very large and clean samples, solved inorganic carbon at the mineral–solution interface 5714 A.K. Tripati et al. / Geochimica et Cosmochimica Acta 74 (2010) 5697–5717 is generally indistinguishable from local isotopic equilib- APPENDIX A. SUPPLEMENTARY DATA rium, and appears to be consistent with the hypothesis that solution pH may be recorded in the solid phase Supplementary data associated with this article can be (Zeebe, 1999, 2007; Usdowski and Hoefs, 1993). This found, in the online version, at doi:10.1016/j.gca.2010. may occur if dissolved inorganic carbon is quantitatively 07.006. precipitated from reservoirs such as seawater vacuoles or 2 at the site of calcification, if CO3 À and HCO3À groups chemisorbed to the mineral surface are present in propor- REFERENCES tion to their abundance in the adjacent reservoir of dis- Adkins J., Boyle E., Curry W. and Lutringer A. (2003) Stable solved inorganic carbon, or if HCO À incorporation in 3 isotopes in deep-sea corals and a new mechanism for “vital lattice maintains local charge balance to compensate for effects”. Geochim. Cosmochim. Acta 67, 1129–1143. lattice defects. The balance of the timescale for equilibra- Adkins J., Henderson G., Wang S.-L., O’Shea S. and Mokadem F. tion of dissolved inorganic carbon species relative to the (2004) Growth rates of the deep-sea scleractinia Desmophyllum rate of precipitation could govern whether a kinetic iso- cristagalli and Enallopsammia rostrata. Geochim. Cosmochim. tope effect is expressed in ‘clumping’. More work is needed Acta 227, 481–490. to investigate the controls on D47 in biogenic carbonates Affek H. P., Bar-Matthews M., Ayalon A., Matthews A. and Eiler precipitated in environments where the timescale for equi- J. M. (2008) Glacial/interglacial temperature variations in librium of the dissolved inorganic carbon system is slow Soreq cave speleothems as recorded by ‘clumped isotope’ relative to typical carbonate precipitation rates, we note thermometry. Geochim. Cosmochim. Acta 72, 5351–5360. Anand P., Elderfield H. and Conte M. H. (2003) Calibration of that similar patterns in D , d18O and d13C (relative to 47 Mg/Ca thermometry in planktonic foraminifera from a sedi- inorganic calcite) have been observed in molluscs, brachio- ment trap time series. Paleoceanography 18, 28–31. pods, deep-sea corals, and bioapatite in teeth from mam- Arbuszewski J., Demenocal P. and Kaplan A. (2008) Towards a mals, reptiles, and fish (Ghosh et al., 2006a; Eiler, 2007; global calibration and validation of the G. ruber (white) Mg/Ca Eagle et al., 2010). As a result, we suggest that the same paleothermometer. Fall AGU Meeting. #PP41F-02 (abstr.). processes likely govern isotopic fractionation in these Beck W. C., Grossman E. L. and Morse J. W. (2005) Experimental other carbonate-precipitating organisms. If true, this may studies of oxygen isotope fractionation in the carbonic acid limit the applicability of the carbonate ‘clumped isotope’ system at 15°, 25°, and 40 °C. Geochim. Cosmochim. Acta 69, thermometer in some settings. 3493–3503. Bell N. and Smith J. (1999) Coral growing on North Sea oil rigs. Nature 402, 601. AUTHOR CONTRIBUTION STATEMENT Bemis B. E., Spero H., Bijma J. and Lea D. W. (1998) Reevaluation of the oxygen isotopic composition of planktonic foraminifera: A.T. designed the project, prepared and measured most experimental results and revised paleotemperature equations. of the samples, analyzed the results, interpretated the data, Paleoceanography 13, 150–160. and wrote and revised the manuscript. R.E. prepared and Benthien A. and Mueller P. J. (2000) Anomalously low alkenone measured samples, provided input to the analysis of the temperatures caused by lateral particle and sediment transport data, and drafted some of the figures. N.T. prepared and in the Malvinas Current region, western Argentine Basin. Deep measured samples. H.B. provided some of the samples used Sea Res. Part I 47, 2369–2393. in this study. P.H. cultured the coccoliths that were used in Boyle E. and Hester K. (1982) Water chemistry control of cadmium content in recent benthic foraminifera. Nature 298, 260–262. this study. J.E. trained A.T. in analysis and interpreta- D47 Boyle E. A. and Keigwin L. D. (1982) Deep circulation of the tion. All co-authors contributed to discussions of this work. North Atlantic over the last 200,000 years: geochemical evidence. Science 218, 784–786. ACKNOWLEDGMENTS Came R., Eiler J., Veizer J., Azmy K., Brand U. and Weidman C.

(2007) Coupling of surface temperatures and atmospheric CO2 Support for this project was provided by the UCLA Division concentrations during the Palaeozoic era. Nature 449, 198– of Physical Sciences, NERC, NSF, and Magdalene College. Sam- 201. ples for this study were obtained from the Godwin Laboratory Coplen T. (1995) Guidelines for the reporting of stable hydrogen, sample archives through David Thornalley, Harry Elderﬁeld, carbon, and oxygen isotope-ratio data. Paleoceanography 11, Mike Hall, and James Rolfe; from Lowell Stott, the Scripps Insti- 369–370. tute of Oceanography collections through Richard Norris, and the Das Sharma S., Patil D. and Gopalan K. (2002) Temperature

Ocean Drilling Program. We would like to thank Jeannie Booth dependence of oxygen isotope fractionation of CO2 from and Linda Booth for assistance with sample preparation; Jeremy magnesite–phosphoric acid reaction. Geochim. Cosmochim. Boyce, Andy Cohen, Simon Crowhurst, Allegra LeGrande, Gavin Acta 66, 589–593. Schmidt, and Laurence Yeung for discussion of this work and Delaney M., Be A. and Boyle E. (1995) Li, Sr, Mg, and Na in comments on a draft of this manuscript; Thibault de Garidel- foraminiferal calcite shells from laboratory culture, sediment Thoron for provision of published data from MD97-2138; and traps, and sediment cores. Geochim. Cosmochim. Acta 49, 1327– Jess Adkins, Daniel Cicerone, David Dettman, Harry Elderﬁeld, 1341. Jonathan Erez, Weifu Guo, Katherine Hungtington, Ted Dennis K. and Schrag D. (2010) Clumped isotope thermometry of McConnaughey, Nele Meckler, Jay Quade, Michael Reddy, carbonatites as an indicator of diagenetic alteration. Geochim. Edwin Schauble, Philippe Van Cappellen, and Bruce Watson for Cosmochim. Acta 74, 4110–4122. discussion of this work. We also thank three anonymous reviewers and the associate editor for their constructive comments on this manuscript. 13C–18O isotope signatures and ‘clumped isotope’ thermometry 5715

Dickson A. G., Sabine C. L. and Christian J. R. (2007) Guide to Guo W. and Eiler J. (2007) Temperatures of aqueous alteration

best practices for ocean CO2 measurements. PICES Special and evidence for methane generation on the parent bodies of Publication 3, 191p. the CM chondrites. Geochim. Cosmochim. Acta 71, 5565– Duffy D. M., Travaille A. M., van Kempen H. and Harding J. H. 5575. (2005) Effect of bicarbonate ions on the crystallization of calcite Guo W. (2008) Carbonate clumped isotope thermometry: applica- on self-assembled monolayers. J. Phys. Chem. B 109, 5713– tion to carbonaceous chondrites and effects of kinetic isotope 5718. fractionation. Ph.D. thesis, California Institute of Technology. Eagle R., Schauble E., Tripati A., Tutken T., Hulbert R. and Eiler Guo W., Daeron M., Niles P., Genty D., Kim S.-T., Vonhof H., J. (2010) Body temperatures of modern and extinct vertebrates Affek H., Wainer K., Blamart D. and Eiler J. (2008) 13C–18O from 13C–18O bond abundances in bioapatite. Proc. Natl. Acad. bonds in dissolved inorganic carbon: implications for carbonate Sci. USA 107, 10377–10382. clumped isotope thermometry. Geochim. Cosmochim. Acta 72, Eiler J. M. (2007) “Clumped-isotope” geochemistry—the study of A336. naturally-occurring multiply-substituted isotopologues. Earth Guo W., Mosenfelder J. L., Goddard W. A. and Eiler J. M. (2009) Planet. Sci. Lett. 262, 309–327. Isotope fractionations associated with phosphoric acid diges- Eiler J. and Schauble E. (2004) 18O13C16O in Earth’s atmosphere. tion of carbonate minerals: insights from first-principles theo- Geochim. Cosmochim. Acta 68, 4767–4777. retical modeling and clumped isotope measurements. Geochim. Elderfield H. and Ganssen G. (2000) Past temperature and d18O of Cosmochim. Acta 73, 7203–7225. surface ocean waters inferred from foraminiferal Mg/Ca ratios. Huntington K., Eiler J., Affek H., Guo W., Bonifacie M., Yeung Nature 405, 442–445. L., Thiagarajan N., Passey B., Tripati A., Daeron M. and Came Elderfield H., Greaves M., Barker S., Hall I., Tripati A., Ferretti P., R. (2009) Methods and limitations of ‘clumped’ CO2 isotope Booth L. and Daunt C. (2010) A record of bottom water (D47) analysis by gas-source isotope-ratio mass spectrometry. temperature and seawater d18O for the Southern Ocean over the J. Mass Spectrom. 44, 1318–1329. past 440 kyr based on Mg/Ca of benthic foraminiferal Huntington K. W., Wernicke B. P. and Eiler J. M. (2010) The Uvigerina spp.. Quatern. Sci. Rev. 29, 160–169. influence of climate change and uplift on Colorado Plateau Elderfield H., Yu J., Anand P., Kiefer T. and Nyland B. (2006) paleotemperatures from carbonate clumped-isotope thermom- Calibrations for benthic foraminiferal Mg/Ca paleothermometry. Tectonics. doi:10.1029/2009TC002449. etry and the carbonate ion hypothesis. Earth Planet. Sci. Lett. Keigwin, Jr., L. D. (1980) Palaeoceanographic change in the Pacific 250, 633–649. at the Eocene–Oligocene boundary. Nature 287, 722–725. Emiliani C. (1955) Pleistocene temperatures. J. Geol. 63, 538–577. Keigwin, Jr., L. D. (1982) Isotopic paleoceanography of the Epstein S., Buchsbaum R., Lowenstam H. and Urey H. (1951) Caribbean and East Pacific: role of Panama uplift in Late Carbonate–water paleotemperature scale. Bull. Geol. Soc. Am. Neogene Time. Science 217, 350–352. 62, 417–425. Kim S.-T. and O’Neil J. R. (1997) Equilibrium and non-equilib- Erez J. (2003) The source of ions for biomineralization in rium oxygen isotope effects in synthetic carbonates. Geochim. foraminifera and their implications for paleoceanographic Cosmochim. Acta 61, 3461–3475. proxies. Rev. Mineral. Geochem. 54, 115–149. Kim S.-T., Hillaire-Marcel C. and Mucci A. (2006) Mechanisms of Erez J., Bentov S., Grinstein M. and De Beer D. (2008) A novel equilibrium and kinetic oxygen isotope effects in synthetic carbon concentrating mechanism for foraminiferal calcification aragonite at 25 °C. Geochim. Cosmochim. Acta 70, 5790–5801. and its potential effects on paleoceanographic proxies. Geochim. Kim S.-T., O’Neil J. R., Hillaire-Marcel C. and Mucci A. (2007) Cosmochim. Acta 72, A245. Oxygen isotope fractionation between synthetic aragonite and Erez J. and Luz B. (1982) Temperature control of oxygen-isotope water: influence of temperature and Mg2+ concentration. fractionation of cultured planktonic foraminifera. Nature 297, Geochim. Cosmochim. Acta 71, 4704–4715. 220–222. Kısaku¨rek B., Eisenhauer A., Bo¨hm F., Garbe-Scho¨nberg D. and Feng J., Lee Y. J., Reeder R. J. and Phillips B. L. (2006) Erez J. (2008) Controls on shell Mg/Ca and Sr/Ca in cultured Observation of bicarbonate in calcite by NMR spectroscopy. planktonic foraminiferan, Globigerinoides ruber (white). Earth Am. Mineral. 91, 957–960. Planet. Sci. Lett. 273, 260–269. Ferguson J., Henderson G., Kucera M. and Rickaby R. (2008) Lea D. (2003) Elemental and isotopic proxies for marine temper- Systematic change of foraminiferal Mg/Ca ratios across a ature. In Oceans and Marine Geochemistry, vol. 6 (ed. H. strong salinity gradient. Earth Planet. Sci. Lett. 265, 153– Elderfield). Elsevier–Pergamon, Oxford, pp. 365–390. 166. Lea D. W., Mashiotta T. and Spero H. (1999) Controls on Gay C. V. and Mueller W. J. (1973) Cellular localization of magnesium and strontium uptake in planktonic foraminifera as carbonic-anhydrase in avian tissues by labeled inhibitor auto- determined by live culturing. Geochim. Cosmochim. Acta 63, radiography. J. Histochem. Cytochem. 21, 693–702. 2369–2379. Ghosh P., Adkins J., Affek H., Balta B., Guo W., Schauble E., Lea D. W., Pak D. and Spero H. J. (2000) Climate impact of late Schrag D. and Eiler J. (2006a) 13C–18O bonds in carbonate Quaternary Equatorial Pacific sea surface temperature varia- minerals: a new kind of paleothermometer. Geochim. Cosmo- tions. Science 289, 1719–1724. chim. Acta 70, 1439–1456. Lear C., Elderfield H. and Wilson P. (2000) Cenozoic deep-sea Ghosh P., Garzione C. and Eiler J. (2006b) Rapid uplift of the temperatures and global ice volumes from Mg/Ca in benthic Altiplano revealed through 13C–18O bonds in paleosol carbon- foraminiferal calcite. Science 287, 269–272. ates. Science 311, 511–515. LeGrande A. N. and Schmidt G. A. (2006) Global gridded data set Ghosh P., Eiler J., Campana S. and Feeney R. (2007) Calibration of the oxygen isotopic composition in seawater. Geophys. Res. of the carbonate ‘clumped isotope’ paleothermometer for Lett. 33. doi:10.1029/2006GL026011. otoliths. Geochim. Cosmochim. Acta 71, 2736–2744. Levitus S., Burgett R. and Boyer T. (1994) World Ocean Atlas Grossman E. (1984) Stable isotope fractionation in live benthic 1994. NOAA Atlas NEDIS: US Department of Commerce, foraminifera from the Southern California Borderland. Palae- Washington, DC. ogeogr. Palaeoclimatol. Palaeoecol. 47, 301–327. 5716 A.K. Tripati et al. / Geochimica et Cosmochimica Acta 74 (2010) 5697–5717

Lipp J. S., Morono Y., Inagaki F. and Hinrichs K.-U. (2008) and intertest Mg/Ca variability in Globigerinoides ruber. Pale- Significant contribution of Archaea to extant biomass in marine oceanography 23. doi:10.1029/2007PA001452. subsurface sediments. Nature 454, 991–994. Schauble E., Ghosh P. and Eiler J. (2006) Preferential formation of Lynch-Steglitz J., Curry W. and Slowey N. C. (1999) Weaker Gulf 13C–18O bonds in carbonate minerals, estimated using first- Stream in the Florida Straits during the Last Glacial Maximum. principles lattice dynamics. Geochim. Cosmochim. Acta 70, Nature 402, 644–648. 1439–1456. MARGO Project Members (2009) Constraints on the magnitude Schouten S., Hopans E. C., Schefu E. and Sinninghe Damste´ J. and patterns of ocean cooling at the Last Glacial Maximum. (2002) Distributional variations in marine crenarchaeotal Nat. Geosci. 2, 127–132. membrane lipids: a new tool for reconstructing ancient sea Mathien-Blard D. and Bassinot F. (2009) Salinity bias on the water temperatures? Earth Planet. Sci. Lett. 204, 265–274. foraminifera Mg/Ca thermometry: correction procedure and Shackleton N. J. (1974) Attainment of isotopic equilibrium implications for past ocean hydrographic reconstructions. between ocean water and the benthonic foraminifera genus Geochem. Geophys. Geosyst. 10. doi:10.1029/2008GC002353. Uvigerina: isotopic changes in the ocean during the last glacial. McConnaughey T. (1989) 13C and 18O isotopic disequilibrium in Colloques Int. C.N.R.S. 219, 203–209. biological carbonates: II. In vitro simulation of kinetic isotope Shackleton N. J. and Opdyke N. D. (1973) Oxygen isotope and effects. Geochim. Cosmochim. Acta 53, 163–171. palaeomagnetic stratigraphy of equatorial pacific core V28–238: McCrea J. M. (1950) On the isotopic geochemistry of carbon- oxygen isotope temperatures and ice volumes on a 105 year and ates and a paleotemperature scale. J. Chem. Phys. 18, 849– 106 year scale. Quatern. Res. 3, 39–55. 857. Sharma T. and Clayton R. N. (1965) Measurement of 18O/16O Mix A. C. (2006) Running hot and cold in the eastern equatorial ratios of total oxygen of carbonates. Geochim. Cosmochim. Acta Pacific. Quatern. Sci. Rev. 25, 1147–1149. 29, 1347–1353. Miyamoto H., Miyashita T., Okushima M., Nakano S., Morita T. Spero H. J., Bijma J., Lea D. W. and Bemis B. E. (1997) Effect of and Matsushiro A. (1996) A carbonic anhydrase form the seawater carbonate concentration on planktonic foraminiferal nacreous layer in oyster pearls. Proc. Natl. Acad. Sci. USA 93, carbon and oxygen isotopes. Nature 390, 497–500. 9657–9660. Swart P. K., Burns B. J. and Leder J. J. (1991) Fractionation of the Muller P., Kirst G., Ruhland G., von Storch I. and Rosell-Mele A. stable isotopes of oxygen and carbon in carbon dioxide during (1996) Calibration of the alkenone paleotemperature index the reaction of calcite with phosphoric acid as a function of U37K0 based on core-tops from the eastern South Atlantic and temperature and technique. Chem. Geol. (Isotope Geoscience the global ocean (60°N–60°S). Geochim. Cosmochim. Acta 62, Section) 86, 89–96. 1757–1772. Tripati A. K. and Elderfield H. (2004) Abrupt hydrographic Nuernberg D., Bijma J. and Hemleben C. (1996) Assessing the changes in the equatorial Pacific and subtropical Atlantic from reliability of magnesium in foraminiferal calcite as a proxy for foraminiferal Mg/Ca indicate greenhouse origin for the thermal water mass temperatures. Geochim. Cosmochim. Acta 60, 803– maximum at the Paleocene–Eocene boundary. Geochem. Geo- 814. phys. Geosyst. 5. doi:10.1029/2003GC000631. Quade J., Garzione C. and Eiler J. (2007) Paleoelevation recon- Tripati A. K. and Elderfield H. (2005) Deep-sea temperature and struction using pedogenic carbonates. Rev. Mineral. Geochem. circulation changes at the Paleocene–Eocene thermal maxi- 66, 53–87. mum. Science 308, 1894–1898. Risk M. J., Heikoop J. M., Snow M. G. and Beukens R. (2002) Tripati A. K. and Zachos J. C. (2002) Late eocene tropical sea Lifespans and growth patterns of two deep-sea corals: Primnoa surface temperatures: a perspective from Panama. Paleocea- resedaeformis and Demosphyllum cristagalli. Hydrobiology 471, nography 17, 1032. doi:10.1029/2000PA000605. 125–131. Tripati A. K., Backman J., Elderfield H. and Ferretti P. (2005) Rollion-Bard C., Chaussidon M. and France-Lanord C. (2003) pH Eocene bipolar glaciation associated with global carbon cycle control on oxygen isotopic composition of symbiotic corals. changes. Nature 436, 341–346. Earth Planet. Sci. Lett. 215, 275–288. Tripati A. K., Delaney M. L., Zachos J. C., Anderson L. D., Kelly Romanek C. S., Grossman E. L. and Morse J. W. (1992) Carbon D. C. and Elderfield H. (2003) Tropical sea-surface temperature isotope fractionation in synthetic aragonite and calcite: effects reconstruction for the early Paleogene using Mg/Ca ratios of of temperature and precipitation rate. Geochim. Cosmochim. planktonic foraminifera. Paleoceanography 18, 1101. Acta 56, 419–430. doi:10.1029/2003PA000937. Rosenthal Y., Boyle E. and Slowey N. C. (1997) Temperature Tripati A., Elderfield H., Booth L., Zachos J. and Ferretti P. (2006) control on the incorporation of magnesium, strontium, fluorine, High-resolution benthic foraminiferal stable isotope stratigra- and cadmium into benthic foraminiferal shells from Little phy across the Oligocene–Miocene boundary at Site 1218, ODP Bahama Bank: prospects for thermocline paleoceanography. Leg 199 Scientific Results 199, 1–13. Geochim. Cosmochim. Acta 61, 3633–3643. Tripati A. K., Roberts C. D. and Eagle R. A. (2009) Coupling of

Sabine C. L., Key R. M., Kozyr A., Feely R. A., Wanninkhof R., CO2 and ice sheet stability over major climate transitions of the Millero F. J., Peng T. -H., Bullister J. L. and Lee K. (2005) last 20 million years. Science 326, 1394–1397. Global Ocean Data Analysis Project: Results and Data. ORNL/ Tripati A. K., Zachos J. C., Marincovich, Jr., L. and Bice K. (2001) CDIAC-145, NDP-083. Carbon Dioxide Information Analysis Late Paleocene Arctic coastal climate as inferred from mollus- Center, Oak Ridge National Laboratory, US Department of can stable and radiogenic isotope ratios. Palaeogeogr. Palaeo- Energy, Oak Ridge, Tennessee, 110p. doi:10.3334/CDIAC/ climatol. Palaeoecol. 170, 101–113. otg.ndp083. Turich C., Freeman K. H., Bruns M. A., Conte M., Jones A. D. Sachs J., Schneider R., Eglinton T., Freeman K. H., Ganssen G., and Wakeham S. G. (2007) Lipids of marine Archaea: patterns McManus J. and Oppo D. (2000) Alkenones as paleoceano- and provenance in the water-column and sediments. Geochim. graphic proxies. Geochem. Geophys. Geosyst. 1, 2000GC00059. Cosmochim. Acta 71, 3272–3291. Sadekov A., Eggins S., De Deckker P. and Kroon D. (2008) Turner J. V. (1982) Kinetic fractionation of C-13 during calcium Uncertainties in seawater thermometry deriving from intratest carbonate precipitation. Geochim. Cosmochim. Acta 46, 1183– 1191. 13C–18O isotope signatures and ‘clumped isotope’ thermometry 5717

Usdowski E. and Hoefs J. (1993) Oxygen isotope exchange between Zeebe R. E. (1999) An explanation of the eﬀect of seawater carbonic acid, bicarbonate, carbonate, and water: a re-exam- carbonate concentration on foraminiferal oxygen isotopes. ination of the data of McCrea (1950) and an expression for the Geochim. Cosmochim. Acta 63, 2001–2007. overall partitioning of oxygen isotopes between the carbonate Zeebe R. E. (2007) An expression for the overall oxygen isotope species and water. Geochim. Cosmochem. Acta 57, 3815–3818. fractionation between the sum of dissolved inorganic carbon Van Cappellen P., Charlet L., Stumm W. and Wersin P. (1993) A and water. Geochem. Geophys. Geosyst. doi:10.1029/ surface complexation model of the carbonate mineral–aqueous 2003GC001663. solution interface. Geochim. Cosmochim. Acta 57, 3505–3518. Zeebe R. E. (2009) Hydration in solution is critical for stable Volkman J. K. (2000) Ecological and environmental factors oxygen isotope fractionation between carbonate ion and water. aﬀecting alkenone distributions in seawater and sediments. Geochim. Cosmochim. Acta 73, 5283–5291.

Geochem. Geophys. Geosyst., 2000GC000061. Zeebe R. E. and Wolf-Gladrow D. (2001) CO2 in Seawater: Watson E. B. (2004) A conceptual model for near-surface kinetic Equilibrium, Kinetics, Isotopes. Elsevier Oceanography Series, controls on the trace-element and stable isotope composition of vol. 65, 346p. abiogenic calcite crystals. Geochim. Cosmochim. Acta 68, 1473– Ziveri P., Stoll H., Probert I., Klaas C., Geisen M., Ganssen G. and 1488. Young J. (2003) Stable isotope ‘vital eﬀects’ in coccolith calcite. Wilson J. B. (1979) Patch development of the deep-water coral Earth Planet. Sci. Lett. 210, 137–149. Lophelia pertusa on Rockall Bank. J. Mar. Biol. Assoc. UK 59, 165–177. Associate editor: Clark Johnson York D. (1969) Least-squares ﬁtting of a straight line with correlated errors. Earth Planet. Sci. Lett. 5, 320–324.