Carbonate “clumped” isotope signatures in aragonitic scleractinian and calcitic gorgonian deep-sea corals

. Deep-sea corals are a potentially valuable archive of the temperature and ocean chemistry of intermediate and deep waters. Living in near-constant temperature, salinity, and pH and having amongst the slowest calciﬁcation rates observed in carbonate-precipitating biological organisms, deep-sea corals can provide valuable constraints on processes driving mineral equilibrium and disequilibrium isotope signatures. Here we report new data to further develop “clumped” isotopes as a paleothermometer in deep-sea corals as well as to investigate mineral-speciﬁc, taxon-speciﬁc, and growth-rate-related effects. Carbonate clumped isotope thermometry is based on measurements of the abundance of the doubly substituted isotopologue 13 C 18 O 16 O 2 in carbonate minerals, analyzed in CO 2 gas liberated on phosphoric acid digestion of carbonates and reported as (cid:49) 47 values. We analyzed (cid:49) 47 in

47 values were compared to in situ temperatures, and the relationship between 47 and temperature was determined for each group to investigate taxon-specific effects.We find that aragonitic scleractinian deep-sea corals exhibit higher values than high-Mg calcitic gorgonian corals and the two groups of coral produce statistically different relationships between 47 -temperature calibrations.These data are significant in the interpretation of all carbonate clumped isotope calibration data as they show that distinct 47 -temperature calibrations can be observed in different materials recovered from the same environment and analyzed using the same instrumentation, phosphoric acid composition, digestion temperature and technique, CO 2 gas purification apparatus, and data handling.
There are three possible explanations for the origin of these different calibrations.The offset between the corals of different mineralogy is in the same direction as published theoretical predictions for the offset between calcite and aragonite although the magnitude of the offset is different.One possibility is that the deep-sea coral results reflect high-Mg and aragonite crystals attaining nominal mineral equilibrium clumped isotope signatures due to conditions of extremely slow growth.In that case, a possible explanation for the attainment of disequilibrium bulk isotope signatures and equilibrium clumped isotope signatures by deep-sea corals is that extraordinarily slow growth rates can promote the occurrence of isotopic reordering in the interfacial region of growing crystals.We also cannot rule out a component of a biological "vital effect" influencing clumped isotope signatures in one or both orders of coral.Based on published experimental data and theoretical calculations, these biological vital effects could arise from kinetic isotope effects due to the source of carbon used for calcification, temperature-and pH-dependent rates of CO 2 hydration and/or hydroxylation, calcifying fluid pH, the activity of carbonic anhydrase, the residence time of dissolved inorganic carbon in the calcifying fluid, and calcification rate.A third possible explanation is the occurrence of variable acid digestion fractionation factors.Although a recent study has suggested that dolomite, 1 Introduction "Clumped"-isotope paleothermometry is an approach to determining carbonate mineral formation temperatures (Ghosh et al., 2006;Schauble et al., 2006;Eiler, 2007); like the oxygen isotope thermometer before it, it is founded on thermodynamic predictions of the distribution of isotopes.Crucially, instead of relying on an isotopic exchange reaction between different phases (e.g., CaCO 3 and H 2 O), clumped isotope thermometry relies on internal isotopic exchange between isotopes in a single phase (Schauble et al., 2006).This means that, in theory, all that is needed to determine mineral formation temperatures is the clumped isotope composition of the solid and not the water from which it grew (Schauble et al., 2006;Eiler, 2007).When considering carbonate minerals, statistical thermodynamics predicts that heavy stable isotopes of carbon ( 13 C) and oxygen ( 18 O) will increasingly bond or "clump" in a mineral as temperature decreases and will conform to isotopic equilibrium constants for reactions such as where X refers to cations such as Ca 2+ , Mg 2+ , Sr 2+ , and Ba 2+ .In practice, the determination of 13 C-18 O bonds in carbonate minerals is accomplished by measurement of mass 47 CO 2 (predominantly 13 C 18 O 16 O) liberated by phosphoric acid digestion (Ghosh et al., 2006). 13C-18 O bonding is reported as a per mil enrichment from that which would be expected in the liberated CO 2 if the sample had a stochastic distribution of C and O isotopes among all isotopologues and is designated by the parameter 47 .Following the initial calibration of synthetic calcites and some coral taxa (Ghosh et al., 2006), several studies have focused on in-depth calibrations of biogenic carbonates that represent potential paleoclimate proxies.Proxy material calibrations thus far have included aragonitic scleractinian zooxanthellate corals (Ghosh et al., 2006;Saenger et al., 2012;Tripati et al., 2015), aragonitic scleractinian non-zooxanthellate deep-sea corals (Ghosh et al., 2006;Thiagarajan et al., 2011), aragonitic otoliths (Ghosh et al., 2007), calcitic and aragonitic foraminifera (Tripati et al., 2010), mollusks and brachiopods (Came et al., 2007(Came et al., , 2014;;Eagle et al., 2013;Henkes et al., 2013) and land snails (Zaarur et al., 2011;Eagle et al., 2013), calcitic speleothems (Affek et al., 2008;Daëron et al., 2011), bioapatite (Eagle et al., 2010), and calcitic microbialites (Petryshyn et al., 2015).
It has been noted that calibration data on different biogenic carbonates generated in the same laboratory at Caltech and using similar analytical methods produce a relationship between temperature and 47 values that was similar to the initial inorganic calcite calibration (Ghosh et al., 2006;Eagle et al., 2013).However, there are differences in inorganic calcite calibrations produced in different laboratories (Ghosh et al., 2006;Dennis and Schrag et al., 2010;Zaarur et al., 2013;Tang et al., 2014;Wacker et al., 2013;Kluge et al., 2015) that are thought to either reflect methodological differences in clumped isotope measurements or differences in how synthetic carbonates were precipitated (Tripati et al., 2015).It is clear from studies in mollusks (Henkes et al., 2013;Eagle et al., 2013;Douglas et al., 2014;Petrizzo et al., 2014) and brachiopods (Henkes et al., 2013;Came et al., 2014) that calibrations of the same types of materials in different laboratories can yield different relationships between temperature and 47 values.
More recent studies have discussed some possible sources of methodological effects on clumped isotope calibrations relating to data handling and mass spectrometric effects (Wacker et al., 2013;Petrizzo et al., 2014;Defliese et al., 2015).It is also possible that phosphoric acid digestion technique and temperature may be playing a role (Wacker et al., 2013;Defliese et al., 2015), as is known to be the case for conventional δ 18 O measurements (Swart et al., 1991).Nonetheless, carbonate standards and a deep-sea scleractinian coral were found to yield broadly comparable results in an inter-lab comparison study between four different laboratories (Dennis et al., 2011).
To date, there are relatively little published data attempting to resolve whether different biogenic calibrations may originate from methodological differences between laboratories by simply measuring multiple types of biogenic carbonates interspersed with each other using the same methods.Therefore, it remains difficult at this time to distinguish between methodological differences and genuine differences in clumped isotope compositions of different materials, for example biologically mediated fractionations or "vital effects".The little work that has been done includes Porites corals that appear to exhibit a kinetic isotope effect that drives their 47 values significantly out of equilibrium (Ghosh et al., 2006;Saenger et al., 2012) and a temperate coral Oculina arbuscula cultured at the same temperature and variable pH that exhibits large kinetic effects (Tripati et al., 2015).
Here we explore 13 C-18 O bond abundances in deep-sea coral species from the orders Scleractinia and Gorgonacea and precipitate aragonite and high-Mg calcite, respectively.As these corals were recovered from the same environment, analyzed by the same methods and mass spectrometer during the same analytical period, and subject to the same data processing, they represent an opportunity to explore the potential role of mineralogy and biology in governing clumped isotope signatures.We discuss results in the context of recent studies including of acid digestion fractionation (Defliese et al., 2015) and theoretical models and experimental data that have been used to constrain differences in clumped isotope com-position between different dissolved inorganic carbon (DIC) species (Hill et al., 2014;Tripati et al., 2015), as it has been suggested that in cases where a carbonate mineral may inherit an isotopic signature of DIC, then factors effecting DIC speciation such as pH, salinity, and temperature may also influence carbonate clumped isotope signatures (Hill et al., 2014;Tripati et al., 2015).

Deep-sea corals in paleoceanography
Deep-sea corals represent a potentially valuable archive of intermediate and deep-ocean temperatures and have been a target for models of stable isotope fractionation (Adkins et al., 2003;Gaetani et al., 2011).These archives could give valuable insight into the natural variability of regions of the ocean that play an active role in large-scale climate dynamics.With fossil deep-sea coral recorded for at least the last 225 000 years (Robinson et al., 2007;Thiagarajan et al., 2013), it is a proxy with the potential to extend our observations of ocean physics and climate into the Pleistocene with decadal to centennial resolution.It has been established that deep-sea corals have significant skeletal vital effects (disequilibrium stable isotope fractionations) that compromise the classic δ 18 O paleotemperature method (Emiliani et al., 1978;McConnaughey, 1989aMcConnaughey, , 2003;;Adkins et al., 2003;Smith et al., 2000;Rollion-Bard et al., 2003, 2010;Lutringer et al., 2005;Hill et al., 2011;Kimball et al., 2014).Attempts to circumvent these kinetic effects have focused on the "lines" method that recovers an average temperature over the lifetime of a coral (Smith et al., 2000;Lutringer et al., 2005;Hill et al., 2011;Kimball et al., 2014) and a Rayleigh-based multielement (e.g., Mg / Ca, Sr / Ca, Ba / Ca) paleotemperature method (Gaetani et al., 2011).
Despite deep-sea coral species showing significant disequilibrium in δ 13 C and δ 18 O values, sometimes termed vital effects, initial calibration work using clumped isotopes in scleractinian corals (Thiagarajan et al., 2011) revealed a good agreement between aragonitic scleractinian deep-sea coral 47 and the inorganic calcite calibration of Ghosh et al. (2006).Therefore, at least some species that show significant vital effects on δ 13 C and δ 18 O yield apparent equilibrium 47 values, an observation also made on foraminifera and coccoliths (Tripati et al., 2010).Recently scleractinian deep-sea coral 47 measurements have been used in applied paleoceanographic reconstructions (Thiagarajan et al., 2014).

Samples
Thirteen live-collected specimens of deep-sea coral were examined.One specimen, PV 703-7 was mostly dead with patches of living tissue.Specimens were collected by deepsea submersible diving on Warwick Seamount, Gulf of Temperature data from Warwick Seamount is averaged from two Seabird conductivity-temperature-depth (CTD) casts taken at the dive location during the dive, while in situ temperature was measured on DSRV Pisces V and IV at the time of collection for Hawaiian and Line Island corals.Temperature ranges from Hawaiian Island coral were calculated from Hawaiian Ocean Time Series (HOTS) (data collected at Station ALOHA during the years 1990-2012.Although Station ALOHA is located about 328 miles east of our samples, a comparison between the HOTS vertical temperature profile matches closely with CTD profiles taken from the sample locations.Additionally, Station ALOHA was originally chosen as a monitoring station because it is believed to be representative of the North Pacific subtropical gyre and we take it to represent conditions experienced by Hawaiian corals.Temperature ranges from Warwick Seamount are estimated from CTD casts taken at approximately 48.02 • N, −130.66 • W during 11 cruises during 1972-1998 at different times of the year.Temperature data was extracted from National Oceanic Data Center at 634 ± 5 m, 704 ± 5 m, 720 ± 5 m, and 872 ± 5 m to assess temperature variability at Table 1.Thirteen scleractinian and gorgonian deep-sea corals were collected live by submersible diving, with depth and temperature measured at the time of collection.Temperature ranges are mean with standard deviation and come from Hawaii Ocean Time Series (Karl and Lukas, 1996) and cruises extracted from the National Ocean Database (Boyer et al., 2013)

Sample preparation
Disks were cut from near the base for all coral skeletons, except PV 703-7, which was dead at the base.For this specimen, a disk was cut from the living branch for sampling.Disks were cleaned by the simple method of sonication in nanopure water and air-drying at room temperature (∼ 25 • C).All specimens lacked visual organic contamination of any kind, and following the findings of Thiagarajan et al. (2011) and Eagle et al. (2013), who found cleaning steps to be unnecessary in deep-sea corals and mollusks, respectively, no further cleaning was performed.For most corals, sample powders of ca.50-70 mg were milled using a Merchantek micromill from the outer portions of disks which represent the most recently accreted part of the skeleton (Fig. 2).In one sample, PV-703-5, a sample was milled from both the center and outer edge of the disk.In a few cases (PIV 148-2, PV 703-2 PIV 146-6, PV 694-3, PV 694-13) samples were first milled from the outer edge and later, to obtain additional sample, whole disks were ground into powder using a mortar and pestle.Although milling was preferred, in some cases, disks were small enough that the entire disk was necessary to yield the required weight of sample for replica 47 measurements.
In order to explore intra-coral heterogeneity and sampling effects, specimen PIV 146-6 is sampled both along the outer edge by micromilling as well as by grinding and homogeniz- ing an entire disk, while PV 703-5 was micromilled in the center and outer edge (Fig. 2).In this way, intra-specimen reproducibility and the effects of sampling were assessed.

Notation and reference frame
The parameter 47 is a measure of the enrichment in per mil of 13  Standard gases of different bulk compositions were equilibrated at different temperatures (25 or 1000 • C) and measured each day.Values of 47 are reported in the absolute reference frame (ARF) and calculated as described in another publication using equilibrated gases (Dennis et al., 2011).Published data are presented in the absolute reference frame with conversion of data from Thiagarajan et al. (2011) and Ghosh et al. (2006) to the ARF, as reported in Eagle et al. (2013).

Analytical measurement of 47
Isotopic measurements were conducted on a dual inlet Thermo Scientific MAT 253 mass spectrometer coupled to a custom-built semiautomated sample digestion and purification system located at the University of California, Los Angeles.The setup is modeled on that used at Caltech and Johns Hopkins, which are described in Passey et al. (2010) and Henkes et al. (2013).Samples were run during two separate time periods (winter: January 2013-March 2013; summer: May 2013-July 2013) and samples of scleractinian (aragonite) and gorgonian (high-Mg calcite) were run interchangeably during both time periods.For each analysis (previously referred to as extraction) 8-10 mg of material was used to yield sufficient gas to maintain a steady signal at extended counting times.Also, recent results suggest that the digestion of samples that are significantly smaller might undergo secondary equilibration with water, resulting in elevated 47 values compared to > 7 mg size samples (Wacker et al., 2013).
Mass spectrometric configuration was set to measure ion beams corresponding to m/z = 44, 45, 46 (amplified by 3 × 10 8 to 1 × 10 11 resistors) and 47, 48, 49 (amplified by 10 12 resistors) with the ion bean m/z = 44 fixed at ∼ 16 V.Each analysis contains eight measurements of seven cycles between sample and reference gas with 26 s of integration per cycle.The total integration time of 1456 s is sufficiently long that errors should be able to approach shot noise error predictions (Thiagarajan et al., 2011;Huntington et al., 2009).Internal precision of 47 was ca.0.005-0.012‰, 1σ (based on eight measurements of seven cycles of sample and reference gas comparison within a single analysis, (Table 2, Supplement Table S1) and external precision was ca.0.004-0.008‰, 1σ (determined from repeat analyses of standards, Table 3).To increase external precision, samples were measured three to seven times each, which resulted in standard errors for 47 of 0.002-0.010‰ for the corals in this study (Table 2).
The semiautomated sample digestion system allows for a sample to be introduced as either a carbonate powder or a gas.Carbonate powders are introduced via a Costech autosampler which drops samples in vacuo into a common ∼ 103% phosphoric acid (ρ ≈ 1.91 mg mL −1 ) bath held at 90 • C and allowed to react for 20 min.Evolved CO 2 from acid digestion is then passed through a cooled ethanol trap (−78 • C) and collected in a liquid nitrogen trap (−200 • C).CO 2 is liberated by warming with the same ethanol trap (−78 • C) and, with a helium carrier gas, passes through a Porapak Q 120/80 mesh GC (gas chromatograph) column at −20 • C and silver wool, which removes organic contaminants and scavenges sulfur compounds, respectively.After GC passage, CO 2 is again collected in liquid nitrogen and undergoes one final cryogenic purification step in vacuo before introduction to the mass spectrometer.Gaseous CO 2 can also be prepared on a vacuum line and introduced using quartz tubes.A tube cracker which leads into the second cooled ethanol trap prior to the GC step allows CO 2 to pass through the autoline in the same way as carbonate samples, minus the common acid bath.
In addition to carbonate powders, on most days, at least one gaseous "equilibrated CO 2 " sample is analyzed for use in defining the absolute reference frame for 47 measurements.Producing equilibrated CO 2 gases with varying bulk isotopic (δ 13 C and δ 18 O) composition is accomplished by utilizing a very depleted CO 2 (δ 13 C = −25 ‰, δ 18 O = 3.6 ‰ Vienna Standard Mean Ocean Water (VSMOW)) and equilibrating Oztech CO 2 (δ 13 C = −3.6 ‰ Vienna Pee Dee Belemnite (VPDB), δ 18 O = 23.6 ‰ VSMOW) with isotopically "heavy" water ( 18 O enriched and produced by boiling house DI).These two isotopical end member gases are then either heated at 1000 • C for 2 h to produce a nearly stochastic distribution of isotopes among isotopologues or equilibrated in a water bath held at 25 • C. Equilibrated CO 2 is then cryogenically purified on a vacuum line and captured into quartz tubes.This procedure produces four isotopically unique gaseous CO 2 samples which define the empirical transfer function (ETF) used to covert 47 values to the ARF scale (Dennis et al., 2011).The carbonate standards Carrara marble, Carmel chalk, TV01, and 102-GC-AZ01 were run in concert with samples and were shown to have values that were indistinguishable from those determined in the Caltech lab (Table 3).
An acid fractionation factor of 0.092 ‰ was applied to all data and to theoretical predictions to normalize to previously reported data at 25 • C (Henkes et al., 2013).In tables and figures, we also report results calculated using a value of 0.082 ‰ (Passey et al., 2010;Defliese et al., 2015) as well as other acid digestion fractionation factors (e.g., Wacker et al., 2013).

Scleractinian coral 47
When compared with previously reported deep-sea coral data the aragonitic scleractinian coral results from this study agree very closely with those measured in Thiagarajan et www.biogeosciences.net/13/6487/2016/Biogeosciences, 13, 6487-6505, 2016  and 4).Measured and predicted coral isotopic measurements are presented in Table 4. Since specimen PV 703-7 was mostly a dead coral except in small sections when collected and therefore could have had a more complicated life history, it was excluded from calibration analysis.For the purpose of calibration, we focus on data from live-collected specimens to compare with in situ temperature measurements to eliminate any uncertainty in growth temperatures or post-formation dissolution when considering proxy systematics.When the scleractinian corals are combined with the data from Thiagarajan et al. ( 2011), a similar 47 -T relationship to that originally reported is found.In a plot of 47 vs. 10 6 /T 2 the linear regression of the 11 corals from Thiagarajan et al. (2011) (slope: 0.0643 ± 0.008; intercept: −0.029 ± 0.103; R 2 = 0.934) is almost identical to that of the combined data set of 14 scleractinian corals (slope: 0.0582 ± 0.008; intercept: 0.0452 ± 0.103; R 2 = 0.81).As noted in Thiagarajan et al. (2011), this is closely similar to that of the Ghosh et al. (2006) calibration of inorganic calcite (slope: 0.0620; intercept: −0.0021).

Gorgonian coral 47
Gorgonian deep-sea corals precipitate skeletal carbonate in the form of high-Mg calcite, with 5-10 mol % MgCO 3 (Noé and Dullo, 2006;Kimball et al., 2014).Compared to scle- ractinian deep-sea coral and all other previously reported biogenic carbonates, 47 is depleted in gorgonian deep-sea corals.In a plot of 47 vs. 10 6 /T 2 , the linear regression through data derived from nine gorgonian deep-sea corals analyzed in this study gives a significantly shallower slope (slope: 0.025 ± 0.01; intercept: 0.403 ± 0.129, R 2 = 0.48) and 47 offset of ca.0.04-0.07‰ in the temperature range of 3.2-11.2• C compared to scleractinian corals as well as compared to the Ghosh et al. (2006) calibration.It is, however, more similar to the inorganic calibration of Dennis and Schrag (2010) (Fig. 3).Temperature sensitivity of the gor- gonian coral calibration was not examined due to poorness of fit (R 2 = 0.48).

Intra-specimen sampling
Intra-specimen measurements from sampling different portions of the skeleton gives insight into the 47 heterogeneity present in an individual coral.Specimen PIV 146-6 was sampled both along the outer edge by micromilling as well as by grinding and homogenizing an entire disk, while PV 703-5 was micromilled in the center and outer edge.In both corals, aliquots produced from a given sampling method are significantly different in bulk isotopes.PIV 146-6 has differing bulk isotopic compositions between the two aliquots of more than 1 ‰ in δ 13 C and 0.4 ‰ in δ 18 O.PV 703-5 has almost a 3 ‰ difference in δ 13 C and 0.5 ‰ in δ 18 O (Table 5).47 of the aliquots are approximately 0.03 and 0.02 different from each other in PIV 146-6 and PV 703-5, respectively.For PIV 146-6, the average of all seven measurements is 0.791 ± 0.008 and for the two aliquots the averages are 0.779 ± 0.007 and 0.803 ± 0.01.For PV 703-5 the average of all six measurements is 0.744 ± 0.007, and the averages for the two aliquots are 0.734 ± 0.005 and 0.754 ± 0.010.

Trends in deep-sea coral data
The scleractinian corals analyzed in this study are solely Enallopsammia sp.In contrast, the deep-sea corals presented in Ghosh et al. (2006) and Thiagarajan et al. (2011) Thiagarajan et al. (2011).In contrast, the gorgonian corals show significantly lower 47 values compared to the scleractinian corals for the same temperature range.An analysis of covariance (ANCOVA) between the scleractinian and gorgonian coral data gives a P value < 0.01, which indicates that the differences between the two groups are statistically significant.This 0.04-0.07‰ offset between scleractinian and gorgonian corals is observed in non-acid-digestion-corrected 47 values.Possible explanations for this offset are discussed below.

Comparison with theoretical predictions
Theoretical modeling of 63 for different carbonate minerals (Schauble et al., 2006;Hill et al., 2014;Tripati et al., 2015) has been combined with theoretical acid digestion fractionations to give predicted 47 values for a variety of carbonate minerals.These predictions generally agree with measured inorganic and biogenic studies; however, they are more depleted in 47 for a given temperature.Further, offsets due to mineralogy, while predicted in these theoretical modeling studies, have not been resolved in measured samples.Compared to deep-sea coral 47 , theoretical predictions are more www.biogeosciences.net/13/6487/2016/Biogeosciences, 13, 6487-6505, 2016   Kim and O'Neil (1997) for calcite and Kim et al. (2007) for aragonite.c δ 47 is relative to intra-laboratory working gas with a known isotopic composition.d 47 predicted value is calculated from the inorganic calibration of Ghosh et al. (2006); similar values are predicted from other inorganic calibrations including Zaarur et al. (2013) and Tripati et al. (2015).depleted for aragonite; however, the calcite prediction agrees well with the calcitic gorgonians from this study (Fig. 5).The offset of 0.03 ‰ between aragonite and calcite is at the lower end of what we observe.A recent theoretical modeling effort (Hill et al., 2014) has produced similar predicted 47 values for aragonite and calcite although with more depleted (less agreement) values between predicted and observed aragonite 47 and a smaller offset between aragonite and calcite.

Acid digestion fractionation factor uncertainties
It is possible that uncertainties in the acid digestion fractionation factor for carbonates digested at 25 and 90 • C may account for some of the above offsets.Three values have been proposed for the acid digestion fractionation factor (AFF) for reporting data at 90 • C and converting them to 25 • C on the absolute reference frame.To explore the sensitivity of results to the choice of acid digestion fractionation factor, Figs.6-9 show calculations using a range of acid digestion fractionation factors of isotope ratios and apparent deviations of measured isotopic ratios from isotopic equilibrium.Passey et al. (2010) calculate and use a value of 0.081 ‰, whereas Henkes et al. (2013) suggest a value of 0.092 ‰ ± 0.012 ‰, which is within error of the Passey et al. (2010) value.Defliese et al. (2015) propose a value of 0.082 ‰ ± 0.014 ‰, which is within error of both values,.In a recent study Wacker et al. (2013) suggest there may be mineral-specific AFF, which is also predicted in the theoretical study by Guo et al. (2009).Application of the Wacker et al. (2013) 25−90 correction to the aragonite and calcite Table 5.Two coral specimens were analyzed using different portions of their skeletons.PIV 146-6 was sampled by micromilling the outer portion of a disk (outer) as well as grinding an entire disk (whole).PV 703-3 was micromilled in both cases but was milled in both the center and outer portion of a disk.All data and averages are reported as well as standard error, 1σ .experimental (Wacker et al., 2013;Defliese et al., 2015) and theoretical (Guo et al., 2009) studies suggest that acid digestion differences due to mineralogy may not be large enough to explain the ca.0.04-0.07‰ offset that we observe.Deep-sea coral 47 was corrected using 25−90 AFF of 0.082 ‰ (Defliese et al., 2015) -identical to the value from Passey et al., 2010-0.092 ‰ (Henkes et al., 2013) and 0.066 ‰ (Wacker et al., 2013, aragonite), and 0.075 ‰ (Wacker et al., 2013, calcite).Figure 6 shows the result of the AFF sensitivity analysis and indicates that although the absolute values of the 47 change with the application of AFF values, the offset between aragonite and calcite is unaffected except in the case of the application of mineral-specific AFF from Wacker et al. (2013).The application of Wacker et al. (2013)  in the offset between the aragonite and calcite deep-sea coral 47 of only 0.009 ‰, therefore not altering our conclusions.However, we acknowledge that the kinetics of acid digestion and the effects on clumped isotope signatures on 47 values is relatively poorly constrained.It is possible there are variable acid digestion fractionation factors that are influencing these observations, possibly due to cation substitution into carbonates (Guo et al., 2009), mineral bulk isotope composition (Guo et al., 2009), the presence of organic matter, as well as other factors.Until systematic studies of acid digestion are undertaken to investigate each of these factors, the contribution of acid digestion fractionation to the observed differences between aragonitic scleractinian and high-Mg calcitic gorgonian corals cannot be accurately quantified.

Analytical procedure, data comparison
Since its inception, the analytical measurement of 47 has improved in precision and accuracy due to advances in mass spectrometric analysis (e.g., development of automated sample digestions and purification systems, abbreviated to autoline, and higher counting times) and standardization of reporting procedures (Huntington et al., 2009;Dennis et al., 2011).Because the methods involved in producing and reporting 47 have changed over time, the comparison of data between laboratories and even within laboratories can be an issue.For relevant comparisons between previously reported data and this study's data, we refer to the recalculations of 47 performed in Eagle et al. (2013) using standard values to project onto the absolute reference frame.Still, we cannot exclude the possibility that analytical artifacts remain.For instance, Zaarur et al. (2013) noted that the shorter counting times and fewer replications of the trailblazing work of Ghosh et al. (2006) may contribute to any discrepancies between data sets.We assess interlaboratory comparability from standards, which indicate accuracy and precision comparable to what has been reported for other laboratories (Table 2; Dennis et al., 2011).Furthermore, we note that good agreement between the aragonitic scleractinian corals of this study and that of Thiagarajan et al. (2011) reinforces the comparability of our data to the wider body of measurements reported from other laboratories.
Interlaboratory comparability should not be a critical issue for the interpretation of data presented herein since the offset between aragonitic and calcitic corals was observed in data generated within our lab, while running samples interchangeably using the same analytical procedure.Standards were also run concurrently during these time periods (Table 3) and found to be in good agreement with accepted values.We therefore conclude it is unlikely that the offset is due to internal analytical variability.

Mixing and intra-specimen 47 heterogeneity
It has been shown that both scleractinian and gorgonian deepsea corals exhibit significant disequilibrium in δ 13 C and δ 18 O of several per mil within an individual coral (eg.Smith et al., 2000;Adkins et al., 2003;Rollion-Bard et al., 2003, 2010;Lutringer et al., 2005;Kimball et al., 2014).Because of the subtle saddle-shaped curvature in 47 relative to δ 13 C and δ 18 O (Eiler and Schauble, 2004;Thiagarajan et al., 2011;Defliese and Lohmann, 2015), 47 is sensitive to the mixing of end member δ 13 C and δ 18 O values and will result in non-conservative mixing (Fig. 10).Deep-sea corals, with their several per mil range in δ 13 C and δ 18 O values will be particularly susceptible to this effect.In fact, both banding and spatial variability in δ 13 C and δ 18 O values within deepsea coral skeletons vary on the scale of ten to hundreds of microns (Smith et al., 2000;Adkins et al., 2003;Rollion-Bard et al., 2003, 2010;Lutringer et al., 2005;Kimball et al., 2014).Any small-scale spatial variability makes the possibility of sampling within a skeletal band or an area with homogenous δ 13 C and δ 18 O values for 47 measurements unfeasible, and therefore there is the potential for mixing effects in 47 which we address here.
Both Thiagarajan et al. (2011) and Henkes et al. ( 2013) have addressed mixing in biogenic carbonates by calculating the 47 effect from mixing samples of differing bulk isotope composition.The end members from the Thiagarajan et al. ( 2011) calculations are similar to the most extreme values that we observe (δ 13 C = 2 ‰, δ 18 O = 5 ‰ and δ 13 C = −10 ‰, δ 18 O = −2 ‰).Using these end member values, Thiagarajan et al. (2011) calculated a positive 47 increase as large as 0.02 ‰ for the resulting mixture if 50 % of each end member is used.They note that while this mixing effect is not consistent with their data, mixing and then re-equilibration could potentially result in the trend seen in their data.Since sampling an isotopically homogenous area of a deep-sea coral is unfeasible (especially given the sample size requirements for clumped isotopes), we estimate the maximum artifacts in our data arising from mixing.The estimate for mixing artifacts from Thiagarajan et al. (2011) is based on observed variability in stable isotope values in scleractinian deep-sea coral, and represents an extreme example for this study, as they use δ 13 C and δ 18 O values of end members that differ by 12 and 7 ‰, respectively.Such differences in scleractinian coral represent systematic isotopic depletions in centers of calcification compared to fibers (Adkins et al., 2003), as well as variability associated with density bands;  Kim and O'Neil (1997) and Kim et al. (2007) for calcite and aragonite, respectively.2006) paired with theoretical acid digestion fractionations from Guo et al. (2009).δ 18 O residuals are calculated from average measured δ 18 O values relative to the calibrations of Kim and O'Neil (1997) and Kim et al. (2007) for calcite and aragonite, respectively.no centers of calcification have been reported in gorgonian corals.The calculations from Thiagarajan et al. (2011) represent an estimate of the maximum artifact to arise from mixing for either taxon given the range of carbon and oxygen isotope values observed in scleractinian and gorgonian deepsea coral.We also observed little variability in δ 13 C and δ 18 O of different preparations of the same coral, with standard deviations of < 1 ‰.
As a further investigation into intra-specimen heterogeneity, effects on 47 portions of two coral samples that differed in bulk isotopic composition were analyzed.In two corals -PIV 146-6 (scleractinian) and PV 703-5 (gorgonian) -δ 13 C, δ 18 O, and 47 were measured from intra-specimen aliquots of carbonate produced from sampling portions of the skeleton which differed significantly in their bulk isotopic composition (Table 5).Although 47 of the aliquots are approximately 0.02 ‰ different from each other in both corals, this is similar to that reported in other studies between a few replicates (Huntington et al., 2009;Thiagarajan et al., 2011).Additionally, no consistent trend is observed between δ 18 O and 47 ; as for coral PIV 146-6, the aliquot with the more positive δ 18 O also has the higher 47 , but the opposite is observed for coral PV 703-5.We therefore interpret the difference in 47 values between the aliquots as noise in the measurements, which further reinforces the need to run many replicates.

Diffusion
Both Thiagarajan et al. (2011) and Saenger et al. (2012) have suggested the diffusion of CO 2 across the calicoblastic membrane in corals as a possible mechanism for the fractionation of 47 .A generalized schematic of a scleractinian coral-calcifying region is shown in Fig. 11.Thiagarajan et al. (2011) calculated 0.7 and 1.6 ‰ decreases in δ 18 O and δ 13 C, respectively, and a 0.036 ‰ increase in 47 for liquid phase diffusion of CO 2 (Fig. 10).For this 47 fractionation to be heritable by the mineral, however, the mineral must preserve the isotopic state of ordering of the DIC rather than obtain internal equilibrium within the mineral carbonate itself.Additionally carbonate mineral growth would have to occur before isotope clumping could reach equilibrium.Furthermore, this fractionation only applies to DIC produced from diffused CO 2 and therefore the ratio of DIC from diffused DIC to DIC derived from seawater "leaking" directly to the calcifying environment would also have to be known.Both Thiagarajan et al. (2011) andSaenger et al. (2012) found fractionation associated with CO 2 diffusion across the calicoblastic layer or, additionally proposed by Saenger et al. (2012), across a boundary layer developed between the growing mineral and calcifying fluid to be an unlikely explanation of the trends observed in their data.Additionally, gorgonian corals are not known to have calicoblastic cells (Noé and Dullo, 2006) Solid-state reordering Interfacial reordering?
Figure 10.Schematic illustrating general trajectories for different processes on stable isotope signatures ( 63 or 47 values and δ 18 O), relative to an arbitrary point.Equilibrium temperature dependence is shown (Ghosh et al., 2006).Note that mixing of DIC from different sources results in curved trajectories (in δ 18 O space, mixing is linear, whereas in 47 space can be nonlinear) (Eiler and Schauble, 2004;Thiagarajan et al., 2011;Defliese and Lohmann, 2015).Diffusion results in enrichment of clumped species and depletion in δ 18 O (Thiagarajan et al., 2011).CO 2 hydration reactions are strongly temperature pH-dependent and also result in enrichment of clumped species and depletion in δ 18 O (Affek, 2013;Tang et al., 2014;Tripati et al., 2015).At any given temperature, equilibrium DIC species have differing clumped and oxygen isotope signatures, potentially also giving rise to pH effect (Hill et al., 2014;Tripati et al., 2015).Solid-state diffusion processes (Passey and Henkes, 2012;Henkes et al., 2014), including hypothesized reordering in the interfacial region (Tripati et al., 2015) only influence clumped isotope signatures and not bulk isotopic signatures (e.g., δ 18 O).
if there is a similar functional cell playing a role in this for gorgonian corals.
It is possible the observed pattern could be explained if scleractinian deep-sea corals contain a larger fraction of diffused CO 2 compared to gorgonian coral if this CO 2 then underwent relatively little isotopic equilibration in the calcifying fluid (Fig. 10).In cross plots of δ 18 O and 47 residuals, scleractinian corals do exhibit elevated 47 and depleted δ 18 O values, relative to Zaarur et al. (2013), Dennis and Schrag (2010), and theoretical aragonite (Schauble et al., 2006;Guo et al., 2009) in accordance with calculated diffusion effects.However, since scleractinian deep-sea corals generally conform to the majority of reported biogenic and synthetic calibrations, this would have to be a widespread effect occurring in other materials reported as well.

Mineralogy and 47
Despite Epstein et al. (1953) originally finding that the nacreous layer (aragonite) and prismatic layer (calcite) of mol- . Schematic of the scleractinian coral-calcifying region (modified from McConnaughey, 1989a;Adkins et al., 2003) illustrating processes that are hypothesized to result in stable isotope signatures observed in corals.Calcium pumping across the cell wall establishes a pH gradient between seawater and the extracellular calcifying fluid.Within the calcifying fluid, carbon can be sourced from CO 2 that is diffusively transported across the cell wall and/or through seawater leakage.The hydration and/or hydroxylation of CO 2 are slow reactions, particularly at low temperature and at high pH, driving disequilibrium in δ 18 O and in clumped isotope signatures.The enzyme carbonic anhydrase will catalyze this reaction if present and sufficiently active.At a given temperature, both HCO − 3 and CO 2− 3 ions have distinct clumped isotope and δ 18 O signatures, and therefore changes in extracellular calcifying fluid pH may also affect the isotopic composition of minerals, as explored through various models including the pH, surface entrapment, and interfacial models mentioned in the text.Surface kinetic and interfacial processes and properties (e.g., surface speciation, attachment or detachment rates for isotopologues, the occurrence of defects, the nature of protein and saccharide macromolecules) also affect crystal growth and chemistry.lusks showed no oxygen isotopic offset, it has since been recognized that aragonite is enriched in 18 O and 13 C relative to calcite both in theory and experiment (Tarutani et al., 1969;Kim and O'Neil, 1997;Kim et al., 2007;Mavromatis, 2013).An analogous mineralogical offset is predicted for 47 (with aragonite predicted to have a greater abundance of 13 C-18 O bonds than calcite that formed at the same temperature) (Schauble et al., 2006;Guo et al., 2009;Hill et al., 2014;Tripati et al., 2015).
The corals in this study differ in mineralogy, with scleractinian corals precipitating a fully aragonitic skeleton and gorgonian corals secreting carbonate in the form of high-Mg calcite.To date, a combination of aragonitic and calcitic biogenic materials have been analyzed.Calcitic specimens have been represented by certain species of mollusks (Eagle et al., 2013;Henkes et al., 2013), brachiopods (Came et al., 2007;Henkes et al., 2013), and foraminifera (Tripati et al., 2010).
Despite not observing this offset between calcite and aragonite 47 in synthetic or biogenic carbonates, it has been predicted in theoretical studies.By combining a previous theoretical model of 13 C-18 O clumping in carbonate minerals (Schauble et al., 2006) with their theoretical model of kinetic isotope effects associated with phosphoric acid digestion, Guo et al. (2009) proposed predicted 47 -T relationships for a number of carbonate minerals.For the temperature range 0-30 • C, 47, aragonite-calcite ≈ 0.02 and 47, aragonite-dolomite ≈ 0.04 ‰, with the direction of fractionation in agreement with the offset that we observe (Fig. 5).A similar pattern was reported by Hill et al. (2014) and Tripati et al. (2015).If these models of mineral-specific 47 -T relationships are correct, then this could potentially explain a substantial portion of the offset we observe between the aragonite and high-Mg calcite corals in this study.We discuss possible reasons for observed mineralogical effects in deep-sea corals and not in other types of carbonates in more detail below.

Calcification in scleractinian deep-sea corals
Nonequilibrium partitioning of oxygen and carbon isotopes has been observed in numerous calcifying organisms and inorganic precipitation experiments and linked to calcification processes (e.g., McConnaughey, 1989a;Spero et al., 1997;Zeebe, 1999;Adkins et al., 2003: Watson, 2004;DePaolo, 2011;Gabitov et al., 2012;Watkins et al., 2014).Clumped isotopes represent another tool that can be used to probe mechanisms for disequilibrium isotopic signals in biological carbonates and to determine when equilibrium precipitation has occurred, through the coupling of theoretical calculations and inorganic precipitation experiments with culturing studies (Thiagarajan et al., 2011;Tripati et al., 2015).Figure 10 shows a framework for how multiple paired stable isotope measurements can be used to trace the origin of disequilibrium effects.
In scleractinian corals, precipitation of carbonate is believed to occur in a space between the calicoblastic layer and the hard skeleton (Fig. 11).It is likely that calcifying fluid chemistry (pH, salinity) as well as growth rate and DIC equilibrium can all be potential sources of nonequilibrium signals preserved in the solid.The chemical composition and DIC sources from which the carbonate mineral forms have been a main focus for understanding nonequilibrium δ 13 C and δ 18 O values observed in corals.Biomineralization in shallow symbiotic scleractinian corals has been well-studied over the past few decades (see Cohen and McConnaughey, 2003, for a review).Early studies of shallow symbiotic scleractinian corals revealed time-independent δ 18 O and δ 13 C offsets from theoretical carbonate-water fractionation curves.This was addressed through species-specific calibration studies, which allowed offsets to be corrected for, revealing the underlying environmental signals (Weber and Woodhead, 1972;Dunbar and Wellington, 1981).In contrast, linear trends between δ 18 O and δ 13 C with depletions of several per mil are observed in an individual deep-sea coral.Living in nearconstant temperature, salinity, and pH environments, these large disequilibrium signals are attributed to their biomineralization mechanism (i.e., vital effects).Different mechanistic models aimed at explaining nonequilibrium δ 13 C and δ 18 O values seen in deep-sea corals have been proposed to account for these so-called vital effects.
The first of these models, proposed by Mc-Connaughey (1989a), calls on a kinetic fractionation during the hydration and hydroxylation steps of DIC speciation in which precipitation outpaces the ability of the system to reach isotopic equilibrium ("Kinetic model" in Fig. 11).The range of δ 13 C and δ 18 O values is therefore a reflection of a spectrum of varying amounts of equilibrium being attained amongst DIC species.The master variable in this case being time, McConnaughey (1989b), using calcification rates, calculated that calcification could outpace the time it took for isotopic equilibrium to be reached.Henkes et al. (2013) further noted that at lower temperatures, days to weeks are required for oxygen isotopic equilibrium to be reached in DIC.
The second model proposed by Adkins et al. (2003) extensively examines potential sources of DIC to the extracellular calcifying fluid and introduces the concept of a variable pH environment as the master controller on skeletal δ 13 C and δ 18 O ("pH model" in Fig. 11).For oxygen isotope processes, they draw on work characterizing the oxygen isotopic fractionation between DIC species (i.e., CO 2 , H 2 CO 3, , HCO − 3 , CO 2− 3 ) (McCrea, 1950;Usdowski et al., 1991;Zeebe and Wolf-Gladrow, 2001;Beck et al., 2005;Zeebe, 2007;Rollion-Bard et al., 2011).Their model suggests that the mineral preserves the oxygen isotopic ratio of HCO − 3 / CO 2− 3 .Because each DIC species has a unique oxygen fractionation relative to water, the oxygen isotope fractionation between the sum of DIC (DIC sum ) species and water will vary depending on the proportion of the DIC species.For seawater, at high pH (pH > 9), DIC will be dominated by CO 2− 3 , which has the lowest fractionation factor and will therefore result in a more depleted δ 18 O value of the DIC sum from which the mineral precipitates.At lower (pH 6-9) fluid pH, as the HCO − 3 / CO 2− tween DIC species is very rapid.δ 13 C, therefore, is proposed to be a function of two sources of CO 2 , membrane-diffused and seawater-leaked, and the proportion of them, a function of pH gradients across the membrane.The "break" that they uncover in the most depleted isotopic values, in which δ 18 O continues to decrease but δ 13 C plateaus, is interpreted as the point at which the maximum flux of diffused CO 2 across the membrane is reached (during rapid calcification).Qualitatively their model captures the isotopic trends that they see in their corals and although numerical modeling is able to reproduce the break and range of δ 18 O and δ 13 C that they report, it fails to reproduce the numerical value of the δ 18 O vs. δ 13 C slopes that they observe.δ 18 O vs. δ 13 C slopes have been observed in other studies and appear to be a conservative property of biomineralization in deep-sea corals (Smith et al., 2000;Adkins et al., 2003;Rollion-Bard et al., 2003, 2010;Lutringer et al., 2005;Hill et al., 2014;Kimball et al., 2014).Discrepancies between the biomineralization model of Adkins et al. (2003) and boron isotope data are yet to be resolved (Blamart et al., 2007;Rollion-Bard et al., 2010).Additional biomineralization models have been proposed in shallow symbiotic corals which incorporate the presence and activity of zooxanthellae (Cohen and McConnaughey, 2003;Rollion-Bard et al., 2003;Allison et al., 2010) and for deep-sea corals (Rollion-Bard et al., 2010;Gaetani et al., 2010).However, only those models that examine effects on δ 18 O and δ 13 C are considered with respect to the data presented herein.

Calcification in gorgonian deep-sea corals
The growth structure, mineralogy, and biomineralization of gorgonian coral skeletons is distinctly different from scleractinian corals (Noé and Dullo, 2006).Additionally, the organic matrix-mediated calcification process is largely unexplored and unknown for most gorgonian families.The biomineralization models described for scleractinian corals, therefore, are not valid when interpreting potential vital-effect signals in gorgonian corals.
An exhaustive study of the growth structure and fabric of deep-sea isidid gorgonians in the same family as those investigated here, was done by Noé and Dullo (2006).Their findings are based on the characterization of the isidid skeletal micro-and ultrastructure in combination with absolute age determinations.From their reconstructions of growth mode and fabric, they propose a biomineralization model.With no calicoblastic cells observed in gorgonians, they propose that gorgonian produced from the endodermal epithelium acts as the structural framework, while a viscous slime surrounding the skeleton acts as the matrix, facilitating and regulating mineral growth.Their model, however, relies on physical observations and lacks geochemical data to inform their interpretations.So while offering some insight into the uniqueness of isidid biomineralization, the findings are not useful in interpreting the geochemical data observed in this study.
Despite the lack of proposed biomineralization models in gorgonian corals, the trends in δ 13 C and δ 18 O have been explored and found to be similar in pattern to scleractinian corals, but with an offset that agrees with theoretical mineralogical fractionation predictions.The similarity in δ 13 C, δ 18 O, and δ 47 patterns suggests that while the particular biomineralization process is necessarily different in the two groups, there is a functional similarity that results in the conserved patterns in both gorgonian and scleractinian corals.

4.11
47 in deep-sea coral skeletal carbonate: biomineralization model implications The 47 of a mineral is predicted to conform to the K eq of a homogenous isotope exchange reaction (Reaction R1) and be independent of the fluid composition from which is forms, as well as the bulk isotopic composition of the crystal.Recent experimental and theoretical studies have suggested that mineral equilibrium with respect to 47 might not always be attained, however (Saenger et al., 2012;Tripati et al., 2015).Because of this, 47 can be utilized as an additional geochemical skeletal signal which could give more insight into the mechanisms involved in skeletal precipitation in deep-sea corals and other biological carbonates.
In a recent study (Tripati et al., 2015), the impact of DIC speciation on mineral 47 signatures was examined with respect to 47 .Contrary to initial theoretical modeling by Hill et al. (2014), who calculated differences of ∼ 0.03 ‰ in the 47 values of DIC species at equilibrium, Tripati et al. (2015), using experimental results, found that HCO − 3 63 could be as much as 0.05 to 0.07 ‰ higher than CO 2− 3 .They further propose that this fractionation could be partially or fully inheritable by the solid.Central to their mechanism of the preservation of disequilibrium mineral 63 signals is the concept of differing mineral isotopic archival effects occurring during "slow" and "fast" crystal growth, with slow and fast being defined as particular to a specific temperature, salinity, and pH growth environment, rather than conforming to a particular value.Depending on the proportion of HCO − 3 and CO 2− 3 incorporated into the growing crystal and the ability of the crystal lattice to conform to K eq (Reaction R1), disequilibrium in 63 and therefore 47 could be preserved.Only during slow crystal growth, they argue, is the mineral able to fully conform to K eq (Reaction R1).In this scenario, disequilibrium in 63 and bulk isotopes are not correlated and it would therefore be possible to attain equilibrium with respect to 63 , but not in bulk isotopes (Tripati et al., 2015).
Deep-sea corals, being one of the slowest calcifiers, are a likely potential candidate for the attainment of 63 mineral equilibrium if the mechanism proposed by Tripati et al. (2015) is correct.Bulk isotopes, as previously mentioned, vary greatly within an individual deep-sea coral, while 47 values are reproducible within instrumental error.This indicates that while bulk isotopes are controlled by processes such as in the biomineralization models outlined above, 47 values are independent and possibly conform to equilibrium values.While a complete quantitative biophysical model has yet to be proposed, one possibility is that the range in bulk isotopes is controlled by processes related to the chemical environment (e.g., calcifying solution pH, kinetics), while the slow growth rates in deep-sea corals allow isotopic reordering at the crystal-solution interface resulting in the attainment of nominal mineral equilibrium for 47 (Fig. 11).Adkins et al. (2003), however, find that in the more dense (interpreted as faster calcification) growth bands, δ 18 O is depleted compared to the less dense (more slowly growing) bands.In the Tripati et al. (2015) model, they predict that during rapid calcification both δ 18 O and 47 would be more positive.Relative to the calcification rates required for inheritance of these signals, however, perhaps even at the most elevated calcification rates, deep-sea corals, are still able to reach mineral isotopic equilibrium with respect to 47 , while bulk δ 18 O and δ 13 C is still under the control of processes associated with the calcifying fluid environment.
Cross plots of δ 18 O and 47 residuals (Figs. 7, 8, and 9) can be used to assess conformation to proposed biomineralization models.For the McConnaughey et al. (2003) model of coral biomineralization, a range of depletions in δ 18 O is expected, with coral δ 18 O approaching or potentially reaching mineral equilibrium δ 18 O, depending on the calcification rate.Similarly in the Adkins et al. (2003) model, a range of pH would result in a range of depletions in recorded mineral δ 18 O, assuming that the proportion of HCO − 3 / CO 2− 3 incorporated into the skeleton is independent of calcification rate (Fig. 10).Although neither model addresses 63 , in both cases, all else being equal, and assuming a mechanism for inheritance of DIC 63 by mineral 63 , a decrease in δ 18 O would be tied to a decrease in 63 .For the McConnaughey et al. (2003) model, an increase in calcification rate resulting in only partial equilibration of the DIC-water system would result in depleted δ 18 O and 63 values of HCO − 3 and CO 2− 3 .Similarly, in Adkins et al. (2003) model, varying solution pH would have the same effect on δ 18 O and 63 with both values decreasing (increasing) as pH increases (decreases).Comparing residual cross plots from various 47 calibrations, however, fails to capture a positive relationship between δ 18 O and 47 , and it instead appears that δ 18 O and 4.12 Other biomineralization effects Unlike scleractinian zooxanthellate and azooxanthelate corals, gorgonian corals have been much less studied.In particular, biomineralization in gorgonian corals has yet to be addressed from a geochemical perspective.The fact that gorgonians precipitate high-Mg calcite, however, in contrast to the aragonitic skeletons of scleractinians, is a simple indication of differing biomineralization mechanisms (Ries, 2010).Additionally, although it has been shown that deep-sea gorgonian corals exhibit δ 18 O and δ 13 C trends similar to their scleractinian counterparts, they have not been shown to exhibit the break that is observed in Adkins et al. (2003) (Kimball et al., 2014).When growing inorganic carbonate precipitates, mineralogy can be controlled through the Mg / Ca ratio of the precipitating fluid, with molar Mg / Ca > 2 producing aragonite and high-Mg calcite and molar Mg / Ca < 2 producing calcite (reviewed by Ries, 2010).Current seas have ca.Mg / Ca = 5.2 (Ries, 2010), and the biomineralization mechanism responsible for precipitation of high-Mg calcite in gorgonians is unknown.
One possibility is that biomineralization is regulated in gorgonians through an organic template.The presence of an insoluble organic matrix in both isidid bamboo and Corallium sp.(Allemand et al., 1994;Ehrlich et al., 2006;Noé and Dullo, 2006) supports the possibility that an organic template may be directing mineral precipitation.The presence of an organic matrix is unique to gorgonian corals (has not been observed in scleractinians) and only preliminary studies have been undertaken to characterize the amino acid composition of the organic matrixes (Allemand et al., 1994;Ehrlich et al., 2006).
It seems plausible that the organic template, in addition to directing lattice structure, could also be influencing isotopic fractionation within a growing mineral, as has recently been shown for Mg / Ca ratios in carbonates (Wang et al., 2009).Therefore, in addition to theoretical depletions expected in 63 due to mineralogy, the presence of an organic template might act as an additional control on bond ordering.Carbonate anhydrase has also been implicated in scleractinian hermatypic coral biomineralization models, and the presence or absence and variable activities of carbonic anhydrase may be another factor to be considered.

Suggestions for future work
In-depth experimental studies, such as Wacker et al. (2013) and Defliese et al. (2015), are needed to explore the role of multiple processes in acid digestion kinetics.Possible factors influencing acid digestion fractionation factors that still need to be examined include (but are not limited to) organic matter content, the influence of cation substitution (e.g., high vs. low Mg calcite), variations in bulk composition (δ 13 C, δ 18 O), the type of reaction vessel and cleanup method used, and phosphoric acid composition.
Direct constraints on the impact of variable cation concentrations on the clumped isotopic composition of carbonate minerals from theory and synthetic experiments may be a useful tool to distinguish between whether these different deep-sea coral orders in fact are recording equilibrium mineral-specific offsets.Additionally, the measurement of coral-calcifying fluid pH and the use of labeled carbon sources during growth experiments would provide important constraints on whether the observed patterns reflect kinetic effects.
A unifying model which explains the observed patterns of isotopic and trace element variability that have been observed could also help to test the different explanations posed in this study and other publications and could help us advance our understanding of biomineralization in scleractinian and gorgonian deep-sea corals.For clumped isotopes, such a model would need to include the factors that influence equilibrium fractionation, i.e., temperature and well-constrained mineral-specific fractionation factors.Additional factors that need to be included to accurately model kinetic isotope effects include temperature, the carbon source used for calcification (the relative fraction of seawater-derived DIC vs. CO 2 diffused across cellular membranes), DIC speciation as controlled by multiple factors (pH, temperature, salinity), and the timescale for DIC equilibration vs. precipitation rate (affected by the residence time of DIC in the calcification space), as modulated by the occurrence and activity of the enzyme carbonic anhydrase.

Conclusions
Two groups of deep-sea corals with differing mineralogy (aragonite and calcite) were analyzed for δ 13 C, δ 18 O, and 47 .Aragonitic scleractinian deep-sea coral exhibit a nearly coincident 47 -T relationship to previously reported scleractinian deep-sea corals (Thiagarjan et al., 2011) and can be used to predict temperatures to ±2 • C at the mean temperature and ±3-5 • C at the extreme ends of the data set.High-Mg calcitic gorgonian deep-sea corals show lower 47 values.This offset between taxa is consistent with theoretical predictions for mineralogic differences, albeit significantly larger in magnitude than expected.Therefore, the offset between gorgonian and scleractinian deep-sea corals may be reflected, at least in part, as a difference in mineralogy that is only observed in deep-sea corals and not in data published for other carbonates simply because of their very slow growth rates as compared to all other studied carbonates.Another equally likely explanation is that differences in carbon sources and biomineralization pathways in coral taxa result in distinctive clumped isotope vital effects in gorgonian and scleractinian deep-sea coral.Finally, we cannot preclude that offsets may reflect variable acid digestion kinetics for the different types of materials that we have surveyed.

Information about the Supplement
All raw data for samples and standards and equilibrated gas lines are provided in the data tables in the Supplement.
The Supplement related to this article is available online at doi:10.5194/bg-13-6487-2016-supplement.

Figure 1 .
Figure 1.Sampling locations of three expeditions from which coral specimens were collected by deep-sea submersible diving.Locations approximate Warwick Seamount, Gulf of Alaska (DSRV Alvin, 2004), Hawaiian and Northwest Hawaiian Islands (DSRV Pisces V, 2005 and 2007), and Line Islands (DSRV Pisces IV, 2006).

Figure 2 .
Figure 2. (a) A bamboo coral from Warwick Seamount, ALV 3808-4.The bottommost internode is cut and the disk used for milling along the outer, most recently accreted portion.(b) A portion of a radial disk from PV 703-5 in which samples were milled from the center and outer portion.

Figure 3 .
Figure 3. Scleractinian and gorgonian deep-sea coral compared to deep-sea corals reported in Thiagarajan et al. (2011) and average Figure 4

Figure 9 .
Figure 9. 47 residuals are calculated from average measured 47 values for each coral relative to the theoretical calibration of Schauble et al. (2006) paired with theoretical acid digestion fractionations fromGuo et al. (2009).δ 18 O residuals are calculated from average measured δ 18 O values relative to the calibrations ofKim and O'Neil (1997) andKim et al. (2007) for calcite and aragonite, respectively.
the depths the corals were collected at in the same oceanic region.Similarly temperature ranges from the Line Islands are estimated from CTD casts extracted from National Oceanic Data Center taken at approximately 6 • N, −160 • W during 9 to 27 cruises (depending on depth) during1972-1998.
• C results) (‰ ARF; adj. 25 • C 1 ) (‰ ARF; adj. 25 • C 2 ) al. (2011) andGhosh et al. (2006) (Figs. 3 Thiagarajan et al. (2011)yllum, Enallopsammia, and Caryophyllia sp.All scleractinian species precipitate a skeleton composed of aragonite, although growth habits vary between species.Desmophyllum sp., also known as cup corals, grows in a rosette shape, whereas Caryophyllia and Enallopsammia sp. have a dendritic growth habit.The E. rostrata specimens measured inThiagarajan et al. (2011)allow comparison between deep-sea corals of the same species measured in this study.The scleractinian deep-sea coral data from this study have values that overlap with those reported by
a Winter.b Summer 2013.

Table 4 .
Coral specimens analyzed for δ 13 C, δ 18 O, and 47 .Average 47 for all replicate measurements is reported relative to the absolute reference frame (ARF).δ 47 is reported relative to intra-laboratory working gas (WG).
a δ 18 O water values are from WOCE and HOTS.b δ 18 O predicted values are calculated using the equations reported in Thiagarajan et al. (2011)an corals from this study andThiagarajan et al. (2011)are compared to the theoretical predictions for 63 for aragonite, calcite, and dolomite.Acid fractionation values were added to convert to 47 .
Figure 8. 47 residuals are calculated from average measured 47 values for each coral relative to the biogenic calibrations of Eagle et al. (2013).δ 18 O residuals are calculated from average measured δ 18 O values relative to the calibrations of , so this is relevant to scleractinian corals or