PII S0016-7037(99)00422-6

U-Th dating of deep-sea corals

HAI CHENG,1 JESS ADKINS,2,* R. LAWRENCE EDWARDS,1 and EDWARD A. BOYLE3

1 Minnesota Isotope Laboratory, Department of Geology and Geophysics, University of Minnesota, Minneapolis, MN 55455 USA2 Geochemistry 62, Lamont-Doherty Earth Observatory, Rt. 9W, Palisades, NY 10964 USA

3 Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139-4307 USA

(Received February24, 1999;accepted in revised form November16, 1999)

Abstract—230Th, 232Th, 234U and238U compositions of several deep-sea solitary corals, mainly the speciesDesmophyllum cristagalli, were determined by thermal ionization mass spectrometry (TIMS). It is possible toobtain high precision ages on modern pristine corals that have low [232Th] (5 to a few hundred ppt). However,because older deep-sea corals tend to have higher [232Th] compared to surface corals, and the initial230Th/232Th ratio is uncertain, older deep-sea corals have larger age uncertainties (6several hundred years forsamples with a few thousand ppt232Th). Therefore, the key hurdle for precise U-Th dating is to remove oraccount for contaminants which contain elevated232Th and associated230Th not due to closed system decaywithin the coral lattice. A modification of the trace metal cleaning methods used for foraminifera and surfacecorals can significantly reduce this contamination. By counting the visible growth bands and measuring themean age of a single septum, the extension rate ofD. cristagalli was estimated to be between 0.1 and 3.1mm/year. In a mean sense, bands appear to be precipitated annually, but this estimate has a large uncertainty.If appropriate tracer calibrations can be established, these corals are therefore suitable to record decadal orsub-decadal oceanographic changes over the course of their lifetime. Thed234U values of all modern samplesfrom different localities and different depths are similar (mean 145.56 2.3‰) and indistinguishable from thedata obtained from surface corals. At a precision of about62‰, we find no structure in the oceanic profileof d234U ratios over the top 2000 m of the water column.Copyright © 2000 Elsevier Science Ltd

1. INTRODUCTION

Uranium series dates from surface corals provide constraintson several late Quaternary climate processes. Past sea levelestimates froma-counted230Th dates on raised coral terracesfrom Barbados supported the Milankovitch hypothesis on therelation between glacier ice volume and insolation (Broecker etal., 1968; Mesolella et al.,1969). Later, more precise and ac-curate thermal ionization mass spectrometry (TIMS) confirmedthis result in a variety of locations (Edwards et al., 1987a,b;Gallup et al., 1994; Muhs et al., 1994; Stirling et al., 1995;Szabo et al., 1994). Drill cores of submerged corals that havebeen precisely dated by230Th methods contain our most de-tailed and continuous record of sea-level since the last glacialmaximum (Bard et al., 1993; Bard et al., 1996; Edwards et al.,1993). Coupled high precision radiocarbon and uranium seriesdates from surface corals constrain the history of atmosphericD14C beyond the tree ring calibration (Bard et al., 1993; Bardet al., 1990; Edwards et al., 1993). In addition, TIMS datesprovide precise ages for coral tracer-based studies of pastoceanographic conditions (Beck et al., 1997; Beck et al., 1992;Gagan et al., 1998; Guilderson et al., 1994; McCulloch et al.,1996).

Analogous studies of deep-sea corals promise to constrain anumber of deep ocean processes (Adkins et al., 1998; Manginiet al., 1998; Smith et al., 1997). Several cosmopolitan genera ofdeep-sea scleractinia inhabit all major ocean basins. Coral hasbeen dredged from the ocean floor since at least the days of theChallenger expedition, so the world’s collections contain thou-

sands of samples. While reported depth habitats of deep-seacoral range from 60–6000 meters (Cairns and Stanley, 1981),most specimens are found between 500–2000 meters. Thisdepth range makes deep-sea corals ideal for studies of inter-mediate and upper deep water masses. The great potential ofthis archive stems from the fact that the density banded coralskeletons are not subject to bioturbation and the fact that theskeletons are potentially datable by uranium-series techniques(Cheng et al., 1995; Goldstein et al., 1996; Smith et al., 1997).Thus, deep-sea corals can provide absolute-dated records withtemporal resolutions not generally attainable in deep-sea sedi-ments and are a new archive of deep circulation rates bycoupling U-series ages with14C dates (Adkins and Boyle,1998; Mangini et al., 1998). However, realization of this po-tential requires evaluating the U-Th dating systematics (Lo-mitschka and Mangini, 1999) and growth rates of deep-seacorals. This study focuses on the ubiquitous pseudo-colonialspecies,Desmophyllum cristagalli, using both a modern and afossil sample set.

Several previous studies have examined deep-sea coralgrowth rate and its relationship to species specific densitybanding. Duncan (1877) and Pratje (1924) reported extensionrates of 6.8 and 7.5 mm/year for specimens of the genusLopheliathat were attached to underwater transatlantic cables.Teichert (1958) also estimated growth rates of 7.5–15 mm/yearfor Lophelia. Grigg (1974) determined a value of about 20mm/year vertically and 3 mm/year horizontally for two gorgo-nains by tagging colonies ofMuricea californicaandMuriceafurticosa from a relatively shallow depth of 14–20 m. Hesuggested that the periodicity of growth ring formation wasannual based on matching the estimated ages to the number ofgrowth bands. Using the210Pb dating method, Druffel et al.

*Author to whom correspondence should be addressed: MS 100-23Caltech 1200 E. California Blvd., Pasadena, CA 91125-0001 (jess@gps.caltech.edu).

Pergamon

Geochimica et Cosmochimica Acta, Vol. 64, No. 14, pp. 2401–2416, 2000Copyright © 2000 Elsevier Science LtdPrinted in the USA. All rights reserved

0016-7037/00 $20.001 .00

2401

(1990) calculated a mean value of 0.116 0.02 mm/year inwidth for the calcitic deep-sea gorgonianCorallium niobe,which grew at 600 m depth. In that study, the periodicity ofgrowth band formation did not appear to be annual. Our ap-proach is to use U-series disequilibria to constrain dating andgrowth systematics in several deep-sea corals.

The behavior of238U, 234U, 230Th, and232Th in reef-build-ing corals and the suitability of reef-building corals for230Thdating is the subject of an extensive literature starting withBarnes et al. (1956) and continuing up to the present, seeIvanovich and Harmon (1992) and references therein. Thebehavior of uranium-series nuclides in surface-dwelling soli-tary corals has also received attention (Ku and Kern, 1974;Muhs et al., 1994; Stein et al., 1991; Szabo, 1985). However,similar studies on deep-sea corals are generally lacking (Adkinset al., 1998; Cheng et al., 1995; Goldstein et al., 1996; Smith etal., 1997; Lomitschka and Mangini, 1999). In many respects,one would expect the U-Th dating systematics of deep-seacorals to be similar to those of their surface-dwelling counter-parts. However, deep-sea coral systematics should differ in atleast two key respects. First, deep ocean waters have muchhigher 230Th concentrations than surface waters. Thus, deep-sea corals may incorporate a significant amount of unsupported230Th during growth. If so, a correction for initial230Th mustbe made, and the precision and accuracy of the230Th age willdepend on the precision and accuracy of the correction. Second,the environments in which deep-sea corals age are quite differentfrom those in which surface corals age. Thus, diagenetic processesmay affect corals in the two settings in different ways.

In this study, we investigate the U-Th dating systematics ofseveral species of aragonitic deep-sea corals, focusing onDes-mophyllum Cristagalli, which is the most abundant coral in ourcollection of dredged specimens. We establish growth rates andinitial chemical and isotopic characteristics by analyzing238U,234U, 230Th, and232Th concentrations in modern samples. Weconstrain diagenetic shifts in the pertinent nuclides by compar-ing the isotopic characteristics of the modern specimens tothose of fossil samples, and by analyzing coatings on thespecimens. By analyzing different sub-samples of the samecoral, we show that deep-sea corals generally have significantamounts of initial and/or added thorium. We investigatesources of this thorium, as well as methods to remove or correctfor initial 230Th. Our results are specifically applicable to the

study of Adkins et al. (1998), which established that deep-ocean circulation changed dramatically in less than severaldecades at 15.4 ka. The time and the duration of the circulationchange is established with uranium-series data and interpreta-tions presented here. Beyond this specific application, we en-vision that our results will be more generally applicable tofurther studies in the emerging field of deep-sea coral pale-oceanography.

2. SAMPLES AND METHODS

2.1. Samples

Deep-sea coral samples were provided by the Smithsonian Institu-tion and the Woods Hole Oceanographic Institution dredge collection.Five modern samples ofD. cristagalli were collected by dredgingprograms in the Pacific, the Atlantic, and the Southern Oceans over thepast 30 years (Table 1 and Fig. 1). Sample depths range from 420 to2200 m. Samples were judged to be modern either because of intactorganic matter, noted upon recovery, or very fresh looking preservationof the septa in the dredge collections. Prior to cleaning, the modernsamples often were covered with a yellowish organic coating. Six fossilsamples were dredged from the Atlantic between depths of 1700 to2000 m (Table 1). All specimens areD. cristagalli except for oneSolenosmilia sp.specimen and one unidentified sample. As opposed tothe modern samples, the fossil samples are almost always covered witha black crust. Energy-dispersive electron microprobe analysis indicatesthat this crust is composed of iron and manganese oxides mixed withdetrital aluminosilicates (Fig. 2a). It is also possible to find discretemetal sulfide deposits in the crust’s matrix (Figure 2b).

2.2. Sample Cutting

Samples were cut into small slabs parallel to the radially symmetricsepta, which are connected around the outer rim by a thick-walled theca(Fig. 3). Septa in individual deep-sea corals are classified on the basisof their relative cross-sectional areas in a radial plane, where the S1septa have the largest cross-sectional areas (Fig. 3, refer to (Cairns,1981) for classification of septa). Each slab generally consists of onewhole S1 septum associated with either one or two smaller septa onboth sides of the S1. S1 septa typically have tens of density bandssub-parallel to their upper surfaces (Fig. 4). The thickness of individualsepta are about 0.1–0.7 mm, with the S1 being the thickest. Eachseptum’s thickness decreases vertically from a bulge near the top andincreases from the interior towards the theca. After cutting, the slabswere set aside for one of two cleaning procedures described below.Slabs of the five modernD. cristagalli samples were further sub-sampled into smaller pieces roughly parallel to their growth bandsusing a small blade (Fig. 3). Final weights of these modern coralsub-samples ranged from 28.9 to 143.3 mg.

Table 1. Deep sea coral samples used in this study.

Sampleset

Samplenumber

Analysisnumber

Collectiontime

Depth(m)

LocalityLat Long Coral species

Solitaryor colonial

Modern47413 DC-2 a, b Jan. 21, 1964 421 50°389S 167°389E Desmophyllum cristagalli Solitary84820 DC-4 a, b Nov. 23, 1986 806 0°149N 91°369W Desmophyllum cristagalli Solitary85080 DC-3 a, b Feb. 25, 1990 990–1150 43°47S 150°299E Desmophyllum cristagalli Solitary48740 DC-5 Oct. 28, 1973 1420–1470 48°409N 10°549W Desmophyllum cristagalli Solitary78459 DC-1 Apr. 16, 1987 2110–2180 38°459N 72°399W Desmophyllum cristagalli Solitary

PaleoJFA-2 2 1684–1829 42°N 29°W ? ?JFA-17 17 1684–1829 42°N 29°W Solenosmillia sp. ColonialJFA-24C 24 1784 38°N 60°W Desmophyllum cristagalli SolitaryJFA-20A 20A 1954 38°N 62°W Desmophyllum cristagalli SolitaryJFA-20B 20B 1954 38°N 62°W Desmophyllum cristagalli SolitaryJFA-20C 20C 1954 38°N 62°W Desmophyllum cristagalli Solitary

2402 Cheng et al.

2.3. Ultrasonic Cleaning

All sub-samples of the five modern samples (Table 2) and somesub-samples of the six fossil corals (labeled “N” in Table 3) werecleaned with the following ultrasonic cleaning technique. Under abinocular microscope, samples were checked for organic coatings or

iron-manganese oxide crusts, which were scrubbed with small dentaltools. Samples were put in a plastic bottle, ultrasonically cleaned indeionized water for 10 min. and then completely rinsed. This processwas repeated several times until the coral looked clean (free of crusts)under the microscope. The above procedure was then repeated threetimes in a Teflon beaker with 5 min. of ultrasonic cleaning eachtime.Finally, the sample was dried in an oven at about 70°C and set aside forisotopic analysis. Some of the yellowish organic material from modernsamples 85080 and 78459 (labeled “D” in Table 2) and some of theiron-manganese crustfrom fossil samples JFA 2 and JFA 24C (alsolabeled “D” in Table 3) were also saved for isotopic analysis.

2.4. Chemical Cleaning

Early in this project it became clear that one of the major limitingfactors in U-Th dating of deep-sea corals would be the ability toaccurately and precisely account for initial and added thorium. Re-moval of thorium on or near coral surfaces can help solve this problem.We therefore modified chemical and physical cleaning techniquesdeveloped some years ago at M.I.T. for trace metal analysis (Boyle,1981; Boyle and Keigwin, 1985/6; Shen and Boyle, 1988) and appliedthem to the remaining modern and fossil sub-samples. As has beendemonstrated previously for surface corals and foraminifera, thesetechniques can remove exterior contaminants from fossil samples andimprove accuracy and reproducibility. We followed the procedure ofShen and Boyle (1988) except that the pre-cleaning step was altered tobetter remove the black crusts. After scrubbing with a brush anddeionized water, samples were placed in plastic tubes with cleandistilled H2O and ultrasonicated for 15 min. Corals were then sub-merged in a 50/50 mixture of 30% H2O2 and 1M NaOH for 15 min.with ultrasonication. This step was repeated several times until therewas little crust left. Occasionally samples were scrubbed with a brushbetween oxidative leaches to remove loose material. The last step wasa brief (30 sec. to 2 min.) rinse in a 50/50 mixture of 30% peroxide and1% HClO4. This step effectively removes all organic stains left on thecoral but also removes about 5–10% of the skeleton. The cleaningsolution in this last step always had a pH, 2. Samples were thenthoroughly rinsed with clean distilled water and left to dry in a HEPAfiltered laminar flow bench. The visible iron-magnesium oxide crusts areeasily removed with these oxidizing steps, flaking off in small sheets. Thissuggests that the crust is bound by an organic “glue”, possibly remnantpolyp organic matter. Subsequent cleaning followed the method of Shenand Boyle (1988) and used clean reagents throughout.

Fig. 1. Sample locations used in this study. Sample depths, in parentheses, are in meters.

Fig. 2. Electron micro-probe analysis of black crusts from a fossilsample ofD. cristagalli. The bulk crust (a) is a mixture of detritalminerals and authegenic iron and manganese oxides. Discrete sulfidemineral phases (b) are also found in the crust matrix. This figure ismeant as a description of some of the contaminants found in coralcrusts, not as a quantitative measure of contamination.

2403U-Th dating of deep-sea corals

2.5. Thermal Ionization Mass Spectrometry

Procedures for chemical separation and instrumental analysis ofuranium and thorium are modifications of those previously describedfor surface corals (Chen et al., 1986; Edwards et al., 1993; Edwards,1988; Edwards et al., 1987b). Ultrasonically or chemically cleanedsamples were slowly dissolved in nitric acid. This was sufficient todissolve all samples completely, except for a residue in the iron-manganese oxide crust samples. The residue, presumably detrital alu-minosilicate contained in the crust, dissolved readily in hydrofluoricacid. The solutions were spiked with233U-236U (233U/236U 51.010527) and229Th solutions of known concentration. Following thechemical separation of thorium and uranium, the thorium fraction wasloaded on a graphite-coated single Re filament, which had been previ-ously checked for its thorium blank. The filament blank for232Th isgenerally 100–150 counts per second at around 1700°C and increaseswith increasing temperature. This is a significant fraction of the total232Th beam in typical sample runs, particularly for the runs on smallsamples with relatively low232Th concentrations. We corrected for thisblank assuming an uncertainty of675% in the value of the filamentblank. In order to minimize the filament blank, we measured232Th attemperatures under 1700°C; nevertheless, as has been shown previ-ously (Edwards et al., 1987b), the uncertainty in the filament blankcorrection is the main source of error in the232Th measurements. Theuranium fraction was loaded on a Re filament without graphite and runwith the double-filament technique. Uranium and thorium were mea-sured on the Minnesota Isotope Laboratory’s Finnigan-MAT 262-RPQmass spectrometer on the first stage electron multiplier prior to thestatic quadrupole second stage. Tails were accounted for by measuringcount rates at half masses.

3. RESULTS AND DISCUSSION

3.1. Reproducibility

Uranium and thorium isotope compositions of modern andfossil samples are presented in Tables 2 and 3 respectively. Wemeasured a number of samples in replicate. In one case weanalyzed separate aliquots of the same solution (Table 2: DC-1T(I) and DC-1 T(II)). These analyses agreed within error for the230Th age and all uranium and thorium concentrations andisotope ratios. In four cases (Table 3), we analyzed two sepa-rate septa from the same fossil coral that were cleaned using thesame method. Three pairs were treated using the ultrasoniccleaning method; JFA 2 (N) I and II, JFA 24C (N) I and II, andJFA 20C (N) I and II, and one pair was treated using thechemical cleaning method; JFA 24C I and II. In all four cases,the 238U and 232Th concentrations did not replicate withinanalytical errors, indicating that concentrations of these nu-clides differ in separate sub-samples of the same coral. In threeof the cases the230Th ages, uncorrected for initial230Th, agreedwithin analytical errors. In the fourth case (the JFA 24C (N)pair), the uncorrected230Th ages differed significantly. In thiscase, the232Th concentrations differed by more than a factor oftwo and the230Th ages quite likely differed because of differ-ing initial 230Th contents (see discussion below). In all fourcases the measuredd234U values (defined as ((234U/238U)sample/(234U/238U)standard21) 31000) replicated within errors. In sum,

Fig. 3. (a) A typical side view of the deep-sea solitary coralDesmophyllum cristagalli. The drawing is adapted from(Cairns, 1981). (b) The location of sub-samples from the five modernD. cristagalli. Each sub-sample includes a mainseptum (S1) and one or two smaller septa. The fine dashed lines indicate the pattern of growth bands. The sample analysisnumbers are on the right side of each sub-sample and the heavy dashed lines are their boundaries. Where there are multiplesepta analyzed from the same coral (see Table 2), septum “a” is the one pictured. The middle piece of 85080 was lost duringsampling.

2404 Cheng et al.

replicate analyses agreed within analytical error in all caseswhere one would expect agreement.

3.2. Modern Samples

3.2.1. Uranium concentrations and isotopic ratios

Ku et al. (1977) and Chen et al. (1986) demonstrated thatwithin several percent uranium is conservative in seawater. Ourmeasured uranium concentrations in modernD. cristagalli, onthe other hand, vary from 2963 to 5531 ppb (Table 2). Theseuranium values are higher than those of most hermatypic corals(e.g., 2–4 ppm, (Burnett and Veeh, 1992)) and are similar tothose of other ahermatypic corals (Stein et al., 1991; Thompsonand Livingston, 1970; Lomitschka and Mangini, 1999). Insurface corals, uranium variations are correlated to changes insea surface temperature (Min et al., 1995; Shen and Dunbar,1995) . However, for a single septum ofD. cristagalli, themeasured range of uranium concentration is larger than thesurface coral data, while the deep water temperature variationsare much smaller. Clearly there is a combination of otherenvironmental parameters, such as temperature, pH, carbonateion concentration, growth rate, and/or a biologically induced“vital effect” affecting theD. cristagalli uranium concentration(Gvirtzman et al., 1973). In general, the primary uraniumconcentrations in deep-sea corals are higher than those of mostother biogenic deep-sea carbonates. For example, foraminifera

shells have primary uranium concentrations about 200 timeslower than deep-sea corals (Delaney and Boyle, 1983; Hender-son and O’Nions, 1995; Ku, 1965; Russell et al., 1994) . Thesehigh uranium contents make deep-sea corals ideal candidatesfor uranium-series dating.

The d234U value for modern hermatypic corals measured inthe Minnesota lab is 145.86 1.9‰ (2s) (Cheng et al., 1999;Edwards et al., 1993). This ratio is about 3‰ lower than earlierreports because we have used the newl234value determined byCheng et al. (1999). Each modern initiald234U value for ourdeep-sea corals is within error of the surface coral value (Fig.5, Table 2). The mean and two sigma error of all 20 values is145.36 2.3‰. As the deep-sea samples were collected fromnine different depths across the Pacific, Atlantic, Indian andSouthern Oceans, these data indicate that the sea waterd234Uvalue is conservative in the upper 2000 meters of the world’soceans within about62‰. In addition, surface coral dataindicate that since the last glacial maximum (Bard et al., 1993;Edwards et al., 1993) and for sea level high stands over the past200,000 years (Gallup et al., 1994; Gallup and Edwards, 1997)thed234U of surface seawater has been the same as the modernvalue within several per mil (Henderson et al., 1993). Sinceseawaterd234U appears to be temporally and spatially constant,the d234U values in fossil deep-sea corals may be used as acheck of diagenetic alteration.

Fig. 4. Transmitted light image of the side view of modernD. cristagalli sample number 47413. This single septum wasplaced in a standard photographic enlarger to expose the density banding structure. X-rays were not required to see thebands. Alternating light and dark density bands are visible sub-parallel to the top edge of the septum. The white area on thebottom is the thickened portion where the septum connects to the rest of the skeleton around the outer edge of the coral.White flecks are from dust on the sample. Bands can be seen with the naked eye but are more clear in transmitted light.

2405U-Th dating of deep-sea corals

3.2.2. Initial 230Th

As opposed to uranium, thorium is not conservative in sea-water. Since both230Th concentrations and230Th/232Th ratiosincrease with depth (Anderson et al., 1983; Bacon and Ander-son, 1982; Cochran, 1992; Cochran et al., 1987; Guo et al.,1995; Huh and Beasley, 1987; Huh et al., 1989; Moran et al.,1995; Nozaki and Horibe, 1983; Nozaki et al., 1987; Roy-Barman et al., 1996), we expect much higher initial230Thconcentrations in deep sea corals than in surface corals. There-fore, we need to estimate the initial230Th/238U ratio for deep-seacorals in order to calculate their230Th ages. We can write the230Th age equation, including the term for initial230Th/238U:

230Th238U

5 1 1 S S 230Th238U D 0

2 1D e2l230t

1d234U~0!

1000 S l230

l230 2 l234D ~1 2 e~l2342l230!t! (1)

where all isotope ratios are activity ratios, thel’s are decayconstants, t is the230Th age, (230Th/238U)° is the term for initial230Th/238U, andd234U(0) is as defined previously (Edwards etal., 1987b). We can substitute (232Th/238U)(230Th/232Th)° for(230Th/238U)°:

230Th238U

5 1 1 S S 232Th238U DS 230Th

232ThD0

2 1D e2l230t

1d234U~0!

1000 S l230

l230 2 l234D ~1 2 e~l2342l230!t! (2)

As (232Th/238U) can be measured, our problem reduces to oneof estimating the initial230Th/232Th ratio, an issue that we willdiscuss at length below. Because (230Th/232Th)° cannot bedetermined exactly, the error in its estimation introduces errorin t, the 230Th age. We will show that this uncertainty is themain source of error in the230Th age for most deep-sea corals.

In general one might expect two possible sources of initialthorium: one from the water column, similar to the “hydroge-nous” component of Lin et al. (1996), and one from continen-tially derived detrital particles. Both dissolved and particulatesea water230Th/232Th generally increase with depth and watermass age. Surface waters are as low as about 5–103 1026

(atomic ratio) (Cochran et al., 1987; Guo et al., 1995; Nozaki etal., 1987), while maximum values in the deep Pacific are.700 3 1026 (Moore, 1981). On the other hand, the thoriumisotopic composition of detrital clays is expected to be close tothe bulk earth value of 43 1026, which is based on a232Th/238U value of 3.8 and the assumption of secular equilibrium

Table 2. U and Th isotopic composition and230Th ages of modern solitary deep sea corals.

Samplenumber

Analysisnumber

Weight(mg)

238U(ppb)

232Th(ppb)

230Th/232Th(atomic, ppm)

d234U(measured)

230Th/238U(activity)

Age(yearb)

d234U(initial)

47413 DC-2aT 129.19 33056 2 56 18 42966 15050 145.16 1.5 0.0004066 0.000047 7.76 4.5 145.16 1.5

M1 114.24 29636 2 406 21 5706 304 145.86 1.6 0.0004616 0.000058 12.96 5.5 145.86 1.6M2 100.13 30886 2 526 24 5016 242 147.26 1.1 0.0005076 0.000058 17.26 5.5 147.26 1.1B 81.09 32576 2 1536 31 2216 56 145.46 2.9 0.0006296 0.000069 28.96 6.5 145.46 2.9

DC-2bT 43.87 30036 2 636 41 3446 244 144.66 1.9 0.0004376 0.000124 10.76 11.8 144.66 1.9M 47.57 30246 15 696 49 3866 292 143.86 5.2 0.0005596 0.000138 22.36 12.7 143.86 5.2

84820 DC-4aT 72.04 52326 2 1396 35 10286 259 144.56 1.2 0.0016556 0.000072 149.56 6.8 144.66 1.2M 66.05 53356 3 1076 37 11386 397 145.16 1.7 0.0013806 0.000063 123.36 6.0 145.16 1.7B 139.90 48986 2 766 27 16676 607 143.66 1.3 0.0015586 0.000096 140.46 9.1 143.66 1.3

DC-4bT 41.30 55316 2 1406 63 11346 512 145.56 1.3 0.0017436 0.000097 157.86 9.2 145.56 1.3B 40.30 48936 3 1096 58 12826 689 144.06 1.5 0.0017246 0.000096 156.26 9.1 144.06 1.5

85080 DC-3aT 135.71 37296 2 666 17 3736 103 145.06 1.2 0.0003996 0.000037 33.06 3.5 145.06 1.2B 58.60 37796 2 22936 42 1546 5 145.46 1.4 0.0056726 0.000150 535.56 14.3 145.76 1.4D 0.81 211046 36 1.383 105 6 3 3 103 90.36 21.1 108.16 36 0.0357056 0.008356

DC-3bT 78.06 35206 2 216 30 8976 1303 146.06 2.2 0.0003236 0.000068 25.86 6.5 146.06 2.2B 94.52 32086 2 1596 26 3166 63 146.66 1.4 0.0008896 0.000114 79.56 10.8 146.66 1.4

48740 DC-5T 95.62 36746 2 9716 37 916 4 146.46 1.5 0.0014566 0.000093 117.16 8.8 146.46 1.5B 28.89 37696 4 21966 102 696 5 142.86 3.2 0.0024256 0.000183 210.06 17.0 142.96 3.2

78459 DC-1T(I) 354.60 37116 2 14386 67 1226 9 146.36 1.2 0.0028716 0.000150 265.36 14.2 146.46 1.2T(II) 37106 2 13746 24 1326 4 145.76 1.1 0.0029676 0.000064 274.66 6.1 145.86 1.1

B 143.30 35166 2 11126 35 1286 5 156.76 1.2 0.0024456 0.000048 224.66 4.7 146.86 1.2D 13.00 38156 221 43 105 6 3 3 103 126 1 1436 126 0.077756 0.00701

Uranium series data of modernD. cristagalli used in this study. Samples labeled “a” and “b” are different septa from the same individual. T(I) andT(II) are different aliquots of the same sample after it was dissolved. T, M and B refer to the Top, Middle and Bottom of septa respectively as shownin Figure 3. (D)—organic material scrubbed off the surface of the coral. Half-lives of234U and230Th used in the calculations are 245,2506 490 yrand 75,6906 230 yr respectively (Cheng et al., 2000). The decay constant for238U is 1.5513 10210 yr21 (Jaffey et al., 1971). Ages are calculatedaccording to equation 1 with no correction for initial230Th. All errors are 2s. b Age before collection.

2406 Cheng et al.

Tab

le3.

Uan

dT

his

otop

icco

mpo

sitio

nan

d2

30 T

hag

esof

old

deep

sea

cora

ls.

Sam

ple

num

ber

Ana

lysi

snu

mbe

r

23

8 U(p

pb)

23

2 Th

(ppt

)

23

0 Th/

23

2 Th

(ato

mic

310

26)

d23

4 U(m

easu

red)

23

0 Th/

23

8 U(a

ctiv

ity)

Age

(BP

year†

)d2

34 U

(initi

al)

Cor

.ag

e(B

Pye

ar1)

Cor

. d2

34 U

(initi

al)

JFA

22

3752

62

1723

626

4936

678

141.

361.

50.

1373

160.

0006

813

9626

7514

7.06

1.6

1373

6622

314

6.96

1.6

2(N

)(I

)40

596

242

176

4921

956

2814

2.56

1.3

0.13

8426

0.00

076

1406

561.

314

8.36

1.3

1355

6648

814

8.16

1.4

2(N

)(I

I)41

966

267

496

4314

216

1614

1.56

1.4

0.13

8146

0.00

124

1404

6613

514

7.26

1.5

1325

8676

014

6.96

1.5

2(N

)(D

)41

146

475

7300

693

5010

262

143.

263.

01.

1357

60.

0172

JFA

1717

3851

61

877

622

9372

623

214

3.66

1.3

0.12

9236

0.00

057

1306

3663

149.

061.

312

9516

120

149.

061.

3JF

A24

C24

(I)

3740

63

2901

626

3317

631

144.

562.

80.

1558

160.

0005

815

9326

7315

1.26

2.9

1555

2636

315

1.06

2.9

24(I

I)38

586

132

556

3630

206

3514

1.36

1.3

0.15

4356

0.00

059

1582

0668

147.

861.

415

4066

396

147.

661.

424

(N)

(I)

3293

62

7960

653

1084

69

144.

862.

10.

1587

060.

0006

416

2456

7715

1.66

2.2

1505

8611

3115

1.16

2.2

24(N

)(I

I)35

216

519

2316

176

5026

514

5.26

4.1

0.16

6126

0.00

092

1705

8612

115

2.46

4.4

1435

9626

0715

1.26

4.4

24(N

)(D

)29

416

4045

1100

0637

800

486

120

6.36

794.

4194

60.

0687

JFA

20A

20A

4524

62

1623

624

2347

0634

311

6.66

1.2

0.50

9996

0.00

133

6578

7625

414

0.46

1.5

6561

0623

914

0.46

1.4

20A

(N)

4935

63

8472

648

4943

637

119.

761.

30.

5138

860.

0024

866

2006

445

144.

361.

665

3536

863

144.

061.

6JF

A20

B20

B45

076

148

096

4368

346

6411

9.76

1.2

0.44

1606

0.00

151

5420

9624

813

9.56

1.4

5368

2653

213

9.36

1.4

20B

(N)

4473

62

6555

635

5045

629

120.

961.

30.

4478

160.

0013

555

1096

229

141.

361.

554

3866

706

141.

061.

5JF

A20

C20

C31

146

285

66

1730

9206

612

128.

861.

00.

5148

560.

0016

665

5896

296

155.

061.

365

4556

246

155.

061.

220

C(N

)(I

)31

096

121

086

3612

5016

213

142.

761.

40.

5133

460.

0014

464

1806

266

171.

161.

763

8546

356

170.

961.

720

C(N

)(I

I)33

196

2530

436

6990

866

212

140.

368.

70.

5046

160.

0046

262

9146

1025

167.

6610

.262

4716

708

167.

4610

.4

Ura

nium

serie

sda

taof

foss

ilco

rals

anal

yzed

inth

isst

udy.

Sam

ples

labe

led

with

an(N

)w

ere

ultr

ason

ical

lycl

eane

d.A

llot

her

sam

ples

wer

ech

emic

ally

clea

ned

asde

scrib

edin

the

text

.(I

)an

d(I

I)ar

ese

para

tepi

eces

ofth

esa

me

sam

ple.

(D)—

detr

italm

ater

ials

crub

bed

off

the

surf

ace

ofth

eco

ral.

All

erro

rsar

e2

s.

†A

gebe

fore

1950

.1C

orre

cted

usin

gan

initi

al23

0 Th/

23

2 Th

ratio

of85

680

310

26.

Err

ors

are

due

only

toth

eun

cert

aint

yin

the

initi

alra

tio.

See

text

.

2407U-Th dating of deep-sea corals

(Taylor and McLennan, 1985). Thus, for deep-sea corals, onemight expect (230Th/232Th)° to be a mixture between these twoend-members.

Our modern deep-sea corals have large variations in their[232Th] and 230Th/232Th ratios (Table 2). The lowest232Thconcentrations (47413, 84820, 85080) are similar to surfacecorals (several tens or a few hundreds of ppt, (Edwards, 1988;Gallup et al., 1994)). These samples also have the highest230Th/232Th ratios, suggesting that the aragonite skeletonsthemselves have relatively low232Th contents (generally in thelow 100’s of ppt). Supporting this idea is the fact that the fourmodern sub-samples with concentrations significantly higherthan 1000 ppt all contained visible remnants of yellowish,presumably organic material, on their exterior surfaces afterultrasonic cleaning. Direct measurement of this yellowish ma-terial (Table 2, analyses DC-1-D and DC-3a-D) yielded232Thconcentrations two to four orders of magnitude higher than thebulk samples.

3.2.3. Development Diagrams

Determination of initial230Th/232Th is not as straightforwardas making an accurate measurement on a “modern sample”. Ifthe coral was alive at the time of collection, we know the ageof the youngest portion of the skeleton. However, each sub-sample has a finite mass and therefore contains material thatgrew over some interval of time potentially ranging from thetime of collection back to the initiation of growth. A priori, theamount of radiogenic230Th that formed during the sampleinterval is not known. To calculate an initial230Th/232Th ratiofrom Eqn. 2 one needs to know the age (or more exactly, themean age weighted by mass) of the sample. So, for eachsub-sample, we have one equation (Eqn. 2) and two unknowns(initial 230Th/232Th and age).

The classic approach to solving this type of problem is

through isochron methods, in which the sub-samples havedifferent parent-daughter ratios, but the same age and initialdaughter isotope composition. We use a variation of this ap-proach because our assumptions do not always fit those used inthe isochron method and the stratigraphy of samples from thesame septum are an added constraint on the age. Instead, weuse development diagrams (Fig. 6, see Faure, 1986 or DePaolo,1981 for analogous examples with other isotope systems), inwhich we plot (230Th/232Th)° versus age for each sub-sample.Assuming closed system behavior, each sub-sample is repre-sented by a curve (in age vs. (230Th/232Th)° space) which is thesolution to Eqn. 2. For modern samples, this curve closelyapproximates a straight line over the time scale of interest. Thiscan be demonstrated by taking the zero and first order terms ofa MacLaurin Expansion of Eqn. 2 (about t), assuming thatdecay the of initial230Th and the ingrowth of234U are insig-nificant, and simplifying:

S 230Th232ThD

m

< S 238U232ThD

m

S 234U238UD

m

l234t 1 S 230Th232ThD

0

(3)

Here all isotope ratios are atomic ratios, the subscript “m”refers to the measured value, and the superscript “°” refers to aninitial value. Because the measured230Th/232Th value is fixed,older ages correspond to smaller initial230Th/232Th ratios, andvise-vera. Pairs of ages and initial ratios lie along a line forwhich we know the slope and a single point. We can transformthe abscissa of this line such that the zero time reference pointis the measurement date rather than the initiation of growth. Inthis case, we are extrapolating back from measured230Th/232Thratios instead of extrapolating forward from initial230Th/232Thratios. Here the variable t9, the new x-axis, is equal to2(t 1sample age). Eqn. 3 then becomes:

S 230Th232ThD

0

< 2 S 238U232ThD

m

S 234U238UD

m

l234t9 1 S 230Th232ThD

m

(4)

By assuming that the initial230Th/232Th ratios in all sub-samples of the same coral have the same value, we can use thisequation to constrain the initial230Th/232Th value (Fig. 6). Ifthe sub-samples have the same stratigraphic position, the ageand initial230Th/232Th value are represented by the intersectionof the lines of the sub-samples. If the stratigraphic sequence ofages is known, the initial230Th/232Th value must fall within arange that gives a sequence of ages which agrees with theknown stratigraphic order. No values of initial230Th/232Th thatgive ages younger than the time of collection are possible.

The development diagram for septum 47413a is shown inFigure 6. Data, including the 2s error bars, for the bottomsub-sample of 47413a describes the dark gray wedge. Onlyvalues of initial230Th/232Th that lie within the wedge and to theright of the time of collection are possible. Similarly, the lightgray area describes the range of acceptable pairs of ages andinitial 230Th/232Th ratios for the top of 47413a. The sides of thelight gray area are curved because of measurement error prop-agation through the various isotopic ratios in Eqn. 4. Due to thestratigraphic order of the top and bottom pieces, the pointmarked “A” is the first stratigraphically acceptable initial ratio.Values higher than this intersection (younger ages) make thedark gray wedge younger than the light gray area. However,because point “A” is younger than the time of collection, the

Fig. 5. Plot of the meand234U values of modern deep-sea coralsversus depth. Number of measurements per point are in parentheses.Error bars show 2s analysis errors. The top dark gray rectanglecorresponds to the meand234U value of 20 modern surface corals fromthe Minnesota Isotope Lab (Cheng et al., 1999) . The long light grayrectangle represents the mean and 2s error for thed234U values of allfive modern individuals (20 total measurements).

2408 Cheng et al.

largest stratigraphically acceptable atomic (230Th/232Th)° valueis really 1473 1026. All values between 0 and 1473 1026

give ages for 47413a sub-samples that agree with stratigraphicconstraints and do not give negative ages for the septum. Oneof the two middle pieces from this septum is shown for com-pleteness (black lines), but does not further constrain the sys-tem. In this manner, each coral septum in Table 2 can beanalyzed for both the minimum and maximum allowable initial230Th/232Th ratios.

Due to lack of space, we do not show all data in graphicalform. Instead, we summarize the results of this analysis inTable 4 and Figure 7. Outside of one coral, all samples haveinitial isotopic compositions that fall between 0 and 16031026 (the black dashed line in Fig. 7). Sample 84820 gives avery large range of possible initial230Th/232Th values andcorrespondingly old ages. This sample was probably alreadydead when dredged from the ocean bottom and the largeisotopic ratio is due to radiogenic230Th rather than a highinitial ratio (see below).

3.3. Growth Rate and Band Periodicity fromMeasurements on Modern Samples

In order to use individualD. cristagalli specimens as ar-chives of oceanographic time series, we need to constrain their

growth rate and band periodicity. Given the constraints oninitial 230Th/232Th deduced in the previous section, we can usethe age equation and the mass weighted mean of our isotopedata to calculate the mean septal age for each sample (Table 4).Ages for all sub-samples are less than 150 yr., and, for the mostpart, in the range of tens of years or even less. Furthermore, theage differences between stratigraphically older and youngersub-samples of the same septa (Table 2) are generally less than20 yr. For four of the five specimens, the youngest sub-sampleof each septum has an uncorrected age within error of 0 to 20yr., and a whole septum mean age less than 80 yr. However, forboth septa analyzed from sample number 84820, the ages of theyoungest sub-samples are much greater than 20 yr. (1506 7and 1586 9 yr., uncorrected for initial230Th). Either this coralgrew more slowly than the others, or it had a very high230Th/232Th initial ratio, or it died prior to collection. Becausethe difference in age between the top and bottom portions issimilar to the same difference in the other 4 specimens, weconclude that coral 84820 died long before it was collected. Inthis case we can not constrain the (230Th/232Th)° or the meanage of the sample so we do not discuss the calculated growthrate for 84820.

For the other samples, we calculate growth rates from mean

Fig. 6. Development diagram for septum A of the modernD. cristagalli sample 47413. Top, middle and bottom piecesof this septum are represented by wedges of light gray, black lines and dark gray respectively. Boundaries for the “wedges”are calculated by propagating errors through Eqn. 4. Sample ages and initial230Th/ 232Th ratios are constrained by theintersection of the wedges coupled with stratigraphic constraints. See text for a full discussion. The y-intercept is themeasured thorium isotopic composition and the x-intercept is the age if the initial230Th/232Th ratio were zero, the maximumpossible age. The slope is a function of the238U/232Th ratio, the higher the value, the steeper the slope.

2409U-Th dating of deep-sea corals

septal ages by assuming that the rate of mass accumulation isconstant with time and that the coral died when it was collected.Given these assumptions, twice the mean age divided by the

septum’s length is the average growth rate. The mass-weightedmean age for a septum is calculated by summing the individualmass-adjusted238U, 234U, 230Th, and232Th concentrations ofthe sub-samples and then recalculating the age from equation 4(see Table 4 caption for which (230Th/232Th)° values wereused). Growth rates range from 0.1–3.1 mm/yr. The youngestsample, 47413, has the largest error bars because its relativeage errors are large. For sample 85080a, only the top portion ofseptum a was used in Table 4. In the bottom portion of thisseptum, the232Th concentration and the corresponding initial230Th are so large that the age is not very well constrained.However, this bottom piece does place strong constraints on theupper value of the initial230Th/232Th ratio. To account forusing only the top piece to calculate the mean age, we adjustedthe total length and number of bands accordingly. On averagethe growth rate forD. cristagalli seems to be about 1 mm/yr,but there is evidence that separate specimens can have differentrates. The growth rates measured here indicate that deep-seacorals may be used as recorders of deep-sea changes at annualto centennial time scales (Smith et al., 1997). Adkins et al.(1998) used this range of growth rates to constrain the life spanof individual D. Cristagalli corals that recorded a deep-oceancirculation change at 15.4 ka to be under 160 yr. While thefractional errors on our calculated growth rates are large, theminimum rates, and therefore maximum ages of individualcorals, are well constrained. Measured230Th concentrations aresmall in our modern data set and can not lead to growth ratesslower than our minimum estimates where there is no initial230Th assumed.

Growth band frequency can also be calculated by dividingthe total number of bands in a septum by 2 times the mean age(Table 4). Band counts were estimated using a transmited lightmicroscope. For the four corals that were alive when collected,band frequency is between 0.3 and 3.0 bands/year. Taken atface value, 2 of the 4 samples are not consistent with annualbanding. However, given uncertainties in both mass accumu-lation rate and especially band counting, we cannot rule outannual banding. At this point, we can say that banding fre-quency is within a factor of a few of 1 pair/yr. Given the water

Fig. 7. Plot of calculated initial230Th/ 232Th ratio vs. sample number.Upper and lower bounds (gray and black circles respectively) for eachseptum were calculated using development diagrams (see Fig. 6 andtext). Nearby water column values (open squares) and measured detritalvalues (open circles) are shown where data exist. Sample 85080 has ameasured detrital component that lies in the middle of the estimatedrange of initial ratios. The range for the deep North Atlantic sample78459 is more restricted because initial230Th/232Th ratios below 5031026 violate the stratigraphic order. Reported isotopic compositions ofthorium in North Atlantic filtered sea water range from 50 to 15031026 (Moran et al., 1997 and Hoff et al., unpublished data). Sample78459’s range of initial230Th/232Th values is consistent with theseseawater data. Because the top of the coral must be younger than thebottom of the sample, sample 48740’s initial230Th/232Th ratio must beless than 703 1026. Water values from the area around the anomaloussample 84820 are much lower than our estimated range, lending sup-port to the argument that this sample died before collection. In general,where data to make such comparisons exist, the initial thorium isotoperatios of the aragonitic portion of the specimens are consistent with seawater and detrital values. The overall range of possible initial230Th/232Th ratios is between 0 and 1603 1026.

Table 4. Summary of evolution isochron data, growth rates and band periodicity.

Samplenumber

Analysisnumber

230/232

Mean septal Agea

Length(mm)

Growth Rate

Bands# err

Periodicity

Error Error Error

min max Years 1 2 mm/yr 1 2 bands/yr 1 2

47413 DC-2a 0 147 7.5 6 5 41 2.7 4.8 1.3 40 8 2.7 4.7 1.3DC-2b 0 994 6.6 39 6 41 3.1 84 2.7 40 8 3.0 82 2.7

85080 DC-3ab 0 155 2.5 10 10 11b 0.4 0.3 0.1 17b 10 0.7 0.6 0.4DC-3b 0 155 41 7 6 33 0.4 0.1 0.1 50 10 0.6 0.2 0.2

48740 DC-5 0 67 72 9 9 19 0.1 0.02 0.01 — — — — —78459 DC-1 50 130 53 10 10 28 0.3 0.06 0.04 35 8 0.3 0.1 0.1- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -84820 DC-4a 183 1375 129 10 7 25 0.1 0.01 0.01 30 6 0.1 0.02 0.02

DC-4b 0 2060 146 19 9 25 0.1 0.01 0.01 30 6 0.1 0.02 0.02

Summary of initial230Th/232Th ratio results from the development diagrams (see Fig. 6). Mean septal ages are calculated using a weighted averageof all U-series data from a single septum. Growth rates/periodicities are calculated by dividing the length/bands by twice the mean septal age.a Agesuse an initial atomic230Th/232Th ratio of 806 80 3 1026 except for 48740 and 78459 where the ratio is bettter constrained by the developmentdiagram data. Ages are also corrected for time since collection.b Includes data from only the top piece. Length and bands are adjusted accordingly.

2410 Cheng et al.

depths of coral growth, banding cannot be controlled by sea-sonal variations in sunlight, temperature or other climate vari-ables. The banding pattern may be governed by endogenousphysiological rhythms (Emiliani et al., 1978) or by the rainingfood supply from above (Deuser et al., 1981).

3.4. Fossil Samples

3.4.1. Sources of232Th and diagenetic Uranium

Ultimately we would like to use data from modern samplesto correct fossil coral ages for initial and added thorium. Fossilsamples are clearly enriched in232Th relative to modern sam-ples (compare Tables 2 and 3). Largely this enrichment is dueto the black crusts that coat older corals and are elevated in232Th by about 104 times over modern samples (labeled “D” inTable 3). As these crusts also contain230Th associated with the232Th, they represent a clear source of contamination to U-series ages. Our strategy has been to use chemical cleaningtechniques to remove a significant fraction of the added tho-rium. Then, using both Eqn. 2 and isochrons, we try andaccount for residual added as well as initial thorium to calculatean age. Chemically cleaned fossil corals have232Th concentra-tions ranging from 856 to 4809 ppt, about 30–80% lower thantheir untreated counterparts. As lower232Th concentrationsimply smaller age errors in Equation 2, and a larger spread inisotopic ratios for the isochrons (where dirtier samples are alsobeing measured), the chemical leaches are an important part ofobtaining precise and accurate ages on deep-sea corals. Withoutremoving the added thorium in black crusts, we could not placeany reasonable constraints on fossil coral ages (Lomitschka andMangini, 1999). In the sections that follow, we will show thatthis cleaning does not significantly bias the calculated ages.

The elevated232Th in fossil corals raises the issue of opensystem behavior for uranium. Concentrations for the fossilsub-samples are similar to the range in modern corals, indicat-ing that none of the fossil corals have undergone large netdiagenetic gains or losses of uranium. In contrast to the situa-tion with 232Th, the two Fe-Mn rich crusts have uraniumconcentrations comparable to the coral values. Thus, any re-sidual crust remaining after either the ultrasonic or chemicalcleaning procedure is not as likely to contribute a significantfraction of uranium to the sample.

Uranium isotopic values for our fossil data set fall into twocategories. The eight sub-samples of coral that have ages be-tween 10 and 20 kyr also have initiald234U values between 146and 152‰, equal to or within several per mil of the modernmarine value (145.86 1.9). It is known from analysis of sur-face corals that thed234U of the surface ocean between 10 and13 ky ago was indistinguishable from the modern value (Ed-wards et al., 1993). Thus, this slight elevation ind234U abovethe modern value is quite likely due to small amounts ofdiagenetic exchange of uranium. The seven sub-samples ofcoral in the older range also have initiald234U values close tothe modern value, but have a larger spread: 139 to 171‰. In atleast one case (JFA 20C), we can resolve thed234U values oftwo sub-samples of the same coral. While the value of marined234U in this time interval is unknown, models suggest that itis not likely to differ by large amounts from modern values(Edwards, 1988; Gallup and Edwards, 1997; Richter and

Turekian, 1993), nor is it likely to have shifted significantlyover time scales of several thousand years or less. Thus, therange ind234U, the fact that most of the values are distinguish-able from the modern value, and the differingd234U withinspecimen JFA 20C all suggest diagenetic exchange of uranium.The degree of exchange is more extensive in the older coralsthan in the younger samples. Although the source of the diage-netic uranium is unknown, sea water and pore fluid uraniumcould plausibly provide uranium of the appropriate isotopiccomposition to cause these effects. In the following sections,we investigate and compare two calculation methods that cor-rect for these potential problems to U-series dating in fossildeep-sea corals.

3.4.2. Correcting fossil coral ages: Equation 2

In section 3.2.2 we established that all samples in our mod-ern data set have an initial atomic230Th/232Th ratio lower than160 3 1026 (Table 4, Fig. 7). Using Eqn. 2 and a value of806 803 1026 for (230Th/232Th)°, we can calculate a conser-vative estimate for the age range of a fossil sample. However,as discussed above, initial thorium is not the only component ofnon-radiogenic thorium in our fossil corals. If we assume thatthe added thorium has the same isotopic composition as theinitial thorium, within the broad bounds that we have set, andthat the thorium was added soon after coral growth, then Eqn. 2is a solution to the age equation. The actual timing of thoriumaddition, whether episodic or continuous, early or late is notcritical for corals significantly younger than the half-life of230Th (75 ky), but becomes an issue for samples older than agood fraction of a230Th half-life.

Given the measured modern initial Th isotopic range, we canevaluate the sensitivity of calculated age errors to the amount ofmeasured232Th. Figure 8is a plot of error in age versus age fora (230Th/232Th)° value equal to 806 803 1026 and a uraniumconcentration of 4 ppm. The thick black line represents typicalanalytical errors as a function of age. The solid circles are theage and 2s error in age for all sub-samples, uncorrected fornon-radiogenic thorium. The open circles represent the sameanalyses corrected for non-radiogenic230Th using Eqn. 2 andthe 806 80 value. For these, uncertainties are a combination ofanalytical error and the assumed range in the (230Th/232Th)°value. As the open circles plot above the solid circles, in somecases by more than an order of magnitude, the dominant sourceof error is clearly not analytical but the range in intial thoriumisotopic composition. Thin black lines are the calculated totalerror in age as a function of age, contoured in232Th concen-tration. Where these lines are horizontal, age errors are domi-nated by the (230Th/232Th)° ratio. The thick dashed line marksthe region where analytical errors begin to contribute to the ageerror as the influence of the error on the (230Th/232Th)° ratiofalls away. These contours illustrate quantitatively the pointmade above, that all other factors being equal, the lower the232Th concentration, the lower the error in age. For paired230Th-14C studies aimed at determining the ventilation age ofpast water masses, one would like to determine230Th age to6102 years or better (Adkins et al., 1998; Mangini et al., 1998).Figure 8 shows that to achieve this, one must clean deglacialage corals to the level of about 1000 ppt. Table 3 indicates thatour cleaning procedure has achieved these levels in some cases,

2411U-Th dating of deep-sea corals

but it is clear that improvements to remove more exteriorthorium would further reduce the errors.

3.4.3. Correcting fossil coral ages: Isochrons

In addition to the method of correcting for initial230Th usingEqn. 2 and our estimate of initial230Th/232Th, we have alsocalculated ages by applying isochron methods to differentfragments of the same coral. Isochrons, in essence, allow one tocalculate a radiogenic end member isotopic composition from aset of two or more materials that lie on a two-componentmixing line between radiogenic and non-radiogenic compo-nents. Given the isotopic composition of the radiogenic com-ponent, one can calculate an age. Inherent in the calculation isthe assumption that there are only two end members, an as-sumption that deep sea corals potentially violate. For example,each coral is deposited over a finite interval of time so that theradiogenic component of the oldest part of the coral may havea different isotopic composition than the radiogenic componentof the youngest part of the coral. Because the lifetime of thecoral (typically tens to 200 yr based on our growth rate deter-minations) is similar to our analytical error in age (for coralsolder than about 10 ka), this is not a significant problem forcorals older than about 10 ka. Also, non-radiogenic compo-nents are added at different times (as evidenced by fossil corals

with generally higher232Th concentrations than modern cor-als). Earlier generations of non-radiogenic component mayhave changed isotopic composition by radioactive decay andingrowth by the time subsequent generations of non-radiogeniccomponent are added, again violating the two-component as-sumption. Isochron methods do not require that the non-radio-genic component be present initially, but do require that thenon-radiogenic component is introduced at one instant in time.For samples significantly younger than the half-life of230Th,the addition of a radiogenic component over a period of timeshould not introduce significant inaccuracy. Therefore the tim-ing of addition should not be an issue for our 10 to 20 ka oldsamples. A third potential problem is the fact that there are atleast two possible non-radiogenic thorium components, thehydrogenous and detrital components discussed above. Afourth potential problem relates to the issue of the cleaning ofsome of our samples by chemical (as opposed to physical)techniques. This set of procedures could potentially shift U/Thratios either by preferential leaching from the solid or bypreferential adsorption onto the solid. To a certain degree, wecan test for all of these potential problems with isochrons. Ifthere are three components present in different proportions indifferent sub-samples, the isotopic composition of the sub-samples would likely deviate from a line in an isochron plot. Ifcleaning techniques shift U/Th ratios, chemically treated sam-ples should also deviate from a line defined by untreatedsamples.

Results of isochron calculations (Ludwig, 1993; Ludwig andTitterington, 1994) are shown in Table 5 and Figure 9. For allthree corals with enough data points, the MSWD statisticexceeds one. As a measure of the linear fit of each isochron,this result implies that one or more of the assumptions outlinedabove is not valid. However, this statistic is sensitive to theerrors assigned to each measurement. Doubling the232Th un-certainties reduces the MSWD to near or below one for eachsample. As232Th errors are difficult to evaluate due to theuncertainty in the filament blank, underestimating this value isthe likely reason for elevated MSWDs on our regressions.Alternately, the slightly elevated MSWDs may reflect slightdeviations from linearity due to natural processes or chemicaltreatment. If so, these deviations are on the order of analyticalerror, suggesting that the isochron approach is valid at about thelevel of our analytical precision. Additionally, chemicallytreated and untreated sub-samples are co-linear at about thelevel of our analytical precision, suggesting that chemicalcleaning has not shifted U/Th ratios significantly.

Corals in Table 3 fall into two age ranges, 10–20 ka and50–70 ka. Assuming that the non-radiogenic component wasadded early for the younger samples JFA-2 and JFA 24C, wecan calculate initial230Th/232Th ratios for both corals from theintercepts in Figure 9. For each sample these values (193 1026

for JFA-2 and 373 1026 for JFA-24) fall within the 806 8031026 range that we established with modern corals. Thus, forthese younger samples, both the isochron and initial230Th/232Th methods appear to be valid. In each case, it is importantthat at least one sub-sample have low232Th concentration toget a reasonable age estimate. In practice, this requires the useof some sort of chemical cleaning technique to remove surficialcontaminants.

For the three samples (JFA-20A, JFA-20B, and JFA-20C)

Fig. 8. Age plotted versus its 2s error. Solid points are the uncor-rected deep-sea coral data and are close to the typical analyticaluncertainty (solid black line). After correction (open circles) using Eqn.2, points are shifted to younger ages and larger errors. Each row inTables 2 and 3 represents one pair of points, one corrected (open) andone uncorrected (solid). The relation between measured [232Th] and theerror in age for an assumed initial230Th/232Th ratio of 806 803 1026

is given by the thin lines. Higher232Th results in a larger absolute ageuncertainty. For young ages, the error is dominated by uncertainty inthe initial Th correction. The older the age and the smaller the measured[232Th] the lower the relative age error. In the region between thedashed and thick black lines, age uncertainty is jointly determined byanalytical and initial Th errors. It is clear that further cleaning of232Thfrom older corals can significantly lower age uncertainties.

2412 Cheng et al.

between 50 and 70 ka old, we have also calculated ages usingboth correction schemes (Tables 3 and 5). Corrected and un-corrected ages for each sample differ by no more than a fewthousand years, suggesting that the ages are accurate to aboutthis level. However, there are some consistencies and someinconsistencies in the details. Sample JFA 20A gives consistentcorrected ages (Table 3). Sample JFA 20B yields an isochronage that is slightly younger than the ages corrected using the80 6 80 3 1026 initial value, while sample JFA 20C gives anisochron age that is older. However, the initial230Th/232Thratio calculated for this sample, assuming that the non-radio-genic component was added early, is negative (Fig. 9). Thisobservation is consistent with a non-radiogenic component that

had a230Th/238U ratio significantly less than that of the primarycoralline aragonite at the time of addition; and that was addedsignificantly after original precipitation of the primary corallinearagonite. Thus, it appears that both uranium and thorium couldhave been added to this sample. The uranium isotopic compo-sition of the sub-samples supports this idea. The sub-sampleswith the highest232Th concentrations have initiald234U valuesthat are furthest from the marine value. The correction forinitial 230Th using the 806 80 3 1026 value is not designedto correct for uranium addition and therefore yields an inaccu-rate age in this case. The isochron method, on the other hand,may well yield an accurate age. In addition to co-linearity of thesub-sample points, the initiald234U value calculated using the

Table 5. Results of the isochron calculations for fossil deep sea corals.

Samplenumber

Datapoints

Corrected(230Th/238U)act

Corrected(234U/238U)act

Correctedage (year)

95% Conf.error (yr)a

Correctedd234U(T)

MSWDvalue#

JFA-2 4 0.133226 0.00048 1.14186 0.00084 135106 50 6280 147.36 1.0 15.1JFA-24C 5 0.152416 0.00033 1.14286 0.00103 155806 40 6130 149.26 1.1 7.39JFA-20A 2 0.508966 0.00178 1.11586 0.00159 686806 340 139.46 1.9JFA-20B 2 0.424996 0.00647 1.11636 0.00549 518306 1100 134.66 6.4JFA-20C 3 0.517186 0.00270 1.11956 0.00212 668006 500 62200 144.36 2.6 4.39

Results of the isochron calculations for fossil deep-sea corals. Calculations are based on “Rosholt Type-II diagrams” (Ludwig and Titterington,1994) and were performed using the UISO program (Ludwig, 1993).¶ Age before 1950.# Mean Square of Weighted Deviates of data from theregression line.

Fig. 9. Isochrons for three fossil deep-sea corals. These plots are half of a traditional Rosholt “Type II” isochron diagram.Calculation of isochrons and error ellipses follow the method of (Ludwig and Titterington, 1994) using the program of(Ludwig, 1993). Normally isochron plots use activity ratios, however we use atomic ratios to emphasize the implied initial230Th/ 232Th represented by the y-intercepts. The lowest points in JFA 2 and JFA 24C are from detrital pieces scraped fromthe coral exterior. Gray scale points at the high end of each plot were chemically cleaned.

2413U-Th dating of deep-sea corals

isochron method is indistinguishable from the marine value. Italso appears that our chemical cleaning method removed mostof the added uranium as well as most of the added thorium. Thetwo sub-samples cleaned only by ultrasonic methods, have highd234U (and high232Th), whereas the sub-sample that was alsochemically cleaned, has ad234U much closer to the modernvalue (and low232Th).

In sum, both isochron methods and methods that use modern230Th/232Th values to correct for initial230Th give consistentresults for samples in the 10 to 20 ka age range. Samples in the50 to 70 ka age range give consistent results within a fewthousand years, but in at least one case show clear evidence fordiagenetic addition of both uranium and thorium. In this case,the isochron method appears to have corrected for this diage-netic component. For all fossil samples, chemical cleaningmethods remove added thorium and uranium. This cleaningstep is important in reducing errors in U-Th dating of fossilcorals. Further improvements in cleaning techniques wouldfurther reduce dating errors.

4. CONCLUSIONS

Uranium rich deep-sea corals are suitable for precise datingby the 238U-234U-230Th decay scheme. Measurements on asuite of modernD. cristagalli constrain the growth rate of thispseudo-colonial species to be between 0.1 and 3.1 mm/yr.Because of uncertainties in the amount of initial230Th, relativeerrors on growth rates appear large. However, maximum ages,and minimum growth rates, are well constrained because of therelatively low amounts of measured230Th. Combining agedeterminations with counts of density bands provides an esti-mate of mean band periodicity of between 0.3 and 3 bands/yr.d234U data from modern deep-sea corals indicate that thisspecies is conservative in the upper 2000 meters of the watercolumn. Because this sample set comes from a variety of openocean settings,d234U initial measurements in fossil corals canbe an indicator of diagenetic alteration. Age errors on fossilsamples exceed the analytical uncertainty because of the rangeof assumed initial230Th/232Th ratios, 806 80 3 1026. Re-ducing the amount of measured232Th, by using more rigorouscleaning techniques, is the most promising way to improve theprecision of fossil ages. Isochron ages for samples of deglacialage agree with our232Th corrected ages indicating that theproposed method properly accounts for all initial230Th. Olderspecimens with more diagenetic alteration, may require bettercleaning and/or isochron measurements to obtain accurate andprecise ages.

Acknowledgments—D. Richards and J. Doral provided stimulatingdiscussions about many of the issues raised in this work. S. Cairnshelped identify the coral samples. T. Kleindinst of WHOI produced thephotograph in Figure 4. We would like to thank G. Henderson and D.Muhs for helpful reviews of the manuscript. JFA thanks the UCARPost-Doctoral Fellowship Program and LDEO for support during thewriting of this paper.

REFERENCES

Adkins J. F. and Boyle E. A. (1997) Changing atmosphericD14C andthe record of paleo-ventilation ages.Paleoceanography12, 337–344.

Adkins J. F., Cheng H., Boyle E. A., Druffel E. R. M., and Edwards

R. L. (1998) Deep-sea coral evidence for rapid change in ventilationof the deep North Atlantic 15,400 years ago.Science280,725–728.

Anderson R. F., Bacon M. P., and Brewer P. G. (1983) Removal ofTh-230 and Pa-231 from the open ocean.Earth Planet. Sci. Lett.62,7–23.

Bacon M. P. and Anderson R. F. (1982) Distribution of Thoriumisotopes between dissolved and particulate forms in the deep sea.J.Geophys. Res.87, 2045–2056.

Bard E., Arnold M., Fairbanks R., and Hamelin B. (1993)230Th-234Uand14C ages obtained by mass spectrometry on corals.Radiocarbon35, 191–199.

Bard E., Hamelin B., Arnold M., Montaggioni L., Cabioch G., FaureG., and Rougerie F. (1996) Deglacial sea-level record from Tahiticorals and the timing of global meltwater discharge.Nature 382,241–244.

Bard E., Hamelin B., Fairbanks R. G., and Zindler A. (1990) Calibra-tion of the14C timescale over 30,000 years using mass spectrometricU-Th ages from Barbados corals.Nature345,405–410.

Barnes J. W., Lang E. J., and Potratz H. J. (1956) Ratio of Ionium toUranium in coral limestone.Science124,175–176.

Beck J. W., Recy J., Taylor F., Edwards R. L., and Cabioch G. (1997)Abrupt changes in early Holocene tropical sea surface temperaturedrived from coral records.Nature385,705–707.

Beck W. J., Edwards L. R., Ito E., Taylor F. W., Recy J., Rougerie F.,Joannot P., and Henin C. (1992) Sea-Surface temperature from coralskeletal strontium/calcium ratios.Science257,644–647.

Boyle E. A. (1981) Cadmium, zinc, copper, and barium in foraminiferatests.Earth Planet. Sci. Lett.53, 11–35.

Boyle E. A. and Keigwin L. D. (1985/6) Comparison of Atlantic andPacific paleochemical records for the last 250,000 years: Changes indeep ocean circulation and chemical inventories.Earth Planet. Sci.Lett. 76, 135–150.

Broecker W. S., Thurber D. L., Goddard J., Ku T.-L., Matthews R. K.,and Mesolella K. J. (1968) Milankovitch hypothesis supported byprecise dating of coral reefs and deep-sea sediments.Science159,297–300.

Burnett W. C. and Veeh H. H. (1992) Uranium-series studies of marinephosphates and carbonates. InUranium-series disequilibria, appli-cations to earth, marine and environmental sciences.(ed. M. Ivano-vich and R. S. Harmon). Oxford Science Publications.

Cairns S. D. (1981) Marine flora and fauna of the Northern UnitedStates. Scleractinia.NOAA Technical Report NMFS Circular438,1–15.

Cairns S. D. and Stanley G. D. Jr. (1981) Ahermatypic coral banks:Living and fossil counterparts.Proc. of the Fourth Intnl. Coral ReefSymp., Manila1, 611–618.

Chen J. H., Edwards R. L., and Wasserburg G. J. (1986)238U, 234U and232Th in seawater.Earth Planet. Sci. Lett.80, 241–251.

Cheng H., Adkins J. F., Edwards R. L., and Boyle E. A. (1995)230Thdating of deep-sea solitary corals.AGU, F292.

Cheng H., Edwards R. L., Hoff J., Gallup C. D., Richards D. A., andAsmerom Y. (2000) The half-lives of Uranium-234 and Thorium-230.Chem. Geol.in press.

Cochran J. K. (1992) The oceanic chemistry of the uranium- andthorium-series nuclides. InUranium-series disequilibrium, appli-cations to earth, marine and environmental sciences (2nd edition)(ed. M. Ivanovich and R. S. Harmon). Oxford Science Publica-tions.

Cochran J. K., Livingston H. D., Hirschberg D. J., and SurprenantL. D. (1987) Natural and anthropogenic radionuclide distributionsin the norhtwest Atlantic Ocean.Earth Planet. Sci. Lett.84,135–152.

Delaney M. L. and Boyle E. A. (1983) Uranium and thorium isotopeconcentrations in foraminiferal calcite.Earth Planet. Sci. Lett.62,258–262.

DePaolo D. J. (1981) Nd isotopic studies: Some new perspectives onearth structure and evolution.EOS62, 137–140.

Deuser W., Ross E. H., and Anderson R. F. (1981) Seasonality in thesupply of sediment to the deep Sargasso Sea and implications for therapid transfer of matter to the deep ocean.Deep-Sea Res.28, 495–505.

Druffel E. R. M., King L. L., Belastock R. A., and Buesseller K. O.

2414 Cheng et al.

(1990) Growth rate of a deep-sea coral using210Pb and other iso-topes.Geochim. Cosmochim. Acta54, 1493–1500.

Duncan P. M. (1877) On the rapidity of growth and variability of someMadreporaria on an Atlantic cable with remarks upon the rate ofaccumulation of foraminiferal deposits.Ann. Mag. Nat. Hist.20,361–365.

Edwards L. R., Beck W. J., Burr G. S., Donahue D. J., ChappellJ. M. A., Bloom A. L., Druffel E. R. M., and Taylor F. W. (1993) Alarge drop in atmospheric14C/12C and reduced melting in theYounger Dryas, documented with230Th ages of corals.Science260,962–967.

Edwards R. L. (1988) High-Precision Thorium-230 ages of corals andthe timing of sea level fluctuation in the late Quaternary. Ph.D.Thesis, California Institute of Technology.

Edwards R. L., Chen J. H., Ku T.-L., and Wasserburg G. J. (1987a)Precise timing of the last interglacial period from mass spectro-metric determination of Thorium-230 in corals.Science236,1547–1553.

Edwards R. L., Chen J. H., and Wasserburg G. J. (1987b)238U-234U-230Th-232Th systematics and the precise measurement of time overthe past 500,000 years.Earth Planet. Sci. Lett.81, 175–192.

Emiliani C., Hudson J. H., Shinn E., and George R. Y. (1978) Oxygenand carbon isotopic growth record in a reef coral from the FloridaKeys and a deep-sea coral from Blake Plateau.Science202, 627–629.

Faure G. (1986) Principles of isotope geology (2nd edition). Wiley andSons.

Gagan M. K., Ayliffe L. K., Hopley D., Cali J. A., Mortimer G. E.,Chappell J., McCulloch M. T., and Head M. J. (1998) Temperatureand surface-ocean water balance of the Mid-Holocene tropical West-ern Pacific.Science279,1014–1018.

Gallup C. D., Edwards L. R., and Johnson R. G. (1994) The timingof high sea levels over the past 200,000 years.Science263,796 – 800.

Gallup C. D. and Edwards R. L. (1997) Constraints from the uraniumisotopic composition of well-preserved corals on past changes in theflux or isotopic composition of riverine uranium.AGU Fall Meeting,San Francisco.

Goldstein S. J., Murrell M. T., Lea D., Chakraborty S., and KashgarianM. (1996)231Pa and230Th dating of deep-sea coral.EOS: Transac-tions, AGU, 1996 Spring Meeting, F291.

Grigg R. W. (1974) Growth rings: Annual periodicity in two Gorga-onian corals.Ecology55, 876–881.

Guilderson T. P., Fairbanks R. G., and Rubenstone J. L. (1994) Trop-ical temperature variations since 20,000 years ago: Modulating in-terhemispheric climate change.Science263,663–665.

Guo L., Santschi P., Baskaran M., and Zindler A. (1995) Distributionof dissolved and particulate230Th and232Th in seawater from theGulf of Mexico and off Cape Hatteras as measured by SIMS.EarthPlanet. Sci. Lett.133,117–128.

Gvirtzman G., Friedman G. M., and Miller D. S. (1973) Control anddistribution of Uranium in coral reefs during diagenesis.J. Sediment.Petrol. 43, 985–997.

Henderson G. M., Cohen A. S., and O’Nions R. K. (1993)234U/238Uratios and230Th ages for Hateruma Atoll corals: Implications forcoral diagenesis and seawater234U/238U ratios. Earth Planet. Sci.Lett. 115,65–73.

Henderson G. M. and O’Nions R. K. (1995)234U/238U ratios inQuaternary planktonic foraminifera.Geochim. Cosmochim. Acta59,4685–4694.

Huh C.-A. and Beasley T. M. (1987) Profiles of dissolved and partic-ulate thorium isotopes in the water column of coastal southernCalifornia.Earth Planet. Sci. Lett.85, 1–10.

Huh C.-A., Moore W. S., and Kadko D. C. (1989) Oceanic Th-232: Areconnaissance and implication of global distribution from manga-nese nodules.Geochim. Cosmochim. Acta53, 1357–1366.

Ivanovich M. and Harmon R. S. (1992) Uranium-series disequilibrium,2nd Edition. Oxford University Press.

Jaffey A. H., Flynn K. F., Glendenin L. E., Bently W. C., and EsslingA. M. (1971) Precision measurement of half-lives and specific ac-tivities of 235U and238U. Phys. Rev. C4, 1889–1906.

Ku T.-L. (1965) An evaluation of the234U/238U method as a tool fordating pelagic sediments.J. Geophys. Res.70, 3457–3474.

Ku T.-L. and Kern J. P. (1974) Uranium-series age of the upperPleistocene Nestor Terrace, San Diego, California.GSA Bulletin85,1713–1716.

Ku T. L., Knauss K. G., and Mathieu G. G. (1977) Uranium in the openocean: Concentration and isotopic composition.Deep-Sea Res.24,1005–1017.

Lin J. C., Broecker W. S., Anderson R. F., Hemming S., RubenstoneJ. L., and Bonani G. (1996) New230Th/U and14C ages from LakeLahontan carbonates, Nevada, USA, and a discussion of theorigin of initial thorium. Geochim. Cosmochim. Acta60, 2817–2832.

Lomitschka M. and Mangini A. (1999) Precise Th/U dating of smalland heavily coated samples of deep sea corals.Earth Planet. Sci.Lett. 170,391–401.

Ludwig K. R. (1993) UISO-A program for calculation of230Th-234U-238U isochrons.USGS Open File Report93-531.

Ludwig K. R. and Titterington D. M. (1994) Calculation of230Th/Uisochrons, ages and erors.Geochim. Cosmochim. Acta58, 5031–5042.

Mangini A., Lomitschka M., Eichstadter R., Frank N., Vogler S.,Bonani G., Hajdas I., and Patzold J. (1998) Coral provides way toage deep water.Nature392,347–348.

McCulloch M., Mortimer G., Esat T., Xianhua L., Pillans B., andChappell J. (1996) High resolution windows into early Holocenclimate: Sr/Ca coral records from the Huon Peninsula.Earth Planet.Sci. Lett.138,169–178.

Mesolella K. J., Matthews R. K., Broecker W. S., and Thurber D. L.(1969) The astronomical theory of climatic change: Barbados data.J.Geol.77, 250–274.

Min G. R., Edwards R. L., Taylor F. W., Recy J., Gallup C., and BeckJ. W. (1995) Annual cycles of U/Ca in coral skeletons and U/Cathermometry.Geochim. Cosmochim. Acta59, 2025–2042.

Moore W. S. (1981) The thorium isotope content of ocean water.EarthPlanet. Sci. Lett.53, 419–426.

Moran S. B., Charette M., Hoff J. A., and Edwards R. L. (1997)Distribution of Th-230 in the Labrador Sea and its relation toventilation.Earth Planet. Sci. Lett.150,151–160.

Moran S. B., Hoff J. A., Buesseler K. O., and Edwards R. L. (1995)High precision230Th and232Th in the Norwegian Sea and Denmarkby thermal ionization mass spectrometry.Geophys. Res. Lett.22,2589–2592.

Muhs D. R., Kennedy G. L., and Rockwell T. K. (1994) Uranium-seriesages of marine terrace corals from the Pacific coast of North Americaand implication for last-intergalcial sea level history.Quat. Res.42,72–87.

Nozaki Y. and Horibe Y. (1983) Alpha-emitting thorium isotopes innorthwest Pacific deep waters.Earth Planet. Sci. Lett.65, 39–50.

Nozaki Y., Yang H.-S., and Yamada M. (1987) Scavenging of thoriumin the ocean.J. Geophys. Res.92, 772–778.

Pratje O. (1924) Korallenbanke in tiefen und kuhlen Wasser.Cebtralblf Min. Geol., 410–415.

Richter F. M. and Turekian K. K. (1993) Simple models for thegeochemical response of the ocean to climatic and tectonic forcing.Earth Planet. Sci. Lett.119,121–131.

Roy-Barman M., Chen J. H., and Wasserburg G. J. (1996)230Th-232Thsystematics in the central Pacific Ocean: The sources and fates ofthorium.Earth Planet. Sci. Lett.139,351–363.

Russell A. D., Emerson S., Nelson B., Erez J., and Lea D. W. (1994)Uranium in foraminiferal calcite as a recorder of seawater uraniumconcentrations.Geochim. Cosmochim. Acta58, 671–681.

Shen G. T. and Boyle E. A. (1988) Determination of lead, cadmiumand other trace metals in anually-banded corals.Chem. Geol.67,47– 62.

Shen G. T. and Dunbar R. B. (1995) Environmental controls ofUranium in reef corals.Geochim. Cosmochim. Acta59, 2009 –2024.

Smith J. E., Risk M. J., Schwarcz H. P., and McConnaughey T. A.(1997) Rapid climate change in the North Atlantic during theYounger Dryas recorded by deep-sea corals.Nature 386, 818 –820.

2415U-Th dating of deep-sea corals

Stein M., Wasserberg G. J., Lajoie K. R., and Chen J. H. (1991)U-series ages of solitary corals from the Californian coast by massspectroscopy.Geochim. Cosmochim. Acta55, 3709–3722.

Stirling C. H., Esat T. M., McCulloch M. T., and Lambeck K. (1995)High precision U-series dating of corals from Western Australia andimplications for the timing and duration of the last interglacial.EarthPlanet. Sci. Lett.135,115–130.

Szabo B. J. (1985) Uranium-series dating of fossil corals from marinesediments of the United States Atlantic coastal plain.GSA Bulletin96, 398–406.

Szabo B. J., Ludwig K. R., Muhs D. R., and Simmons K. R. (1994)Thorium-230 ages of corals and duration of the last interglacialsea-level high stand on Oahu, Hawaii.Science266,93–96.

Taylor S. R. and McLennan S. M. (1985) The Continental Crust: ItsComposition and Evolution. Blackwell, Oxford.

Teichert C. (1958) Cold and deep-water coral banks.Bull. Am. Assoc.Petrol. Geol.42, 1064–1082.

Thompson G. and Livingston H. D. (1970) Strontium and uraniumconcentrations in aragonite precipitated by some modern corals.Earth Planet. Sci. Lett.8, 439–442.

2416 Cheng et al.