Sampling and Analytical Methodology:
Forty to sixty µg splits of drilled coral samples were analyzed for carbon and oxygen stable isotopes without chemical pretreatment or roasting. Samples were analyzed on a Thermo Delta V+ with dual inlet and Kiel IV device at the Lamont-Doherty Earth Observatory (LDEO), with values reported in per mil relative to the Vienna Pee-Dee Belemnite (VPDB). The 1σ analytical accuracy for δ¹³C and δ¹⁸O based on replicate analyses of laboratory standards are ±0.03‰ and ±0.06‰, respectively. When available, replicate analyses of individual coral samples were precise to within ±0.05‰ for δ¹³C and ±0.08‰ for δ¹⁸O (1σ).

For boron isotope analyses, 1 to 2 mg of calcite powder was cleaned of organic material with 1% H₂O₂ buffered in 0.1N NaOH at 80°C for twenty minutes, then rinsed five times with boron-free MilliQ water under ultrasonication. Recovery varied from 70 to 90%, as fine-grained material was invariably lost to suspension during rinse steps. Cleaned coral powders were dissolved in 2N HCl immediately prior to analysis; Fisher Optima® grade chemicals were used for all treatments. A sufficient volume of analyte to obtain a minimum of 1 ng of B (typically 1.5 to 2 µL depending on sample B concentration) was loaded with 1.0 µL of a boron-free seawater matrix solution onto degassed rhenium filaments.

Boron isotope ratios were measured on a Thermo TRITON multicollector thermal ionization mass spectrometer at LDEO in negative-ion mode (NTIMS, see detailed methodology in Foster et al., 2013). Average electric potentials for ¹¹BO₂^- ranged between 120 and 300 mV. Boron isotope ratios are reported in delta notation (δ¹¹B) relative to the NIST 951 boric acid standard reference material. Mass 26 (¹²C¹⁴N^-) was monitored immediately prior to analyses, to check for isobaric interference on ¹⁰BO₂^- from organic material. No samples were excluded based on this criterion, as mass 26 counts were below previously determined thresholds for organic matter contamination (Foster et al., 2013). Boron concentrations were determined in a representative sample for each coral by isotope dilution with 5 ppm NIST 952 boric acid reference material. A minimum of three acceptable repeat analyses was required for each reported δ¹¹B value, where analyses with >1‰ within-run fractionation were discarded. Although only two acceptable repeat analyses were obtained for five samples, these numbers are reported here because of their close correspondence with surrounding measurements (italicized δ¹¹B values in dataset).

Uncertainty in δ¹¹B measurements is reported as the larger value of either twice the measurement standard error (2se=2σ/√n, where n is the number of repeat analyses on a single sample solution), or the external error, which is given as the 2se on repeat analyses of an in-house standard of NIST 951 precipitated in vaterite (see Foster et al., 2013 and Penman et al., 2013).

BC_11B13C18O.csv