Astronomically calibrated ⁴⁰Ar/³⁹Ar age for the Toba supereruption and global synchronization of late Quaternary records

Volcanic ash is widespread in the Lenggong Valley and has been correlated to the 74-ka ignimbrite erupted from the Toba caldera, based on its biotite composition (12). At the archaeological site of Kota Tampan (16), located on an ancient terrace of the Perak River, in situ ash occurs among and above stone artifacts that may have been manufactured by anatomically modern humans (17). The ash is up to 5 m thick, irregularly distributed, mostly very fine grained [<100 μm (12)], and not suitable for ⁴⁰Ar/³⁹Ar dating. In 2011, however, drilling of a bore hole 6 km north of Kota Tampan revealed a crystal-rich, coarser facies to the ash, about 1.3 m thick and 5 m above the metamorphic basement rocks (Fig. 1). Sanidine crystals up to 2 mm in length were handpicked from a sample of this ash for ⁴⁰Ar/³⁹Ar dating experiments and were analyzed as 45 separate single- and multiple-grain aliquots using an NU Instruments Noblesse multi-collector noble-gas mass spectrometer. The enhanced performance characteristics of this instrument and the high signal-to-noise ratio of the ion-counting detectors lend themselves to increased accuracy and precision of ⁴⁰Ar/³⁹Ar ages for late Quaternary samples. Details are given in Materials and Methods and in SI Text.

The 45 aliquots gave model ⁴⁰Ar/³⁹Ar ages of between 67.0 and 167 ka, all but four of which fall in a narrow age range (71.1–78.7 ka), forming a Gaussian distribution in a probability plot (Fig. 2). The weighted mean of this main population is 74.0 ± 0.3 ka [1σ analytical errors, including neutron fluence monitor; mean square of the weighted deviates (mswd) = 1.92, probability of fit (prob.) = 0.0004, n = 41]. We applied an outlier-rejection scheme to the main population to discard ages with normalized median absolute deviations of >1.5 (Dataset S1) (18), resulting in a weighted mean age of 73.88 ± 0.32 ka (1σ; mswd = 0.95, prob. = 0.56, n = 36). An inverse isochron plot gives a statistically identical age of 73.6 ± 0.5 ka (1σ; mswd = 1, prob. = 0.12, n = 39) (Fig. S1). The ⁴⁰Ar/³⁶Ar intercept of 300 ± 3 is statistically indistinguishable from the atmospheric ratio of 298.6 ± 0.3 (19), thus indicating limited influence from excess argon or xenocrysts in the aliquots and further supporting the weighted mean age result.

These ⁴⁰Ar/³⁹Ar ages have been cross-calibrated against the astronomically dated A1 tephra (A1T) from Crete [6.943 ± 0.0025 (1σ) Ma (20)] using R-values previously determined on the Roskilde Noblesse (Table S1), where R is ⁴⁰Ar*/³⁹Ar_K (sample)/⁴⁰Ar*/³⁹Ar_K (standard) (21, 22). Because the A1T age is known independently of the K-Ar system, the ⁴⁰Ar/³⁹Ar age of the Toba sanidine crystals calculated relative to the A1T requires knowledge only of the ⁴⁰K total decay constant (equation 5 of ref. 22). As highlighted in ref. 23, this approach avoids incorporating the uncertainty associated with the ⁴⁰K branching ratio and also is relatively insensitive to the chosen ⁴⁰K total decay constant and its uncertainty of (5.464 ± 0.107) × 10^−10 y^−1 (24). Including the latter, as well as the uncertainty on the astronomical age of A1T [0.036% at 1σ (20)], does not increase the ± 0.32-ka uncertainty on the weighted mean age of the Toba sanidine crystals at two decimal places (Fig. 2).

The high-resolution NGRIP ice core records ∼6 y of high sulfate concentration with a spike at 2,548-m depth, near the start of the prolonged coldest part of the stadial between D-O 20 and 19 (Fig. 3). This sulfate anomaly has been attributed to the Toba eruption (9–11), but the proposed correlation cannot be confirmed because of the ± 5-ka uncertainty associated with the original Greenland Ice Sheet Project 2 (GISP2) ice core model age estimate of 71 ka (9) and the absence of microtephra (10). Some support for the Toba origin of the sulfate peak in the Greenland ice cores is given by the recovery of glass shards from Arabian Sea sediments; these shards have been geochemically associated with the Toba eruption and were deposited close to the Marine Isotope Stage 4/5 boundary (25).

High-precision U-Th ages of 75.92 ± 0.15 ka and 71.75 ± 0.11 ka (1σ) for the D-O 20 and 19 warming events, respectively, have been obtained for speleothems from the northern rim of the Alps (NALPS), a region that shares a dominant Atlantic influence and common oxygen isotopic signal with Greenland (15). These ages are systematically younger by about 400 to 600 y than those derived for these D-O events using the NGRIP GICC05modelext timescale (26), but the latter age estimates are an order-of-magnitude less precise (10). The timing of the NGRIP sulfate peak at 2,548-m depth, relative to the D-O 20 and 19 warming events, is resolved to ± 50 y using the GICC05modelext timescale. When combined with the NALPS ages for D-O 20 and 19, the age of the sulfate peak in the NGRIP core can be constrained to 73.7 ka, with a 1σ uncertainty ± 0.2 ka or better (Table 1). This age is within the uncertainty of our ⁴⁰Ar/³⁹Ar age for the Toba eruption, so these events cannot have been separated by more than a few centuries. This correlation affirms the original hypothesis that the large sulfate spikes in the Greenland ice cores between D-O 20 and 19 are most likely related to the Toba eruption (9).

Recent identification of an identical pattern of sulfate spikes in the Antarctic EDML core (11) now permits comparison of ice-core records from the two poles to elucidate leads and lags in the climate system and to test the hypothesis of a bipolar see-saw (27) contemporaneous with D-O and AIM events 20 and 19. Our precise ⁴⁰Ar/³⁹Ar age for the Toba eruption provides a well-constrained, radioisotopic-based calibration point for geological archives that lie beyond the range of ¹⁴C dating and contain traces of Toba ash or sulfates in primary depositional context, thus facilitating the synchronization of ice, marine, and terrestrial records of past environments.

Our ⁴⁰Ar/³⁹Ar age also has implications for the evolution and dispersal of Homo sapiens. The 74-ka Toba eruption has been variously implicated or exonerated in causing human population bottlenecks in Africa, mammal extirpations in Southeast Asia, and environmental changes in India (4–8), where stone tools attributed to anatomically modern humans are buried by YTT ash (4). Stone tools also are buried within and below YTT ash at Kota Tampan (16, 17) and are considered the handiwork of H. sapiens (17). If confirmed, then this evidence would support genetic estimates (28) for the first wave of dispersals of modern humans out of Africa and into Asia before 74 ka and argue against the initial exit occurring only after the eruption. As in India, the absence of associated human fossils at Kota Tampan precludes a definitive verdict (7), and it also is possible that the Kota Tampan artifacts were manufactured by Denisovans or modern humans who had exchanged genes with Denisovans in Southeast Asia (28, 29). If the descendants of these toolmakers survived the Toba supereruption, they could have spread eastwards during the following 2 millennia of cooler climate and lowered sea level, taking advantage of the newly exposed continental shelf and land bridges.

Angular but occasionally idiomorphic sanidine crystals of up to ∼2 mm in length, identified using a Bruker Tornado micro-XRF spectrometer, were extracted from the >500-μm fraction of the P3/D4 ash sample from bore hole BH1-2011 (Fig. 1). The sanidine crystals were hand picked and then ultrasonically leached in cold 10% (vol/vol) hydrofluoric acid for ∼1 min to remove adhering volcanic glass, followed by ultrasonic rinsing in deionized water. Approximately 80 mg of glass-clear sanidine crystals were loaded into a single well of a seven-well, 18-mm–diameter aluminum sample disk for ⁴⁰Ar/³⁹Ar dating (Fig. S2). Alder Creek sanidine (ACs), acting as the neutron fluence monitor and as an intermediate internal standard, was loaded into four evenly spaced wells on the sample disk, which then was wrapped in aluminum foil and encapsulated in a heat-sealed quartz tube. Fast neutron irradiation was carried out in the General Atomics TRIGA reactor using the Cadmium-Lined In-Core Irradiation Tube (CLICIT) facility at Oregon State University for 0.25 h on December 5, 2011. Argon isotopic analyses of the gas released by laser fusion of single- and multiple-grain sanidine aliquots (Datasets S1 and S2) were made on a fully automated, high-resolution Nu Instruments Noblesse multi-collector noble-gas mass spectrometer (Nu Instruments) at Roskilde University, using previously documented instrumentation and procedures (20, 30) that are summarized here.

Before fusion, crystals were gently degassed of loosely adhering argon by heating with a defocused low-power beam (0.3 W) from a 50-W Synrad CO₂ laser. Sample gas cleanup was through an all-metal extraction line, equipped with a −130 °C cold trap, to remove H₂O, and two water-cooled SAES Getters GP-50 pumps to absorb reactive gases. Analyses of unknowns, blanks, and monitor minerals were carried out in identical fashion during a fixed period of 400 s in 14 data-acquisition cycles, in which ⁴⁰Ar and ³⁹Ar were measured on the high-mass ion counter (HiIC), ³⁸Ar and ³⁷Ar on the axial ion counter (AxIC), and ³⁶Ar on the low-mass ion counter (LoIC), with baselines measured every third cycle. Measurement of the ⁴⁰Ar, ³⁸Ar, and ³⁶Ar ion beams was carried out simultaneously, followed by sequential measurement of ³⁹Ar and ³⁷Ar. Beam switching was achieved by varying the field of the mass spectrometer magnet and with minor adjustment of the quad lenses. All signals incorporate a small to insignificant correction for detector deadtime using the following deadtime constants: HiIC, 27 ns; AxIC, 37 ns; LoIC, 24 ns. The data collection and reduction were carried out using the program “Mass Spec” (by A. Deino, Berkeley Geochronology Center, Berkeley, CA). Individual fusion analyses were bracketed by one or multiple blank analyses. The precision on the blank measurement for the low-abundance isotope ³⁶Ar is better than ±0.5% (1σ). In comparison with single-collector peak-switching measurements, multicollection allows more data to be gathered in a fixed time, but for accurate and reproducible age determinations the method requires that the relative efficiencies of the different detectors be well known. As previously described for the Roskilde Noblesse (20, 30), by reference to the atmospheric argon isotopic composition (19), correction factors that combine detector efficiencies (detector intercalibration) and mass fractionation into single terms are based on the measurement of a time series of measured atmospheric argon aliquots delivered from a calibrated air pipette using the following detector configurations: (⁴⁰Ar/³⁶Ar)_HiIC/LoIC, (⁴⁰Ar/³⁸Ar)_HiIC/AxIC, and (⁴⁰Ar/³⁶Ar)_HiIC/AxIC (Fig. S3). Decay and other constants, including correction factors for interference isotopes produced by nucleogenic reactions, are given in Table S1.

We thank Saiful Shahidan and Shyeh Sahibul Karamah for help during fieldwork; Paul Renne for supplying monitor mineral ACs-2; Matt Heizler for organizing the 2008 EARTHTIME intercalibration experiment; and Kim Mogensen for assisting in the drafting of Fig. 1. We thank Tiffany Rivera and other members of the GTSnext Marie Curie Initial Training Network (funded by the European Community's Seventh Framework Programme) for valuable input. The Quaternary Dating Laboratory at Roskilde University is funded by the Villum Foundation. R.G.R. is supported by the Australian Research Council and M. Saidin by Apex University grants, Universiti Sains Malaysia.