A Tibetan ice core covering the past 1,300 years radiometrically dated with ³⁹Ar

Significance

Ice cores from high mountain glaciers outside the polar regions are unique archives of the past climate, but their use is often hampered by the lack of reliable dating, especially in the age range of 100 to 500 anni (a). The extremely rare noble-gas isotope ${}^{39}$Ar is ideally suited for radiometric dating in this age window. Advances in the detection method, atom trap trace analysis, have allowed us to apply ${}^{39}$Ar dating to an ice core from the Tibetan Plateau, yielding a timescale that covers the past 1,300 a. This work enables the use of further ice cores from polar regions for firn studies or from high mountain glaciers for retrieving climate records of the past millenium and beyond.

Abstract

Ice cores from alpine glaciers are unique archives of past global and regional climate conditions. However, recovering climate records from these ice cores is often hindered by the lack of a reliable chronology, especially in the age range of 100 to 500 anni (a) for which radiometric dating has not been available so far. We report on radiometric ³⁹Ar dating of an ice core from the Tibetan Plateau and the construction of a chronology covering the past 1,300 a using the obtained ³⁹Ar ages. This is made possible by advances in the analysis of ³⁹Ar using the laser-based detection method atom trap trace analysis, resulting in a twofold increase in the upper age limit of ³⁹Ar dating. By measuring the anthropogenic ⁸⁵Kr along with ³⁹Ar we quantify and correct modern air contamination, thus removing a major systematic uncertainty of ³⁹Ar dating. Moreover, the ⁸⁵Kr data for the top part of the ice core provide information on firn processes, including the age difference between the ice and its enclosed gas. This first application of ³⁹Ar and ⁸⁵Kr to an ice core facilitates further ice cores from nonpolar glaciers to be used for recovering climate records of the Common Era, a period including pronounced anomalies such as the Little Ice Age and the Medieval Warm Period.

Ice cores from mountain areas in mid and low latitudes such as the Andes, the Alps, the Caucasus, or the Tibetan Plateau are valuable archives for regional and global climate history (1–7). In particular, they complement polar ice cores in recovering trends in the regional climate variability and provide a reference for intercomparison with other records in the area. A prerequisite for retrieving a climate record from those ice cores is to establish a depth–age relationship. The construction of a reliable chronology by layer counting can be hampered by the lack of a clear seasonal signal in the physiochemical parameters (such as stable water isotopes, soluble ions, and dust) or because the depth resolution of their measurements is larger than the layer thickness, especially in the lower part of the core (8–10). Moreover, the complex interpretation of ${\delta }^{18}$O and global gas markers, like CH₄ or CO₂, in nonpolar ice cores hinders dating via stratigraphic matching with polar ice cores (11, 12). Glacier flow models can assist in the timescale construction. However, due to the complexity of glacier flow in high mountains, the uncertainty in the model input parameters can lead to large uncertainties in the age results, particularly if constraints are sparse (3, 13, 14).

Where layer counting is not possible or ambiguous and reference horizons from volcanic eruptions cannot be identified, radiometric dating can provide absolute age constraints for the construction of a chronology. The radiometric dating range of an atmospheric radioisotope is given by its half-life and its atmospheric history. For very old ice in the age range of 30,000 to 1.3 million anni (a), the noble gas radioisotope ${}^{81}$Kr can be used (15, 16). For ice of 500 to 20,000 a, ¹⁴C dating is available on the water-insoluble organic carbon (17–19), the dissolved organic carbon (20), or occasionally on trapped plant or animal residue if such happens to be found (2, 18). For ice of the past 100 a ³H, ¹³⁷Cs, and ²¹⁰Pb can be used (21, 22). In between, for ice cores in the age range of 100 to 500 a, a suitable radiometric dating method has so far been lacking. The cosmogenic isotope ³²Si with a half-life of around 150 a is in principle suitable for this time range. However, the large uncertainty of its half-life as well as the difficulty to analyze it have so far hampered its use in the earth sciences (23, 24). Continuing efforts to redetermine the half-life of ³²Si and to improve its analysis may facilitate its use for dating ice in the future.

The noble gas radioisotope ${}^{39}$Ar with a half-life of 268 ± 8 a (25, 26) has long been identified as an ideal dating isotope for water and ice in the range of 50 to 1,800 a given its chemical inertness and uniform distribution in the atmosphere (27, 28). However, due to its extremely low isotopic abundances (${}^{39}$Ar/Ar) of ${10}^{-17}$ to ${10}^{-15}$ in the environment, the analysis of ${}^{39}$Ar has posed a major challenge. In the past, it could only be measured by low-level counting requiring several tons of water or ice (29). Nevertheless, a number of studies using ${}^{39}$Ar were conducted on ice samples with known ages from Greenland, Antarctica, and Devon Island (29). The large amount of gas needed for analysis was extracted in situ from boreholes (30).

In recent years, the sample size for ${}^{39}$Ar dating has been drastically reduced by the emerging method, atom trap trace analysis (ATTA). This laser-based technique was originally developed for ${}^{81}$Kr and ${}^{85}$Kr (15, 16, 31–34) and was later also realized for ${}^{39}$Ar dating of groundwater and ocean water (35–37). Recently, ${}^{39}$Ar dating has also been demonstrated on ~5-kg glacier ice blocks that were cut out with a chainsaw from artificial glacier caves (38), obtaining an age uncertainty of 25 to 30% for samples with ages of 190 to 530 a and an age uncertainty of >70% for a sample of 1,126 a.

In this work, we have obtained an ${}^{39}$Ar profile for an ice core from the Tibetan Plateau providing ages covering the past 1,300 y. Due to major improvements in the analytical method, including increased ${}^{39}$Ar detection efficiency and reduced background (39, 40), we reach uncertainties of 7 to 16% for ages of 250 to 1,370 a (excluding contaminated samples). Moreover, parallel progress in the analysis of the anthropogenic ⁸⁵Kr (41) using ATTA now allows us to measure samples as small as 30 nL standard temperature and pressure (STP) of krypton, which corresponds to ~30 mL STP of air or 1 to 3 kg of high-elevation glacier ice from the Tibetan Plateau. This has made it possible to measure the short-lived ⁸⁵Kr (half-life $=$ 10.8 a) along with ³⁹Ar in the same ice sample to quantify and correct contamination with modern air and to determine the age difference between the ice and the gas phase [Δage (42, 43)] at different depth. These advances in the ³⁹Ar dating method allow us to assess a timescale for the ice core, previously constructed based on layer counting, and to develop a new timescale by constraining a glacier-flow model with the obtained ³⁹Ar ages.

Site Description and Ice Core Sampling

The details of the study site and ice coring are provided in ref. 44. The Qiangtang Glacier No. 1 is a valley glacier located in the middle of the Tibetan Plateau (Fig. 1). It is situated approximately in the boundary region of the Westerlies and the Indian monsoon (45), receiving accumulation mostly during the summer monsoon season (46, 47). The in-land plateau glaciers are stagnant to climate change (48) and hence survive longer. The small glacier has two branches, with the east branch covering an area of 2.415 km² and reaching a maximum depth of 132 m (49). The average air temperature is –10.7 ^∘C and the annual precipitation is 486 mm as recorded by a weather station on top of the glacier (46).

Fig. 1.

Location of Qiangtang Glacier No.1 shown together with locations of other ice coring sites on the Tibetan Plateau.

In 2011, a 17-m shallow ice core was drilled on the saddle of the glacier (33.30^∘N, 88.69^∘E, 5,890 m above sea level [a.s.l.]). The top of this Core2011 consists of fresh snow, providing a surface age horizon (44). In May 2014, two parallel bedrock ice cores were retrieved from the same site (~2 m distance), both reaching a depth of ~109 m (44). Different from the 2011 core, the top of the 2014 cores consists of ice. Already at a depth of ~70 cm the density reaches a constant value (see density profile in SI Appendix). The 2014 ice cores were packed and sealed in polyethylene bags at the field site and then transported to a freezer facility in Lhasa where the cores were stored at –20 ^∘C. Core2014-1 was analyzed for stable water isotopes, radioisotopes (³H, ¹³⁷Cs, and β-radioactivity), and glacio-chemical parameters such as soluble ions, dust, and nitrate isotopes (44, 47, 50, 51). The comparison of the isotopic records of Core2011 and Core2014-1 indicates that the surface of Core2014-1 originates from the year 2011. This means that ~3 a of accumulation were lost (44).

A timescale for Core2014-1, named QT-timescale1, based on layer counting of Core2014-1 was constructed using the bomb peak signature in the radioisotopes (for depth $\le$ 20 m), seasonality in the isotopic signal (for 20 m < depth < 79 m), and seasonality in the visual stratigraphy (for 79 m $\le$ depth $\le$ 109 m). The timescale reaches an age of 1,345 ± 66 a at the bottom (this age and all following ages are provided relative to the year 2014 when the ice cores were drilled), in reasonable agreement with glacier-flow models (44). The depth relation of Core2014-1 and Core2014-2 is compared using several pronounced dust layers in the visual stratigraphy between 92 and 109 m (SI Appendix). The dust layers in Core2014-2 are lower than the ones in Core2014-1 by ~1.2 m, perhaps due to accumulated inaccuracies in the length determination of the core pieces or due to topographical differences at the two drilling sites. When transferring results obtained for Core2014-2 to Core2014-1, the depth for Core2014-2 is converted and its uncertainty increased as described in SI Appendix.

Results and Discussion

In 2019 to 2021, Core2014-2 samples were processed, degassed, and analyzed for ${}^{39}$Ar and ${}^{85}$Kr as described in Materials and Methods. The recently developed dual separation of argon and krypton with high recovery and purity (52) has allowed us to measure ${}^{39}$Ar and ${}^{85}$Kr in the same gas extracted from an ice sample. The full dataset for the measured ice samples is compiled in Table 1. The ice samples weigh between 3 and 7 kg and contain gas amounts of 37 to 137 mL STP. The resulting air contents are 8 to 20 mL STP/kg, which is on the lower side of the typical range for ice samples from the Tibetan Plateau (15, 53, 54). The raw, measured ${}^{39}$Ar and ${}^{85}$Kr abundances vs. depth (given in Table 1) are plotted in Fig. 2. The ${}^{85}$Kr abundance decreases from 42 dpm/cc (decay per minute per cubic centimeter STP of krypton) at 1.4 m depth to <1.3 dpm/cc at 30 m where our detection limit for ${}^{85}$Kr is reached. Therefore, any detectable ${}^{85}$Kr below 30 m indicates contamination with modern air, which is the case for several samples. The contamination has presumably occurred either when the ice was degassed or during sample gas storage. The contamination may potentially also be due to postcoring entrapment of modern air in open porosity due to melting at the surface and subsequent refreezing (55, 56). However, as the top of the core consists of ice and as the contamination is not limited to the upper part of the core, this contamination mechanism does not seem to be dominant. The ${}^{39}$Ar abundance decreases from around 100 pMAr (percentage of the modern ³⁹Ar level) at the top of the core down to values close to 0 pMAr at the bottom, indicating that the ice core covers a large part of the ${}^{39}$Ar dating range. The ${}^{39}$Ar abundance of 6.4 pMAr for the sample at 109 m is higher than that of the samples above, reflecting contamination with modern air as identified with ${}^{85}$Kr.

Table 1.

The ${}^{39}$Ar and ${}^{85}$Kr data for the Qiangtang ice-core samples

Depth,   m	Weight, kg	Air, mL STP	Air content, mL STP/kg	Argon, mL STP	Krypton, nL STP	85Kr, dpm/cc	85Kr age, a*	39Ar, pMAr	39Ar corr, pMAr	39Ar age, a*
2	5.6	98	17.5	0.9	70	41.67 ± 1.12	14.2 ± 0.5	${102.5}_{-8.5}^{+8.8}$	${102.5}_{-8.5}^{+8.8}$	<56
5	5.8	86	15.0	0.7	60	24.28 ± 1.7	12.3 ± 0.9	${90.0}_{-7.9}^{+8.3}$	${90.0}_{-7.9}^{+8.3}$	<121
10	2.9	51	17.6	0.5	30	9.55 ± 1.55	${23.2}_{-1.9}^{+1.7}$	${94.1}_{-8.5}^{+8.8}$	${94.1}_{-8.5}^{+8.8}$	<101
30	3.8	58	15.3	0.5	40	<1.3	>60	${56.7}_{-6.1}^{+6.4}$	${56.7}_{-6.1}^{+6.4}$	${239}_{-44}^{+49}$
40	6.0	86	14.3	0.9	60	7.93 ± 0.92	—	${53.2}_{-6.0}^{+6.6}$	${48.3}_{-6.8}^{+7.5}$	${307}_{-59}^{+49}$
50	4.4	37	18.4	0.3	30	3.95 ± 1.11	—	${38.2}_{-3.6}^{+3.8}$	${35.1}_{-4.2}^{+4.4}$	${422}_{-44}^{+47}$
57	7.1	72	10.1	0.7	60	1.21 ± 0.49	—	${35.9}_{-3.0}^{+3.1}$	${35.0}_{-3.1}^{+3.3}$	${423}_{-36}^{+36}$
63	6.0	93	15.6	0.8	50	4.04 ± 0.71	—	${28.2}_{-3.6}^{+3.9}$	${24.5}_{-3.9}^{+4.2}$	${549}_{-51}^{+54}$
69	5.8	45	17.8	0.5	50	<0.1	—	${20.8}_{-2.7}^{+2.9}$	${20.8}_{-2.7}^{+2.9}$	${599}_{-41}^{+55}$
75	7.0	109	15.6	1.1	90	<0.63	—	${11.6}_{-2.0}^{+2.3}$	${11.6}_{-2.1}^{+2.3}$	${809}_{-67}^{+77}$
80	6.2	75	12.1	0.7	80	1.29 ± 0.63	—	${5.4}_{-1.4}^{+1.8}$	${3.9}_{-1.6}^{+2.0}$	$1,{233}_{-158}^{+195}$
85	6.0	73	12.2	0.7	80	<0.7	—	${8.5}_{-1.6}^{+1.8}$	${8.5}_{-1.7}^{+1.8}$	${934}_{-78}^{+76}$
90	5.9	93	15.9	0.9	80	<0.24	—	${4.0}_{-1.1}^{+1.5}$	${4.0}_{-1.1}^{+1.5}$	$1,{222}_{-125}^{+120}$
95	7.0	122	17.5	1.2	120	<0.42	—	${2.6}_{-0.6}^{+0.9}$	${2.6}_{-0.7}^{+0.9}$	$1,{369}_{-99}^{+110}$
100	6.6	102	15.5	1.1	110	2.12 ± 0.78	—	${2.6}_{-0.9}^{+1.3}$	<2.5	>1,387
106	6.7	137	20.4	1.2	140	<0.5	—	${2.8}_{-0.8}^{+1.1}$	${2.8}_{-0.9}^{+1.1}$	$1,{353}_{-115}^{+145}$
109	6.0	78	12.9	0.8	90	2.78 ± 0.62	—	${6.4}_{-2.5}^{+3.2}$	<8.7	>925

The given errors are $1\sigma$ SDs, whereas upper limits are given at a 90% confidence level. The ⁸⁵Kr abundance drops below the detection limit at a depth of 30 m. Therefore, the measurable ⁸⁵Kr abundances below 30 m, originating from contamination, are not converted to age. The raw ³⁹Ar abundances as well as the ³⁹Ar abundances corrected for modern air contamination (for depth >30 m) are provided. The shown ³⁹Ar ages are based on the corrected ³⁹Ar abundances (see SI Appendix for details). The given error for the ³⁹Ar ages is based on the ³⁹Ar counting statistics as well as the error from the contamination correction. In addition, there is a systematic age uncertainty of 3% due to the error of the ³⁹Ar half-life, which would shift all ³⁹Ar ages up or down together. This error can be corrected and improved in the future with a more precise measurement of the ³⁹Ar half-life. Moreover, there is a systematic age uncertainty of about 20 a due to the uncertainty of the atmospheric ³⁹Ar history (40). corr, corrected; dpm/cc, decay per minute per cubic centimeter STP of krypton; pMAr, percentage of the modern (year 2018) ${}^{39}$Ar level.

^*Relative to the year 2014 when the ice core was drilled.

Fig. 2.

The ${}^{39}$Ar and ${}^{85}$Kr abundances vs. depth of the ice core. The lines are guides to the eye. dpm/cc, decay per minute per cubic centimeter STP of krypton; pMAr, percentage of the modern ${}^{39}$Ar level, using 2018 as the year of reference (40).

The raw, measured ${}^{39}$Ar abundances (given in Table 1) below 30 m are corrected for modern air contamination using the measured ${}^{85}$Kr abundances (Materials and Methods and SI Appendix). The ${}^{39}$Ar ages are calculated from the corrected ${}^{39}$Ar abundances based on its radioactive decay as well as its atmospheric history (40). The ${}^{39}$Ar ages are provided relative to the year 2014 when the ice cores were drilled. The ${}^{39}$Ar ages at 100- and 109-m depths acquire a large uncertainty due to the contamination correction (Table 1), so that only lower age limits can be obtained. The ${}^{39}$Ar ages are shown in Fig. 3 together with QT-timescale1. Down to 70 m the ${}^{39}$Ar ages are close to QT-timescale1, although they lie consistently above it. Below 70 m, the ${}^{39}$Ar ages start to deviate significantly by 100 to 300 a. For depth <79 m QT-timescale1 is based on the seasonality in the stable water isotope signal whereas for depth >79 m it relies on the visual difference between winter and summer layers. A systematic undercount of annual layers based on the isotopic signature, particularly between 70 and 79 m, is a possible explanation for the discrepancy. Another hypothesis is that there may have been a systematic loss of ice layers due to melting in the time period covered by the ice core at 70 to 80 m depth. According to ${}^{39}$Ar, the age span covered in that depth is 600 to 900 a, which roughly coincides with the Medieval Warm Period. However, the warmest period in the past 2,000 a has been the present century and the ice of that period is well preserved, so a loss of 100 to 300 a of layers during the Medieval Warm Period seems unlikely. Similarly, it also does not seem probable that there has been an extended dry period that would lead to the absence of so many layers.

Fig. 3.

The ${}^{39}$Ar ages of the gas (given in Table 1) vs. depth of the ice core compared to QT-timescale1 (44), which is based on layer counting as well as the bomb peak horizon. The ages are relative to the year 2014, when the ice cores were drilled. The dashed lines indicate upper/lower ${}^{39}$Ar age limits as given in Table 1. The new ice timescale is based on a glacier flow model constrained by the ${}^{39}$Ar gas ages converted to ice ages using the Δage (see main text) and the bomb peak horizon. The shaded green area indicates the 1σ confidence band.

To obtain information about processes in the firn including the ice–gas age difference, we use the ${}^{85}$Kr data for the top 30 m of the core. The ${}^{85}$Kr ages are calculated from the measured ${}^{85}$Kr abundances based on its radioactive decay as well as its atmospheric history (57). The resulting ${}^{85}$Kr ages (provided relative to the year 2014 when the ice cores were drilled) are shown in Fig. 4 for the upper 30 m of the core together with the ${}^{39}$Ar ages and QT-timescale1 (44). Interestingly, the ${}^{85}$Kr ages (gas) at 2, 5, and 10 m differ significantly from QT-timescale1 (ice), which for the top 20 m has a high degree of confidence due to the bomb peak horizon. Moreover, the age difference between QT-timescale1 and the ${}^{85}$Kr ages (i.e., the Δage) increases with depth. The Δage is related to firn thickness and accumulation (42, 43, 58). The increase in Δage with depth may thus be caused by a decrease in accumulation or by a faster firn densification. Both processes can be induced by higher temperatures. A decrease in accumulation is supported by the observation that the surface year of Core2014-1 is 2011, indicating ~3 a of lost accumulation. This is in line with other studies of dramatic loss of glacier accumulation area on the Tibetan Plateau caused by the warming conditions on the Tibetan Plateau in the past 60 a (4). Also, note that already at a depth of ~70 cm Core2014-1 reaches a constant density (SI Appendix). We obtain Δage $=$ ${32}_{-5}^{+2}$ a at a depth of 10 m, where we have included the mean ${}^{85}$Kr contamination below 30 m in the error to take into account that during processing the ${}^{85}$Kr may have been altered by contaminant krypton (which for the top 30 m cannot be distinguished from sample krypton).

Fig. 4.

The ${}^{85}$Kr and ${}^{39}$Ar ages of the gas (given in Table 1) vs. depth compared to QT-timescale1 of the ice for the top 30 m of the ice core. The ages are relative to the year 2014 when the ice cores were drilled. The top ice layer originates from the year 2011 so its age relative to the year 2014 is 3 a (44). The dashed lines indicate upper/lower limits as given in Table 1.

The ${}^{39}$Ar ages suggest that QT-timescale1, especially below 70 m depth, is at least slightly biased toward too old ages. Based on the ${}^{39}$Ar ages as well as the already available bomb peak horizon a new timescale was constructed. To convert the gas ages obtained from ${}^{39}$Ar to ice ages, the Δage is needed for all sampled depths. Using the above hypothesis that the increase of the Δage with depth reflects the warming conditions on the Tibetan Plateau, the Δage at 2 m can be considered as a lower limit given that the present temperatures are likely the highest in the past 1,300 a. As an upper limit, we use the Δage of 50 a obtained for the Himalayan ice core at Dasuopu (7,200 m a.s.l.) (59). Within this confidence range we use the Δage obtained at 10 m to convert ${}^{39}$Ar ages to ice ages, i.e., using Δage $=$ ${32}_{-24}^{+18}$ a. We establish the depth–age relationship by applying a two-parameter glacier flow (2p) model (60, 61) using the converted ${}^{39}$Ar ages and the bomb peak horizon as constraints. The resulting timescale and its uncertainty are shown in Fig. 3, indicating an ice age of $1,{573}_{-204}^{+231}$ a at 100 m depth. The ${}^{39}$Ar ages at 80 and 106 m depth differ significantly from the new timescale, but still lie within 2σ statistical uncertainty. The comparatively young age at 106 m may hint at disturbed flow close to the bedrock. However, there is no evidence for that in the glaciology, the visual stratigraphy, or the glacio-chemical analysis.

Conclusions and Outlook

Advances in the analysis of ${}^{39}$Ar using atom trap trace analysis have led to a substantial decrease in the uncertainties of the ${}^{39}$Ar ages (Fig. 5). For the ${}^{39}$Ar ages obtained on the Qiangtang ice core in this study, we reach uncertainties of 7 to 16% (excluding contaminated samples) for ages of 250 to 1,370 a (Table 1). Moreover, we infer that samples with an age of 1,800 a can be measured with 20% age uncertainty (Fig. 5). This is a twofold extension of the ³⁹Ar dating range compared to previous efforts (38). Besides identifying contamination, the ⁸⁵Kr data allow for obtaining the ice–gas age difference, yielding Δage $=$ ${32}_{-5}^{+2}$ a at the depth of 10 m. In comparison, Δages of 50 and 30 a have been estimated for the Himalayan ice cores at Dasuopo (7,200 m a.s.l.) and East Rongbuk (6,518 m a.s.l.), respectively, based on measurements of bubble volume and ice density (53, 59). The increase in the Δage with depth observed in the Qiangtang ice core is in agreement with and supports previous findings of past century warming conditions experienced in this region (4, 62). These findings on the Δage suggest that further studies of the Δage in Tibetan ice cores are of interest, using gas parameters such as nitrogen isotopes (43, 63), ⁸⁵Kr, bubble volume, air content, and density in comparison with ice parameters such as ${\delta }^{18}$O, ³H, and ¹³⁷Cs.

Fig. 5.

Calculated ${}^{39}$Ar age uncertainty as a function of sample size and sample age. The ice amount on the right y axis is calculated based on an air content of 20 mL/kg, typical for alpine glaciers. The calculated uncertainties are confirmed by the actual measurements of this work (Table 1), indicated as red dots (contaminated samples not included).

With the demonstrated ${}^{39}$Ar dating of an ice core, a powerful tool is available to constrain chronologies of ice cores for the past 1,800 a, bridging the radiometric dating gap between ²¹⁰Pb and ¹⁴C. This will considerably enhance the value of nonpolar ice cores as an archive for the climate of the Common Era, not only on the Tibetan Plateau but also in other high mountain areas such as the Alps, the Andes, and Central Asia. This is particularly relevant given the spatiotemporal incoherence of climate anomalies in that time period, such as the Medieval Warm Period and the Little Ice Age (64, 65). Furthermore, the anthropogenic influence on climate and atmospheric chemistry, which mainly occurred during the past few hundred years, is recorded in those ice cores (66–69).

As an absolute tracer for the gas age, ${}^{39}$Ar may also be a useful future tool to further investigate the air-trapping process in polar firn and ice (42, 43, 58, 70). High-resolution methane records can provide precise timescales for polar ice cores through correlation techniques with well-dated records on a centennial timescale (71). The ${}^{39}$Ar may assist the methane records to quantify smoothing of ice core gas records due to diffusion. The change in the ${}^{39}$Ar/${}^{38}$Ar ratio due to gravitational fractionation in the firn is ~0.05% (72), two orders of magnitude smaller than the typical measurement error. Since the air content of polar ice is ~100 mL STP/kg, 5 to 10 times higher than that of the Tibetan ice core used in this study, a few hundred grams of polar ice would be sufficient for ${}^{39}$Ar analysis.

In this study, ice samples between 3 and 7 kg have been used due to their low air content. As shown in Fig. 5, for ages of 400 to 1,200 a, a sample size of 0.5 to 1 kg is already sufficient to obtain an ${}^{39}$Ar age with ~20% precision. The continuing development on the analysis of ³⁹Ar with ATTA is expected to lower the required sample size and to increase the precision further, which will enable even stronger constraints on the chronologies of ice cores and their archived climate records.

Materials and Methods

Extraction of Air from the Ice Samples.

For ${}^{39}$Ar and ${}^{85}$Kr analysis, the air trapped in the ice has to be extracted. Prior to extraction, the surface of the ice samples is scraped and cleaned to remove any layers or flaky debris that may contain modern air. The ice is then transferred to a stainless-steel container (Fig. 6), which is thereafter sealed and evacuated for about 30 min by a scroll pump through a water trap (stainless-steel bellow immersed in ethanol at –115 ^∘C).

Fig. 6.

Apparatus for extracting air from ice samples for ${}^{39}$Ar and ${}^{85}$Kr analysis.

Since during pumping the chamber is constantly being flushed by the water vapor released by the sublimating ice, the remaining atmospheric gas in the container is rendered negligible. After evacuation, the container is heated by a stove for 20 to 30 min (depending on the ice mass) until the ice has completely melted. The gas released from the ice passes through the water trap and is trapped on activated charcoal at liquid nitrogen temperature. The air content of the ice sample is determined based on the pressure after expanding the trapped gas into a calibrated volume. Extraction efficiencies higher than 99% and contamination with modern air below 1% are achieved with this degassing method. It takes ~1 h to process a sample.

³⁹Ar and ⁸⁵Kr analysis.

The extracted gas from the ice samples has been sent to the University of Science and Technology of China (USTC) for dual argon/krypton purification and for ATTA analysis of ${}^{39}$Ar and ${}^{85}$Kr. Argon and krypton are separated from the extracted gas using a purification system based on titanium gettering and gas chromatography (52), yielding purities and recoveries higher than 99% for argon and 90% for krypton, respectively. The ${}^{85}$Kr is measured to identify and quantify any air contamination during ice sampling and processing. Moreover, it is used to obtain information about the ice/gas age difference in the ice core. We use a recently improved ATTA system for krypton at USTC (15, 16) to measure the ${}^{85}$Kr abundances. The stable and abundant ${}^{83}$Kr is also measured to cancel drifts in the counting efficiency. The resulting ${}^{85}$Kr/${}^{83}$Kr ratio of the samples is compared to the corresponding ratio of a reference krypton gas to derive the ${}^{85}$Kr abundance. The amount of krypton in the ice samples presented in this study is only 30 to 140 nL STP, which is an order of magnitude lower than the smallest krypton samples analyzed in previous works (15, 16). A thorough cleaning procedure with xenon (>60 h washing time) is used to suppress cross-sample contamination in the system.

The ${}^{39}$Ar measurements are performed with the latest ${}^{39}$Ar-ATTA instrument at USTC (39), where individual ${}^{39}$Ar atoms are selectively laser cooled and then detected in a magneto-optical trap. The stable and abundant ${}^{38}$Ar is also measured for normalization. It is measured every 10 min for 20 s during the ${}^{39}$Ar measurement (SI Appendix). The resulting ${}^{39}$Ar/${}^{38}$Ar ratio of the sample is compared to the corresponding ratio of a reference argon gas to derive the ${}^{39}$Ar abundance. The reference gas is an argon sample enriched in ${}^{39}$Ar with an ${}^{39}$Ar abundance of 979 ± 31 pMAr. It is measured before and after each sample for about 4 h (SI Appendix). A high ${}^{39}$Ar age limit is desired to date ice cores as far back in time as possible. The upper ${}^{39}$Ar age limit is determined by the ${}^{39}$Ar count rate as well as the level of contamination in the ATTA system. Similar to the ${}^{85}$Kr measurements, we employ a thorough cleaning procedure with xenon (>60 h washing time) before each analysis to suppress cross-sample contamination from the system. Moreover, we determine the residual background after the washing procedure with a ${}^{39}$Ar-free sample for correction (39). A typical remaining background after the washing procedure is 0.4 ± 0.2 pMAr (depending on sample amount and washing time) (SI Appendix). We use the Feldman–Cousins method (73) for the error analysis of the measured ${}^{39}$Ar abundances. For the ice core samples below 30 m, the ${}^{39}$Ar abundances and their errors shown in Table 1 are corrected for contamination with modern air using the measured ${}^{85}$Kr abundances relative to the atmospheric ${}^{85}$Kr value of 82 dpm/cc (measured in China in the year 2020). The ${}^{39}$Ar abundances corrected for contamination with modern air are shown in SI Appendix, Table S2. The ${}^{39}$Ar ages are then deduced using the known radioactive decay rate as well as the reconstructed atmospheric history of ${}^{39}$Ar (40). More specifically: If $R_{39}^{\text{s,corr}}$ denotes the measured ${}^{39}$Ar abundance of a sample corrected for contamination with modern air and $R_{39}^0(t)$ denotes the ${}^{39}$Ar abundance in the atmosphere at time t, then the best solution for the time t_s (i.e., the ${}^{39}$Ar age of the sample) is sought that fulfills the radioactive decay equation $R_{39}^{\text{s,corr}}=R_{39}^0(t_s)·\text{exp}(-t_s/{\tau }_{39})$, where ${\tau }_{39}=387\hspace{0.17em}\text{a}$ is the mean lifetime of ${}^{39}$Ar. This equation can be solved numerically using the reconstructed data of $R_{39}^0(t)$ provided in ref. 40 to obtain the ${}^{39}$Ar ages provided in Table 1 (see SI Appendix for further information). The initial atmospheric ${}^{39}$Ar abundances $R_{39}^0(t_s)$ obtained for the Qiangtang samples are provided in SI Appendix. The data for the argon/krypton purification as well as ${}^{39}$Ar and ${}^{85}$Kr analysis are compiled in Table 1. QT-timescale1 is provided in SI Appendix.

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.