Combined uranium-series and electron spin resonance dating from the Pliocene fossil sites of Aves and Milo’s palaeocaves, Bolt’s Farm, Cradle of Humankind, South Africa

Aves cave

Aves Cave (7120889.838N, 571642.968E) is located at the northern end of the Greensleeves property, on the boundary between Greensleeves and Klinkerts and was first excavated in 2011. What was originally defined as Pits 5 (Smithy Cave), 8 (Rodent Cave), and 14 (Benchmark Pit) by Peabody was referred to as Aves Cave 4, Aves Cave 2 and Aves Cave 1 complex by Pickford & Gommery (2016), but as other names such as Smith Cave, Rodent Cave and Pit 15 in other publications (detail in: Edwards et al., 2019). Pit 15, as noted on Peabody’s map, is a small pit to the west of Pit 14 and Pit 13 appears to just be a lime miners dump. Pits 5, 8 and 14 have now been shown to have been dug into different parts of the same palaeocave deposits now simply known as ‘Aves Cave complex’ (Edwards et al., 2020).

The stratigraphy of all three pits has been described by Edwards et al. (2020), and the deposits were previously dated by U-Pb and palaeomagnetism to 3.03–2.61 Ma, making it terminal Pliocene in age. A basal flowstone in Pit 5 was U-Pb dated to 2.41 ± 0.74 Ma, while a capping flowstone in Pit 14 was U-Pb dated to 2.67 ± 0.30 Ma suggesting a depositional age not younger than 2.37 Ma and not older than 3.15 Ma. Sediments from Pit 5 and 14 have a normal magnetic polarity indicating they could not have formed younger than the Gauss-Matuyama Boundary at ~2.61–2.58 Ma (Singer, 2014; Ogg, 2020). The age is therefore sometime between ~3.03 and ~2.58 Ma in C2An.1n (Pliocene, Gauss normal Chron in geomagnetic polarity timescale), or between 3.15 and 3.12 Ma in C2An.2n, although the later age estimate was considered more unlikely (Edwards et al., 2020; Fig. 2). This age is consistent with macrofaunal remains recovered from the deposit, specifically Metridiochoerus andrewsi Stage I (Edwards et al., 2020), (also defined as Potamochoeroides hypsodon by Pickford & Gommery (2020)) which has been recovered from the Makapansgat Limeworks at 3.03–2.61 Ma and Drimolen Makondo at ~2.61 Ma, and is not known from eastern Africa before 3.4 Ma (Herries et al., 2013, 2018).

Only Pit 14 was sampled for US-ESR dating and so only this stratigraphic sequence is described here (Fig. 2B). The Pit 14 sequence starts with a basal flowstone (Facies E) that is considered equivalent to the dated flowstone in Pit 5. This is overlain by Facies A that consists of large angular boulders surrounded by a fine-grained sandstone and siltstone matrix representing gravity roof collapse. This is overlain by Facies B that is a massive sandy siltstone with few clasts that represents a significant flood event. This is interlayered with thin lenses of Facies B1 that consists of sandy siltstone with more clasts and bone. Part way throughout the Pit 14 sequence Facies B1 begins to dominate and the sequence transitions to Facies F that consists of red brown laminated siltstone. The sequence is then capped by the upper 2.67 ± 0.30 Ma U-Pb dated flowstone. Bovid teeth for US-ESR dating were sampled from Facies B1. Sample ESR-01 was recovered from the base where Facies B is more prevalent just below palaeomagnetic sample AV08 (Fig. 2). ESR-02 and 03 are the ESR dated samples come from the upper part of the sequence where Facies B1 dominates between palaeomagnetism samples AV08 and AV10. The aim of this study is to conduct US-ESR dating of teeth from Aves Cave to contrast against the consistent biochronological, U-Pb and palaeomagnetic data and to test the upper age limit of US-ESR dating in the region.

Milo’s cave

Milo’s Cave (7120526.445N, 571103.527E) is situated in the middle of the Klinkert’s Property of Bolt’s Farm, just to the east of Pit 16 (Edwards et al., 2019). Because Milo’s Cave was not mined for speleothem it is a relatively intact cave remnant however it has been heavily eroded, and is covered by a Makondokarren (Fig. 3) (Brink & Partridge, 1980; Herries & Shaw, 2011). The vertical stratigraphy at the site is extremely limited, but there is significant lateral variation. The US-ESR samples were collected from a bone rich breccia deposit on the NE corner of the site that was closer to an entrance. The western part of the deposit consists of interlayered flowstone speleothem, sandstone and siltstone deposits that represent material winnowed to the rear of the cave. The ages of Milo’s cave are mostly indicated by biochronology. Gommery et al. (2012) identified Potamochoeroides hypsodon (defined by others as Metridiochoerus andrewsi Stage I) from Milo’s (A), which would suggest a late Pliocene age as with Aves Cave, however, Pickford & Gommery (2020) subsequently defined another fossil from the site as a later stage (III) of Metridiochoerus andrewsi that they use to suggest the deposit is much younger at ~1.8 Ma.

Sample preparation

The experiments were carried out on fossil bovid teeth extracted from breccia samples from Bolt’s Farm, South Africa. A series of tooth enamel fragments and enamel powder were prepared and analysed follow the standard ESR dating methods. The teeth used in the analysis are housed at the Ditsong National Museum of Natural History in Pretoria. Detailed information on the analysed samples of each pit is shown in Tables 1 and 2. The methodology of powder follows that established by Joannes-Boyau (2010) and fragment analysis follows Joannes-Boyau (2013), Fig. 4.

Among the seven teeth, five teeth had enough enamel powder volume. For each of these teeth, two fragments were cut from the fossil tooth, avoiding areas showing obvious diagenetic process, and cleaned from dentine and tartar using a diamond saw. One fragment was crushed into powder and split into 10 aliquots (about 50–70 mg each) then sent for gamma irradiation, while the other fragment was kept for X-ray irradiation (Fig. 4A). For the remaining two fossil teeth without enough enamel volume, only non-destructive US-ESR dating on enamel fragments was carried out (Joannes-Boyau, 2013; Yu et al., 2022). Such methods make it possible to date teeth by US-ESR dating even when the tooth samples have not met the minimum volume of tooth enamel for powder analysis (Fig. 4B).

Irradiation settings

Gamma irradiations were performed for powder samples irradiated with a calibrated 60 Co gamma source with a dose rate of 23.8 Gy/min at the Australian Nuclear Science and Technology Organisation (ANSTO), Australia. Powder samples received gamma irradiation steps of 50, 100, 250, 600, 1,200, 2,400, 4,000, 8,000, 15,000 Gy.

X-ray irradiations were performed for fragments at Southern Cross University (SCU) on a Freiberg X-ray irradiation chamber, which contains a Varian VF50. Irradiation parameters: 40KV voltage and 0.5mA current. The samples were mounted onto a Teflon sample holder, which directly exposed the fragment to the X-ray source with no shielding. The X-ray irradiation protocol followed a classic exponential dose step distribution with the dose steps: 90, 360, 900, 1,800, 3,600, 7,200 and 14,400 s (with an average dose rate of 0.22 to 0.25 Grays (Gy/s)) (details see Joannes-Boyau, 2013). The X-ray emission received by the fragment samples was calibrated using a known gamma irradiation dose performed at ANSTO.

Equivalent dose (D_E) determination

ESR intensities of the aliquots were measured by a Frieberg MS5000 ESR spectrometer at SCU and all the measurements were performed with the following measurement conditions: 2 mW microwave power, 100 kHz modulation frequency with 1,024 points resolution, 0.1 mT modulation amplitude, 45 s conversion time, 12 mT sweep width, 21 s sweep time. The ESR intensities of enamel powder were determined followed Duval & Grün (2016) and the ESR intensities of enamel fragments followed Grün, Joannes-Boyau & Stringer (2008), Joannes-Boyau, Grün & Bodin (2010), Joannes-Boyau (2013). D_E values and all dose response curves (DRC) were obtained by fitting a single saturated exponential (SSE) function because it provides a better fit for our data in the Matlab based fitting program McDose 2 (Joannes-Boyau, Duval & Bodin, 2018).

U-series data

High resolution U-series analyses of the enamel and dentine were performed from the remaining tooth. The concentration of ²³⁸U, ²³⁴U and ²³⁰Th was determined by ESI NW193 ArF Excimer laser ablation multi collector inductively coupled plasma mass spectrometry (LA-MC-ICP-MS). A raster or ablation track of laser ablation measurements were performed twice across both the dentine and enamel of each sample. The ablation rate of LA-ICP-MS was 20 Hz and the scan speed was 5 μm/s. ²³⁴U and ²³⁰Th were measured simultaneously, with uranium in the center faraday cup coupled with a secondary electron multiplier (SEM) and thorium on the L3 faraday cup coupled to an ion counter (IC). All other faraday cups were set to use high-gain 1,011 Ω amplifiers. The cup configuration was as follows: L3/IC (230); L2 (232); L1 (233); C/SEM (234); H1 (235); H2 (236); H3 (238). Baselines and drifts were corrected using NIST 610 and NIST 612 glass standards. In contrast, two corals (the MIS7 Faviid and MIS5 Porites corals from the Southern Cook Islands) and a rhinoceros tooth with a known concentration were used to correct ²³⁴U/²³⁸U and ²³⁰Th/²³⁸U ratios and assess the accuracy of measurements. The direction of the scan in a series of transects followed the growth axis of the enamel and dentine. An average value from the U-Th data was calculated for each dental tissue and used for the optimum U-uptake model calculation in the US-ESR model.

Age calculation

Combined US-ESR ages follow the equation: Age = D_E/Da and were calculated using the USER program (Shao et al., 2012) with the following parameters: alpha-efficiency of 0.13 ± 0.02 (Grün & Katzenberger-Apel, 1994), the water content of 3 percentage by weight (wt%) in the enamel, 5 wt% with 2 wt% error in the dentine, and 0% in cementum. It is not possible to measure the water content of these open cave sites, and so a fixed value of 10% ± 5 wt% was assumed for the US-ESR age calculation. The external dose rate was determined by the radioisotopes U, Th, and K content in the attached sediments collected from the site at the Environmental Analysis Laboratory (EAL). The gamma spectrometer Inspector 1,000 was used to measure the in situ deposits surrounding the teeth at Bolt’s Farm. The cosmic dose rate was estimated by the site variation over the burial time. The error of the cosmic dose rate was set to 80% because of the complicated deposition history of Bolt’s Farm. Rn and Ra loss were not able to be measured within the dental tissue, and thus a state of equilibrium was assumed in this study. This is because the highly mineralized teeth are typically not thought to be affected by Rn and Ra loss from experience (Dirks et al., 2017; Herries et al., 2020). The mathematical algorithm for calculating the ages in the USESR program followed Shao et al. (2014). The Early Uptake (EU) model and Linear Uptake (LU) model generated by the DATA program were also applied as a reference to the age as per Grün (2009).

Aves cave

As shown in Table 1, the US-ESR and EU models are most closely aligned with each other. In contrast LU models give slightly higher median age estimates, but the ages from all three models overlap with each other in all of the teeth. Tooth AV-ESR-02 gives the youngest ages of between ~2.6 and ~2.3 Ma (powder ages), while AV-ESR-01 and 3 give older median ages of between 3.1 and 2.8 Ma for powder and 3.5 to 3.2 Ma for fragments. However, the fragment ages have large errors and overlap within error of the powder ages.

Fragment ages were not initially able to be calculated for teeth AV-ESR-3 and AV-ESR-02. However, when this sample was tested with no isotopic error ratio allowed the generation of an unrealistic age of 4.9 million years. This is further confirmed by the AV-ESR-02 fragment age data was problematic, and cannot generate consistent results, thus exclude from the final age estimation (Table S1, Supplemental Materials). When looked at stratigraphically the sample from the deepest part of the site AV-ESR-01 give the oldest ages and the one from the youngest part gives the youngest ages AV-ESR-02. All three provide ages within the error of the U-Pb and palaeomagnetic age estimate of 3.0−2.6 Ma (Edwards et al., 2020).

Milo’s cave

As presented in Table 2, for teeth fragment ages, two of the samples (MA-ESR-02 and MA-ESR-04) give similar age estimates while the third one (MA-ESR-05) provides a much younger result. Fragment and powder ages for sample MA-ESR-04 are very similar ranging between 3.14 and 2.74 Ma with US-ESR ages being the youngest (2.79–2.74 Ma) and LU ages giving the oldest age estimates (3.14–3.04 Ma). Fragment ages for sample MA-ESR-02 are also consistent with these age estimates between 3.06 and 2.95 Ma, but powder ages are significantly younger (2.40–2.29 Ma). Age results for sample MA-ESR-03 are excluded from the final age estimations due to the high uranium concentration (See discussion).

D_E difference between enamel fragments and powder

Because fragment ages and powder ages are from the same tooth (same Da), the D_E difference determines the age difference. This section aims to explore the D_E differences between enamel fragments and powder.

The dating results indicate the systematic D_E underestimation from powder to fragments. Two of three D_E estimated by fragments are significantly higher (17∼30%) than the one obtained with powder (Fig. 5). All have larger discrepancies between D_E than the associated errors, making each paired result statistically distinct. Except for sample MA-ESR-04, which shows similar D_E for enamel powder and enamel fragments, all the other D_E results obtained from enamel powder are lower than those obtained from enamel fragments.

Such differences could be due to the following reasons. Firstly, this difference could be due to the presence of the two different types of CO₂^-radical (AIRCORs/NOCORs). These radicals dominant the radiation-induced ESR signal in tooth enamel, and only fragment samples can measure the percentage of AICORs/NOCORs, in agreement with the previous research of ESR age systematically underestimation on enamel powder analysis (Grün, Joannes-Boyau & Stringer, 2008; Joannes-Boyau & Grün, 2011; Joannes-Boyau, 2013). In this consideration, fragments might offer a better estimate of the true age. However, at Aves Cave the enamel fragments appear to give overestimations for the age of the site when compared to U-Pb and palaeomagnetism, although they are still within error of this age estimate because of the larger uncertainties in the dates (±600 − 500 ka). Moreover, powder analysis can provide a bulk age for the enamel because the powder was divided into multiple aliquots for irradiation and measurements. In contrast, each fragment requires repeat measurement and can suffer more from potential localized diagenesis. The AV-ESR-03 fragment age likely suffers from localized diagenesis, because the calculated internal dose of AV-ESR-03 is under zero (−363 ± 810) μGy a-1 and is thus excluded from the final age estimation. Localized diagenesis is more relevant to older sites that have complicated past depositional histories, and imperfect spectra decomposition of fragment analysis can also generate large errors in US-ESR dating. Moreover, spectra decomposition, required for enamel fragments age evaluation, and imperfect spectra merging can also induce significant error. Alternatively, some fragments could be contaminated as there are huge differences between the original isotopic error ratio and no error ratio (Table S1, Supplemental Material).

Secondly, it is also worth noting that powder samples were irradiated by gamma irradiation while fragments were irradiated by X-rays. This difference in protocol explains why the D_E obtained on fossil fragments is frequently higher than those obtained on powder. Gamma irradiation sources have traditionally been used as the main irradiation sources for ESR dating. Research has shown that X-ray sources are a promising irradiation source (Yu, Herries & Joannes-Boyau, 2022). However, the main problem for using an X-ray source is that calibration appears to be a much more complex task than expected. Furthermore, gamma emission is mono-energetic, while X-ray emission by most instruments would have a broader spectrum, with high to low energies photons; which complicates greatly the calibration of the source (Grün, Mahat & Joannes-Boyau, 2012). Another problem is that the ionisation difference between X-rays and Gamma rays is significant. Finally, it is also possible but unlikely that the differences, including the D_E estimation, might be linked to the powdering of the sample.

Scatter ages between models

Although EU and LU models have many problems (Grün, 2009; Shao et al., 2012), ages are calculated by both the US-ESR model in the program USESR (Shao et al., 2014) and the EU model and LU model in the DATA program (Grün, 2009) for simple comparison reasons (Fig. 6). A large discrepancy between different models can be observed. The reason for the large variability in the age estimation for different samples is probably due to the complicated deposition history.

LU models evidently give the oldest ages and EU models seem to give a similar age estimation to the US-ESR model (Fig. 6). Yet, previous studies have preferred the LU model as their best age estimate for ESR dating in the CoH (Curnoe, 1999; Curnoe et al., 2001), correlating well with other methods at Sterkfontein (Herries, 2022). Theoretically, EU models provide what could be seen as the minimum age. It has been argued that LU models suit the CoH context better because of low background dosimetry, and the small uranium concentration in dentine (Dirks et al., 2017; Herries et al., 2020). With small uranium concentration uptake within the fossils, the difference between EU and LU is less when compared to high concentrations models. These scatter ages indicate significant uncertainties in the deposition environment (U-series ages). Bolt’s Farm has experienced important changes in terms of climatic and geological conditions. Climatic changes introduced massive erosion in the whole area and added to postdeposition in the karst events. All these processes have probably led to significant changes in the deposition environment and taphonomy process in the Bolt’s Farm area. The heterogeneity of the depositional environment possibly links to enamel modification during burial and fossilization. Some of the samples have been partly eroded out of the surrounding sediments by surface erosion and the formation of Makondo features into the deposits. While it is likely that teeth from a completely decalcified context, like a Makondo, could result in high uranium content in enamel, and thus age underestimation as seen for some samples from Sterkfontein (Herries & Shaw, 2011), it is possible that partial erosion could have similar effects for parts of the teeth. While the samples were carefully chosen in that they were still mostly encased in indurated sediments, the process of fossilisation and adjacent decalcified could have altered their geochemical parameters (e.g., density) significantly.

The fossil tooth samples from Milo’s Cave and Aves Cave have also evidently experienced some degree of uranium leaching (U-leaching), which could also be triggered by these processes. This could explain why some of the US-ESR ages from the same caves at Bolt’s Farm are producing some variation in the ages. The ²³⁰Th/²³⁴U activity ratio affects the beta dose rate calculation and U-leaching behaves as a higher ²³⁰Th/²³⁴U ratio than equilibrium in a given volume (Duval et al., 2011, 2012). Half of the analysed Bolt’s Farm dental tissues had ²³⁰Th/²³⁴U ratios higher than equilibrium, indicating that uranium had probably leached from these tissues. For this reason, sample AV-ESR-02 was excluded for age calculation due to the abnormal uranium patterns when modelling.

The complicated deposition history may also have led to the high uranium concentration in enamel. For example, MA-ESR-03 shows a significantly higher uranium concentration in enamel, at 14.05 ± 0.71 ppm. The high U-concentrations in the enamel generate the calculation of massive internal dose value, contributes to more than 89% of the total dose rate. Previous research shows low uranium concentrations in sample enamel is essential to the successful application of US-ESR dating of tooth enamel and high uranium concentrations in enamel often cause age underestimations (Duval et al., 2011, 2012). This is because high U-concentrations bring uncertainties of U-uptake and significant alpha efficiency (Grün et al., 2008) and thus it is not possible to model the correct uranium-uptake. MA-ESR-03 is thus excluded from the age results estimation for this reason.

Final age estimates

The ESR ages from Aves Cave appear to provide ages that are in stratigraphic age order in terms of median powder ages, with the lowest sample (AV-ESR-01) giving the oldest ages around 2.9 − 2.8 Ma, AV-ESR-03 the next oldest ~2.8 Ma, followed by AV-ESR-02 at 2.6 – 2.3 Ma. While there maybe issues with sample AV-ESR-02 that are causing underestimation of its age, it still falls within error of the U-Pb and palaeomagnetic chronology of Edwards et al. (2020) at 3.0 − 2.6 Ma, although only at the upper limit of their error range. Fragment ages for AV-ESR-01 suggest ages (3.5 − 3.2 Ma) older than this, although they are still within the error of the 3.0 − 2.61 Ma age estimate.

The ESR ages for Aves Cave and comparisons to other dating methods clearly indicate that ESR is able to date deposits that are at least 3.0 − 2.6 Ma in age. In terms of Milo’s Cave, which has yet to be U-Pb dated, the ages give scattered results for both powder samples ranging from ~3.1 − 2.8 and 2.4 to 2.3 Ma depending on the sample and model used to between 3.1 and 2.1 Ma for fragment ages, despite coming from the same level in the site. In this sense there is some consistency in the ages for both fragments and powder samples overall but often not on the same samples. The only sample that gives similar ages for both powder and fragment is MA-ESR-04 which suggests an age of between 3.0 and 2.7 Ma depending on the model used. The LU model gives the greatest consistency across all the fragment samples at around 3.1 to 2.8 Ma, which is again consistent with the ages for MA-ESR-04 and as such this age is likely the best age estimate.

Preliminary palaeomagnetic analysis on speleothem and sediments indicates a predominance of intermediate polarity at the site with high inclinations characteristic of reversal behaviour, but the short stratigraphic sequence at the site has made defining the nature of the polarity sequence difficult. If this intermediate polarity does suggest the occurrence of a reversal, then there are a number of reversals occurring in the time window suggested by the US-ESR at ~3.03 and 3.11 Ma. Based on the US-ESR ages the site may date to either side of the upper Kaena event reversal at ~3.11 Ma, with the flowstone underlying the fossil deposits correlating to Pickering et al’s (2019) FGI1 stage between 3.19 and 3.08 Ma and the fossil deposits dating to Pickering et al’s (2019) SED1 phase between 3.08 and 2.83 Ma. To date, the Hoogland site (Adams et al., 2010) in the very NE of the CoH and the base of AV01 have been dated to this phase of formation. If the ~3.11 Ma age is correct, this suggests that Milo’s maybe slightly older than Aves Cave, at post 3.03 Ma. However, an age of ~3.03 Ma could also be suggested.

As mentioned above, Gommery et al. (2012) initially suggested the occurrence of Metridiochoerus andrewsi Stage 1 at Milo’s Cave, which would suggest that it is of a similar age to Aves Cave and no older than 3.4 Ma based on correlations with eastern Africa. In South Africa this species is also found at Drimolen Makondo at ~2.61 Ma and the Makapansgat Limeworks at 3.03 to 2.61 Ma (Herries, 2022). The US-ESR ages are thus consistent with this biochronological age estimate. However, Pickford & Gommery (2020) have more recently suggested that fossils of a later stage (III) of Metridiochoerus andrewsi occur at Milo’s Cave, which they suggest would indicate an age of ~1.8 Ma. Pickford & Gommery (2020) note that the defining of the teeth to M. andrewsi (stage III) is based on the overall size of the canine but that it is still encased in breccia and so cannot be measured and analysed fully. Another molar was worn too much to be certain, while another molar fragment is suggested to be consistent in size with M. andrewsi stage III (Pickford & Gommery, 2020). However, a recent reanalysis of this tooth is instead consistent with M. andrewsi (Justin Adams, personal communication). The US-ESR ages are certainly more consistent with the original interpretation of Gommery et al. (2012) that Milo’s Cave dates to the late Pliocene, rather than early Pleistocene.