Accelerator mass spectrometer facility for heritage radiocarbon dating at NRICH, Korea

Abstract

In 2021, the National Research Institute of Cultural Heritage (NRICH) installed a new accelerator mass spectrometer (AMS) to date Korean national heritage. Before conducting radiocarbon dating, the facility was set up and its performance was assessed. AMS system parameters have been optimized. Measurement of standard and blank samples was verified. Intercomparison analyses were also performed using heritage samples. The F¹⁴C value of NIST 4990C was 1.3406 and the background level was 0.0024. Both IAEA-C7 and C8 were confirmed to match the reference value within the 1-σ range. The NIST 4990C measurement results confirmed high precision and accuracy, with 1-σ values below 2‰. In the intercomparison, the error for each sample was 2‰. Thus, based on this study, NRICH plans to launch a dating service specializing in cultural heritage.

Introduction

Radiocarbon dating is a fundamental tool in cultural heritage research because it facilitates the passage of time through the remains of the target material. However, in Korea, no institution specializes exclusively in radiocarbon dating for cultural heritage research. In August 2021, the accelerator mass spectrometer (AMS) was established in Korea to conduct specialized archaeological dating research through the systematic analysis of organic samples excavated from Korean ruins and establish a database. The Conservation Science Division of the National Research Institute of Cultural Heritage (NRICH) analyzes organic materials, such as wood, bone, textile, and organic residues, and inorganic materials, such as pottery, bronze, and iron, using methods such as dating, genetic information, and dietary studies to contribute to the restoration of ancient life and culture and the creation of future value of cultural heritage. In Korea, radiocarbon dating using AMS has been widely employed in archaeology since the early 2000s. This method has facilitated various archaeological applications, including period classification, material culture, technology diffusion, and changes in population distribution (Hwang et al. 2022). It has grown significantly and accumulated more than 17,000 radiocarbon dating data points to date. Consequently, the field of Korean archaeology and cultural heritage is currently plagued with the challenge of reviewing and adjusting existing dates, such as the beginning of the Korean Bronze Age, and the temporal bias of dating results in certain periods, such as the early Iron Age in Korea (Hwang 2021).

Libby (Arnold and Libby 1951; Libby et al. 1949) introduced radiocarbon dating for organic cultural heritage sites. This method has been applied to various sample types. Several studies have focused on cultural heritage materials, such as charcoal, wood, and bone. Wood from culturally significant sites has been used for dating in combination with dendrochronology (Piotrowska et al. 2023). Organic materials such as textiles made from annual plants (Quiles et al. 2023; Hajdas et al. 2014; Nakamura et al. 2019) and seeds (Kutschera et al. 2012), as well as tools made of bones or excavated bones, have been utilized to determine the historical period of a site (Rimkus et al. 2023; Choy et al. 2021; White et al. 2020).

AMS has been utilized for radiocarbon dating of organic cultural heritage (Hellborg and Skog 2008; Jull et al. 2024; Libby et al. 1949). It has been applied for analyzing ultra-trace amounts in the range of 1.0E-12 to 1.0E-15 and can theoretically measure approximately 50,000 BP (Kutschera 1993; Kutschera et al. 2023; Stenström 1995). It has been used for radiocarbon dating studies because it can be measured with as little as 1 mg C, which is 1/1,000 the size of the sample size used in the liquid scintillation counting (LSC) method (Stenström et al. 1996). Therefore, the application of AMS dating to heritage samples offers the advantage of minimizing artificial damage, although it is a destructive analysis.

This study assessed the accuracy and precision of data as indicators of system stabilization. We established certain criteria for tuning and consequently optimized the operating conditions of the equipment using standard samples. The accuracy and precision of the AMS measurement were verified using standard samples from NIST and IAEA. Quality tests were conducted 4 times per year, as per a 3-month cycle, to ensure the accuracy of the data produced by the entire process of our radiocarbon dating system; AMS, reduction system, and pretreatment methods. The quality control study has been conducted along with the commissioning of external organizations since 2021 (Park and Kong 2022). This study discussed the results of quality control using AMS by NRICH from June 2022 to November 2023. In addition, we aimed to ensure the reliability of the dating results via an intercomparison study using cultural heritage samples, to facilitate the research institute’s goals regarding cultural heritage dating research. Testing the stability and performance of the equipment is an important process for ensuring reliable data for future application research. This study utilized the MICADAS made by Ionplus in addition to the EA (Elemental Analyzer), CHS 2 (Carbonate Handling System 2), and AGE3 (Automated Graphitization Equipment 3) (Wacker et al. 2010c). The MICADAS has a maximum acceleration voltage of 200 kV (Synal et al. 2007; Wacker et al. 2010a).

This paper presents the details of the AMS system stabilization in the following order:

1. 1. Optimization of the AMS system.

2. 2. Measurement of the NIST 4990C (modern) for calibration, and phthalic acid for background monitoring.

3. 3. Conduction of system quality tests.

4. 4. Intercomparison study using heritage samples.

Materials and methods

AMS facility

The graphitization system, including the processes of EA, CHS 2, and AGE3, was installed in December 2019, followed by the construction of the AMS laboratory in July 2021. The installation of the AMS at NRICH was completed in August 2021. Our AMS is a MICADAS (Figure 1) manufactured in Switzerland and has a tandem accelerator with a maximum voltage of 200 kV and a Cs-sputter-type ion source (Wacker et al. 2010a). The ion source was equipped with 40 cathodes in a magazine, and He was used as a stripper gas between the accelerator tubes. Two 90° magnets were built into the AMS, with low-energy (LE) and high-energy (HE) magnets located in front and behind the accelerator tube, respectively. Because the measurements are not simultaneous, an LE magnet with an insulated magnet box (pulse) acted as a switch. Carbon ions were measured using Faraday cups for ¹²C⁺ and ¹³C⁺ and a gas ionization chamber with isobutane for ¹⁴C⁺ in MICADAS system (Müller et al. 2015). The measurement reference was set to approximately 50–70 μA of ¹²C LE current. Sample measurements were performed in 17 runs of 15 cycles of 20 s each for monitoring purposes during the instrument optimization study; the first two cycles of the first run were excluded from the analysis for cleaning purposes. Theoretically, a statistical error of 2‰ could be obtained with 250,000 counts of ¹⁴C in NIST 4990C, which was achieved in approximately 4 runs.

Figure 1. MICADAS at NRICH.

The ¹⁴C/¹²C and ¹³C/¹²C ratios measured with AMS can be utilized to calculate F¹⁴C values that are corrected for isotope fractionation effects and used in dating. This normalization was realized using the BATS program (Wacker et al. 2010b).

For beamline tuning, we used NIST 4990C (oxalic acid; Ox2) as a modern standard, and phthalic acid as a blank to check the measurement limits. Both samples were used as indicators of the measurement conditions to establish the appropriate environmental conditions through instrument tuning. The standard was set such that the ¹⁴C/¹²C ratio of Ox2 was within the range of 1.49 $ \times $ 10⁻¹²–1.54 $ \times $ 10⁻¹² and ¹³C/¹²C was within the range of 1.07–1.08%.

To evaluate the optimization of the equipment, the beam transmission was checked, and the measurement conditions were established accordingly. The beam transmission can be checked using two ¹²C measurement Faraday cups for measuring ¹²C behind the LE and HE magnets, respectively. Permanent magnets are used in AMS. The bending radius is 25 cm in LE magnet, and 35 cm in HE magnet (Synal et al. 2007).

Sample preparation

Five standards, one blank, and two heritage samples (Table 1) were used in this study. The heritage samples were collected from the Four-Fold Screen of the Sun, Moon, and Five Peaks (SMFP) in Changdeokgung Palace. As a painting embodying the principles of Eastern Yin Yang and the Five Elements philosophy, the SMFP symbolizes the universe. It was used as a screen behind the throne during the Joseon Dynasty (1392∼1910 AD), symbolizing the majesty of the monarchy and the prosperity of the dynasty. It was placed behind the royal throne or royal portrait of the palace. In the backing paper, it was noted 1840 AD. Ox2, IAEA-C7, C8, and phthalic acid were directly combusted using EA and then graphitization of sample CO₂, using AGE3, was performed. In total, 122 phthalic acid samples were measured. The number of measurements for Ox2, IAEA-C7, and C8 were 140, 45, and 68, respectively. Both cultural heritage samples, the “wood sample” and “paper sample,” were dated by our laboratory and another analytical laboratory. Figure 2 shows a wood sample. Only the second ring section was used in this study after removing the most recent ring. Laboratory quality control utilized IAEA-C9 and C5 samples, wherein both acid-base-acid (ABA) and acid-base-acid-bleaching (ABA-B) treatment methods were used (Table 2). These methods were adapted from the Oxford Radiocarbon Accelerator Unit (ORAU) (Brock et al. 2010). They are commonly used by researchers, depending on the condition, amount, or material of the sample. This study aimed to assess the stability of the system using both pretreatment methods.

Table 1. Sample list

Figure 2. Collected wood sample from SMFP Screen in Changdeokgung. (a) Full shot of the sample, including the wood frame with the residual backing paper. (b) and (c) Wood sample used for this study. The red dashed circle in (a) indicates the sampling point. (c) Photograph with a 1 mm scale bar.

Table 2. Two treatment methods, ABA and ABA-B

The ABA treatment ([a] in Table 2) was chosen based on prior research on various pretreatment methods for dating heritage samples (Park et al. 2023). After each step, the samples were rinsed with Milli-Q water for neutralization. Consequently, the samples were dried in a vacuum oven or lyophilizer and stored. The code numbers for each sample are presented in Table 3.

Table 3. Wood and paper from heritage sample code list

EA-AGE3 (Figure 3) was used for the preparation of resulting graphite, and the cathode was fabricated using PSP (Pneumatic Sample Press). To extract carbon from the sample, 4 mg of Fe powder was prepared in the reactor and AGE3 was used to convert it into Fe₂O₃ by reacting with O₂ in air at 500°C. Then, it was preheated with H₂ gas to leave only pure Fe. Subsequently, the sample was divided into small portions and combusted with oxygen using an EA. Approximately 1 mg C was extracted by reducing the resulting CO₂ gas with H₂ and Fe catalyst at 580 °C in the AGE3 reactor (Wacker et al. 2010c). The resulting graphite with a Fe catalyst was stored at a relative humidity of less than 10%. Finally, the target was compressed into the cathode using the PSP and measured using the AMS.

Figure 3. (a) Picture of AGE3 and (b) Diagram of combustion using EA and graphitization using AGE3.

Results

To stabilize the system, we optimize, review the standard sample data, estimate reliability periodically, and test heritage samples.

The measurement conditions were optimized in the first step. This was achieved by tuning the beam line from the ion source to the detector to ensure that the ion beam was centered along the z-axis in the direction of the ¹⁴C ion. The beam trajectory was verified by scanning the magnetic-field strength and adjusting the current in LE and HE magnets. Figure 4 shows the results of the LE and HE magnet scans. Under these conditions, Ox2 had ¹⁴C/¹²C ratios ranging as 1.4922–1.5522 E-12 and ¹³C/¹²C ratios ranging as 1.076–1.088%. We investigated the minimum number of ¹⁴C counts required to achieve a measurement accuracy of 2‰ for Ox2. Our results indicated that more than 750,000 ¹⁴C counts were required.

Figure 4. Results of magnet scanning. (a) Results of LE Magnet and (b) HE Magnet. We optimized the condition to center the beam in the beamline. ^xC indicates stable carbon isotopes, ¹²C and ¹³C.

Next, the accuracy and precision of the data and background level of the system were checked using standards and a blank sample before performing the intercomparison study. The graphs in Figures 5–8 show the red line as the standard value, and the dashed line represents the standard error. The measurement result is represented by black squares (▪) and order is represented on the x-axis; the most recent data are presented on the right.

Figure 5. AMS results for NIST 4990C (Ox2). The red dashed line represents the 2‰ error range.

Figure 6. AMS results of background.

Figure 7. AMS results of IAEA standard samples (IAEA-C7&C8). The red dashed line represents the 1-σ error range.

Figure 8. Results of F¹⁴C values of IAEA-C5 and C9 for system quality testing. We divide a year into four quarters. They were measured in the 3rd quarter of 2022 and 2023(22Q3–23Q4).

The 1-σ for all NIST-4990C was approximately 1–2‰, and the average value for F¹⁴C was approximately 1.3406, which closely matched the reference value (Figure 5). Each measurement value was within the 1-σ. The average background level (phthalic acid) was 0.0024 (± 0.0001 $ \pm 0.0001$ ) F¹⁴C (Figure 6). The dating limit was calculated by substituting Equation (2) into Equation (1) (Stuiver and Polach 1977). The dating limit calculated using our system was 53,951 BP.

(1) $${\rm{t}} = - 8033 \times \ln \left( {10 \times {\sigma _{bl,\;ave}}} \right)$$

(2) $${\sigma _{{bl},\;{ave}}} = {{\sum\nolimits_\;^N {{\sigma _{bl}}} } \over {N - n}}$$

Where N is the number of total samples, n is the number of measurements, and σ is the measurement error of each sample. The number of measurements (n) was 19 and the total number of samples (N) was 122.

The AMS results of the IAEA-C7 and C8 are shown in Figure 7. IAEA-C7 yielded an overall average value of 0.4946, which was within 1.9‰ of the certified value. Whereas, IAEA-C8 yielded an average of 0.1503, which was within 1.5‰ of the certified value. Both samples exhibited values that closely matched those of the standard sample within the 1-σ error range, and the error from the certified value was within 2‰. This indicated proper management of the system.

The results of the quality testing are presented in Figure 8. Both the treatment methods produced data consistent with the standard value within a 1% margin of error. On the IAEA-C5 results (in Figure 8), the reference data consists of 49 samples (n = 49), whereas our dataset includes only 5 samples (n = 5). We suspect that there may be differences influencing our results, and continuous monitoring and additional measurements are necessary to better understand and address these potential discrepancies. Despite this, we calculated the sigma value for our data and found that the values fall within the 1-σ range which is standard deviation. Specifically, the 1-σ value is 0.0014, and our data fits within this error margin. The red dashed line in Figure 8 represents 1-σ range.

To ensure the reliability of the analysis, we conducted an intercomparison study on the heritage samples following the standard samples (Figure 9). For wood, we compared the results of all four measurements with the externally analyzed ‘SW001’ sample for intercomparison. All data were consistent within 2‰ of the mean. The differences between the mean and each measurement ranged from 0.0001 to 0.0037. For the study sample, we obtained consistent results within the margin of error, as it was a single-year sample. The difference between the average and each measurement in this study was 0.0012. Calibrated data obtained using OxCal are shown in Table 4 (Reimer et al. 2020).

Figure 9. Results of heritage samples as F¹⁴C. y-axis is F¹⁴C, x-axis is the sample codes. The red solid line represents the mean of data, and the red dotted line represents an error of 2‰.

Table 4. Dating results for heritage samples calibrated by OxCal v4.4.4

Conclusion

In August 2021, the NRICH established an AMS for dating research on Korean Cultural heritage and conducted stability tests. This study confirmed the reliability of the production data, indicating the successful establishment of a dating research laboratory. Background level measurements showed that F¹⁴C was 0.0024, which is equivalent to 53,951 BP. The measurement results of IAEA-C7 and C8 confirmed the high accuracy of the data, with errors within 2‰ of the certified value and a precision within 2‰. To obtain high precision in the shortest possible measurement time, we found that conducting 10 runs of 15 cycles of 20 s each was sufficient to achieve 750,000 counts in our measurement. This method will be incorporated in future measurements.

The reliability of NRICH AMS’s data was confirmed by conducting an intercomparison to validate the operating parameters. Utilized cultural heritage samples, confirmed the high accuracy of data. We obtained precise results with an uncertainty within 1‰ for intra-laboratory testing. There was a variance of approximately 2‰ compared with the intercomparison results.

The background levels were found to be inconsistent, with certain readings being 1.5 times higher than the others. The maintenance of a stable background level is crucial for the dating of heritage studies. In future studies, we will investigate the cause of this inconsistency and attempt to maintain stable levels with high precision.

NRICH has studied general pretreatment methods for wood, charcoal, bone, cremated bone, paper, textiles, and lacquer for radiocarbon dating. Our research will continue by examining various heritage samples and conducting further verifications. In addition to verifying various heritage samples, we plan to study pretreatment methods for seeds and other materials to expand the scope of radiocarbon dating techniques. Based on this, we aim to conduct reliable dating studies to contribute to the scientific analysis of Korean cultural heritage.