The Micro-Specific Graphitization Reactor

The micro-specific graphitization reactors are based on the micro conventional furnace (MCF) design at ANSTO (Yang et al. Reference Yang, Smith and Hua2013; Yang and Smith Reference Yang and Smith2017). The MCF is a cost-effective solution for producing graphite from carbon dioxide sample gas and is connected to a vacuum line for the synthesis of graphitic carbon in 1.2 mL fixed reaction tubes using hydrogen reduction. Reducing the internal volume of reactors is an effective approach for graphitizing smaller samples while minimizing isotopic fractionation (Pearson et al. Reference Pearson, McNichol, Schneider, Von Reden and Zhang1998; Santos et al. 2007; Yokoyama et al. Reference Yokoyama, Koizumi, Matsuzaki, Miyairi and Ohkouchi2010). Our device comprises two small-volume graphitization reactors connected by quartz manifolds and stainless steel cold fingers suitable to trap water vapor during the reaction. Samples with 10–300 μg C can be synthesized in a tube furnace at 600°C, adding reduced iron powder and ultrapure hydrogen (H₂), these reactors can also produce a few micrograms of graphite (<10 μg C).

The performance of the micro-specific graphitization reactors was assessed by analyzing ultra-small standard samples, including the National Institute of Standards and Technology (NIST) standard oxalic acid II (OxII; SRM 4990C) as well as secondary standards - International Atomic Energy Agency (IAEA) ¹⁴C reference materials (IAEA-C2, IAEA-C3, IAEA-C6, IAEA-C7) and background samples (calcite, Sigma Aldrich graphite, 99.99% pure, -100 mesh, anthracite, CO₂ background gas). Many smaller CO₂ standard gases were obtained by combustion or hydrolyzed of large quantities of the standard materials with 85% orthophosphoric acid (H₃PO₄) and subsampled to equal 5–150 µg C for graphite synthesis on the ultra-small reactors.

Performance of Micro-Specific Graphitization Reactors

We analyzed the effects of different catalysts and reducing agents involved in the graphitization of microgram carbon samples based on Scanning electron microscope (SEM) images and AMS measurements to determine appropriate conditions for the synthesis of ultra-small mass graphite.

Efficiency of Graphitization with Different Fe Catalysts

Hydrogen-catalyzed reduction reactions were conducted using four Fe catalysts: (1) Sigma FeO: Fe₂O₃ powder (Sigma Aldrich, ≥99.995% pure), (2) Acros FeO: Fe₂O₃ powder (Acros organics, 99.999% pure, -100 mesh), (3) Sigma Fe: Fe powder (Sigma Aldrich, 97% pure, -325 mesh), and (4) nano FeO: Fe₂O₃ powder. These Fe catalysts were processed by first reducing Fe or Fe₂O₃ in H₂ at 600°C before graphitization. Subsequently, AMS measurement and micro-scale morphological analyses were conducted to determine optimal reagents. This section aims to establish the optimal reagents and suitable mass of carbon for micro-specific graphitization reactors by synthesizing and analyzing graphite samples with different carbon masses.

First, we examined the graphitization reaction curves of CO₂ gas samples with various carbon masses using the four different Fe catalysts described above. The reaction curves for the first three Fe catalysts are shown in Figure 1 as residual gas pressure during graphitization. Nano-FeO is not shown due to low yields (20–30%). Sample masses in Figure 1 range from 34 -5.82 μg C. “0” on the time-scale in Figure 1 denotes the moment when the graphitization reaction tube was pushed into the 600 °C furnace, indicating the warming up period before the graphitization reaction, during which the gas pressure first rises rapidly due to the temperature increase. All of the curves reach a peak within one minute. From this point, the pressure of all the gases begins to drop, which means the true beginning of all graphitization reactions, two types of curves are observed: (1) a rapid graphitization that is completed within 15 minutes (Acros FeO and Sigma FeO); and (2) a relatively slow graphitization that is completed within 45 minutes (Sigma Fe). The reaction time also depends on the sample mass. A sample with 5 μg C can be processed in five minutes. We conclude that Acros FeO and Sigma FeO have the highest efficiency with nearly 100% reaction yields.

Figure 1 Graphitization reaction curves for a variety of CO₂ gases with varying carbon size catalyzed by various Fe reagents. The solid line for Acros FeO, the dashed line for Sigma FeO, and the dotted line for Sigma Fe are used to depict the three Fe catalysts, and the carbon size is decreased sequentially from top to bottom.

SEM Morphological Analysis of Different Fe Catalysts

Sigma Fe catalyst is used routinely at the Xi’an lab for graphitization in standard ¹⁴C analyse. However, for ultra-small samples, FeO catalysts appear to perform better. SEM analysis of graphites produced in our experiments was performed on the Fe catalysts and Fe-C mixtures using a Zeiss EVO18 multipurpose SEM analysis system. The results are shown in Figure 2 with 2-μm resolution. Graphite for these tests was synthesized from the IAEA-C2 reference standard with carbon mass above and below 20 micrograms. Another graphite with a similar C mass was synthesized and measured by AMS to measure ¹²C³⁺ currents and fraction modern C (F¹⁴C) values. SEM images of several activated Fe powders revealed that most exhibit relatively smooth surfaces with both types of FeO. These showed a relatively large surface area, as compared to Sigma Fe which may explain the relatively slow reaction time of the latter.

Figure 2 The SEM images of iron powder before and after graphitization.

A black graphite-coated iron powder is produced during the graphitizations and this form contributes to strong and reliable ion currents for AMS analysis along with accurate and precise F¹⁴C values (Kim et al. Reference Kim, Kelly and Clifford2008). Although the mass of carbon is small, a visible amount of graphite was observed on the surface of Fe powder in the SEM images (Figure 2), no apparent difference exists between the images of different carbon masses, however, ¹²C³⁺ beam current and F¹⁴C values (reference value: 0.4114 ± 0.0003) of the IAEA-C2 reference standard targets measured by AMS were significantly different. All the targets >20 µg C had relatively high ¹²C³⁺ currents and F¹⁴C values close to their reference values, whereas all the targets < 20 µg C produced relatively low ¹²C³⁺ currents and F¹⁴C values, further from their reference values. Based on the SEM images, Sigma FeO and Acros FeO catalysts appear to be better than Sigma Fe (based on surface area), and they outperformed Sigma Fe, based on reaction rate analysis, with varying carbon sizes.

The Sigma FeO catalyst was selected as our catalyst of choice since it demonstrated superior reaction rates and consistent F¹⁴C values for samples<20 µg C. With this dedicated trace graphitization system, all CO₂ gas samples above 5 µg C can be synthesized, and the optimal amount of carbon for the analyzed samples lies within the range of 10–200 µg C.

AMS Measurements

All samples were analyzed using the 3 MV multi-nuclide Accelerator Mass Spectrometer at the Xi’an AMS Center, Institute of Earth Environment, Chinese Academy of Sciences. Under routine ion source conditions producing∼20 μA ¹²C⁻ from graphite injected into the tandem accelerator, the ion injection energy used was 35 k eV, and a fast 100 Hz cycling frequency was routinely employed in the Fast Sequential Injection (FSI) mode that alternates ¹²C^–, ¹³C^– and ¹⁴C^– injections. The isotopes: ^{12, 13, 14}C³⁺ were measured in the high-energy end of the machine with a terminal voltage of 2.5 MV. The best background of ¹⁴C/¹²C was 1.84 × 10^–16 with a long-term daily ¹⁴C/¹²C procedural background range for 1 mg C in the range of 9 × 10^–16 to 2 × 10^–15 (Figure 3; Zhou et al. Reference Zhou, Zhao, Lu, Liu, Wu, Cheng, Zhao and Huang2006, Reference Zhou, Lu, Wu, Zhao, Huang, Li, Cheng and Xin2007, Reference Zhou, Wu, Lange, Lu, Cheng, Xiong, Cruz, Liu, Fu and Zhao2012).

Figure 3 Plot of the average ¹²C³⁺ beam current versus the sample sizes. The linear fitting of ¹²C³⁺ beam current and the carbon content (%) for three ranges of sample sizes (<20 μg C, 20–100 μg C, and 100–200 μg C) is shown in small figure. The black dots and black line for the range of 100–200 μg C, the blue squares and blue line for the range of 20–100 μg C, the pink triangles and pink line for the range of <20 μg C. (Please see online version for color figures.)

According to the carbon mass of samples, the analyzed graphite samples were divided into wheels of >20 µg C targets and <20 µg C targets, and targets with activated Fe₂O₃, Fe powder, and Nd powder were inserted into each batch of samples to monitor ¹⁴C counting interference. Sample batches consist of a set of small standards with the same size as the unknown samples, analyses of IAEA reference materials and background materials of the same type as unknown samples, as well as OxII standards that are used for quality and consistency control and standards normalization. Measured ¹⁴C/¹²C ratios are reported as F¹⁴C. The data were analyzed and performed the correction for isotope fractionation (normalization to δ¹³C = –25‰ VPDB). Uncertainties of our data are fully propagated for each correction.

Beam Current Investigation

Figure 4 shows the ¹²C³⁺ beam currents that we measured on the AMS. The ¹²C³⁺ beam current intensity of 1 mg C samples was 20–30 µA and decreased to a few µA or even below 1 µA as the carbon mass of the sample decreased (Figure 4). In general, we fix the C/Fe mass ratio for conventional mg-scale samples to 1:2 (Zn method) or 1:3 (H₂ method). To make it easier to load ultra-small mass samples into the holder, we used a constant amount of iron (1 mg Fe or 1.43 mg Fe₂O₃) regardless of carbon mass. Hence, the C concentration in the targets decreases with carbon mass. This dilution caused the ¹²C³⁺ beam current to decrease with decreasing sample mass. We achieved a maximum ¹²C³⁺ beam currents of approximately 0.1 µA per µg of carbon. The carbon concentration in conventional 1 mg graphite targets was about 25% (calculated by the C/Fe mass ratio of 1:3), corresponding to an anion beam current of 20–30 µA. For samples of 200 µg C or more, the ¹²C³⁺ beam current was stable and above 20 µA, which is consistent with regular 1 mg C targets. For samples of less than 200 µg C, the ¹²C³⁺ beam currents decreased sharply with decreasing sample mass (from 20 µA→1 µA or less), which was primarily attributed to dilution of the carbon concentration in the graphite target (from 25%→1%), and ¹²C³⁺ beam currents correlated linearly with sample mass. The maximum observed ¹²C³⁺ beam current to mass ratio was approximately 0.1 μA/1 μg C. A relatively uniform distribution of ¹²C³⁺ beam currents for samples was observed ranging from 20 to 200 µg C, and the distribution was linear for samples with masses between 1 and 20 µA. For ultra-small samples of <20 µg C, ¹²C³⁺ beam currents were more dispersed, especially for samples smaller than 10 µg C, where the ¹²C³⁺ beam current was below 1 µA and did not correlate well with sample mass.

Figure 4 (a) Fraction modern C values for four ¹⁴C-free blanks samples from 0.005 mg C to 0.3 mg C sample size are shown to quantify modern carbon contamination (MCC). The four ¹⁴C-free blanks represent the MCC from four different processes for the ¹⁴C micro-samples analysis: graphitization (¹⁴C-free CO₂ gas), surface leaching (carbonates: calcite), combustion and purification (organics: graphite), conventional acid/base/acid washes process (organics: anthracite). (b) ΔF¹⁴C values for small and ultra-small OxII samples from 0.003 mg C to 1 mg C measured against 1 mg C normalizing OxII standards are shown to quantify the dead carbon contamination (DCC). Note that uncorrected ¹⁴C results from smaller OxII samples are always depleted. The solid lines in both plots represent fixed amounts of carbon contamination from 0.1 to 2 µg C.

By linear fitting of ¹²C³⁺ beam current and the C% for three sample sizes range (<20 μg C, 20–100 μg C, and 100–200 μg C), it is found that all of them had a strong linear relationship, especially for the sample sizes of 100–200 μg C, for which the linear relationship is the most significant (R²=0.9894). These linear relationships become weaker as the sample size decreases, it is possible that be related to the uncertainty of analytical results caused by the small sample size, the smaller the sample size, the greater the uncertainty introduced into the analytical results. It could be brought on by uncertainties of various apparatus such as pressure gauge, analytical balance, and others. Therefore, a miniature pressure transducer and a 1.2 mL fixed reaction tube are used to minimize this uncertainty of measurement in our reactor, and the uncertainty introduced by this measurement was considered to be small. It is also possible that by the sample purity, the introduction of contamination can also influence the accuracy of sample mass. While the exogenous contamination introduced during the analysis can be solved by taking multiple measurements of different small standard materials at the same time and applying a suitable correction model.

Quantity of Exogenous Carbon Contamination

Exogenous carbon contamination is generally made up of two major components: (1) modern carbon contamination (MCC) and (2) ¹⁴C-dead carbon contamination (DCC), both of which are introduced during sample preparation. The contamination can come from chemical reagents, quartz tubes for loading samples, pretreatment protocols, and poor vacuum, and the uncertainty introduced by this measurement was considered to be small. The proportion of carbon contamination increases as the sample mass decreases, especially for targets with <0.1 mg C, amplifying the effect of contaminations and causing ¹⁴C results and δ¹³C data to significantly deviate from true values. A blank correction has become an indispensable step in the analysis of ultra-small samples due to the increased uncertainties associated with ultra-small samples using AMS.

Several researchers have proposed models for this correction calculation. For instance, Santos et al. (Reference Santos, Moore, Southon, Griffin, Hinger and Zhang2007a) developed formulas to separately correct the contribution of MCC and DCC from unknown samples as small as 0.001 mg C (Santos et al. Reference Santos, Southon, Griffin, Beaupre and Druffel2007b, Reference Santos, Southon, Drenzek, Ziolkowski, Druffel, Xu, Zhang, Trumbore, Eglinton and Hughen2010). By analyzing systematic trends of standard and background normalization results from a similar batch, the MCC content was quantified using smaller ¹⁴C-free background samples (3–200 µg C) in the same batch. For this purpose, four background materials were analyzed representing four different chemical pretreatment procedures, including anthracite, calcite, and graphite, and ¹⁴C-free CO₂ gas (Figure 5a). The trend of the four background F¹⁴C values with varying sample sizes are discussed, the solid lines represent the linear effects of several fixed amounts of MCC contamination ranging from 0.1 to 2 µg C, revealing an average MCC of 0.2–1.2 µg. The fraction modern carbon values of the four backgrounds reflect different sources of MCC contamination introduced from conventional ¹⁴C-AMS experimental processes (e.g., acid-base-acid pretreatment of organics, combustion and purification, acid-hydrolysis of carbonates, and graphitization). The estimated quantity of our MCC involving different procedural stages are presented in Table 1. The chemical pretreatment process of ¹⁴C-AMS samples must be carefully monitored for ultra-small samples. DCC is calculated from ΔF deviation values of the measured fraction modern carbon for small aliquots OxII samples from the primary standard (1 mg C OxII) (Figure 5b). Similarly, several black solid lines represent the effects of several fixed amounts of DC contamination. Our DC contamination is in the range of 0.1–1.5 µg C.

Figure 5 The fitting curves of the reciprocal relationships between our F¹⁴C results and sample size for various standards. (a) OxII standards from 0.003 mgC to 1 mgC. All FC results have been corrected for fractionation using δ¹³C AMS measurements. (b) Four background materials: anthracite, graphite, calcite, and ¹⁴C-free CO₂ gas.

Table 1 Summary of carbon contamination masses in our ¹⁴C-AMS analysis of ultra-small samples.

The MCC of the KCCAMS laboratory was 0.6 ± 0.3 µg C, and DCC was 0.3 ± 0.15 µg C (Santos et al. Reference Santos, Southon, Drenzek, Ziolkowski, Druffel, Xu, Zhang, Trumbore, Eglinton and Hughen2010). The values of the single-stage AMS laboratory at Tokyo University were 0.37 ± 0.13 µg C (MC) and <1.62 µg C (DC) (Masako Yamane et al. Reference Yamane, Yokoyama, Hirabayashi, Miyairi, Ohkouchi and Aze2019). By comparing with data published by other laboratories, it is found that our DC blank contribution was quite significant, this could be attributed to our vacuum system for pumping samples and the chemical reagents used. It is suspected that the use of oil in our vacuum pumps has a negative impact, which, however, needs to be investigated in detail. The ¹⁴C content of extraneous carbon in CuO (an oxidation reagent) can vary significantly between batches and suppliers. CuO promotes more contamination than Fe powder (Q. Hua et al. Reference Hua, Zoppi, Williams and Smith2004) because more CuO is used for sample preparation, therefore, having a much higher contamination potential. In the future, oil-free pumps and higher purity reagents will be considered to adopt to minimize these contamination contributions during sample processing at the Xi’an AMS center.

Furthermore, we explore the constant contamination model to precisely quantify the F¹⁴C of the contamination (F¹⁴C_c) and the mass of the contamination (m_c), the F¹⁴C measured by AMS (F¹⁴C_m) can be expressed as:

(1) $${\rm{F}}_{}^{14}{C_m} = {{{\rm{F}}_{}^{14}{C_s} \times {m_s} + {\rm{F}}_{}^{14}{C_c} \times {m_c}} \over {{m_m}}}$$

Where the F¹⁴C_s is the actual F¹⁴C of the sample, and m_m denotes the total measured C mass, which is the sum of the actual mass of the sample (m_s) and m_c. From Equation (1) and derivation, we obtain the following equation:

(2) $${\rm{F}}_{}^{14}{C_m} = {\rm{F}}_{}^{14}{C_s} + {m_c} \times {{{\rm{F}}_{}^{14}{C_c} - {\rm{F}}_{}^{14}{C_s}} \over {{m_s}}}$$

The impact of constant contamination is reciprocally related to the sample size, its mass is assumed to be much smaller than that of the sample (mc ${\rm{\;}} \ll $ ms) (Hua et.al. Reference Hua, Zoppi, Williams and Smith2004). Since the values of F¹⁴C_s、F¹⁴C_c and m_c are constant, F¹⁴C_m is inversely related to m_s, here F¹⁴C_s and m_c × (F¹⁴C_c-F¹⁴C_s) can be regarded as the offset and coefficient of the F¹⁴C_m and m_s relationships, respectively. The reciprocal curves of OxII and four background samples between our F¹⁴C_m results (corrected for δ¹³C fractionation) and sample size are fitted (Figure 6), and the quantity of our m_c and F¹⁴C_c involved different procedural stages are calculated and presented in Table 1. Based on a two-component mixing model explored in the publication of Hua et.al. (Reference Hua, Zoppi, Williams and Smith2004), the F¹⁴C of our several IAEA reference materials are corrected, the corrected results are illustrated in Figure 8 and discussed in the section “Accuracy and Reproducibility.”

Figure 6 The ¹⁴C counts rate (CPS, counts per second) and ¹²C³⁺ current values of the reagents used in the ultra-small graphite holders. The blue bar chart represented the CPS values of before and after reduced Fe₂O₃ reagents, the brown curves represent ¹²C³⁺ current values of before and after reduced Fe₂O₃ reagents. The x-axis represents many measurements of targets which were made of Fe and Fe₂O₃ powder in measured wheels.

Results of F¹⁴C_c and m_c for the different procedural stages of our ¹⁴C-AMS analysis are obtained more precisely based on the mass-balance model with the F_c and m_c of the four background standard materials in Table 1. These results include a carbon contamination mass of 0.07 ± 0.01 μg of C with an F¹⁴C_c value of 1.33 ± 0.37 introduced from the acid-base-acid chemical pretreatment process for organics, a mass of 0.11 ± 0.02 μg of C with an F¹⁴C_c value of 1.39 ± 0.37 introduced from the combustion process, a mass of 0.04 ± 0.008 μg of C with an F¹⁴C_c value of 1.36 ± 0.19 introduced from acid-hydrolyzation process for carbonates, and a larger mass of 0.54 ± 0.10 μg of C with an F¹⁴C_c value of 0.44 ± 0.09 introduced from graphitization and AMS measurements process. It is concluded that m_c of graphitization is the most, and more dead carbon contamination is introduced from this process (F¹⁴C $ \ll $ 1).

Iron catalyst is a key participant in the entire graphitization and AMS measurements, the most probable source of dead carbon contamination can be iron catalyst (Cherkinsky et al. Reference Cherkinsky, Prasad and Dvoracek2013). Correction for contamination of dead and modern carbon which could be present in the catalyst or from cross-contamination in the ion source also plays a significant role, it should not be neglected especially in the ¹⁴C-AMS analysis of ultra-small samples (<20 μg C). Therefore, our measurement results of small samples are performed using a correction method of ¹⁴C counts from iron.

Correction of ¹⁴C Counts from Fe Powder

In the measured wheels of ultra-small samples, targets made of Fe powder, Nb powder, and Fe₂O₃ were also added. Fe powder and Fe₂O₃ powder were prepared in two ways: (1) pre-activated and (2) post-activated. Pre-activated targets were heated to 600°C for 2 hr. Post-activated targets were reduced with H₂ at 600°C. We measured the ¹⁴C count rate and ¹²C³⁺ beam currents for these targets. The ¹⁴C count rate of Fe powder and ¹²C³⁺ beam currents of the pre-activated and post-activated targets are shown in Figure 7. The ¹⁴C count rate and ¹²C³⁺ beam currents have changed before and after activated Fe₂O₃. Post-activated targets had higher ¹⁴C count rates. This implies that atmospheric ¹⁴CO₂ (of modern origin) could have contaminated the catalyst during the reaction process. Therefore, we introduce a correction for the exogenous carbon contribution during graphitization by subtracting ¹⁴C counts from Fe powder alone. The corrected F¹⁴C was obtained from the subtraction of ¹⁴C counts for all samples measured in the same batch respectively, which also includes various standards measured in the same batch the measured, and further calculations with formulas involved in the conventional ¹⁴C-AMS analysis.

Figure 7 (a) F¹⁴C results for multiple ¹⁴C-AMS measurements of OxII samples between 0.004 mg C and 1 mg C are shown. The uncorrected F¹⁴C values are represented by the black hollow triangle and the corrected F¹⁴C values by the black solid triangle. The dashed line represents the consensus value of OxII. The horizontal dark gray (light gray) band indicates the 1σ (2σ) uncertainty calculated from all F¹⁴C results corresponding to the range of each sample size. (b) F¹⁴C results for background samples between 0.003 mg C and 0.2 mg C from multiple ¹⁴C-AMS measurements are shown. The uncorrected F¹⁴C values are represented by hollow symbols, and the corrected F¹⁴C values by colored solid symbols. Different materials have different symbols and colors: anthracite with square and blue), graphite with Rhombus and yellow, calcite with circle and gray, ¹⁴C-free CO₂ gas with triangle and orange. The dashed line represents several ¹⁴C age boundaries corresponding to F¹⁴C values respectively, including 10 ka, 30 ka, 40 ka, and 50 ka. All results have been corrected for fractionation using δ¹³C AMS measurements, also compared the F¹⁴C values before and after ¹⁴C counts calibrated from Fe powder for all background samples.

After correction, F¹⁴C values of the OxII standard and four background materials with various carbon masses are compared with measured values (Figure 7). Figure 7a shows the variation in F¹⁴C values of the OxII standard with varying carbon masses, as well as corrected F¹⁴C values. All results have been corrected for fractionation using δ¹³C-AMS measurements (including machine-induced isotopic fractionation). To precisely clarify the reliability of measurements, the results with three ranges of sample size are analyzed alone. F¹⁴C results ranging from 0.1 mg C to 1 mg C mostly fall on the line of the reference value and lie within the 1σ error of all results of measurements in this range. The average F¹⁴C of 1.3410 ± 0.0051 analyzed the corrected results of sample size with 0.1–1 mg C agrees with the consensus value of F¹⁴C =1.3407 ± 0.0005, with a precision of 3.8‰, which is almost no big difference with of the conventional ¹⁴C-AMS analysis. For the range of 0.01–0.1 mg C, the F¹⁴C values have the larger error and are more scattered, F¹⁴C results generally fall within the 2σ error of all results of measurements in this range. The average F¹⁴C of 1.3375 ± 0.0186 (n = 30) analyzed the corrected results of sample size with 0.01–0.1 mg C has some deviation from the consensus value of OxII standard, with a precision of 1.39%. For ultra-small samples of <0.01 mg C, improvement range is larger after correction, but the deviation from consensus remains larger. The ¹⁴C count correction from Fe powder does not completely solve the correction problem of C contamination for this range, it is implied that preparing samples of <0.1 mg C is very challenging, probably it is necessary to improve the accuracy via strictly controlling the procedural background in the future.

Figure 7b illustrates the variation in F¹⁴C values of all background samples with varying carbon masses, and the corrected background F¹⁴C results. The F¹⁴C values of small background samples increase with decreasing carbon masses. For small samples more than 0.1 mg C, the background ¹⁴C ages remain above 40 ka even for ¹⁴C–free CO₂ gas close to 50 ka after ¹⁴C counts correction. Before the correction, the background ¹⁴C results for small samples of 0.1–0.02 mg C generally ranged from 30–40 ka, whereas the ¹⁴C results of ¹⁴C-free CO₂ gas samples appeared to be higher, mostly around 40 ka; as carbon mass decreases, the background F¹⁴C values gradually increase, with the ¹⁴C age of about 30 ka. With correction, extraneous carbon contamination is removed during the graphitization process, resulting in smaller F¹⁴C values and higher background ¹⁴C ages, specifically for background gas. For ultra-small samples of <0.02 mg C, the background F¹⁴C values rapidly increase with decreasing carbon masses, which is generally scattered between 20 ka and 35 ka. After correction, the ¹⁴C age was close to 40 ka, emphasizing the importance of ¹⁴C count correction from Fe powder, which is required urgently for ultra-small samples, specifically those near background levels of less than 0.02 mg C. Consequently, the good agreement of the measurements of OxII and four background standards with their reference values demonstrates the improvement of contamination correction and a good precision of ultra-mass graphite preparation and AMS measurement for small samples of >0.01 mg C at Xi’an AMS center.

Accuracy and Reproducibility

To assess the accuracy and reproducibility of our micro-scale ¹⁴C-AMS analysis, we prepared and measured four reference materials from the International Atomic Energy Agency (IAEA), IAEA-C2 (travertine, 7135a BP), IAEA-C3 (cellulose, modern sample), IAEA-C6 (ANU sucrose, modern sample), and IAEA-C7 (oxalic acid, 5644a BP), covering a range of 0.01–0.1 mg C. The IAEA series can help in evaluating different correction methods and measurement performance of small samples. The corrected F¹⁴C values applying three methods for our IAEA reference materials are compared, including (1) the traditional background subtraction: The F¹⁴C values of samples are calculated by the direct subtraction of the F¹⁴C values of background measured in the same batch; (2) the constant contamination model: Based on a two-component mixing model explored in the publication of Hua et al. Reference Hua, Zoppi, Williams and Smith2004; and (3) our correction of ¹⁴C counts from iron. The three corrected F¹⁴C results with the same sample sizes are illustrated in Figure 8. The corrected results with method (2) are mostly higher than the other corrected results, and the corrected results with method (1) and method (3) are close to each other. In comparison, the corrected F¹⁴C values based on our correction of iron counts are mostly closer to their corresponding consensus values and lie within 1σ uncertainty of all corrected results, which indicates good reproducibility for samples with >0.01 mg of C. Overall, ¹⁴C analysis of small IAEA reference standard samples containing different sample types and age ranges below 10 ka BP show that our ¹⁴C analysis of ultra-small samples of 10–100 μg C obtained accurate and reproducible results.

Figure 8 F¹⁴C corrected values of IAEA standards ranging 10–100 μg C from multiple ¹⁴C-AMS measurements. (a) Different symbols show different correction methods: the traditional background subtraction (black solid squares), constant contamination model (red solid circle), and our correction of ¹⁴C counts from iron (blue solid triangle). The black dashed lines represent the F¹⁴C consensus values of IAEA standards, the black solid lines and the horizontal gray band represent the average and its associated onesigma uncertainty of all corrected results.

Case Study

Chronological Framework for Marine Sediment Based on ¹⁴C Analysis of Small Foraminiferal Samples

Marine sediments are important scientific carriers for studying major issues including paleoclimate change, sea level trends, and geochemical cycles. Understanding the above issues require a precise chronological frame for marine sediments. Planktonic foraminifera is frequently used as ¹⁴C-AMS dating material for marine sediment. We have a collection of foraminifer samples from the South China Sea that have been classified into the genera G.ruber and G.sacculifer, respectively. We aim to determine a basic chronological sequence for sediment cores from the South China Sea. In practice, there exist few representative and reliable dating materials in sediments, and many harbor exogenous carbon, making the final ¹⁴C results unreliable and difficult to interpret. As a result, foraminiferal shells in marine sediments are selected for dating; it is even more crucial to select a single species of foraminifera for accurate dating. Nevertheless, the quality of single species of foraminifera that can be obtained is frequently limited. Therefore, establishing a chronological framework of marine sediment cores requires reliable ¹⁴C-AMS analysis of small foraminifer samples, even ultra-small samples.

In the first stage, four high-quality foraminifer samples were selected for analysis, i.e., 110–112 cm (B56F), 180–182 cm (B91F), 370–372 cm (B186F), and 453–455 cm (B228F). Through carbonate hydrolysis pretreatment, each sample was separated into many small samples of CO₂ (0.01–0.2 mg C) for graphitization and AMS dating. Meanwhile, we prepared 1 mg C graphite of each foraminifer sample for AMS measurement to study if ultra-small foraminifera samples as low as 0.01 mg C could be precisely dated as 1 mg C samples. As shown in Figure 9, the ¹⁴C-AMS outcomes of small foraminifer samples have been corrected for isotopic fractionation with δ¹³C-AMS values and ¹⁴C counts of Fe powder. In contrast with the ¹⁴C ages of the 1 mg C sample as the reference values, the ¹⁴C ages of the four foraminifer samples were all dispersed around the corresponding reference values and are within the 3σ error range of the reference value, which is comparable with the ¹⁴C results of 1 mg C sample. This illustrates that our small foraminifer samples of 10–200 µg C can also yield stable and reliable ¹⁴C ages, which can help in the dating of important layers lacking effective C fractions and emphasize the importance of correctly constructing the chronological framework of marine sediment cores.

Figure 9 The ¹⁴C age results of four foraminifera samples between 10 µgC and 1 mgC from multiple ¹⁴C-AMS measurements. All results have been corrected for isotopic fractionation with on-line δ¹³C AMS values and ¹⁴C counts of Fe powder. The black solid squares were represented ¹⁴C age values and, the black hollow triangles were represented ¹²C³⁺ beam currents, the gray zones were represented the range of 3σ errors of ¹⁴C age of 1 mg C sample, the dotted lines were represented the linear relationship between ¹²C³⁺ currents and the sample size.