Technical note: Studying lithium metaborate fluxes and extraction protocols with a new, fully automated in situ cosmogenic ¹⁴C processing system at PRIME Lab

Extraction procedures for in situ cosmogenic ¹⁴C (in situ ¹⁴C) from quartz require quantitative isotopic yields while maintaining scrupulous isolation from atmospheric and organic ¹⁴C. These time- and labor-intensive procedures are ripe for automation; unfortunately, our original automated in situ ¹⁴C extraction and purification systems, reconfigured and retrofitted from our original systems at the University of Arizona, proved less reliable than hoped. We therefore installed a fully automated stainless-steel system (except for specific borosilicate glass or fused-silica components) incorporating more reliable valves and improved actuator designs, along with a more robust liquid nitrogen distribution system. As with earlier versions, the new system uses a degassed lithium metaborate (LiBO₂) flux to dissolve the quartz sample in an ultra-high-purity oxygen atmosphere, after a lower-temperature combustion step to remove atmospheric and organic ¹⁴C.

We compared single-use high-purity Al₂O₃ against reusable 90 %Pt $/$ 10 %Rh ($\mathrm{Pt}/\mathrm{Rh}$) sample combustion boats. The $\mathrm{Pt}/\mathrm{Rh}$ boats heat more evenly than the Al₂O₃, reducing procedural blank levels and variability for a given LiBO₂ flux. This lower blank variability also allowed us to trace progressively increasing blanks to specific batches of fluxes from our original manufacturer. Switching to a new manufacturer returned our blanks to consistently low levels on the order of (3.4 ± 0.9) × 10^4 14C atoms.

We also analyzed the CRONUS-A intercomparison material to investigate sensitivity of extracted ¹⁴C concentrations to the temperature and duration of the combustion and extraction steps. Results indicate that 1 h combustion steps at either 500 or 600 ^∘C yield results consistent with the consensus value of Jull et al. (2015), while 2 h at 600 ^∘C results in loss of ca. 9 % of the high-temperature ¹⁴C inventory. Results for 3 h extractions at temperatures ranging from 1050 to 1120 ^∘C and 4.5 h at 1000 ^∘C yielded similar results that agreed with the nominal value and published results from most laboratories. On the other hand, an extraction for 3 h at 1000 ^∘C was judged to be incomplete due to a significantly lower measured concentration. Based on these results, our preferred technique is now combustion for 1 h at 500 ^∘C followed by a 3 h extraction at 1050 ^∘C. Initial analyses of the CoQtz-N intercomparison material at our lab yielded concentrations ca. 60 % lower than those of CRONUS-A, but more analyses of this material from this and other labs are clearly needed to establish a consensus value.

1 Introduction

Extracting in situ cosmogenic ¹⁴C (in situ ¹⁴C) from quartz is challenging in that minute quantities of ¹⁴C must be extracted and purified from quartz samples while preventing contamination by ubiquitous atmospheric and organic ¹⁴C. These extraction and purification procedures are time-consuming and labor-intensive when done manually – as such they are attractive targets for automation. Lifton et al. (2015) presented results from the initial automated in situ ¹⁴C extraction and purification systems at the Purdue Rare Isotope Measurement Laboratory (PRIME Lab), reconfigured and retrofitted from our original glass systems at the University of Arizona. As hoped, the automation of key components of our in situ ¹⁴C lab indeed led to increased throughput and reproducibility. While the overall timeline of the extraction, purification, and graphitization was still ca. 3 d, a single person was able to operate these systems simultaneously, boosting sample throughput significantly over the purely manual systems.

These automated systems comprised two independent extraction systems and a separate CO₂ purification system. A separate system for converting CO₂ to graphite was not automated. This required custom design and implementation of equipment to automate three key aspects of the systems: servo-based valve actuators, temperature control for cryogenic gas purification, and liquid nitrogen (LN) transfer. While these automated systems improved throughput over our original purely manual systems, they also required manual transfer of sample gas between separate extraction, purification, and (manual) graphitization systems.

However, in terms of overall reliability of operation, the limitations of retrofitting our original designs ultimately became apparent. For example, the glass high-vacuum valves are not precision components – no two are precisely the same. The valve actuators thus had to adapt to differences in resistance to motion arising from variations in valve stem and valve bore diameters, as well as to different lengths of travel to adequately seat each valve. As originally designed, the valve actuators accommodated these variations well, but the mechanical settings at which each operated properly tended to creep over time, such that sometimes during active processes individual valves might not indicate that they are closed or open, or might indicate a closed position but not be seated properly and allow leakage across the valve. Without actively checking on the system status when this happened, the sample gas could be pumped away by accident or a process could be interrupted (which could lead to system damage).

Similarly, the LN distribution system in that system was ultimately problematic. LN was transferred from a pressurized 200 L supply dewar through insulated Teflon tubing to fill dewars on various cold traps. Filling and emptying of individual dewars was controlled using LN level sensors comprising three resistors in series, positioned with resistors at empty, nominal, and full levels within each dewar. During the processes, certain cold traps needed to be alternately filled and emptied. Dewars stationed on those traps were emptied using a small shop vacuum cleaner via a drain manifold fitted with cryogenic solenoid valves. Particularly at times of high humidity (not as much of an issue in arid Arizona as in Indiana), ice condensation in those dewars could cause the drain tubing to clog and interrupt the process sequence. In addition, sometimes when a particular dewar was filled and emptied multiple times in a process sequence, the resistor string would not register the proper voltage during a fill cycle to trigger shutoff of LN flow, and the dewar would overflow continuously unless an operator was present to close the main supply dewar valve manually. Thus, although sample throughput and repeatability was considerably improved over manual operation, system reliability was not at the point where one could generally leave a system in unattended operation.

We thus recently purchased and installed a customized Carbon Extraction and Graphitization (CEGS) system from Aeon Laboratories, LLC, similar to that of Goehring et al. (2019) at Tulane University. The largely stainless-steel system (except for specific sections requiring borosilicate glass or fused-silica components) incorporates more reliable valves and improved actuator designs compared to our original system, as well as a robust and efficient liquid nitrogen distribution system (see Goehring et al., 2019, for additional details). The new system, controlled by a flexible and extensible modular software package written in C#, follows a similar procedure to that of Lifton et al. (2015), using a degassed lithium metaborate (LiBO₂) flux to dissolve the quartz sample in a research purity (RP) O₂ atmosphere. In addition, all sections of the new system are connected, meaning that one can extract all evolved carbon species as CO₂ from a quartz sample, purify and precisely measure the resulting gas yield, and convert the CO₂ to graphite for AMS analysis – all without human intervention. Below we describe key differences relative to the system of Goehring et al. (2019), and we then present baseline results from the now fully operational system, including procedural blanks and analyses of established intercomparison materials, for both our original single-use high-purity aluminum oxide (Al₂O₃) and new reusable 90 %Pt $/$ 10 %Rh sample boats.

2 Purdue CEGS design and operation

The Purdue CEGS (PCEGS) comprises three main modules: two extraction modules and a collection–purification–graphitization module (main CEGS module) (Fig. 1), following the general design of Goehring et al. (2019) but with an additional extraction module. However, the PCEGS differs from the latter system in two key aspects. First, the two PCEGS extraction modules (Tube Furnace 1, TF1, and Tube Furnace 2, TF2) are connected in parallel, each accommodating a high-temperature resistance furnace with a mullite furnace tube and evacuated by separate vacuum systems distinct from the main CEGS vacuum system (Fig. 1). On the other hand, the Goehring et al. (2019) system comprises one tube furnace in series with the CEGS, evacuated by a single vacuum system. Our design allows each PCEGS extraction module to run processes independently of those controlled by the other modules, enabling increased flexibility in system operations. The other key difference is that condensable gases evolved on the PCEGS during an extraction procedure in either furnace are trapped in a compact borosilicate glass coil trap held at LN temperature (−196 ^∘C) instead of the variable temperature trap (VTT) used for this purpose on the Goehring et al. (2019) system (Figs. 1, 2). The compact coil trap (ca. 3.5 cm diameter × ca. 10 cm tall) derives from our previous larger coil trap designs (e.g., Lifton et al., 2001; Pigati et al., 2010; Lifton et al., 2015), which consistently demonstrated quantitative trapping of minute CO₂ quantities from O₂ carrier gas. This compact design ensures similarly reliable CO₂ trapping through a 9 mm o.d. × 7 mm i.d. inlet downtube, delivering process gases directly to the base of the trap before passing through a constriction connecting the downtube to a 6 mm o.d. × 4 mm i.d. coiled section and outlet tube (Fig. 2). The total length of the trap submerged in LN when operating is ca. 55 cm (ca. 5 cm of the downtube and the ca. 50 cm coil).

Figure 1 (a) Schematic and (b) photo of the Purdue Carbon Extraction and Graphitization System (PCEGS).

Figure 2 Compact borosilicate glass coil trap, consisting of a 9 mm o.d. × 7 mm i.d. inlet downtube (on the left), connected to a tightly coiled 6 mm o.d. × 4 mm i.d. section with subsequent outlet tube. The scale on bottom is in centimeters.

Other than incorporating a U-shaped secondary oxidation furnace (9 mm o.d. × 7 mm i.d. filled with 2 mm quartz beads, held at ca. 900 ^∘C) from the Lifton et al. (2015) system instead of an inline granular quartz oxidation furnace of the Goehring et al. (2019) design, the rest of the PCEGS utilizes similar hardware to that of the latter. However, our six-reactor graphitization manifold is configured in front of the main purification and measurement process path to achieve a shorter footprint than the linear configuration of the Tulane system (Fig. 1), allowing the first extraction module and the CEGS module to fit onto our existing lab frame and benchtop.

We implement a 2 d extraction procedure with the PCEGS similar to those of Lifton et al. (2015) and Goehring et al. (2019), utilizing a LiBO₂ flux to dissolve the quartz and release the in situ ¹⁴C at 1100 ^∘C (Table 1). The first day's procedures involve degassing the LiBO₂ flux and preparing the purified quartz for extraction, while the second day is the extraction, purification, and graphitization procedure. Once started, the day 1 LiBO₂ degassing process operates on the selected extraction module (either TF1 or TF2) completely independently of the main CEGS module. The day 2 process, on the other hand, requires control from the main CEGS module to allow sample collection, purification, measurement, dilution, and graphitization. In practical terms, we execute a day 1 process on one extraction module, then the next day start a day 1 process on the second extraction module. The day 2 process for the first extraction module can then be run without interruption from the main CEGS module. The day 1 and day 2 processes are then subsequently cycled between the two extraction modules. This comfortably allows for PCEGS throughput of 4–5 samples per week.

Table 1 Procedural flow for in situ ¹⁴C extraction at PRIME Lab. See text for abbreviations.

On day 1, a quartz sample is pretreated with 50 % (v : v) HNO₃ : 18 MΩ water for at least 90 min in an ultrasonic bath, rinsed thoroughly in 18 MΩ water, then dried in a vacuum oven overnight. A sample boat (either single-use high-purity Al₂O₃ or reusable 90 %Pt-10 %Rh) containing ca. 20 g of pre-fused LiBO₂ beads (melting point 845 ^∘C) is placed inside a flame-cleaned fused-silica sleeve in the mullite furnace tube (with borosilicate glass o-ring ball joint end seals) using flame-cleaned implements. The 24 in. long (60.96 cm) fused-silica sleeve (replaced after every sample) extends beyond the furnace hot zone, protecting the furnace tube from LiBO₂ vapors that evolve from the fused sample at high temperature. The aggressively reactive vapors etch the interior of the sleeve within the hot zone of the furnace instead of the furnace tube itself (Fig. 3). To minimize intrusion of atmospheric CO₂ or other contaminant gases into the furnace tube each time it is opened, the tube is first backfilled with research purity He (99.9999 %) to 20 torr above ambient atmospheric pressure. The He is then slowly bled through the tube while open to atmosphere. Once closed again, the furnace tube is evacuated to $<\mathrm{5}\times {\mathrm{10}}^{-\mathrm{3}}$ torr, isolated, and 50 torr of RP O₂ is subsequently added. The furnace is then heated to the extraction temperature (typically 1100 ^∘C) for 1 h while O₂ is bled through with a mass flow controller and automated metering valve to maintain the tube pressure and to flush out any evolved contaminants to the vacuum pump. The tube is then cooled to <800 ^∘C to ensure LiBO₂ solidification before evacuation overnight.

Figure 3 Comparison of quartz sleeve corrosion from LiBO₂ for (a) Al₂O₃ boats and (b) $\mathrm{Pt}/\mathrm{Rh}$ boats (inset) after the high-temperature fusion step (3 h) at the temperatures indicated. Note the significantly greater corrosion associated with the $\mathrm{Pt}/\mathrm{Rh}$ boats vs. the Al₂O₃, indicating more aggressive and uniform heating in the former and noticeably milder corrosion from the 1050 and 1000 ^∘C runs. The Al₂O₃ boat on the left side of (b) is holding the sleeves in place but also serves as a comparison to (a). We speculate that the greater corrosion from LiBO₂ in the $\mathrm{Pt}/\mathrm{Rh}$ boats reflects higher thermal conductivity of the metal boats vs. the sintered ceramic boats.

On day 2, after backfilling the tube with He as before, the boat with degassed LiBO₂ is removed (again with flame-cleaned implements) to a HEPA-filtered laminar flow bench. Approximately 5 g of the pretreated quartz sample is evenly distributed over the now-solid LiBO₂ in the boat and the boat and sample are returned to the furnace, evacuated to $<\mathrm{5}\times {\mathrm{10}}^{-\mathrm{3}}$ torr, isolated, and 50 torr RP O₂ is added again. The sample is then heated to 500 ^∘C for 1 h to combust and remove atmospheric and organic contaminants, while bleeding O₂ across the sample as before and exhausting to the vacuum system. After that hour, the 500 ^∘C tube furnace is evacuated to $<\mathrm{5}\times {\mathrm{10}}^{-\mathrm{3}}$ torr. Subsequently, 50 torr of RP O₂ is admitted into the tube furnace and the sample and flux are heated to 1100 ^∘C and held at the high temperature for 3 h to dissolve the quartz and release any trapped carbon species. During extraction, the O₂ pressure in the tube typically rises to ca. 60 torr.

After the extraction procedure completes, the evolved gases are bled with RP O₂ through the secondary oxidation furnace to ensure any carbon species released during extraction are completely oxidized to CO₂ before collection in the compact coil trap cooled with LN. During this step, the tube pressure at the end of the extraction step is maintained during the bleed (to prevent excess LiBO₂ vaporization) while the furnace cools to <800 ^∘C (to ensure complete melt solidification), before shutting off additional O₂ inflow and slowly evacuating all tube gases through the secondary furnace and coil trap. The condensed gases are then transferred to the purification section to remove water, halogens, and nitrogen and sulfur oxides. The gas is transferred cryogenically with LN first into the variable temperature trap (VTT), and the incondensable gases are evacuated. The VTT is then warmed to −145 ^∘C for 10 min to remove sulfur oxides, water, and other contaminant gases from the sample CO₂. The evolved CO₂ is then passed through a Cu mesh $/$ Ag wool trap held at 600 ^∘C (removes nitrogen oxide, halogen, and sulfur oxide contaminants) and frozen with LN into the volumetric measurement chamber (MC) (Fig. 1). The CO₂ yield is then measured manometrically as equivalent mass of C (µg) and typically diluted to ca. 300 µg C with ¹⁴C-free CO₂. If the undiluted CO₂ yield is ≥300 µg C, no dilution step occurs. A ca. 9 µg C split (ca. 3 % of the total C mass) is collected in a pre-evacuated Exetainer^® vial for stable C isotopic analysis offline, and the remaining sample is transferred cryogenically to one of the six graphite reactors (Fig. 1). The sample then undergoes hydrogen reduction (with research purity H₂ – 99.9999 %) to filamentous C (graphite) on an Fe catalyst, with water trapped by Mg(ClO₄)₂ (Southon, 2007; Santos et al., 2004). Procedural background samples are run after approximately every 7–10 unknown samples, using identical procedures without adding quartz.

Finally, the graphite is packed into an Al cathode for ¹⁴C measurement by accelerator mass spectrometry (AMS) at PRIME Lab. Sample ${}^{\mathrm{14}}\mathrm{C}{/}^{\mathrm{13}}\mathrm{C}$ ratios are measured relative to Oxalic Acid II (NIST-4990C). Stable carbon isotopic ratios were measured at the University of California at Davis Stable Isotope Facility (https://stableisotopefacility.ucdavis.edu, last access: August 2023) using isotope ratio mass spectrometry (Lifton et al., 2015). Measured in situ ¹⁴C concentrations are calculated from the resulting ${}^{\mathrm{14}}\mathrm{C}/{\mathrm{C}}_{\mathrm{total}}$ after subtracting representative procedural background ¹⁴C values, following Hippe and Lifton (2014). Measurement uncertainties are presented at the 1σ level unless otherwise noted.

3 Initial experiments

Once the PCEGS was operational, we began to characterize its performance in terms of procedural blank (background) values and measurements of intercomparison materials such as CRONUS-A (Jull et al., 2015). We also characterized the mass dependence of graphitization blanks. Since the publication of Lifton et al. (2001), we and other labs using LiBO₂ for extraction (e.g., Goehring et al., 2019; Lamp et al., 2019; Fülöp et al., 2010) had used single-use high-purity sintered Al₂O₃ combustion boats for our flux + samples. On the other hand, laboratories that implemented flux-free in situ ¹⁴C extractions have either used Pt (e.g., Hippe et al., 2009, 2013; Lupker et al., 2019) or fused-silica vessels (Fülöp et al., 2015, 2019). The labs using flux-free processes typically report blanks on the order of 1–3 × 10^4 14C atoms (e.g., Lupker et al., 2019; Fülöp et al., 2019), while the labs using flux-based extractions have reported blanks on the order of 1–2 × 10^5 14C atoms (e.g., Lifton et al., 2015; Goehring et al., 2019; Lamp et al., 2019). Goehring et al. (2019) deduced that the differences in ¹⁴C backgrounds between the flux and flux-free extraction systems lay at least in part with the sintered Al₂O₃ boats reacting with the flux to release small and variable amounts of persistent contaminant ¹⁴C during the extraction process. They described assessing boats of alternate construction and reported a promising process blank result of ca. 4×10^4 14C atoms from an initial experiment with a reusable 90 %Pt : 10 %Rh alloy boat.

Our initial PCEGS experiments utilized the single-use Al₂O₃ combustion boats, but in the meantime we also obtained a set of 90 %Pt : 10 %Rh (hereafter $\mathrm{Pt}/\mathrm{Rh}$) combustion boats from Heraeus Precious Metals North America LLC (http://www.pt-labware.com, last access: August 2023). We thus compared results using both types of boats for both blanks and intercomparison samples. The solidified LiBO₂ + sample melt is cleaned from the $\mathrm{Pt}/\mathrm{Rh}$ boats between samples by overnight ultrasonication at 40 ^∘C in 10 % (v : v) reagent grade HNO₃ : 18 MΩ water in sealed 1 L polypropylene bottles, followed by thorough rinsing in 18 MΩ water and drying in a gravity oven.

3.1 Graphitization blanks

The mass dependence of the PCEGS graphitization blanks was assessed by graphitizing aliquots of ¹⁴C-free CO₂ in masses ranging from ca. 50 to 1000 µg C (Table 2). As with previous studies using either Zn or H₂ as the reducing agent for CO₂ to C (e.g., Donahue et al., 1990; Lifton et al., 2001, 2015; Goehring et al., 2019), we observe an inverse relationship between sample mass (in µg C) and the measured ${}^{\mathrm{14}}\mathrm{C}/{\mathrm{C}}_{\mathrm{total}}$. If one assumes a constant modern contaminant contribution to the graphitization blank from the graphitization reactor, independent of the sample mass, then one would observe an inverse mass dependence of the blank: higher blanks for lower-mass samples (e.g., Donahue et al., 1990). This relationship is well-characterized by the following equation (adjusted R²=0.994):

$$\begin{array}{}\text{(1)}& \begin{array}{rl}{B}_{\mathrm{g}}& =\left(\mathrm{1.243}\pm \mathrm{0.045}\right)\times {\mathrm{10}}^{-\mathrm{13}}/\mathrm{mass}\\ & +\left(\mathrm{1.301}\pm \mathrm{0.050}\right)\times {\mathrm{10}}^{-\mathrm{15}}.\end{array}\end{array}$$

Correction of the measured sample ${}^{\mathrm{14}}\mathrm{C}/{\mathrm{C}}_{\mathrm{total}}$ for the graphitization blank ${}^{\mathrm{14}}\mathrm{C}/{\mathrm{C}}_{\mathrm{total}}$ (B_g) follows Eq. (6) of Donahue et al. (1997).

Table 2 Graphitization blanks.

^a PRIME Lab ID. ^b δ¹³C averages −45.6 ± 0.2 ‰_VPDB.

3.2 Procedural blank comparison

Initial experiments with the new system involved procedural blanks with our original single-use Al₂O₃ boats in concert with measurements of intercomparison materials (Sect. 3.3). Subsequently, we switched to reusable 90 %Pt $/$ 10 %Rh sample boats, with associated measurements of procedural blanks and intercomparison materials for a range of experimental conditions.

3.2.1 Al₂O₃ boats

The first set of blanks and intercomparison samples processed on the PCEGS with Al₂O₃ boats involved a more aggressive than normal day 2 combustion step to more thoroughly remove any potential organic C that might remain on the etched sample grains. This was motivated by Nichols and Goehring (2019), who found evidence of modern ¹⁴C contamination by laurylamine used in froth flotation mineral separation techniques that was not removed completely by their original etching procedure. Although we had not observed evidence of this issue with in situ ¹⁴C results from our lab, we tested a low-temperature combustion procedure of 2 h at 600 ^∘C, reasoning that Hippe et al. (2013) utilized a 2 h at 700 ^∘C combustion step with no apparent demonstrable effects on their results relative to combustion for 1 h at 500 ^∘C. This more aggressive combustion step was then followed by our normal 1100 ^∘C flux fusion for 3 h.

Initial procedural blank experiments largely utilized TF1 and progressively increased from ca. 6.50×10⁴ to 1.03×10^5 14C atoms with a mean of (8.79 ± 1.64) × 10^4 14C atoms, while a single blank from TF2 yielded ca. 1.14×10^5 14C atoms (Table 3, Fig. 4). The source of the time-dependent increase was not identified before switching to the $\mathrm{Pt}/\mathrm{Rh}$ boats, but these values still represent an improvement over blank values presented in Lifton et al. (2015) of ca. 30 %–70 %.

Figure 4 Procedural blank results for Al₂O₃ and $\mathrm{Pt}/\mathrm{Rh}$ boats (1σ uncertainties). All blanks using Al₂O₃ boats used a 2 h at 600 ^∘C combustion step followed by a 3 h extraction at 1100 ^∘C (1120 ^∘C for Tube Furnace 2 due to a miscalibration at the 1100 ^∘C set point). All $\mathrm{Pt}/\mathrm{Rh}$ Tube Furnace 1 runs were a 1 h at 500 ^∘C combustion step followed by a 3 h extraction step at 1100 ^∘C (except as indicated). Tube Furnace 2 combustions with $\mathrm{Pt}/\mathrm{Rh}$ boats were also 1 h at 500 ^∘C, but extractions were at 1120 ^∘C due to the miscalibration (except as indicated).

Table 3 Al₂O₃ procedural blanks. All analyses used 2 h combustion at 600 ^∘C and 3 h extraction at 1100 ^∘C (unless otherwise noted).

All blanks here used Claisse C-0611-00, Batch C-10001 – “Batch 1” ultrapure LiBO₂. ^* 1120 ^∘C extraction due to furnace miscalibration at 1100 ^∘C set point.

3.2.2 Pt $/$ Rh boats

On switching to the $\mathrm{Pt}/\mathrm{Rh}$ boats, we also reverted to our original procedure utilizing a 500 ^∘C combustion step for 1 h. It was immediately obvious that the $\mathrm{Pt}/\mathrm{Rh}$ boats heat much more uniformly than the Al₂O₃ based on dramatic differences in the flux's corrosive effects on the quartz sleeves between the two types of boats (Fig. 3). The sleeves used with the Al₂O₃ boats were corroded mainly above and below the boat and at the ends of the heated zone where the LiBO₂ vapor condenses in ca. 5 cm wide bands (Fig. 3a). The rest of the heated portion of the sleeve is only lightly corroded and remains transparent. However, when using the $\mathrm{Pt}/\mathrm{Rh}$ boats, the LiBO₂ more evenly corrodes the sleeve interior over the entire hot-zone length (Fig. 3b). Most of the boat heating likely occurs via conduction from the bottom edges of the boat in contact with the quartz sleeve. It thus appears that the more efficient heat conduction of the metal boats leads to more even and aggressive heating of the flux and sample than in the Al₂O₃ boats. Experiments with the $\mathrm{Pt}/\mathrm{Rh}$ boats at extraction temperatures of 1000 and 1050 ^∘C resulted in significantly less corrosion of the sleeve than at 1100 ^∘C (Fig. 3b).

Initial procedural blanks using the $\mathrm{Pt}/\mathrm{Rh}$ boats were dramatically lower than those using the Al₂O₃ boats, with much better reproducibility, averaging (4.08 ± 0.66) × 10^4 14C atoms (1σ) (Table 4, Fig. 4). Different combinations of combustion (500 and 600 ^∘C for 1 h) and extraction temperatures and times (1100 ^∘C for 3 h; 1000 ^∘C for 3 and 4.5 h) were investigated as well (Table 4) (corresponding to intercomparison experiments described in Sect. 3.3), with no significant effect on blank results. This supports the hypothesis of Goehring et al. (2019) that a significant component of Al₂O₃ procedural blanks derived from the sintered ceramic boats themselves. The improved blank reproducibility using the $\mathrm{Pt}/\mathrm{Rh}$ boats allows us to identify background signals that previously we were unable to resolve. After this initial set of analyses depleted most of the bottle of ultrapure-grade LiBO₂ (Claisse C-0611-00, Batch C-10001 – “Batch 1”), we switched to a new bottle of pure-grade LiBO₂ (Claisse C-0610-00, Batch C-17000-10 – “Batch 2”). We reasoned that pure and ultrapure grades only differ in metal impurity content – both are pre-fused, spherical beads and thus should be essentially functionally equivalent for our application.

Table 4 $\mathrm{Pt}/\mathrm{Rh}$ procedural blanks. All analyses used 1 h combustion at 500 ^∘C and 3 h extraction at 1100 ^∘C (unless otherwise noted).

^a 1000 ^∘C extraction, 3 h. ^b 1000 ^∘C extraction, 4.5 h. ^c 600 ^∘C combustion, 1 h. ^d 1100 ^∘C extraction, 4.5 h. ^e 1120 ^∘C extraction due to furnace miscalibration at 1100 ^∘C set point. ^f 1050 ^∘C extraction. ^g Claisse C-0611-00, Batch C-10001 – “Batch 1” ultrapure. ^h Claisse C-0610-00, Batch C-17000-10 – “Batch 2” pure. ⁱ Claisse C-0610-00, Batch C-19000-10 – “Batch 3” pure. ^j Claisse C-0611-00, Batch C-19001-10 – “Batch 4” ultrapure. ^k SPEX Certi-Prep FFB-0000-03, Lot 240920D-2904 ultrapure.

However, subsequent blanks with the new bottle increased in both CO₂ yield (ca. 1.5 µg to ca. 4 µg C equivalent) and ¹⁴C content (ca. (1.51 ± 0.31) × 10^5 14C atoms) (Table 4, Fig. 4). Although these values were higher than the initial measurements, they were reproducible on both TF1 and TF2, and thus we continued with normal system operation. Subsequently, the CO₂ yields and ¹⁴C content inexplicably jumped again to new “stable” values of ca. 6.6 µg C equivalent and (2.66 ± 0.07) × 10^5 14C atoms, respectively, using TF1, with a similar but slightly lower result with TF2. At that point we tested a second bottle of pure-grade LiBO₂ (Claisse C-0610-00, Batch C-19000-10 – “Batch 3” – purchased at the same time as Batch C-17000-10) on procedural blanks in TF1 (PCEGS-94) and TF2 (PCEGS-95), with even higher results of (3.21 ± 0.10) × 10⁵ and (3.63 ± 0.15) × 10^5 14C atoms, respectively. The higher blanks from Batch C-19000-10 also exhibited higher CO₂ yields (ca. 8–9.5 µg C equivalent). In fact, the CO₂ yields from each extraction module tracked the ¹⁴C atoms quite linearly for all of these experiments, with similar regression fits to each (R² values of 0.955 and 0.970 for TF1 and TF2, respectively; see Fig. 5).

Figure 5 Linear fit between CO₂ yield (in µg C) vs. procedural blank (in ¹⁴C atoms) for tube furnaces 1 and 2 (1σ uncertainties shown).

At that point we paused normal system operations and conducted more basic experiments to try to isolate the source of the increased blanks – was it in the system overall or the LiBO₂? Two procedural blanks with everything except for the LiBO₂ (boat and sleeve only) – one boat cleaned in 10 % v : v HCl and the other in 10 % v : v HNO₃ – both yielded ca. 2.0×10^4 14C atoms. This indicated that the LiBO₂ was the source of the high blank, although the nature of that source and why the blank increased with time is unclear. This is particularly puzzling since the low blanks with $\mathrm{Pt}/\mathrm{Rh}$ boats (compared to the boat-and-sleeve-only blanks) indicate that the degassing step on day 1 should be effective at minimizing atmospheric CO₂ contamination from the flux.

We then obtained a new bottle of ultrapure-grade LiBO₂ (Claisse C-0611-00, Batch C-19001-10 – “Batch 4”); two blanks from that bottle from TF2 (PCEGS-98 and 99) yielded values comparable to PCEGS-95 – ca. 3.6–3.7 × 10^5 14C atoms and ca. 8–9 µg C-equivalent yields (Fig. 4, Table 4). Finally, we tried a blank with the remainder of the original bottle of ultrapure-grade LiBO₂ (Claisse C-0611-00, Batch C-10001 – “Batch 1”). This experiment (PCEGS-100) exhibited CO₂ yield and ¹⁴C content comparable to our original tests: 2.2 µg C equivalent and 4.76 ± 1.12 × 10^4 14C atoms. In consultation with Claisse technical support, we were unable to identify any chemical change in their product or manufacturing process that could have led to the progressively increasing blanks. As such, we identified another vendor, SPEX CertiPrep. We purchased a similar pre-fused ultrapure-grade LiBO₂ from them (FFB-0000-03, Lot 240920D-2904) and ran a blank on each extraction module. CO₂ yields were comparable to those of the original Claisse ultrapure batch, and ¹⁴C contents were slightly improved over that material: ca. 2.5 µg C equivalent and ca. 3.6×10^4 14C atoms (Table 4, Fig. 4). Subsequent blanks with the new SPEX LiBO₂ were generally comparable to or better than those initial measurements, ranging from ca. 2.4×10⁴ to 5.0×10^4 14C atoms (mean: (3.38 ± 0.92) × 10^4 14C atoms), and similar to recently published blank measurements from other in situ ¹⁴C laboratories using Pt sample boats (e.g., Lupker et al., 2019; Goehring et al., 2019) (Table 4, Fig. 4). Regardless of the ultimate cause of the unexplained blank behavior with the more recent bottles of Claisse LiBO₂, we are proceeding with the SPEX ultrapure LiBO₂ as our preferred flux.

Late in this process we also discovered that the temperature controller for TF2 was miscalibrated at high-temperature set points, reading 1120 ^∘C on an independent Type S thermocouple probe when set to 1100 ^∘C. Independent measurement of the lower temperatures for the combustion steps in TF2 agreed with the set points – only the extraction temperatures exhibited the offset. We subsequently adjusted the set point temperatures for extractions to achieve the desired temperature on that furnace (1080 ^∘C set point for 1100 ^∘C actual and 1035 ^∘C set point for 1050 ^∘C actual). No such problem was observed with TF1. Results from both blanks and intercomparison materials (Sect. 3.3 below) do not appear to indicate any significant effect from the 20 ^∘C excess temperature in the affected TF2 experiments (Figs. 4 and 6, Tables 4 and 5).

Figure 6 CRONUS-A results with experimental details (1σ uncertainties) from this study, with mean values from other studies for comparison. The PCEGS mean value includes all samples except PCEGS-47. Other laboratories' results are listed as follows: ETH (Lupker et al., 2019), ANSTO-Wollongong (Fülöp et al., 2019), LDEO (Lamp et al., 2019), Tulane (Goehring et al., 2019), and PRIME 2015 (Lifton et al., 2015).

Table 5 Intercomparison samples. All analyses used 1 h combustion at 500 ^∘C and 3 h extraction at 1100 ^∘C (unless otherwise noted).

^a 2 h at 600 ^∘C combustion. ^b 4.5 h at 1000 ^∘C extraction. ^c 3 h at 1000 ^∘C extraction. ^d 1 h at 600 ^∘C combustion. ^e 3 h at 1120 ^∘C extraction. ^f 3 h at 1050 ^∘C extraction. ¹ TF1. ² TF2.

3.3 Extraction experiments with intercomparison materials

While we worked to isolate and understand the source(s) of the time-dependent procedural blanks on our new system, we also set out to better understand the effects of different combustion temperatures and durations on the amount of ¹⁴C extracted from the well-studied CRONUS-A intercomparison material derived from Antarctic sandstone bedrock of long exposure duration (Jull et al., 2015). In addition, since the more uniform heating of the $\mathrm{Pt}/\mathrm{Rh}$ boats rendered the LiBO₂ flux more broadly aggressive toward the fused-silica sleeves at 1100 ^∘C, we tested whether it would be possible to lower the extraction temperature and still achieve full ¹⁴C recovery from CRONUS-A. We also initiated measurements at PRIME Lab of the in situ ¹⁴C content of the CoQtz-N intercomparison material (e.g., Binnie et al., 2019) using both types of boats. CoQtz-N is derived from a boulder of vein quartz in Namibia, which is again of long exposure duration (Binnie et al., 2019).

3.3.1 CRONUS-A – Al₂O₃ boats

Initial experiments with the Al₂O₃ boats used CRONUS-A to test whether the more aggressive combustion procedure described in Sect. 3.2.1 (2 h at 600 ^∘C) followed by a 3 h fusion at 1100 ^∘C might affect the measured in situ ¹⁴C concentrations significantly. Results from both TF1 and TF2 yielded ¹⁴C concentrations on the order of 10 % below the consensus value for the material and outside the uncertainty band (Table 5, Fig. 6), suggesting diffusive loss of in situ ¹⁴C during the more aggressive low-temperature combustion step. We thus subsequently abandoned that more aggressive procedure in favor of the original 1 h at 500 ^∘C combustion step of Lifton et al. (2001) (also Sect. 3.2.2); results from Lifton et al. (2015) using Al₂O₃ boats with this original procedure are shown for comparison in Fig. 6.

3.3.2 CRONUS-A – Pt $/$ Rh boats

Our efforts with the $\mathrm{Pt}/\mathrm{Rh}$ boats largely focused on optimizing extraction temperature and time, again using CRONUS-A as a benchmark (Table 5, Fig. 6). We varied combustion and extraction temperatures and durations using corresponding background corrections appropriate for the procedures used and allowing for the observed procedural blank time dependence.

The experiments with extractions for 3 h at 1100 ^∘C and 1120 ^∘C and 4.5 h at 1000 ^∘C (PCEGS-44, 46, 50, 90, 104, 105, 106, 133; see Table 5, Fig. 6) yielded a mean and standard deviation of (7.08 ± 0.17) × 10^5 14C atoms g⁻¹ (1σ). An additional extraction test for 3 h at 1000 ^∘C (PCEGS-47) yielded a ¹⁴C concentration about 8 % lower than this mean but still within the nominal range of results in Jull et al. (2015). However, we judge this extraction as likely to be incomplete as it is outside of the 2σ uncertainty in our mean $\mathrm{Pt}/\mathrm{Rh}$ CRONUS-A analyses, and as such do not consider this further. Another test with a combustion step of 1 h at 600 ^∘C and a normal 1100 ^∘C extraction (PCEGS-50) yielded a result ca. 4 % below the mean above using a 500 ^∘C combustion temperature but within 2σ of that mean and still well within the Jull et al. (2015) range. Excluding PCEGS-50 from the mean above does not significantly change the mean nor these conclusions. After discovering the furnace controller miscalibration for TF2, we also tested CRONUS-A results for TF2 at 1100 ^∘C (PCEGS-105) and found them indistinguishable from the $\mathrm{Pt}/\mathrm{Rh}$ mean. Finally, given the less aggressive corrosion of the quartz sleeve from tests at 1050 ^∘C (Fig. 3b), we also tested CRONUS-A extraction for 3 h at 1050 ^∘C (PCEGS-106), with results being indistinguishable from our overall $\mathrm{Pt}/\mathrm{Rh}$ mean (Table 5, Fig. 6). We thus have switched to a 3 h at 1050 ^∘C extraction temperature and duration going forward.

Our CRONUS-A results are consistent with the consensus value and range of Jull et al. (2015), i.e., (6.93 ± 0.44) × 10^5 14C atoms g⁻¹, as well as with the mean of our previous results at PRIME Lab (Lifton et al., 2015), i.e., (6.89 ± 0.04) × 10^5 14C atoms g⁻¹. In addition, these new results are consistent with recent measurements by Lupker et al. (2019), Fülöp et al. (2019), and Lamp et al. (2019) (Fig. 6). Like those other studies, they also disagree with the CRONUS-A measurements of Goehring et al. (2019), for reasons yet to be determined (Fig. 6).

3.3.3 CoQtz-N

Our three results for the CoQtz-N intercomparison material spanned the period discussed in this work (Table 5). An initial analysis using an Al₂O₃ boat and the more aggressive 2 h combustion at 600 ^∘C returned a lower concentration (2.48 ± 0.06) × 10^5 14C atoms g⁻¹ than the two $\mathrm{Pt}/\mathrm{Rh}$ experiments at 500 ^∘C/1100 ^∘C (TF1) and 500 ^∘C/1120 ^∘C (TF2), which agree within 1σ measurement uncertainties and yield a mean value of (2.62 ± 0.04) × 10^5 14C atoms g⁻¹. Interestingly, the Al₂O₃ result with the more aggressive combustion step is only about 5 % lower than the $\mathrm{Pt}/\mathrm{Rh}$ mean CoQtz-N result (uncertainties overlap at 2σ), while the Al₂O₃ analyses of CRONUS-A are 9 % lower than the corresponding nominal value. The source of this difference is not clear, but it likely reflects intrinsic differences in diffusive properties of the quartz from each sample.

We only found one other study in which in situ ¹⁴C had been measured in CoQtz-N (Schiffer et al., 2020), but that study provides incomplete experimental details and only a plot of concentrations vs. quartz mass without any tabulated data. The four measured values presented for 1 g of CoQtz-N appear to span concentrations ca. 3×10⁵ to over 4×10^5 14C atoms g⁻¹ – well above our measured values. The source of this discrepancy merits further investigation but is currently difficult to evaluate without complete experimental details.

4 Conclusions

This study details key characteristics of and procedures in use for the new in situ ¹⁴C extraction system at PRIME Lab (PCEGS), and presents results of initial testing of procedural blanks and intercomparison materials. We compare results using the original single-use Al₂O₃ sample boats employed since Lifton et al. (2001) with those from a new set of reusable 90 %Pt $/$ 10 %Rh alloy sample boats.

It is clear from these experiments that the reusable $\mathrm{Pt}/\mathrm{Rh}$ boats provide distinct advantages over the Al₂O₃ boats, supporting the suggestions of Goehring et al. (2019). First, the $\mathrm{Pt}/\mathrm{Rh}$ boats appear to heat much more aggressively than the sintered Al₂O₃ ceramic boats, likely leading to more uniform heating of the contents. The $\mathrm{Pt}/\mathrm{Rh}$ boats also appear to reduce or eliminate a significant component of the blank variability associated with the sintered ceramics, perhaps associated with small amounts of atmospheric carbon potentially incorporated into the ceramics during manufacture. Taken together, the aggressive uniform heating and purity of the $\mathrm{Pt}/\mathrm{Rh}$ alloy allow for improved analytical reproducibility, allowing robust identification of systematic influences on background signals that we were previously unable to resolve with the Al₂O₃ boats.

Using the $\mathrm{Pt}/\mathrm{Rh}$ boats, we demonstrated that time-dependent increases in procedural blanks were tied directly to specific batches of LiBO₂ fluxes manufactured by Claisse. The time dependence did not appear to reflect flux purity, but instead some presently unknown characteristic of the Claisse fluxes appears to have changed since the original batch we used for our early experiments. Subsequent analyses with LiBO₂ from an alternate supplier, SPEX CertiPrep, yielded consistently low procedural blanks on the order of (3.4 ± 0.9) × 10^4 14C atoms, and we have switched to that flux going forward.

We also analyzed two intercomparison materials as part of our initial experiments to confirm compatibility with earlier results from this lab and from others. Using both Al₂O₃ and $\mathrm{Pt}/\mathrm{Rh}$ boats, we focused mainly on CRONUS-A but also made initial measurements for our laboratory of the newer CoQtz-N intercomparison material. We first tested CRONUS-A in Al₂O₃ boats using a more aggressive combustion procedure than typically used (2 h at 600 ^∘C vs. 1 h at 500 ^∘C) and found significantly lower ¹⁴C concentrations from the high-temperature extraction relative to the nominal value of Jull et al. (2015), likely due to diffusive loss during the more aggressive low-temperature step. Abandoning that aggressive procedure in favor of the shorter 500 ^∘C combustion and switching to the $\mathrm{Pt}/\mathrm{Rh}$ boats, we then explored various time–temperature combinations for the high-temperature extraction step with CRONUS-A. Results for 3 h extractions at temperatures ranging from 1050 to 1120 ^∘C and 4.5 h at 1000 ^∘C yielded similar results, in agreement with the consensus value and with published results from most laboratories, including those using our previous extraction system (Lifton et al., 2015). On the other hand, an extraction for 3 h at 1000 ^∘C yielded a significantly lower concentration than the other analyses in this study, suggesting incomplete extraction for those conditions. Based on these results, our preferred technique is now combustion for 1 h at 500 ^∘C followed by a 3 h extraction at 1050 ^∘C.

The initial analysis of CoQtz-N at PRIME Lab used the more aggressive combustion step but displayed less diffusive loss (relative to our analyses with $\mathrm{Pt}/\mathrm{Rh}$ boats) than CRONUS-A did with that procedure, suggesting variable low-temperature diffusion behavior among samples. Subsequently, internally consistent results were achieved with CoQtz-N using $\mathrm{Pt}/\mathrm{Rh}$ boats, with approximately 60 % lower ¹⁴C concentrations than CRONUS-A. However, additional analyses of this material from this and other labs are clearly needed to work toward a consensus value.