INTRODUCTION
The Bristol Radiocarbon AMS (BRAMS) facility was established as a joint enterprise between the faculties of Arts and Science at the University of Bristol to provide radiocarbon (14C) analytical capabilities to the archaeological, earth- and life-science communities. It was created to facilitate 14C-based research for internal and external users alike. The establishment of the BRAMS facility was the realization of an ambition evolved from nearly fifty years of research into ancient organic materials by the Organic Geochemistry Unit (OGU). This background of analytical biomolecular archaeological and environmental chemistry provides the ideal footing for analytically rigorous radiocarbon-based research, particularly with a molecular focus. The limiting factor in the accuracy and precision of 14C data is no longer limited by the AMS instrumentation, but by the sample pretreatment chemistry and graphite preparation. One of the major aims in establishing the BRAMS Facility was to continue refining these aspects of 14C analysis.
The BRAMS Facility is situated in a dedicated suite of laboratories housed in the department of Anthropology and Archaeology at the University of Bristol. The BrisMICADAS at the heart of BRAMS is a compact 200kV MICADAS AMS (Figure 1) developed and built by the Laboratory of Ion Beam Physics, ETH, Zurich. It is equipped with helium stripper and permanent magnet technologies and a gas-capable ion source interfaced to both an elemental analyzer (EA) and a carbonate handling system (CHS). Our sample preparation laboratory is equipped with an IonPlus AGE3 graphitization system, able to graphitise CO2 produced online either by EA combustion or by acid digestion using an IonPlus CHS. This is further supported by a suite of preparative chromatographic systems (GC and HPLC) and a range of complementary analytical instrumentation (e.g. IRMS, GC-MS).
The BrisMICADAS AMS at the University of Bristol.
The installation, commissioning, and acceptance of the BrisMICADAS was completed in January of 2016. Since then, we have undertaken an extensive pretreatment development and validation process using pretreatment methods based upon long-established and published protocols. Our routine pretreatment procedures and other sample preparation methods are outlined below alongside data obtained from a range of standards and intercomparison samples as part of our validation program and measured during routine analyses.
SAMPLE PREPARATION METHODS
Pretreatment Methods
Every routine pretreatment method employed within the BRAMS facility is assigned a short pretreatment code. Appropriate pretreatment methods for submitted samples are identified after discussions with the submitter and the pretreatment code is tied to the analysis of that sample and reported alongside the 14C determination. All sample, pretreatment, graphitization, and measurement data are stored in a central database within the laboratory management package (LMP) developed at ETH, Zurich and IonPlus.
All samples are pretreated in batches of up to ca. 30 samples, containing at least one true replicate where possible, in addition to processing standards and blanks. Our sample submission form asks for an approximate expected age range to assist with the selection of appropriate standards and for information regarding any potential contaminants (such as glues, consolidants, varnishes etc.) to highlight the need for any additional physical or chemical cleaning steps required before pretreatment.
All pretreatment batches and AMS magazines contain appropriate 14C blanks. Wherever possible, these are matrix-matched and undergo identical pretreatment procedures to unknown samples. A chemical blank (Phthalic anhydride, Sigma Aldrich) is also included in all AMS magazines to monitor blank contributions from EA combustion and graphitization. Similarly, wherever possible, matrix- and age-matched standards are employed to ensure the accuracy of the 14C dates obtained. The blanks and standards currently in use are outlined in Tables 1 and 2, respectively.
Current 14C blanks in use at BRAMS.
a Scott et al. (Reference Scott, Harkness and Cook1997).
b IAEA-C9 and FIRI A (Boaretto et al. Reference Boaretto, Bryant, Carmi, Cook, Gulliksen, Harkness, Heinemeier, McClure, McGee, Naysmith, Possnert, Scott, van der Plicht and van Strydonck2002).
c (Cook et al. Reference Cook, Higham, Naysmith, Brock, Freeman and Bayliss2012).
Determined F14C values for a range of standards and intercomparison samples.
“n” gives the number of replicate analyses. In the case of intercomparison samples, these are true replicates including all pretreatment steps. “Sig” denotes whether the determined value and the reference value are: ** indistinguishable at the 2σ level; * indistinguishable at the 1σ level.
a Scott et al. (Reference Scott, Harkness and Cook1997).
b Scott et al. (Reference Scott, Cook, Naysmith, Bryant and O’Donnell2007).
c Scott et al. (Reference Scott, Cook and Naysmith2010a).
d Scott et al. (Reference Scott, Cook and Naysmith2010b).
e Scott et al. (Reference Scott, Naysmith and Cook2017).
† weighted mean if n>1.
During pretreatment, all steps are performed in acid washed and furnaced (480°C, >4 hr) borosilicate culture tubes. The use of plastics is avoided during all wet chemical procedures wherever possible to avoid potential introduction of exogenous C (this is unavoidable when using ultrafilters). All steps are carried out at room temperature unless otherwise stated. All laboratory equipment including lyophilizer, centrifuge and heating blocks is dedicated to the preparation of samples for 14C and archaeological science analyses.
Bone, Tooth Dentine, Antler and Ivory (BC, BCU)
Bone samples are taken avoiding areas of archaeological, pathological and/or aesthetic significance. The area around the sampling location is surface cleaned with a rotary tool and samples are either drilled or cut and crushed to obtain a coarse bone powder.
Bone collagen extraction and purification follows a modified Longin method (Longin Reference Longin1971) as outlined by Brock et al. (Reference Brock, Higham, Ditchfield and Ramsey2010). Briefly, coarse bone powder is demineralized in 10 mL 0.5 M HCl (~18 hr) before rinsing three times with ultrapure (18.2 MΩ·cm) MilliQTM deionized water (henceforth, ultrapure water) and subsequent removal of “humic acids” in 0.1 M NaOH (30 min). Samples are further rinsed as before and an additional acid wash (0.5 M HCl, 30 min) performed. Samples are rinsed again before gelatinization (pH 3, 75°C, 20 hr). Gelatin solutions are filtered through pre-combusted tightly packed glass fiber (11 μm, Assistent, Germany) plugs before being lyophilized. Samples are dated if the collagen yield is above 1% and the C:N ratio of the collagen is between 2.9 and 3.5 (Ambrose Reference Ambrose1990; van Klinken Reference van Klinken1999). Typically, no ultrafiltration step is employed unless specifically requested (BCU), in which case, the method of Brock et al. (Reference Brock, Higham, Ditchfield and Ramsey2010) is employed after careful precleaning of ultrafilters according to the method described by Brock et al. (Reference Brock, Ramsey, Higham, Bronk Ramsey and Higham2007).
Organics, Charcoal, Plant Macrofossils (ABA, A)
Samples are treated with either an acid-base-acid (ABA) or with a gentler acid only (A) method depending on the size, fragility and preservation of the specimens. Our standard ABA pretreatment procedure consists of an acid wash (1 M HCl, 75°C, 20 min, or until effervescence stops) to remove any “fulvic acids” and carbonate present. Samples are then rinsed three times with ultrapure water before a base wash to remove “humic acids” (0.2 M NaOH, 75°C, 20 min). This step may be repeated until the solution is colourless to ensure complete removal of “humic acids.” Samples are rinsed again and a final acid wash is performed (1 M HCl, 75°C, 1 hr) to remove any base-liberated “fulvic acids” and any atmospheric CO2 absorbed during the base wash. After a final rinsing process, samples are lyophilized. For small or fragile samples, the base wash may be omitted, and/or the temperatures and durations of the washes reduced to preserve sufficient sample for dating.
Wood (BABAB)
Cellulose extraction from wood samples follows the base-acid-base-acid-bleach cellulose extraction method described by Němec et al. (Reference Němec, Wacker, Hajdas and Gäggeler2010). Briefly, thin wood shavings are subjected to an initial base wash (1M NaOH, 75°C, overnight) to open up the porous structure of the wood, an acid wash (1M HCl, 75°C, 1 hr) to remove any carbonates and “fulvic acids” present, a second base wash (1M NaOH, 75°C, 1.5 hr) to remove “humic acids” and ligninous material, a further acid wash (1M HCl, 75°C, 1 hr) and a final bleaching step (2 hr in 5% NaClO2 adjusted to pH 2 with HCl, held at 75°C for 2 hr before placing in an ambient ultrasonic bath for 15 min) to obtain the holocellulose fraction. Samples are washed with 3 × 10 mL ultrapure water between each step and the cellulose lyophilized before combustion and graphitization.
Carbonized Organic Residues on Potsherds (CR)
Residues are gently removed from sherds using a scalpel. Due to the likelihood of a significant fatty acid component in the charred residues and their small sample sizes no base wash is employed; an acid only pretreatment is applied with ultrasonication as described in Brock et al. (Reference Brock, Higham, Ditchfield and Ramsey2010). Briefly, samples are demineralized in 1M HCl for 1 hr before being placed in an ultrasonic bath for 15 min. Samples are rinsed with 4× ultrapure water and untrasonicated in fresh ultrapure water for 5 min. This final ultrasonication is repeated until the supernatant is colourless. Samples are briefly acidified (1 M HCl, 5 min), rinsed in ultrapure water and lyophilized prior to combustion and graphitization.
Carbonates and Calcined Bone (AH, AHN, AHO)
Care must be taken when dating shells that aragonitic species are targeted. If necessary, shells are tested with Fiegl’s solution (Friedman Reference Friedman1959) prior to analysis to determine the presence of potentially recrystallized calcite. If necessary, coral and shell samples are surface abraded using a rotary tool and coarsely cut or crushed before cleaning in MilliQ water with ultrasonication. Samples are then etched using 0.2 M HCl to remove the outer layer (ca. 20%) of potentially recrystallized carbonates. Samples are dried and transferred to acid-washed and precombusted exetainers. The exetainers are then sealed and the headspace replaced with helium gas using an IonPlus CHS (Wacker et al. Reference Wacker, Fülöp, Hajdas, Molnár and Rethemeyer2013). Orthophosphoric acid (1 mL 85% v/v) is injected through the septa and samples heated at 70°C until CO2 evolution has ceased. The CO2 generated during the acid hydrolysis is transferred to the AGE3 graphitization system in a stream of helium.
If no acid etching step is performed (e.g. in the case of foraminifera or sedimentary carbonate samples), the pretreatment code given is AHN.
Calcined bone analysis (AHO) follows a similar procedure, but with a post-acid hydrolysis oxidation. Calcined bone samples are cleaned using a rotary tool before being coarsely crushed. Samples are then transferred to acid washed and precombusted exetainers and treated with 1 M acetic acid at room temperature for 24 hr to remove any carbonates originating from the burial environment before rinsing three times with ultrapure water and lyophilization (Snoeck et al. Reference Snoeck, Staff and Brock2016). The exetainers are then sealed and the headspace replaced with helium gas. 1–2 mL 85% orthophosphoric acid is injected through the septa and samples are heated to 70°C until CO2 evolution has ceased. The CO2 generated during this process is transferred from the exetainer headspace to the AGE3 graphitization system in a stream of helium. However, in addition to CO2, the products of acid digestions of cremated bones often contain sulfur-containing species amongst other contaminants which are known to inhibit the graphitization reaction, so an online combustion/scrubbing system in the transfer line from the Carbonate Handling System to the Automatic Graphitization System is employed. This system will be fully described in a forthcoming publication.
Removal of Consolidants or Conservation-Related Contamination
If the analysis of conserved artifacts cannot be avoided, depending on the nature of the conservation treatment, several pretreatment methods can be employed prior to the commencement of the standard pretreatment processes. If the nature of the treatment is unknown, an acetone-methanol-chloroform pretreatment is employed as described in Brock et al. (Reference Brock, Higham, Ditchfield and Ramsey2010). Where the conservation treatment is known, methods tailored to the removal of those contaminants are employed as described in Brock et al. (Reference Brock, Dee, Hughes, Snoeck, Staff and Bronk Ramsey2017).
After any such procedure, careful attention is paid to the elemental composition of the samples upon combustion, as these data could highlight the incomplete removal of conservation-related contamination. As with all pretreatment procedures, matrix-matched standards and blanks are treated alongside samples.
Compound-Specific 14C Analyses
We have been developing various pretreatment and sample preparation techniques for the compound-specific 14C analysis of a range of sample types, including lipids absorbed within archaeological potsherds. These methods are described elsewhere (Casanova et al. Reference Casanova, Knowles, Williams, Crump and Evershed2017, Reference Casanova, Knowles, Williams, Crump and Evershed2018).
Graphitization
Samples are graphitized using an IonPlus automatic graphitization equipment (AGE3) system, as described in (Wacker et al. Reference Wacker, Němec and Bourquin2010c). Briefly, after either combustion using an Elementar Vario Isotope Select elemental analyzer or acid digestion using an IonPlus CHS, CO2 from samples is transferred to the AGE3 system in a stream of helium carrier gas and trapped on a zeolite trap. The CO2 is then thermally desorbed into one of 7 graphitization reaction tubes containing a conditioned (oxidized and subsequently reduced) iron catalyst. H2 gas is introduced and graphitization is performed by heating the tubes to 580°C for 2 hr. Water is removed from the reaction volume by cryogenic trapping using a Peltier cooler. Pressures and temperatures are recorded for each reactor over the duration of the conditioning, loading, and graphitization processes. Typically, full-sized samples contain around 1 mg C, but samples of 200 μg C produce reliable dates when analyzed alongside size-matched samples and blanks.
Graphite samples are pressed into cathodes (targets) using an IonPlus PSP. Targets are pressed from the rear, ensuring a smooth and clean sputtering surface every time. The pneumatic press ensures the same level of compaction for all samples.
Accelerator Mass Spectrometry
14C analysis is performed on the BrisMICADAS, designed and built by the Laboratory of Ion Beam Physics, ETH, Zurich. Details of the accelerator and analytical approach are given in Synal et al. (Reference Synal, Stocker and Suter2007) and Wacker et al. (Reference Wacker, Bonani, Friedrich, Hajdas, Kromer, Němec, Ruff, Suter, Synal and Vockenhuber2010a). A full AMS magazine contains 39 cathodes including samples, standards and blanks. Typical ion currents during routine analysis of graphite targets are around 30 μA 12C+. Samples are analyzed such that the OXAII standards achieve at least 500,000 counts, but up to 1,000,000 is commonplace. For a full magazine, this takes ca. 2 days. Data reduction is performed using the software package BATS, as described in Wacker et al. (Reference Wacker, Christl and Synal2010b). All magazines contain at least 3–4 NIST SRM-4990C Oxalic Acid II targets as normalization standards in addition to our in-house graphitization blank (phthalic anhydride), matrix-matched blanks, matrix- (and, where possible, age-) matched standards and graphitization standards (IAEA-C7 and/or -C8). All standards and blanks are size-matched to the samples in each magazine.
Standards, and Data Quality Assurance
Reference materials and 14C blanks are run with every batch of processed samples and in every magazine. Careful monitoring of the long-term data from these ensures accurate dates are obtained. These standards and blanks are chosen to match the sample type (and in the case of standards, wherever possible, approximate age) and undergo identical pretreatment procedures to the unknown samples. Furthermore, every batch of samples contains at least one true replicate sample, whereby the submitted sample is subsampled twice, and both subsamples undergo independent pretreatment. Each batch (including all standards, replicates and blanks) is treated with the same batches of reagents. Archives of solvents and reagents are retained at least until analysis is complete.
Wherever possible, matrix- and age-matched processing standards are analyzed alongside unknowns. These consist of either samples from previous laboratory intercomparison exercises, or in-house standards comprising large samples which have been dated many times alongside other standards to obtain a reliable reference date. We are currently in the process of establishing robust dates for these in-house standards. Matrix-matched 14C dead “blanks” are also employed wherever possible. These include bone, wood, lignite and carbonate blanks in addition to our chemical blank.
Total analytical uncertainty is calculated within the BATS data reduction software and represents the combined uncertainty resulting from counting statistics, isotopic fractionation correction, blank subtraction uncertainty and a sample scatter factor (Wacker et al. Reference Wacker, Christl and Synal2010b). Appropriate values for the sample scatter factor are determined by ensuring that multiple long-term replicate determinations on reference materials and standards with uncertainties including this sample scatter factor pass a chi-squared test and its magnitude is adjusted to achieve a right-tailed p-value (α) of as close to 0.5 as possible. The suitability of this sample scatter factor is regularly re-evaluated for each sample type. This sample scatter factor is then incorporated into the total measurement uncertainty in the BATS software package using a sum-of-squares based approach.
RESULTS AND DISCUSSION
F14C values from IAEA-C7 and C8 oxalic acid standards (in measurement order) are shown in Figure 2 (A) and (B) alongside a graph of all F14C data for all full-sized IAEA C7 and C8 targets analyzed at BRAMS since the commissioning of the accelerator, normalized such a value of 1 represents the true value of the standards (C). It is clear that the determined values for the standards agree well with the reference values for these materials and that the long-term accuracy and precision are good. The normalized F14C values demonstrate that the observed scatter is in agreement with that which would be expected on a purely statistical basis.
(A) F14C values determined for IAEA-C7 oxalic acid to date arranged in order of analysis; (B) F14C values determined for IAEA-C8 oxalic acid to date arranged in order of analysis; (C) normalized F14C values from IAEA-C7 and -C8 targets arranged in order of determined normalized F14C value. Horizontal black lines represent literature values. The red line indicates the expected trend from normally distributed data. (Please see electronic version for color figures.)
As part of the commissioning of the BRAMS facility, our combustion, carbonate digestion, graphitization and sample pretreatment protocols have been validated by analyzing many samples with a range of known (or consensus) 14C ages covering a range of sample types and pretreatment methods. A summary of the results of these analyses using the methods described herein are given in Table 2.
Where more than one replicate analysis has been performed, the weighted mean and associated reduced error are reported, in addition to indicators of whether these measurements lie within 1σ or 2σ of the reference (known/consensus) value. Our results show good agreement with the accepted F14C values for all samples, covering a range of sample types and ages, with the proportion of values lying within 1σ and 2σ uncertainties being slightly better than would be expected on probabilistic grounds.
Typical values obtained for our blanks are given in Table 3. We are in the process of characterising our lignite blank and anticipate that it will be in routine use in the coming months.
Mean blank values measured at BRAMS.
SUMMARY
This paper represents a snapshot of the methods employed at BRAMS and the results of an extensive program of method validation and 14C determinations on intercomparison samples and standards. Our aim in presenting this work is to demonstrate our ability to reliably perform 14C measurements on a range of routine sample types of varying ages and also to serve as a starting point from which it is possible to document improvements and innovations in the pretreatment and processing of samples in the future.
Ongoing and planned work includes: (1) investigations into exogenous carbon introduction and removal of post-depositional environmental contaminants during a range of different pretreatment methods using the high-field NMR approach developed during work by Casanova et al. (Reference Casanova, Knowles, Williams, Crump and Evershed2017) alongside organic mass spectrometric methods, (2) continue the acquisition and characterization of in-house standards, (3) continue to review sample scatter values for the full range of pretreatment methods as more standards are analyzed, and (4) continue work to develop and adapt methodologies for the 14C analyses of small samples.
ACKNOWLEDGMENTS
The authors would like to thank H-A Synal, L Wacker, S Fahrni and members of LIP, ETH, Zurich and IonPlus for their help and advice before, during and after the commissioning of the BrisMICADAS and F Brock for helpful discussions regarding pretreatment protocols. NERC, BBSRC, and the University of Bristol are thanked for their financial contributions which enabled the creation of the BRAMS Facility.
