Optimization of the LLNL/CAMS gas-accepting ion source and 1 MV compact AMS for natural abundance radiocarbon analysis of CO₂

Abstract

The Lawrence Livermore National Laboratory – Center for Accelerator Mass Spectrometry (LLNL/CAMS) 1 MV AMS system was converted from a biomedical AMS instrument to a natural abundance ¹⁴C spectrometer. The system is equipped with a gas-accepting hybrid ion source capable of measuring both solid (graphite) and gaseous (CO₂) samples. Here we describe a series of experiments intended to establish and optimize ¹⁴CO₂ measurement capabilities at natural abundance levels. A maximum instantaneous ionization efficiency of 8 % was achieved with 3 % CO₂ in helium at a flow rate of approximately 220 μL/min (3.5 μg C/min). For modern materials (e.g., OX I) we measured an average of 240 ± 50 14C counts/μg C, equivalent to a total system efficiency of approximately 3 %. Experimental CO₂ samples with F¹⁴C values ranging from 0.20 to 1.05 measured as both graphite and directly as CO₂ gas produced equivalent values with an average offset of < 2σ.

1. Introduction

The Lawrence Livermore National Laboratory Center for Accelerator Mass Spectrometry (LLNL/CAMS) 1 MV compact AMS system was repurposed from a biomedical AMS system to a natural abundance ¹⁴C spectrometer [1]. The system is based around a 1 MV National Electrostatics Corporation (NEC) Pelletron tandem accelerator and a heavily modified NEC gas-accepting hybrid ion source (MCGSNICS). Prior to its conversion, the gas-accepting source was coupled to a moving wire interface and used extensively to measure both discrete liquid drops containing ¹⁴C-labeled materials and the eluent stream from a high-pressure liquid chromatography system [2,3]. While the discrete injections of small (sub μg) quantities of CO₂ gas produced by the moving wire interface were sufficient for the detection of ¹⁴C at the required accuracy and precision of these biomedical studies [4], this approach is not feasible for the measurement of natural abundance materials, particularly if analytical precision comparable to graphite ¹⁴C-AMS (<1 % uncertainty) is desired. As a result, since its conversion, this system has been used exclusively for the routine measurement of solid (graphite) samples. However, direct measurements of CO₂ gas are desirable as they offer several benefits over analyses requiring the reduction of sample CO₂ to graphite, including reduced sample preparation time and costs, reduced risk of contamination with exogenous C, and size-independent ion currents and source performance.

Here, we describe a series of experiments intended to establish and optimize ¹⁴CO₂ measurement capabilities at natural abundance levels including: 1) optimization of CO₂ gas concentrations and flow rates, 2) optimization of ion source and beamline parameters, 3) validation of measurement capabilities using background and standard materials, and 4) comparison of gas and graphite results from splits of experimental CO₂ samples.

2. System configuration

2.1. AMS configuration

The current configuration of this system is based around a NEC gas-accepting ion source and NEC 3SDH-1 1 MV Pelletron tandem accelerator, described in detail in Broek et al., 2021 [1]. The low energy (LE) side consists of two sets of electrostatic horizontal and vertical steerers, an einzel lens, insulated vertical magnetic steerer, electrostatic acceleration gap (bouncer), 90-degree injection magnet, and off-axis Faraday cup for measurement of LE ¹²C⁻ ions. The high energy (HE) side consists of a 90-degree analyzer magnet, off-axis Faraday cup for measurement of ¹³C⁺ ions, vertical magnetic steerer, 90-degree electrostatic spherical analyzer, and silicon surface barrier particle detector for ¹⁴C⁺ ions (Canberra FD 150–18–300 RM). ¹⁴C measurements are made relative to ¹³C⁺ current measured at the HE off-axis Faraday cup and normalized to the measured ¹⁴C/¹³C ratio of NBS oxalic acid I (OX I; SRM-4990B).

2.2. Ion source design

The NEC gas-accepting ion source (Multi-Cathode Gas Source of Negative Ions by Cesium Sputtering; MCGSNICS) was heavily modified based on the work of Southon et al., [5–7] to increase ion output and accept components that are interchangeable with the “Livermore” sources of the LLNL/CAMS 10 MV AMS. More detailed discussion of these modifications can be found in Ognibene and Salazar 2013 and Broek et al., 2021 [1,4]. Most relevant to this study is the reduction of material in the interior of the source to increase pumping efficiency and the addition of an extra turbomolecular pump affixed to the extractor/ einzel assembly section of the source beamline to better handle the higher source pressures resulting from the high carrier gas (helium) flow during gas sample introduction.

2.3. Gas targets

The targets used for these measurements were custom designed and manufactured to fit into a standard 39 sample NEC wheel. A detailed description of the initial target design and optimization is described in Salazar and Ognibene 2013 [8]. The targets consist of an aluminum shell with equivalent shape and exterior dimensions as the solid sample targets and a titanium piece which is inserted and held firmly by the aluminum shell. The rear of the aluminum shell is conical and contains a 1.09 mm (0.043”) ID hole which mates to the rounded tip of the gas feed tubing. The titanium inserts are pre-cleaned in a solution of hexane and dried completely prior to target assembly. Following cleaning there is still some residual surface carbon contamination which can be removed by pre-sputtering the target for approximately 2 min prior to sample injection. Gas is delivered to the back, flows through the target, and is forced out the front of the cathode where it is sputtered by energetic cesium ions [4]. When completely assembled, the front surface of the titanium insert is positioned at the same approximate depth as the graphite surface of a typical solid sample target.

2.4. Gas sample preparation

Samples were converted to CO₂ gas and purified prior to introduction to the gas source as follows: OX I, secondary standards, and backgrounds materials equivalent to approximately 10 mg C each were weighed into 7 mm OD quartz tubes with 200 mg CuO, sealed under vacuum, and combusted in an oven at 900C for 5 h. Combusted samples were individually cracked into a vacuum line typically used for graphitization, water was removed with a dry ice / isopropanol slush trap, non-condensable gasses were pumped away, and the residual CO₂ gas was expanded into 12 separate volumes and allowed to equilibrate for 20 min. Each of the 12 gas splits, containing approximately 500 μg C, were then sealed into individual 6 mm OD quartz tubes of equal length for use in the tube cracker system (see section 2.5).

2.5. Tube cracker for sample introduction

We designed and built a manually operated tube cracker system (Fig. 1) to introduce pre-purified CO₂ to the source based on the general configuration in Wacker et al., 2013 [9]. Scored 6 mm OD quartz sample tubes were inserted into a flexible section of PTFE tubing and the entire system volume was evacuated and flushed with 3 volumes of ultra-high purity (UHP) helium. The sample tube was then cracked by flexing the PTFE tubing, and the released CO₂ was measured manometrically. The CO₂ was then expanded into a 10 mL gas-tight syringe and diluted to approximately 3 % by volume (determined by gas pressure) with the appropriate volume of UHP helium. The syringe was then compressed to achieve the desired injection pressure, and the 4-port 2-position valve was actuated to introduce the diluted sample gas to the ion source through a 75 μm ID × 2.6 m long capillary. The syringe was continually compressed to maintain a constant inlet pressure and consistent gas flow. As a result of the relative volumes of the cracker (1.7 cm³), transfer lines (3.2 cm³), and gas syringe (10.0 cm³), a maximum of 65 % of the sample CO₂ can be introduced to the source.

Figure 1.

Schematic of tube cracker / gas handling system.

3. Results and discussion

3.1. Hybrid ion source considerations

The NEC MGSNICS is designed to measure both graphite and gas samples, and, in theory, both can be measured interchangeably on the same sample wheel by simply opening or closing a valve on the gas inlet. However, in practice, we found that measurement of gas samples significantly impedes the ability to subsequently conduct high quality graphite analyses. During gas measurement, the ionizer becomes coated in titanium from sputtering of the gas target insert. Starting with a new ionizer, this titanium coating will reduce the current output of the source by 10–20 % over the course of a few days of measurements. With gas samples, this reduced current can be mostly counteracted by increasing the temperature of the cesium reservoir from the normal operating temperature of 160C to approximately 170–175C. This effect can also be mostly reversed by grinding the titanium off the surface of the ionizer. However, for graphite measurements, we found the reduction in current to be both more dramatic and seemingly irreversible. An ionizer used for gas measurements will produce as little as 50–60 % of the current from a graphite target as an unused ionizer, and no amount of grinding or increased cesium temperature will fully recover the lost current. As a result, we found it necessary to have separate gas and graphite designated ionizers, which would be swapped between measurement of solid and gaseous samples, necessitating an opening and full cleaning of the ion source between sample types.

3.2. System optimization

3.2.1. Gas flow optimization and ionization efficiency

Four CO₂ / helium mixtures (1, 3, 5, and 10 %) were tested at a range of flow rates (~100 to 300 μL/min) to determine the optimum sample introduction conditions (Fig. 2). We found that 3 % CO₂ (v/v) at a flow rate of approximately 220 μL/min (3.5 μg C/min) resulted in the most efficient production of C⁻ ions while maintaining a reasonable source vacuum. Ion source pressures greater than 5×10⁻⁶ Torr (equivalent to a gas flow rate of approximately 250 μL/min) significantly impacted instrument backgrounds, beyond the levels discussed below, and put undesirable strain on the system’s turbomolecular pumps. Underthese optimized conditions, the source produced an ion beam with ≥ 15 μA of ¹²C⁻ current, representing an instantaneous conversion of approximately 3.5 % of CO₂ gas flowing into the source to C⁻ ions. Subsequent measurements with newly manufactured gas targets at this CO₂ concentration and flow rate more than doubled the current output, resulting in ¹²C⁻ currents up to 30 μA, representing up to 8 % ionization efficiency. It is uncertain what led to this increased ion output as there was no intentional change to the target design. However, it is possible that slight differences in the Ti surface depth, entrance aperture, or Ti purity could influence the ionization efficiency. Alternately, it is conceivable that target storage conditions could lead to a change in performance, as the first set of targets were stored in atmosphere for several years prior to use and new targets were stored under vacuum following the initial hexane cleaning and assembly.

Figure 2.

Relationship between CO₂ concentration, flow rate, C ion current, and ionization efficiency. Each point represents the average of three 30 s measurements and error bars represent the standard deviation of the three values. The top panels show the total volumetric gas flow (CO₂ and He carrier gas) and the bottom panels show the calculated mass flow of C (in μg). Red shading indicates gas flow rates which led to undesirably high source pressure. The highest currents were produced from 10% CO₂ at relatively low flow rates, whereas the most efficient ionization occurred with 1% CO₂ at high flow rates. However, the conditions which produced the most efficient ionization also resulted in undesirably high source pressure. Based on these results, all subsequent measurements were made using 3 % CO₂ at a flow rate of approximately 220 μL/ min (3.5 μg C/min).

Our average current of approximately 25 μA of C is equivalent to the maximum reported current of the gas-accepting ion source of the Ion Plus MICADAS [10], however, it is significantly higher than the typical reported averages of the MICADAS (5–15 μA; e.g., [10,11]), other NEC MCGSNICS sources (<12 μA; e.g., [12]), and the HVEE ions source in gas operation mode (12 ± 1 μA; e.g., [13]).

3.2.2. Ion source optimization

In order to assess the optimum Cs sputter energy, the ¹²C⁻ current (measured at the low-energy offset Faraday cup) was monitored while adjusting both the cathode and bias voltages (to maintain the same total C ion energy) (Fig. 3). Ion currents (and ionization efficiency) scaled with cathode voltage to a maximum at 9 kV. Above 9 kV, the average current decreased and became more variable. This behavior is likely partially attributable to the cesium focus on the titanium surface. We found that ion output and ionization efficiency were highly sensitive to cathode positioning, particularly in the z-direction (along the beam axis), with the highest currents produced when the focal point of the cesium was well centered and slightly beyond the front surface of the titanium insert.

Figure 3.

Relationship between cathode voltage (Cs⁺ ion energy) and CO₂ ionization efficiency. For this test, ¹²C⁻ current was monitored at the low-energy offset Faraday cup while adjusting both the cathode and bias voltages at 0.5 kV increments to maintain the same total C ion energy. Each point represents the average of five 60 s measurements and error bars represent the standard deviation of the five values. Based on these results, a 9 kV cathode voltage was used for all subsequent measurements.

3.2.3. Stripper optimization

To optimize the Ar gas-stripper pressure we measured CO₂ from a combusted coal sample while sweeping the Ar pressure, as measured at the exhaust side of the recirculating turbos, from approximately 35 to 70 mTorr. We found that an Ar stripper pressure of approximately 45 ± 1.5 mTorr resulted in the lowest instrument backgrounds (Fig. 4). This pressure is approximately 10–15 mTorr lower than the optimized range for graphite, and improved measured background ages by 2000 to 3000 years over measurements made at the higher, graphite optimized, pressure range. This result suggests that fewer carbon hydrides are formed during the production of C⁻ ions from CO₂ than from elemental carbon. We hypothesize that this could be related to the increased levels of oxygen present in the source and its availability to react with protons produced during cesium sputtering.

Figure 4.

Influence of the Ar stripper gas pressure on raw ratio (¹⁴C/¹³C), measured ¹⁴C age, and transmission through the tandem accelerator (measured as the ratio of high energy ¹³C⁺ to low energy ¹²C⁻ and corrected for the known ¹³C/¹²C ratio) of CO₂ gas from an untreated coal sample. Each point represents an approximately 15-minute measurement of pre-purified splits of CO₂ from a single combustion of untreated coal. Dark grey shading represents the ideal stripper pressure for gas source measurements (45 ± 1.5 mTorr). Light grey shading represents the ideal stripper pressure for graphite measurements with the same system [1].

3.2.4. Backgrounds

During gas measurements, there was a significantly higher background compared to graphite. Samples of untreated coal analyzed as graphite on this system have an average measured age of 40 ± 2 ka, whereas CO₂ from the same sample materialmeasured with optimized parameters had an average apparent age of approximately 25 ± 2 ka. We hypothesize that this is caused by angular scattering of ions resulting from the increased ion source and low-energy beamline pressure, which could lead to more non-¹⁴C ions being measured in the rare isotope detector. The detector is a single anode silicon surface barrier detector and therefore does not allow for any particle identification. We found that winding in the three sets of adjustable apertures (slits) on the system to the approximate edge of the beam improved measured background ages by 3000 to 4000 years without any changes to measurement accuracy or precision. However, even with this improvement, significant background corrections are currently required to produce accurate data (see section 3.4). Installation of additional apertures or baffles which can both prevent carrier gas from reaching further into the beamline and block scattered non-¹⁴C ions from reaching the detector, especially on the low energy injection side where beamline pressures are elevated, are expected to further improve backgrounds, allowing for the measurement of older materials.

3.3. Sample analysis

In this study, standards and backgrounds were measured by injecting three replicate tubes of pre-purified CO₂ gas onto a single gas target in succession for approximately 15–20 min each (Fig. 5; Supplemental Fig. 1). Following injection, there is a substantial amount of residual C on the target available for continued sputtering and measurement. Therefore, in this study, sample measurement was typically continued for approximately 30 min post-injection to both increase the number of acquired ¹⁴C counts and to investigate the differences in data quality and system performance with and without gas flow into the source. This residual C present on the gas target following sample injection also likely explains the increase in ion current with each subsequent injection in the three-injection scheme, where the measured C⁻ current during the second and third injections represent ions generated from both newly injected CO₂ and residual C.

Figure 5.

Typical sample injection scheme for samples measured in this study. Top panel: raw measured ratio (¹⁴C/¹³C) of each individual injection (blue) and the post- injection phase (green). Error bars represent the counting statistics derived uncertainty. Black lines represent the average and standard deviation of measured raw ratios of OX I graphite over the last year. Bottom panel: ¹²C⁻ ion current and corresponding ¹⁴C counts during and after injection of n = 3 splits of OX I CO₂.

During sample injection, the raw ¹⁴C/¹³C ratio is always elevated relative to graphite measurements of the same material, which, like the elevated backgrounds, we hypothesize is due to the higher ion source pressures and resulting angular scattering (Fig. 5). However, during the continued measurement period (post sample injection) there is no gas flow into the source and the ion source pressure is comparable to typical pressures during graphite measurement (~1 × 10⁻⁷ Torr). As a result, the raw ratio is also comparable to that of typical graphite measurements of the same material. In addition, this results in a ~5000-year improvement in the measured age of background materials (Supplemental Fig. 2). While it is possible that some other effect within this system such as fractionation in the ion source or preferential loss of ¹³C in the accelerator could produce elevated ratios, these effects are not consistent with the observed differences between the gas injection period and sputtering of the residual graphite.

3.4. Measurement accuracy and background corrections

Three secondary standard materials: ANU sucrose, oxalic acid II (OX II), and Belfast pine (TIRI B) were measured following the protocol described in section 3.3 (Fig. 6). When normalized to the measured ratio of OX I CO₂, the calculated F¹⁴C values were on average >1σ different than the long-term average values of these materials measured as graphite. These offsets were driven by the elevated raw ratios described previously. Two different corrections were tested to account for this. First, a “simple” background correction was used, which subtracted the measured ratio of CO₂derived from a combusted ¹⁴C-free coal sample measured with the gas source. This correction generally improved measurement accuracy, but in some cases overcorrected the value due to counts attributable to actual ¹⁴C present in the coal material being erroneously subtracted. Next, a correction was applied that attempted to account for only the extra counts added by the scattered non-¹⁴C ions (henceforth referred to as a “scatter” correction). This correction subtracted the difference in ratio between coal-derived CO₂ measured with the gas source (which includes both scattered non-¹⁴C ions reaching the detector and actual ¹⁴C present in the sample) and a split of the same coal-derived CO₂ measured as graphite with the LLNL-CAMS 10 MV FN AMS system (which includes only actual ¹⁴C present in the sample). Due to a combination of factors, including measurement in the 4+ charge state, more charge and energy filtering elements, and a multi-channel gas ionization detector, the 10 MV FN AMS system is less susceptible to the influence of scattered ions and has significantly better backgrounds than the 1 MV compact AMS system, and is therefore better suited to determine the levels of actual ¹⁴C present in the ¹⁴C-free coal. In all cases, this “scatter” correction produced F¹⁴C values within 1σ of the typical graphite value. However, although the values were more accurate following correction, propagation of the background uncertainty caused the uncertainty of the measured values to increase from an average of ~3‰ to ~1%. More precise measurement of the instrument background would reduce the propagated uncertainty of the corrected values to closer to the Poisson uncertainty.

Figure 6.

F¹⁴C of three secondary standard materials measured as CO₂ gas and compared to the average value for graphite measurements over the last year. Each value is calculated from the combined active CO₂ injection phases of three individual injections. Blue points represent the uncorrected value with error bars representing the counting statistics derived uncertainty. Orange points were calculated with a “simple” background correction, which subtracted the measured ratio of a back- ground coal sample measured with the gas source. Green points were calculated using the “scatter” correction by subtracting the difference in ratio between coal CO₂ measured with the gas source and a split of the same coal CO₂ measured as graphite with the LLNL-CAMS 10 MV FN AMS system (to account for actual ¹⁴C present in the coal sample). Error bars on corrected values represent the propagated uncertainty of both the sample and the coal used for the correction. Black lines represent the average and standard deviation of measured raw ratios of each material as graphite over the last year.

3.5. Analysis of analytical unknowns and comparison to graphite

CO₂ generated via microbial respiration of soil organic matter in an ongoing soil warming experiment at LLNL/CAMS was purified and analyzed for ¹⁴C as both graphite and CO₂ gas (Fig. 7). Graphite targets were prepared from approximately 1 mg of respired C. For gas measurements, aliquots of the same CO₂ used for graphitization ranging from 600 to 2000 μg C were split into between two and five individual quartz tubes of 300 to 500 μg C each and introduced to the source using the tube cracker system in the same manner as the standards and background samples, with multiple tubes of the same sample gas injected onto each target. Of the eight samples measured, the absolute offset in F¹⁴C between graphite and gas samples ranged from 3 to 110 ‰ for uncorrected values (average = 45 ± 40 ‰), with higher offsets seen in lower F¹⁴C (older) samples (Fig. 7; right panel). Following the “scatter” correction which accounts for spurious ¹⁴C counts, the offsets decreased to a range of 1 to 65 ‰ (average = 22 ± 25 ‰). In some cases, the F¹⁴C values calculated from the post-injection measurement (see section 3.3) were more similar to the graphite values than the corrected values (offset range = 1 to 56 ‰; average = 22 ± 19‰). This demonstrates that future efforts to reduce the level of scattered ions reaching the ¹⁴C detector have the potential to greatly improve the accuracy of these measurements.

Figure 7.

Comparison of measured gas and graphite values of CO₂ produced by microbial respiration of soil organic carbon demonstrating the differences in accuracy following the different correction and measurement protocols described in the text. Error bars on uncorrected CO₂ and post-injection CO₂ values represent the counting statistics derived uncertainty and error bars on “scatter” corrected values represent the propagated uncertainty as described in the text.

4. Conclusions and future work

Maximum ionization efficiency of 8 % (average = 6.8 ± 1.0 %) was achieved by introducing a stream of 2.5 to 3.5 % CO₂ in UHP helium to the ion source at a flow rate of approximately 200 to 220 μL/min (2.7 to 4.2 μg C/min). For modern materials (e.g., OX I), we measured 240 ± 50 ¹⁴C counts / μg C, equivalent to a total system efficiency of approximately 3 %. Therefore, we calculate that the following sample sizes would be required to achieve specific levels of measurement uncertainty: 10 μgC ≈2%, 40 μgC ≈1%, 200 μgC ≈5‰, 1 mgC ≈2‰.It should be noted that the ionization efficiency numbers quoted throughout represent the instantaneous conversion of CO₂ gas flowing into the source to C⁻ ions, as measured at the off-axis low-energy ¹²C⁻ Faraday cup. Therefore, these values do not account for the substantial ion current which continues for an indeterminate amount of time after the injection of CO₂ gas has ceased. While the total amount of C available for continued sputtering and measurement was not assessed in this study, it represents a significant source of additional ¹⁴C which can increase measurement precision.

In order to assess the accuracy of measurements made with this system, a suite of secondary standards and experimental samples were analyzed. Following a “scatter” correction, which removes extra counts associated with non-¹⁴C ions reaching the ¹⁴C detector, the F¹⁴C value of secondary standard materials was within 1σ of typical graphite values. Experimental CO₂ samples with F¹⁴C values ranging from 0.20 to 1.05 measured as both graphite and directly as CO₂ gas produced values within 2σ, with an average absolute offset of 22‰.

With future modifications to the system and the collection of additional data, we expect significant improvements to the accuracy of these measurements and increased confidence in our ability to measure samples over a larger range of ¹⁴C concentrations. Possible future work includes: 1) the addition of beamline apertures to reduce non-¹⁴C ions at detector, 2) better quantification of instrument backgrounds to improve correction protocols, 3) reduction of the volume of the tube cracker system to allow a larger proportion of sample to be introduced to source, and 4) focused development of hardware and protocols to interface the gas handling system with peripherals to allow introduction of other sample types and reduce sample preparation requirements.