Modifying a sealed tube zinc reduction method for preparation of AMS graphite targets: Reducing background and attaining high precision

The sealed tube zinc reduction method for converting CO 2 to graphite for AMS 14 C measurements was originally developed for rapid production of graphite in biomedical tracer experiments. The method was usually thought to have low precision and a high background. We have modiﬁed the zinc reduction method originally outlined in Vogel [J.S. Vogel, Radiocarbon 34 (3) (1992) 344] by carefully controlling the amounts of reagents (zinc, titanium hydride and Co or Fe catalyst) and now routinely obtain a precision of 2–3 & and a relatively low background of (cid:2) 50,000 14 C years when analyzing for 14 C at the Keck Carbon Cycle AMS facility at UC Irvine. Fractionation of carbon isotopes does occur during graphitization and depends on the graphitization yield, which can be aﬀected by the amounts of reagents used and other conditions. The d 13 C of our zinc-reduced graphite is usually lighter by 2–3 & than the CO 2 from which it is made, but this is corrected for in our system by simultaneous measurement of 13 C/ 12 C along with 14 C/ 12 C by the spectrometer. This method is suitable for 14 C enriched samples, as well as natural abundance 14 C samples, especially those with modern 14 C contents. With improved precision and background, we believe that many disciplines can beneﬁt from this technique because of its low cost and rapid production of graphite.


Introduction
The production of uniform and high quality graphite targets is essential for the optimal performance of any AMS system. Graphite production for AMS 14 C measurements is most commonly achieved by hydrogen reduction of CO 2 with iron or cobalt catalyst [2,3]. The method has proved to be precise, accurate and have a low background while being suitable for all types of natural samples. The graphitization yield can be monitored during the reaction and is often close to 100%. At yields close to 100% isotopic fractionation between the original CO 2 and the graphite produced is very small, although large fractionation can occur with low yields. Samples as small as 2 lg carbon can be successfully converted to graphite [4]. Graphite produced by hydrogen reduction is also well behaved in the cesium sputter ion source, producing high and long-lasting current. However, building a hydrogen reduction line is relatively costly because of the requirements for pressure monitoring and the need for multiple ovens for hydrogen reactors. Hence, the number of reaction vessels is usually a limiting factor for the throughput of samples using hydrogen reduction. Jull et al. [5] and Slota et al. [6] developed zinc reduction methods for precise AMS 14 C measurements. Both techniques use a Zn finger at 400-500°C to reduce CO 2 to CO and an Fe finger at 700-750°C to covert CO to graphite. The main difference between the two methods is that Slota et al. eliminated the use of H 2 . Both methods were reported to be precise and accurate, but again, the graphite production was limited by the number of reaction units on the vacuum line. Slota and Taylor [7] experimented using a small sealed tube with Fe and Zn in the same tube to make graphite from CO 2 and claimed that the graphite was equivalent to the graphite produced in the two-fingered apparatus. However, a very long time was required for reduction.
The sealed tube zinc reduction method we modified is adapted from one described in Vogel [1]. This method was originally developed for converting CO 2 to graphite for biomedical research materials that had elevated 14 C from labeling experiments. To avoid sample-to-sample contamination, Vogel [1] used disposable gas manifolds for CO 2 transfer, trapped water in the combustion tube and sealed CO 2 into Pyrex tubes pre-loaded with zinc, TiH 2 , and a smaller tube with Fe or Co catalyst. Graphitization was accomplished by putting the entire sealed tube into a 550°C oven. This method allows rapid production of large numbers of graphite samples without cross-contamination ($300 samples/week/person). Ognibene et al. [8] modified the method by replacing the Pyrex tubes (which needed to be produced as well as sealed) with septa-sealed vials and eliminating the use of TiH 2 as a source of H 2 pressure. Instead, they used H 2 O directly from the combusted biochemical sample without separating it from CO 2 . With the modification, the authors reported that a single technician could prepare approximately 150 samples in an 8 h day. However, this sealed tube method has too low precision (10-20& reported by Vogel [1] and 9.8& by Ognibene et al. [8]) and a background ($1 pMC and variable, reported by Vogel [1]) too high to be used routinely for natural abundance radiocarbon samples.
Ten years ago, we adapted the original sealed tube zinc reduction method from Vogel [1] for samples that have relatively high amounts of 14 C (i.e. with bomb 14 C to a few thousand years old), and which are not smaller than 0.5 mg C [9]. We achieved a precision of 5-6& and a background of $40,000 14 C years when our zinc-reduced graphite was measured for 14 C at LLNL CAMS facility, which measures only 14 C/ 13 C ratio and not 13 C/ 12 C. In the last several years, we have modified the original method by carefully controlling the amounts of reagents (zinc, titanium hydride and Co/Fe) and most importantly, by using the simultaneously measured AMS d 13 C for correction of isotopic fractionation during graphite production. We now routinely obtain a precision of 2-3& and a relatively low background of 50,000 14 C years when analyzing for 14 C at the Keck Carbon Cycle AMS facility at UC Irvine.
In this paper, we describe our modified Zn reduction procedure, the experiments done to investigate carbon iso-tope fractionation and report the precision and background obtained in our lab for measurements made over the last three years. With improved precision and background, we believe that many disciplines can benefit from this technique because of its low cost and rapid production of graphite.

Procedures
Dry organic solid samples equivalent to about 1-1.5 mg of C are combusted with clean CuO in 6 mm O.D. quartz tubes at 900°C for 2 h. CO 2 from the combustion is extracted cryogenically using the simple vacuum line shown in Fig. 1(a). The vacuum system consists of a mechanical pump and a liquid nitrogen trap between the pump and the line that prevents oil from back streaming to the line. The quartz tube with CO 2 is cracked open in the tube cracker. A dry ice and ethanol slurry is placed on the first trap in the line to freeze out water. A dewar filled with liquid nitrogen is placed on the second trap for freezing down CO 2 . Any non-condensable gases, mostly N 2 , are pumped away. The purified CO 2 is transferred to a known-volume reservoir and quantified by measuring the CO 2 pressure. This is used to calculate the yield of sample combustion, which we compare with an elemental analyzer (FISONS NA-1500) to check the volume of our measuring system.
Approximately 1 mg C as CO 2 is frozen into the reaction tube using liquid nitrogen (see Fig. 1(b) and next paragraph). The remainder is expanded into a finger on the left of the vacuum line where we insert a gas-tight syringe through the septum on a three way Ultra-Torr tee to take a subsample of the CO 2 that is injected into a He-filled vial for d 13 C measurement by Gas Bench coupled with IRMS (Thermo Electron Delta Plus).
The reactor tube with frozen CO 2 is sealed using a torch. The total length of the sealed tube is $110 mm long; a shorter tube may help graphitizing small samples [10]. The tube is placed upright in a metal heating block capable of withstanding temperatures of 550°C. The block and tubes are then placed inside a furnace for graphitization. The furnace is set to 500°C for 3 h and 550°C for 4 h (then cools slowly, usually overnight). Filamentous graphite forms at the surface of the catalyst.
Graphite samples are stored in the reaction tubes until immediately prior to pressing into sample holders for AMS measurement. To remove the sample from the tube without introducing glass fragments, we gently slide the inside 6 mm tube with graphite/catalyst mixture horizontally to the top end of the tube, then crack open the outside 9 mm tube near the bottom, just above the Zn/TiH 2 powders. Graphite is hammered into the central 1 mm diameter hole in the aluminum target holder [11]. The graphite targets are analyzed for 14 C at Keck Carbon Cycle AMS facility at UC Irvine [12,13].
The samples we discuss here were combusted, but we also use the same vacuum lines to purify CO 2 from heated molecular sieve traps [14,15], or from larger air samples from canisters [14]. These kinds of samples are more time consuming (for example, baking out and reactivating a molecular sieve trap will take $1.5 h per sample). Therefore, our approach is to have multiple extraction lines and to freeze the purified CO 2 directly into the reaction tube at the end of each extraction, with no need to store it in a break-seal tube first as would be the case if we transferred the sample to another line for H 2 -reduction of graphite.

Reaction tube preparation
The reaction tube preparation is as described in Vogel [1]; the major difference is that we have modified the amounts of Zn and TiH 2 reagents. The reaction tube is a 152 mm long, 9 mm O.D. Pyrex tube that is sealed at one end, with a small indentation $2 cm above the bottom (see Fig. 1(b)). Tubes are pre-baked at 500-550°C for 7 h. We use 30-35 mg zinc (Aldrich, #324930) and 10-15 mg titanium hydride (Alfa Aesar, #12857) that are placed in the bottom of the 9 mm tube. The catalyst -we now use Fe (Alfa Aesar, 350 mesh, #39813) [16], in a ratio of 3-5 mg catalyst per gram of C -is weighed separately into in a pre-baked, 50 mm long, 6 mm O.D. Kimble culture tube which is handled with gloves. The 6 mm tube is then inserted into the 9 mm tube so that it is suspended above the Zn and TiH 2 powders by the indentation (see Fig. 1(b)). The presence of zinc in graphite decreases sputtering yield in the ion source [8].
We use three different sizes of Chalazion curettes to measure the amount of reagents quickly and relatively accurately without weighing them out individually. With practice, a precision of ±1-2 mg can be achieved. The amounts of agents given above here are optimal for 1 ± 0.2 mg C even though samples as small as 0.1 mg C have been graphitized successfully under the same conditions.
Although we used Co as a catalyst for many years, we recently switched to Fe. Both catalysts make good graphite, but graphite formed with Fe performs better in the ion source of our AMS system. The Zn/Fe graphite is less fluffy than the Zn/Co graphite and thus easier to get inside the target holder. At the moment, we are not pre-cleaning any of the reagents, but are still able to obtain satisfactory background ($50,000 14 C years). In addition, we store prepared tubes -loosely covered with Al foil -in ambient lab air for up to a week or two without apparent adverse effects on background or precision.

Experiments
As pointed out by Vogel et al. [1], mass-dependent fractionation of carbon isotopes does occur during the zinc reduction graphitization process. The authors reported an average of 2.2& lighter in d 13 C for the finished graphite with 80% conversion yield than the initial CO 2 at reaction temperature of 435°C. For graphite with lower yields, the fractionation is larger, but still usually less than 5&. In order to investigate the degree and conditions of isotopic fractionations involved in this zinc reduction graphitization method and to search for the optimum graphitization conditions, we graphitized a set of OX1 and ANU standards using different amount of reagents (Zn from 8 to 51 mg, TiH 2 from 4 to 29 mg and Co from 2 to 8 mg). The amount of sample graphitized is either 1 mg C or 0.5 mg C. Most of the graphite was used for 14 C measurement and a small portion of the graphite was also measured for d 13 C and C% by EA-IRMS on a Thermo Electron Delta Plus. The graphitization yield is calculated as follows: where C% is the percentage of C in the graphite-catalyst mixture as determined by elemental analyzer, the graphite/catalyst mixture mass (mg) is determined by weighing before pressing the sample and C added , the amount of C (reported as mg C) in CO 2 originally sealed in the reduction tube, was determined using the pressure of CO 2 in a known volume on the vacuum line. Errors involved in the graphitization yield calculation come from three sources: the calibration of the measured volume for CO 2 , possible physical loss during transfer of the graphite-catalyst mixture for weighing and the C% measurement by EA. The relative errors for these three sources are estimated at 2.0%, 1.3% and 1.0%, respectively, yielding a total estimated error for an average graphitization yield of 86% (Table 1) of about ±2% (1r). Another method for determining gravimetric yield, based on comparing the weight of the 6 mm inner tube before and after graphitization consistently yielded mass increases in excess of what was expected from 100% conversion of CO 2 to C on the graphite. We attribute the extra weight to the deposition of ZnO on the glass wall. Results are shown in Table 1 and Figs. 2 and 3. Both OXI and ANU behave similarly in our tests. As expected, overall fractionation (d 13 C difference between graphite and initial CO 2 ) is minimized when graphitization yield is higher (Fig. 2). The minimum fractionation observed was 1&. Fractionation decreases when the Zn/TiH 2 ratio increases ( Fig. 3(a)). Too much TiH 2 will increase fractionation ( Fig. 3(b)), likely because CH 4 may form when excess H 2 is present. Also shown in Fig. 3(b), too much Zn (>50 mg/mg C) may increase the fractionation. The effect of catalyst amount on 13 C fractionation is not significant in the range we investigated. Based on the experiments described above, we set the amount of reagents and catalysts we currently use for a 1 mg C sample to be: 30-35 mg Zn, 10-15 mg TiH 2 and 3-5 mg Fe. Using these amounts, we typically obtain 85-90% yield, with an average fractionation of 2-3& in d 13 C.
Increasing graphitization time from 4 h at 550°C to 8 h had no effect on yield or isotopic fractionation. We did not attempt to stop the reaction early to study the effect of incomplete graphitization, or to test whether we could shorten the graphitization time, because it is convenient to leave the reaction overnight.
[10] on their sealed tube H 2 /Fe system. Other data points are from Vogel [1] (sealed tube Zn/TiH 2 system, shown as ''n'') and Verkoueren et al. [17] (sealed tube H 2 /Fe + Mn system, shown as ''+''). The dashed line calculated using a graphite-CO 2 of 0.972 is the best fit for data of Verkoueren et al. [17]. The open circles (s) are data from this work. The best-fit a graphite-CO 2 for our data set is 0.9868 ± 0.0013, shown as solid line with dashed lines for the 95% confidence bands. The smaller a graphite-CO 2 indicates that the actual graphitization reactions are influenced by kinetic fractionation.  13 C fractionation and reductant/C ratios. It shows that too much TiH 2 and too much Zn (>50 mg/mg C) may increase 13 C fractionation. The dashed straight line is the best linear fit for the Ti/C data, excluding two points of Zn = 10 mg and Zn = 18 mg (R 2 = 0.688). The dashed curve line is the approximate trend for the Zn/C data (except for the points of either very high or low Ti H 2 ).
Previous studies show that the presence of H 2 helps increase the rate of graphite formation [1]. McNichol et al. [18] observed that the reaction rate of CO to graphite is more rapid in the presence of H 2 (compared to Zn reduction alone with Fe as catalyst). As shown in reaction (1), TiH 2 provides the source of H 2 . Strictly speaking, our method is a combination of both H 2 and Zn reduction and the function of Zn here is not only for reducing CO 2 to CO, but most importantly, recycling the H 2 O back to H 2 (Reaction 6). Ognibene et al. [8] reported using H 2 O from combustion of biochemical samples (without separating it from CO 2 ) as the H 2 source, which simplified the CO 2 extraction greatly. However, the amount of water is highly variable in different types of natural samples and would be difficult to control in our application. Some samples may not have enough water to provide the source of H 2 while others may overwhelm the capacity of the Zn.
Other possible reactions within the sealed reaction tube [5,[17][18][19] are: As we can see, CH 4 could form in the reaction tube, especially in the presence of excess H 2 . McNichol et al. [18] show that in their H 2 /Co system, CH 4 rose to a low pressure of 0.03-0.08 atm when CO started to form and remained for the rest of the reaction (8 h). CH 4 was also present in the Zn/Fe method. Verkouteren et al. [18] and Verkouteren and Klouda [19] also demonstrated the correlation of CH 4 production and the presence of H 2 . For our Zn reduction method, the formation of CH 4 may be one of the causes for low graphitization yield and isotope fractionation, especially with high amounts of TiH 2 .
Graphitization condition is set at 500°C for 3 h and 550°C for 4 h in a Muffle furnace in our method. The temperature settings were adapted from Vogel [1]. The only change we made is prolonging the time at 550°C from 2 to 4 h, which helps small samples to graphitize. In the Zn reduction method with Fe as catalyst (no H 2 added) by Slota et al. [6], the temperature was set to 500°C for the Zn finger and 750°C for the Fe finger. The Arizona AMS facility's current setting for the Fe finger is 600°C. The temperature settings for the Zn reduction method (with small amounts of H 2 present) at the ANTARES AMS Center are 400°C for Zn and 600°C for Fe, respectively [20]. The lower temperature of 500°C in the first step may favor the CO formation and 550°C in the second step may favor the graphite formation over Fe. McNichol et al. [18] studied the effect of temperature, different catalysts and reducing agent on graphitization by studying the gas composition changes of CO 2 , CO and CH 4 during the formation of graphite. Their studies show that conversion of CO 2 to CO with H 2 is most rapid at 750°C, but conversion of CO to graphite is impossible at that temperature due to possible sintering of the catalyst. Also Fe appears to be a better catalyst (faster than Co) at temperatures between 575 and 650°C. We keep the reaction temperature between 500 and 550°C because we use Pyrex tubing, which is much cheaper and much easier to torch seal than quartz tubing.

Carbon isotope fractionation
Mass-dependent fractionation of carbon isotopes does occur during this sealed tube zinc reduction graphitization, depending on the amount of reagents used and other conditions. d 13 C of graphite can be lighter by 1-8 & than the initial CO 2 under the conditions we investigated. Fig. 2 illustrates the measured and modeled 13 C isotopic fractionation as a function of graphitization yield. The calculated fractionation shown as solid lines is from a simple closed system Rayleigh model [21]: where D graphite-CO 2 is the difference of d 13 C between graphite (product) and CO 2 (reactant) as &, F is the fraction of graphite formation and a graphite-CO 2 is the isotopic fractionation factor between graphite and CO 2 at the reaction temperature. The theoretical a graphite-CO 2 [22,23] is 0.9900 and 0.9913 at 500°C and 550°C respectively. A value of a graphite-CO 2 less than 1 indicates that the product graphite is depleted in 13 C relative to CO 2 at equilibration. If the reaction is 100% complete, i.e. all CO 2 is converted to graphite, then there should be no isotopic fractionation (D graphite-CO 2 = 0). Our experimental data shows that isotope fractionation exists when the graphitization yield is low. The best-fit a graphite-CO 2 is only 0.9868 ± 0.0013 (with 95% confidence and a chisq of 29.8), significantly less than the theoretical fractionation factor. In reactions governed by kinetics, the light isotopes are more concentrated in the products. Thus the deviation of the experimental data from the theoretical prediction suggests some influence of kinetically controlled fractionation. The curve of 0.995 is the best fit of data from Kitagawa et al. [10] on their sealed tube H 2 /Fe system. Other data points are from Vogel [1] and Verkoueren et al. ( [17], sealed tube H 2 /Fe + Mn system). The solid line calculated using a graphite-CO 2 of 0.972 is the best fit for data of Verkoueren et al. [17]. The much smaller a graphite-CO 2 is due to the fact that some of those highly fractionated reactions were stopped early to intentionally obtain the incomplete graphitization fractionation. Also, the actual graphitization process consists of complicated multi-step reactions, which may not be represented by the simple Rayleigh model. Isotopic fractionation can become much higher for small samples when graphitization cannot be completed [4,17,20,24,25]. The kinetic effect is much more significant for these reactions.
The amount of Zn is a major control of the quality of the graphite produced by our sealed-tube method. The fractionation of C isotopes increases with amounts of Zn smaller than 10 mg (Fig. 3(b)), while too much Zn (>50 mg/mg C) may increase the fractionation with no clear explanation (Fig. 3(b)). Fig. 3(b) also clearly demonstrates that excess TiH 2 will increase fractionation, which supports the theory of formation of CH 4 when excess H 2 is present. The least fractionation was observed when 5 mg of TiH 2 was used. For the H 2 reduction method, a H 2 /C ratio of 2:1-3:1 is often used [26]. If we adopt the same approach, then the amount of TiH 2 should be between 8.3 and 12.5 mg. Fig. 4 (using Fe as catalyst) shows that fractionation is relatively constant over the range of natural d 13 C level under our regular graphitization conditions, which gives us confidence that this fractionation can be corrected using simultaneously measured AMS d 13 C as long as all the samples and standards are prepared in the same way.

Graphite performance in ion source
The graphite targets prepared using this method generally perform well in our AMS ion source. The graphite made with Co as catalyst normally yields $40 lA 12 C + current at the high energy Faraday cup in our AMS system. Switching catalyst from Co to Fe has significantly improved the current to $50-60 lA, compatible to that of H 2 /Fe graphite run in the same system. Graphite made from small samples of 0.1-0.2 mg C also produces stable long lasting current, but usually only about 70-80% of the output from a large sample. The ''burn-in time'' (time for current to come up to a stable level) is usually longer for Zn reduced graphite than H 2 reduced graphite. Using Fe as catalyst has shortened this burn-in time. Currents from Zn/Fe graphite come up about 50% faster than Zn/ Co graphite. An advantage in using the low energy AMS at UCI is the lack of problems from di-lithium (7Li 2 ), which was sometimes a problem when our Zn reduced targets were measured at CAMS-LLNL; these problems were usually traced to impurities in specific batches of the small inner Pyrex tubes.

Precision and accuracy
Our primary standard for 14 C measurement is OXI. We normally use seven OXI standards in a 40-position wheel. All unknown samples and secondary standards are normalized to OXI. We also use at least three secondary standards in each wheel for quality control. Usually we use two OXII, one ANU, one TIRI wood B or one FIRI J (Barley) as secondary standard in one wheel. Simultaneously measured AMS d 13 C is used for isotope fractionation correction in 14 C result calculation. The use of AMS d 13 C corrects isotope fractionation that occurred in natural processes, as well as the fractionation that happens inside the AMS system and during graphite formation. This has significantly improved the precision of our Zn reduced graphite. Even though the analysis of CO 2 d 13 C by stable isotope mass spectrometry is a more precise measurement it will not take into account any fractionation that occurred during graphitization and AMS measurement. Also, it should be noted that many AMS systems lack the capability to measure 12 C and 13 C on the high-energy side and some systems do not make as precise measurement even if they have the capability. For those systems, this method will give slightly less precise results ($5-6&) if only IRMS measured d 13 C is used for fractionation correction. Fig. 5 is an example of measurements of our secondary standard OXII in the last three years. Table 2 is the summary of all measured secondary standards compared to the consensus values. Our measured values overlap the consensus values within the analytical error. Our accuracy is good and our precision has been improving over time, we now routinely achieve a precision of 2-3&.

Background
Background for 14 C measurement can come from contamination of sample, combustion, graphitization and AMS machine background. As shown in Fig. 6, the background of our measurements has improved over the three years. We now routinely obtain a satisfactory background of about $50,000 14 C years for 1 mg C sample, without pre-distillation of Zn or pretreatment of other reagents. We have found that pre-distillation of Zn can further improve background (Shuhui Zheng, unpublished data), as well as reducing the amount of Zn used. Machine background improvement [12], use of an ABA (acid-base-acid) pre-cleaned USGS coal (a bituminous coal from the Pocahontas #3 coal bed, collected in Buchanan County, Virginia in 1986), pre-combustion of sieved CuO and switching to Fe as catalyst all helped to reduce the background. We have also found that maintaining low backgrounds depends on the 'freshness' of the Zn powder. We put small aliquots of the Zn and TiH 2 powders into small vials for day-to-day use to decrease exposure to air.
Backgrounds for 0.5 mg C samples are about 43,000 14 C years and one measurement of 0.25 mg coal gave a background of 41,000 14 C years. Blank levels could become problematic for very small (<0.1 mg) and old samples. We plan to further investigate if pre-distillation of Zn powder will improve the background for small samples and thus expand the application of this method to small samples.

An example of high precision 14 C measurements
We present here the time series of atmospheric 14 CO 2 at a coastal site at Point Barrow, Alaska (71°N, 157°W) from July, 2003 to August, 2005 (Fig. 7). Air samples were collected into evacuated 6 L canisters and pressurized to about 2 atm. Upon returning to the lab, CO 2 was extracted cryogenically on a vacuum line and graphitized using the method described above and then analyzed for 14 C at our at AMS facility at UC Irvine. The high precision measurements ($2&) allow us to clearly see seasonal cycles in the distribution of 14 C with time: a broad minimum around February and a maximum in September with an amplitude of 7&. In this period, D 14 C decreased by 5.5&/year, to $57& in mid-2005. The peaks may be due to the joint effects of stronger stratosphere injection in April and May, and higher soil respiration with enriched 14 CO 2 between May and August; and the valleys may be due to less soil respiration and more fossil fuel burning in the winter months. This seasonality was also observed at Jungfraujoch by Levin and Kromer [27]. Simulations with the model of atmospheric transport and chemistry (MATCH) will be presented and discussed in another paper elsewhere [28]. This work has proved that 14 C is a unique tracer for discriminating sources between fossil and biosphere carbon emissions, and the sealed tube Zn reduction method for graphite preparation is suitable for high precision 14 C measurements for carbon cycle studies.

Conclusions
The modified sealed tube Zn reduction method has been shown to produce high quality graphite for high precision (2-3&) and high accuracy 14 C measurements by AMS, with a relatively low background of $50,000 14 C years on 1 mg C. The precision is as good as that achieved with the more widely used H 2 reduction method for 1 mg C samples if reagents are measured carefully and most importantly, if simultaneously measured AMS d 13 C is used for isotopic fractionation correction.
The major advantage of the Zn reduction method is that the system is relatively inexpensive and easy to set up (Table 3). A single technician can process 40 graphite samples from purified CO 2 or 20 from combusted samples in an 8 h day. However, the method may not work well with very small (<0.1 mg) and old samples. Distilling zinc should improve our background level and we will investigate this possibility so that we may expand the application of this method to smaller samples in the future. We believe that many disciplines can benefit from this technique because of its low cost and rapid production of graphite.  Table 3 Advantages and disadvantages of zinc sealed tube reduction method

Advantages
• Low cost -just a simple vacuum line and an oven • Rapid -a large number of samples can be prepared in the same day • Convenient -graphite can be saved for a long time in sealed tubes • High precision (2-3&) -if AMS d 13 C is used for fractionation correction • Acceptable background -50,000 years without pre-cleaning any metal powders

Disadvantages
• No way to monitor the reaction • Reaction can be strongly fractionating -precision will not be as good (5-6&) if simultaneously measured AMS d 13 C is not available • Background still higher than H 2 reduction method -not appropriate for very small or very old samples