Quantification of extraneous carbon during compound specific radiocarbon analysis of black carbon.

Radiocarbon ( 14 C) is a radioactive isotope that is useful for determining the age and cycling of carbon-based materials in the Earth system. Compound speciﬁc radiocarbon analysis (CSRA) provides pow-erful insight into the turnover of individual components that make up the carbon cycle. Extraneous or nonspeciﬁc background carbon (C ex ) is added during sample processing and subsequent isolation of CSRA samples. Here, we evaluate the quantity and radiocarbon signature of C ex added from two sources: preparative capillary gas chromatography (PCGC, C PCGC ) and chemical preparation of CSRA of black carbon samples (C chemistry ). We evaluated the blank directly using process blanks and indirectly by quantifying the difference in the isotopic composition between processed and unprocessed samples for a range of sample sizes. The direct and indirect assessment of C chemistry + PCGC agree, both in magnitude and radiocarbon value (1.1 ( 0.5 µ g of C, fraction modern ) 0.2). Half of the C ex is introduced before PCGC isolation, likely from coeluting compounds in solvents used in the extraction method. The magnitude of propagated uncertainties of CSRA samples are a function of sample size and collection duration. Small samples collected for a brief amount of time have a smaller propagated 14 C uncertainty than larger samples collected for a longer period of time. CSRA users are cautioned to consider the magnitude of uncertainty they require for their system of interest, to frequently evaluate the magnitude of C ex added during sampling processing, and to avoid isolating samples e 5 µ g of carbon.

Radiocarbon ( 14 C) is a radioactive isotope that is useful for determining the age and cycling of carbonbased materials in the Earth system. Compound specific radiocarbon analysis (CSRA) provides powerful insight into the turnover of individual components that make up the carbon cycle. Extraneous or nonspecific background carbon (C ex ) is added during sample processing and subsequent isolation of CSRA samples. Here, we evaluate the quantity and radiocarbon signature of C ex added from two sources: preparative capillary gas chromatography (PCGC, C PCGC ) and chemical preparation of CSRA of black carbon samples (C chemistry ). We evaluated the blank directly using process blanks and indirectly by quantifying the difference in the isotopic composition between processed and unprocessed samples for a range of sample sizes. The direct and indirect assessment of C chemistry+PCGC agree, both in magnitude and radiocarbon value (1.1 ( 0.5 µg of C, fraction modern ) 0.2). Half of the C ex is introduced before PCGC isolation, likely from coeluting compounds in solvents used in the extraction method. The magnitude of propagated uncertainties of CSRA samples are a function of sample size and collection duration. Small samples collected for a brief amount of time have a smaller propagated 14 C uncertainty than larger samples collected for a longer period of time. CSRA users are cautioned to consider the magnitude of uncertainty they require for their system of interest, to frequently evaluate the magnitude of C ex added during sampling processing, and to avoid isolating samples e5 µg of carbon.
Radiocarbon dating of bulk organic and inorganic carbon reservoirs has allowed the average residence time of carbon in most carbon pools to be calculated. However, many of these reservoirs are comprised of complex, heterogeneous mixtures whose components have different residence times from bulk radiocarbon values. Initially, the heterogeneous mixtures were studied via compound class radiocarbon analysis (CCRA). 1 Introduction of compound specific radiocarbon analysis (CSRA) allowed the 14 C measurement of a single compound. 2 CSRA usually involves a multiple-step purification procedure that culminates in the collection of a single compound (or group of compounds) of high purity. The applications of CCRA and CSRA range from source apportionment of atmospheric particles, 3,4 biomarkers with paleoclimatic implications, 5-7 microbial incorporation of fossil material, 8,9 and compound class studies in marine sediments 10 and marine dissolved organic carbon. 11,12 New developments in accelerator mass spectrometry (AMS) have decreased the sample size requirements for CSRA. Ultrasmall samples 13 and online 14 C measurements 14 enable CSRA as small as 2 µg of C. Preparation of CSRA samples requires two sets of laboratory protocols, sample isolation, and 14 C analysis, each of which introduce extraneous or nonspecific background carbon (C ex ). Thus a CSRA sample of 2 µg of C may have a large uncertainty associated with its isotopic composition. To date, few studies have quantified C ex . 15 Accounting for C ex has largely been avoided by processing samples large enough so as to overwhelm the C ex . However, not all environmental CSRA techniques allow for the preparation of large sample sizes because the compound of interest may be in low abundance.
Constraining the uncertainty of 14 C measurements is done by evaluating the mass and variability of C ex added during sample preparation. Here we assess the mass and radiocarbon signatures of C ex specific to the chemical oxidation of organic matter for quantifying black carbon using PCGC. We employed the benzene polycarboxylic acid method that * To whom correspondence should be addressed. E-mail: lorized@ gmail.com.

METHODS
Natural and synthetic vanillin (4-hydroxy-3-methoxybenzaldehyde, Table 1) were used as standards to assess the extraneous carbon added during PCGC isolation. Black carbon (BC) reference materials were used as process standards to quantify C ex added throughout the entire isolation procedure (Table 1). 16,17 Chemical Oxidation. To minimize carbon contamination, all glassware and quartz filters that came in contact with the samples and standards were baked at 550°C for 2 h prior to use. Samples were processed using a modification of the benzene polycarboxylic acid (BPCA) method. 18,19 Process materials, wood char, and hexane soot (Table 1) were oxidized in 2 mL of concentrated nitric acid (grade ACS, Fisher Scientific) in quartz tubes inside a high-pressure digestion apparatus at 180°C for 8 h. Postdigestion, the samples were filtered through quartz fiber filters (27 mm diameter, 0.8 µm pore diameter), and 15 mL of Milli-Q water was used to rinse any remaining BPCAs from the filter. The filtrate was collected and freeze-dried overnight.
Dried samples were redissolved in 5 mL of methanol, and the internal standard, biphenyl-2,2′-dicarboxylic acid (1 mg mL -1 in methanol), was added. Samples were derivatized by titration with 2.0 M trimethylsilyl diazomethane in ethyl ether (Sigma Aldrich). Derivatization was considered complete when the solution retained the yellow color of the trimethylsil-diazomethane. Methanol was dried with in stream of ultrahighpurity nitrogen. A fixed volume of dichloromethane was added.
The derivatized oxidation products were separated and quantified on a Hewlett-Packard 6890N outfitted with a Gerstel cooled injection system, a DB-XLB capillary column (30 m × 0.53 mm i.d., 1.5 µm film thickness), and a flame ionization detector (FID), and a Gerstel preparative fraction collector (PFC). After injection, the column temperature was maintained at 100°C for 1 min, then raised at 25°C min -1 to 250°C followed by a 5°C min -1 ramp to 280°C for 10 min, and then raised to 320°C for 5 min of bake out (Figure 1). The FID temperature was 300°C. The splitless injection volume was 1 µL for all samples in this study.   Approximately 1% of the flow eluting from the capillary column was diverted to the FID, and 99% was sent to the PFC, which consists of a zero-dead-volume valve in a heated interface (320°C ) and seven 200 µL glass U-tube traps (six sample traps and a waste trap). The PFC transfer was kept constant at 320°C for all samples processed. U-tubes were supported in isopropyl alcohol cooled units (-10°C). The autoinjector, CIS, and trapping device are programmable and computer controlled, and FID data was acquired using Chemstation software.
BPCAs were identified by comparison of their retention times with those obtained for a commercially available mixture, and they were also verified using gas chromatography/mass spectrometry (GC/MS). All methylated BPCAs were quantified relative to the biphenyl-2,2′-dicarboxylic acid internal standard.
Radiocarbon Analysis of Isolated Samples. To avoid cross contamination from previously injected samples (e.g., memory), the U-tubes (and the material collected within) for the first 10 injections were replaced with clean, baked U-tubes. Unless otherwise noted, trapped samples were collected from 50 injections of each sample. To avoid possible isotope fractionation of isolates, 20 care was taken to trap the entire peak.
After PCGC isolation, the U-tubes containing trapped samples were rinsed with 700 µL of dichloromethane into prebaked GC autosampler vials. Samples were evaluated by GC-FID for purity and yield. Samples were then transferred to 6 mm quartz tubes using an additional 700 µL of dichloromethane, and the solvent was removed in a stream of UHP nitrogen. CuO and silver wire were added, and the sample tube was evacuated to 10 -6 Torr and flame-sealed under vacuum. Tubes were then heated to 850°C for 2 h. The resulting CO 2 was purified, quantified, and reduced to graphite according to standard procedures. 21 Measurements of 14 C were made at the Keck Carbon Cycle Accelerator Mass Spectrometry Laboratory at University of California Irvine. In all cases, radiocarbon analysis are reported as fraction modern, which is the deviation of a sample from 95% of the activity in 1950 A.D., of National Bureau of Standards (NBS) oxalic acid 1 normalized to δ 13 C ) -25 with respect to Pee Dee Belemnite. 22,23 All fraction modern values reported within this article have been corrected for combustion and graphitization 13,21 and mass dependent isotope fractionation by reporting all data to a common δ 13 C value of -25. 23

RESULTS
Carbon Mass Balance and Corrections. Once corrected for graphitization and combustion, the mass of carbon graphitized in CSRA samples (C reported ) originate from four sources: the mass of carbon in the compound of interest isolated from the sample (C sample ), the mass of derivative carbon (C derivative ), the mass of extraneous carbon added during chemical extraction (C chemistry ), and subsequent isolation via PGCG (C PCGC ). The compounds of interest in this study, BPCAs, contain functional groups that require derivatization to adjust their polarity and volatility to enable separation by PCGC. The derivatization adds a methyl group (-CH 3 ) to each carboxylic acid group, and this additional carbon alters the 14 C signature of the sample. Since the isotopic composition of the derivative carbon is assumed to be 14 C-free (FM derivative ) 0), the reported isotopic signature is known, and the amount of added derivative carbon is known, the radiocarbon composition of the parent BPCA compound can be calculated via mass balance.
When samples are corrected for C derivative , eq 1 is simplified to To provide accurate isotopic values of C sample+chemistry+PCGC , the mass and isotopic composition (FM) of C ex must be determined. Here we evaluated two sources of C ex : added during chemical extraction (C chemistry ) and during PCGC isolation (C PCGC ). Reported values of CSRA samples (C reported ) need to be corrected for C ex . For the purposes of estimating the C ex via process materials, samples had to be corrected for derivative carbon before estimating C ex , which assumes that all C ex has been derivatized. Extraneous Carbon Added during PCGC Isolation (C PCGC ). Two methods were used to evaluate the mass and FM of C ex from PCGC isolation. First, the direct approach was used to collect a process blank over a 7 min retention time window (from 18 to 25 min, Figure 1) from 400 dry injections (direct C PCGC ). No solvent was injected during the dry injections, that is, there was no needle in the autosampler, and all other GC parameters (i.e., carrier gas, oven temperature) were maintained. This sample yielded 7.6 ± 0.4 µg of C and had a FM PCGC of 0.125 ± 0.034. Because the sample collection window varied with sample type (Figure 1), we normalized the amount of C ex (µg of C) to collection duration (in minutes) and 50 injections. Normalizing the C ex to time assumes the majority of C ex is due to column bleed (sample history and/or breakdown of the GCcolumn stationary phase) and that the bleed does not change over time or temperature. To standardize this nonspecific background correction, all subsequent collections maintained the same injection volume and number of injections; only the collection time and injected materials varied for the samples reported here. We normalized all samples that evaluated C ex , even samples that included the C chemistry . Thus evaluated directly, the C PCGC added in the dry injections was 0.1 ± 0.05 µg of C min -1 per 50 injections.
The second method of evaluating the mass and FM of C PCGC used various sizes of isolated process standards of known FM values. It was assumed that the sample was diluted with a constant mass and isotopic signature of C ex , and the presence of C ex would cause a deviation in the consensus 14 C value. The FM values of samples were expressed by the following equation: where FM sample is the radiocarbon value of the sample corrected for C PCGC , FM reported is the measured radiocarbon value of the sample uncorrected for C PCGC , and FM PCGC is the radiocarbon value of the extraneous carbon added during PCGC isolation. Because C ex was assessed as a combination of both dead and modern material, FM ex fell between 0.0 and 1.0. Therefore, small samples of modern isotopic composition were lower and samples of 14 C-depleted composition were higher in radiocarbon (e.g., Figure 2).
With the use of the above approach, the PCGC isolation sizeseries of modern vanillin (FM sample ) 1.052, Table 1) samples revealed that C PCGC (FM PCGC ) 0.0) was 0.4 ± 0.2 µg of C min -1 per 50 injections. The PCGC isolation of a series of different sized samples of 14 C-free vanillin (FM sample ) 0.002) revealed an additional 0.2 ± 0.1 µg of C min -1 per 50 injections was added with an assumed FM PCGC ) 1.0. Combined, these two blanks revealed the total indirect C PCGC of 0.6 ± 0.3 µg of C with an average FM PCGC ) 0.3 ( Table 2).
The difference of 0.5 µg of C min -1 per 50 injections of C ex added to isolated vanillin samples calculated using standard materials (0.6 ± 0.3 µg of C min -1 per 50 injections) as compared to the dry injections (0.1 µg of C min -1 per 50 injections) may be due to several factors. First, no solvent was injected into the GC column during dry injections. It is likely that when solvent is present in the GC column, more C ex is mobilized than during the absence of solvent. The FM ex values for vanillin (FM ex ) 0.3 ± 0.1) and that for the dry injections (FM ex ) 0.125 ± 0.034) were similar suggesting the same source of C ex . Other possible explanations are that C PCGC and its isotopic signature vary with time and the presence of sample memory and/or contamination of the injector port. Therefore, we estimate that for each minute of collection on the PCGC, at least 0.6 µg of C with a FM ) 0.3 was being added to samples from PCGC. Possible sources of C PCGC are column bleed, break down of non-GC column tubing, post-PGCG sample handling, and residual solvent in the programmable fraction collector trap.
Extraneous Carbon Added during Chemical Oxidation and PCGC Isolation (C chemistry+PCGC ). CSRA samples are typically subjected to extensive chemical extraction procedures prior to isolation by PCGC and consequently it is likely that extraneous carbon is added during these procedures. In theory, isolation of compounds by PCGC should removed any extraneous carbon added during sample preparation. Similar to the evaluation of C ex added during PCGC isolation, we evaluated the mass and FM of C ex added during the chemical methods and PCGC isolation (C chemistry+PCGC ) using both an indirect and direct approach. To evaluate C ex directly, the chemical oxidation, derivatization, and PCGC isolation steps were carried out with no sample added. In this way, direct analysis of the C ex was 1.1 ± 0.2 µg of C min -1 per 50 injections and FM ) 0.200 ± 0.054 ( Table 2).
The C ex was evaluated indirectly by quantifying the deviation in FM sample+ex from the unprocessed material for radiocarbon dead (hexane soot) and modern (grass char) of different sizes. Samples of modern grass char (2-16 µg of C) were chemically oxidized, derivatized, and isolated by PCGC. We found that 0.80 ± 0.40 µg of C min -1 per 50 injections of an assumed FM ex ) 0.0 was added in chemical oxidation and PCGC isolation. Fossil hexane soot revealed 0.15 ± 0.08 µg of C min -1 per 50 injections of an assumed FM ex ) 1.0 was added in sample processing. The total indirect method C ex was then calculated to be 1.0 ± 0.5 µg of C min -1 per 50 injections and with FM ex ) 0.15.
When evaluated directly and indirectly the mass and isotopic composition of the time normalized C ex added during sample processing and isolation was the same even though the collection window varied from 0.4 to 7 min. This suggests that C chemistry can be scaled with time. If the C ex for indirect assessment was much larger than the direct method, the source of the C ex may be a matrix effect of the oxidation process. The agreement of the two methods suggests that the C ex is not associated with matrix effects during the processing of a sample.
The magnitude of the C ex added during chemical oxidation (C chemistry ) 0.5 µg of C min -1 per 50 injections) is approximately equal to that added during PCGC isolation (C PCGC ) 0.6 µg of C min -1 per 50 injections). This was determined by the difference between C chemistry+PCGC and C PCGC . Since all samples were treated to the same post-PCGC handling, our data suggests that only half of the nonspecific background (C ex ) is originating from PCGC isolation (e.g., column bleed) and post-PCGC sample handling. The dry injection blank was very small (0.1 µg of C) suggesting that post PCGC handling is likely a small part of C PCGC . The remainder, C chemistry , is likely from coeluting compounds in the reagents and solvents used in the oxidation and derivatization processes. Reagents and solvents can become contaminated over time and with use. Therefore, if the source of this additional carbon is coeluting compounds, it is essential to frequently evaluate the C ex (e.g., every 2 to 5 samples) to ensure reagent and solvent purity.

10159
Analytical Chemistry, Vol. 81, No. 24, December 15, 2009 Correcting for Extraneous Carbon and Associated Uncertainties. Radiocarbon measurements are typically reported with an uncertainty of the AMS measurement alone. As we have shown above, the corrected radiocarbon value of a CSRA sample is dependent on the mass and FM of the C ex . In our work, if the sample was g50 µg of C, the C ex was insignificant. However most of samples were small so the FM of small CSRA samples required a correction for the presence of C ex . The uncertainties of all terms needed to be considered when reporting the uncertainty of the CSRA FM value. To determine the propagated total mathematical uncertainty of FM sample (e.g., eq 3), we applied the following equation: where σ FM reported is the AMS uncertainty of FM reported (machine uncertainty), σ FM ex is the uncertainty for FM ex , σ m reported is the uncertainty for C reported (uncertainty in graphitization), and σ m ex is the uncertainty for C ex . The total uncertainty of the direct process blank (C chemistry+PCGC in Table 2) was used for FM ex and C ex .
To illustrate how correcting CSRA samples for C ex affects the isotopic values and associated uncertainties, 10 we corrected modern grass char and fossil hexane soot samples using the direct process blank determined in Table 2. For grass char, a modern BC standard, 16 the measured FM reported values for 7 small samples without C ex correction (average FM reported ) 0.824 ± 0.128, Table 3) were all significantly lower than the FM value of the unprocessed material (FM ) 1.056 ± 0.002, Figure 2). After correction for C chemistry+PCGC , the FM sample (average 1.098 ± 0.221) agreed with that of the unprocessed material. For hexane soot, a dead BC standard, the measured FM values without correc- 1.0 ± 0.5 0.15 ± 0.08 a Materials were subjected to varying treatments to determine the mass and source of C ex . The uncertainty of the mass of extraneous carbon was estimated at 50% of the sample mass. The uncertainty of FM ex determined by the indirect approach was estimated at 50% of the FM value. All estimates of C ex are scaled to the width of the collection window per 50 injections to facilitate comparison between direct and indirect estimations of C ex . b BPCA includes the chemical oxidation of BC into BPCAs and their subsequent derivatization, see text for details. c All samples processed through the PCGC were subjected to the same post-PCGC handling procedures. d The mass of C ex is normalized to per minute per 50 injections.  ) is assumed to be 1.1 ± 0.2 µg of C per minute of collection for a 50 injection run with a FM ) 0.2 ± 0.054 (see Table 2). The uncertainty associated with the FM reported is the AMS machine uncertainty, and the uncertainty associated with FM sample is the propagated uncertainty. The collection window duration was varied to collect individual BPCAs or ΣBPCAs. b After diazomethane correction. c Determined using eq 3. tion for C ex (average FM reported ) 0.061 ± 0.55, Table 3) were significantly higher than the unprocessed material (FM ) 0.005 ± 0.001). After correction for C chemistry+PCGC , the FM reported (average 0.036 ± 0.056) was within error of the FM of the unprocessed material. The deviation of low mass corrected hexane soot samples from the consensus value indicated that the mass or FM signature of C chemistry+PCGC was different for samples smaller than 5 µg of C. Because this phenomenon was not observed for low mass grass char samples, it appeared that the FM ex value of ultrasmall samples (5 µg of C) contained additional modern carbon. To avoid this complication, we avoided processing samples smaller than 5 µg of C.
These results demonstrate that the uncertainties of FM sample associated with the preparation and isolation of samples by CSRA are significantly larger than the machine error. Propagated total uncertainty of modern process materials is much higher than 14 C depleted process materials due to (1) the logarithmic nature of radioactive decay and (2) the FM ex was more 14 C depleted than modern. Each system will be distinct, therefore each user needs to evaluate the C ex and FM ex values specific for their system.
Thus, when considering CSRA applications, one must consider the magnitude of uncertainty required to provide useful information about the system being studied. For example, our interest in CSRA of BPCAs is to examine the BC in marine dissolved organic carbon (DOC). 19 Bulk DOC, which is comprised of a wide range of organic molecules of varying 14 C ages, typically ranges from FM ) 0.8 to 0.5. 12 The BC in marine DOC has been postulated to be more depleted in radiocarbon. Provided that BC extracted from marine DOC has a propagated total uncertainty for FM less than 0.10, the results should provide valuable information about this pool of recalcitrant carbon. However, if one was interested in studying the removal of BC from soils over a few centuries, much larger samples than those presented here are required in order to ensure that the contribution of C ex to the FM reported is insignificant. Regardless of the application, it is equally important that CSRA users assess their ability to duplicate CSRA measurements. In some cases, the variation of duplicate analyses of CSRA samples may be larger than the propagated uncertainty, which may not be precise enough for certain applications. The mass and isotopic composition of C ex should ideally be evaluated with each batch of samples, as we found that the mass of C ex varied by over 50% over the course of 6 months. 19

CONCLUSIONS
Half of the C ex was added during PCGC isolation and half was added during the chemical oxidation and derivatization. Extraneous carbon added during PCGC isolation of CSRA samples was found to be a function of collection duration on the GC. The estimates of extraneous or nonspecific background carbon presented here are specific to the BPCA chemical isolation technique. After background correction, CSRA samples 5 µg of C were unreliable with respect to accuracy and precision. Another facility using the same chemical extraction technique would need to determine the extraneous carbon (C ex ) introduced to samples that they process. Different GC columns, solvents, and users will likely produce more or less C ex carbon, with unique FM ex signatures.