Anal. Chem. 2007, 79, 3469-3475
An Approach for Assessing Total Instrumental Uncertainty in Compound-Specific Carbon Isotope Analysis: Implications for Environmental Remediation Studies Barbara Sherwood Lollar,* Sarah K. Hirschorn, Michelle M. G. Chartrand, and Georges Lacrampe-Couloume
Stable Isotope Laboratory, Department of Geology, University of Toronto, 22 Russell Street, Toronto, Ontario, Canada M5S 3B1
Determination of compound-specific carbon isotope values by continuous flow isotope ratio mass spectrometry is impacted by variation in several routine operating parameters of which one of the most important is signal size, or linearity. Experiments were carried out to evaluate the implications of these operating parameters on both reproducibility and accuracy of δ13C measurements. A new method is described for assessing total instrumental uncertainty of routine compound-specific δ13C analysis, incorporating both accuracy and reproducibility. These findings have important implications for application of compound-specific isotope analysis in environmental geochemistry and in particular for the rapidly developing field of isotopic investigation of biodegradation and remediation of organic chemicals in contaminant hydrogeology. Conventional dual inlet gas source mass spectrometry using offline sample preparation techniques involves optimization of analytical parameters to maximize precision (hereafter referred to as reproducibility). By alternating measurements between an isotopic standard in one inlet and the sample in the second inlet multiple (typically 12-20) times during a single data acquisition run, both accuracy and reproducibility are achieved. An additional important aspect of optimizing reproducibility is that sample and standard are balanced (set to the same signal size) to minimize linearity effects. Typical reproducibility for CO2 samples run by this approach is in the second decimal place as long as the signal size is maintained within the linear range for the instruments typically ∼0.2 to 7 V for the instrument used in this study. For signal sizes above and below the linear range, both accuracy and reproducibility can be compromised. In contrast, continuous flow gas source mass spectrometry systems introduce both sample and standard gases into the mass spectrometer source in a helium gas stream.1,2 Continuous flow * Corresponding author. Phone: 416-978-0770. Fax: 416-978-3938. E-mail:
[email protected]. (1) Matthews, D. E.; Hayes, J. M. Anal. Chem. 1978, 50, 1465-1473. (2) Hayes, J. M.; Freeman, K. H.; Popp, B. N.; Hohan, C. H. Organic Geochemistry In Advances in Organic Geochemistry; Durand, B., Behar, F., Eds.; Pergamon Press: Oxford. U.K., 1990; Vol. 16, pp 1115-1128. 10.1021/ac062299v CCC: $37.00 Published on Web 03/29/2007
© 2007 American Chemical Society
isotope ratio mass spectrometry (CF-IRMS) provides a number of advantages vis a vis conventional dual inlet mass spectrometry, including ease and efficiency of online sample preparation and transfer, the ability to measure all compounds in a complex sample mixture in a single data acquisition run, and significantly reduced sample size requirements (as low as nanomoles of CO2). Offsetting these advantages is a loss of reproducibility due to a variety of factors including a higher source pressure, higher backgrounds, and residual water vapor due to the online sample combustion and transfer system, the effects of chromatographic separation, and the reduction in the number of alternating standard and sample measurements in a given data acquisition run, etc. A detailed discussion of these factors is beyond the scope of this paper but can be found in Freedman et al.,3 Hayes et al.,2 Merritt et al.,4 and Ricci et al.5 The end result is that a CO2 sample that can be characterized with a 2σ reproducibility in the second decimal place by conventional dual inlet mass spectrometry, can be characterized by routine CF-IRMS methods only to the first decimal placesan order of magnitude loss in reproducibility that is generally considered a small price to pay for the quantum advantages of CF-IRMS.4,5 POTENTIAL FOR OPTIMIZATION OF REPRODUCIBILITY WITHIN CF-IRMS While many factors such as the higher source pressure and backgrounds inherent to continuous flow approaches cannot be eliminated, even within CF-IRMS certain steps can be taken to optimize reproducibility by customizing critical components of the hardware4 and/or software and data processing.5 Even within routine systems and software, certain steps can be taken to improve reproducibility. Figure 1a shows a chromatogram for a continuous flow gas chromatograph-isotope ratio mass spectrometry (GC-IRMS) δ13C analysis of a sample containing vinyl chloride (VC), 1,2-dichloroethane (1,2-DCA), trichloroethene (TCE), and perchloroethene (PCE). This chromatogram incor(3) Freedman, P. A.; Gillyon, E. C. P.; Jumeau, E. J. Am. Lab. 1988, 20, 114119. (4) Merritt, D. A.; Brand, W. A.; Hayes, J. M. Org. Geochem. 1994, 21, 573583. (5) Ricci, M. P.; Merritt, D. A.; Freeman, K. H.; Hayes, J. M. Org. Geochem. 1994, 21, 561-571.
Analytical Chemistry, Vol. 79, No. 9, May 1, 2007 3469
Figure 1. (a) Typical chromatogram for continuous flow gas chromatograph isotope ratio mass spectrometry (GC-IRMS) δ13C analysis of a ground water sample containing compounds VC, 1,2-DCA, TCE, and PCE. As described in the methods, after GC separation, each compound in the mixture is separately combusted to CO2, H2O vapor is removed, and the resulting CO2 derived from each compound sequentially enters the mass spectrometer for δ13C analysis. While the peaks in a are labeled with the individual compound names, the analyte in all cases is thus CO2. Intensity (in volts) is the signal on the major Faraday cup that measures mass 44 (12C16O16O). Acquisition runs typically start with injection of two peaks of the same CO2 isotopic reference standard gas of which the second (CO2*) is typically designated as the standard against which all other measurements are calibrated (see text). Panel b demonstrates how linearity effects could be minimized by conducting multiple acquisition runs where the peak (signal) size of the CO2 standard is modified to match, or balance, the smallest signal size (for VC in this example). This could be referred to as a pseudo-dual inlet (balanced) mode compared to the normal continuous flow CSIA approach of analyzing all sample peaks within the instrument’s linear range in the same acquisition run all calibrated to a CO2 peak set to a mid- to low-range (1-2 V), as in a. Panel c demonstrates the same pseudo-dual inlet (balanced) approach where the peak (signal) size of the CO2 standard is modified to match, or balance, the largest signal size (for 1,2-DCA in this example).
porates many features typical of a CF-IRMS analysis. Typically the run begins with at least two injections of the CO2 reference gas standard, of which one (CO2*; Figure 1a) is designated as the standard in the instrument acquisition software and hence designated as the measurement against which all other peaks are to be characterized via the following equation:
δ13C (‰) )
[
]
(13C/12C)sample - (13C/12C)standard (13C/12C)standard
× 1000 (1)
Additional CO2 reference gas standard peaks can be added throughout the run as long as they do not interfere with a sample 3470
Analytical Chemistry, Vol. 79, No. 9, May 1, 2007
compound peak eluting at the same time, but typically any additional CO2 standard peaks are used as blind standards (as an additional cross-check). One of the principal differences between CF-IRMS and conventional dual inlet mass spectrometry is that in CF mode the standard value from eq 1 is based on only one measurement of the standard per run, rather than on multiple iterations between standard and sample in the same run as carried out in dual inlet mass spectrometry. The other critical difference illustrated in Figure 1a is that while dual inlet mass spectrometry is based on running standard and sample in balance (at the same signal size), in CF mode, while the standard is typically set at a mid- to low-range signal size (∼1-2 V), sample compound peaks will vary significantly ranging from well under 1 V all the way up
to the linear range of the instrument (typically ∼7 V for the mass spectrometer used to produce this chromatogram). Under the terminology used for dual inlet mass spectrometry, this approach of running standard and sample at different signal sizes is termed a misbalancesa pejorative term reflecting what was once seen as an undesirable deterioration of measurement quality. Specifically, maximum accuracy between standard and sample is best achieved by running in balanced mode. When standard and sample are run at different signal sizes, the linearity of the instrument results in small but significant absolute differences in the δ13C value measured. Hence running with a “misbalance” between standard and sample is inherently less accurate. Furthermore if the reproducibility of multiple sample compound measurements run at one signal size are compared to the reproducibility for multiple measurements run over a range of signal sizes, this linearity effect will also be reflected, in the latter case, in a larger standard deviation (reduced precision). In the era of CF-IRMS, this is no longer seen as an inherent problem as the measurement quality deterioration that results is a small disadvantage compared to the enormous advantage of isotopic characterization of multiple sample compounds of significantly different signal size all within the same acquisition run. The quality of the measurement can be improved by the simple approach illustrated in Figure 1b,cswhere misbalance is eliminated by characterizing the different sample compounds in different runs, and in each run fine-tuning the CO2 reference gas standard signal size to match each successive sample peak. The same objective can be achieved by introducing different amounts of the sample mixture in each successive run, so each sample compound is successively tailored to be in the same 1-2 V signal size range as the CO2 reference standard peak. While this pseudodual inlet (balanced) mode is occasionally used for specific applications and for constraining optimal performance using CFIRMS,4 it is not a typical approach since it is time-consuming and negates many of the advantages of CF-IRMS. More commonly, routine practice is to take full advantage of CF-IRMS by setting the CO2 reference standard peak to some acceptable mid- to lowrange value (typically 1-2 V) and analyzing all sample compounds within an acceptable range above and below that, as in Figure 1a. The first objective of this paper is to examine what constitutes an acceptable range and the implications of that for the total instrumental uncertainty (accuracy and reproducibility) associated with typical routine compound-specific carbon isotope analysis by CF-IRMS. A second objective is to demonstrate a new method for evaluating total instrumental uncertainty on the basis of these principles. The third objective is to demonstrate that these analytical issues are not only important for the field of compoundspecific isotope analysis in general but also are particularly critical for the developing field of research that uses compound-specific isotope analysis determined by CF-IRMS to identify and quantify biodegradation of organic contaminants in the environment. EXPERIMENTAL SECTION To develop laboratory isotopic working standards, pure-phase liquid samples of 1,2-DCA, PCE, and TCE were prepared offline and analyzed using dual inlet isotope ratio mass spectrometry. A 5 µL aliquot of each compound to be isotopically characterized was injected under vacuum (3 × 10-3 mbar) into a separate quartz ampule containing ∼1 g of copper, ∼2 g of cupric oxide, and a
3.3 × 5 mm piece of silver and flame sealed after Schimmelman DeNiro.6 The sealed ampules were combusted in a muffle furnace at 850 °C for 1 h. The temperature was decreased to 600 °C for 1 h and then decreased to room temperature. This preparation method allowed for the simultaneous production of CO2 and H2O and removal of any chlorine by formation of silver chloride. Under vacuum, the CO2 was cryogenically separated from the H2O, and the CO2 was transferred to an evacuated pyrex ampule, flamesealed, and stored in the freezer until analyzed. Stable carbon isotope analyses of the prepared samples were performed using a Finnigan MAT 252 dual inlet isotope ratio mass spectrometer. Carbon isotope ratios (δ13C) for each sample analyzed are expressed relative to the international IAEA standard V-PDB (Vienna-Peedee Belemnite) for carbon following eq 1. On a daily basis the CO2 reference gas standard is a laboratory working standard gas that has been cross-calibrated against V-PDB and is then used for daily routine standardization of sample acquisition runs per normal practice. For continuous flow compound-specific isotope analysis, the previously isotopically characterized standard materials were then used to create dissolved aqueous solutions at concentrations ranging from 1 to 500 mg/L. All dissolved aqueous samples were then analyzed by CF-IRMS using headspace analysis after the method of Slater et al.7 The GC/C/IRMS (gas chromatograph/ combustion/isotope ratio mass spectrometry) system consisted of a Varian 3300 GC coupled to a Finnigan MAT 252 mass spectrometer via a combustion interface. For 1,2-DCA, a 30 m × 0.32 mm DB 624 column was used, with an oven temperature program of 45 °C for 1 min, increasing at 15 °C/min to 90 °C and holding for 10 min. For PCE and TCE, the same column was used, with an oven temperature program of 50 °C initially, increasing at 15 °C/min to 180 °C, and holding for 2 min. For p-xylene in groundwater samples, the compound was extracted by pentane extraction after the method of Dempster et al.8 A Bentone 34 din-decylphthalate fused silica SCOT column was used. GC parameters were 60 °C for 4 min, increasing at 5 °C/min to 90 °C, and holding for 11 min, followed by increasing at 30 °C/min to 135 °C and holding for 20 min. RESULTS AND DISCUSSION Parts a and b of Figure 2 show the results of investigating the effect of signal size differences between the standard and sample (linearity effects) carried out by introducing the CO2 reference standard peak at 1-2 V and injecting previously characterized isotopic working standard materials over a range of signal sizes. For the 1,2-DCA laboratory isotopic working standard, the mean of the δ13C values determined by CF-IRMS in Figure 2a is -27.0 ( 0.4‰ (n ) 109). This value is in good agreement with the value of -27.2 ( 0.2‰ determined for the same 1,2-DCA working standard by offline sample preparation and conventional dual inlet mass spectrometrysa result typical of other studies that have compared the δ13C values of pure-phase liquids and aqueous-phase (dissolved in water) organic compounds.9 The data show two (6) Schimmelmann, A.; DeNiro, M. J. Anal. Chem. 1993, 65, 789-792. (7) Slater, G. F.; Dempster, H. S.; Sherwood Lollar, B.; Ahad, J. Environ. Sci. Technol. 1999, 33, 190-194. (8) Dempster, H. S.; Sherwood Lollar, B.; Feenstra, S. Environ. Sci. Technol. 1997, 31, 3193-3197. (9) Jochmann, M. A.; Blessing, M.; Haderlein, S. B.; Schmidt, T. C. Rapid Commun. Mass Spectrom. 2006, 20, 3639-3648.
Analytical Chemistry, Vol. 79, No. 9, May 1, 2007
3471
δ13C
Figure 2. (a) Results from experiments measuring values of isotopically characterized laboratory working standard 1,2-DCA by CF-IRMS. CO2 isotopic reference standard gas is set at 1-2 V, while the signal size for 1,2-DCA was varied in the range from 0.2 to 3.5 V. The mean value of all analyses is -27.0 ( 0.4‰ (indicated by the solid black line through the data) and by the hatched lines above and below. Variance in δ13C values for different signal size intervals is 2 V (0.036). (b) Results from experiments measuring δ13C values of isotopically characterized laboratory working standard PCE by CF-IRMS. CO2 isotopic reference standard gas is set at 1-2 V, while the signal size for PCE was varied in the range from 0.3 to 6 V. The mean value of all analyses is -30.0 ( 0.5‰ (indicated by the solid black line through the data) and by the hatched lines above and below. Variance in δ13C values for different signal size intervals is 2 V (0.035).
important aspects of the linearity effectsincreased variability at low signal size (
