Anal. Chem. 1998, 70, 2737-2744
Water-Induced Errors in Continuous-Flow Carbon Isotope Ratio Mass Spectrometry Kristen J. Leckrone† and John M. Hayes*
Biogeochemical Laboratories, Departments of Chemistry and of Geological Sciences, Indiana University, Bloomington, Indiana 47405-1403 Formation of HCO2+ from CO2 and background H2O in isotope ratio mass spectrometers has been examined in detail. The process is troublesome because its product is not resolved from 13C16O2+. The resulting, artifactual enhancement of the mass 45 ion current (and analogous enhancement of the mass 46 ion current by transfer of hydrogen to mass 45 species) can cause systematic errors in analyses of 13C based on measurement of ion current ratios in the mass spectrum of CO2. Such errors are neutralized when isotopic analyses are based on differential comparisons in which ion currents and background water levels are precisely equal during admission and ionization of both sample and standard gases. In continuous-flow systems, however, that requirement is generally not met. The resulting systematic error is proportional to the 18/44 ion current ratio. When the widely used MAT 252 mass spectrometer is tuned to yield maximum sensitivity, the constant of proportionality is 26 ( 2‰ (i.e., the error will be 0.26‰ if the mass 18 ion current is 100 times smaller than that at mass 44). Errors can be reduced 5-fold when the ion-source residence time of CO2+ is decreased by use of stronger ionextraction potential gradients. Under those same conditions, sensitivity is decreased by 60%. For operation at highest sensitivity, carrier gas dew points on the order of -70 °C are required to obtain errors e0.1‰ for samples yielding mass 44 ion currents of 10 nA. Carrier gas dew points e-80 °C are conveniently reached by use of a Nafion dryer operated at ≈0 °C. Variations in the abundances of the stable isotopes of hydrogen, carbon, nitrogen, oxygen, and sulfur in natural materials are customarily reported using a differential parameter based on a comparison of the analyte (the sample) to a reference material (the standard). In the case of carbon, for example
δ13C ≡ (13Rsa/13Rst) - 1
(1)
where sa and st denote sample and standard and 13R ≡ 13C/12C. Because isotope ratios commonly vary by a few parts per * Corresponding author’s present address: Woods Hole Oceanographic Institution, MS 8, Woods Hole, MA 02543-1539. † Present address: Department of Geology, University of Georgia, Athens, GA 30602-2501. S0003-2700(98)00343-6 CCC: $15.00 Published on Web 06/02/1998
© 1998 American Chemical Society
thousand, the right side of eq 1 is usually multiplied by 1000 and δ is expressed in per mil (‰). Until recently, conditions for sample-standard comparisons have been so rigidly controlled that these analyses could be described as “truly differential”. Pressures of sample and standard in the ion source of the mass spectrometer have been equalized and held constant; total ion currents have accordingly been matched. Sample and standard gases have been admitted alternately, and this repetition of comparisons has yielded a precision of better than 0.01‰ while requiring micromoles of sample and minutes of time. For many applications, speed and sensitivity are more important than the last increment of precision. In newly developed, continuous-flow techniques, a carrier gas is used to lead tens of picomoles of analyte through an ion source in a few seconds.1-4 Although the precision can approach the maximum theoretically obtainable (i.e., the shot-noise or ion-statistical limit),5 the performance routinely obtained in practical applications often falls short of that obtained in instrumental tests. This deficit, usually evident both in terms of low precision and systematic errors, often persists even after obvious problems such as (in the case of isotope ratio monitoring GC/MS) poor chromatographic resolution are eliminated.6-11 A fundamental point thus presents itself, since systematic errors ought to be eliminated by differential techniques. Something is wrong, and it is that conditions of analysis for the sample and standard are not truly identical. In continuous-flow analyses, the measurements of ion current ratios for sample and standard gases may be widely separated in time, may occur in the presence of differing background gases (in the ion source), and may involve different profiles of ion current as a function of time. The result is still reported using the δ notation, (1) Matthews, D. E.; Hayes, J. M. Anal. Chem. 1978, 50, 1465-73. (2) Barrie, A.; Bricout, J.; Koziet, J. Biomed. Mass Spectrom. 1984, 11, 583-8. (3) Freeman, K. H.; Hayes, J. M.; Trendel, J.-M.; Albrecht, P. Nature 1990, 343, 254-6. (4) Merritt, D. A.; Hayes, J. M. J. Am. Soc. Mass Spectrom. 1994, 5, 387-97. (5) Merritt, D. A.; Hayes, J. M. Anal. Chem. 1994, 66, 2336-47. (6) Brenna, J. T. Acc. Chem. Res. 1994, 27, 340-6. (7) Wong, W. W.; Hachey, D. L.; Zhang, S.; Clarke, L. L. Rapid Commun. Mass Spectrom. 1995, 9, 1007-11. (8) Bakel, A. J.; Ostrom, P. H.; Ostrom, N. E. Org. Geochem. 1994, 21, 595602. (9) Merritt, D. A.; Freeman, K. H.; Ricci, M. P.; Studley, S. A.; Hayes, J. M. Anal. Chem. 1995, 67, 2461-73. (10) Guilluy, R.; Billon-Rey, F.; Pachiaudi, C.; Normand, S.; Riou, J. P.; Jumeau, E. J.; Brazier, J. L. Anal. Chim. Acta 1992, 259, 193. (11) Eakin, P. A.; Fallick, A. E.; Gerc, J. Chem. Geol. (Isot. Geosci. Sect.) 1992, 101, 71-9.
Analytical Chemistry, Vol. 70, No. 13, July 1, 1998 2737
but the measurement is much less truly differential than its conventional antecedent. When CO2 serves as the analyte for continuous-flow analyses of 13C, water can transfer hydrogen to CO2 according to
C16O2+ + H2O f H12C16O2+ + OH•
12
(2)
or related reactions.9,12,13 The hydrogen-bearing product has m/z ) 45 and corrupts the 45/44 ion current ratio attributed to (13C16O2+ + 13C17O16O+)/(12C16O2+). Reactions between water and isotopically substituted CO2 similarly alter 46/44 ratios. Here we show that failure to maintain constant H2O/CO2 ratios during measurement of sample and standard ion-current ratios is a cause of significant systematic errors in continuous-flow analyses of 13C. This is true even when levels of H2O, present as a component of the background or as an incompletely removed component of the sample matrix, are low enough that they could readily be tolerated in conventional analyses. In the present study, we have (i) examined the effect of variations in water backgrounds on the stability of observed ioncurrent ratios, (ii) quantified errors in measured values of δ when samples of fixed size are analyzed in the presence of differing levels of background water, and (iii) similarly quantified errors when water backgrounds are held constant and sample size is varied. From these observations, a general relationship between the observed error in δ and the ratio of water to CO2 has been derived and experimentally evaluated. A final set of experiments has determined whether errors could be reduced by varying mass spectrometric conditions such as extraction potential and source conductance, both of which can affect the extent of ion-molecule reactions.
Figure 1. Experimental setup for measuring water-induced errors in continuous-flow carbon isotope ratio mass spectrometry.
EXPERIMENTAL SECTION Gas-Flow Pathway. A diagram of the experimental setup is shown in Figure 1. Two continuous flows of helium were supplied by a single zero-grade tank. (i) A flow of 3.24 mL of He/min (20 psi; 30 cm × 0.11 mm i.d. restrictor capillary) passed through a liquid nitrogen trap (150 cm; 0.5 mm o.d. × 0.32 mm i.d.) to a tee junction. At the junction, 240 ( 20 µL/min was drawn through a capillary leak (2.5 m × 0.11 mm i.d.) and into a set of changeover valves, which directed the gas to either the ion source or the waste vacuum line of a mass spectrometer. The remainder of the dry helium purged a Nafion dryer (described below). (ii) A second flow of 5.38 mL of He/min (20 psi; 50 cm × 0.11 mm i.d. restrictor capillary) was humidified by bubbling through deionized water at 23 ( 2 °C and then directed to a six-port valve and an open split. At the split, 260 ( 30 µL of He/min was drawn through a Nafion dryer, through a capillary leak (also 2.5 m × 0.11 mm i.d.), and into the waste vacuum or ion source of the mass spectrometer via the changeover valves. In addition to providing a vent for excess helium, the split isolated the mass spectrometer from small pressure variations introduced by the bubbler. Humidity Control. The liquid nitrogen trap dried the purge gas to a nominal dew point of -195 °C, while the carrier gas was initially saturated (at 23 °C) by the bubbler. The carrier gas was
then dried to desired levels by varying the temperature of a selectively permeable Nafion dryer. The residual humidity of Nafion-dried gas depends on the vapor pressure of bound water in the Nafion membrane and is therefore inversely related to the temperature at which the dryer is operated. The temperature dependence of Nafion dryers as well as the construction and performance of the dryer used in this study have been described previously.14 Briefly, the dryer consisted of a 60-cm length of 0.51 mm o.d. × 0.33 mm i.d. tubular Nafion membrane (PermaPure, Inc., Toms River, NJ) mounted coaxially in a stainless steel sheath that was purged by a countercurrent flow of dry gas. The initially saturated carrier could be dried to dew points from -43 to -94 °C when passed through the Nafion dryer, which was maintained at temperatures from 50 to -20 °C (( 2 °C) by immersion in constant-temperature baths. In some experiments, the carrier was dried by immersing coils of the capillary leak downstream from the Nafion dryer into a dry ice/acetone (-79 ( 3 °C) or liquid nitrogen/ethanol slush (-117 ( 4 °C). Introduction of CO2. As shown in Figure 1, under conditions of continuous flow, CO2 was added to the gas stream entering the ion source using one of two procedures. (i) A six-port sampling valve injected CO2 into an He stream of controlled humidity. Adjustment of the CO2 + He mixture directed to the six-port sampling valve allowed insertion of peaks of varying sizes. After accounting for the split, mixtures of 0.1-1.0% CO2 in He yielded approximately Gaussian peaks of 0.2-2.0 nmol of CO2 at the ion source. (ii) A special mixing device15 added “square waves” of CO2 into an auxiliary flow of He that was mixed with one of the two gas streams described above; the changeover valves were manipulated to select either the dry or the controlled-
(12) Brand, W. A. J. Mass Spectrom. 1996, 31, 225-35. (13) Meier-Augenstein, W.; Hess, A.; Hoffmann, G. F.; Rating, D. Isotopenpraxis 1994, 30, 349-58.
(14) Leckrone, K. J.; Hayes, J. M. Anal. Chem. 1997, 69, 911-8. (15) Merritt, D. A.; Brand, W. A.; Hayes, J. M. Org. Geochem. 1994, 21, 57383.
2738 Analytical Chemistry, Vol. 70, No. 13, July 1, 1998
humidity helium stream. The mixing device and the six-port valve were connected to a single tank of 1.0% CO2 in He (Indiana Oxygen Co., Indianapolis, IN; δ13CPDB ) -14.11‰). In the absence of helium carrier, pure CO2 with δ13CPDB ) -10.64‰ was introduced via the variable-volume inlet of the mass spectrometer (not shown in Figure 1). Mass Spectrometer. A Finnigan MAT 252 isotope ratio mass spectrometer was used to monitor amounts of water in either gas stream and to measure the concentration and isotopic composition of CO2. The ion source was differentially pumped, producing firststage (source housing) pressures of 7.9 × 10-6 and 7.6 × 10-6 mbar during introduction of variable humidity and dry helium streams. Second-stage (analyzer) pressure was 3 × 10-7 mbar during introduction of either gas stream. Ions were generated by electron impact at 70 eV, accelerated using a potential of 10 kV, and collected in Faraday cups. The mass spectrometer was tuned for maximum sensitivity (voltage difference of 100-120 V between the ionization chamber and the extraction lens; trap potential of 50 V) unless otherwise noted. Under these conditions, the efficiency of the mass spectrometer was 2400 ( 200 CO2 molecules per mass 44 ion collected, and 3000 ( 400 H2O molecules per mass 18 ion collected.14 Water and CO2 were quantified based on ion currents at mass 18 (18i) and mass 44 (44i), while 13C was calculated by vendorsupplied software (Finnigan Isodat 5.0) from ion currents at 44, 45, and 46 (44i, 45i, and 46i). Ion currents were converted to voltages via feedback electrometers and recorded digitally or on a stripchart recorder (18i). The feedback electrometers had resistors of 300 MΩ (44i), 30 GΩ (45i and 18i), and 100 GΩ (46i). Thus, a 1-nA current at mass 44 produced a signal of 0.3 V, while only 10 pA at mass 18 was required to produce the same 0.3-V signal. Characteristics of the mass spectrometer made it impossible to collect 18i simultaneously with 44i, 45i, and 46i. Consequently, mass 18 ion currents were observed immediately before and after each CO2 and δ measurement and were confirmed by continuously monitoring 18i during otherwise identical measurements. Procedures. Five separate experimental procedures were used to investigate the effects of water backgrounds on instrument stability and to determine the effects of water background, sample size, and mass spectrometric conditions (extraction potential and source conductance) on the magnitude of water-induced errors. Details of individual procedures are outlined separately at the beginning of each section in the following discussion of results. RESULTS AND DISCUSSION Water Backgrounds and Stability of Ion Current Ratios. Effects of background water on the stability of measured values of 13C were examined in three sets of measurements in which the rate of drift of observed values of δ was determined for repeated measurements of a CO2 standard. In the first set of measurements, no helium entered the ion source, and pure CO2 was repeatedly introduced from the variable-volume inlet of the mass spectrometer. Variations in δ thus reflected electronic instabilities and any other factors not related to continuous-flow measurements. In the second set, the mixing device added CO2 to a flow of dry helium that flowed continuously through a capillary leak connected to the changeover valves. Finally, to duplicate circumstances encountered in most continuous-flow systems, the
Table 1. Stabilities of δ Values Measured under Varying Source Conditions source conditions no helium
σb (‰)
σsnlc (‰)
0.01 -0.01 -0.02 -0.04
0.04 0.04 0.04 0.03
0.010 0.011 0.012 0.011
0.00055 0.00079 -0.00077 0.00009 0.00062
-1 -3 -1 -2
-0.011 -0.026 -0.014 -0.016 0.018
-0.06
0.04
0.014
-0.00079
-14
-0.114
-0.06 -0.07 -0.01
0.03 0.03 0.02
0.014 0.011 0.011
-0.00028 -0.00002 -0.00018 0.00043
-6 -4 -4
-0.051 -0.031 -0.030 0.066
-0.27
0.15
0.016
-0.0043
-208
-1.74
-0.26 -0.17 0.05
0.10 0.08 0.07
0.016 0.016 0.016
-0.0026 -0.0016 0.0021 0.0028
-193 -124 38
-1.60 -1.03 0.31 1.30
rms helium via changeover valves
rms helium via on-off valve
rms
∆13C driftd (‰/min)
av ∆13Ca (‰)
∆18i e (pA)
∆18i drifte (pA/min)
a Average error for 35 observations made over 120 min and standardized to the first observation. b Standard deviation of the population of observed δ values. c Shot noise limit or best attainable precision, calculated as in ref 5. d Calculated by linear regression of ∆13C vs time over a 120-min interval or, for on-off data, over the first 60-min interval. e ∆18i ) 18i(initial) - 18i(final), so that negative ∆18i denotes a declining water background over the 120-min measurement interval. Water drift rates ≡ ∆18i/t.
CO2 + He observations were repeated except that the gas stream entered the source through an on-off valve connected downstream from the changeover valves. In contrast to the changeover valve manifold, which was continuously maintained at high vacuum while the instrument was not in use, gas-handling systems upstream from the on-off valve were isolated or in some instances vented between measurements. Chances for accumulation of sorbed water were thus much higher for this subsystem. For each set of conditions, CO2 was repeatedly introduced, and values of δ were observed over a 120-min interval. Background levels of H2O were monitored by observation of mass 18 ion currents immediately before and after each series of measurements. To be sure that those spot checks were representative, water levels were also monitored continuously during a 120-min interval equivalent to each of the experiments. Results are summarized in Table 1. Observations are reported in terms of ∆13C, the difference between δ observed at time t and δ observed at time zero. The change in 18i over the measurement interval is given as ∆18i, the difference between 18i observed at the end and beginning of the interval. The rate of drift in the water background is ∆18i/t. Negative values of ∆18i or of ∆18i/t indicate instances in which the water background declined over the course of the measurement interval. Observed Drift Rates. In the absence of helium carrier gas, source pressures were 3-5 × 10-7 mbar when CO2 entered the source and 1-2 × 10-8 mbar between observations. Water backgrounds were low (18i < 5 pA) and were stable or declined very slightly during each experiment, with variations of less than -0.03 pA/min. For four different series of measurements, rates of drift in ∆13C (Table 1) were comparable to the drift rate of Analytical Chemistry, Vol. 70, No. 13, July 1, 1998
2739
0.0005‰/min reported for this instrument under similar conditions by Merritt and Hayes.5 Values of ∆13C scattered randomly around zero, and standard deviations of the populations of observations were 3-4 times higher than the theoretical (shot noise) limit of precision.5 When CO2 was carried into the ion source by a stream of He admitted via the changeover valve manifold, rates of drift in ∆ and standard deviations of the populations of individual values of ∆ (both shown in Table 1) were similar to those observed in the absence of helium. However, drifts were consistently negative, leading to average errors of -0.01 to -0.07‰. Water backgrounds were higher than in the helium-free measurements (25-40 pA) and declined during each experiment, with drifts of -0.11 to -0.03 pA/min. The experiment with the highest drift rate also showed the largest decrease in background water. Since the transfer of hydrogen to CO2+ increases observed values of δ, declining water backgrounds are consistent with negative drift in ∆13C. As shown in Table 1, drifts in ∆ were higher and more variable when helium was introduced through the external system that was not continuously purged or evacuated, although the flow rate and the source pressure were the same as in the preceding measurements. In fact, the rms drift rate observed for the first 60 min under these conditions was 8-12 times higher than that observed in the other experiments and is comparable to the 0.0045 ‰/min drift rate reported by Merritt and Hayes for this instrument in continuous-flow mode.5 Water backgrounds were high and variable (64-272 pA), with drifts of 0.31-1.74 pA/min. Increasing and decreasing water backgrounds corresponded to positive and negative drifts in ∆13C, suggesting that drifts in water backgrounds were responsible for observed drifts in ∆13C. Relationship between Drift Rates and Background Water Levels. Linear regression of the drift rate for ∆13C (in ‰/min) on the drift rate for ∆18i (in pA/min) yields a quantitative estimate of the effects of background water on observed values of δ (expressed in ‰/pA). This calculation is approximate because both drift rates are only roughly linear. The values summarized in Table 1 are nevertheless highly correlated (r2 ) 0.96). The resulting value of 0.0040 ( 0.0003 ‰/pA indicates that shifts greater than 0.1‰ can be expected whenever water backgrounds vary by as little as 25 pA. Observations with Humidified Carrier Gases. To further examine the effects of H2O, peaks of CO2 from the mixing device were introduced while dry helium, then helium of fixed humidity, and once again dry helium were directed to the source. Each square peak corresponded to 3.3 nmol of CO2 and generated a mass 44 ion current of 7 nA. The experimental series was repeated for helium dried with an ethanol/liquid nitrogen slush, a dry ice/acetone slush, and a Nafion dryer operated at -20, 0, 23, 35, and 50 °C. Water backgrounds were measured at the beginning and end of each experimental series and were confirmed by continuously monitoring 18i during otherwise identical measurements. Results are summarized in Figure 2, in which the lines depict the trends in δ values for 30 successive peaks of CO2 admitted at 1-min intervals. In each series, the switch to notperfectly-dry carrier gas (5 min into each experiment) led to increases in apparent values of δ13C, even though the actual isotopic composition of the CO2 was unchanged. 2740 Analytical Chemistry, Vol. 70, No. 13, July 1, 1998
Figure 2. Effect of water background on δ13C observed for a CO2 sample. At times before 5 and after 20 min, CO2 was introduced using dry helium carrier gas. At intermediate times, CO2 was introduced using a helium carrier of controlled humidity, resulting in backgroundcorrected water signals (∆18i) ranging from 27 to 406 pA.
Two trends were observed. First, 45i/44i increased significantly and caused significant changes in calculated values of δ13C as the water content of the carrier gas increased. This observation is consistent with the transfer of hydrogen from H2O to CO2+, resulting in excess ion current at mass 45. Similar effects were observed at mass 46. Second, values of δ responded slowly to changes in background water levels, particularly when the gas streams differed widely in humidity. The same slow response was observed for mass 18 ion currents and is probably due to sorption and desorption of water along the sample flow pathway. To minimize the influence of such lags, measurements were continued until a minimum of five successive peaks observed with a humidified carrier gas yielded values of δ differing by less than 0.1‰; these were averaged to provide δH, a value characteristic of the effects of hydrogen transfer at a given water background. Quantitative Relationship between 13C and 18i. To determine quantitatively the effect of background water, the systematic error was calculated for each experimental sequence represented in Figure 2 according to
∆13C ) δH - δ
(3)
where δ is the isotopic composition observed on a dry background. Since mass 18 ion currents during introduction of the dry purge gas were low (
