Anal. Chem. 1995, 67, 2486-2492
High-Precision D/H Measurement from Hydrogen Gas and Water by Continuous-Flow lsotope Ratio Mass Spectrometry Herbert J. Tobias, Keith J. GOOdman,t Craig E. Blacken, and J. Thomas Brenna*
Division of Nutritional Sciences, Cornel1 Universify, Ithaca, New York 14853
Two instrumental approaches are described for continuous-flowhigh-precisiondeterminationsof D/H ratios from hydrogen gas or via on-linereduction of water. In the first system, Ar is used as a carrier gas, with a Ni reduction furnace and a water trap to remove minor levels of unreduced water that are a potential source of memory effects. Precisions of SD < 10%0(GDs~ow)over a 6W%o range from -55 to +532%0 are obtained for liquid water (0.4pL). Iinearity is excellent over 4 orders of magnitude of D concentration in tap water (9> 0.9999),although precision degrades at enrichments GDs~ow> 5000%0. In the second system, a heated Pd metal foil functions as a mter to admit purified hydrogen into the mass spectrometer. Hydrogen gas injections are made into flowing Ar and are directed to the Pd filter (-330 "C) which passes hydrogen isotopes only while divertingthe carrier flow to waste. Precisions of these measurements are SD < 6%0 over the D enrichment range -213 to 340%0,with excellent linearity (1.2 > 0,9999) and accuracy ( 0.999) over 3 orders of magnitude of D concentration. Neither system shows any sign of memory effects when water is analyzed. 'Ihe data indicate that either one of these systems is a useful means for continuous-flowIRMS of D/H isotope ratio determinations. High-precision gas isotope ratio mass spectrometry (GIRMS) has been the technique of choice for determination of isotopes of the light elements, H, C, N, 0, and S for several decades.' The classical dual-inlet approach is now accompanied by continuousflow methods, first developed for the interface of the isotope ratio mass spectrometer (WvlS) with gas chromatography (GC)? but now available for analysis of products of elemental analyzers3and other mixtures such as breath: and more recently liquid chromatograph~.~The precision of continuous-flow methods a p proaches that of dual-inlet with more rapid analysis and often without the requirement for cumbersome pretreatment6 + Present address: Department of Food Science and Human Nutrition, Iowa State University, Ames, IA 50011. (1) Brenna, J. T. Acc. Chem. Res. 1994,27, 340-346. (2) Matthews, D. E.; Hayes, J. M. Anal. Chem. 1978,50, 1465-1473. (3) Preston, T.; Owens, N. J. P. Analyst 1983,108, 971-977. (4)Preston, T.; McMillan, D. C. Biomed. Enuiron. Mass Spectrom. 1988,16, 229-235. (5) Caimi, R. J.; Brenna, J. T. Anal. Chem. 1993,65,3497-3500. (6) Menitt, D. A.; Hayes, J. M. Anal. Chem. 1994,66, 2336-2347.
2486 Analytical Chemistty, Vol. 67, No. 14, July 15, 1995
Continuous-flow methods have been developed for all the classic IRMS elements except for H, with chromatographic methods available for C, and recently for N73 An inert carrier gas is required to move bands of analyte gas through on-line separation (e.g., GC), through treatment (e.g., combustion, reduction, and drying), and into the IRMS. Moreover, it maintains constant pressure, independent of sample level, in the IRMS ion source. Helium is the carrier of choice because of its chemically unreactive character. D/H analysis is of intense interest for studies of variation due to natural processes and for tracer studies involving enriched samples. Of particular current interest is the analysis of D/H originating in water for studies of natural isotope fractionation and for analysis of artificial enrichment in tracer experiments. Conventional high-precision IRMS methods require the conversion of the hydrogen in water to Hz and HD using cumbersome and time-consuming chemical reduction schemes, although there are reports of an equilibration system for automated analysis with precision approaching that of conventional off-line analysis? Besides rapid analysis, a continuous-flow system with on-line chemical reduction would reduce considerably the possibility of contamination during off-line preparation. There are three principal reasons why continuous-flow D/H analysis is a challenge using a conventional He-based carrier scheme. First, the abundance sensitivity required to eliminate the tail of m/z 4,due to an overwhelming amount of 4He, in the m/z 3 detector (for HD) is beyond the capability of commercial IRMS instruments. As an additional mass interference, common sources of laboratory He include -1 ppm 3He, which interferes directly with the detection of HD at m/z 3. These limitations in resolution translate into limitations in sensitivity. Second, hydrogen isotopes are highly subject to fractionation, which is usually observed during gas transport. Third, attempts to reduce water on-line for D/H analysis frequently suffer from severe memory effects due to exchange with sources of ambient water, which significantly degrades the analysis.1° Here, we describe two interfaces for continuous-flow IRMS determinations of D/H from hydrogen gas, and from water with on-line reduction, at high precision and accuracy. The &st system ("Ar carrier") relies on Ar carrier gas for introduction into the ion source and replaces He as a conventional carrier gas." The second system ("Pd filter") uses a hydrogen filter, in the form of (7) Menitt, D. A; Hayes, J. M. J. Am. Sot. Mass Spectrom, 1994,5, 387-397. (8) Preston, T.: Slater, C. PYOC. Nutr. SOC.1994,53, 363-372. (9) Coplen, T. B.; Wildman, J. D.; Chen, J. A n d . Chem. 1991.63, 910-912. (10) Wong, W. W.; Cabrera, M. P.; Klein, P. D. Anal. Chem. 1984,56, 18521858. (11) Tobias, H. T.; Goodman, K J.; Brenna, J. T. Int. Mass Spectrom. ConJ 1994, 302.
0003-2700/95/0367-2486$9.00/0 0 1995 American Chemical Society
Injectors
Oven
A.
W
I
'
Combustion
1
Combustion
fl ,/ fl n
Reduction Reaczy
Amolifiers
-
T
V I
v
Water Trap
Injectors
Trap
/
Open Spht
Finnigan MAT 2 5 2
Magnetic Sector
Figure 1. Schematic of the Ar carrier system for D/Hanalysis after on-line reduction of water samples. A Varian 3400 GC is interfaced to a Finnigan MAT 252 GIRMS instrument via a combustion reactor filled with CuO, a reduction reactor filled with Ni metal, a Nafion water trap, and an open split.
a heated Pd foil, to pass hydrogen isotopes while retaining carrier gas and residual impurities. EXPERIMENTAL SECTION
A Finnigan MAT 252 high-precision GIRMS instrument, equipped with multiple Faraday cup detectors, amplifiers, and digitizers for simultaneous monitoring of m/z 2 for Hz and m/z 3 for HD, is the mass spectrometer used in this work. The instrument is operated at 10 kV accelerating potential and with a source chamber pressure of -1 x Torr for the Ar carrier system and -4 x Torr for the Pd filter system. The absolute sensitivity for Hz was determined to be -5 x 103 molecules/ion. Ar Camier System. A GIRMS instrument is interfaced to a Varian 3400 gas chromatograph according to the schematic shown in Figure 1. This system is set up to analyze D/H from water samples and potentially from organic materials. For this work, a split/splitless injector, in the split mode (split ratio 100:1), is used exclusively to introduce 0.4 pL of water into a stream of highpurity Ar flowing at -15-20 cm/s. The injector temperature is held at 250 "C while the GC oven is held at 210 "C. The interface consists of two sequential furnaces placed in-line after a 10 m length of deactivated fused silica capillary (10 m x 0.32 mm) in the GC oven. The furnaces can serve as either combustion or reduction reactors. For this work, the first furnace is a conventional combustion furnace,I2loaded with CuO and Pt metal held at 850 "C, which combusts organic molecules to COZand HzO. This unit was placed in-line for future experiments involving combustion of organics with subsequent reduction of the resulting water and has no apparent effect on injected water samples in the present experiments. The second furnace is held at -850 "C, is loaded with Ni metal, and is used to reduce water to hydrogen gas. The two furnaces are connected by a heated capillary to facilitate transfer of water vapor. In initial experiments in which the furnace output was admitted directly to the IRMS ion source, analyte peak shapes noticeably degraded and showed memory effects with repeated use. This poor performance appeared to be characteristic of the adsorption of unreduced residual water in the transfer line. A Nafion (Dupont, Wilmington, DE) water trap installed following the second furnace produced uniform peak shapes and eliminated memory effects and was used in all subsequent experiments reported here. As the data will show, nonquantitative reduction did not have a substantial effect on the measurements. The
'L
Open To Atm
Cryogenic Trap
\ U s;l;:;
Steel
Figure 2. Instrument schematic of the Pd filter system for continuous-flow analysis of D/H ratios from hydrogen gas samples. The online reduction interface used in the Ar carrier system, consisting of a combustion furnace, reduction furnace, and water trap, can be also be used with the Pd filter for on-line reduction of water for D/H analysis.
stream of hydrogen bands in Ar carrier emerging from the water trap passes through an open split made from a glass "Y" connector (Restek, Bellefonte, PA) and finally into the IRMS valve block via a transfer capillary (0.1 mm i.d. x 4 m). Pd Filter System. A Pd foil used as a hydrogen filter was incorporated into a second interface to facilitate continuous-flow admission of pure sample hydrogen into the mass spectrometer. This system consists of a GIRMS instrument interfaced with a Varian 3400 GC according to the schematic shown in Figure 2. The split/splitless injector (250 "C) in the splitless mode was used to introduce a range of enriched hydrogen gas into a carrier of high-puntyAr gas flowing at 50 cm/s. With the GC oven at room temperature, the carrier transports sample gas through a 10 m deactivated fused silica capillary (10 m x 0.32 mm). The capillary end is placed in nominal contact with a 0.127 mm thick, l/g in. diameter, round, 99.9%pure Pd foil (Aldrich Chemical, Milwaukee, WI) housed within a l/16 in. to l/4 in. stainless steel union and maintained at -330 "C by use of a resistive heater surrounding the exterior. The capillary side of the foil is open to the atmosphere and serves as an open split, with carrier gas and untransmitted hydrogen flowing to waste. A stainless steel tube (l/16 in. i. d., 1 / 4 in. 0.d.) is machined such that the Pd foil can be compressed via a knife edge to create a vacuum tight seal on the l/4 in. union for direct connection to one of the IRMS valve blocks. At 330 "C, hydrogen gas, principally Hz and HD, is absorbed into the foil while all other species are directed to waste. The vacuum sets up an Hz and HD gradient that drives a net flow into the IRMS. The pressure on the vacuum side of the Pd foil is nominally identical to the pressure in the ion source when no sample is present. A minor background level of Hz and HD is associated with the membrane when heated, presumably due to atmospheric hydrogen. Initial experiments showed that the m/z 18 signal was significantly elevated as the temperature of the foil was increased. This appears to be due to water desorption from the heated stainless steel on the vacuum side of the foil. To eliminate this potential source of interference, a cryotrap consisting of a loop of tubing placed into a liquid Nz bath was inserted between the Pd foil and the valve block. This drying stage eliminated the m/z 18 Analytical Chemisty, Vol. 67, No. 14, July 15, 1995
2487
signal with no obvious peak broadening, probably because this transfer region is held at high vacuum. Hydrogen gas samples over -600% range, from -213 to +380?h, were prepared in Vacutainers (Becton Dickinson, Rutherford, NJ) by mixing 98%pure HD gas (Kor Isotopes, Cambridge, MA) with laboratory grade hydrogen gas, at a pressure compatible with dual-inletanalysis. Vacutainers were sampled and D/H ratios determined by conventional dual-inlet IRMS. A few hours later, the same samples were backfilled with high-purityAr gas to bring the overall pressure to -1 atm. Twenty-five microliters of this gas mixture, containing -1 pmol of hydrogen, was injected into the carrier stream of the Pd filter interface using a gas-tight syringe, for continuous-flow analysis. The Pd filter system was also set up to analyze liquid water directly. As shown in the upper inset of Figure 2, combustion and reduction furnaces and a Nafon water trap are inserted between the GC oven and the Pd filter. As with water reduction in the Ar carrier system, the combustion furnace is in-line for future studies of D/H ratios in hydrocarbons and has no observable effect on water. All experimental parameters for the GC, furnaces, and water trap are identical to those described for the Ar carrier system. Water standards were calibrated for D/H ratios off-line by clyogenic distillation onto Zn turnings (Biogeochemical Laboratories, Bloomington, IN) and roasting at 600 "C in sealed tubes with subsequent dual-inlet analysis. Data Aquisition. Our versions of the vendor-supplied software, ISODAT 4.11 and 5.2, do not permit continuous monitoring and recording of the m/z 2 and 3 signals required for these experiments. For this reason, a homebuilt data acquisition system was constructed to acquire signal from the Faraday cup amplifier outputs. Signal was connected to unity-gain amplifiers and passed to DT2802 chromatography boards (Data Translation, Marlboro, MA) set to acquire data 22 bits deep during 267 ms acquisition times. Digitized m/z 2 and 3 data are then uploaded to ISODAT for peak detection, integration, and isotope ratio calculation. The m/z 2 channel is used for peak detection in these experiments. Since the vendor-provided data system is unable to store continuously acquired hydrogen measurements, a simultaneous comparison of analytical figures of merit of the two systems is not possible. Noise levels were therefore evaluated in several indirect ways. First, the systems were compared for COZ measurements by simultaneous acquisition of the ampuer output, splitting the signal between the two acquisition systems. Analysis of 1.8 V, 30 s wide pulses admitted from an adjustable volume gave 6I3C = -40.82 i 0.043% and -40.76 f 0.038% (mean i SD, n = 28) for the vendor and homebuilt systems, respectively. Second, root-mean-square (rms) noise for simultaneous static background measurements was determined for the two systems. In this configuration, signal was split to the home-built system and simultaneously acquired using ISODATs amplilier evaluation "UFC" diagnostic. This program acquires data continuously and outputs the mean and standard deviation, to two decimal places, for a user-defined number of acquisitions and integrations. For 80 consecutive points acquired at -0.25 s duration in the m/z 2 channel, both systems yielded rms noise of 0.02 mV in the m/z 3 channel, rms noise values of 0.08 and 0.06 mV were obtained for the vendor and homebuilt systems, respectively. Third, using the homebuilt system alone, 10 replicate hydrogen pulses from the bellows yield an SD = 1.4% for a 1.4 V plateau at m/z 2. Last, the SD for standard pulses placed between eluting peaks for the 2488 Analytical Chemistry, Vol. 67,No. 74, July 15, 7995
memory experiment reported below gave SD < 3% at 300-500 mV signal levels. Taken together, these results indicate that the home-built system does not contribute significantly to noise in the present measurements. Notation. High-precision D/H ratio data are usually expressed using the standard GDs~ownotation and defined as the relative difference in isotope ratio in parts per thousand (%) between the sample and a standard:
where R, refers to the D/H of the sample (SPL) or a standard. In the case of hydrogen, the material known as standard mean ocean water (SMOW) with isotope ratio IDI/[Hl = 0.OOO 155 76 is commonly used as an international standard.13 HdHD Detector Response. In conventional high precision dual-inlet IRMS analysis, sample gas is admitted to the ion source continuously from a volume and measurements are recorded when the signal has plateaued. Under these conditions, matching of the time constants associated with the amplifier/detector systems is not critical, since the average signal level changes very slowly, principally because of slow depletion of the sample. However, for continuous-flow applications, bands of sample gas swept into the ion source by the carrier gas cause much more rapid changes in signal level. Under these conditions, matching of time constants for the relevant mass channels is a critical consideration. The amplification in the m/z 2 and 3 channels on the MAT 252 corresponds to the 100Gfold difference in abundance of H2 and HD, such that outputs of 1 and 1000 mV/pA beam current are obtained, respectively. The nominal time constants for the RC circuits of these feedback amplifiers are 0.15 and 2 s, respectively. This order of magnitude difference in time constants is expected to precipitate a substantial difference in peak shapes, in which the slower responding channel will fail to record signal represented in the faster responding channel. The definition of peak start/stop and calculation of peak area ratios may also be affected. To characterize this effect, a 30 s hydrogen pulse was admitted to the ion source from one of the dual-inlet volumes. As expected, the m/z 3 signal was observed to rise and fall at a slower rate than that of m/z 2. To evaluate the effect of detector response for on-line analysis, hydrogen gas was admitted from the bellows at varying pulse widths to roughly simulate peaks of varying widths. Figure 3 presents data for hydrogen standard gas pulses, with isotope ratios calculated compared to the longest pulse (30 s). Deviations in parts per thousand calculated relative to this pulse are plotted along the left ordinate and RSD for each replicate set (n = 5) is plotted along the right ordinate, with both plotted against pulse widths along the abscissa. Isotope ratio accuracy degrades dramatically below 10 s pulse width, without substantial degradation in precision. As expected, the origin of this effect is a highly reproducible underestimation of the m/z 3 peak area, which results in smaller area ratios m/z [3]/[2] ratios and smaller 6D values. The mismatch in time constants is the major contributor to this phenomenon, although time-of-flight effects due to isotope effects from H2 and HD may also be a minor contributor. (12) Goodman, IC J.; Brenna, J. T. Anal. Chem. 1992, 64, 1088-1095. (13) Ehlennger, J. R.; Rundel, P. W. In Stable Isotopes: Histoly, Units, and Instrumentation; Rundel, P. W.. Ehleringer, J. R, Nagy, K. A,, Eds.; SpnngerVerlag: New York, 1988.
-20
f
I
. I
-40 Y
Table I.D/Hfrom On-Line Water Analysis via the Ar Carrier System' 2.0
3 -60 E
.$
-80
11.0
a
'5 -100
s
0.5
-120
--
-140 0
5
10
+ 0.0
15
20
25
30
H2 Pulse Width (s) Figure 3. Data presented for hydrogen standard gas pulses at varying pulse widths and isotope ratios and calculated relative to the longest pulse. Pulse widths are plotted along the abscissa. Deviations from the 30 s pulse reference values are plotted along the left ordinate and the % RSD from the longest pulse is plotted along the right ordinate. The solid squares represent the YO RSD at each pulse width and the open squares represent the isotope ratio measured at each pulse.
I
I
1.0
I
0.2
I
0.0
\
i- -! 0
5
.
+i
10
15
20
I
25
30
Time (s) Figure 4. Simultaneous response of the m/z 2 and 3 detectors as a band of hydrogen gas in Ar carrier passes through the ion source. For convenient comparison, the signals were normalized by subtracting a calculated baseline from each plot with subsequent scaling by appropriatefactors. The slower response of the m/z3 channel is due to the larger time constant of the amplifier system.
These data are consistent with our preliminary measurements that produced poor accuracy and precision for very sharp peaks analyzed using the Pd filter system. Broader peaks require slower response times and permit the slower responding channel to keep up with the faster channel, yielding more accurate and precise analysis. For subsequent analyses, conditions were maintained to give broader peak widths than can be routinely obtained with either system. RESULTS AND DtSCUSSION
Ar Carrier System. The Ar carrier system was assessed using samples of 99.9%DzO diluted serially with tap water and analyzed by on-line Ni reduction. Figure 4 is a plot of the peak shapes for the m/z 2 and 3 channels as one representative peak elutes. The raw m/z 2 signal is nominally %fold greater than the m/z 3 for isotope ratios near that of SMOW because of the arbitrary amplification difference between the channels, which makes comparison of peak shapes difficult. Therefore, data are presented in parameterized form, where the baseline of each channel is set to zero using a calculated offset, and the heights are made equivalent by multiplying the resulting chromatogram by an appropriate factor. This figure shows that the m/z 3 signal rises more slowly, peaks -2 s later, and is -20% broader than the m/z 2 signal. These differences are consistent with the difference in time constants of the feedback amplifiers between the channels. The peak detection/integration algorithm included
DzO dilution
mean area ratio m / z [3]/[2]
isotope ratio*
tap water 1 x 10-6 I 10-5 1 10-4 1 10-3 1 x 10-2
0.2734 0.2779 0.2943 0.4433 1.7208 15.915
-55.00 & 4.94 -39.47 f 9.76 17.36 f 9.60 532.26 f 7.25 4947.9 35.7 54012 f 289
*
a No outliers excluded. * Mean f SD (mean area ratio m / z [31/[21 of tap water used as standard to calculate isotope ratios, ~DSMOW, '??I.
in ISODAT 4.11 was used with these and the rest of the data in this report. No obvious effects of the m/z 3 broadening were apparent. Should this become significant in future work, peak shift algorithms, such as discussed by Ricci et al.,14may be used to correct for different widths of the m/z 2 and 3 peaks. Table 1 shows means of five replicates of the water samples over 4 orders of magnitude of D concentration in tap water. Peak heights were -300 mV. Precisions of SD < lo??are observed over a 600%0 range from GDSMOW = -55 to 53%. A linear leastsquares fit of the area ratio m/z [31/[21 plotted against DzO dilution yields a correlation coefficient of 72 > 0.9999 over the 4 orders of magnitude range (slope 1565.44 f 7.22; intercept 0.2516 f 0.0324). No outliers were excluded from this data set. These data indicate that the precision and linearity of on-line water reduction in the Ar carrier system approach that of dual-inlet IRMS for samples of D/H near SMOW. Precisions (SDs) of up to 7% are reported and are acceptable in many types of studies in biological and geological samples,13J5 particularly if analysis is sufficiently rapid. During Ar carrier measurements we noted an increase in the high-voltage leakage current. Examination of the mass spectrometer optical elements indicated a substantial amount of metal deposition on some ceramic insulators. We interpret this as resulting from sputtering of metal lens elements by the intense Ar beam, estimated to be -50 nA It is likely that this beam, consisting of mass 40 u ions, efficiently sputters the metal lenses upon which it impinges. Steps to minimize this serious problem will be required for routine use of Ar as a continuous-flowcarrier. IRMS Nonlinearity Using the Pd Filter System. In conventional dual-inlet GIRMS analysis, sample and standard pres sures are balanced before the measurement to ensure that differences in instrument response related to ion source pressure are minimized. For on-line measurements of D/H isotope ratios with Ar carrier flowing into the IRMS, the ion source pressure remains constant despite differences in H2 peak intensities. This feature is similar for standard GCC/IRMS analysis of C or N where errors due to nonlinearities are minimized with He carrier.6 In contrast, the Pd filter passes only hydrogen, so the source pressure varies with flow of analyte into the source. An additional nonlinear interference, specific to hydrogen, is the well-known contribution to the m/z 3 signal from H3+ resulting from pressuredependent ion/molecule reactions. It was therefore necessary to evaluate these effects before making continuous-flowisotope ratio measurements using the Pd filter system. (14) Ricci, M. P.; Memtt, D. A; Freeman, K H.; Hayes, J. M. Org. Geochem. 1994,21(6/7), 561-572. (15) Schoeller. D. A; Taylor, P. B.; Shay, K Obesity Res. 1995, 3 (Suppl. l ) , 15-20.
Analytical Chemistry, Vol. 67, No. 14,July 15, 7995
2489
-25
P
0.0
-
d 0.5
1.o
1.5
2.0
m/z 2 Signal (V)
0.0
Figure 5. Ion source nonlinearity for D/H ratios measured as a function of source pressures. The percent error from an arbitrarily chosen reference ratio, 2 V in the m/z 2 channel, is plotted vs the /n/" 2 signal intensity.
Overall source nonlinearity was characterized by introducing hydrogen gas over a range of pressures from a variable volume. For each determination, sample gas was admitted to the ion source and integrated over a 2 s window after allowing a 25 s settling time. The bellows was then compressed slightly and the measurement repeated. F i e 5 shows the results of this experiment, with each point representing a single measurement and the area ratio m/z [31/[21 of -0.38, at 2 V for m/z 2, chosen arbitrarily as a reference value. An apparent 6D range of -300%0 is observed relative to the strong signal at 2 V and is not linearly related to signal. These data indicate that 6D calculations for continuously varying pressures, such as those encountered in the Pd filter system, will suffer from substantial nonlinearity. Further, this nonlinearity increases in severity as the point-by-pointamplitudes within a peak diverge from the reference level. To adjust for these severe pressuredependent effects, a correction equation was generated by fitting a curve to the m/z 3 vs 2 plot for-standardhydrogen gas admitted from the bellows at signal levels from 0 to 2 V. The data could be adequately modeled with a thiid-order polynomial (v = -2.37 x 0.2968~ 0.0741s - 0.01431x3, ?=0.999 98). This equation was used to adjust the m/z 3 signal through a point-by-pointcorrection for the on-line data generated subsequently. This correction was first tested on standard hydrogen pulses of 30 s duration collected continuously and admitted by valve opening and closing, with data collected continuously using the homebuilt data acquisition system. The line shape for these pulses resembles a square wave more closely than a chromatographic peak. Raw files were corrected for nonlinearity using the third-order polynomial before uploading to ISODAT for peak integration. F i i r e 6 illustrates the effect of correction on pulse data. The raw area ratios and corresponding 6D (%) are plotted as a function of signal along with corrected data. The uncorrected points show the expected large systematic increase in apparent GDSMOW with signal, over a range of -169 to +2% using the ratio at 1.75 V (m/z 2) as the arbitrarily chosen reference. The pointby-point correction resulted in dramatically improved data, giving a range from -19 to +1%. This correction was also applied to hydrogen gas injected into the continuous-flow system and resulted in appreciable improvement. The uncorrected points show the expected large systematic increase in apparent 6Ds~owwith signal, over a range of -119 to o%o using the ratio at 1.75 V (m/z 2) as the arbitrarily chosen reference. Again, the point-by-point calibration resulted in improved data, giving a range from -16 to +1%. These results suggest that bias due to fluctuations in peak heights degrades the measurement. To test this hypothesis, we
+
2490 Analytical Chemistry, Vol. 67, No. 14, July 15, 1995
+
1.0
0.5
1.5
2.0
2.5
m/z 2 Pulse Height (V) Figure 6. Measured hydrogen isotope ratios for standard 30 s wide hydrogen gas pulses, collected continuously using the home-built acquisition system, and plotted as a function of signal. Isotope ratios were referenced to the ratio measured at a pulse intensity of -1.75 V in the d z 2 channel. The striking nonlinearity is diminished considerably by correcting the m/z 3 chromatogram point by point before calculation of isotope ratios. The correction is a polynomial equation calculated by applying a least-squares fit to the data of Figure 5. E 4
P .F3
I
I
901.-
I
0
40
20
80
60
100
SD Isotope Ratio (permil) Figure 7. Effect of unmatched peak intensities on the precision of qeasured isotope ratios of hydrogen gas passing through the Pd filter. The SD of the N z 2 signal is plotted on the ordinate against the SD of the measured isotope ratio for five or more replicate samples on the abscissa. The positive relationship indicates that matching of signal levels is important for optimal precision.
4
1.0 1 Pd
i0.8
*a 0.6
'
;I
c.'
t
0.4
Ar
,
I
1 1I
0.2 "
,
\
'.-. --.-,---
0.0 0
10
20
30
40
1 50
60
70
Time (9) Figure 8. Comparison of typical d z 2 peak shapes generated by the Ar carrier or the Pd filter system. Data were normalized as in Figure 4. No major difference in peak width or tailing is apparent.
analyzed hydrogen gas with varying amounts of gas injected to simulate fluctuations in peak heights. Figure 7 shows a plot of the SD of m/z 2 peak heights vs the SD of the isotope ratio measured for continuousflowanalysis. The slope of a regression line through these points is strongly positive and confims the hypothesis. In addition, acceptable precision (SD < 10%)results when peak heights are controlled within easily attainable limits (A100 mv). Correction of unreported preliminary data from the Pd filter system did not result in acceptable precision (~10%) where m/z [2] peak intensities differed by more than flOO mV. For this reason, subsequent analyses are reported without correction and peak intensities were controlled within this limit (&lo0 mv), unless otherwise stated. This is easy to achieve if little or
Table 2. D/H from Hydrogen Gas Analysis via Pd Filter System
dual-inlet analysis sample gas isotope rationsb 1 2 3 4 5d 6
Pd filter system analysis isotope ration,c deviation from regression line ( O h )
mean area ratio m/z [31/[21
-212.77 f 0.48 -86.20 f 0.28 48.40 f 0.68 152.02 f 0.80 341.94 f 0.90 380.47 f 0.94
-212.58 f 5.72 -85.15 f 3.11 47.97 f 5.50 150.16 f 2.86 341.00 f 3.91 382.46 f 10.06
0.2227 0.2860 0.3521 0.4029 0.4976 0.5182
0.19 1.05 -0.43 -1.86 -0.95 1.99
Mean f SD (GDs~ow,O h ) . Single aliquot, eight standard-sample comparisons. c Replicate gas injections: 1 (n = 5), 2-4 and 6 (n = 4), and 5 (n = 3). d One outlier excluded. Table 3. D/H from On-Line Water Analysis via Pd Filter System.
dual-inlet analysis isotope ratiobat
sample
DzO dilution
a be
tap water 1.00 x 10-6 1.00 10-5 3.25 10-5 5.50 10-5 7.75 10-5
C
d
e f @
a (n
-55.00 f 1.40 -46.87 f 0.67 -3.56 f 4.63 111.25 f 2.10 315.75 f 1.90 464.28 i 12.66
mean area ratio m/z [3]/[2] 0.2305 0.2341 0.2381 0.2620 0.2992 0.3171
Pd filter system analysis isotope ratiobad deviation from regression line (%) -57.78 f 9.61 -37.66 f 3.27 -26.35 f 8.99 120.74 f 8.29 333.28 f 4.60 450.83 f 5.43
-2.78 9.21 -22.79 9.49 17.53 -13.44
One outlier excluded for each sample determination. Mean f SD (BDSMOW, %). c Replicate reduction tubes, n = 2. d Replicate water injections: = 15), b (n = 5). and c-f (n = 4). e No Outliers excluded.
no split is used at the injector. Pd Filter System. Figure 8 is a comparison plot of typical m/z 2 peak shapes for the Pd filter and Ar carrier systems, presented in normalized form. Although precise peak shapes are sensitive to experimental conditions, the plot shows that there are no major differences in peak width or tailing between the two systems. The Pd filter system was first tested by analyzing samples of laboratory grade hydrogen gas with increasing enrichments of D over the natural abundance range, at a signal level of -600 mV. Each sample of hydrogen gas was analyzed first by dual-inlet analysis and second using the Pd filter system. Data from these experiments are calibrated to the dual-inlet results and presented in Table 2. A linear least-squares plot of measured area ratios m/z [3]/[2] vs dual-inlet analyses reveals excellent linearity in the range -213 to +380?? (12 > 0.9999; slope 4.966 x f 0.015 x lo-*; intercept 0.3283 f0.0004), with one outlier excluded (from sample 5). An average of four replicates for each sample results in precisions of SD < 6%, except for sample 6 with SD = 10%. Deviations from the least-squares regression line are < P ?and indicate the accuracy of these analyses is comparable to the precision. This experiment demonstrates that the there are no major D/H analysis problems associated with the Pd filter system. The fraction of hydrogen entering the IRMS vacuum system through the Pd filter was calculated from the integrated intensity of the m/z 2 channel for injection of a known amount of hydrogen gas and the measured absolute sensitivity. The overall transfer of gas through the filter is -0.1%, with 99.9% flowing to waste. This figure was confirmed by separate measurements using water injected into the Ar carrier system to check the sensitivity level. Samples of 99.9% DzO diluted serially with tap water over 3 orders of magnitude were analyzed using the Pd filter interfaced with the on-line water reduction apparatus. The data approach the results of the previous Pd filter experiment. The overall signal level was double that of the Ar system for equivalent sample
injections. The isotope ratios were measured relative to a SMOW standard gas admitted into the instrument a few times during each set of sample replicates. However, the depletion of the standard from the bellows causes the signal to decrease steadily throughout the analysis, resulting in a steady decrease in apparent isotope ratio due to nonlinearity. A linear least-squares plot of the uncorrected m/z [31/[21 data referenced to these uncorrected pulses vs GDs~owresulted in an 12 > 0.9995 (slope 0.52712 5 0.00400; intercept -302.97 f 8.96). The correction routine was then applied only to the pulses, and a least-squares fit to the resulting plot, in which the m/z [3]/[2] data were referenced to corrected pulses, yielded an Z > 0.9998 (slope 0.5346 f 0.003; intercept -296.82 i 6.86). This improvement may be small because the peak amplitudes in this data set are relatively well matched. Presented in Table 3 are results of analysis of several water samples of known D/H ratio, near the natural-abundance range. These samples were analyzed by both dual-inlet analysis and the Pd filter system. The data were referenced against uncorrected standard hydrogen gas admitted from the bellows and calibrated to the dual-inlet results. The precision is uniformly
