Anal. Chem. 1884, 56, 1852-1858
1852
LITERATURE CITED (1) Corey, E. J.; Nlwa, H.; Falck, J. R. J. Am. Chem. SOC. 1979, 101,
1586-1587. (2) Falck, J. R.; Manna, S. Tetrahedron Left. 1982, 23, 1755-1756. (3) Corey, E. J.; Marfat, A.; Falck, J. R.; Albrlght, J. 0. J. Am. Chem. SOC. 1080, 102, 1433-1435. (4) Chacos, N.; Falck, J. R.; Wlxtnom. C.; Capdevlla, J. Biochem. Biophys. Res. Common. 1982, 104, 916-922. (5) Chacos, N.; Capdevila, J.; Falck, J. R.; Manna, S.;Martln-Wixtrom, C.; Gill, S. S.;Hammock, B. D.; Estabrook, R. W. Arch. Blochem. BioPhyS. 1983, 223,839-648. (6) Snyder, G. D.; Capdevih, J.; Chacos, N.; Manna, S.;Falck, J. R. froc. Nafl. Acad. Sci. U . S . A . 1983, 8 0 , 3504-3507. (7) Oliw, E. H.; Guengerlch, F. P.; Oates, J. A. J. 8/01.Chem. 1981, 257, 3771-3781. (8) Li, Y. H.; Herman, J. A.; Harrison, A. G. Can. J. Chem. 1981, 5Q, 1753-1759.
(9) Harrison, A. G. "Chemical Ionization Mass Spectrometry"; CRC Press: Boca Raton, FL, J983; pp 12-15 and references clted. (10) Kiser, R. W. Introduction to Mass Spectrometry and I t s Appllcatlons"; Prentlce-Hall: Englewood Cliffs, NJ, 1965; pp 163-166. (11) Subba Rao, S. C.; Fenselau, C. Anal. Chem. 1978, 50, 511-515. (12) Sieck, L W. Anal. Chem. 1979, $ 1 , 128-132. (13) Sieck, L. W. Anal. Chem. 1983, 55,38-41. (14) Jelus, B.; Munson, B.; Fenselau, C. Anal. Chem. 1974, 46, 729-730. (15) Jardlne, I.; Fenselau, C. Anal. Chem. 1975, 4 7 , 730-733. (16) Still, W. C.; Kahn, M.; Mltra, A..J. Org. Chem. 1978, 43, 2923-2925. (17) Brundle, C. R.; Turner, D. W. Int. J. Mass Specfrom. Ion fhys. 1965, 2. 195-220.
RECEIVED for review December 12, 1983. Accepted April 9, 1984.
Evaluation of a Dual Mass Spectrometer System for Rapid Simultaneous Determination of Hydrogen=2/Hydrogen-1 and Oxygen- 18/0xygen- 16 Ratios in Aqueous Samples William W. Wong,* Mercedes P. Cabrera, and Peter D. Klein
USDAIARS Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Texas Children's Hospital, Houston, Texas 77030
The commercial prototype of a twln mass spectrometer system (AquaSIRA, VG Instruments, Ltd.) has been evaluated for Its Ilnearlty, accuracy, and preclslon In the measurement of 2H/k and leO/i'O ratlos In a serles of water standards and urine samples. Wlth an Instrument correctlon, (2H/'H)-d = [(2H/1H),, 13.23]/0.9139, the 'H/'H ratlos (89-696 ppm) were determined wIth a mean preclslon of 0.5 ppm and a mean accuracy of 0.87 f 0.67 ppm. The ''O/"O ratlos (1894-2684 ppm) were determlned to be linearly proportional to the true values wlth a mean preclslon of 0.6 ppm and a mean accuracy of 2.7 f 2.9 ppm. The ablllty of the system to measure absolute ratios of 'H/'H and '*O/"O accurately and precisely In aqueous samples at hlgh enrlchrnent levels of 2H and "0 suggests that this system wlll be Important In nutrltlonal and cllnlcal studles employlng 2H,1'0.
-
The measurement of 2H/1H and lsO/leO isotope ratios is done customarily in two separate mass spectrometers following the conversion of water to H2gas for 2H/1H values (1,2) and equilibration of the water oxygen isotopes with carbon dioxide values (2). Each type of gas to form C1eO1sOfor 1s0/160 sample is introduced into a dual inlet isotope ratio mass spectrometer and its isotopic abundance is compared to that of a reference gas. Because the inlet pressures can be adjusted independently for both standard and sample gases, the ionization source pressure conditions and major ion beam intensities can be matched closely to minimize the measurement errors involved in the comparison. The requirement for separate sample preparation and mass spectrometric analysis for 2H and lSO in a single sample is time-consuming, but until the need arose to analyze hydrological samples in large numbers, there was little incentive to develop an instrument to carry out the facile and rapid measurement of both isotopes in the same water sample. In order to eliminate the sample preparation step for the oxygen isotope measurement, Majzoub and Nief (3) developed the
first mass spectrometer in which the determination of lsO in water samples was carried out directly on water vapor. Subsequently, Hagemann and Lohez (4) described a twin mass spectrometer system in 1978 in which a furnace, for conversion of water to hydrogen gas, was added for the simultaneous determination of 2H/1H and ls0/160in water samples. To date, however, water analyses with this instrument reportedly have been limited to those having natural abundances of 2H and/or lSO. The design used by Hagemann and Lohez, which is the basis of the Aqua-SIRA instrument, has several radical elements which include a pronounced and significant memory of previous samples, the use of dynamic (vs. static) isotope ratio determination techniques, and, most importantly, the direct determination of absolute abundances in the place of differential measurements with reference standards. The latter was of particular concern because analysis of samples significantly enriched in 2Hand lSOwould be required, and the linearity of response at levels 2- and 3-fold above natural abundance had not been established. Moreover, it would be necessary to inject biological samples of salt-containing fluids, such as urine, directly into the inlet system and the effects of sample matrix on the isotope ratio measurements was unknown and could be resolved only after an extensive evaluation had been undertaken. The description of the Aqua-SIRA instrument system, its characteristics and mode of operation, its performance, and the limits of its accuracy and precision are the subjects of this paper, Although this paper is limited necessarily to a specific commercial product, the question of interest is the following: Are the analytical results obtained from this type of mass spectrometer system adequate to support biomedical and clinical studies with 2H and lSO? On the basis of the performance levels obtained in the use of this system, a positive answer is justified. EXPERIMENTAL SECTION Physical Characteristics. Aqua-SIRA is a mass spectrometer system (VG Isogas, Limited, Cheshire, England) designed for
0003-2700/84/0356-1852$01.50/00 1984 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 56, NO. 11, SEPTEMBER 1984
F] rotary
Figure 1. Schematic diagram of the Aqua-SIRA system.
direct simultaneous determination of ,H and l80abundances in aqueous samples. The system consists of two mass spectrometers and a common inlet system. Mass Spectrometer. Deuterium abundance is determined on a 6 cm radius 90" sector analyzer with a permanent magnet in which H2H+and Hz+ions are focused onto a double collector designed for masses 3 and 2. Oxygen abundance is analyzed in a second 6 cm radius 90"sector analyzer with a permanent magnet of higher field strength in which Ha60+ (mass 18) and H2l80+ (mass 20) are focused onto a special double collector. The source block of the oxygen analyzer is made of copper and the entire ion source assembly is heated and maintained at 150 "C. The analyzer system is pumped with two oil diffusion pumps (Edwards, E02) at 170 L/s and a back-up oil diffusion pump (Edwards, E01) at 8 L/s followed by a rotary pump (Edwards E2M5) at 113 L/min. For deuterium analysis, the filament current, trap current, electron voltage, and ion accelerating potential are set at 2.8 mA, 0.2 mA, 80 eV, and 3.2 kV, respectively. For oxygen analysis, the filament current, trap current, electron voltage, and ion accelerating potential are set at 3.6 mA, 0.2 mA, 68 eV, and 3.8 kV, respectively. The input resistors for the major and minor collectors on both analyzers have values of 1 X lo9 and 1 X 10" Q, respectively. All the DC amplifiers are of Analog Devices, type 3105. Inlet System. The inlet system, shown in Figure 1,consists of two expansion chambers, B1 (Pyrex glass, 250 mL) and B2 (copper, 50 mL); four pneumatically operated bellow valves, V1, V2, V3, and V4 (Nupro "BK" valves); two copper capillaries, C1 (0.2 mm i.d. X 2 cm long) and C2 (0.2 mm i.d. X 12 cm long); a uranium furnace; a pirani gauge; and a rotary pump (Edwards, E2M5). With the exception of the pirani gauge, rotary pump, and uranium furnace, the entire inlet system is heated and maintained at 150 "C. The uranium furnace is heated and maintained at 620 "C. Principle of Isotope Ratio Measurement. With the exception of ion source tuning and sample injection, the operation sequence and the data acquisition and manipulation of the Aqua-SIRA are controlled by a Hewlett-Packard HP85 computer. The ion source focusing parameters of the two mass spectrometers are stored on tape by the HP85 computer and are retrieved from the tape to setup the instrument after routine shut down. Operation Sequence. With V1, V2, V3, and V4 closed, 0.5 p L of an aqueous sample is injected into B1 through the septum. A Hamilton 1-pL syringe with a Chaney adapter and stop (7001-NCH) is used to ensure injection of a constant sample volume. After the aqueous sample has vaporized in B1 for 10 s, V4 is opened to allow the water vapor to enter B2. Part of the water vapor travels through C1, enters the uranium furnace, and is converted to hydrogen gas. The hydrogen gas enters the hydrogen ion source and H2+and H2H+ions are formed. At the same time, another portion of the water vapor moves through C2, enters the heated water ion source, and is ionized to form HZ16O+ and Hz"O+ ions. Following a 45-5 equilibration period, the ion beam intensities are measured simultaneously for 20 s. A second series of ion beam intensities then is determined at a reduced source pressure. In this automated sequence, V4 is closed and a portion of the sample is isolated in B1 prior to opening V2 and V3 to pump out the sample contained in B2 and in the lines between B2 and the two ion sources. At the end of 30 s, the water vapor in B1 is allowed to expand into B2 and into the two ana-
1853
lyzers by closing V2, V3, and opening V4. The four ion beam intensities again are measured at the end of the 45-5 equilibration period. Finally, VI, V2, and V3 are opened for 60 s to pump down the inlet system for the next injection. An aliquot of the sample is again injected into the inlet system until constant 'H2H/'Hz and HJ@O/H260ratios are obtained. Depending on the difference in isotopic abundance between the prior and current samples, the number of injections required to reach plateau values varies between 6 and 20, with an average of seven injections per sample. Quantitation of Isotopic Abundances. The absolute ,H/'H ratios of a sample are reported by the HP85 computer and 1@O/160 after a series of equations have been solved. The individual equations and their functions are as follows: First, the raw lH2H/'Hz and H;@O/Hi60 ratios from the initial measurements are calculated and reported according to the equations lH2H/'Hz ratio (R,) = (U, - Z , ) / ( U , - 21) X lo', pprn H 2 8 0 / H i 6 0 ratio (R,) = (U5- Z5)/(U4 - 2,)
X
lo6, ppm
(1)
(2)
where U,, U,,U4,and U5are the measured ion beam intensities and Z1, Z2,Z4,and Z5are the background ion beam intensities of masses 2,3,18, and 20, respectively. Following expansion of the sample and establishment of a lower source pressure, a new set of 'H2H/lHz and H2180/H260ratios are calculated according to the equations 'H2H/lHz ratio (R4) = (W, - Z,)/(Wl - 21) X lo6, ppm
(3)
H21sO/H~60 ratio (R5)= ( W5- Z5)/(W4 - Z4) X lo6, ppm (4) where W,, W,, W4, and W5are the new ion beam intensities of masses 2, 3, 18, and 20, respectively, and Z1, Z, Z4,and Z5are the background ion beam intensities defined earlier. Pressure gradients are used to interpolate the observed ratios to the ratios at preselected target intensities. The target intensities are the major ion beam intensities, Hz+( T I )and H260 (T,), chosen as reference points for all measurements of lH2H/'Hz and H2sO/ H2160ratios. The gradients are calculated as follows:
C1= (R, - R4)/(U1- W,), ppm/l
X
A
(5)
A Gz = (R, - R 5 ) / ( U 4- W4), ppm/l X (6) where GIand Gz are the pressure gradients for the interpolation of lH2H/'Hz and H;80/Hz160 ratios, respectively. The existence of pressure gradients and the need to apply a correction to the rneasurementa are due to the variation of isotope ratios with ion source pressures and to the formation of H3+(mass 3) and H2H160+in the ion sources. Since the formation of these ions is a linear function of inlet pressure, the contribution of H8/H2 to the observed 'H2H/lHz ratio and of H3170/Hz160and H~H160/Hz160 to the observed H21sO/H2160ratio at the preselected target intensities (T,,T,)can be interpolated from the pressure gradients (G1, G,) in the following manner: C1 = G1T1,ppm (7) c2 = G2T2,PPm (8) where C1is the mass 3 correction for lH2H/lH, ratio measurmenta and Czis the mass 20 correction for 1sO/160ratio measurements. These corrections are applied to the interpolated 'H2HJ1H, and H~180/H2160 ratios prior to their conversion to absolute abundances. At the same time, the use of the target intensites defines the reference points at which standards and samples are compared. Second, the factors used to convert the measured isotope ratios to absolute abundances are derived from the measured ion beam intensities of a "working" standard of known 2H/1H (RT,) and 180/160 (RT,) ratios as follows: RS1 = (US2 - Zz)/(USl - 2,) x lo6, ppm (9) RS2 = (US5 - z5)/(us4- Z,) x
lo6, ppm
(10)
where RS1 and RSz are the measured 'H2H/lHz and Hz180/HJ60 ratios of the "working" standard, respectively, US1, US,, US,, and US5 are the measured ion beam intensities, and Z1, Z, Z, and Z5are the background ion beam intensities of masses 2,3,18, and T,), 20, respectively. Interpolating to the target intensities (T,,
1854
ANALYTICAL CHEMISTRY, VOL. 56, NO. 11, SEPTEMBER 1984
the following 'H2H/'H2 (RS,') and Hz180/H21e0(RS,') ratios for the working standard are obtained: RS; = RS1+ - (US1 - 21)) (11) RS,' = RSz + G2(Tz- (US4- 2,)) (12) The conversion factors for 2H/1H (F,) and l80/la0(Fz) abundances are
F1 = R T l / ( R S i - C,) F2 = RTz/(RS,' - C2)
(13) (14)
Third, the absolute 2H/1H and 1sO/160 ratios of a sample are finally calculated in the following sequence: the measured ratios (Rl, Rz) are calculated based on the actual ion beam intensities of masses 2(U1- Z1), 3(Uz - 2 2 ) , 16(U4- Z4),and 18(U5- Z6); next, the ratios (R{, R,') at the target intensities (Tl, Tz)are obtained by interpolation according to the following equations: Rl'(1H2H/1H2) = R1+ Gi(T1- (UT - 2,))
(15)
R,'(Hz'sO/HzleO) = Rz + G2(T2- (U4- 2,))
(16)
and finally, the contributions of H3+,lH2H170+,and 2H,'60+ are taken out of these interpolated ratios and they are converted to their true ?-H/IH (TR1) and l80/l60(TR2) ratios as follows: TR1 = (R: - C1)Fi
(17)
TR2 = ( R l - CJF2 (18) Calibration Standards and Enriched Standards. The accuracy and precision with which the Aqua-SIRA system could measure the %/lH and l80/l6O ratios at natural abundance levels were evaluated by using two international water standards, VSMOW and SLAP, and two National Bureau of Standards water standards, NBS-1 and NBS-1A. These water standards are distributed by R. Gonfiantini at the International Atomic Energy Agency, Vienna, Austria. Enriched samples were prepared by adding known amounts of 2Hz0or Hz180 to a tap water of known isotopic composition. The isotopic composition of the tap water was determined by using a National Bureau of Standards water standard, NBS-1, as the reference. The Hz180 solution was purchased from Monsanto Research Gorp. (Miamisburg, OH). This solution had been normalized to natural deuterium levels and had an enrichment of 10.4 atom % leg. The 2Hz0solution, obtained from Argonne National Laboratory (Argonne, IL), had an enrichment of 99.55 atom % deuterium. Water standards were prepared as follows: ten with 2H/1H ratios which ranged from 164 to 696 ppm and eight with 180/1*0 ratios which ranged from 2082 to 2684 ppm. Minimization of Memory Effect. Since the magnitude of memory effect and the rate with which it disappeared were dependent on the difference between the initial isotopic content in the inlet system and that of the sample, the fraction of sample carry-over in the inlet system should have been a function of the number of injections. Assuming the time between injections was constant, this relationship could be expressed mathematically as 1 - (Ri - Ro)/ (Rf - BO)= f(0
(19)
where R, is the initial 'H2H/lH2 or H280/H260ratio in the inlet system or the final ratio of the last sample. Ri is the observed ratio following the Ith injection, and Rf is the fiial or plateau ratio of the sample. f(0is an exponential term of the form Ae-Bx,and the coefficient and exponent could be evaluted by using the "curve" fitting program of the HP85 computer on data obtained from the analysis of two samples with large differences in isotopic composition such as V-SMOW and SLAP. Assuming that this relationship was characteristic of this particular Aqua-SIRA system, it became possible to predict the final ratios of a sample Ro,Ri, and I were known. The values within a few injections iff(0, of the coefficient and exponential terms for f(0were obtained from three seta of data selected at random on different dates over a period of 1month. Since the preexponential and exponential constants which governed the decay of memory effect should have been applicable to both of the hydrogen and oxygen systems, their overall mean values w$re-used to test the accuracy of the extrapolation procedure. A series of urine samples previously analyzed by 6 to 14 injections provided the opportunity to test the
i
l 1 MIN
Figure 2. Peak shapes (A) for m l e 2 and 3 at a scanning rate of 0.4 kVlmin and (B) for m l e 18 and 20 at a scanning rate of 0.4 kV/min.
ability of the equation to extrapolate to final values of early samples. Specifically,the first three values were omitted and the extrapolated values and their standard deviations were calculated from the fourth, fifth, and sixth injection values. Statistics. Because water standards and urine samples with ratios were used to evaluate large differences in 2H/1Hand l80/le0 the precision and accuracy of the Aqua-SIRA system, the mean precision and mean accuracy for the determination of these ratios were calculated as follows: mean precision = (C(SD)2/N)1/2 mean accuracy = C(accuracy)/N where SD is the standard deviation for each set of ratio measurements and N is the total number of sets of water standards and/or urine samples used in the evaluation. Accuracy is defined as the difference between the measured ratio and the predicted ratio.
RESULTS Physical Characteristics. Mass Spectrometer. In the absence of a sample, the pressure in the analyzer systems was 8X mbar and rose to 1 X mbar with the injection of 0.5 p L of a sample into the system. Under the present ion focusing conditions and with a sample size of 0.5 pL, ion 2X 2.7 X and 5.6 X 10-l2 currents of 5.4 X A were obtained for masses, 2,3, 18, and 20, respectively. As shown in Figure 2, both the major peaks (ion masses 2 and 18) and minor peaks (ion masses 3 and 20) peaks were symmetrical and flattopped. Figure 3 shows tracings of the ion beam intensities of masses 2 and 18 during a pressure gradient measurement. Both major ion beam intensities rose rapidly after injection and reached plateau values within 45 s. Gradient Parameters. The target intensities for Hz+(T,) and H2lSO+( T z )in this system were set at 6 X and 2.5 X lo4 A, respectively. Under identical ion focusing conditions and over periods up to 5 days, the pressure gradients for the to their target interpolation of 'H2H/lHz and H2180/H2160 intensities were determined to have mean values of 5.5 f 0.1 A for G2. ppm/l X A for GI and 2.9 f 0.5 ppm/l X On a day-to-day basis, the pressure gradient G1 deviated less than 3% from the mean daily value; however, G2 fluctuated more; in one case the fluctuation was 14% from the mean daily value. Mass 3 and Mass 20 Corrections. By use of eq 7 and 8 and the mean values for G1and G,, the mass 3 and mass 20 corrections were calculated to be 33 ppm (CJ and 7.25 ppm (Cz), respectively. Since the ratios were always of the order of 340 ppm for C1 and 2050 ppm for Cz, the C1 correction amounted to 10% or less of R,' and the Cz correction amounted to less than 1% of R;. Converson Factors. The 'H/lH (RT,) and 1sO/160(RT2) ratios of NBS-1 were calculated to be 148.48 and 1989.4 ppm, respectively, using the average @H and 6 l s 0 values reported
ANALYTICAL CHEMISTRY, VOL. 56, NO. 11, SEPTEMBER 1984
1855
_-e
E
'
rnle=2 8.02
400-
380:
0
A
2
CLOSE v 4 )PEN V2.V3
a
E
*1
2
360340
-
320-
cu -8 2.13 x10 A
300-
z
rnle=2
a c
280260-
240
'
, . . . . , . . . . , . . . . ) . . . . ( . . . . ( . . . . I . . . . ,
5
0
10
20
15
25
30
1
35
INJECTION NUMBER
rn/e=lS 1.62 x 10-'A
:LOSE It, v2. 13. v 4 NJECT SAMPLE
I
OPEN V4
Figure 3. Tracings of ion beam intensities of mle 2 and m / e 18 during a pressure gradient determination. Sample size was 0.5 pL.
by Gonfantini (5). With NBS-1 as the working standard, the conversion factors were determined to have mean values of 0.4930 f 0.0253 for Fl and 0.9701 f 0.0093 for F2 over a 30-day period. Analysis of Accuracy and Precision. Memory Effect. The magnitude of the memory effect on isotope ratio measurements is demonstrated in Figure 4 in the alternation between two water standards, V-SMOW and SLAP. These water standards were chosen because of the differences in their 2H/1H and l e 0 / 1 6 0 ratios. V-SMOW has a 2H/1H ratio of 155.76 ppm and an 1 8 0 / 1 6 0 ratio of 2005.2 ppm as compared ratio of 1894.0 to a 2H/1Hratio of 89.02 ppm and an 180/160 ppm in SLAP (5). In spite of these large differences in isotopic composition (66.74 ppm for 2H and 111.2 ppm for leg),carry-over between samples diminished to less than 1% of the final 2H/1H and Hz1eO/H2160ratios after seven successive injections of the same samples into the system. All measurements reported below were made after the new equilibrium had been obtained and five to seven additional injections had been made. Instrumental Corrections. The linearity with which the Aqua-SIRA system can measure 2H/1H ratios in water at high abundance levels is shown in Table I. The system consistently underestimated the true abundance values. The amount of underestimation also increased progressively with increasing 2H/1H ratios. A linear regression analysis of the predicted and the measured ratios yielded a correlation coefficient of 1.0000 with an intercept of 13.42 ppm and a slope of 0.9121. Thus it was necessary to introduce a correction
R, = (R,
- 13.42)/0.9121
(20)
to the measured 2H/1H ratios, where R, is the corrected ratio and R, is the measured ratio. Because the deuterium concentration determined by NMR confiied the original values of 99.55% 2H20,the discrepancy was not due to preparation of the standards. A number of
INJECTION NUMBER Flgure 4. Memory effect on (A) 2H/'H ratios and (B) H2180/H2's0ratios as determined by V-SMOW and SLAP.
Table I. Linearity of Aqua-SIRA for Measurement of 2H/1H Ratios in Water Spiked with 2Hz0 predicted, PPm
164.99 241.75 316.06 460.37 695.71
2H/1H ratios measd f std dev, trialn PPm 1 2 1 2 1 2 1 2 1 2
163.80f 0.11 (6)b 163.97 f 0.23 (8) 233.77 f 0.09 (5) 233.78 f 0.09 (6) 300.85 0.39 (5) 302.47 f 0.14 (5) 433.20 f 0.18 (5) 434.35 f 0.35 (5) 648.50 f 0.42 (5) 647.10 rt 0.30 (5)
*
accuracy, PPm -1.19 -1.02 -7.98 -7.97 -15.21 -13.59 -27.17 -26.02 -47.21 -48.61
mean precision: 0.26 "The first set of data was collected on 6/7/83and the second set of data was collected on 6/23/83. bNumbers in parentheses represent total number of final ratios used to calculate the means and standard deviations. possibilities were considered for the origins of what appeared to be a previously unrecognized isotope effect in the determination of 2H/1H abundances. Refilling the quartz tube in the uranium furance with new uranium turnings had no effect on the measured 2H/1H ratios. Changing the filament or retuning the mass spectrometer by an adjustment of the position of the permanent magnet also had no effect on the measured 2H/1Hvalues. Because changes in pressure gradient during analysis can affect the interpolation as well as the H3+ correction, four water samples with 2H/1H abundance ratios of 150, 165,371, and 696 ppm were analyzed with a modified program in which a pressure gradient determination was performed for each measurement. The results indicated that
1856
ANALYTICAL CHEMISTRY, VOL. 56, NO. 11, SEPTEMBER 1984
Table 11. Linear Regression Analysis Parameters between the Theoretical and Observed 2H/1H Ratios under Different Conditions
conditions
Nn
no change new uranium turnings new filaments and retuning gradient for each measurmt increase uranium furnace temp from 620 to 700 "C mean std dev
intercept, ppm
slope
corr coeff
10 (19) 12.46 0.9155 10 (11) 13.55 0.9140
3 (4) 4 (4) 4 (4)
1.0000
13.34 0.9152 12.53 0.9176 14.25 0.9073
1.0000 1.0000 1.0000 1.0000
13.23 0.9139 0.75 0.0039
1.0000 0.0000
" N is the number of water standards used in the evaluation process. The numbers in parentheses represent the total number of analyses done on these water standards which have 2H/1H ratios ranging from 150 to 696 ppm.
Table 111. Linearity of Aqua-SIRA for Measurement of 180/1s0 Ratios in Water Spiked with H21s0 lsO/lEO ratios
predicted, PPm 2082.5 2133.2 2294.1 2486.6 2684.2
trial"
measd f std dev, PPm
2081.6 f 0.4 (12)b 2080.9 f 0.9 (5) 2133.2 f 0.9 (13) 2132.5 f 0.4 (5) 1 2295.5 f 0.1 (10) 2 2290.0 f 0.1 (10) 1 2482.9 f 0.9 (10) 2 2480.4 f 0.8 (5) 1 2678.2 f 0.7 (10) 2 2675.2 f 1.0 (10) mean precision: 0.7 1
2 1 2
accuracy, PPm -0.9 -1.6 0.0
-0.7 1.4 -4.1 -3.7 -6.2 -6.0 -9.0
"The first set of data was collected on 6/7/83 and the second set of data was collected on 6/23/83. bNumbers in parentheses represent total number of final ratios used in the calculation of means and standard deviations. the pressure gradients had remained constant during the analysis of the four samples. Incomplete conversion of water vapor to hydrogen gas at the uranium furnace was considered as a possible factor in the instrumental correction. An increase in the temperature of the furnace from 620 "C to 700 "C, however, had no effect on the 2H/1H ratios measured for the same four water samples. When the calibration values for the 2H/1H abundances were determined over the range of 150 to 696 ppm (Table 11),the five series had the same slope (0.9139 f 0.0039) and intercept (13.23 f 0.75 ppm) and each yielded a correlation coefficient of 1.0000. We concluded, therefore, that this fractionation phenomenon was genuine, reproducible, and consistent and was unlikely to affect the outcome of the determinations when corrections were introduced as described above. All 2H/1H ratios reported below were corrected accordingly. The linearity with which the Aqua-SIRA measured water at high enrichment levels of l80/l6O is displayed in Table 111. As shown, the measured values are linearly proportional to the true abundance values. A linear regression analysis of the predicted and the measured ratios yielded a correlation coefficient of 1.oooO with an intercept of 24.0 ppm and a slope of 0.9884. Within the limits of analysis, no instrumental correction was required for ls0/160 ratios as determined from water samples with l80/l6O ratios which ranged from 2083 ppm to 2684 ppm. Accuracy and Precision. The accuracy and precision with which the Aqua-SIRA instrument could measure water
Table IV. Accuracy and Precision of Aqua-SIRA for Measurement of aH/lH Ratios in Water Standards after the Instrumental Correction
predicted, PPm 89.02 (SLAP) 127.41 (NBS-1A) 148.48 (NBS-1) 164.99 184.19 241.75 265.83 316.06 370.83 460.37 545.03 624.64 695.71
2H/1H ratios measd mean, mean precision, accuracy, PPm PPm PPm
93.47 129.60 148.90 164.60 183.73 241.05 266.09 315.37 372.07 459.29 545.17 625.86 694.10
0.52 (1,13)' 1.70 (1,13) 0.15 (1,5) 0.19 ( 5 , 29) 0.12 (3, 16) 0.12 (3, 16) 0.19 (3, 16) 0.28 (3, 16) 0.33 (6, 32) 0.24 (3, 15) 0.32 (3, 17) 0.33 (3, 16) 0.38 (6, 37) 0.55c
4.45b 2.19 0.42 -0.39 -0.46 -0.70 0.26 -0.69 1.24 -1.08 0.14 1.22 -1.61 0.87d 0.67e
The first number in parentheses represents the number of times the same water standard was analyzed over a period of 11/2 months. The second number in parentheses represents the total number of final ratios used to calculate the measured mean ratio and its standard deviation. Not included in average. Precision of the mean = [E(mean precision)2/N]1/2where N is the total number of sets of mean standard deviation used in the calculation. dAccuracy of the mean. e Standard deviation. standards with 2H/1Hratios which ranged from 89 to 696 ppm after the instrumental correction (eq 20) are shown in Table IV. A mean precision of 0.55 ppm and an accuracy of 0.87 f 0.67 ppm were obtained. Even though the 2H/1H ratio of SLAP deviated significantly from the reported mean value of 89.02 ppm (5),the measured value of 93.47 ppm was within the range of values (86.24-94.37 ppm) recently reported in an interlaboratory calibration which involved 34 institutes (personal communication from R. Gonfiantini). The l80/l6O ratios of three water standards (SLAP, NBSlA, and NBS-1) were determined with the Aqua-SIRA system and were found to be 1894.0 f 0.5, 1955.5 f 0.8, and 1989.1 f 0.2 ppm, respectively. V-SMOW was used as the reference. These measured values compared well with those of the reported values of 1894.0 ppm for SLAP, 1956.5 ppm for NBS-lA, and 1989.4 ppm for NBS-1 (5). These results, together with those presented in Table 111,indicate that a mean precision of 0.6 ppm and accuracy of 2.7 f 2.9 ppm could be ratios using the obtained in the determination of 180/160 Aqua-SIRA system. A chronological record of accuracy and precision of the Aqua-SIRA system in the measurement of 2H/1H and ls0/l6O ratios in a water sample with natural abundances of 'H and l80is shown in Table V. Over the course of 11/2months, the system measured these ratios with a mean precision of 0.16 ppm for deuterium and 0.7 ppm for "0. Minimization of Memory Effects. The preexponential and exponential constants governing the decay of memory effect, as derived from three different sets of 2H/1Hand 180/160 data, are summarized in Table VI. The projected 2H/1H and 180/160 ratios, based on the overall mean values of the preexponential and exponential constants, were compared with those based on values obtained at equilibrium and are shown in Tables VI1 and VIII, respectively. Results in Table VI1 show that the extrapolation procedure for measuring 'H/'H ratios in urine samples had a mean precision of 0.40 ppm which was slightly higher than a mean precision of 0.26 ppm for the measured values. Assuming the mean measured %/lH ratios were the "true" values for these urine samples, a mean
ANALYTICAL CHEMISTRY, VOL. 56, NO. 11, SEPTEMBER 1984
1857
Table V. Chronological Record of Accuracy and Precision of Aqua-SIRA in Measurment of 2H/1Hand 1sO/160Ratios in a Water Sample with Natural Abundances of Deuterium and l80
date
NO
2H/1H ratios f std dev, ppm
1sO/160ratios f
1995.4 f 0.6 1995.2 f 0.5 1995.4 f 0.7 1995.2 f 1.8 1995.6 f 0.5 1994.9 f 0.6 1996.7 f 0.6 1995.5 f 0.3 1995.0 f 0.5 1995.1 f 0.4
6/07/83 6/14/83 6/15/83 6/16/83 6/21/83 6/22/83 6/23/83 6/24/83 6/28/83 7/15/83
10
10 6 6
150.28 f 0.17 151.79 f 0.13 150.65 f 0.19 151.04 f 0.07 150.97 f 0.21 151.13 f 0.35 151.05 f 0.10 150.69 f 0.08 151.07 f 0.09 150.95 f 0.15
7/21/83
5 5
150.93 f 0.12 150.66 0.08
1995.8 f 0.5 1995.5 f 0.1
150.93 0.16
1995.4 0.7
5
16 9 7 6 8
*
mean ratio mean precision
comments
std dev, ppm
new uranium
new uranium, new filaments, moved magnets gradient for each measurment furnace 700 O C
" N is the total number of final ratios used in the calculation. NBS-1 was used as the standard.
Table VI. Preexponential and Exponential Constants Governing the Decay of Memory Effect as Derived from Three Different Sets Each of 2H/1H and laO/lsO Datan 2H/1H
A
B
2 3
0.352 0.156 0.169
0.870 0.417 0.278
mean std dev
0.226 0.110
0.522 0.310
1
overall mean 0.091
f
A
B
2 3
0.084 0.191 0.134
0.232 0.464 0.346
mean std dev
0.136 0.054
0.347 0.116
180/160 1
std dev: preexponential constant, A = 0.181 f
exponential constant, B = 0.435 f 0.230 'Eauation: 1- (Ri - R,)/(R, - R,) = Ae-B'. Table VII. Comparison of 2H/1H Ratios Measured from a Set of Urine Samples Requiring an Average of 10 Consecutive Injections per Sample with Those Ratios Obtained by Extrapolation from the Fourth, Fifth, and Sixth Injections
sample 1
2 3 4
5
6 7 8
9
2H/1H ratios f std dev measured, extrapolated, PPm PPm
142.73 f 0.12 192.15 f 0.15 214.59 f 0.07 255.86 f 0.22 265.12 f 0.32 274.20 f 0.33 284.31 f 0.27 295.02 0.24 301.52 f 0.44
*
0.26b
142.85 f 0.16 192.81 f 0.27 214.91 f 0.14 255.52 f 0.33 265.16 f 0.23 273.70 f 0.79 284.50 f 0.36 295.12 f 0.49 301.66 f 0.38
Table VIII. Comparison of 1sO/160Ratios Measured from a Set of Urine Samples Requiring an Average of 10 Consecutive Injections per Sample with Those Ratios Obtained by Extrapolation from the Fourth, Fifth, and Sixth Injections
sample
180/160 ratios f std dev measured, extrapolated, PPm PPm
1993.7 f 0.3 2048.3 f 0.3 2084.6 f 0.6 2163.1 f 0.4 2182.4 f 0.4 2194.8 f 0.6 2212.7 f 0.4 2237.1 f 0.2 2250.5 f 0.5 0.4b
1993.9 f 0.4 2049.2 f 0.4 2084.8 f 0.3 2162.9 f 0.2 2183.1 f 0.1 2196.6 f 0.8 2213.0 f 0.6 2235.6 f 0.7 2250.6 f 0.1
accuracy," PPm 0.2 0.9 0.2 -0.2 0.7 1.8
0.3 -1.5 0.1
0.5b
0.3c 0.9d Accuracy = extrapolated ratio - measured ratio. Mean precision. Mean accuracy. Standard deviation.
accuracy: PPm 0.12 0.66 0.32 -0.34 0.04 -0.50
0.19 0.10 0.14
O.4Ob 0.08C 0.34d
OAccuracy = extrapolated ratio - measured ratio. Mean precision. Mean accuracy. dStandard deviation.
*
accuracy of 0.08 0.34 ppm was obtained by using the exratios trapolation procedure. Similarly, the predicted 180/1e0 shown in Table VI11 has a mean precision of 0.5 ppm and an accuracy of 0.3 0.9 ppm using the extrapolation procedure.
*
DISCUSSION Physical Characteristics. Due to the oxidative and corrosive nature of the water vapor and the volatile organic constituents of the urine samples, the tungsten filament of
the water ion source had a limited lifetime of approximately 20 working days. The septum of the glass expansion chamber, B1, is made of a silicone elastomer and has a lifetime of approximately 400 injections. Placement of the copper capillary C1 at the inlet of the uranium furnace prolongs the lifetime of the uranium turnings to 1500 injections or 250 samples and increases the ion beam intensities for H;80+ and H2lS0+obtained from a 0.5-pL sample by restricting the sample flow into the hydrogen mass spectrometer. Analysis of Accuracy and Precision. The Aqua-SIRA system uses a novel concept in isotope ratio measurements, namely, interpolation of measured ratios to preselected target intensites using pressure gradient terms. The accuracy and precision of the isotope abundances might be expected to depend heavily on the reproducibility of the pressure gradient values. As stated earlier, the pressure gradients were determined to be approximately 6 ppm for G1and 3 ppm for G2 The values for G1were highly reproducible (*3% of daily value) while values for Gz showed greater variation. By careful adjustment of the injection volume, as well as the position and the length of the capillaries preceding the ion sources, the actual major ion beam intensities obtained were consistently within 0.2 X l@A of the target intensites. Since the observed ratios were typically 340 ppm for RS1 and 2050 ppm for RS2,
1858
ANALYTICAL CHEMISTRY, VOL. 56, NO. 11, SEPTEMBER 1984
the product [G,X (Tl - (US1- Z,)] had a value of less than 1% of the observed ratio RS,, even if the pressure gradient G1had deviated by 50% from the mean value of 6 ppm/l x lo4. This effect was diminished further for RSz, which was 6-fold larger than RS,. An additional reduction in the effect of variations in G1and Gzoccurred during the conversion of the interpolated ratios to their true ratios by the conversion factors, F, and F,, which had values of 0.4930 and 0.9701, respectively. Thus, the correction terms in eq 11, 12, and 15-18 attenuated any analytical variance introduced by the pressure gradients. The consistent underestimation of the true 2H/1H ratios by the Aqua-SIRA system is intriguing. At the present time, however, we can only postulate that the underestimation is due to one or both of the following factors: (1) isotope fractionation in the hydrogen capillary inlet system, because lHZ0 has a molecular velocity 5.4% faster than 2H20,and (2) isotope fractionation at the ion source, because lH2H has a slightly smaller ionization cross section than lH’H. By placing a water reference between every four water samples and using only the fourth and fifth measurements from each sample, Hagemann and Lohez (4) were able to obtain a precision of 0.2%0for the determination of 62H and 6l80 values in ten water samples which had 62H and 6l80 values similar to those of the water reference, SMOW. According to the results in Table V, a water sample which had a 2Hand l80content similar to that of the water standard (NBS-1) could be measured by the present system with a precision of 0.16 ppm (l.l%o) and 0.7 ppm (0.4%0),respectively, over a period of one and one-half months. The precision and ratios accuracy for the determination of 2H/1H and 180/160 decreased when water samples with high enrichments of 2H and l80were included in the evaluation (Tables 111, IV). It is important to emphasize that in the present analysis procedure, a water reference was used only a t the beginning of the day for the determination of pressure gradients (G,,G,) and conversion factors (F,, F,). With more appropriate spacing of water references between samples, better precision apd accuracy for the determinations of 2H/1H and l80/l60 ratios in aqueous samples a t high enrichment levels should be possible. The use of eq 19 to calculate the limiting isotopic abundances in the presence of a memory effect has a significant impact on the accuracy and precision of the final ratios shown in Tables VI1 and VIII. In spite of the large variations in the values of the coefficient (A) and the exponential terms (B) shown in Table VI, the residual difference between predicted and limiting values was sufficiently low after three injections to begin collection of accurate data. The sample processing time, therefore, was reduced to approximately 20 min and a
total sample volume of 3 p L was required. These results demonstrated that the consistency of the memory effect (which appeared to be characteristic of the inlet system of the Aqua-SIRA instrument) could be exploited to predict the final ratios of aqueous samples with high accuracy and precision and to increase sample processing speed. As shown in Tables VI1 and VIII, eq 19 can be used to ratios in a set of urine samples predict the ,H/lH and l80/l6O with high accuracy and precision. Because the preexponential and exponential constants that govern the decay of memory effect for the ,H/lH ratio, as shown in Table VI, were derived from water standards, the potential for urine matrix effect on the memory and instrumental correction is negligible. Even though the number of injections per sample required to reach the equilibrium value is increased by the buildup of residue in the expansion chambers and transfer lines of the inlet system, our experience indicates that a periodic (6-month interval) cleaning of the chambers and transfer lines is sufficient to eliminate this problem. On the basis of the assessments described above, this instrument system can be expected to provide reliable, accurate and precise measurements of 2H/1H and l80/l6O abundances at highly enriched levels as well as at natural abundance levels. As such, it will provide a significant impetus to the use of these isotopes in nutritional and clinical studies (6). ACKNOWLEDGMENT The authors wish to thank MSD Isotopes, Inc., Montreal, Canada, for performing the NMR spectroscopic analysis on the 2Hz0solution, E. O’Brian Smith for his valuable statistical assistance, C. S. Irving for his helpful discussions, E. R. Klein for editorial review, and E. M. Jenkins and M. Boyd for preparation of the manuscript. Registry No. 2H, 7782-39-0; lH,12184-88-2;lSO, 14797-71-8; “0, 7782-44-7. LITERATURE CITED (1) Halllday, D.; Miller, A. G. Biomed. Mass. Spectrom. 1070, 4, 62-87. (2) Schoeller, D. A.; van Santen, E.; Peterson, D. W.; Dletz, W.; Jaspan, J.; Klein, P. D. Am. J. Clin. Nutr. 1080, 33, 2666-2693. (3) Majzoub. M.; Nief, G. Adv. Mass Spectrom. 1968, 4, 511-520. (4) Hagemann, R.; Lohez, P. Adw. Mass Spectrom. 1078, 7, 504-508. (5) Gonfiantlni, R. Nature (London)1078, 271, 534-538. (6) Klein, P. D.; James, W. P. T.; Wong, W. W.; Irving, C. S.;Murgatroyd, P.; Cabrera, M.; Dallasso, H.; Klein, E. R.; Nichols, B. L. Hum. Nutr.: Clin. Nutr. IS84,38C,95-106.
RECEIVED for review January 27,1984. Accepted February 23, 1984. This work was supported by the USDA/ARS Children’s Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine and Texas Children’s Hospital.
