Detecting small isotopic shifts in two-isotope elements using thermal

Accurate Measurement of Ruthenium Isotopes by Negative Thermal Ionization ... Determination of the Isotopic Compositions of Samarium and Gadolinium by...
0 downloads 0 Views 520KB Size
Anal. Chem. 1992, 64, 2216-2220

2216

Detecting Small Isotopic Shifts in Two-Isotope Elements Using Thermal Ionization Mass Spectrometry Jason Jiun-San Shent and Typhoon Lee'*$ Institute of Earth Sciences, Academia Sinica, Box 23-59, Taipei, Taiwan 10764, Republic of China

Chau-TingChangt Institute of Atomic and Molecular Sciences, Academia Sinica, Taipei, Taiwan 10764,Republic of China

Because tw+Wope elements have no extra ratio to normalize In order to correct the Instrumental mass fractlonatlon effect, it has been difficult lo measure lhelr isotoplc compositionwith high precislon. We propose a new technique to overcome lhls by analyzing monooxide Ions and normalizingto the oxygen Isotopic ratlos in order to correct the Instrumental mass fractionation effect. UslngLa0 as an example, we were able to show that mass fractionation causes both 1soLa1*O/1s8La180 and 1seLa180/1SgLa180 to vary almost llnearly with 1soLa180/ 1soLa180.By normalizing 1soLa180/1soLa180 to l80/l8O, we can detect the isotoplc shffl in 1s8La/1sgLa from both lSoLal8O/ 1seLa180 and 1s8La180/1SgLa180 with a 95 % confldenceInterval varles by of 10.08 % whlie the uncorrected 1soLa180/1s9La180 more than 1 2 % and 1s8La180/1SgLa180 varies by 10.2 %. The ablilty lo detect m a i l isoloplc shffls In two-Isotope elements WHI have many appllcatknsfor naturalisotopk varlalkn stuch and artificial tracer work.

INTRODUCTION In thermal ionization mass spectrometry, the mass fractionation effect is one of the most important factors affecting the precision of the isotope ratio data. For elements with more than three isotopes, one can normalize one ratio between one pair of isotopes to a standard value and correct the mass fractionation effect on other ratios. Nd and Sr are good examples of an effective use of this method. The resulting precision for the corrected ratios can be as good as 0.0015%. But with two-isotope elements, since there is only one ratio, it is not possible to apply the internal normalization method. The isotopic ratio can thus have large variations both within one mass spectrometer run and from run to run. Some workers have attempted to form high-mass molecular ions to minimize the mass fractionation effect. For example Spivack and Edmondl measured boron isotope ratios using Cs2B02+ions. Others have used an exhaustion method to overcome the problem of the mass fractionation effect, e.g. K2 and La.3 In this method one simply wears out the entire sample, integrates all the ions received, and then takes the ratio. Still others have tried to use a time factor in their correction models to correct the mass fractionation effect (e.g. K and Pb4s5). Furthermore some workers have used silica gel to diminish

* Corresponding author.

+ Also at Department of Chemistry, National Tsing-Hua University. t Also at Department of Geology and Department of Physics, National

Taiwan University. (1) Spivack, J.; Edmond, M. Anal. Chem. 1986,58, 31-35. (2) Eberhardt, A.; Delwiche, R.; Geiss, J. 2.Naturforsch. 1964, i 9 A , 736-740. (3) Makishima, A.; Shimizu, H.; Masuda, A. Mass Spectrom. 1987,35, 64-72. (4) Kanno, H. Bull. Chem. SOC.Jpn. 1979, 52 (8), 2299-2302. (5) Habfast, K. Int. J.Mass Spectrom. Ion Phys. 1983,51, 165-189.

the mass fractionation effect into their requirement.6J When long-lived artificial isotopes are available the double spike method can be used to resolve the same problem, e.g. U.8 None of these methods are entirely satisfactory. It is thus important to devise more analytical techniques to overcome the mass fractionation effect, especially for two-isotope elements. Many elements (e.g. light rare-earth elements) form monoxide ions with ionization efficienciesmuch higher than metal ions. For an element with only two isotopes, oxide ions may appear a t up to six different masses. This opens the possibility of normalizing the ratio between monoxide ions of the same metal isotope to that between the oxygen isotopes. For instance, M180+/M160+can be normalized to l80/l6O. In this paper we follow up on our earlier preliminary works and explore the use of this method for lanthanum in more detail. An interesting and related study was that of Makishima and NakamuralO who analyzed CeO+ ions to obtain a highprecision measurement of 13sCe/136Ce. These authors normalized 142Ce/136Ceto remove the mass fractionation effect while at the same time corrected the contribution of 1@Ce18O to 142Ce160 by determining the variable la0/l60from lWe18O/ 142Ce160 during the analysis. The determination of small isotopic shifts in two-isotope elements is potentially important to a variety of interesting problems. The impetus of our study came from the need in cosmochemistry and nuclear astrophysics to use elements such as Li, B, V, La, and Eul1-l3 as sensitive indicators of early solar system irradiation by energetic particles from the primeval sun. A more precisely determined La ratio would also contribute to the improvement of the 138La-138Ce radiometric dating method.14

EXPERIMENTAL SECTION La has only two natural isotopes at masses 138 and 139 amu with an 138/139 isotope ratio of about O.OOO9. Lanthanum monoxides appear at 154,155,156,and 157amu (Table I). Their relative intensities are roughly 2:2000:1:5 since 1aO/160is about 0.002 and 1 7 0 / 1 6 0 is about 0.OOO 37. The isotopic ratio of purified La was determined in our MAT2626 thermal ionization solid-source mass spectrometer. It ~~~~

~

(6) Tera, F.; Wasserburg, G. J. Anal. Chem. 1975,47,2214-2220. (7) Kelly, W. R.; Tera, F.; Wasserburg, G. J. Anal. Chen. 1978, 50, 1279-1286. (8) Chen, J. H.; Wasserburg, G. J. Anal. Chem. 1981,53, 2060-2067. (9) Shen, J. J.;Lee,T.; Chang, C. T. Chem. Ceol. 1988,70,26 (abstract). (10) Makishima, A.; Nakamura, E. Chem. Geol. Isotope Geosci. 1991, 94, 1-11. (11) Fowler, W. A.; Greenstein, J. L.; Hoyle, F. J. Geophys. 1962, 6, 148-220. ~~. (12) Balsiger, H.; Geiss, J.; Lipschultz, M. E. Earth Planet. Sci. Lett. 1969,6, 117-122. (13) Lee, T . Astrophys. J. 1978,224, 217-226. (14) Tanaka, T.; Masuda, A. Nature 1982,300, 515-518. ~~

0003-2700/92/0364-2216$03.00/0 0 1992 Amerlcan Chemical Society

ANALYTICAL CHEMISTRY, VOL. 64,

Table I. Contributing Species for Masses 154-157 mass (amu) sDecies 154 155 156 La0 138+ 16 138+ 17 138+8 ~~~

10

157

9

~~

CeO Sm

Pro BaO

136 + 18 138 + 16 154 138 + 16

139 + 16

139 + 17

138 + 17

138 + 18 140 + 16

139 + 18

8 7

140 + 17 141 + 16

e

NO. 19, OCTOBER 1, 1992 2217

4 0 0 0 mV

2 . 6 ~ 1 Cps 0 ~

f

t

1

h

>

E '

v

4

3

employs a single magnetic sector with an equivalent radius of curvature of 62 cm. It is equipped with a seven Faraday cup variable multicollector system, a secondary electron multiplier (SEM) system in the pulse-counting mode, and a COMPAQ3861 20e computer m i n g HT-BASIC. A highly sensitivegas pressure regulator (VARIAN Model 951-5106) is attached to the source chamber in order to leak oxygen into the mass spectrometer via a PFA tubing to the filament position. An oxygen partial pressure of 2 X lo-* mbar in addition to a base pressure of 6 X lo+ mbar was maintained during La analyses. Strips of zone-refined Re filament (from H. Cross) were outgassed at 4.5 A for 45 min at pressure between 10-8 and mbars prior to use. The possible interfering impurities BaO+, Lao+,CeO+,and Pro+ in outgassed filaments were found to be lessthan 2 counts/s in the SEM system under running conditions. The double-filament method was used. The sample was loaded on one Re filament which was then oxidizedby raising the current to 1.5 A in air for 1 min. The other Re filament facingthe sample evaporation filament was used for ionization. The ionization filament current was first raised to 3 A. The evaporation filament current was then raised to 1.2 A, held for 10 min, and then raised again gradually to about 1.6 A (about 1200 "C), at which point the l39La"3O+ current is usually 3 X lW1A. The signal then grows by a factor of 2 in 1 h, during which time isotopic ratios were measured. Signals at 154,155, and 157 amu were measured using Faraday cups in the static multicollecting mode. At the same time background and interference were monitored at 152 amu using the SEM ion-counting system. The baselines were measured at 0.5 m u below each mass. A typical run consisted of 15 blocks of 10 ratios. Each ratio was integrated for 16 a. An entire run took 60-70 min. For a typical sample size of 60 ng of La, this corresponds to a detected ions over loaded atoms ratio of around 3%.

Background signal from the tails of large nearby peaks is an important source of error especially for small peaks. Since the signal at 155 amu is more than lo00 times larger than that of 154 m u , the background of the 155 signal at 154 requires special attention. At our working vacuum of 2 x 10-8 mbars for the source chamber and 6 X lo+ mbars for the flight tube, the magnitude of the background signal at 154 amu is about 1ppm of the lS6LaOsignal (Figure 1). It means that the intensity of backgroundversus the peak intensity of the '"La0 signal is about 1.1% which is significant. Since the tail at 154 corresponds to about 4 X 10-17 A while the Faraday amplifier has a noise equivalent to 4 X A for 1-s integration, the noise was 10 times higher than the tail signal. The baseline for '"La0 was taken at 153.5 amu on the Faraday cup, integrating for 16 s before each block of data. The total baseline integration time for an entire run was thus 16(15) = 240 s resulting in a final noise/signal ratio of 10/(240)1/2= 65%. This is so imprecise that special effort must be made to better determine the tail at 154 amu. We have considered taking tail readings at both 153.7 and 154.3amu using the side Faraday cup, with which the 154.0 amu signal was measured and then taking the geometrical mean. However, even if we had integrated the tail for as long as we integrate the peak signal (2400 s = 16 s X 15 blocks X 10 cycles),we still end up with a precision of only about 20 % but would have to double the data acquisitiontime. This seemsrather inefficient. Instead, we used the axial multiplier to monitor the signal at 152 amu when the side Faraday cups were integrating La0 ions. The counts at 152 amu came from three sources, Sm, CeO, and the tail of 1MLa0.

1 1

0

153

147

154

148

15

149

156

157

150

158

151

159

152

153

154

15;

Mass (amu)

Flgure 1. La0 mass spectrum of rock standard (3-2 (solid line) and a pure metal standard (dotted line) obtained by using an Ion counter and a Faraday cup (inset). Note the baseline at 154 amu, whlch comes from the tail of 155 amu, is about 1% of the 154 amu slgnai. The sharpfeature at 153.3 amu is presumablycaused by scattering%a0 ions with an efficiency of Small Nd, Sm, and CeO peaks are present at a level of less than lo-' of that of Lao. No signal was detected at 153 amu, Indicating the absence of Be0 interference.

Spectrum scans before and after data acquisition showed that the proportion of these three components stayed constant to a precision of about IO%,judging from lreSm, "CeO, and tail readings on both sides of the 152 amu peak. We thus used a correction factor combining these three effects on the 154 amu signal by monitoring 152 counts. This approach is the most efficient since it requires no additional time. The total correction was about 2% with an uncertainty less than 0.4%. The disadvantages of this method is that the tail was not measured directly using the side Faraday cup where the '"La0 signal was taken. It cannot be completelyruled out that there are scattering features on the side cup, which do not appear on the axial multiplier. There are many possible interferences for our mass region, many of which are shown in Table I. In order to guard against artifact caused by interferences, we carefully scan the spectrum between 148 and 164 amu using the SEM ion-counting system both before and after data acquisition. When our normal La standard prepared from high-purity La metal from the Ames Laboratory (Iowa)was measured,we could not find any interfering signals except a negligibly small amount of lr2Ce160. In natural geological samples, small signals of NdO, Sm, CeO, and Eu were sometimes present (Figure 1). However, Sm and CeO contributions at 154 amu were less than 0.2%and have been monitored at 149,158, and especially 152 amu as described above. Because Ba is an abundant element, we have carefully examined signals at 153,152, and 151 amu to look for oxides of barium-135, -136, and -137. The upper limit for the oxide of 137Bawas 10 counts/s, implyingthat barium-138oxide could not have contributed more than 0.3% of the ions at mass 154. Since Pr is monoisotopic, it is virtually impossible to monitor Pro. No 141Pr ions were detected, and the upper limit was 5 counts/s, but it is difficult to convert this to a limit on Pro. However, we note that Pro contributes only to 157 amu; hence ita presence will affect the two methods of obtaining La isotope ratios described below in completely different manners. Therefore, the consistency between the two ways to obtain 1sLa/13eLaratios described below suggests that the P r o interference was insignificant.

2218

ANALYTICAL CHEMISTRY, VOL. 64, NO. 19, OCTOBER 1, 1992

finite number of ions would be about 0.014 % ,assuming Poisson statistics. The rms electronic noise for a block can be estimated from the measured rms noise current of 4 X 10-16 A (1-8 integrations) to be 0.06%. So the statistical limit of our uncertainty per block should be about 0.07 5%. The observed standard deviation for one block measurement is between 0.07 and 0.2 % , indicating that we are approaching the statistical limit for our instrument. So it is not important to consider mass fractionation effects within a data block which takes about 4 min to accumulate. However, in a 1-h run consisting of 15 blocks of data, if there were no mass fractionation, the error should have converged to become 34-fold smaller. The observed l a error for an entire run is between 0.05 and 0.22 % ,suggesting significant contribution from mass fractionation effects. So mass fractionation correction is essential in order to improve the precision. This point is abundantly demonstrated in Figure 2 which is a plot 1 of 6(157/154) versus 6(157/155) for block means from three -representative runs covering the extreme range of the I I I I I I I I I I 1 1 1 1 1 1 1 1 1 1 1 1 l 1 1 1 1 variation. First, we note that within each run these two ratios -20 -10 0 10 20 30 40 show large but well-correlatedvariations. Furthermore, there are even larger variations from run to run which follow a similar correlation line. T o better determine the correlation lines caused by mass Figure 2. Varlatlon of 6(13%a180/138La160) plotted against that of fractionation, we plot the grand means from all runs of our 6(13%a~80/~3@LaIBO) for three representatlveruns covering the extreme range of the varlatlons. Each polnt Is the mean of a block consistlng standard in Figure 3. Per mil fractional deviations (6) of of 10 ratios of 1 6 s integration. The typical 2a error bar is shown. The 157/154 and 154/155 are plotted against that of 157/155. The regression line from Flgure 3 Is plotted to show that variation during equations for these two regression lines are Y = 0.933 X one run follows approximately the same correlatlon line determined 0.004 13 (R= 0.999) and Y 0.0659X - 0.0111 (R = 0.908). from different runs. Note that, over a range of variation for 157/155 of 40460,157/ 154 also varies by about 40460, but 154/155 only changes by ,3"' ( d 4%. When these correlation lines are plotted in Figure 2, it is obvious that within each run the fractionation effects also seem to follow these two correlation lines. These correlations provide the means to correct for mass 30 fractionation in analogy to the internal normalization method used for multiisotope elements. Geometrically, this method is equivalent to sliding the block mean data along a line with the same slope as the correlation line to a fixed but arbitrary 20 157La0/155La0 ( =l80/l6O) value, for which we use Nier's value of 0.00204515 and obtain a corrected lS7LaO/lULa0and 1uLa0/155La0(=138La/139La).We apply this correction to 10 all our block means and then obtain the grand mean of the corrected 15TLaO/1"La0 and 1"La0/155La0 (=13*LaPLa)for each run. When these are plotted in Figure 4a,b for 16 analyses of standard, we find a 2a uncertainty per analysis of k0.8460 0 in 138LaPLa. In order to test our technique on natural samples, we have analyzed two standard rocks from the US.Geological Survey: -10 basalt BCR-1 and granite G-2. Their results are also plotted in Figure 4a,b. These analyses have precisionssimilar to those of our pure metal standard runs. Their corrected ratios agree with the pure metal standards within the uncertainty range -20 estimated from repeated runs of the metal standard. These -20 -10 0 10 20 30 analyses imply that our method for correcting instrumental fractionation can be applied to geologicalsamples. To further test our ability to resolve small isotopic shifts, we artificially Figure 3. (a, Bottom) per mli fractional varlatlon of 13sLa180/138La180 enriched our normal standards by adding a known amount plotted against that of 13%a180/13gLa160 for the grand means of 16 of '%La spike. The isotopic shift expected from gravimetry mass spectrometer runs of our La standard. A shalght line is fltted is 1.63460, which agrees exceedingly well with our measured by linear regression. The zero point for the X-axis Is the laO/leO ratlo (0.002 045) reported by Niern3 The zero polnt of the Y-axls is the shift of 1.74460. This enriched standard is also included in 139La180/136La160 ratio (2.2525) for the polnt on the regression line Figure 4. The agreement again demonstrates the validity of correspondlng to Nier's 180/160. (b, Top) slmilar plot except for our correction procedures. 6(138La160/138La180). The zero polnt for the Y-axls is 0.000 907 9. d

h

INTERNAL NORMALIZATION USING THE ls0/l6O RATIO For a 154La0beam of 5 X A and an integration time of 160 s for each block, the rms noise associated with the

DISCUSSION The oxygen isotope ratio 180/160 can be directly obtained from 157La0/155La0,since the contribution to l55LaO from (15)Nier, A. 0.Phys. Reu.1950,77, 789-793.

ANALYTICAL CHEMISTRY, VOL. 84, NO. 19. OCTOBER 1, 1992

(per mil)

2219

Table 11. Comparison of Published Oxygen Isotopic ComDositions source 180/'60 method Nier16 0.002 045 O+electron bombardment Wasserburgle 0.002 11 NdO+thermal ionization Makishima3 0.002 129 & 10 Pro+thermal ionization this study 0.002062 15 Lao+ thermal ionization ~~

0

1

-

-

I

4 -

o Standard 0 G-2 m BCR-1

-

Enriched

-

6 -

I

I

I I-o+ I I

-

-

*

l-c-i

-

L-o--

0.163%

c'

I

-

-

I

Table 111. Comparison of Published La Isotopic Compositions 1nghra"7 Whit4318

i

d

Makishima3

this study 0

157

L ~ o / ~ ~(Corrected) ~ L ~ o

6 ( 1 6 4 ~ a ~ / 1 5 5(per ~a~ mil) ) 0

I

-

-

Kn

-

$ E

10-

a,

12-

R Ld

F: 4

14 16

BCR-1

I

,

,I

-

-

w I-

-

w v+-

-

-

-

-

7-

-

w

I I

* I

I

I l l I 0.000908

I

0.000006

-

-

I i-b-

I

-

-

Enriched 0.163%

I

k w

20 -

22

H

J-3-

18

Standard

A2-i

-

3

0

CCnI

-

g

I G-2

I

KH

6 CI

r"

nn

4-

I

1

2 -

-

I

I

bo+

-

I I

I

I

I

I

0.000910

I

0.000012

154

La0/155La0 (Corrected)

Flgure 4. (a, Top) grand means of 13sLa180/138La180 for 18 runs of our La standard, each corrected for fractionation by normalizing to Nier'soxygen ratio usingthe correlation line from Figure 3a. The 95 % confidence interval per run (dotted lines) 1s slightly larger than *0.8%. (b) Similar plot for 1381ai80/139La180. In this figure, two rock standards from the U.S. Geological Survey (BCR-1, Q-2) and an artiflclally enriched standard are measured. These analyses have preclslons similar to those of our pure metal standard runs.

8.9 & 0.1 8.93 f 0.15 9.025 & 0.005 9.079 0.002

Errors are two standard deviations of the mean.

138La170is only 3.5 ppm. The average ratio of la0/160 is 0.002 062f O.OO0 015 (24over 12runsof purified La standard samples. The total range observed so far is from 0.002 013 to 0.002 100. Table I1 compares our oxygen isotope ratio with those of Nier,15 Wasserburg et al.,16 and Makishima et al.3 Our mean value for l80/l6Ois similar to Nier's but much lower than that of Wasserburg or Makishima, presumably due to relative mass fractionation. In principle 170/180can also be used for normalization. However, in this study we cannot reliably measure 139La170. This is because the interference from 140Ce160at 156 amu is serious for the small lS9Lal70signal. Since our corrected 138L4139La depends on the choice of the normalizingvalue of 157LaO/1"La0, it is difficultto obtain the absolute 138La/13QLa ratio from our study. Table I11 is a comparison between our La data and those of Inghram et al.,17 White et al.,18 and Makishima et aL3 The 138L4139La ratio from this study is 1% higher than the ratios reported previously. This could be a systematic difference because our measurement was relative. The correlated variations we observed are not in accord with the prediction of the simplest type of model in which L a 0 ions fractionate as a whole. In that case the effect should be proportional to the mass differences of various L a 0 ions. So we would expect the slopes to be 3/2 and -l/2 in Figure 3a,b, respectively, because of the 3 amu mass difference between 157La0and 154La0,etc. The slopes we observed were 0.933 and 0.066,respectively. We note that the sum of the two slopes is 1 for both the observed case and the case predicted by the simple L a 0 fractionation model. This is because there are only three species (i.e. La0 154, 155, and 157) yielding two independent ratios. Therefore, when there are two linear relations connecting the three possible ratios, there must be amathematical identity relating the two relationsto eliminate one degree of freedom. It is easy to show that the slopes of the two linear relations must add to one. This mathematical identity has nothing to do with the mechanism of mass fractionation. The fractionation mechanisms that produced the observed correlationsthus remain unclear a t the moment. It is also conceivablethat the correlation may be different for different instruments. Therefore, the correlation lines obtained here may not be applicable in general. The shallow slope of 0.066 for the b(1MLa0/166La0) versus 6(157La0/155La0) correlationindicatesthat La isotopes stayed relativelyunfractionated when L a 0 ions varied greatly. This (16)Wasserburg, G.J.;Jacobsen, S. B.; DePaolo, D. J.; McCulloch,M. T.; Wen, T. Geochim. Cosmochim. Acta 1981,45, 2311-2323. (17)Inghram, M.G.;Hayden, R.J.; Hess, D. C. Phys. Reo. 1947,72, 967-960. (18)White, F. A.; Collins, T. L.; Rourke, F. M. Phys. Rev. 1966,101, 1786-1791.

2220

ANALYTICAL CHEMISTRY, VOL. 64, NO. 19, OCTOBER 1, 1992

suggests that oxygen isotopes were pkobably responsible for most of the fractionation. Therefore, *38La/139La can be obtained h o s t directly from 154La0/155LaO without using the correlation to remove fractionation. Our correction procedure merely improves the precision from f2 to fO.8960. Nevertheless, it is useful to be able to detect a small shift in 138La/139Lafrom also the highly variable 157La0/1MLa0 using

our correction procedure. The agreement between the two approaches provides a consistency check for two-isotope elements, which is valuable for guarding against artifacts such as interferences*

R E C E ~ for D review December 31, 1991. Accepted julY io, 1992.