Calibration of stable isotopic data: An enriched δ 18 O standard used for source gas mixing detection and correction

Transcription

1 Calibration of stable isotopic data: An enriched δ 18 O standard used for source gas mixing detection and correction 1 D. R. Ostermann and W. B. Curry Department of Geology and Geophysics Woods Hole Oceanographic Institution, Woods Hole, Massachusetts Abstract We present empirically based calibrations of our measurements made on a Finnigan MAT252 equipped with a Kiel Device to Vienna Pee Dee belemnite, using an enriched δ 18 O standard. Calibrations include corrections for biases caused by the differences in isotopic composition of carbonate standards measured on the two parallel extraction lines of the Kiel Device and for decreases in the isotopic difference between the reference and sample gas caused by mixing in the source. After correcting for these biases, the precision of 2200 NBS19 analyses ( µg )is ±0.07 for δ 18 Oand±0.03 for δ 13 C. We have shared our standard enriched in δ 18 Owith18 laboratories engaged in paleoceanographic research, producing the first large-scale interlaboratory calibrations for this community. Using correction procedures reported here, water mass reconstructions using data produced on multiple mass spectrometers may now be possible with a precision approaching the level necessary to reconstruct temperature-salinity and density variability in the deep ocean.

2 2 1. Introduction Mass spectrometer computer automation has enabled paleoceanographers to assemble large data sets from multiple laboratories. Such an automated mass spectrometer (Finnigan MAT252 with a Kiel Device) was purchased by the paleoceanography research group at the Woods Hole Oceanographic Institution (WHOI) in 1992 and entered routine use in March We have since analyzed over 59,000 unknown samples of CaCO 3 and over 12,000 weighed CaCO 3 standards. In addition to analyzing carbonate standards NBS18 and NBS19 as suggested by Coplen [1996], we also routinely analyze a benthic solitary coral with an oxygen isotopic composition similar to late Pleistocene benthic foraminifera ( 3 o / oo Vienna Pee Dee belemnite (VPDB)). After 5 years of analyzing various isotopic standards, we have detected several phenomena which affect precision and accuracy as a function of the size and isotopic composition of our samples. We have found clear evidence for a size-dependent difference in isotopic enrichment that we can document because we have measured the mass of each standard analyzed. We have gone to this effort because we routinely analyze samples that are very small; a typical sample consisting of one benthic foraminiferal test has a median size of 40 µg [eg. Sloweyand Curry, 1995; Curryand Oppo, 1997]. Given these small sample requirements, it was necessary to analyze standards in the same size range to document that we have converted our data to VPDB with accuracy and to know at what level isotopic differences in our data exceed the precision of our measurements. In this paper we detail our approach to detecting and evaluating these phenomena and report on an informal interlaboratory calibration which has included many of the mass spectrometer laboratories actively pursuing paleoceanographic research problems. 2. Methods 2.1. Analytical Procedure The Kiel Device we obtained in 1993 runs a carousel containing 46 reaction vessels attached to two parallel extraction lines (designated A and B). Carbonate is loaded into each individual reaction vessel with a spatula for powder or a brush for foraminifers. When loading foraminifers, we routinely add a few microliters of methanol to each vial to facilitate the transfer of the foraminifers to the vessel. Additionally, we slightly crush the foraminifers directly in each vial with a glass rod to assure a complete reaction. We do not vacuum roast sample or standard material. Each carousel is left in the Kiel Device oven at 70 C for at least 24 hours while the previous carousel is being run to evaporate water and methanol. Each reaction vessel is placed on the vacuum system for a 10 min reaction with 70 C, 100% phosphoric acid made using a modified Coplenetal., [1983] formula. Acid is dropped onto the carbonate sample through the acid drip valve (one for each line) which, when open, allows the vacuum in the reaction vessel to draw phosphoric acid through a capillary and onto the sample. The number of acid drops is controlled by the system software. The size of the acid drop is controlled by changing the gap between the end of the capillary and an electrode that determines when the drop has been released. Attached to each extraction line is a distillation system which consists of two traps in series, connected by appropriate isolation valves and tubing. Each trap is capable of attaining temperatures of between -190 C and +130 C. During the 10 min reaction time all evolved gases are trapped at liquid nitrogen temperatures (-190 C). When the reaction is complete, the noncondensible products are pumped away and the first trap is warmed to -100 C, releasing the CO 2 but trapping the water. The pressure of evolved CO 2 is measured and recorded, and the temperature of the second trap is lowered to -190 C. At the end of the CO 2 transfer from the first trap to the second, the second trap is warmed to room temperature and the gas is expanded if necessary before introduction into the mass spectrometer through a capillary. All traps are baked at +130 C after each reaction to remove contaminating water vapor. We typically run samples between 20 and 100 µg of carbonate, although smaller and larger sizes are possible. This translates to a 1-6 volt mass 44 signal in the mass spectrometer. Because of the small capacity of the mass spectrometer bellows the reference gas aliquot is automatically replaced from a 5 L tank after every seven samples. Our reference gas is made by room temperature acidification (using concentrated HCl) of planktic foraminifers collected from deep-sea sediment cores. Using this method, we produce a reference gas with an isotopic composition close to the average values of Quaternary planktic and benthic foraminifers.

3 NBS, AtlantisII, and Carrara Marble Standards We analyze three different standards in the A and B lines during each run: at the beginning, middle, and end of the carousel. These standards (B1, NBS19 or Carrara and AtlantisII, our in-house coral standard) span a δ 18 O range of 5.6 o / oo (-2.2to3.4 o / oo VPDB), typical of most planktic and benthic foraminiferal carbonates. In addition, we periodically run NBS18 to constrain our intralaboratory and interlaboratory calibration [Coplen, 1996]. In March 1994 we began processing various benthic solitary corals obtained from the WHOI dredge collection to produce an enriched δ 18 O daily working standard. One particular coral from dredge Atlantis ( N, W, 260 m) was chosen for further analysis. Iron and manganese coatings were removed from the coral using the citrate dithionite bicarbonate (CDB) leach as described by Ruttenberg [1992]. CDB reductively dissolves ferric iron and manganese oxyhydroxides, and some ferric oxides, and keeps the dissolved metals in solution via chelation. CDB is the second step in the sequential extraction scheme optimized by Ruttenberg [1992] to selectively dissolve different forms of phosphorus, including ferric iron-bound phosphorus, from sediments. The MgCl 2 washes which typically follow the CDB extraction were omitted since the objective of the present work was limited to removing the metal oxide coatings. This coral was further analyzed by X-Ray diffraction and was found to contain 100% calcite. The final step in preparing this material to become our AtlantisII daily working standard was to grind and sieve the coral to <150 µm to produce a small uniform grain size. Aliquots of this standard have been analyzed by 20 US and international mass spectrometers from 18 laboratories (Table 1). In March 1995 we began analyzing ground andsieved(<150 µm) Carrara Marble (originally collected by Stan Margolis and obtained from Howard Spero) to replace NBS19 as one of our daily working standards. At present, we continue to run NBS19 in every sixth carousel. 3. System Performance 3.1. A Line - B Line Differences The first evidence that there are biases in our system comes from the observation of differences in isotopic composition between carbonate standards measured on the two parallel extraction lines of the Kiel Device before correction to VPDB (Table 2). For all NBS19 standards >0.8 volts the difference between the A line and B line is o / oo for δ 18 O and o / oo for δ 13 C with the B line producing the more enriched values. As a result, we treat each extraction line separately for converting to VPDB, using only the standards analyzed on each line to make the conversion. We developed postprocessing programs on UNIX-based machines to deal with this and other statistical biases detailed below [Ostermann, 1999] Reference and Sample Gas Mixing in the Source There is a small but measurable decrease in isotopic difference between the reference gas and sample gas caused by reference and sample gas mixing in the source. We knew about gas mixing from the previous experience of K.C Lohmann (University of Michigan) and John Hayes (then at Indiana University, now at WHOI). We believe this mixing happens for two reasons: (1) The turbo pumping system is too small for the size of the source, and (2) flaws in the source fabrication can allow gases to be briefly trapped in interstitial volumes or sorbed on surfaces inside the source. Simply lengthening the delay before analyzing a gas is not a solution because the system is tuned to run very small samples, so longer delays waste too much sample gas. The reference and sample gas mixing in the source is a function of the size and the isotopic composition of the sample and the reference gas. Figure 1 shows the magnitude of reference and sample gas mixing as a function of sample size for 8120 standards run through 1997 (NBS18, NBS19, Carrara, B1, and AtlantisII). To understand these mixing phenomena, we first calculated the average δ 18 O value for each standard relative to the reference gas for large samples (those >5 volts). We initially assumed that at voltages >5 we would see no gas mixing in the source. We then normalized each standard using the following parameter: δ 18 O measured δ 18 O >5volts δ 18 O >5volts This normalized data is dimensionless and independent of standard composition; it quantifies the fractional decrease in isotopic composition between the reference gas and each individual standard analysis, assuming that the correct enrichments occur

4 4 when the largest samples are analyzed. The gas mixing relationship shows that the difference in δ 18 O composition is reduced for smaller samples by 4%, corresponding to a 0.04 o / oo reduction for each 1 o / oo difference in isotopic composition between the reference gas and the unknown gas. With no gas mixing the slope of the data should be zero. The slope and intercept of the linear best fit (0.008 and , respectively) allow us to correct for the isotopic mixing in the source based upon the sample size using the following equations: δ 18 O unmixed = δ 18 O measured (0.008SV) (1) values [Coplenet al., 1983: Stichler, 1995]. This disagreement, we believe, comes from our incorrect assumption of zero mixing for large samples >5 volts. After correcting for gas mixing in the source, our isotopic difference between NBS18 and NBS19 was still 4% smaller than the published values. This 4% difference, we assume, must be the minimum value for mixing in the source even for the largest samples we measure. When we accounted for this in our mixing equations (3) and (4), the measured differences between NBS18 and NBS19 became the same as published data. In addition, our measured values for AtlantisII are the same as the values measured by most other laboratories when we apply equations (3)and (4) (Table 3) to correct all data as follows: δ 13 C unmixed = δ 13 C measured (0.008SV) (2) δ 18 O correct = δ 18 O measured [(0.008SV) D+1.0] (3) where δ 18 O measured and δ 13 C measured are the raw enrichments from the mass spectrometer with respect to WHOI reference gas, δ 18 O unmixed and δ 13 C unmixed are the calculated values for an unmixed sample, and SV is the sample voltage. Specific A and B line corrections are then required to convert these data to VPDB. The isotopic difference between the reference and sample gas is always reduced for smaller samples, without regard to sample composition (i.e., sample isotopic compositions are always biased toward the reference gas by this process.) The isotopic mixing also affects both A and B line samples equally as both capillaries enter the changeover block assembly at the same place. Carbon and oxygen isotopic compositions are equally affected, although the range of δ 18 Ovalues is generally much greater, so the absolute value of any correction is often larger for δ 18 O. Unfortunately, this correction for gas mixing in the source (equations (1) and (2)) does not bring all isotopic standards analyzed in our laboratory into agreement with their published values (Table 3 and Figure 2) or with other laboratories which analyzed our standards. We had initially assumed that the mixing effect decreased to zero as the mass 44 beam increased to >5 volts (which equals 80 µg of carbonate in our system). However, we observed that not only were our AtlantisII δ 18 O values 0.21 o / oo depleted compared to other laboratories, but our NBS18 values were 1.05 o / oo enriched compared to published δ 13 C correct = δ 13 C measured [(0.008SV) D+1.0] (4) where D = , the effect of mixing in the largest samples and δ 18 O measured and δ 13 C measured are the raw enrichments from the mass spectrometer with respect to WHOI reference gas. The δ 18 O correct and δ 13 C correct are the calculated values for an unmixed sample, and SV is the sample voltage. Specific A and B line corrections are then required to convert these data to VPDB as shown in Tables 1, 3 and 4. For samples with isotopic compositions in the range of late Quaternary planktic and benthic foraminifera the bias introduced by this mixing is <0.15 o / oo for δ 18 O because we are using a reference gas that is close in oxygen isotopic composition to Holocene benthic foraminifera. However, the bias would become quite large (e.g., 0.5 o / oo bias in δ 18 O for a sample of - 10 o / oo ) were we to analyze samples of diagenetically altered limestones with very low δ 18 Oandδ 13 Cvalues Mixing Variability Following the recommendation of Coplen [1996], we have found that by analyzing an enriched δ 18 O standard such as AtlantisII, along with NBS18 and NBS19, gas mixing in the source can be detected and corrected. We analyze all of these standards on a regular basis to monitor the performance of

5 5 the system. By assuming that the isotopic difference between NBS18 and AtlantisII must be constant through time, any deviations from this constancy can be constrained and will lead to real improvements in our ability to distinquish between paleoceanographic signals and machine variability Precision and Sample Size After we correct all data run on the Finnigan MAT252 for the two different problems (gas mixing in the source using equations (3) and (4) and separate A and B line differences), we are in a position to show our best determination of mass spectrometer analytical performance (Table 4). After correcting for these biases, the precision of 2200 NBS19 analyses (all sizes from 10 to 300 µg )is± 0.07 for δ 18 Oand± 0.03 for δ 13 C. We are close to the theoretical efficiency of our mass spectrometer of ± 0.05 for δ 18 Oand± 0.03 for δ 13 C[MerrittandHayes, 1994, J. Hayes, personal communication, 1999]. For samples between 10 and 20 µg the precision is poorer, but still an acceptable ± 0.10 for δ 18 Oand± 0.04 for δ 13 C. It is clear that there is a decrease in precision as a function of size for all NBS19 data (Figure 3). 4. Interlaboratory Calibration Greatly improved interlaboratory precision is possible by analyzing several isotopic standards covering a large isotopic range. The δ 18 O range between NBS18 and NBS19 is 21 o / oo. As recommended by Coplen [1996] every laboratory reporting stable isotopic results should at least report their values of NBS18 and NBS19. We believe it is equally important to analyze an enriched δ 18 O standard to increase this range to >26 o / oo, which covers the entire range of isotopic observations reported in paleoceanographic research. This paper introduces the initial calibration and empirical use of such a standard. However, the values reported here are preliminary, as every laboratory participating in this study runs different standards on machines of various vintages and may or may not make various corrections to their isotopic data. All of the laboratories use different reference gases. There may also be differences in the software routines by which users covert from raw data to VPDB. Some laboratories bake and/or crush their standard material, which may be sieved to varying sizes. All these unknowns may singly or in combination contribute to the observations of Zahnand Mix [1991] that there may be as much as a o / oo difference in the δ 18 O isotopic composition from similar samples run on different mass spectrometers. There is a spread of 0.27 o / oo in the δ 18 O and 0.11 o / oo in the δ 13 C data from Table 1. Empirical intercalibration is, we believe, the simplest way to bring our results into agreement with other laboratories and is an objective means of correcting several biases in our own data. The corrections we use (equations (3) and (4)) affect the data presented in several previous publications, including Sloweyand Curry [1995]andCurryand Oppo [1997]. The effects are small (Table 3) for the planktic and benthic foraminiferal oxygen isotopic data (<0.1 o / oo and <0.2 o / oo, respectively) in those papers and neglible for the planktonic and benthic foraminiferal carbon isotopic data (<0.03 o / oo ). In all cases the corrected δ 18 O values are slightly more enriched than the published values. We believe that our data are now directly comparable to other stable isotopic laboratories at the quoted level of analytical precision and to a degree that would permit comparisons of benthic foraminiferal oxygen isotopic values unattainable before. Zahnand Mix [1991] used previously published δ 18 O benthic foraminiferal data to constrain the present-day and Last Glacial Maximum thermohaline circulation in the Pacific and Atlantic Oceans. Data from seven laboratories worldwide showed differences of only o / oo between Atlantic 2-4 km and >4 km waters and similar differences between Pacific and Atlantic deep water. These differences were near the interlaboratory variability they observed and so they were unable to ascribe the origin of the water mass gradients either to real oceanographic changes or simply to interlab machine variability. Using these new correction procedures, water mass reconstructions using data produced on multiple mass spectrometers may now be possible with a precision approaching the level necessary to reconstruct the thermohaline circulation. 5. Summary We have found that analyzing an enriched δ 18 O standard such as AtlantisII is an essential part of tracking machine-related changes that affect the stability of isotopic data produced on the Finnigan MAT252 at WHOI. By running AtlantisII, NBS19, and NBS18 standards which cover a δ 18 O range of 26 o / oo, mass spectrometer laboratories can be alerted to the extent of gas mixing in the source or other nonideal machine performance and moniter those those changes over time. The values for AtlantisII reported

6 6 in Table 1 are preliminary. The paleoceanographic mass spectrometer community needs to continue this intercalibration effort. In the future it will be most important for laboratories to agree on the isotopic difference between NBS18 and some positive CaCO 3 standard the community wishes to use. The analysis of two such standards will reduce interlaboratory variability, which can be as large as 0.3 o / oo for δ 18 O values in the range of Quaternary benthic foraminifera. Acknowledgments. We wish to thank WHOI, NSF, and ONR for the joint funding necessary to purchase this machine. We are forever indebted to Finnigan Engineer Peter Haubold who had our machine up and running to factory specifications within 2 weeks of the installation. Without many conversations with K. C Lohmann and John Hayes we would never have thought to check for gas mixing in the source. Early drafts of the manuscript were reviewed by John Hayes and Delia Oppo and, after submission, benefited greatly from reviews by L. Peterson and T. Coplen. We also wish to thank Jean Lynch-Stieglitz from LDEO (while a postdoctoral scholar at WHOI, she ran many samples on our machine) who developed her own set of equations using our uncorrected standard data. These equations, using a different set of assumptions, showed a similar gas mixing response. Thanks also to investigators around the world who ran AtlantisII and read early drafts of the paper: T. Bickert, H. Ehrlenkeuser, M. Gagan, W. Howard, L. Keigwin, L. Labeyrie, B. Linsley, K. C Lohmann, A. Mix, D. Murray, W. Patterson, M. Raymo, N. Shackleton, H. Spero, and J. Zachos. We wish to thank Marti Jeglinski for keeping the machine running so that we are able to analyze over 10,000 samples a year.

7 7 References Coplen, T. B., More uncertainty than necessary, Paleoceanography, 11 (4), , Coplen, T. B., C. Kendall, and J. Hopple, Comparison of stable isotopic reference samples. Nature, 302, , Curry, W. B., and D. W. Oppo, Synchronous, highfrequency oscillations in tropical sea surface temperatures and North Atlantic deep water production during the last glacial cycle, Paleoceanography, 12 (1), 1-14, Merritt, D. A., and J. M. Hayes, Factors controlling precision and accuracy in isotope-ratio-monitoring mass spectrometry, Anal. Chem., 66 (14), , Ostermann, D. R., Inter-laboratory mass spectrometer calibration programs, IGP PAGES/World Data Center- A for Paleoclimitology Contrib. Ser , ftp://ftp.ngdc.noaa.gov/paleo/contributions by author /ostermann2000, Natl. Geophysics. Data Cent., Boulder, CO., Ruttenberg, K. C. Development of a sequential extraction method for different forms of phosphorus in marine sediments, Limnol. Oceanogr., 37 (7), , Slowey, N. C., and W. B. Curry, Glacial-interglacial differences in circulation and carbon cycling within the upper western North Atlantic, Paleoceanography, 10 (4), , Stichler, W., Interlaboratory comparison of new materials for carbon and oxygen isotope ratio measurements, Rep. IAEA-TECHDOC-825, pp , Int. At. Energy Agency, Vienna, Zahn, R., and A. C. Mix, Benthic foraminiferal δ 18 Ointhe ocean s temperature-salinity-density field; constraints on ice age thermohaline circulation, Paleoceanography, 6 (1), 1-20, D. R. Ostermann and W. B. Curry, Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, MA (dostermann@whoi.edu;wcurry@whoi.edu) Received May 25, 1999; revised February 18, 2000; accepted February 22, This preprint was prepared with AGU s LATEX macros v5.01. File paleo-final-pp formatted April 4, 2000.

8 Figure 1. The effect of gas mixing in the source based on 8120 δ 18 O standards run between 1993 and All standards normalized to >5 volt data using equation (1) clearly show gas mixing in the source. The intercept would be zero with no gas mixing in the source. The slope and intercept are used in gas mixing equations (1) and (2) to correct the data using postprocessing programs. Figure 2. Published standard data plotted against measured standard data from Table 3. (top left) The δ 18 O data before and after applying gas mixing corrections (equations (1) and (2)). AtlantisII data were depleted by 0.21 o / oo, and NBS18 data were enriched by 1.05 o / oo. (bottom left) The δ 13 C data before and after applying gas mixing corrections (equations (1) and (2)) which are not greatly affected (<0.03 o / oo for late Quaternary planktic and benthic foraminifers) by gas mixing in the source. Note that the published δ 13 C scale is a third that of the published δ 18 O scale (top left). Triangles connected by the dashed line are measured data. Circles connected by the solid line are published data. The arrows in Figure 2 (top and bottom left) indicate our average reference gas composition. (top and bottom right) The anomaly relationship between published and measured standard data is specific to our mass spectrometer. Figure 3. Precision of 2200 NBS19 standard analyses for the A (solid circle) and B (open circle) lines based upon 10 µg weight bins: (top) is δ 18 O and (bottom) is δ 13 C. These data are for all analyses >0.8 volts and whose ratio of sample to standard voltage (balancing efficiency) are between 0.8 and 1.2, corrected using equations (3) and (4). 8

9 9 Table 1. AtlantisII Standard Data Produced by 20 Stable Isotope Mass Spectrometers Relative to VPDB Laboratory Machine Manufacturer and Model δ 13 C Std. δ 18 O Std. N Dev. Dev. Brown University VG Sira U. C. Santa Cruz VG Prism MIT VG Prism III Isocarb WHOI VG Prism Cambridge U. VG Prism Multiprep U. of Tasmania VG Optima Gilson Multiprep U. Albany SUNY VG Optima Gilson Multiprep U. South Carolina VG Optima WHOI VG 602E U.C. Davis VG Optima - Common acid bath CNRS-GIF Finnigan MAT251 Kiel-I (avg A and B)* CNRS-GIF Finnigan MAT251 Kiel-I (avg A and B)* CNRS-GIF Finnigan Delta Plus Kiel-II (avg A and B)* WHOI Finnigan MAT252 Kiel-I (avg A and B) Kiel University Finnigan MAT252 Kiel-I (avg A and B) U. Bremen Finnigan MAT251 Kiel-I (avg A and B) U. Michigan Finnigan MAT251 Kiel-I (avg A and B) Oregon State U. Finnigan MAT251 Fairbanks device Australian National U. Finnigan MAT251 Kiel-I (avg A and B) Syracuse University Finnigan MAT252 Kiel-III Average AtlantisII Reference gas composition, reaction temperature, standard baking or size fraction, and software version have not been reported. * After applying the CNRS mixing corrections. After applying mixing corrections (equations (3) and (4)). See text for explanation. After applying University of Michigan normalizations corrections.

10 10 Table 2. A and B line Standard Data Relative to WHOI Reference Gas, Corrected for 17 O and Gas Mixing in the Source, From a Block of Analyses Numbers through Num- Line ber δ 13 C δ 18 O NBS19 A ± ±0.07 B ± ±0.06 A-B B1 A ± ±0.06 B ± ±0.05 A-B AtlantisII A ± ±0.07 B ± ±0.05 A-B Carrara A ± ±0.09 B ± ±0.06 A-B NBS18 A ± ±0.12 B ± ±0.19 A-B Note the differences in the isotopic composition of the standards measured on the two parallel extration lines. A-B equals the A line value minus the B line value.

11 11 Table 3. Standard Data Before and After Various Gas Mixing Corrections to VPDB, Averaged for the A and B Lines Eq. (1) and (2) Eq. (3) and (4) Eq. (1) and (2) Eq. (3) and (4) Published Values Minus Published Minus Published Standard δ 13 C δ 18 O δ 13 C δ 18 O δ 13 C δ 18 O δ 13 C δ 18 O δ 13 C δ 18 O NBS NBS Carrara B AtlantisII Equations (1) and (2) and equations (3) and (4) corrected data come from a compilation of standard data converted to VPDB. Published values for AtlantisII come from the 20 machine average reported in Table 1. The published values for Carrara come from H. Spero (personal communication, 1998); All other published values are from Coplenetal., [1983] and Stichler [1995].

12 12 Table 4. Standard Data From a Block of Analyses Numbers through as in Table 2, Corrected to VPDB Using Equations (3) and (4) Line Number δ 13 C δ 18 O NBS19 A ± ±0.07 B ± ±0.06 B1 A ± ±0.06 B ± ±0.05 AtlantisII A ± ±0.07 B ± ±0.05 Carrara A ± ±0.09 B ± ±0.06 NBS18 A ± ±0.12 B ± ±0.19 NBS19 minus Table 1 AII Avg. A B NBS18 minus Published A B AtlantisII minus Table 1 AII Avg. A B These data are for all analyses >0.8 volts and whose ratio of sample to standard voltage (balancing efficiency) are between 0.8 and 1.2. Table 1 AII Avg. is the average value for AtlantisII from Table 1.