Anal. Chem. 1993, 65, 2125-2130
2125
Measurement of Iron Isotopes (54Fe,56Fe,57Fe,and 58Fe)in Submicrogram Quantities of Iron Paul R. Dixon,' Richard E. Perrin, Donald J. Rokop, Reinhold Maeck? David R. Janecky, and Joseph P. Banar Los Alamos National Laboratory, Los Alamos, New Mexico 87545
Iron isotopes offer an excellent opportunity for detailed tracing of natural processes because of the range of stable isotopes (54,56,57,58) and the importance of iron redox transitions in biochemical and inorganic processes. In this paper, we describe a significantly improved mass spectrometric technique for iron isotopic analysis that utilizes ion pulse counting detection and is calibrated with an absolute iron isotope abundance standard. All four stable iron isotopes are quantified in a single analysis to -0.1% precision and accuracy using a low-temperature(1200 "C)silica gel/boric acid technique on an improved source filament assembly composed of platinum and ceramic. Filament loads of 100-1000 ng can be utilized through implementation of an ultraclean work environment, improved chemistry, and a source designed specifically to minimize isobaric interference in the mass 50-60 region.
INTRODUCTION Stable isotope fractionation is a powerful indicator of igneous, metamorphic, geothermal, sedimentary, and biologic processes in terrestrial samples, as well as cosmological processes in extraterrestrial samples. Isotopes of oxygen, carbon, sulfur, and hydrogen have been extensively used in such studies.112 In addition to the isotopes used in such wellestablished studies, there are other major elements with several stable isotopes which may also be significantly fractionated, including iron. Several factors have led us to develop the techniques necessary to measure iron isotopic fractionation. Iron is ubiquitous in terrestrial systems (fourth most abundant element in the earth's crust) and is an important biochemical nutrient, and its transformations between redox states are used as an energy source or storage mechanism by many organisms.- While inorganic fractionation of terrestrial iron has not been observed, biologic fractionation is a much stronger possibility based on comparisons with other elements. Isotopic fractionation of selenium, an element with a larger percentage difference in isotopic mass than iron (10?6 vs 7 96 ) has been observed.s An important factor that has been identified with measurable isotope fractionation is the t Central Bureau for
Nuclear Measurementa, Geel, Belgium. (1) Hoefs, J. StableZsotope Geochemistry;Springer-Verlag: New York, 1973. (2) Valley, J. D.,Taylor, H. P., Jr., ONeil, J. R., Eds. Reviews in Mineralogy; Mineralogical Society of America: 1986; Vol. 16. (3) Farina, M.; Esquivel, D. M.; Barros, H. Nature 1990,343,256-258. (4) Mann, S.;Sparks, N.; Frankel, R. B.; Bazylinski, D. A.; Jannasch, H. W. Nature 1990,343,268-261. (5) Fassbinder, J. W.; Stanjek, H.; Vali, H. Nature 1990,343,261-263. 0003-2700/93/0385-2125$04.00/0
structural environment of the atoms. For example, sulfur isotopes in processes which involve sulfide are strongly fractionated, while those processes involving only sulfate molecules do not result in significant sulfur isotopic fractionation. Inorganic iron behaves similarly to sulfate, magnesium, and inorganic silicon (all chemical species that are strongly tied to oxygen) and may not be significantly isotopically fractionated. However, in biologic or organic systems the isotopes of sulfur, nitrogen, selenium, and silicon have observed 1-50% fractionations.1$2*wThisis particularly true for reactions involving redox transitions (e.g., Se and S). Our efforts have, therefore, been directed toward enhancing the analytical techniques for iron isotopic analysis so that an analytical survey of small isotopic differences (on the order of 1-10%) can be conducted to assess the existence and occurrence of iron isotope fractionation in a variety of materials (Dixon et al., in preparation and ref 9). Thermal ionization mass spectrometry and ultraclean chemical separation techniques have been the primary focus of our efforts for iron isotopic analysis. Previous work using similar techniques involved relatively large filament loads (10-50 reg) and higher temperature ionization (1350-1450 "C) to facilitate analysis with faraday cup detectors.10-1' Isobaric corrections for chromium and nickel have typically been relatively large (0.1-0.5% of the iron isotopic peakbatio). Furthermore, investigators have used double-spike techniques (e.g., "Fe and aFe) with internal normalization to obtain iron isotoperatios, leadingto a total loss of fractionation information about the isotopes used for spiking in nature. In this paper, we describe improved mass spectrometric analysis utilizing ion pulse counting detection and clean separation techniques for unspiked samples that allow measurement of all four stable iron isotopes, reduced filament loads, reduced (6) Rashid, K.; Krouse, H. R.; McCready, R. G. L. (1978) US.Geol. Survey Open-File Rept. 1978, No. 78-701, 347-348. (7) Ohmoto, H.;Rye, R. 0.Geochemistryofhydrothermal ore deposits, 2nd ed.; Wiley: New York, 1979; pp 509-567. (8) Douthitt, C. G. Geochim. Cosmochim. Acta 1992,46,1449-1458. (9) Dixon, P. R.;Janecky, D. J.; Perrin, R. E.; Rokop, D. J.; Unkefer, P. L.; Spall, W. D Proceedings of the 7th International Symposium on Water Rock Interaction, WRZ-7. 1992; Vol. 1, pp 915-918. (10) Maeck, R. Atomic Weight and Absolute Isotopic Composition of a Natural Iron Isotopic Reference Material. Ph.D. Thesis, University of Antwerpen, Belgium, 1992. (11) Volkening, J.; Papanatassiou, D. A. Astrophys. J. 1989,347,L43L46. (12) Gotz, A.; Heumann, K. G. Znt. J. Mass Spectrom. Zon Processes 1988,83, 319-330. (13)Davis, A. M.; Clayton, R. N.; Mayeda, T. K.; Brownlee, D. E. Proceedinas of the Lunar and Planetary Science ConferenceXXZZ. 1991: pp 281-2@2. . (14) Turnlund, J. R.;Keyes, W. R. J. Micronutr. Anal. 1990, 7,117145. (15) Garner, E. L.; Dunatan, L. A. The Twenty-Third Annwl Conference on Mass Spectrometry and Allied Topics, Houston, T X , 1975; 53-54. (16) Garner, E. L.;Dunstan, L. A. Adv. Mass. Spectrom. 1978, 7A, 481-485. (17) Valley, G. E.;Anderson, H. H. J. Am. Chem. SOC.1947,69,18711875. 0 1993 American Chemical Society
2128
ANALYTICAL CHEMISTRY, VOL. 05, NO. 15, AUGUST 1, 1993
isobaric interferences, and stable ionization efficiencies a t reduced temperatures. EXPERIMENTAL SECTION Instrumentation. An NBS-designed, 30.5 cm radius, 90° deflection single-sector thermal ionization mass spectrometer equipped with ion counting was used in this study. Details of the instrumentation have been published elsewhere.1"20 The detection system is linear and accurate to 0.06 % .21 Ion count rates are corrected using a 10-ns dead time. The dead time was determined using certifieduranium standards (IsotopicReference Material IRM 072/1-15, available from the Central Bureau for Nuclear Measurements, B-2440, Geel, Belgium) using the procedure detailed in Efurd.22 The counting times used for an iron isotope analysis are 15 s for masses 52 (Cr), 54 (Fe, Cr), 56 (Fe), 57 (Fe), and 58 (Fe, Ni) and background. Mass 60 (Ni), a low abundance peak (1-5 counts/s) is counted for 20 s. An individual data set consists of the average of five ratio measurement sets, while the final ratio is the mean of four data sets. Reagents and Materials. Aqueous solutions are prepared from Teflon subboiling distilled l&MQ (Milli-Q)water. Ultrapure HCl and HNOs are obtained from Seastar (Sidney, BC, Canada). High-purity H2Oz is obtained from E.M. Reagents, and glass distilled acetone is from J. T. Baker. Gold-label boric acid of 99.999% purity is obtained from AESAR. The anionexchange resin used was AG l-X8, 100-200 mesh, obtained from Bio-Rad. Silica gel (Aerosil-300) is used as an ionization enhancer on the filament. It is cleaned prior to use by the following procedure. Stock silica gel (10 g) is weighted into a 50-mL Teflon centrifuge tube with 10 mL of 7 M HNOs. The mixture is gently shaken for 10 min and then centrifuged, and the acid decanted off. This step is repeated three times. The above procedure was repeated for a Milli-Q water rinse, a 1.0 M HBr wash, a Milli-Q water rinse, a 8 M HCl wash, and a final Milli-Q water rinse. The silica gel is then diluted with Milli-Q water to an approximate concentration of 5 mg/L. The exact concentration is determined by measuring the density of the final silica gel solution. Microcapillary Teflon tubing is used for loading samples onto the filaments. This tubing must be cleaned to remove Ni and Fe residues on the surfaces of the tubing. The tubing is cleaned using a small peristaltic pump (Lamda pump, Instech Labs Inc.), to draw the cleaning solution through the microbore (15.2" i.d. X 43.2 mm 0.d.) tubing. Successive 12-18-h cleanings are done with 6 M HC1 and Milli-Q water, followed by class 10 air (4-6 h) for drying. This cleaning procedure eliminates all detectable Ni and Fe previously introduced by uncleaned tubing. As part of a collaboration between the Los Alamos National Laboratory (LANL) and the Central Bureau for Nuclear Measurements (CBNM), R. Maeck provided high-purity iron isotope spikes of MFe (99.8437%)and 6BFe(99.9013%),as well as several calibrated mixtures of the MFe and MFe isotope spikes. Using these standards and mixes, calibration of machine fractionation was accomplished. The absolute isotopic abundance of all four stable iron isotopes can be determined on any sample. Details of the enriched isotope purification, as well as preparation of the MFe/'Te isotope mixtures, are discussed in Maeck.lo Sample Processing. Natural samples, standards, reagents, and filaments are handled in a class 100 or better cleanroom environment. Reagents, samples, and standards contact rigorously cleaned Teflon ware. The cleaning procedures are an adaptation of those described by Zief and Mitchell%and include soaking the ware for 24-36 h in room temperature Micro Soap, ~~~
(18) Rokop, D. J.; Perrin, R. E.; Knobeloch, G. W.; Armijo, V. M.; Shields, W. R. Anal. Chem. 1982,54, 957. (19) Rokop, D. J.; Schroeder, N. C.; Wolfsburg, K. Anal. Chem. 1990, 62,1271-1274. (20) Perrin, R. E.; Knobeloch, G. W.; Armijo, V. M.; Efurd, D. W. Int. J. Mass Spectrom. Ion Phys. 1985, 64, 17-24. (21) Goldstein, S . J.; Murrell, M. T.; Williams,R. W. Phys. Rev. C. 1989,40, 2793-2795. (22)Efurd, D. W. Rocky Flats Pond Water Characterization and Treatment (LATO/EG&G-92-022) Quarterly Progress Report. Los Alamoe National Laboratory Report Number: LA-UR-92-1050; 1992. (23) Zeif,M.;Mitchell, J. W. ContamirurtionControlinTraceElement Analysis; Wiley and Sons, Inc.: New York, 1976.
a nonionic detergent from VWR. The ware is removed from the soap bath and thoroughly rinsed with Milli-Q water. This is followed by sequential treatments with aqua regia, 6 M HCl (Baker reagent grade), 8 M HNOa (Baker reagent grade), and Milli-Q water. Each treatment consists of soaking the ware 2436 h at 80 OC and rinsing with Milli-Q water. After the final water rinse, the lab ware is air dried in a class 10 hood and stored in cleaned LDPE containers. Iron in samples must be separated from alkali metals, chromium, nickel, and hydrocarbons before it can be run on a mass spectrometer. The initial cleaning of the iron isotope spikes, standards,and samples is described in Maeck.lo Further cleaning of the spikes, standards, and samples was accomplished at LANL using a modified Kraus and MooreU and Samuelson%rocedure that nearly eliminates alkalis, chromium, nickel, and hydrocarbons from the sample. Aliquots of the iron solutions (-300-400 pg of iron) were taken to dryness in a class 10 hood under an infrared heat lamp and redissolved in 200 pL of 6 M HCL. Before and after chemical purification, iron concentrations were determined spectrophotometrically with a Hach spectrophotometer using the ferrozine method. Anion resin (2.5 mL) is placed in a Bio-Rad 15-mL disposable polyethylene column. The anion resin is pretreated and cleaned using sequential 1.5-mLrinses of Milli-Q water, 6 M HC1,4 M HC1,0.5 M HC1, and Milli-Q water. After cleaning, the iron solution (200pL) is loaded onto the anion resin for purification. The iron is purified by sequential 2.25mL washes of 6 M HC1,4 M HC1, and 2.5 M HC1, followed by a 0.5-mL wash with 0.5 M HC1. Following the washes, the iron is eluted with 1.5 mL of 0.5 M HCl. The eluted iron sample is taken to dryness and redissolved in 6 M HCl and run through a second anion column using the procedures described above. The eluted iron from the second column (-95% recovery from complete procedure) is dried with an infrared heat lamp in a class 10 hood and redissolved in 1.5 M HCl to a concentration of 1 pg/mL for filament loading. Using the MFe spike, the chemistry blank from this procedure was determined to be less than 20 ng (Dixon et al., in preparation). The purification of iron from geologicand biologic samples also uses these procedures and is discussed in ref 9 and in Dixon et al., in preparation, and will not be detailed in this paper. Filament Preparation. A special source filament assembly is being used for the thermal ionization of iron isotopes. The filament material is zone-refined platinum ribbon (H. Cross, 99.999%) cut to size (0.025 X 0.76 X 17.5 mm). The cleaning of the fiiamenta is accomplished using the followingprocedure. First, sequential ultrasonic baths with distilled acetone and Milli-Q water are performed. The ribbon is then heated to 80 OC for 8-12 h in sequential baths of 3 M HNOs and Milli-Q water. Following the last water bath, the ribbon is placed in an ultrasonic bath with distilled acetone, the acetone is decanted, and the filaments are dried in a class 10 hood under an infrared heat lamp. The filament posts are high-purity platinum (H. Cross, 99.99%) cut to size (0.97 x 30 mm) and cleaned using the same procedure. The filament insulator material is a high-density ceramic from Coors (AD96) that has been drilled to hold the platinum posts. The insulator block was designed by J. Banar to fit a modified NBS filament holder. The ceramics are f i i t cleaned in sequential ultrasonic baths was distilled acetone and Milli-Q water. Then the ceramics are heated to 80 "C for 8-12 h in sequential baths of aqua regia, 3 M HNOs, and Milli-Q water. Following the last water bath, the ceramics are placed in an ultrasonic bath with distilled acetone, the acetone is decanted, and the ceramics are air dried under an infrared heat lamp in a class 10 hood. Once dry, the ceramics are baked at 400 OC in a quartz dish for 8 h in a vacuum oven at a pressure of -2.0 X 1o-B Torr. The cleaning procedures decribed for all the materials in the source filament assembly were designed to eliminate or substantially reduce potential hydrocarbon, alkali and alkaline earth element, and transition metal element contamination. Prior to use, the platinum posts are inserted into the ceramic insulator (24) Kraus, K. A.; Moore, G. E. Contrib. Oak Ridge Natl. Lab., Chem. Diu. 1953, 75, 1460-1462. (25) Samuelson, 0. Ion &change Separations in Analytical Chemistry; Wiley and Sons: New York, 1963.
ANALYTICAL CHEMISTRY, VOL. 65, NO. 15, AUGUST 1, 1993
Table I. Bakeout and Heating Patterns and Count Rate for a 1-pg Iron Load time (min) temp ("0 counts/s of S6Fe 0-4 900 0-200 (LN? in cold finger) 8 1000 100-500 10 1050 200-700 12 1100 1300-4000 32 0 0 (cold finger to 25 "C) 44 900 0-200 (LN,in cold finger) 48 1000 100-500 50 1050 200-700 52 1100 1300-4000 54 1150 4000-9000 56 1180 (0.25-1.0) x 105 >62 1200-1215 (1.0-5.0) X 105
and bent into a special design developed by Delmore (see ref. 26 and Figure 3c). Finally the filaments are molded and welded to the filamentposts, and another jig dimples the top of the filament to aid in sample loading. All of the jigs for making this special filament were designed by Banar. The assembled filaments are degassed at 3.9-4.2A ( 1300-1400"C) in an evacuated chamber (lo0OOO counts/s at 1200 "C), final source focusing is performed, as well as a scan from mass 51 to mass 61 (Figure 1). Typical base-line count rates range from 1-7 counts/s at mass 51 to 0-2 counts/s at mass 61 (Figure 1). It has been noted that these backgrounds can be increased 1 order of magnitude or more if water adsorption in the ion source chamber is not controlled by adequate dry nitrogen flushing and pumping. To reduce the pumping time and displace adsorbed water when the relative humidity is high, the source can is vented twice with dry N2 gas while pumping to high vacuum. Also shown on Figure 1 are peaks for the four iron isotopes (MFe, SGFe,67Fe,and 68Fe)as well as the peaks of the major isobaric elementsthat one has to contend with when doing an iron isotope analysis: chromium (masses 52,53,and 54)and nickel (masses 58,60,61,62,and 64). Isobaric corrections for chromium are made by monitoringthe 52 mass peak, and correctionsfor nickel are made by monitoring the 60 mass peak, assuming natural isotope abundances for chromium and nickel. Isobaric corrections for chromium are less than 0.03% and for nickel less than 0.05%. These isobaric corrections are a factor of 5-10 less than those previously reported by other researchers studying iron isotopes.1c14J7*27,28 The assumption that the isotopic compositions
2128
ANALYTICAL CHEMISTRY, VOL. 65, NO. 15, AUGUST 1, 1993
1*035
PosVlnsulator Material
I
1.030
SST/Glass
f
1.025 c
a L
I
51
52
I
I
1
54
I
I
I
S5
56
57
58
59
60
61
Mass
I 1 2 3 4 . 5 6 7 Number of Filaments After 40 Runs of Natural Fe (CBNM IRM 014) Flgure 2. Isotopic contaminationof the absolute 54Fe/56Feisotope ratios of mix I I I I O after more than 40 analyses of natural iron (CBNM IRM 014). Certified value of mix I11 from Maeck.lo 1.020 f
53
1
0
of Cr and Ni are nonfractionated may not be correct, but because the corrections are very small, even large isotopic fractionations of either element will not effect the measured iron isotope ratios. The peaks at masses 59.1,61.1,62.1,63.1,65.1, and 67.1 (not all peaks are shown) are thought to be CaF+,because the relative abundance of these peaks is the same as the natural abundance of calcium. The peaks off unit mass at 56.15, 57.15, and 58.15 are small organic peaks that remain after bakeout. By reducing the size of these organic peaks by a prerun bakeout, they are totally resolvable from the iron metal ions at masses 56,57, and 58 (see insert on Figure 1). Memory effects between samples of similar isotopic composition do not appear to be detected but have been seen between samples of very different isotopiccomposition. Consecutive 1-pg loadings of "Fe spike (99.8%) or MFe spike (99.9%) and natural isotopic abundance iron demonstrated that within the precision of the analysis there is no cross contamination. After more than 40 filament runs of near natural isotopic abundance iron, a (0.11.0%)shift in the isotopic composition of the 1:lmix was noted, and the precision of the analysis was not affected, making it hard to detect previous sample memory contamination (Figure 2). To eliminate potential errors due to memory, the source is cleaned after every 20 analyses, triplicate external filament runs are done on each sample, and a standard is run every third sample. Source Filament Design Evolution. Initial survey experiments on standard NBS filaments assemblies (stainless steel posts with glass insulators) with silica gel/ boric acid enhancer loads proved unacceptable for iron isotope analysis by pulse counting (Figure 3a). Very high backgrounds (60-250 cps) across the mass range of interest (51-61) were observed. This background was caused by scattering of alkali ions (mainly Ca and K) generated by volatilization from the glass insulators. A secondary, but also serious, problem was ionization of the stainless steel filament posts. Isobaric "Cr and "i, as well as Fe (see SFe peak in Figure 3a), were characteristic. Dramatic improvements resulted when the source filament assembly was changed to platinum posts in a high-density Coors ceramic insulator. Alkali scattering backgrounds were reduced 1-2 orders of magnitude, and Cr, Fe, and Ni backgrounds were reduced (Figure 3b,c).This design also showed that the only contaminant of concern in the high-purity platinum was trace amounts of chromium (see Cr peaks in Figure 3c at mass 52, 53, and 54). Over 90% of our chromium correction comes from surface contamination of the filament material during filament assembly. The initial NBS style filament design shown in Figure 3b-1 had several problems. It was very hard to precisely load the small
