Anal. Chem. 2006, 78, 6109-6114
Automated Determination of Silicon Isotope Natural Abundance by the Acid Decomposition of Cesium Hexafluosilicate Mark A. Brzezinski,* Janice L. Jones, Charlotte P. Beucher, and Mark S. Demarest
Marine Science Institute and the Department of Ecology Evolution and Marine Biology, University of California, Santa Barbara, California 93106 Howard L. Berg
Department of Earth Science, University of California, Santa Barbara, California 93106
A procedure for the automated determination of isotopic abundances of silicon from biogenic and lithogenic particulate matter and from dissolved silicon in fresh or saltwaters is reported. Samples are purified using proven procedures through the reaction of Si with acidified ammonium molybdate, followed by precipitation with triethylamine and combustion of the precipitate to yield silicon dioxide. The silicon dioxide is converted to cesium hexafluosilicate by dissolution in hydrogen fluoride and the addition of cesium chloride. Isotopic analysis is accomplished by decomposing the cesium hexafluosilicate with concentrated sulfuric acid to generate silicon tetrafluoride gas. Silicon tetrafluoride is purified cryogenically and analyzed on a gas source isotope ratio mass spectrometer. Yields of silicon tetrafluoride are >99.5%. The procedure can be automated by modifying commercial inlet systems designed for carbonate analysis. The procedure is free of memory effects and isotopic biases. Reproducibility is (0.03-0.10‰ for a variety of natural and synthetic materials. As the second most abundant element in the Earth’s crust, silicon is a primary element in the formation of rocks, and it has become an essential element for many forms of life.1 There are three stable isotopes of silicon, with masses of 28 (92.229 atom %), 29 (4.683 atom %), and 30 (3.087 atom %).2 Natural variations in the relative abundance of these isotopes are proving to be useful for identifying the origins of rocks,3 weathering patterns in soils,4,5 and the movement of Si through the hydrosphere.6-9 Silicon isotopes have also provided insights into silicon processing in * Corresponding author. E-mail:
[email protected]. (1) Simpson, T. L., Volcani, B. E., Eds. Silicon and Siliceous Structures in Biological Systems; Springer-Verlag: New York, 1981. (2) Rosman, K. J. R.; Taylor, P. D. P. Pure Appl. Chem. 1998, 70, 217-236. (3) Ding, T.; Jiang, S.; Wan, D.; Li, Y.; Li, J.; Song, H.; Liu, Z.; Yao, X. Silicon Isotope Geochemistry; Geological Publilshing House: Beijing, 1996. (4) Ziegler, K.; Chadwick, O. A.; Brzezinski, M. A.; Kelly, E. F. Geochim. Cosmochim. Acta 2005, 69, 4597-4610. (5) Ziegler, K.; Chadwick, O. A.; White, A. F.; Brzezinski, M. A. Geology 2005, 33, 817-820; doi: 810.1130/G21707.21701. (6) Alleman, L. Y.; Cardinal, D.; Cocquyt, C.; Plisnier, P.-D.; Descy, J.-P.; Kimirei, I.; Sinyinza, D.; Andre´, L. J. Great Lakes Res. 2005, 31, 509-519. 10.1021/ac0606406 CCC: $33.50 Published on Web 07/29/2006
© 2006 American Chemical Society
higher plants10,11 and in marine protists, including radiolaria12 and diatoms.13,14 Variations in silicon isotope abundances in diatoms recovered from dated marine sediments are being explored as a proxy to link the marine silicon cycle to climate change in the past.15 However, making measurements of silicon isotope abundance remains difficult, and future progress would benefit from new analytical methods that allow for simple sample preparation and analysis. The most common method used to measure silicon isotopes is the high-temperature fluorination of samples in a BrF5 3 or F2 3,7,9,16-18 atmosphere to produce SiF gas, followed by analysis on 4 a gas source isotope ratio mass spectrometer (IRMS). Highly accurate determinations of Si isotopes have also been made by generating SiF4 by the thermal decomposition of BaSiF6.19,20 All of these methods measure silicon as SiF3+ ions at 85, 86, and 87 m/z on a gas source IRMS with little mass interference by other ions. Recently, secondary ion mass spectrometry has also been used to measure isotopes of Si as Si- ions within individual mineral grains.21 New methods employing multicollector inductively (7) De La Rocha, C. L.; Brzezinski, M. A.; DeNiro, M. J. Geochim. Cosmochim. Acta 2000, 64, 2467-2477. (8) Ding, T.; Wan, D.; Wang, C.; Zhang, F. Geochim. Cosmochim. Actaq 2004, 68, 205-216. (9) Varela, D. E.; Pride, C. J.; Brzezinski, M. A. Global Biogeochem. Cycles 2004, 18, GB1047, doi: 1010.1029/2003GB002140. (10) Opfergelt, S.; Cardinal, D.; Henriet, C.; Andre´, L.; Delvaux, B. J. Geochem. Explor. 2006, 88, 224-227. (11) Ding, T. P.; Ma, G. R.; Shui, M. X.; Wan, D. F.; Li, R. H. Chem. Geol. 2005, 218, 41-50. (12) Wu, S.; Ding, T.; Meng, X.; Bai, L. Chin. Sci. Bull. 1997, 42, 1462-1465. (13) De La Rocha, C. L.; Brzezinski, M. A.; DeNiro, M. J. Geochim. Cosmochim. Acta 1997, 61, 5051-5056. (14) Milligan, A. J.; Varela, D. E.; Brzezinski, M. A.; Morel, F. M. M. Limnol. Oceanogr. 2004, 49, 322-329. (15) Brzezinski, M. A.; Pride, C. J.; Franck, V. M.; Sigman, D. M.; Sarmiento, J. L.; Matsumoto, K.; Gruber, N.; Rau, G. H.; Coale, K. H. Geophys. Res. Lett. 2002, 29, 10.1029/2001GL014349. (16) Taylor, H. P. J.; Epstein, S. Earth Planet. Sci. Lett. 1970, 9, 208-210. (17) De La Rocha, C. L.; Brzezinski, M. A.; DeNiro, M. J. Anal. Chem. 1996, 68, 3746-3750. (18) Epstein, S.; Taylor, H. P. J. Proc. 2nd Lunar Sci. Conf. 1971, 2, 1421-1441. (19) Valkiers, S.; Rube, K.; Taylor, P.; T., D.; Inkret, M. Int. J. Mass Spectrom. 2005, 242, 321-323. (20) De Bievre, P.; Lenaers, G.; Murphy, T. J.; Peiser, H. S.; Valkiers, S. Metrologia 1995, 32, 103-110.
Analytical Chemistry, Vol. 78, No. 17, September 1, 2006 6109
Figure 1. Schematic of modified Kiel III. Numbers identify the location of individual valves. L1 and L2 refer to analytical processing lines 1 and 2. All other abbreviations are defined in the text.
coupled plasma mass spectrometry (MC-ICPMS) are also being developed.22 These instruments measure Si+ ions directly and hold great promise due to low sample size requirements and potentially high sample throughput, but they have the challenge of being sensitive to matrix effects, mass discrimination, and temporal drift in isotopic ratios. Here, we describe a new method for silicon isotope analysis in which SiF4 is generated by the acid decomposition of Cs2SiF6, followed by analysis using IRMS. The advantage of the method is that conversion of either solid or dissolved Si from natural samples to Cs2SiF6 is simple, and the required chemical procedures are known to not fractionate isotopes of silicon.20,23 The procedures do not introduce a detectable blank, and the method is free of significant isotopic biases. Precision is improved by over 200% as compared to the laser fluorination method17 and is comparable to that achieved by the dry plasma MC-ICPMS method.22 The acidification process and mass analysis can be automated by modifying commercially available inlet systems, resulting in relatively rapid and precise automated measurements without the need for the presence of an operator. We have implemented the method using a modified Kiel III “carbonate” device and a MAT 252 gas source IRMS manufactured by Thermo Electron Corporation, Waltham, MA. METHODS The method involves the conversion of solid silica or dissolved silicon to purified SiO2, followed by conversion of the silica to Cs2SiF6. Cs2SiF6 is then decomposed with 98% sulfuric acid in vacuo, generating SiF4 and HF gases. The SiF4 is purified cryogenically and introduced into a dual inlet gas source IRMS for mass analysis. The ratios of 30Si/28Si and 29Si/28Si are determined, and δ30Si values are calculated relative to the quartz standard NBS28 as
δ30Si ) {[(30Si/28Si)Sample/(30Si/28Si)Standard] -1} × 1000
Sample Preparation. Samples of biogenic silica, mineral silica or aqueous Si from either fresh or saltwaters are processed easily (21) Basile-Doelsch, I.; Meunier, J. D.; Parron, C. Nature 2005, 433, 399-402. (22) Cardinal, D.; Alleman, L. Y.; de Jong, J.; Ziegler, K.; Andre, L. J. Anal. At. Spectrom. 2003, 18, 213-218. (23) Lenaers, G. Ph.D. Dissertation, Universiteit Antwerpen, Antwerpen, 1990.
6110
Analytical Chemistry, Vol. 78, No. 17, September 1, 2006
to Cs2SiF6. Solid samples of high purity are converted to Cs2SiF6 by dissolution in 1 mL of 7.5 M HF, followed by the quantitative precipitation of H2SiF6 by the addition of 0.5 mL of 3 M CsCl. The procedure is known to not fractionate isotopes of Si20,23 and results in a high purity product. The Cs2SiF6 is isolated by centrifugation, and the supernatant is removed. The product is washed twice in 100% ethanol and dried at 60 °C for storage at room temperature. Solid samples that contain significant impurities are processed according to the procedures described by De La Rocha and colleagues.17 Briefly, the sample is dissolved in HF, and then the Si is precipitated from solution using a mixture of triethylamine, ammonium molybdate and hydrochloric acid.17 The precipitate is recovered by filtration through a 0.6-µm polycarbonate filter. While still on the filter, the precipitate is washed with a mixture of deionized distilled water and the molybdate/triethylamine reagent (60 mL reagent/100 mL H2O) to further remove impurities. The precipitate and filter are then combusted in a 1-mL platinum crucible to produce solid SiO2 (mainly cristobalite and trydimite).17 The Si in aqueous samples is recovered using this same chemistry, except that the initial step of dissolving the sample in HF is not required. Modification of a Kiel III “Carbonate” Device for Si Isotope Analysis. The Kiel III is designed to decompose carbonates with 100% phosphoric acid. Unfortunately, the use of phosphoric acid to generate SiF4 from Cs2SiF6 results in poor sample yields and inaccurate results (see below). Concentrated sulfuric acid produces excellent results, and the acid has a low vapor pressure, but it is incompatible with the metals used in the construction of the original acid dosing system. The dosing system was redesigned to create a metal-free flow path for the sulfuric acid. The Kiel III has two sample lines with independent acid dosing subassemblies that operate in parallel (Figure 1). Each dosing subassembly was fitted with its own acid reservoir consisting of a modified borosilicate glass chromatography column with Tefzel fittings attached to an all-tetrafluoroethylene (TFE) valve (Valcor Valve), which was in turn mounted on top of the acid delivery subassembly of the Kiel III. The headspace of the acid reservoir is continuously pumped by a rotary vacuum pump. This serves to remove dissolved gases from the acid, and the presence of a vacuum on both sides of the valve slows the flow of
acid during sample dosing. A 1-mm-o.d. (outside diameter) TFE tube conveys the acid from the Teflon valve to the sample vial. The flow of acid through the tube was reduced by reducing the size of the inner diameter of the tube to 0.3 mm. This was accomplished by inserting a 0.3-mm-o.d. wire into the end of the TFE tube, softening the end of the tube with a flame, compressing the softened tube around the wire with pliers and removing the wire after the tube cooled and hardened. The modified acid dosing system is controlled by a custom electronic timing circuit. In the unmodified Kiel III, the addition of acid to samples is initiated by the opening of a pinch valve that allows acid to flow through a flexible tube attached to a metal capillary leading to the sample vial. The amount of acid dispensed is determined by counting drops that bridge metal electrodes and complete an electrical circuit. In our modified system, the pinch valve is no longer used, but the electrical signal that would normally open that valve is used to trigger the custom electronic timing circuit to open the TFE valve, thereby dosing the sample with acid. A potentiometer is used to set the amount of time the valve is open after the trigger signal is received and is adjusted to allow 5-6 drops of acid (∼40-80 µL) to be dispensed. Upon closing the valve after dosing the sample with acid, a simulated drop count signal is supplied to the Kiel control circuitry to signal that the analysis can continue. Occasionally, a drop of acid adheres to the end of the TFE tube. This poses a potential problem should the drop fall and react prematurely with the subsequent sample. This is avoided by briefly attaching a sample vial that is partially filled with 3-mm TFE balls to the sample port when changing from one sample to the next. Drops of acid adhering to the tube are transferred to the balls upon contact. The method requires cold temperatures rather than the high temperatures that the Kiel III is designed to produce. The Kiel III oven was modified to maintain a temperature of 1 °C. The original oven heater was removed, and the oven walls were insulted with sheets of 5-cm-thick Styrofoam insulation. An external circulating chiller pumps a solution of water and antifreeze through a coil of copper tubing placed inside the former oven. A small fan circulates the internal atmosphere inside the chamber past the coil, and the temperature is maintained to within (0.2 °C. Because SiF4 is destroyed by water, precautions must be taken to avoid condensation of water vapor inside the chilled chamber. Nitrogen gas (>99% purity) with dew point of less than -70 °C from a Prism compressed air membrane nitrogen generator (Air Products and Chemicals, Inc.) is fed into the chamber at a flow rate of ∼1 L/min to create a dry, inert atmosphere at positive pressure inside the chamber. We also discovered that exposing the internal surfaces of the acid dosing housing to atmospheric gases while changing samples produced products that were difficult to evacuate and that interfered with the results. Exposure to atmosphere was reduced by filling the chamber with nitrogen gas as described above and by allowing the nitrogen gas purge that is used to detach vials from the sample port to flow continuously whenever a sample vial is not attached to the drip line. Evacuation of the sample vial was improved by the addition of a liquid nitrogen trap on the fore vacuum line. Analytical Procedure. A known mass of Cs2SiF6 (1.2-1.5 mg) is placed into each Kiel III borosilicate sample vial, covered with 50 µL of ethanol, and dried at 60 °C. This procedure adheres
the sample to the bottom of the vial and prevents loss of sample when the vial is attached to the sample port of the Kiel III under the flow of nitrogen gas. Samples in vials are then loaded into the sample carousel and dried at 150 °C for 2-3 h to drive off water vapor prior to placing the carousel into the Kiel device. The sample chamber is then purged with dry nitrogen gas for at least 24 h at room temperature. The chamber is then cooled to 1 °C, and the analysis is initiated. In the following description, we describe the analysis of a single sample on sample line number 1 (Figure 1). In a typical run, analyses alternate between sample lines 1 and 2. The system is initiated with an empty sample vial attached to position 1 of each dosing assembly. The vial on assembly 1 is detached by raising the sample lift piston, closing valve 7, and opening valves 8 and 13 to admit nitrogen gas into the sample vial, thus breaking vacuum and releasing the vial. The lift piston is lowered to remove the vial and drop it back into the sample carousel, with the nitrogen purge gas flowing continuously. The sample carousel is rotated to the position containing the vial with TFE balls. This vial is lifted onto the sample port, valves 8 and 13 are closed, and valve 7 is opened to establish vacuum. Valve 13 is opened to secure the sample against the viton O-ring seal of the sample port. The vial with the TFE beads is held in place for 10 s to remove any drops of acid adhering to the acid-delivery tube. Then the vial with the TFE balls is detached, and the first vial containing Cs2SiF6 is attached using the same valve sequence. The sample vial is evacuated by the rotary vacuum pump for 1 h, then valve 13 is closed, and valves 12, 1, 4, and 9 are opened, and the sample is pumped under high vacuum by a turbo molecular pump for 3 min. To further improve the vacuum and remove contaminating gases, trap 1 (T1) is cooled to -196 °C with liquid nitrogen, and the sample is pumped under high vacuum for 20 min. Valves 1 and 4 are then closed, T1 is warmed to 115 °C, and the pressure of the released gas is measured on vacuum gauge 2 (VM2). Valve 4 is then opened, and the gases are pumped away to high vacuum. This cycle is repeated with the time spent at -196 °C reduced to 5 min. The pressure of residual gases trapped during these procedures is logged automatically and is used for quality control. Just prior to the addition of acid, T1 is cooled to -196 °C with valve 1 open and valve 4 closed. Acid is dropped onto the sample, and the evolved SiF4, HF, and minor amounts of H2SO4 are frozen into T1. Then valve 4 is opened, and noncondensable gases are pumped away to high vacuum. Valves 1 and 4 are then closed, and valve 2 is opened. T1 is then warmed to -125 °C. At this temperature, HF and H2SO4 remain frozen, and SiF4 gas is released quantitatively. The pressure of SiF4 is read on VM2. If the pressure is sufficiently high to result in an off-scale voltage (>10 V) on the faraday cup detector of the mass spectrometer, the sample mass is reduced by closing valve 2 and evacuating the portion of the sample between valves 2 and 3. If further reduction in pressure is required, valve 3 is closed, and valve 2 is opened. Valve 2 is then closed, and valve 3 is opened to once again evacuate the gas from between valves 2 and 3. This process is repeated a sufficient number of times to achieve a pressure in VM2 that has been empirically determined to produce a sample voltage between 3 and 7.5 V. The sample is then transferred to trap 2 (T2) by Analytical Chemistry, Vol. 78, No. 17, September 1, 2006
6111
cooling T2 to -196 °C while maintaining T1 at -125 °C and opening valve 3. The SiF4 is transferred from T1 to T2 for 1.5 min with the sample side of the changeover valve (valve 32, Figure 1) closed and for an additional 1.5 min with valve 32 open. T2 is warmed to 35 °C, releasing the SiF4 into the capillary leading directly to the changeover valve on the mass spectrometer. The 87/85 and 86/85 isotope ratios are read on the MAT 252 against commercial SiF4 that has been calibrated against NBS28.9 When the analysis is complete (typically 16 cycles), the remaining sample is pumped to waste by opening valve 5, and both traps are cleaned by heating to 115 °C under high vacuum. RESULTS AND DISCUSSION Sample Yield. The Kiel III is not capable of determining sample yield accurately. Yields of SiF4 were determined on an all-metal (Monel) manual vacuum line against a calibrated capacitance manometer.17 Acid decomposition was achieved by placing a known mass of cesium hexafluosilicate in the bottom of a 0.25-in. Swagelok tee fitting with a sidearm of stainless steel tubing containing 100 µL of either concentrated phosphoric or sulfuric acid. Samples within these assemblies were attached to the vacuum line17 via a flexible corrugated stainless steel tube and VCR fitting and were pumped overnight by a turbo molecular pump prior to reaction. Then the tee fitting and the sidearm were immersed in a beaker of ice water for 10 min. SiF4 was generated by tilting the submerged tee fitting so that the acid flowed onto the sample. The resulting reaction was allowed to run for 15 min while collecting the evolved gases in a liquid nitrogen trap. SiF4 was separated from HF and H2SO4 cryogenically, the yield of SiF4 was measured by a capacitance monometer, and the sample gas was sealed into a borosilicate tube using the same system as described in ref 17. Sample yields were always >99.5% when Cs2SiF6 was decomposed with 98% sulfuric acid. Yields were 2.5 mg of SiO2 were processed. Samples of SiO2 from the Big Batch of 5 ( 1 and 10 ( 1 mg resulted in δ30Si values of -7.6 ( 0.2‰ and -6.8 ( 0.3‰, respectively. This bias can be avoided by adjusting reagent volumes to keep the concentration of dissolved silicon 1 mg Cs2SiF6 yield consistent results (Figure 2), but samples of 5 °C become highly variable and are generally biased to be more positive than expected (Figure 2), suggesting that higher tem-
peratures promote reactions that preferentially destroy 28SiF4 molecules. The temperature effect bears certain similarities to the sample mass effect described above in that a higher temperature results in generally more positive values and increased variability. Analytical Chemistry, Vol. 78, No. 17, September 1, 2006
6113
Table 3. Test for Memory Effect consecutive sample 1 2 3 4 5
δ30Si Big Batch
δ30Si Diatomite
Trial 1 -10.38 +1.40 -10.46 -10.33 +1.31 Trial 2
1 2 3 4 5 mean SD
+1.36
-10.45 +1.37 -10.37 +1.32 -10.40 0.05
+1.35 0.04
The acid decomposition method is very versatile in that a variety of materials are analyzed easily, including dissolved silicon in seawater; freshwater and soil pore water; and solid silica, including biogenic and mineral materials (Table 2). The agreement between the δ30Si values obtained by acid decomposition and those obtained by laser fluorination are generally better than 0.1‰ for all materials tested (Table 2). Memory Effect. We tested for a memory effect by alternating analyses between samples of the Big Batch sodium metasilicate and diatomite. The isotopic values for these materials differ by 11‰, which exceeds the known range of natural samples from Earth (9.1‰24). The results for both the diatomite and Big Batch did not differ from the long-term average values for these materials (compare Tables 1 and 3), indicating that the method will be free of a memory effect for studies of natural materials.
Our analysis temperature of 1 °C was chosen to fall well below the 5 °C threshold to avoid the adverse effects of warmer temperatures. We found no difference in the isotopic values obtained in stainless steel sample vials versus vials made of borosilicate glass. HF is produced in a 2:1 mole ratio with SiF4 during the decomposition of Cs2SiF6, but the glass vials appear unreactive under the cold dry conditions of the reaction with 98% H2SO4. Precision and Accuracy. We determined the precision and accuracy of the method using Big Batch and a sample of commercial purified diatomite from Lompoc, California, with the brand name Celpure. The standard deviations about the mean value for Big Batch and diatomite were 0.07 and 0.08‰, respectively (Table 1). This level of precision is comparable to that achieved by MC-ICPMS and represents a 2.4-fold improvement in precision over laser fluorination (Table 1).22 The mean δ30Si value for Big Batch is 0.15‰ lower than that obtained with laser fluorination, but it is within 0.01‰ of the mean value for this material obtained by MC-ICPMS with Mg calibration (Table 1).22
SUMMARY The method described involves simple sample preparation procedures that are suitable for a wide variety of natural samples. The precision of the δ30Si measurement is improved over laser fluorination and is comparable to that reported for MC-ICPMS. The isotopic analysis is fully automated, thus allowing unattended sample processing. Currently, 12 samples can be analyzed in a 24-h period, as compared to 6 samples per day for laser fluorination.
(24) Basile-Doelsch, I. J. Geochem. Explor. 2006, 88, 252-256.
AC0606406
6114 Analytical Chemistry, Vol. 78, No. 17, September 1, 2006
ACKNOWLEDGMENT This work was supported by OCE-0350576 from the National Science Foundation. We thank Celite Corporation of Lompoc, California for supplying the Celpure diatomite used in this study. Karen Ziegler and Oliver Chadwick kindly provided the sample of soil water.
Received for review April 6, 2006. Accepted June 29, 2006.
