General UCSC silicate rock analysis methods and results
Jim Gill, May, 2004
The following describes typical procedures used for major and trace element, and isotope analyses of silicate rocks at UCSC. This description supercedes previous publications of UCSC XRF and TIMS methods. It is the first 'publication' of our ELEMENT and Neptune ICPMS methods. Specific projects may differ as noted in publications reporting their results. Comments and questions to email@example.com are welcome.
A. Sample preparation
Whole rock samples ~500 g are the norm for crystalline rocks. Exterior surfaces are removed using a hammer, splitter, or saw, resulting in pieces several cm in maximum dimension which are inspected to avoid visible alteration. The pieces are then mechanically pulverized on a steel or alumina jaw crusher to <5mm pieces and powder. Then a 20 g aliquot is crushed in a ceramic ring mill or alumina ball mill for trace element and isotope analysis, and a separate 20 or 70 g aliquot is crushed to <200 mesh in a WC ring mill for major element analysis.
Submarine samples are sonicated in distilled water until no turbidity is observed, then dried before crushing. Glass samples are hand picked under a microscope. Before final crushing, submarine samples are leached in 1N HCl for 10 minutes at room temperature to remove carbonate. Samples for U-Th isotope analysis are further leached sequentially in (sodium dithionate +citrate) and (0.5M oxalic acid + 2%H2O2).
B. XRF and LOI
Major element concentrations are measured by x-ray fluorescence (XRF) using an automated Philips PW1410 X-ray Spectrometer fitted with a Rh X-ray tube operated at 50 kV and 55 mA. Sample preparation and data reduction strategy are based on the WSU methods of Johnson et al. (1999). Off-peak backgrounds are removed for Mg, Na, Mn, and P. Analyte intensities are normalized to a rock standard run every fourth time as a drift correction standard. The normalized analyte intensities are reduced to oxide concentrations using the fundamental parameters method, as provided in the ÒXRFwinÓ software package by Omni Instruments, Inc. Samples are measured in duplicate and results are averaged. Total analytical time is ~30 minutes per sample for ten elements.
Samples are analyzed as glass pellets that have been diluted 2:1 (flux:sample) with lithium tetraborate. About 3.5g <200 mesh size sample powder is used, and the flux weight is adjusted for its loss on ignition (LOI). The combined powders are thoroughly blended by agitation for 2 minutes using a Spex Mixer/Mill before static fusion in a graphite crucible placed inside a muffle furnace set at 1000oC for 15 minutes. Pellets are then re-crushed in a WC ring mill for at least 40 seconds, and re-fused. Sums can be 1-2 wt% high, especially in silica, when the initial powder or crushed pellet is >200 mesh. Pellets are flattened and smoothed with 600 mesh silicon carbide grit, then cleaned and sonicated in isopropyl alcohol in order to avoid hydration of the freshly polished surface, which can result in low sums.
Calibrations of the instrument response parameter used in the Fundamental Parameters method are based on 12 international rock standards ranging from basalt to rhyolite. Table 1 gives the concentrations assumed for them, and identifies the standards omitted from calibration of some elements. Results are given on an anhydrous basis with all Fe as FeO, which is how we report results. Table 1 also gives the concentrations obtained for the standards when run as unknowns, their reproducibility, and the average difference between duplicate pellets for unknowns. In general, results are accurate and reproducible to within 1-2 relative % or 0.2 wt% whichever is higher. Sums are usually 99.0-100.5 wt% including LOI before normalization to anhydrous values.
LOI for samples is measured by igniting rock powders in alumina crucibles to 1200oC for 10 minutes in a muffle furnace.
Digestions for ICPMS trace element and isotope measurement are made in a suite of Class 1000 clean labs. Each task is performed in a separate room or separate bay in a HEPA-filtered, laminar flow, metal-free environment.
Most digestions use open teflon beakers that have been cleaned by sequential soaking in detergent, 50%HNO3, 50%HCl, and 50%HNO3 using trace metal grade (TMG) acids. Beakers used for U-series or Hf isotope analyses are additionally refluxed using sequential aqua regia and concentrated HF. Beaker volumes in ml typically are ~20% the sample mass in mg.
Typical sample:acid proportions are 100 mg sample to 3 ml conc HF and 1 ml 7N HNO3. Those involving spikes or leading to ICPMS concentration measurements also use 0.5 ml conc HCLO4 to ensure complete evaporation of F which helps prevent formation of insoluble fluoride precipitates during the initial evaporation. (Fluorides can co-precipitate 50-80% of LILE and REE, especially HREE, during digestion of basalt [Yokoyama et al., 1999]). When different sample masses are used, acid volumes are scaled accordingly. This usually is followed by digestion in 8N HNO3 and evaporation to ensure complete conversion of solids to nitrate salts. The residue is taken up in 6N HCl to obtain clear solutions and preceding steps are repeated as necessary to achieve clear solutions. The sample is then converted to dilute nitric for ICPMS analyses, dilute HBr for Pb, or dilute HCl for Sr-Nd or Hf purification. Solutions are centrifuged before being loaded onto ion exchange columns.
For U-series analyses requiring large samples (>300 mg), the HF+HNO3+HClO4 step is dried only to a damp paste, then 2-5 ml of 2.5 N HCl saturated with boric acid is added and refluxed over night before fuming and drying to a hard cake. This cake is then dissolved in 6 N HCl and the solution centrifuged. Any insoluble materials are re-attacked with 2.5 N HCl saturated with boric acid, refluxed, fumed, and dried. This is repeated as often as necessary until nothing solid is seen after centrifuging, and the solutions combined.
Total processing blanks (including column chemistry) usually are <300 pg Sr and Nd, <150 pg Pb, <100 pg Hf, and <40 pg Th and U. Sample masses are adjusted to ensure that blanks are <1%.
When refractory accessory minerals are suspected (e.g., zircon, sphene, garnet, spinel), especially in phaneritic felsic rocks, dissolutions are made in steel-jacketed teflon bombs.
Trace metal grade acids are used for normal ICPMS concentration measurements. For isotope ratio and isotope dilution concentration measurements, the HCl, HNO3, and HF are reagent grade acids that are then double-distilled in sub-boiling quartz or teflon stills. Other acids are Optima grade.
D. Trace element analysis by HR-ICPMS
Concentrations of 31 trace elements are routinely determined using an ELEMENT1 high resolution inductively coupled plasma mass spectrometer (HR-ICPMS). Table 2 lists the isotopes used for each element. Approximately 50 mg of sample are dissolved. Samples are diluted to 103 ml of 1% HNO3 plus 0.004M HF. In, Re, and Bi at a solution concentration of 10 ppb are added for internal normalization to reduce effects of drift. No correction is required for oxide interferences or mass overlaps.
Calibration is against six international rock standards, assuming the concentrations given for them in Table 2 which are mostly those of Eggins et al. (1997) and Kelley et al. (2003). (Some adopted values differ from those given on the website of the sponsoring agency. These are given in italics in Table 2). Solutions of these standards are robust for years, but they require sonication before each use if they have been sitting for more than a couple of months. Unknowns are typically analyzed for the first time within weeks of preparation. Calibration slope and intercepts vary little over years because of internal standardization.
To avoid negative intercepts, and to ensure that low concentration samples can be accurately measured, the intercept is forced through zero for all elements except V, which has a high instrument blank due to a number of interferences. This is especially important for elements that show a narrow range in nature, such as the heavy REE, because the slope would otherwise be poorly constrained and subject to large deviations from one run to another. The only drawback to forcing the intercepts through zero is that the total process plus instrument blank is not subtracted from the analysis, as it would be if a positive intercept were calculated and used (and as is done with V). Instead a process blank is prepared along with the samples, and the measured intensities are subtracted from the sample analyses before concentrations are calculated from the calibration curve information. R2 values for linear calibration curves cannot be calculated from regressions forced through zero, but in measuring the total blank using calibrations not forced through zero, R2 values ranged from 0.98925 to better than 0.99999 for basalt to rhyolite. Analysis of low concentration samples (peridotites, N-MORB, and oceanic arc basalts) usually are calibrated only against the low concentration standards BIR, JB2, and DNC.
Total analytical time for each sample is about 2.5 minutes, which is done in 50 scans. Uptake is typically one minute, with a 10 second wash between samples. The wash solution has the same solution chemistry as a sample, except for the sample matrix (i.e. internal standard, trace HF, etc.). This ensures that equilibrium is maintained between the sample solution and the peristaltic uptake tubing, which allows the instrument to be tuned using the wash solution. Sample gas flow is set to the minimum that gives a stable signal, which is typically between 0.8 and 0.9 liters/minute. Standards are run immediately after tuning the gas flow, followed by the unknowns. After one rack of samples is run, gas flow tuning is re-evaluated. If a change is required then standards are re-run before running the next (or repeating the same) set of samples. Glass ExpansionÕs micromist nebulizer is used with a standard Scott, double-pass spray chamber. The guard electrode is not used so as to minimize oxide and other molecular interferences; this has been found to be more important than the concomitant loss in sensitivity. Drift is monitored by running an in-house standard after every third sample, and can vary from less than one to six or seven relative percent during 2-2.5 hours. Drift correction is accomplished by defining linear changes in instrument response, and interpolating the effect of this change for the samples bracketed by the drift standard (e.g. for a change of 2% in the measured concentration of an element in the drift standard, the first sample would be adjusted by 0.5%, the second by 1%, and the third by 1.5%). Non-linear changes in instrument response are not corrected for, and if instrument behavior appears erratic it is re-tuned and the standards and samples are run again.
Internal precision (RSD values per element per analysis) typically are 1-3% for rock concentrations >1 ppm and 4-8% for concentrations of 10-1000 ppb. Repeated analyses of the same solution during one calibration (external error) typically gives 2s of 1-3% at rock concentrations >100 ppb (Table 2). Repeated dissolutions of the same powder typically gives similar %2s. When double-corrected for drift by normalization to a frequently analyzed solution throughout one analytical session, external error drops to 0.5-2% at rock concentrations as low as10 ppb except for Ni and Ta. This includes all sources of error (solution preparation, instrument drift, and counting statistics).
International standards run as unknowns generally agree with accepted value within external error. Examples are given in Table 2.
E. Column chemistry
Pb is purified using 250 ml of AG1-8x anion resin in teflon columns. Samples are loaded and most elements are removed in 1N HBr. Pb is collected using 6N HCl. This procedure typically is repeated twice.
Sr is purified using 8 ml of AG50-8x cation resin in glass columns. Samples are loaded and Sr is collected in 2.5N HCl. This is repeated twice for samples with high Rb/Sr. The REE are then removed using 6N HCl.
Nd is purified using 3 ml HDEHP resin (Ln-Spec) in glass columns. Samples are loaded and LREE are separated using 0.25N HCl. Less than 0.1% Sm and 1-2% Ce remain in the Nd cut.
Hf is purified using 1 ml of Ln-Spec 100-150m resin in BioRad poly-prep columns. Samples are loaded in 3M HCl and eluted following Munker et al. (2002) where major elements are first removed in 3N HCl, Ti-Nb in weak nitric and citric acid, Zr in (6N HCl + weak HF), and finally Hf in (6N HCl + 0.4N HF).
Th, U, and Ra are purified from one dissolution using mostly AG1-X8 anion resin ('resin' below), plus Sr-Spec resin for the final Ra-Ba separation. First, Th+Ra are separated from U using 10 ml resin in a Biorad column. The sample is loaded in 10 ml 6N HCl, and Th+Ra are eluted in 30 ml 6N HCl. Then U+Fe are eluted in 20 ml H2O followed by 20 ml 0.5 N HCl.
The Th is purified further using the same resin by loading the Th+Ra split in 10 ml 8N HNO3, eluting the Ra in 20 ml 8N HNO3, washing with 30 ml 8N HNO3, and then eluting Th in 5 ml H2O followed by 30 ml 6N HCl. The Th is again purified by repeating this step using the same resin in a 1 ml transfer pipette, followed by a 150 ml teflon microcolumn, with acid volumes scaled accordingly. A drop of boric acid is added at each dry-down to prevent formation of fluorides.
The U split is converted to nitrate and purified further using the same resin in 8N HNO3. The sample is loaded in 1 ml 8N HNO3, Fe is removed in 30 ml 8N HNO3, and U is eluted in 10 ml H2O followed by 30 ml 0.5 N HCl. This step is repeated using a 150 ml teflon microcolumn with acid volumes scaled accordingly, and repeated until the beaker is clear at dryness.
The Ra split is converted to chloride and loaded in 4 ml 6N HCl onto 8 ml of AG50x12 200-400 mesh cation resin in a 10-ml transfer pipette. After washing with 24 ml 6N HCl, Ra is eluted in 32 ml 4N HNO3. It is then loaded in 0.5 ml 4.6N HCl onto 2 ml of the same resin in a Biorad polyprep column. Another 2 ml of 4.6N HCl is added to the beaker and then loaded onto the column. After washing with 16 ml of 4.6N HCl, Ra is eluted in 6 ml of 4N HNO3. Finally, this Ra split is loaded in 1.1 ml HNO3 onto 150 ml Sr-Spec resin in a teflon microcolumn. After washing with 500 ml of 1.1N HNO3 to remove Ba, the Ra is eluted in 1 ml of 1.1 ml HNO3.
Pa column chemistry follows the procedures described by Lundstrom et al. (1998).
F. TIMS Mass spectrometry
Thermal ionization mass spectrometry (TIMS) uses a VG Sector 54 spectrometer equipped with nine Faraday and one ion-counting Daly detector and a Wide Angle Retarding Potential (WARP) lens that achieves abundance sensitivity of <150 ppb at one amu when the analyzer pressure is < 10-8 torr. All Faraday detectors and the housing block were replaced in 1995, as was the measurement system and control interface in 2003.
Sr isotopes are measured by TIMS using single Re filaments and a Ta activator. Measurements are in dynamic mode. Typical loads are 100-500 ng, yielding 88Sr beams of 2-3 V. Typical internal error (2 std error) is <0.00001 for 100 ratios. Data are corrected for internal mass bias to 86Sr/88Sr = 0.1194. Ratios <0.118 are not accepted because they tend to result in 86Sr/88ratios that are too low by >0.00001. Correction is made for 87Rb interference by monitoring 85Rb and assuming 85Rb/87Rb = 2.5926. Results are normalized to 0.710250. Results from mid-2003 onward for NBS 987 are 0.710229 ±23 (2s). As one measure of external error in rock samples, nine analyses of different samples from the same eruption of Arenal volcano had 2s of 0.00001, and probably some of this variation is geological.
Nd isotopes are measured by TIMS using double Re filaments in dynamic mode. Typical loads are 100-500 ng yielding 146Nd beams of 0.5-2V and un-normalized 146/11 ratios = 0.721-723. Typical internal error (2 std error) is <0.000015 for 150 ratios. Data are corrected for internal mass bias to 144Nd/146Nd = 0.7219. Then correction is made for 143Sm interference by monitoring 147Sm and assuming 144Sm/147Sm = 0.20632. Results for La Jolla during 2004 are 0.511851 ±0.000009 (2s). 143/144 and 145/144 are found to be too low in unknowns when Ba is present. Consequently, 138Ba and 145/144 are monitored, and results are not accepted unless 145/155 = 0.34840-42.
Pb isotopes are measured by both TIMS and solution MC-ICPMS. TIMS measurements are made in static mode. >200 ng Pb is loaded together with silica gel and phosphoric acid. Uncorrected results for NBS 981 during 2003 are 206Pb/204Pb = 16.896 ±.056 (2s external reproducibility), 207Pb/204Pb = 15.438 ±.049, and 208Pb/204Pb = 36.537 ±.151. (See next section for accepted values.) Consequently, the mass fractionation correction applied to unknowns is 1.0±0.1 per mil per amu.
226Ra and 231Pa concentrations are measured by ID-TIMS using 228Ra and 233Pa spikes following procedures described most recently by Lundstrom et al. (1998). About 10 fg of 228Ra or 100 fg 233Pa is added. Duplicates typically agree to within 1-2%. Accuracy is assessed by approach to 226Ra-230Th or 231Pa-235U equilibrium in TML. Both are within 1-2% internal error.
G. MC-ICPMS Mass Spectrometry
Multicollector inductively coupled plasma mass spectrometry (MC-ICPMS) uses a ThermoFinnigan Neptune(TM) second generation double-focusing magnetic sector instrument. It is equipped with one fixed and eight moveable Faraday collectors, and a single (fixed) secondary electron multiplier (SEM) for high abundance sensitivity. Its ion detection system is equiped with an retarding potential quadrole (RPQ) and is characterized by low dark noise (<<1 cpm) and abundance sensitivities <50ppb at 2 amu at an SEM/Faraday yield of >75%. This allows measurement of very low abundance nuclides (e.g. 234U, 230Th). Aqueous samples are introduced at 50 mL/min using a Teflon nebulizer fitted to either a glass spray chamber or an MCN 6000 desolvating spray chamber. Typical sensitivities for all isotopes measured are ~55-65V/ppm analyte for the glass spray chamber and >200V/ppm analyte for the MCN spray chamber. Typical on-peak backrounds are less than 0.01mV and therefore much less than 0.001% of the total signal intensities for most elements. Low abundance isotopes of U and all isotopes of Th have on-peak backgrounds less than 0.05% of the total signal intensities typical of samples. No matrix effects (e.g., of Ti or W on Hf) have been found that are significant after our purification methods.
Hf isotopes are measured by solution MC-ICPMS in static mode. Solutions are 0.3M HNO3 plus trace HF which contain 50-100 ppb Hf. Sensitivity is ~200V per ppm Hf in solution. Solutions are introduced using an MCN 6000 desolvating spray chamber and free aspiration with an uptake rate of 50microL/min. Sweep gas and nitrogen flow rates are ~3.5 L/min and 10-20 mL/min, respectively. Rinse time is 5 min. Total data collection time per analysis is 20 minutes, and the total sample consumed is ~1 ml. . Data are collected as a single block of 100 8-second integrations. On-peak backgrounds are measured immediately prior to each analysis and automatically subtracted from all peaks. Mass bias is corrected exponentially using 179Hf/177Hf = 0.7325, and small (<<1%) isobaric interferences of Yb and Lu on 176Hf are corrected using 176Lu/175Lu = 0.2656 and 176Yb/173Yb = 0.7876. JMC 475 is run 4-5 times during each analytical session and results are reproducible within one session to <0.000004 (2s). Daily average values during 2003-4 are 176Hf/177Hf = 0.282146 +/- 0.000006 (2s). Results for each analytical session are normalized to JMC 475 = 0.282160 for that session. JMC-normalized values for unknowns agree between sessions to within 0.000005 which we take as our external reproducability.
Pb isotopes are measured by solution MC-ICPMS in static mode using internal spiking with NBS 997 Tl for mass fractionation correction, assuming 203Tl/205Tl = 0.418911. Solutions are 2% HNO3 which contain 100-150 ppb Pb. Results for standards and unknown are indistinguishable when Pb/Tl ranges from 2 to 10 (Wolff and Ramos, 2003). Results for NBS 981 are 206Pb/204Pb = 16.929 ±0.002 (2s external reproducibility), 207Pb/204Pb = 15.483 ±0.002, and 208Pb/204Pb = 36.671 ± 0. 005. This agrees well with TIMS double-spike results: 206Pb/204Pb = 16.936 ±0.001 (2s external reproducibility), 207Pb/204Pb = 15.489 ±0.001, and 208Pb/204Pb = 36.701 ± 0. 003 External errors for silicate unknowns are 0.001-2 based on triplicate dissolutions run on different days.
U isotopes are measured by solution ICP-MS in static mode with an abundance sensitivity of <25 ppb at 2 amu. Solutions are 2% (0.3N) HNO3 containing ~10-20 ppb U, and introduced using a teflon nebuliser in free-aspiration mode. Sensitivity is 50-100V per ppm U in solution. Rinse time is 5-10 minutes or until backgrounds on 234U are <5 cps. Total data collection time per analysis is ~10 minutes, and the total sample consumed is 0.5-1.0 ml. Data are collected as a single block of 25-50 8-second integrations. Internal errors (2 std errors) on 234U/238U are <0.1 %. On-peak backgrounds are measured immediately prior to each analysis and automatically subtracted from all peaks. Mass bias is corrected exponentially assuming 235U/238U = 137.88. An SEM/Faraday yield is determined before each unknown by measuring 235U sequentially on both detectors. NBS 4321 is run periodically during each analytical session and results are reproducible within one session to 2x10-7 (2s). Its 234U/238U is 53.2 ±0.1ä which agrees with the result of Richter et al. (2003) rather than the certified NBS value. 234U/238U ratios for TML and AThO are within 1% of equilibrium (0.0000549), and disequilibria measured by Neptune is silicates agree with those measured by TIMS to within 0.5%. We estimate that our external error for 234U/238U between days for a silicate unknown is ~1% 2s.
Th isotopes are measured by solution ICP-MS in static mode with an abundance sensitivity of <25 ppb at 2 amu. Solutions are 2% (0.3N) HNO3+0.1% HF containing ~50-100 ppb Th and introduced using a MCN 6000 desolvating nebulizer. Sensitivity is ~200-300V per ppm Th in solution. Rinse time is 10-15 minutes or until backgrounds on 230Th are <5-10 cps. Total data collection time per analysis is 10-20 minutes, and the total sample consumed is 0.5-1.0 ml. Data are collected as a single block of 25-50 8-second integrations. Internal errors (2 std error) on raw 230Th/232Th are <0.1 %. On-peak backgrounds are measured immediately prior to each analysis on a blank solution of 2% (0.3N) HNO3+0.1% HF and automatically subtracted from all peaks. Currently no tail correction from 232Th on 230Th is made because it is <0.3% for the ratios of interest and other sources of error are greater. Mass bias, SEM/Faraday yield, and dead time corrections are addressed altogether in either of two ways. First, the 236U/238U ratio of NBS U010 is measured, normalized to an assumed value of 0.0000693, and this factor is applied linearly to the 230Th/232Th ratio. This is sometimes done before and after an unknown, or the NBS U010 is added to the U-free Th of a sample. 232Th/230Th ratios measured in this way are routinely a few percent too high, but the U-normalized values are reproducible to within 1% 2s during an analytical session. These U-normalized values are then Th-normalized by reference to an international standard run during the analytical session. Alternatively, 232Th/230Th are simply normalized once to the accepted value of a bracketing standard solution such as UCSC ThA (171,500) or Open University ThU (161,917). TML and AThO give accepted results (172000-173000 and 182500-183500, respectively) by both methods when run as unknowns. We estimate that our external error for 232Th/230Th between days for a silicate unknown is ~1% (2s).