Institute for Research on Earth Evolution, Japan Agency for Marine-Earth Science and Technology, 2-15 Natsushima-cho, Yokosuka , Japan

Transcription

1 Geochemical Journal, Vol. 47, pp. 369 to 384, 2013 NOTE Pb isotope analyses of silicate rocks and minerals with Faraday detectors using enhanced-sensitivity laser ablation-multiple collector-inductively coupled plasma mass spectrometry JUN-ICHI KIMURA,* HIROSHI KAWABATA, QING CHANG, TAKASHI MIYAZAKI and TAKESHI HANYU Institute for Research on Earth Evolution, Japan Agency for Marine-Earth Science and Technology, 2-15 Natsushima-cho, Yokosuka , Japan (Received October 12, 2012; Accepted February 13, 2013) We present analytical protocols and the analytical performance of Pb isotope analysis of silicate minerals and rocks at the Pb = ppm level using excimer laser ablation (LA) coupled with an enhanced-sensitivity multiple collectorinductively coupled plasma mass spectrometer (MC-ICP-MS). We applied Faraday cups for all the isotopes analysed, including the minor isotopes 204 Pb and 204 Hg. This device is, so far, the first attempt to analyse samples with low Pb concentrations by LA-MC-ICP-MS. The improved sensitivity of the mass spectrometer, along with careful correction of the Hg overlap on 204 Pb, permits accurate analyses of 206 Pb/ 204 Pb, 207 Pb/ 204 Pb, and 208 Pb/ 204 Pb ratios. This procedure includes determination of the fractionation factor of Hg isotopes among 200 Hg, 202 Hg, and 204 Hg. The analytical error, including precision and reproducibility, was within 1.5 per mil for 5 15 ppm Pb in glassy silicate rocks (groundmass) taken from 500 µm 500 µm 100 µm (0.025 mm 3 ) craters. These results were comparable to those obtained using conventional non-spiked thermal ionization mass spectrometry after chemical separation. Major Pb isotope ratios, including 207 Pb/ 206 Pb and 208 Pb/ 206 Pb, were also analysed to within a 1 per mil error for ppm Pb in groundmass and plagioclase crystals from craters 200 µm in diameter and 100 µm deep. When a larger sample size, i.e. a 500 µm 500 µm 200 µm crater was used, a ~10 per mil error was achievable for samples containing ~0.02 ppm Pb, which allows in situ analysis of low-pb minerals such as clinopyroxene. Melt (glass) inclusions in olivine with a sample size of diameter ~200 µm containing 1 5 ppm Pb were also analysed with 1 10 per mil errors. The achievable accuracy and precision were significantly better for 207 Pb/ 206 Pb and 208 Pb/ 206 Pb, and better for 206 Pb/ 204 Pb, 207 Pb/ 204 Pb, and 208 Pb/ 204 Pb than those obtained by in situ analyses using secondary ion microprobe mass spectrometry and a normal-sensitivity LA-MC-ICP-MS using one or more ion counters. The analytical performance of the present method is useful for many applications connected with geochemical studies on igneous rocks. We present examples for groundmass, plagioclase, clinopyroxene, and olivine melt inclusion analyses of ocean island basalts. Keywords: laser ablation, Pb isotopes, MC-ICP-MS, glass, melt inclusion, plagioclase, clinopyroxene INTRODUCTION Pb isotope analysis of silicate samples using laser ablation-multiple collector-inductively coupled plasma mass spectrometry (LA-MC-ICP-MS) is a versatile tool for the micro-analysis of silicate minerals (Gagnevin et al., 2005; Kent et al., 2005; Mathez and Waight, 2003) and glasses (Kent, 2008a, b; Kent and Dilles, 2005; Kent et al., 2005; Paul et al., 2005, 2011; Peate et al., 2003; Simon et al., 2007; Westgate et al., 2011). Pb has one minor isotope, 204 Pb (~1.4%), and three major isotopes, 206 Pb, 207 Pb, and *Corresponding author ( jkimura@jamstec.go.jp) Copyright 2013 by The Geochemical Society of Japan. 208 Pb. Because of the low concentration of Pb in natural silicate rocks and minerals (typically sub parts per million to tens of parts per million), one or more ion counters (ICs, for Hg isotopes and the minor 204 Pb isotope) and Faraday cup (FC) detectors (for the major Pb isotopes) have typically been used for analyses of isotope ratios (Kent, 2008a; Kent et al., 2008; Paul et al., 2005). When using ICs, careful adjustment and monitoring of the gain is necessary to obtain accurate results (Kent, 2008a; Paul et al., 2005). In contrast to ICs, FCs are suitable for stable and accurate isotope ratio measurements (Kent, 2008b; Kent et al., 2008; Paul et al., 2005). However, the low sensitivities and low signal-to-noise ratios in FCs limit the application of FC detectors, particularly with 204 Pb. Recent improvements in the sensitivity of MC-ICP-MS instru- 369

2 Groundmass analysis A Pre-ablation cleaning B Post-ablation crater 208 Pb / V Photo micrograph of melt inclusion 207 Pb/ 206 Pb 206Pb/ 204 Pb C Laser crater Olivine Melt inclusion analysis D E F Scan Melt inclusion Fig. 1. Photomicrograph of a rectangular laser crater for pre-ablation cleaning (A) and post ablation crater (B), and spot laser crater dug through melt inclusion in olivine crystal (C), and signal intensities ( 208 Pb: D) and isotope ratios ( 207 Pb/ 206 Pb: E, 206 Pb/ 204 Pb: F) obtained by transient Pb isotope signals from olivine melt inclusion. Solid squares indicate signals used for isotope ratio calculations. Abscissas indicate scan numbers (single scan = ~4 s). ments has resulted in a sensitivity increase of about one order of magnitude, e.g., >3000 V ppm 1 (>190 Gcps ppm 1 ) in solution mode using a desolvating nebulizer (Bouman et al., 2008). The improvement was achieved by modification of the sampling and skimmer cone designs and an improved vacuum at the expansion chamber in the ion-sampling interface region (Bouman et al., 2008). A high-gain amplifier with a Ω resistor for FCs also allows accurate amplification of signals at the microvolt level. These improvements encouraged us to use FCs for LA isotope analysis of Pb, including the minor 204 Pb isotope. We report the application of high-sensitivity MC-ICP- MS to LA analysis of Pb isotope ratios in natural silicate rocks and minerals using FC detectors for all the Pb isotopes. Correction of the 204 Hg isobar on 204 Pb is also examined by determining the fractionation of Hg using 200 Hg/ 202 Hg as measured by FCs. Accurate and precise Pb isotope analyses, within ~1.5 per mil, were achieved for 206 Pb/ 204 Pb, 207 Pb/ 204 Pb, and 208 Pb/ 204 Pb, using 500 µm 500 µm 100 µm craters for the groundmass of lava samples containing 5 15 ppm Pb. Major isotope analyses for 207 Pb/ 206 Pb and 208 Pb/ 206 Pb with accuracies/ precisions of <0.5 per mil were achievable from the same sample size. Accuracies and precisions of 1 10 per mil for the major isotope ratios were obtained for craters µm in diameter and µm in depth, even for a Pb sample with as little as 1 5 ppm Pb. Below, we describe the analytical protocols of the measurements and present the application results for natural silicate rocks (groundmass), minerals, and melt inclusions in olivine crystals. 370 J. Kimura et al.

3 Table 1. Instrumental setup of 193 nm ArF excimer laser ablation-multiple collector-inductively coupled plasma mass spectrometer (LA-MC-ICP-MS) Laser ablation system Laser source 193 nm ArF excimer laser (Coherent, ComPex 102) Pulse width <20 ns Pulse energy 200 mj at laser output Laser fluence 20 J cm 2 at sample surface Focusing objective lens Air-spaced doublet UV-grade silica objective lens (OK Laboratory Limited) Beam diameter 50, 100, and 200 µm diameter for spot analysis µm 2 crater for scan analysis resulted in µm 2 crater Mass spectrometer MC-ICP-MS Thermo Fisher Scientific, Neptune (modified) Plasma power 1500 W (27.12 MHz) Guard electrode On Plasma Ar gas-flow rate 13 L min 1 Auxiliary Ar gas-flow rate 0.7 L min 1 Sample Ar gas-flow rate 1.4 L min 1 Sample He gas-flow rate 1.15 L min 1 Sample cone JET-cone (Ni) Skimmer cone X-cone (Ni) Interface pump Pfeiffer UNO35 in tandem with Edwards E2M80 (total pumping speed 115 m 3 h 1 ) Interface vacuum 1.5 mbar with He carrier gas Mass resolution M/ M = 400 (low resolution) Typical sensitivity >3000 V ppm 1 for Pb in solution mode using Aridus desolvating nebulizer Oxide molecular UO + /U + < 40% Detector mode Faraday cup static mode L3 200 Hg Ω amplifier (mass-bias correction for Hg) L2 202 Hg Ω amplifier (mass-bias correction for Hg) L1 204 Pb Ω amplifier (Pb signal, 204 Hg subtraction by L3/L2) CF 205 Tl Ω amplifier (Tl monitor) H1 206 Pb Ω amplifier (Pb signal) H2 207 Pb Ω amplifier (Pb signal) H3 208 Pb Ω amplifier (Pb signal) Data acquisition On-peak background subtraction (60 s before data acquisition) Standard bracketing method (SRM612 standard) w/o external correction 24 times ~4 s scan (total acquisition time 96 s) INSTRUMENTAL SETTINGS Laser ablation We used an excimer LA (ExLA) system (OK- ExLA3000, OK Laboratory Ltd., Tokyo, Japan), which used a ComPex 102 (Coherent, Bremen, Germany) excimer laser source (1 20 Hz, 200 mj output) with imaging laser-focusing optics (Kimura et al., 2012). This allowed focusing of the excimer laser beam on craters 20, 30, 50, 100, and 200 µm in diameter. Application of a large rectangular slit generated a 500 µm 200 µm crater. Line raster analysis using a motor-driven tri-axial stage with a rectangular slit resulted in the 500 µm 500 µm 100 µm crater used in this study (Figs. 1A and B). Spot analyses with a 200 µm diameter crater were also performed without raster mode (Fig. 1C). The repetition rate of the laser was set at 5, 10, and 20 Hz, depending on the size of the Pb signals. This resulted in craters with depths of ~50, 100, and 200 µm, respectively, after ~2 min ablation, irrespective of the crater size and shape (Table 1). Changes in the repetition rate altered the drilling depth of a crater almost exactly in proportion to the repetition rate of the laser pulse. He carrier gas at a flow rate of 1.15 L min 1 was fed (Horn and Günther, 2003) to an in-house-designed sample chamber (~20 cm 3 inner volume) through five fine nozzles, allowing steady-state laminar flow inside the chamber (Kimura et al., 2012). A mixing device with a 150 cm 3 inner volume was used to mix the He carrier gas with the Ar sample gas (~1.4 L min 1 ) immediately in front of the ICP torch (Table 1) (Kimura et al., 2011). Pre-ablation of the surfaces using a 10 Hz laser for 2 s (Fig. 1A) removed any remnants of surface Pb contamination (Paul et al., 2011) that remained after washing with hexane and 0.6 nm hydrochloric acid as necessary (Kimura and Nakano, 2004; Kimura and Chang, 2012). Pb isotope analyses by LA-MC-ICP-MS with Faraday cups 371

4 MC-ICP-MS We used a Thermo Scientific Neptune MC-ICP-MS (Bremen, Germany), which was fitted with a modified ion-sampling interface that included a high-efficiency rotary vacuum pump (E2M80, Edwards, Crawley, West Sussex, UK), which allowed for a higher vacuum of ~1.5 mbar (1 bar = 10 5 Pa) at the expansion chamber (in He + Ar carrier gas) (Kimura et al., 2012). The ion interface cones used were JET-sampling and X-skimmer cones (Thermo Fisher, Bremen, Germany). The guard electrode was kept on during all analyses. The modifications improved the instrument-sensitivity by about one order of magnitude (>3000 V ppm 1 Pb = ~190 Gcps ppm 1 ), using an Aridus II (CETAC Technologies, Omaha, NE, U.S.A.) desolvating nebulizer for solution analysis mode (Table 1). The same refinements were applied to our LA- MC-ICP-MS system, enabling us to achieve the highest sensitivity (Bouman et al., 2008). A high-gain amplifier (Thermo Scientific, Bremen, Germany) of with a Ω resistor was used for the FC, which allowed accurate amplification of low signals. We only used the high-gain amplifier for 204 M (M = mass; L1), in which the minor isotope 204 Hg (6.87% natural abundance) overlaps the minor isotope 204 Pb (~1.4% natural abundance). Other isotopes, i.e., 200 Hg, 202 Hg, 206 Pb, 207 Pb, and 208 Pb were analysed using FCs with a normal amplifier with a Ω resistor (Table 1). 200 Hg and 202 Hg were used to monitor the mass fractionation of Hg and estimate the 204 Hg signals for the overlap correction (for details, see below). For comparison with bulk-rock Pb isotope compositions, basalt lava samples from Pitcairn Island (PC samples) were analysed by solution MC-ICP-MS using an Aridus desolvating nebulizer and a Neptune MC-ICP-MS (Thermo Fisher, Bremen, Germany) with normal samples, and skimmer cones after the Pb separation procedure (Miyazaki et al., 2012). ANALYTICAL PROTOCOLS Standard-sample bracketing method Use of a laser-aerosol and solution-aerosol dualintake system is useful for determination of fractionation in LA-MC-ICP-MS (Horn et al., 2000; Kimura et al., 2012). However, the introduction of additional Pb impurities (e.g., Pb in the blank solution or Pb memory from Aridus II) from the solution-introduction system must be avoided because we are analysing natural silicate rocks (groundmass) and minerals containing such tiny amounts of Pb. Therefore, we used a standard-sample bracketing method, using NIST SRM612 glass as the standard, a method used by previous researchers (Kent, 2008b; Paul et al., 2005). The standard glass contains ~34 ppm Pb (Jochum et al., 2005) and all the Pb isotope ratios have been determined accurately by both solution MC-ICP-MS and LA-MC-ICP-MS in previous studies (Jochum et al., 2005; Kent, 2008b; Weis et al., 2006). Isotope ratios of 206 Pb/ 204 Pb = , 207 Pb/ 204 Pb = , 208 Pb/ 204 Pb = , 207 Pb/ 206 Pb = , and 208 Pb/ 206 Pb = (Kent, 2008a) were used throughout for the experiments. On-peak background subtraction and isobaric overlap correction of Hg The ICP and sample carrier gases (Ar and He) contained Hg impurities. Pb blanks of a few hundred microvolts were also present, mainly because of memories of previous samples and the skimmer cones. We subtracted these blanks using on-peak background measurements for 1 min prior to every analytical run. The Hg impurities from the gas blanks were subtracted using this method (Kent, 2008a; Kent and Dilles, 2005; Paul et al., 2005), but the Hg signals from the ablated silicate samples remained (Paul et al., 2011). We corrected the 204 Hg overlap on 204 Pb using a fractionation factor determined for the 202 Hg/ 204 Hg ratio that is based on the measured 200 Hg/ 202 Hg ratio. Signal acquisition Sample signals were analysed after the on-peak background measurement by ablating the sample for ~2 min. A signal scan of ~4 s was used for the data acquisition, which resulted in datasets of 24 scans for an isotope after background subtraction for most of the analyses. In all sample acquisition scans, the first scan was started 5 s after LA began to avoid the effect of the slow responses of the FC amplifiers to the increasing sample signals, especially for the Ω resistor amplifier used for 204 M. In contrast to the normal analyses, only 4 15 scans (very occasionally 24 scans) of stable signals were available for the melt inclusions of olivines, which had diameters of µm. The effect of the transient signals from the small melt inclusions will be discussed in a later section. Faster scanning using a ~0.2 s acquisition time was also tested, but use of 2SE (standard error) statistical treatment gave the same result in terms of precision. FRACTIONATION OF Hg AND ITS EFFECT ON OVERLAP CORRECTION FOR 204 M In this study, we found that fractionation of Hg was present in the LA-MC-ICP-MS, unlike previous studies, which used IUPAC values to correct the overlap of Hg isotopes (Kent, 2008a; Paul et al., 2005). As noted above, we subtracted Hg impurities in the ablation and ICP gases by on-peak background subtraction. However, Hg was also present in both natural silicate rocks and the SRM612 glass standard. In the case of SRM612, the contribution of 204 Hg to the total 204 M signal was less than 0.1%. There- 372 J. Kimura et al.

5 Table 2. Measured and calculated mass-fractionation factors among Hg isotopes and determination of mass-bias factor for 204 Hg isobaric overlap correction Isotope ratio 202 Hg/ 200 Hg 202 Hg/ 204 Hg IUPAC natural ratio Measured* ± (measured) ± (calculated) M% per mass unit* (measured) (calculated) Measured** (IUPAC) (measured) M% per mass unit** (measured) Note: * 202 Hg/ 204 Hg = ± was calculated based on the measured average mass ratio 202 Hg/ 200 Hg = ± and 204 Pb based isotope ratios from natural rock JA-1 by LA-MC-ICP-MS and the natural IUPAC abundance of 202 Hg/ 200 Hg = 1.293, using the exponential law. Mass-bias factors, M% per mass unit*, are given by these measured and calculated values. **Correction factor 202 Hg/ 204 Hg = was determined empirically by adjusting measured 204 Pb-based Pb isotope ratios of JA-1 to the reference bulk-rock values using 202 Hg/ 200 Hg = (IUPAC) as the normalizing factor for exponential mass-fractionation corrections of other Hg isotopes. Therefore, the mass-fractionation factor of 202 Hg/ 200 Hg at M% per mass unit* = had already been involved in the obtained value of 202 Hg/ 204 Hg. However, an additional correction factor using 202 Hg/ 204 Hg = 4.335, rather than the natural IUPAC value of 4.346, was needed for 204 Hg overlap correction. This indicates a mass-fractionation of Hg isotopes between 200 Hg/ 202 Hg and 202 Hg/ 204 Hg. The fractionation factor was M% per mass unit** = for 202 Hg/ 204 from 202 Hg/ 200 Hg, identical to the measured M% per mass unit* = for 202 Hg/ 204 Hg. We thus used 202 Hg/ 200 Hg = and 202 Hg/ 204 Hg = for mass-fractionation corrections throughout. The above numbers were stable over several months and Hg overlap correction was successfully achieved for Hg-bearing arc lava samples, such as JA-1 and JB-2 (see Fig. 2 and text). Accurate correction was possible by up to 30 50% of total 204 M signal from 204 Hg. fore, the effect was minimal and the overlap corrections using IUPAC factors were valid, either by using the two Hg isotopes for mass-fractionation correction of the Hg overlap via exponential low correction or by applying a mass-fractionation correction factor using a Tl isotope ratio as an external reference with exponential low correction (Kent, 2008a; Paul et al., 2005). However, the effect of the Hg fractionation significantly affected the analytical results of the 204 Pb-based isotope ratios. This was the case for silicate rocks (groundmass) containing large amounts of Hg relative to Pb. We found that some arc-lavas such as JB-2 and JA-1 from the Geological Survey of Japan (GSJ) (AIST, 2010) contained large amounts of Hg (5 7 ppm in JB-2 and ppm in JA-1) (AIST, 2010) than did either of the intraplate-type basalt glasses, BCR-2G and BHVO-2G, provided by the United States Geological Survey (USGS) (Jochum et al., 2005), or ocean island basalt (OIB) lavas from Pitcairn Island (our own PC series samples). Larger amounts of Hg have also been reported from un-degassed submarine OIB samples (Paul et al., 2011). Below, we describe the determination of the fractionation factor of Hg using JA-1 and JB-2. We used original rock chips of JB-2 and JA-1 and also JB-1 igneous rock standards from the archives in GSJ. The JB-2 and JA-1 samples contain 5.6 ppm and 6.4 ppm of Pb, respectively (AIST, 2010). Our analytical results indicated that 20 45% of the 204 M signals of these samples were from 204 Hg, in contrast to <1% found in other samples examined (e.g., JB-1, BHVO-2G, BCR-2G, and Pitcairn Island OIBs). The Pb isotope ratios obtained using isobaric interference corrections based on IUPAC natural abundances of Hg isotopes, 202 Hg/ 200 Hg = and 204 Hg/ 202 Hg = 4.346, led to over-correction of the 204 Hg signals and resulted in 20X Pb/ 204 Pb isotope ratios that were too high for JB-2 and JA-1 (e.g., 206 Pb/ 204 Pb = , 207 Pb/ 204 Pb = , and 208 Pb/ 204 Pb = for JA-1 (Tanimizu and Ishikawa, 2006), and 206 Pb/ 204 Pb = , 207 Pb/ 204 Pb = , and 208 Pb/ 204 Pb = for JB-2 (Baker et al., 2004)). These results clearly indicated that a mass-fractionation that did not follow IUPAC abundances was present among Hg isotopes. We measured the raw (instrumental mass-fractionation uncorrected) 202 Hg/ 200 Hg ratios using JA-1 and obtained ± (n = 50) (Table 2). JA-1 was used because the groundmass composition was fairly uniform among the samples examined, and it contained a sufficient amount of Hg for determining accurate fractionation factors. With the measured JA-1 dataset, we then calculated the appropriate 204 Hg/ 202 Hg ratio by adjusting the measured 20X Pb/ 204 Pb isotope ratios of JA-1 to those of the bulk-rock compositions previously reported ( 206 Pb/ 204 Pb = , 206 Pb/ 204 Pb = , and 206 Pb/ 204 Pb = ) (Tanimizu and Ishikawa, 2006). We then obtained 204 Hg/ 202 Hg = ± during the same analytical runs (Table 2). When the measured 202 Hg/ 200 Hg = value was normalized to the IUPAC value of 202 Hg/ 200 Hg = 1.293, the measured 204 Hg/ 202 Hg ratio was estimated to be rather than the IUPAC predicted value of When the mass-fractionation factor in M% per mass unit ( M% pmu) notation (Hamelin et al., 1985) was applied, the Pb isotope analyses by LA-MC-ICP-MS with Faraday cups 373

6 measured 202 Hg/ 200 Hg had M% pmu = 0.499% from IUPAC ratio and calculated 204 Hg/ 202 Hg also had M% pmu = 0.499% showing identical mass-fractionation factor between the two isotope ratios (Tale 2). We therefore used the empirical correction factor 204 Hg/ 202 Hg = with 202 Hg/ 200 Hg = (IUPAC) for the isobaric overlap corrections of 204 Hg by monitoring 202 Hg/ 200 Hg (Table 2). We also measured raw 206 Pb/ 208 Pb isotope ratios by LA-MC-ICP-MS using SRM612 (n = 53) and calculated mass-fractionation factor using the reported ratio 206 Pb/ 208 Pb = for SRM612 (Kent, 2008b). We then obtained M% pmu = 0.496% (data not shown), in which the mass bias factor F = ± similar with the reported value of F = ± for Pb (Hirata, 1996). The mass-fractionation factor for Pb isotopes was almost similar with that for Hg isotopes (0.499%) and no residual mass-fractionation existed between Hg isotopes. Therefore, we concluded that the mass-fractionation in Hg originates from instrumental mass-fractionation but was not exactly identical with that in Pb as has been suggested in the previous study (Hirata, 1996). Note that part of the origin of the mass-fractionation of Hg can be natural mass-independent fractionation (Estrade et al., 2007; Jackson et al., 2006) but this is beyond the scope of this paper. Because we did not see non-systematic drift in the fractionation factor of Hg, so far, we believe that the fractionation occurred inside the mass spectrometer. The above approach is exactly the same as that used for 176 Yb overlap correction on 176 Hf in Hf isotope analysis, in which necessary isotope ratios among 173 Yb, 174 Yb, and 176 Yb were empirically determined and used for isobaric overlap corrections (Kemp et al., 2009; Kimura et al., 2012). Any drifts in the instrumental massbias factor were monitored by 202 Hg/ 200 Hg and were reflected in the 202 Hg/ 204 Hg ratios through the measured isotope ratios. The analytical results using this protocol were stable, indicating stable mass-fractionation among three Hg isotopes. The evaluation will be given below. TRANSIENT SIGNALS Unlike ICs, FCs have a delayed response against transient signals (Hirata et al., 2003; Pettke et al., 2011). This has been observed in the growing and decaying signals before and after continuous plateau signals during LA- MC-ICP-MS analyses (Hirata et al., 2003; Pettke et al., 2011). We used a Ω resistor amplifier for 204 M, and this had a slower response against transient signals. For a normal ablation protocol, we started data acquisition 5 s after LA started. This eliminated the slow response of the first growing transient signals from LA. LA continued until slightly after 2 min of data acquisition (24 scans), so that relatively flat plateau signals were acquired for the groundmass and mineral samples that had sufficient sample volume. In contrast to the samples with sufficient volume, analyses of melt inclusions in olivines did not always allow flat plateau signals because the sample volumes were limited. In this study, we analysed melt inclusions in olivines from Pitcairn Island lavas. The melt inclusions had diameters ranging from 100 to 300 µm (mostly 200 µm; Fig. 1C), and we ablated them from both the exposed surface of an inclusion and from the surface of the host olivine to drill through the melt inclusions using a 200 µm laser crater. In both cases, all transient signals from the start of LA and the end of total consumption of the melt inclusions were acquired because of the limited sample volume (Figs. 1D F). The repetition rate was adjusted from 5 to 20 Hz, depending on the sizes and depths of the melt inclusions. The exposed melt inclusions (mostly resulting from breakage along the cleavages and cracks of the host olivine crystals) were ablated at 5 Hz because the use of 10 Hz quickly consumed the melt inclusions of diameter 100 µm in 4 5 scans, resulting in poor precision. Instead, use of prolonged signals at 5 Hz (8 20 scans) usually yielded better analytical precision. Melt inclusions moulded inside olivine crystals were ablated with a Hz repetition rate, depending on the depth and size of the inclusions. Note that the host olivine did not contain detectable Pb (or Hg), therefore, co-ablation of the host olivine was not detrimental to the analyses, similar to the analyses of fluid inclusions in quartz crystals (Günther et al., 1997). To examine the effect of transient signals, we observed the isotope ratios of all the transient signals from the melt inclusions. In the moulded melt inclusions, usually the first one or two scans (occurring over ~8 s) were affected by the slow response of the FCs (first scan in Figs. 1D F). Signals during the decay process after total consumption of the melt inclusions should have also been affected by the slow response. However, lowering of the signal intensities led to scattering of isotope ratios as a result of the counting statistics, and this effect hindered the effect of slow responses (scans in Figs. 1D F). In both cases, these data were discarded by monitoring the isotope ratios using 2SE statistics (see Figs. 1D F). Although only 5 15 scans were available for melt inclusions µm in diameter, the obtained isotope ratios exhibited <4 per mil errors for 207 Pb/ 206 Pb and 208 Pb/ 206 Pb, significantly better than those obtained by secondary ion mass spectrometry (SIMS) (<3000 ppm errors) (Saal et al., 2005) and normal-sensitivity LA-MC-ICP-MS (<10 20 per mil errors) (Paul et al., 2011). The isotope ratios of 20X Pb/ 204 Pb gave large errors of 1 27 per mil, with an average of seven errors per mil in 2SE. However, these were still comparable to or better than those obtained by 374 J. Kimura et al.

7 A JA-1/JB-2 BCR-2G D Pb/ 204 Pb Pb/ 204 Pb JA LA-MC-ICPMS Solution MC-ICP-MS 206Pb/ 204 Pb B BCR-2G Pb/ 204 Pb E Pb/ 204 Pb JA-1/JB Pb/ 204 Pb JA Pb/ 204 Pb C Pb/ 204 Pb F 208Pb/ 206 Pb BCR-2G JA-1/JB-2 Mass fractionation trend BCR-2G (Pb = 11 ppm) JA-1 (Pb = 6.55 ppm) JB-2 (Pb = 5.36 ppm) BCR-2G (Reference) JA-1 (Reference) JB-2 (Reference) 207 Pb/ 206 Pb Pb/ 206 Pb JA Pb/ 206 Pb Fig. 2. Pb isotope ratios obtained by LA-MC-ICP-MS for the groundmass of JA-1 and JB-2 and BR-2G glass standards. Panels D F show enlarged rectangular areas in panels A C with JA-1 data only. Errors are shown by thin crosses as 2SE (standard error). Positive correlations shown by thick shaded line in panels D F indicate mass-fractionation trends of the MC-ICP-MS. Pb isotope analyses by LA-MC-ICP-MS with Faraday cups 375

8 Table 3. Analytical results of geological standard glass and ground mass using LA-MC-ICP-MS Sample 206 Pb/ 204 Pb Error (2SE) 207 Pb/ 204 Pb Error (2SE) 208 Pb/ 204 Error (2SE) 207 Pb/ 206 Pb Error (2SE) 208 Pb/ 206 Pb Error (2SE) [Solution MC-ICP-MS/TIMS] BCR-2G (br) [LA-MC-ICP-MS] BCR-2G (gl) / / / / / 1.36 [Solution MC-ICP-MS/TIMS] JB-2 (br) [LA-MC-ICP-MS] JB-2 (gm) / / / / /2.16 [Solution MC-ICP-MS/TIMS] JA-1 (br) [LA-MC-ICP-MS] JA-1 (gm) / / / / / 0.05 Note: Errors are given as 2 standard deviations (2SD). Solution ICP-MS/TIMS values are from our own analytical results and compilation of the previous reports: br, bulk rock; gl, glass; gm, groundmass; Diff., difference; abs., absolute difference;, difference in per mil. See source data set for LA-MC-ICP-MS in Table S1. normal MC-ICP-MS using a single IC ( 204 Pb), which showed 3 30 per mil errors, with an average of nine errors per mil (Paul et al., 2011). ANALYTICAL RESULTS Based on the analytical protocols described above, we analysed the groundmass/basalt glass of JB-2, JA-1, and BCR-2G andesite lavas, groundmass and olivine melt inclusions in Pitcairn Island lavas, plagioclases in a Pitcairn Island lava, and clinopyroxenes in a St. Helena (SH) lava. All the analytical results are given in Tables 3 7 and Supplementary Tables S1 S4 and are given below. Groundmass composition of lavas To test the above analytical protocols, we first analysed the groundmass compositions of JA-1 and JB-2 from 500 µm 500 µm 100 µm craters to test the reproducibility and precision of the method by comparison with the results from reported bulk-rock Pb isotope compositions of JA-1 (Kimura et al., 2003; Koide and Nakamura, 1990; Nohda, 1999; Shimoda and Nohda, 1995; Tanimizu and Ishikawa, 2006) and JB-2 (Baker et al., 2004; Ishizuka et al., 2003; Kimura and Nakano, 2004; Tanimizu and Ishikawa, 2006). Analysis of BCR-2G synthetic glass was also performed and compared with the bulk-rock compositions obtained by MC-ICP-MS (Jochum et al., 2005; Weis et al., 2006). Figures 2A, B, D, and E show that all the 204 Pb-based isotope ratios were in good agreement with the bulk-rock compositions, within 1.4 per mil (2SE) (see Table 3 for representative result; all the data are shown in Table S1). The reproducibility was comparable to the inter laboratory reproducibility of conventional thermal ionization mass spectrometry (TIMS) with unspiked methods (Todt et al., 1996). Inclined positive correlation trends were always observed in the 20X Pb/ 204 Pb ratios indicating (1) residual instrumental massfractionation, or (2) inherited 204 Pb blank instability, or (3) the effect of a relatively low signal-to-noise (S/N) ratio in the IC in MC-ICP-MS or heterogeneity in SRM612 used as bracketing standard (Kent, 2008a, b) (see Figs. 2D and E). Application of correction factors using residual massfractionation of Hg (see above) accurately corrected Hg isobars on 204 M up to 50% of the 204 Pb signals (e.g., JB- 2 and JA-1). BCR-2G did not show significant Hg (<1% of Pb signals) after blank subtraction. Similar reproducibility between Hg-unaffected BCR-2G and Hg-affected JB-2 and JA-1 confirmed the robustness of our protocol. This condition was unchanged over two months, demonstrating that the use of the high-sensitivity instrumental setting was stable enough for the Hg isobar corrections. The mass-fractionation factors of the Pb isotopes measured using SRM612 changed significantly ( M% pmu = ~0.5) as a result of the different conditions at the interface cones and instrumental settings. In fact, the use of a large amount of sample ablation led to severe deposition of the samples on the skimmer cone, and this resulted in deterioration of the instrumental sensitivity and changes in the mass fractionation of Pb, along with changes in the 376 J. Kimura et al.

9 fractionation factors. Even so, a standard bracketing method could reasonably cancel out such changes, as shown by the average values in Figs. 2A F and by the Diff. (abs./ ) in Table 3, which are less than 1.4 per mil difference from the reference values. In this case, the mass-fractionation of Hg isotopes was not corrected for by the bracketing method. However, the overlap of Hg on 204 Pb was readily corrected by mass-fractionation correction of Hg using 200 Hg and 202 Hg and the determined fractionation correction factor between 202 Hg and 204 Hg (see Section Fractionation of Hg and Its Effect on Overlap Correction for 204 M ). Note that instrumental massfractionation of Pb isotopes was corrected for by standard-bracketing, but that for Hg isotopes was not. So, the empirical fractionation correction factor between 202 Hg and 204 Hg was employed. The measured 207 Pb/ 206 Pb and 208 Pb/ 206 Pb ratios were also in very good agreement with the reported bulk-rock ratios for JA-1 (Kimura et al., 2003; Koide and Nakamura, 1990; Nohda, 1999; Shimoda and Nohda, 1995; Tanimizu and Ishikawa, 2006) with Diff. ( ) < 0.05 per mil (Table 3 also see Fig. 2F). However, those for JB-2 (Diff. ( ) < 2.16 per mil in Table 3) and BCR-2G (Diff. ( ) < 1.36 per mil in Table 3) indicated large offsets from the reported bulk-rock values, even though the analytical precisions of the averaged values were less than 0.5 per mil. The deviation originated from the residual instrumental mass-fractionation in JA-1, as shown by the positively correlated variations in the analytical data, was within twice the amount of the analytical errors of JB-2 (<1 per mil; Fig. 2C). The analytical errors originating from both the analytical precision and the massfractionation are, however, smaller than the offsets found in JB-2 and BCR-2G from the preferred values. Therefore, the offsets may reflect real variations in the natural rocks because the variation trends were not in line with the mass-fractionation trend defined by the analyses of JA-1. If the measured rock powders (bulk-solution analysis) and the groundmass/glass (LA-MC-ICP-MS) had dissimilar isotope ratios for JB-2 and BCR-2G, the Pb isotope ratios were heterogeneous in these samples. Isotopic heterogeneity of the BCR-2 powder has also been reported and the heterogeneity was identical with or even greater than the observed variation in this study (Weis et al., 2006), so it is feasible that the BCR-2G fused glass used in this study also inherited such heterogeneity in the source powders. In contrast, JA-1 showed an almost perfect match between LA-MC-ICP-MS and solution results for the 207 Pb/ 206 Pb and 208 Pb/ 206 Pb ratios, indicating homogeneity of the lava. So far, only the JA-1 standard could be used for accessing analytical accuracy of LA-MC-ICP-MS. We need to perform further tests on analytical precision of the Pb 208Pb/ 206 Pb Solution-MC-ICP-MS (bulk rock) LA-MC-ICP-MS (groundmass) Mass fractionation trend PC-2 PC-46 PC Pb/ 206 Pb PC Fig. 3. Pb isotope ratios obtained by solution-mc-icp-ms for bulk-rock and by LA-MC-ICP-MS for the groundmass of Pitcairn Island basalts. All errors are shown as 2SE. Analytical errors for bulk rocks are smaller than the symbols. Shaded lines indicate mass-fractionation trends. isotopes using LA-MC-ICP-MS. To examine the analytical reproducibility further, the groundmass compositions of lava samples from Pitcairn Island were analysed and compared to the bulk-rock Pb isotopes compositions analysed by solution ICP-MS using a Thermo Scientific Neptune at IFREE/JAMSTEC (Miyazaki et al., 2012). The analytical results are shown in Fig. 3 and Table 4 for the bulk-rock and groundmass (representative bulk-rock and groundmass data are shown in Table 4; see all the bulkrock data in Table S2 and groundmass data in Table S3). The bulk-rock and groundmass results show excellent agreement for all isotopes Diff. ( ) < 0.72 per mil for 206 Pb/ 204 Pb, Diff. ( ) < 1.17 per mil for 207 Pb/ 204 Pb, Diff. ( ) < 0.81 per mil for 208 Pb/ 204 Pb, Diff. ( ) < 0.42 per mil for 207 Pb/ 206 Pb, and Diff. ( ) < 0.52 per mil for 208 Pb/ 206 Pb (Table 4). All the 207 Pb/ 206 Pb 208 Pb/ 206 Pb plots match well with the bulk-rock solution MC-ICP-MS results within the error of analytical precision or within the of mass-fractionation deviation (Fig. 3). All the analytical results for the bulk-rock and groundmass of the Pitcairn Island lavas are plotted in Fig. 4. Compared with the bulk-rock compositions from 17 lava samples, almost similar or slightly wider Pb isotopic variations were observed in the groundmass compositions from 49 lava samples (Fig. 4 and Tables S2 and S3). Considering the good reproducibility shown by the double analyses of bulk rock and groundmass, groundmass analysis using LA-MC-ICP-MS is a useful tool for screening the samples prior to bulk-rock analysis because the time required to analyse 50 groundmass spots is about one day. Pb isotope analyses by LA-MC-ICP-MS with Faraday cups 377

10 Table 4. Comparison between bulk rock solution-mc-icp-ms and by groundmass LA-MC-ICP-MS analytical results 206 Sample Pb/ Pb Error (2SE) Pb/ Pb Error (2SE) Pb/ Error (2SE) Pb/ Pb Error (2SE) Pb/ 206 Pb Error (2SE) PC-2 (br) PC-2 (gm) / / / / /0.05 PC-46 (br) PC-46-2 (gm) / / / / /0.21 PC-86 (br) PC-86 (gm) / / / / /0.52 PC-87A (br) PC-87-1 (gm) PC-87-2 (gm) Average* / / / / /0.12 Note: br, bulk rock; gm, groundmass; Diff., difference; abs., absolute difference;, difference in per mil. See all the analytical results by solution-mc-icp-ms in Table S2 and LA-MC-ICP-MS in Table S3. Olivine melt inclusions Melt (glass) inclusions in olivines are useful in studies of source variations in magmas because the geochemical variations in the source magma before mixing in the shallow magma chamber can be found. SIMS and LA-MC-ICP-MS with ICs have been used for this purpose (Paul et al., 2011; Saal et al., 2005). We analysed olivine melt inclusions from Pitcairn Island lavas with the analytical protocols noted above. Melt inclusions in the olivines varied in size, and we analysed melt inclusions with diameters of µm (mostly 200 µm). Transient signals were processed following the analytical protocol noted above (see the example in Figs. 1D F). This method allowed sampling of the melt inclusions without polishing, which saves time in sample preparation and saves the available volume of the melt inclusions for analysis. We did not measure the Pb contents of the melt inclusions because we had consumed all available amounts by using a 200 µm crater. However, the Pb concentrations can be approximated by comparing the 208 Pb signal intensities of the melt inclusions and those of SRM612 and assuming the same ablation rate. The analytical precisions of 206 Pb/ 204 Pb, 207 Pb/ 204 Pb, and 207 Pb/ 204 Pb (Tables 5 and S4) varied from 0.4 to 27 per mil for all three isotope ratios, and the isotopic variations almost overlapped in the bulk-rock and groundmass compositions within 2SE analytical errors (Figs. 4A and B). The highest precision achieved was 0.4 per mil from a melt inclusion of diameter ~300 µm. The lowest precision, ~27 per mil, was from melt inclusions ~100 µm in diameter (Table 5 and Table S4). The isotopic variation in the 206 Pb/ 204 Pb versus 207 Pb/ 204 Pb and the 206 Pb/ 204 Pb versus 207 Pb/ 204 Pb plots showed positive correlations, and the variation in olivine melt inclusions are narrower than all the isotopic variations in the bulk rock and groundmass (Figs. 4A and B). This narrow variation was because of olivine crystals were available from only one rock group of the Pitcairn Island samples (Tedside group, see Figs. 4A F). However, it is notable that the range of variations in the melt inclusions exceeded those of the bulk-rock and the groundmass of the host lavas when compared within the Tedside group (Figs. 4A C). Some melt inclusions with larger errors tended to give an oblique linear trend line relative to the horizontal bulk-rock trend line (Figs. 4A and B). The melt inclusions with larger errors could have been affected by the mass-fractionation lying at an oblique angle to the geochemical trend (see estimated mass-fractionation trend from JA-1, shown by thick solid lines in Figs. 4A C). However, some samples plotted out of range of the analytical uncertainties (2SE), and, thus, the steep trends could have been from natural isotopic variations. This should be explored further, but discussion of the origin of the wide variation is beyond the scope of this paper and will be discussed elsewhere. The analytical precisions of 207 Pb/ 206 Pb and 208 Pb/ 206 Pb (Table 5 and Table S4) varied by ~4 per mil for the two isotope ratios. The isotope compositions for the bulkrock and groundmass compositions clustered within 2SE analytical errors (Fig. 4C). The highest precision achieved was 0.1 per mil (2SE) from a melt inclusion of diameter ~300 µm (Table 5 and Table S4). The lowest precision was 4 per mil from a melt inclusion of diameter ~100 µm. The isotopic variations in the 207 Pb/ 206 Pb and 208 Pb/ 206 Pb plots showed a positive correlation comparable to the variation in the Tedside group (Fig. 4C). The range of variations in the melt inclusions exceeded that in the bulk- 378 J. Kimura et al.

11 16.80 A D Tedside group Tedside group 207Pb/ 204 Pb Pb/ 204 Pb B 206Pb/ 204 Pb Tedside group E 206Pb/ 204 Pb Tedside group 208Pb/ 204 Pb Pb/ 204 Pb C 206 Pb/ 204 Pb Tedside group F 206 Pb/ 204 Pb Tedside group 208 Pb/ 206 Pb Pb/ 206 Pb Pb/ 206 Pb Mass-fractionation trend Bulk rock Groundmass Ol melt inclusion Pb/ 206 Pb Fig. 4. Pb isotope ratios in bulk-rock (open circles), groundmass (open squares with errors), and olivine melt inclusion (solid circles with errors) samples from Pitcairn Island lavas. Panels A C from our results and those in panels D F from Paul et al. (2011) for olivine melt inclusions. Shaded areas indicate ranges of bulk-rock compositions in the Tedside samples. Thick shaded solid bars in panels A C indicate mass-fractionation trends determined by JA-1 (see Fig. 2). Errors are shown as 2SE. Pb isotope analyses by LA-MC-ICP-MS with Faraday cups 379

12 Table 5. Representative analytical results of olivine melt inclusion compositions from Pitcairn lavas Sample 206 Pb/ 204 Pb Error (2SE) 207 Pb/ 204 Pb Error (2SE) 208 Pb/ 204 Error (2SE) 207 Pb/ 206 Pb Error (2SE) 208 Pb/ 206 Pb Error (2SE) [LA-MC-ICP-MS] PC PC PC PC PC PC PC PC PC PC PC PC PC PC PC PC PC PC PC PC PC Note: All the analytical results are given in Table S4. Table 6. Analytical results of plagioclase and groundmass compositions from Pitcairn lavas Sample Pb (ppm) 207 Pb/ 206 Pb Error (2SE) 208 Pb/ 206 Pb Error (2SE) Plagioclase and groundmass of Pitcairn Island OIB PC32pl PC32pl PC32pl PC32pl PC32pl PC32pl PC32gm PC32gm PC32gm PC32gm Note: Errors are given as 2 standard errors (2SE). rock and the groundmass of the host lavas, similar to the 20X Pb/ 204 Pb isotope ratios (Figs. 4A and B). Some melt inclusion samples with larger errors tended to plot to the right of the linear trend line of the bulk-rock. These could reflect inappropriate instrumental mass-fractionation correction, in a manner similar to the 20X Pb/ 204 Pb isotope ratios (Fig. 4C). Pitcairn Island lavas have been analysed by SIMS and LA-MC-ICP-MS using ICs and FCs (Paul et al., 2011; Saal et al., 2005). As noted above, SIMS only provided the 207 Pb/ 206 Pb and 208 Pb/ 206 Pb ratios, and the precisions are lower than those obtained by LA-MC-ICP-MS, which used FCs for the major isotopes. We compared our results with the normal-sensitivity LA-MC-ICP-MS results (Paul et al., 2011). Their sample was from one of the submarine seamounts around Pitcairn Island and our samples were from on-land only, so the comparison would highlight natural variations. The dataset obtained by the two different LA-MC-ICP-MS methods agreed, within analytical uncertainties, with the Pb isotope compositions from our on-land bulk-rock samples (Figs. 4A F). The 208 Pb/ 204 Pb data appeared to be systematically lower than those of the seamount samples (Figs. 4B and E) and the 208 Pb/ 206 Pb data lay obliquely to the on-land lava trend (Figs. 4C and F). The origin of the low 208 Pb isotope composition in the seamount sample may reflect natural vari- 380 J. Kimura et al.

13 Table 7. Analytical results of clinopyroxene and groundmass compositions from St. Helena lavas Sample Pb (ppm) 207 Pb/ 206 Pb Error (2SE) 208 Pb/ 206 Pb Error (2SE) Clinopyroxene and groundmass of St. Helena Island OIB SH35cpx SH35cpx SH35cpx SH35gm SH35gm Bulk rock Note: Errors are given as 2 standard errors (2SE). ation (Fig. 4F); however, discussion of this is beyond the scope of this paper. It is also worth noting that the Tedside group samples had an isotope ratio variation of in 207 Pb/ 206 Pb, and either this variation was reproduced or even wider variations were observed for olivine-hosted melt inclusions in three individual specimens PC-38, PC-40, and PC-87 (Fig. 4, Table 5, and Table S4). The same results were reported by Paul et al. (2011) for one single rock specimen from the Pitcairn seamount and we confirmed this with better resolution in terms of precision (Figs. 4C and F). This indicated that various melt sources in Pb isotopes were commonly present in the magma supply system beneath Pitcairn Island, and the melts were admixed in a shallow magma chamber during crystallization of olivines. Silicate minerals Minerals precipitated from magmas may record different information from that of the bulk-rock and groundmass compositions of the host lavas. This has been used to decode the evolutionary history of the magmas recorded in the chemical zoning of the crystals during crystal growth. This methodology is called isotope stratigraphy (Davidson et al., 2001), and Sr isotope compositions of plagioclase crystals have been a useful tool for this purpose. Application of Pb isotope stratigraphy to feldspar has also been reported (Kent and Dilles, 2005). In this study, we investigated applications of Pb isotope ratios using the analytical protocols developed in this study. Below, we report the analytical results for plagioclase crystals in Pitcairn Island lavas. Furthermore, we report, for the first time the analytical results of Pb isotope ratios in clinopyroxene crystals using a St. Helena lava sample. The analytical results are given in Tables 6 and 7, and Fig. 5. Plagioclase crystals in the Pitcairn Island lava (PC- 32) had a Pb isotope composition of ~ ppm, which was about three to five times lower than that in the groundmass composition of ppm Pb (the partition 208Pb/ 206 Pb 208Pb/ 206 Pb Mass-fractionation trend Plagioclase (PC-32) Groundmass (PC-32) Bulk rock A B Pitcairn Island OIB PC-31 (bulk) PC-32 (groundmass) PC-26 (bulk) 207 Pb/ 206 Pb St.Helena Island OIB PC-32 (plag) Mass-fractionation trend Clinopyroxene Groundmass Bulk Pb/ 206 Pb Fig. 5. Pb isotope compositions of plagioclase, groundmass, and bulk-rock of a Pitcairn lava sample (panel A) and of clinopyroxene, groundmass, and bulk-rock of a St. Helena lava sample (panel B). Errors are shown as 2SE. Shaded lines indicate mass-fractionation trends. Pb isotope analyses by LA-MC-ICP-MS with Faraday cups 381