The development of whole rock analysis of major and trace elements in XRF glass beads by fsla-icpms in GSJ geochemical reference samples

Transcription

1 Geochemical Journal, Vol. 45, pp. 387 to 416, 211 The development of whole rock analysis of major and trace elements in XRF glass beads by fsla-icpms in GSJ geochemical reference samples YOSHIAKI KON,* HIROYASU MURAKAMI, TETSUICHI TAKAGI and YASUSHI WATANABE Geological Survey of Japan, National Institute of Advanced Industrial Science and Technology, 1-1-1, Higashi, Tsukuba, Ibaraki , Japan (Received January 2, 211; Accepted July 5, 211) Using laser-ablation inductively coupled plasma mass spectrometer (LA-ICPMS), we have improved the reliability of the abundance data for trace-elements in geochemical samples using a glass bead ablation method. The glass beads were made of mixture of.1 g sample and 1. g of lithium-tetraborate preliminary prepared for an analysis of major components using a X-ray fluorescence (XRF) technique. The present method has several advantages: 1) higher sensitivity than that achieved by the XRF method, 2) obviation of erroneous measurements due to incomplete dissolution of heavy minerals, and 3) simple, rapid and user friendly sample preparation procedures for the analysis of both the major and trace elements. Development of this method constitute: 1) femtosecond laser-ablation for minimal elemental fractionation during the laser ablation, 2) new software to control all the laser, sample stage movement as well as triggering the data acquisition using the ICP-MS, and 3) a newly designed sample cell to enhance the transport efficiency of the sample aerosol into the ICP. Moreover, to improve the data quality for both the major and trace elements, calibration lines were defined based on the Li-normalized signal intensities and the reported abundance values for the analytes in well distributed GSJ geochemical reference samples. These improvements enabled us to analyze whole-rock compositions at 1 sec/sample. Using this method, the precisions of analyses were better than 1% for Na, Mg, Al, Si, K, Ca, Sc, Ti, V, Mn, Co, Ga, Rb, Sr, Y, Zr, Nb, La, Pr and Nd; 2% for P, Zn, Sn, Ce, Sm, Gd, Tb, Dy, Ho, Er, Tm, Yb, Lu, Hf, Ta, Th, and U, and 3% for Fe, Cs, Ba, and Eu. For Ni and As, precisions of the measurements was not better than 3%. lities of analyses were better than 1% for Na, Mg, Al, Si, K, Ca, Sc, Ti, V, Mn, Co, Rb, Sr, Y, Zr, Cs, Ba, La, Ce, Nd, Sm, Gd, Tb, Dy, Yb, Lu, Hf, Th and U; <2% for Zn, Ga, Nb, Sn, Pr, Sm, Eu, Ho, Er, Tm and Ta, and 3% for P. For Fe, Ni and As, reliability of the measurement was not better than 3%. Keywords: bulk chemistry, major and trace elements, glass beads, femtosecond laser, LA-ICP-MS INTRODUCTION Chemical compositions and the mineral modal compositions has been a principal information for petrological and geochemical studies on igneous rocks. For the analysis of chemical composition of the rock samples, X-ray fluorescence (XRF) spectroscopy on the glass beads has been widely used because of both the high reliability in the resulting data and the high analysis throughput (Norrish and Hutton, 1969). In the case of the XRF analyses using glass beads, mixing ratios of ranging from 1:1 to 1:5 (rock : flux) were widely employed to avoid the incomplete decomposition of heavy minerals (e.g., Norrish and Hutton, 1969; Goto, 1976; Sugisaki et al., 1977). For trace element determinations, the XRF technique utilizing a pressed powder-pellets or low dilution- *Corresponding author ( yoshiaki-kon@aist.go.jp) Copyright 211 by The Geochemical Society of Japan. glass bead (mixing ratio of ca. 1:2) were used (e.g., Sugisaki et al., 1981; Kimura and Yamada, 1996). These analyses, however, require longer analysis time (about 2 minutes per sample), and the reliability of the resulting abundance data for trace elements including the rare earth element (REE) was sometime not high enough to derive the detailed petrological information. To improve the analytical precision of the measurements and also to improve the analysis throughput, solution-based ICP-mass spectrometry (ICP-MS) has been widely used for the trace-element determinations. The major drawback in the solution-based ICP-MS technique is that this analytical approach requires highly complicated and time consuming procedures for chemical decomposition and separation procedures. In the case of the solution-based ICP- MS technique, the mass spectrometric interferences by oxide polyatomic ion (MO + ) can cause systematical error in the resulting abundance values. Moreover, great care must be taken to avoid the incomplete decomposition of refractory minerals, which causes the systematical errors in the resulting abundance values for some elements 387

2 (e.g., Sc, Zn and Eu) (e.g., Longerich et al., 199; Hirata et al., 1988; Imai, 199; Yoshida et al., 1992; Ujiie and Imai, 1995). To overcome these features, laser ablation sample introduction technique combined with highsensitivity ICP-MS instruments has been widely used for the abundance measurements of trace elements (Jarvis and Williams, 1993; Perkins et al., 1993; Nesbitt et al., 1997, Orihashi and Hirata, 23; Kurosawa et al., 26). With the powdered pellets, the analytical uncertainty in REE abundance of igneous rocks was as high as 2% (RSD), due to the effect of elemental inhomogeneity in these pellets (Jarvis and Williams, 1993). In the case of the fused glasses, which are made from entirely of rock powder alone, the sample glasses can become intrinsically homogeneous for the mafic rocks, erroneous measurements caused by the possible sample heterogeneity can be minimized. However, the inhomogeneity of felsic-rock samples still remains when using this fused-glass method, because felsic rocks frequently include refractory minerals such as zircon and monazite, and their melt is highly viscous. Low dilution-ratio glass beads, which are made of lithium tetraborate flux and rock powder, can easily be homogenized even in felsic rocks (e.g., Kimura and Yamada, 1996). Major drawback in these analytical approaches require an internal normalization correction to calibrate the concentration of analytes (e.g., Si, Ca, Sr), and the abundance values for the internal standardization elements must be determined by separate analytical method such as XRF analysis (e.g., Günther et al., 21; Orihashi and Hirata, 23). To overcome these drawbacks, we have developed a new calibration method using a high dilution-ratio glass beads (1 flux to 1 rock, prepared for major element analysis by XRF) for major and trace element using the LA-ICPMS technique. To improve the analytical precision and reliability of the data, abundance values for the analytes were calibrated based on the internal standardization using the Li. Hence, it should be noted that no calibration was required for the Li abundance value. Moreover, we have developed a new laser ablation system using a Ti:S femtosecond laser to improve the analytical sensitivity and precision of the measurements. In this paper, we report the details of this new analytical method and the results of our analyses compared with the recommended values of GSJ geochemical reference samples. ANALYTICAL PROCEDURE Sample and sample preparation We measured trace and REE abundances in twelve rock reference materials (JA1, JA2, JA3, JB1b, JB2, JB3, JG1a, JG2, JG3, JR1, JR2 and ). These 12 reference materials are widely distributed by the GSJ and are well characterized in a wide range of chemical compositions 115 In-normalized signal intensity Analog Pulse % % 1% Nd concentration (ppb) Difference Fig. 1. Relationship between the 115 In-normalized Nd-signal intensity and the Nd concentration of standard solutions and differences from regression values. There were no significant deviations between pulse (open circles) and analog (filled circles) modes. from mafic to felsic rocks (Imai et al., 1995, 1999; Terashima et al., 1998). The split/position numbers of the analyzed reference materials in this study were JA-1 (2/ 58-3), JA-2 (5/75) and JA3 (4/33-4) as andesite, JB1b (5/ 84), JB2 (7/5-5) and JB3 (1/4) as basalt, JG-1a (2/4), JG2 (4/75) and JG3 (6/99) as granite, and JR1 (1/66), JR2 (7/42) and (2/19) as rhyolite. Glass beads (1:1) for the analyses were prepared by mixing.5 g of sample powder with 5. g of lithium tetraborate (Li 2 B 4 O 7, Spectromelt A1) flux. The resulting sample mixture was heated to 1,2 C for 15 minutes in a 95% Pt-5% Au crucible with 3 mm inner diameters, using an automated high-frequency bead sampler (Tokyo-Kagaku Co., Ltd., Tokyo). Instruments and analytical conditions of fsla-icpms The ICP-MS instrument used in this study was a quadruple-based ICP-MS (Agilent 75cx: Agilent Technology, Japan). The in-house laser-ablation system was composed of a Ti:S femtosecond (fs) laser (IFRIT: Cyber Laser Inc., Japan) as a 78 nm NIR (near infrared) light source, a micro-step XYZ electric-stage (CAVE-X series: Suruga Seiki Co., Ltd., Japan), a sample cell and a sample holder (T21K: Suzu-shin Industry, Japan). Both the ICP-MS and fsla system was controlled by the in-house universal operating software (Laser Ablation: OK-Lab., Japan). Prior to the analysis sequence, position of the laser ablation pit can be set and memorized, and the data acquisition can be made automatically. For glass bead analysis, total 15 samples and/or standards were placed on the sample holder at batch. Flow rates of carrier gas and the ion-lens setting of the ICP-MS were tuned to maximize the signal intensity 388 Y. Kon et al.

3 Table 1. LA-ICPMS operating conditions ICP-MS Instrument: Agilent 75cx (Agilent Technology) Forward power: 16 W Gas flow rate: cool 15. l/min auxiliary 1. l/min nebuliser l/min carrior He.8 ml/min Scanning mode: peak jump Analyses mode: time resolved analyses (TRA) Integration time: 1 sec/sample (2 sec for gasblank, 3 sec for ablation, 3 sec for washout and 2 sec for gasblank) Dwell time: Sweep time: Oxide ratio: Monitor: 5 msec for Li, Na, Mg, Al, Si, K, Ca, Fe; 1 msec for Ti, Rb, Sr, Y, Zr, Ba, La, Ce, Nd, Sm, Gd, Dy, Er, Yb; and 2 msec for the other elements.8 sec.1.3% for ThO/Th 6 Li, 23 Na, 24 Mg, 27 Al, 29 Si, 31 P, 39 K, 42 Ca, 45 Sc, 49 Ti, 51 V, 55 Mn, 57 Fe, 59 Co, 6 Ni, 66 Zn, 69 Ga, 75 As, 85 Rb, 88 Sr, 89 Y, 9 Zr, 93 Nb, 118 Sn, 133 Cs, 137 Ba, 139 La, 14 Ce, 141 Pr, 146 Nd, 147 Sm, 153 Eu, 157 Gd, 159 Tb, 163 Dy, 165 Ho, 166 Er, 169 Tm, 172 Yb, 175 Lu, 178 Hf, 181 Ta, 232 Th, 238 U, 248 ThO Detector mode: analog mode for Li, Na, Mg, Al, Si, K, Ca and Fe pulse-counting/analog automatic switching mode for the other elements Laser ablation Instrument: TiS femtosecond (IFRIT, Cyber Laser Inc.) Wave length: 78 nm (NIR) Pulse energy: 1 µj/cm 2 Crater size: 2 µm Repetition rate: 1 Hz for glassbead, 1 Hz for NIST glass Rasetred area: 5 5 µm for glassbead, µm for NIST glass Raster speed: 5 µm/sec for glassbead, 5 µm/sec for NIST glass Sample cell Microstep XYZ-stage Operation software T2K (Suzu-shin Industry) CAVE-X series (Suruga Seiki) Laser Ablation (OK Lab.) of La and to minimize the oxide production rate ( 232 Th 16 O/ 232 Th). We analyzed 1 major elements and 33 trace and rare earth elements for the rock reference materials. The 7 Li signal was monitored for the internal standardization. Dwell times were 5 ms for 6 Li, 23 Na, 24 Mg, 24 Al, 29 Si, 39 K, 42 Ca, and 57 Fe, 1 ms for 49 Ti, 85 Rb, 88 Sr, 89 Y, 9 Zr, 137 Ba, 139 La, 14 Ce, 146 Nd, 147 Sm, 157 Gd, 163 Dy, 166 Er, and 172 Yb, and 2 ms for 31 P, 45 Sc, 51 V, 55 Mn, 59 Co, 6 Ni, 66 Zn, 69 Ga, 75 As, 93 Nb, 118 Sn, 133 Cs, 141 Pr, 153 Eu, 159 Tb, 165 Ho, 169 Tm, 175 Lu, 178 Hf, 181 Ta, 232 Th, 238 U, and 232 Th 16 O, resulting in a relatively long sweep-time of about.8 sec. To minimize signal fluctuation during sweeps, a mixingcell was placed between the sample cell and ICP plasma (Tunheng and Hirata, 24). Using this mixing-cell, the washout time of sample-aerosol was extended from.3 sec to 2 sec. For monitoring the abundant elements such as major elements or Li, the detector mode of the ICP- MS was switched to an analog mode, and the combination of analogue and pulse-counting mode was used for monitoring the minor elements. In the case of monitoring the trace elements, a pulse counting mode was used for the most samples, except for Mn, Rb, Y, Zr, Nb, La and Ce in. For cross calibration between the pulse counting and analog modes, we carefully optimized the operational settings for pulse counting/analog (P/A) factor. Figure 1 shows the correlation between 115 In-normalized signal intensity and the concentration of Nd-standard solution after the analysis of glass beads. Hence, the signal intensity of 115 In ranged from 85, to 9, cps. The difference of 115 In-normalized signal intensity of 148 Nd from the regression value was maintained within 5% both the pulse and analogue analyses. The laser ablation was made under the diameter pit size of ~2 µm, a pulse energy of 1 µj/cm 2, a laser emission repetition rate of Glass bead analysis by fsla-icpms 389

4 1 Hz and total ablation time of 3 s. With the present ablation conditions, the resulting depth of the ablation pit was about 1 µm after the.1 sec (Hirata and Kon, 28), demonstrating the more effective ablation efficiency for glass beads than those achieved by the conventional nano-second lasers. To minimize elemental fractionation during ablation, the ablation spot was moved at 5 µm/sec during the analysis. Total ablation area was 5 5 µm square on the glass bead samples. The estimated ablation volume was about µm 3. For the signal intensity measurement on the NIST SRM61, the pulse repetition rate, raster speed and size of ablation area were re-optimized to 1 Hz, 5 µm/sec and 2 µm square, respectively, to adjust the signal intensities of the analytes. The operational settings employed in this study was summarized in Table. 1. DATA CALCULATION To obtain stable signal intensity profile under the optimal sample sizes, raster-ablation, rather than the single spot ablation was employed throughout the analysis. In order to reduce the contribution of the signal spikes onto the analytical precisions, time resolved analysis (TRA) was employed as the data acquisition mode. Erroneous measurements due to laser ablation of secondary phase or heterogeneous area within the samples could be minimized by screening the signals found in the TRA profile. After the background correction based on the baseline subtraction protocols, Li-normalized signal intensity data for each element were obtained. The repeatability of Linormalized signal intensity was estimated from weighted variance during a single glass-bead ablation (N ~ 5). To quantify the concentration of each element, calibration lines were made by Li-normalized signal intensities and the reference values of each standard. The reliability of analyses for each element was estimated by the averaged difference between each analysis and the reference values. To estimate the repeatability of the calibration lines, we first performed five set of duplicate measurements for 12 glass beads as standards. Subsequently, we re-calculated the calibration lines using only three standards (JB3, JG3 and ) to estimate the repeatability of these analyses for unknown samples. Details of these calculations are shown below. Gas blank subtraction, Li-normalization and data screening Background signal intensity obtained without laser ablation (gas blank intensity) was continuously changed through the analysis sequence. To minimize the erroneous measurement due to changes of the background signals, the gas blank signals were obtained before and after ablation of each sample (Fig. 2a), and the average of these (a) 1e6 Intensity of element "M" (b) Signal ratio (M/ 6 Li) 1e5 1e t Time (sec) GB -2s 2s [M] ave-int GB ablation t [M] cr-int t R 8 1s [M] ave-int ablation GB GB 2 8s R + 2 8s R 2 8s R' + 3SD 2 8s R' 2 8s R 2 8s R' 3SD 2 8s. R 2 8s R 2 t Time (sec) Fig. 2. (a) Time profile of signal intensity. The averaged intensity of the gas blank was calculated from data at 2 sec and 8 1 sec (grey areas). (b) Time profile of signal intensity ratio (black-line plot). The gray-line plot shows signal intensity of element M (no scale). The light-gray area shows the extent of 2 8s R ± 2 8s R. The outliers of this area were rejected for subsequent calculations. The dark-gray area represents the extent of 2 8s R ± 3 SD. The outliers of this area were also rejected. two background signals were used for the gas blank subtraction as expressed by t [M ] int = t [M] int ( 2s [M] ave-int + 8 1s [M] ave-int )/2 (1) where t [M ] int is the blank subtracted signal intensity of the element M, [M] int is the signal intensity of element M at time t, and 2s [M] ave-int and 8 1s [M] ave-int are the averaged intensities of element M during 2 s and 8 1 s, respectively. The gas blank subtraction was followed by the reduction of the contribution of signal spikes. This procedures begins with the calculation of the intensity ratio for analytes to 6 Li. Hence, we introduced a new variance t R defined by 39 Y. Kon et al.

5 t R = t [M ] int / t [ 6 Li ] -int (2) hence 2 8s R represents the M/ 6 Li isotope ratio calculated based on the averaged signal intensity from 2 to 8 s (Fig. 2a). Signal spikes was defined as signal intensity data having large deviation from the averaged ratio values, and hence, the signal intensity data having relatively large deviation (>3 SD) was removed as signal spikes (Fig. 2b). Thus the signal intensity data having 2 8s R 3 SD < t R < 2 8s R + 3 SD (3) were used for the further calculations. It should be noted that the standard deviations were calculated by the weighted mean of the t R, in which weight of the individual data were estimated by the total intensity per sweep. Here, the weight (W i ) of signal-intensity ratio of a sweep (R i ) was determined by total signal-intensity during a sweep (i). 2 8 s 2 ( i) i i ( ) i ( ) 3SD = 3 N R R W N 1 W. 4 After the removal of the signal spikes, Li-normalized signal intensity data for analaytes ([M ] t-int /[Li ] t-int ) were then used for the calculation of abundance values of the analytes. The repeatability of the t R was estimated by weighted standard error (2 SE, N ~ 5). 2 8 s 2 ( i) i i 2SE = 2 R R W N 1 W. 5 i ( ) i ( ) The glass beads analyzed in this study contains 91wt% Li 2 B 4 O 7 (~8 ppm of Li), and this is almost 1 times higher than that for natural rocks. This suggests that the contribution of Li from rock samples was negligibly small (<.25%), and resulting the Li contents for all glass bead samples analysed in this study was assumed to be constant. Thus, the signal intensities of analytes ( t [M ] int ) were normalized by 6 Li signal intensity ( t [ 6 Li ] int ) and the Li content, which was assumed to be constant (8 ppm), and the Li-normalized signal intensity data for the analytes could be calculated by [M] t-int /[Li] t-int = t [M ] int / t [ 6 Li ] int 8, (const.). (6) i Oxide-interference correction To obtain reliable abundance data for trace-elements, great care must be given in the mass spectrometric interference by oxide polyatomic signals (e.g., LiO +, SiO +, and BaO + ) (Longerich et al., 199). To evaluate the possible contribution of the mass spectrometric interferences, we have investigated the formation ratio of the oxide signal (MO + /M + ) by monitoring the 232 Th 16 O/ 232 Th ratio. Although the measured 232 Th 16 O/ 232 Th ratio should be generally lower than.2%, contribution of the interferences can cause the systematical error in the measured abundance values for some trace elements such as 23 Na ( 7 Li 16 O/ 23 Na 2%), 45 Sc ( 29 Si 16 O/ 45 Sc 2%), 6 Ni ( 44 Ca 16 O/ 6 Ni 6%), 66 Zn ( 5 Da 16 O/ 66 Zn 5%; 5 Da = 5 Ti + 5 V + 5 Cr), 75 As ( 59 Co 16 O/ 75 As 1%) or 153 Eu ( 137 Ba 16 O/ 153 Eu 5%). In this study, the oxide-interferences for analytes were corrected by the calculated signal intensity of MO + signals based on the assuming that there were no significant difference in the MO + /M + ratios for all the elements, using the following equation. [ n M 1 ] int = [ n Da] int - [ n-16 M 2 ] int [ 232 Th 16 O] int /[ 232 Th] int(7) where [ n M 1 ] int is the corrected signal intensity of an interfering element, [ n Da] int is the total signal intensity of mass-number n, [ n-16 M 2 ] int is the signal intensity of the interfering elements, and the [ 232 Th 16 O] int /[ 232 Th] int is the oxide production ratio found in the present instrumentations and operational settings. If in the case that the [ n-16 M 2 ] int was not monitored, the signal intensity of [ n-16 M 2 ] int was calculated by the signal intensity data for other isotopes of the same elements M 2. Approximation of calibration lines To determine the concentration of each element, we made calibration lines using Li-normalized signal intensities and the reference value for each standard. We did not measure the blank of the analytes in the LBO flux. However, no correction for the procedural blank was made throughout the study, because the contribution of the blank from the LBO flux should be identical for both the sample and standard. The slope (b) and intercept (a) of the calibration lines were calculated by weighted least-square (LSQ) fitting (Miller and Miller, 21). The x-axis is the reference value of the samples (x i ) and the y-axis is the Li-normalized intensity (y i ). In this study, compilation values (C.V.) based on the previously published data (Imai et al., 1995, 1999; Terashima et al., 1998) were adopted as a concentration values for the analytes. i i i w w i i 2 2 w i i b = w x y nx y w x nx () 8 a y bx =. ( 9) w w Weighted median-points (x w, y w ) were calculated from the Glass bead analysis by fsla-icpms 391

6 Na Mg Al JA Si P K JA1 JA3 JB2 JA Legend Ca Sc Ti V Compilation value Imai (1995, 99) and Terashima et al. (1997) Approximate line Published value Dulski (21) Orihashi and Hirata (23) Awaji et al. (26) Makishima and Nakamura (26) and Lu et al. (27) Shimizu et al. (21) JB Reference value Reference value Reference value Reference value Li-normalized intensity Li-normalized intensity Li-normalized intensity 392 Y. Kon et al.

7 Mn Fe Co Ni 5 4 JB2 JA2 JA1 JA3 JB1b JA JA3 JB2 Zn Ga As Rb Sr Y Zr Nb JG3 JB JG JG3 JG JA2 JG1a Reference value Reference value Reference value Reference value Fig. 3. Calibration line using GSJ geochemical reference samples. Li-normalized intensity Li-normalized intensity Li-normalized intensity Glass bead analysis by fsla-icpms 393

8 Sn Cs Ba La JR1 JG1a JG1a JG1a JG3 JG Pr Eu Ho Yb 1 JG2 JA JG1a JB3 JA JG2 JR2 JR JG2 JR2 JR Hf Ta Th U JG3 JG2 JG1a JA2 JG JG3 JG1a Reference value Reference value Reference value Reference value Fig. 3. (continued). Li-normalized intensity Li-normalized intensity Li-normalized intensity 394 Y. Kon et al.

9 Table 2. Compilation values (C.V.) of GSJ geochemical reference samples, averaged regression values (wt%, ppm), its repeatabilities and trueness Na 2 O C.V. Regression value JB1b ±.5 1.7%.% % JB1b 2.59 ± % 1.5 JB ±.1.5%.2%.1 5.1% JB ± % 2.4 JB ±.8 3.% 1.% % JB3(STD) 2.72 ±.2.8%.4 JA ±.1 2.6%.1%.2 5.2% JA ± %.6 JA ±.3 1.1%.1% % JA2 3.7 ±.9 2.9% 1.3 JA ±.5 1.5%.1% % JA ±.5 1.7% 1. JG1a ±.7 2.% 1.2% % JG1a 3.32 ±.9 2.7% 2.2 JG ±.5 1.4%.4% % JG ±.5 1.4%.5 JG ± % 2.% % JG3(STD) 4.1 ±.8 2.% 1.3 JR ±.7 1.8% 1.5% % JR ±.1 2.5% 2.1 JR ±.6 1.4%.% % JR ±.8 2.% ±.8 1.8%.6% % (STD) 4.64 ±.6 1.3% 1. Mean value 1.8%.6% 5.3% 2.8% 1.2 Slope 2386 ± 474 Slope 222 ± 1551 Precison 5.3 Intercept 4 ± 2252 Intercept 2787 ± 5783 lity 1.2 MgO C.V. Regression value JB1b ± % 2.4% % JB1b 8.35 ± % 2.6 JB ± %.4%.23 5.% JB ± %.6 JB ± %.5% % JB3(STD) 5.23 ±.9 1.8%.7 JA ±.2 1.1%.8%.9 5.6% JA ±.3 1.7%.5 JA ±.7.9%.8% % JA ± % 1. JA ±.9 2.4%.3%.19 5.% JA ± %.5 JG1a ±.2 2.7% 2.7%.3 4.8% JG1a.68 ±.2 3.3% 2.1 JG2.4.4 ± % 2.8%. 9.6% JG2.4 ± % 9.8 JG ±.5 2.6% 1.1%.8 4.3% JG3(STD) 1.78 ±.4 2.4%.8 JR ±. 3.3% 2.2%.1 4.5% JR1.13 ±.1 4.7% 4.5 JR2.4.4 ±. 6.7% 3.6%. 5.6% JR2.4 ±. 4.% ±. 3.9% 5.2%. 5.8% (STD).5 ±..1%. Mean value 3.7% 1.9% 5.4% 3.3% 2.8 Slope 8419 ± 118 Slope 844 ± 177 Precison 5.4 Intercept 22 ± 7 Intercept 1 ± 15 lity 2.8 Al 2 O 3 C.V. Regression value li % JB1b ±.3 2.1% 1.1% % JB1b 14.2 ± % 1.2% JB ±.45 3.% 1.6% % JB ±.6 4.% 1.4% JB ±.4 2.3%.4% % JB3(STD) ± %.1% JA ± %.8% % JA ±.3 2.%.6% JA ± %.4% % JA ± %.5% JA ± % 2.3% % JA ± % 2.1% JG1a ± % 3.5% % JG1a 13.8 ± % 3.5% JG ±.2 1.6%.4% % JG ± %.4% JG ±.31 2.%.1%.7 4.5% JG3(STD) ± %.1% JR ± % 1.2% % JR ± % 1.1% JR ± %.7% % JR ±.2 1.6%.7% ± %.1% % (STD) 11.9 ±.4.3%.% Mean value 2.2% 1.% 4.7% 2.4% 1.% Slope 792 ± 176 Slope 88 ± 859 Precison 4.7% Intercept 1241 ± Intercept 2384 ± lity 1.% Glass bead analysis by fsla-icpms 395

10 Table 2. (continued) SiO 2 C.V. Regression value JB1b ± %.% % JB1b 5.96 ± %.3 JB ± %.7% % JB ± %.8 JB ± %.% % JB3(STD) 5.86 ±.19.4%.2 JA ± % 1.9% % JA ± % 2.3 JA ±.5.9%.3% % JA ± %.2 JA ± % 1.% % JA ± % 1.3 JG1a ± % 1.1% % JG1a ± %.5 JG ± % 2.1% % JG ± % 3.1 JG ± %.1% % JG3(STD) ± %.6 JR ± %.7% % JR ± %.2 JR ± %.1% % JR ± % ±.6.8% 1.4% % (STD) ± %.7 Mean value 2.5%.8% 5.% 3.1%.9 Slope 132 ± 13 Slope 129 ± 13 Precison 5. Intercept 352 ± 359 Intercept 551 ± 87 lity.9 P 2 O 5 C.V. Regression value JB1b ±.5 2.% 1.8% % JB1b.249 ±.8 3.4% 2.7 JB ±.4 3.7%.8%.8 7.6% JB2.96 ±.6 5.8% 4.9 JB ±.1 3.3% 3.7% % JB3(STD).291 ± %.9 JA ±.6 3.8% 1.2% % JA1.16 ±.6 4.% 3.2 JA ±.2 1.3% 4.8% % JA2.147 ±.5 3.4%.4 JA ±.3 2.8%.8%.9 7.6% JA3.112 ±.5 4.6% 3.5 JG1a ±.3 3.5% 5.8%.5 6.1% JG1a.75 ±.4 4.8% 9.6 JG2.2.6 ± % 179.1%.1 2.9% JG2.6 ± % JG ±.9 7.3% 4.8%.1 8.1% JG3(STD).122 ±.7 5.7%.4 JR ±.4 2.6% 13.5% % JR1.18 ± % 15.5 JR ±.1 15.% 29.5%.2 19.% JR2.8 ± % ± % 2.%.2 1.1% (STD).17 ±. 1.5%.2 Mean value 7.5% 2.7% 1.2% 1.7% 21.4 Slope 153 ± 43 Slope 163 ± 7 Precison 1.7 Intercept 6 ± 4 Intercept 5 ± 5 lity 21.4 K 2 O C.V. Regression value JB1b ±.4 2.9% 2.9%.9 6.6% JB1b 1.3 ±.3 2.7% 1.5 JB ±.1 2.5%.6%.2 5.9% JB2.44 ±.3 7.4% 4.9 JB ±.2 2.7% 2.4%.5 6.2% JB3(STD).78 ±..1%. JA ±.1 1.4% 2.1%.4 5.4% JA1.8 ±.2 2.5% 4.4 JA ±.4 2.5% 1.% % JA ±.5 3.%. JA ±.4 2.6%.8%.9 6.2% JA ±.5 3.5% 2. JG1a ± %.9% % JG1a 3.94 ± %.5 JG ± %.4% % JG ± %.6 JG ±.4 1.7%.3% % JG3(STD) 2.65 ±.2.8%.3 JR ± % 1.6% % JR ±.1 2.2% 1.9 JR ±.6 1.2% 1.4% % JR ±.1 2.1% ± % 1.% % (STD) 4.26 ±.8 1.9%.7 Mean value 2.4% 1.3% 6.% 2.8% 1.6 Slope ± 37 Slope ± 738 Precison 6. Intercept 99 ± 331 Intercept 461 ± 761 lity Y. Kon et al.

11 Table 2. (continued) CaO C.V. Regression value JB1b ±.19 2.%.8% % JB1b 9.47 ± % 1.4 JB ± %.%.4 4.1% JB ±.2 2.1%.6 JB ± %.5% % JB3(STD) 9.78 ± %.1 JA ± %.5% % JA ±.17 3.% 1.1 JA ± % 2.5% % JA2 6.1 ± % 3. JA ± %.6% % JA ± %. JG1a ±.6 2.9% 2.2% % JG1a 2.7 ±.6 2.8% 2.6 JG ± % 1.5%.5 6.4% JG2.71 ± % 1.7 JG ±.2.6%.6% % JG3(STD) 3.69 ±.4 1.1%.1 JR ±.6 8.4% 4.9%.3 4.8% JR1.7 ±.7 1.3% 5.2 JR2.5.5 ±.4 7.7%.4%.2 4.4% JR2.5 ±.4 7.4% ± % 6.3%.1 6.9% (STD).9 ±..3%.1 Mean value 5.2% 1.7% 5.2% 4.2% 1.4 Slope 12 ± 2 Slope 12 ± 3 Precison 5.2 Intercept 18 ± 4 Intercept 17 ± 3 lity 1.7 Sc C.V. Regression value JB1b 29.2 ±.4 1.4% % JB1b 28.9 ±.7 2.6% JB ± % 4.3% % JB ± % 3.4 JB ±.7 2.1%.9% % JB3(STD) 33.8 ±..%. JA ±.8 2.8% 2.% % JA ± % 1. JA ±.7 4.% 4.6%.8 4.4% JA ±.3 1.7% 5.4 JA ±.8 3.9% 5.6% % JA3 2.6 ± % 6.3 JG1a ±.4 5.8% 4.%.3 4.3% JG1a 6.4 ± % 3.5 JG2 2.4 n.d. ± JG2 n.d. ± JG ± %.6%.4 4.9% JG3(STD) 8.8 ±..%. JR ±.4 7.8% 1.9%.2 4.6% JR1 5.2 ± % 1.6 JR ±.1 2.7% 2.7%.2 4.1% JR2 5.4 ± % n.d. ± n.d. ± Mean value 4.6% 2.9% 4.5% 9.% 2.7 Slope 9 ± Slope 8 ± Precison 9. Intercept 22 ± 31 Intercept 24 ± 26 lity 2.9 TiO 2 C.V. Regression value JB1b ± % 1.6% % JB1b ±.24 2.% 2.7 JB ± %.3%.6 5.% JB ± %.8 JB ± % 1.5% % JB3(STD) ±.9.6%.3 JA ± % 3.4% % JA1.869 ± % 2.2 JA ± % 2.7% % JA2.669 ± % 1.4 JA ± %.9% % JA3.697 ± %.4 JG1a ±.9 3.9% 4.8%.1 4.2% JG1a.234 ± % 6.5 JG ±.4 7.9% 9.9%.3 6.4% JG2.46 ± % 3.9 JG ± %.9% % JG3(STD).478 ±.6 1.2%.5 JR ±.4 3.7% 1.%.5 5.% JR1.16 ± % 3.9 JR ±.3 4.8% 6.2%.3 5.% JR2.63 ± % ±.7 3.1% 2.3% % (STD).211 ±.2 1.%.4 Mean value 3.5% 3.% 4.9% 7.3% 2.8 Slope 166 ± 64 Slope 1622 ± 57 Precison 7.3 Intercept 9 ± 7 Intercept 5 ± 25 lity 3. Glass bead analysis by fsla-icpms 397

12 Table 2. (continued) V C.V. Regression value JB1b ± % 2.3% % JB1b 27.7 ± % 2.9 JB ± % 1.3% % JB ± %.4 JB ± % 5.% % JB3(STD) ± % 4.1 JA ± % 2.9% % JA ± % 2.5 JA ± 1.1.9% 1.% % JA ± % 1.5 JA ± 1.5.9% 1.7% % JA ± % 1.1 JG1a ±.7 3.1% 5.7% % JG1a 21.8 ±.4 1.9% 3.7 JG ± % 9.5%.4 1.9% JG2 4.7 ± % 24.9 JG ± % 4.3% % JG3(STD) 67.1 ± % 4.3 JR ±.5 6.6% 12.5%.4 5.5% JR1 8.4 ±.4 4.2% 2.6 JR ±.7 2.5% 17.%.2 5.5% JR2 4.1 ± % ±.4 11.% 13.2%.2 6.2% (STD) 4.3 ±..6% 1.2 Mean value 6.% 6.4% 5.4% 9.7% 8.7 Slope 1 ± 1 Slope 1 ± Precison 9.7 Intercept 41 ± 3 Intercept 35 ± 4 lity 8.7 MnO C.V. Regression value JB1b ±.4 2.9% 1.8%.7 4.7% JB1b.143 ±.3 2.1% 3. JB ±.7 3.2%.1% % JB2.214 ±.5 2.3% 1.7 JB ±.5 2.7% 2.1%.9 4.9% JB3(STD).178 ±.1.8%.5 JA ±.3 2.1%.3%.6 4.1% JA1.155 ±.2 1.4% 1. JA ±.1.9%.1%.6 5.4% JA2.17 ±.1 1.%.5 JA ±.1.8% 2.8%.5 4.3% JA3.16 ±.1 1.1% 2.2 JG1a ±.2 3.% 1.5%.3 5.% JG1a.59 ±.2 2.8% 3. JG ±. 1.3%.1%.1 5.2% JG2.18 ± % 13. JG ±.1 1.3%.6%.3 4.5% JG3(STD).72 ±.1 1.8% 1.3 JR ±.2 2.5% 1.4%.4 4.2% JR1.97 ±.3 2.7% 1.7 JR ±.2 1.6% 1.5%.5 4.5% JR2.11 ±.2 1.8% ±.3 3.3% 1.9%.4 4.8% (STD).82 ±.2 2.3% 1.6 Mean value 2.1% 1.2% 4.7% 3.6% 2.6 Slope ± 3471 Slope ± 4234 Precison 4.7 Intercept 144 ± 57 Intercept 344 ± 341 lity 2.6 FeO C.V. Regression value JB1b ± % 7.1%.3 5.5% JB1b 5.58 ±.34 6.% 8.1 JB ± % 1.8% % JB ± % 2. JB ± % 6.4% % JB3(STD) 7.84 ±.1.1%.2 JA ± % 9.7% % JA ± % 3.8 JA ± % 5.9% % JA ± % 3.5 JA ± % 13.7%.25 6.% JA ± % 19.7 JG1a ±.6 5.3% 14.%.7 6.1% JG1a.15 ± % 89.1 JG ±.3 4.7% 3.8%.4 7.3% JG2.63 ± % 29.7 JG ± % 23.4%.16 7.% JG3(STD) 1.5 ± % 17.8 JR ± % 8.4%.3 5.3% JR1.65 ± % JR ±.2 4.9% 4.5%.3 6.% JR2.79 ± % ±.3 1.4% 53.7%.17 6.% (STD) 2.25 ± % 21. Mean value 8.7% 13.5% 6.% 29.9% 73.9 Slope 518 ± 47 Slope 415 ± 26 Precison 29.9 Intercept 15 ± 35 Intercept 558 ± 132 lity Y. Kon et al.

13 Table 2. (continued) Co C.V. Regression value JB1b ±.6 1.5% 1.1% 2. 5.% JB1b 38.4 ±.9 2.4% 4.8 JB ±.7 1.8%.3% % JB ± % 3.9 JB ±.7 1.9% 7.1% % JB3(STD) 35.4 ±.1.3% 3.2 JA ±.3 2.5% 6.8%.7 6.4% JA ±.2 1.5% 9.6 JA ±.4 1.4% 2.1% % JA2 29. ±.4 1.5% 1.6 JA ±.4 2.1%.6% % JA3 2.5 ±.3 1.3% 2.8 JG1a ±.4 7.6% 6.%.3 5.6% JG1a 5.4 ±.4 6.8% 8. JG ±.3 6.9% 31.% % JG2 4.7 ±.3 6.6% 28.7 JG ±.2 1.8% 3.1%.7 6.6% JG3(STD) 11. ±.1 1.2% 6. JR1.8.9 ±.2 2.4% 1.5% % JR1 1. ±.2 21.% 18.6 JR2.5.5 ±. 9.3% 4.1% % JR2.6 ± % ±.1 5.9% 6.% % (STD) 1. ±..4%.9 Mean value 5.3% 6.6% 9.5% 5.1% 9.2 Slope 7 ± 1 Slope 8 ± 1 Precison 9.5 Intercept ± Intercept 1 ± 1 lity 9.2 Ni C.V. Regression value JB1b ± % 7.9% % JB1b ± % 19.8 JB ± % 37.1% % JB2 1.4 ± % 37.4 JB ± % 4.8% % JB3(STD) 37.3 ± % 2.9 JA1 3.5 n.d. ± JA1 2.6 ± % 24.9 JA ± % 11.5% % JA ± % 23.8 JA ± % 3.1% % JA ± % 4.4 JG1a ± % 12.1% % JG1a 15.8 ± % 128% JG ± % 13.8% % JG2 4.9 ± % 12. JG ± % 6.7% % JG3(STD) 13.6 ± % 4.8 JR1 1.7 n.d. ± JR1 7.4 ± % JR ± % 19.6%.2 9.9% JR2 1.3 ± % ± % 61.7%.3 11.% (STD) 1.6 ± %.7 Mean value 38.4% 28.6% 12.1% 127.4% 52.9 Slope 1 ± Slope 1 ± Precison Intercept 9 ± 2 Intercept 11 ± 2 lity 52.9 Zn C.V. Regression value JB1b ± % 7.4% % JB1b 73.3 ± % 8.4 JB ± 1. 1.% 7.% % JB ± % 11.3 JB ± % 2.6% % JB3(STD) 93.2 ± % 6.8 JA ± %.3% % JA ± % 3.4 JA ± % 5.1% % JA ± % 4.9 JA ± % 2.3% % JA ± % 1.9 JG1a ± % 9.1% % JG1a 43.7 ± % 19.8 JG ± %.2% % JG ± % 56. JG ± % 3.9% % JG3(STD) 47.9 ±.6 1.3% 3.1 JR ± % 2.8% % JR1 35. ± % 14.5 JR ± % 11.7% % JR ± % ± % 16.9% % (STD) ± % 5. Mean value 14.3% 5.8% 15.5% 11.6% 13.8 Slope 1 ± Slope 1 ± Precison 15.5 Intercept 12 ± 3 Intercept 6 ± 3 lity 13.8 Glass bead analysis by fsla-icpms 399

14 Table 2. (continued) Ga C.V. Regression value JB1b 23.1 ±.8 3.5% % JB1b 2.7 ±.7 3.2% JB ±.8 5.2% 4.5% % JB ±.8 6.3% 24.1 JB ± % 6.5% % JB3(STD) 18.4 ±.6 3.3% 7. JA ±.7 3.9% 15.% % JA ±.9 5.7% 2.6 JA ±.7 3.9% 6.% % JA ± % 12.4 JA ±.8 4.4% 15.8% % JA ± % 2.5 JG1a ± % 31.1% % JG1a 19. ± % 15.2 JG ± % 13.6% % JG ±.4 3.1% 31.5 JG ± % 23.1% % JG3(STD) 18.4 ± % 7.3 JR ±.4 2.8% 2.4%.9 7.2% JR1 9. ± % 44. JR ±.9 6.5% 22.2% 1. 7.% JR2 1.3 ± % ± % 2.4% % (STD) 37. ±.5 1.4% 1. Mean value 5.3% 14.6% 7.2% 8.8% 17.3 Slope 5 ± Slope 5 ± 1 Precison 8.8 Intercept 5 ± 9 Intercept 75 ± 1 lity 17.3 As C.V. Regression value JB1b ± % 2.6% % JB1b 1.3 ±.3 2.9% 1.4 JB ± % 1.5% % JB2 4.2 ± % 44.7 JB ± % 17.6% % JB3(STD) 1.6 ± % 12.1 JA ± % 12.3% % JA1 4.5 ± % 6.3 JA2.9.6 ± % 3.7% % JA2.3 ± % 65.2 JA ± % 4.5% % JA3 7.2 ± % 53.5 JG1a.4.6 ± % 44.3%.2 25.% JG1a.3 ± % 24.3 JG ± % 83.% % JG2 1.3 ± % 97.5 JG3.4.4 ± % 16.%.1 25.% JG3(STD).1 ± % 78.9 JR ± % 1.8% % JR ± % 55.4 JR ± % 2.% % JR ± % ±.2 8.6% 69.1% % (STD) 2.3 ± % 17.1 Mean value 24.9% 23.8% 16.1% 217.6% 55.2 Slope 1 ± Slope 1 ± 1 Precison Intercept ± Intercept 1 ± lity 55.2 Rb C.V. Regression value JB1b ± % 1.4% % JB1b 37.4 ± % 4.4 JB ±.4 5.4% 5.7%.6 8.6% JB2 5.9 ± % 19.9 JB ±.4 2.7% 4.2% % JB3(STD) 14.7 ±.5 3.2% 2.9 JA ±.4 3.1% 1.1%.8 6.2% JA ± % 9.8 JA ± % 1.4% % JA ± % 8.6 JA ± % 9.% % JA ± % 5.8 JG1a ± % 2.% % JG1a ± % 2.9 JG ± % 2.8% % JG ± % 3.5 JG ± 3. 4.% 12.5% % JG3(STD) 74.4 ± % 1.6 JR ± % 4.6% % JR ± % 5.3 JR ± % 2.1% % JR ± % ± % 4.3% % (STD) 431. ± % 4.9 Mean value 3.8% 5.% 6.8% 6.3% 6.8 Slope 1 ± Slope 1 ± 1 Precison Intercept ± Intercept 1 ± lity Y. Kon et al.

15 Table 2. (continued) Sr C.V. Regression value JB1b ± % 1.8% % JB1b ± % 2.4 JB ± % 2.1% % JB ± % 2.8 JB ± % 2.7% % JB3(STD) ± % 3.3 JA ± % 1.2% % JA ± %.6 JA ± % 2.4% % JA ± % 1.7 JA ± %.5% % JA ± %.1 JG1a ± % 4.4% % JG1a 18.1 ± % 3.7 JG ±.4 2.2% 1.8% % JG ±.2.9% 1.8 JG ± % 4.% % JG3(STD) ± % 3.4 JR ±.9 2.9% 9.2% % JR ± % 11.7 JR ±.4 4.6% 1.4%.6 7.5% JR2 8.8 ±.4 4.1% ±.3 2.6% 5.8%.8 8.% (STD) 1.4 ±..%. Mean value 2.6% 3.1% 6.6% 2.9% 3.4 Slope 5 ± Slope 5 ± Precison 6.6 Intercept 6 ± 3 Intercept 3 ± 4 lity 3.4 Y C.V. Regression value JB1b 22.9 ±.9 4.% % JB1b 22.8 ±.9 3.8% JB ± %.1% % JB2 25. ± %.5 JB ± % 1.6% % JB3(STD) 27.6 ±.7 2.7% 2.7 JA ±.8 2.7%.4% % JA1 31. ±.7 2.1% 1.5 JA ±.5 2.9% 3.3% % JA ±.9 5.% 5.9 JA ±.8 4.%.4% % JA3 2.9 ± % 1.3 JG1a ± %.7% % JG1a 32.6 ± % 1.4 JG ± % 7.1% % JG ± 3. 3.% 13.5 JG ±.2 1.3%.6% % JG3(STD) 16.9 ±.7 4.% 2.2 JR ± % 1.2% % JR ± % 5.1 JR ± % #REF! % JR ± % ± % 7.3% % (STD) ± % 1.1 Mean value 3.2% 2.3% 6.6% 3.7% 3.8 Slope 5 ± Slope 5 ± Precison 6.6 Intercept 3 ± 5 Intercept 6 ± 6 lity 3.8 Zr C.V. Regression value JB1b ± % % JB1b ± % JB ±.9 1.7% 2.7% % JB ± % 8.7 JB ± % 2.4% % JB3(STD) 93.9 ± % 4. JA ± % 3.3% % JA ± % 5.4 JA ± % 4.3% % JA2 11. ± % 5.2 JA ± % 1.3% % JA ± % 2.1 JG1a ± % 1.6% % JG1a ± % 1. JG ± % 16.3% % JG ± % 15.3 JG ± % 8.4% % JG3(STD) ± % 8.7 JR ± % 1.4% % JR ± %. JR ± % 1.8% % JR ± % ± % 4.9% % (STD) ± % 2.2 Mean value 2.8% 4.4% 6.8% 3.9% 4.8 Slope 3 ± Slope 3 ± Precison 6.8 Intercept 11 ± 13 Intercept 2 ± 22 lity 4.8 Glass bead analysis by fsla-icpms 41

16 Table 2. (continued) Nb C.V. Regression value JB1b ± % % JB1b 26. ± % JB ±.7 4.9% 9.4% % JB2.52 ± % 66.9 JB ± % 11.6% % JB ±.8 4.3% 2.5 JA ±.1 4.8% 17.2% % JA ± % 28.4 JA ± % 4.7% % JA ±.18 2.% 7.2 JA ± % 11.% % JA3 3.8 ± % 9.6 JG1a ± % 2.1% % JG1a 11.1 ± % 2.6 JG ± % 2.7% % JG ± % 1.2 JG ± % 8.1% % JG3(STD) 5.88 ±..%. JR ± %.6% % JR ± % 2.6 JR ±.17.9% 2.6% % JR ± % ± % 7.9% % (STD) 51. ±..%. Mean value 3.3% 7.1% 8.4% 4.8% 12.7 Slope 11 ± 1 Slope 1 ± 1 Precison 8.4 Intercept 1 ± 2 Intercept ± 1 lity 12.7 Sn C.V. Regression value JB1b 1.74 ± % % JB1b 1.69 ± % JB ±.6 8.5% 2.8% % JB2.72 ± % 24.6 JB ± % 22.5% % JB3(STD) 1.11 ± % 18.1 JA ± % 7.7% % JA1 1.3 ± % 11.3 JA ± % 9.% % JA ± % 6. JA ± % 21.7% % JA ± % 17.2 JG1a ± % 8.3% % JG1a 4.1 ± % 1.2 JG ± % 4.8%.3 1.5% JG ± % 7. JG ± % 9.3% % JG3(STD) 1.22 ± % 12.6 JR ± % 9.% % JR1 3.5 ± % 6.6 JR ± % 9.3% % JR ± % ±.6 3.3% 3.4% % (STD) 17.7 ±.53 3.% 1.7 Mean value 14.6% 11.4% 12.1% 18.5% 11.1 Slope 3 ± Slope 4 ± Precison 18.5 Intercept 1 ± 1 Intercept 1 ± 1 lity 11.4 Cs C.V. Regression value JB1b.82 ±.6 6.9%.8 1.% JB1b.77 ± % JB ± % 6.2% % JB2.75 ± % 11.2 JB ± %.2% % JB3(STD).89 ±.3 3.8% 5.5 JA ± % 1.9%.6 9.7% JA1.61 ± % 2.3 JA ± % 1.2% % JA ± % 11.1 JA ± % 2.6% % JA ±.7 4.% 11.2 JG1a ± % 1.%.8 7.5% JG1a 9.56 ± % 9.8 JG ±.28 4.% 2.6% % JG ± % 8.1 JG ± % 8.5%.17 9.% JG3(STD) 1.77 ±.3 1.4%.8 JR ± % 3.9% % JR ± % 14.3 JR ± % 4.7% % JR ± % ±.12 1.% 18.3% % (STD) 1.1 ± % 1.2 Mean value 1.3% 4.6% 11.5% 2.5% 8.2 Slope 14 ± 1 Slope 16 ± 3 Precison 2.5 Intercept 1 ± 1 Intercept ± 5 lity Y. Kon et al.

17 Table 2. (continued) Ba C.V. Regression value JB1b 49.9 ± % % JB1b ± % JB ± %.9% % JB ± % 2.8 JB ± % 3.1% % JB3(STD) ± %.7 JA ± % 21.1% % JA ± % 1.2 JA ± % 6.2% % JA ± % 2.2 JA ± % 4.5% % JA ± %.2 JG1a ± % 7.1% % JG1a ± % 2.2 JG ± % 19.1% % JG ± % 21.9 JG ± % 5.2% % JG3(STD) ± %.6 JR ± % 12.4% % JR ± % 6.5 JR ± %.9% % JR ± % ± % 3.2% % (STD) 65.7 ±.4.6%.1 Mean value 26.4% 7.6% 8.1% 3.7% 4.3 Slope 1 ± Slope 1 ± Precison 26.4 Intercept 11 ± 21 Intercept 4 ± 3 lity 7.6 La C.V. Regression value JB1b 41.2 ± % % JB1b 4.9 ± 2. 5.% JB ±.2 8.9% 1.6%.3 1.6% JB2 2.5 ± % 5.8 JB ±.4 4.1%.8%.7 7.6% JB3(STD) 8.8 ±.1.7%.5 JA ±.5 9.6%.2% % JA1 5.3 ±.4 7.8% 1.1 JA ±.9 5.6%.6% % JA ±.7 4.2%.3 JA ±.4 4.% 3.3%.8 7.9% JA3 9.7 ±.6 5.8% 3.5 JG1a ±.8 3.6% 5.4% % JG1a 22.3 ±.8 3.6% 4.8 JG ±.7 3.7% 6.8% % JG ±.8 4.1% 7.2 JG ±.7 3.3% 1.8% % JG3(STD) 2.9 ±.3 1.6% 1.3 JR ±.8 3.9% 1.7% % JR ± % 1.2 JR ±.6 3.5% 3.9% % JR ±.8 4.9% ± %.4% % (STD) ± 1.5.8%.7 Mean value 4.9% 2.4% 8.6% 5.5% 2.8 Slope 6 ± Slope 6 ± 1 Precison 8.6 Intercept 2 ± 1 Intercept 1 ± 4 lity 2.8 Ce C.V. Regression value JB1b 71.8 ± % % JB1b 71.1 ± % JB ±.2 3.7% 1.2%.6 9.4% JB2 5.9 ± % 12.4 JB ±.7 3.% 2.4% % JB3(STD) 21.3 ±.3 1.2% 1. JA ± % 1.% % JA ± % 1.2 JA ± % 2.1% % JA ±.9 2.6%.2 JA ±.6 2.5% 1.9% % JA ± % 5.2 JG1a ± % 3.6% % JG1a 45.9 ± % 2. JG ± % 5.% % JG ± % 6.5 JG ± % 4.5% % JG3(STD) 41.4 ± % 2.6 JR ± %.% % JR ± % 1.6 JR ± % 1.6% % JR ± % ± %.8% % (STD) ± % 1. Mean value 5.1% 3.% 11.5% 5.8% 3.4 Slope 6 ± Slope 6 ± 1 Precison 11.5 Intercept 1 ± 4 Intercept 6 ± 1 lity 3.4 Glass bead analysis by fsla-icpms 43

18 Table 2. (continued) Pr C.V. Regression value *C.V. of JB2 was excluded from the calcuration of calibration line. JB1b 7.73 ± %.6 7.7% JB1b 7.55 ± % JB2 1.1*.62 ± % 61.6%.6 1.1% JB2.87 ± % 13.9 JB ±.9 2.9% 1.5% % JB3(STD) 3.16 ±.4 1.1% 1.7 JA ±.9 5.3% 2.7% % JA ±.3 1.7% 13.1 JA ± % 9.6%.31 9.% JA ± % 7.6 JA ±.4 1.6% 7.3% % JA3 2.7 ± % 12.6 JG1a ± % 7.6% % JG1a 5.17 ± % 8.1 JG ± % 3.7% % JG ± % 4.9 JG ± % 2.9%.4 8.8% JG3(STD) 4.57 ± % 2.7 JR ± % 8.7% % JR ± % 7.2 JR ± % 3.5% % JR2 4.9 ± % ±.9 2.5% 6.5% % (STD) ±.27.8%.9 Mean value 5.% 1.5% 8.8% 4.2% 6.9 Slope 13 ± 1 Slope 14 ± 1 Precison 8.8 Intercept 1 ± 1 Intercept 6 ± 2 lity 1.5 Nd C.V. Regression value JB1b 27.1 ± % % JB1b 26.6 ± % JB ±.5 7.1% 3.4%.6 1.1% JB2 6. ± % 9.5 JB ±.4 2.4% 4.5% % JB3(STD) 15.9 ±.4 2.7% 1.7 JA ±.4 3.7% 3.3% % JA1 1.8 ±.4 3.5%.5 JA ± % 3.7% % JA2 14. ± %.5 JA ± % 1.9% % JA ± % 1.6 JG1a ±.7 3.8% 3.1% % JG1a 19.3 ±.7 3.6% 5.4 JG ±.4 1.7% 6.8% % JG ±.7 2.9% 8.6 JG ±.7 3.9%.8% % JG3(STD) 16.9 ±.6 3.4% 1.8 JR ±.6 2.7% 3.3% % JR ±.8 3.3% 1.2 JR ± % 1.1% % JR ± % ± % 1.% % (STD) 17.2 ±.3.3%.1 Mean value 4.9% 3.% 9.3% 5.3% 3.1 Slope 1 ± Slope 1 ± Precison 9.3 Intercept 1 ± 2 Intercept 1 ± 1 lity 3.1 Sm C.V. Regression value JB1b 5.17 ± % % JB1b 5.22 ± % JB ± % 7.4% % JB ± % 12.8 JB ± % 1.1% % JB3(STD) 4.39 ± % 2.8 JA ± % 3.% % JA ± %. JA ± % 1.% % JA ± % 2.6 JA ± % 7.2% % JA ± % 1.6 JG1a ± % 2.5% % JG1a 4.71 ±.38 8.% 3.9 JG ± % 3.% % JG ± % 2.5 JG ± % 4.2% % JG3(STD) 3.36 ±.2 6.%.9 JR ±.23 4.% 5.7%.57 1.% JR ± % 5.2 JR ± % 3.3% % JR ± % ± % 2.% % (STD) ± %.2 Mean value 9.7% 3.7% 12.6% 9.9% 4.1 Slope 1 ± Slope 1 ± Precison 12.6 Intercept ± Intercept ± lity Y. Kon et al.

19 Table 2. (continued) Eu C.V. Regression value JB1b 1.59 ± % % JB1b 1.52 ± % JB ± % 3.6% % JB2.88 ±.8 9.1% 1.8 JB ± % 1.% % JB3(STD) 1.38 ±.7 5.4% 4.9 JA ± %.9% % JA ± % 1.5 JA ±.7 8.1% 1.2% % JA2.9 ±.6 6.1% 3.1 JA ± % 1.8% % JA3.89 ± % 8.7 JG1a.7.67 ±.5 7.7% 4.5% % JG1a.67 ±.5 7.1% 3.9 JG ±.2 2.7% 15.1% % JG2.17 ± % 66. JG ±.7 7.7% 5.1% % JG3(STD).84 ±.4 5.% 6.3 JR ± % 6.2% % JR1.35 ± % 17.7 JR ± % 11.5% % JR2.17 ± % ± %.2% % (STD).55 ±.2 2.7% 3. Mean value 13.5% 6.3% 14.3% 22.7% 12.8 Slope 8 ± Slope 9 ± 1 Precison 22.7 Intercept ± Intercept ± 1 lity 12.8 Gd C.V. Regression value JB1b 4.38 ± % % JB1b 4.53 ± % JB ± % 1.5% % JB ± % 2. JB ±.4 8.9% 2.6% % JB3(STD) 4.7 ± %.6 JA ± % 7.1% % JA ± % 3.9 JA ± % 1.9% % JA ± % 1.7 JA ± % 16.8% % JA ± % 2.9 JG1a ± % 1.1% % JG1a 4.27 ± % 4.6 JG ± % 13.1% % JG ± % 16.6 JG ± % 2.9% % JG3(STD) 2.94 ± %.8 JR ± % 2.6% % JR ± % 5.9 JR ± % 6.1% % JR ± % ± % 2.7% % (STD) ±.4 2.%.2 Mean value 8.5% 5.3% 12.3% 8.3% 5.5 Slope 1 ± Slope 1 ± Precison 12.3 Intercept ± 1 Intercept ± lity 5.5 Tb C.V. Regression value JB1b.69 ±.6 9.4%.7 1.7% JB1b.69 ±.7 1.6% JB ±.3 4.8% 2.4% % JB2.58 ±.5 8.9% 2.7 JB ±.8 11.% 1.7%.8 1.6% JB3(STD).74 ±.6 7.7% 1.7 JA ±.3 4.5% 2.7% % JA1.73 ±.6 8.2% 2.7 JA ±.4 8.6% 8.6% % JA2.47 ±.3 6.3% 7.8 JA ±.5 9.6%.9%.6 1.9% JA3.52 ± %.4 JG1a ±.2 2.4% 8.4%.8 1.4% JG1a.74 ±.2 2.5% 8.4 JG ±.2 1.% 19.% % JG ±.6 3.3% 2. JG ±.3 6.7% 1.2% % JG3(STD).45 ±.4 7.9% 2. JR ±.6 5.7% 1.7%.1 1.% JR1 1. ±.8 7.8% 1.3 JR ±.6 5.9% 3.8% % JR2 1.6 ±.7 6.2% ± % 1.4% % (STD) 4.28 ± %.3 Mean value 6.% 4.7% 11.% 7.1% 4.6 Slope 15 ± 1 Slope 14 ± 1 Precison 11. Intercept ± 1 Intercept ± 1 lity 4.7 Glass bead analysis by fsla-icpms 45