(Received June 29, 2015; Accepted July 21, 2016)

Transcription

1 Geochemical Journal, Vol. 50, pp. 403 to 422, 2016 doi: /geochemj Simultaneous determination of 58 major and trace elements in volcanic glass shards from the INTAV sample mount using femtosecond laser ablation-inductively coupled plasma-mass spectrometry SEIJI MARUYAMA, 1 * KENTARO HATTORI, 2 TAKAFUMI HIRATA, 2 TAKEHIKO SUZUKI 3 and TOHRU DANHARA 1 1 Kyoto Fission-Track Co., Ltd., 44-4 Oomiyaminamitajiri-cho, Kita-ku, Kyoto , Japan 2 Department of Geology and Mineralogy, Faculty of Science, Kyoto University, Kitashirakawaoiwake-cho, Sakyo-ku, Kyoto , Japan 3 Department of Geography, Tokyo Metropolitan University, Minamiosawa 1-1, Hachioji, Tokyo , Japan (Received June 29, 2015; Accepted July 21, 2016) Four volcanic glass samples on the INternational focus group on Tephrochronology And Volcanism (INTAV) sample mount described by Kuehn et al. (2011) were analyzed to establish a new methodology for measurement of major and trace elements in Quaternary tephras by femtosecond laser ablation-inductively coupled plasma-mass spectrometry (LA- ICP-MS). NIST SRM 610 and 612 glasses were used for calibration of the measurement of 58 elements from Li to U in two rhyolitic glass samples (Lipari obsidian ID3506 and Old Crow tephra) and the phonolitic Sheep Track tephra. In addition to the NIST SRM glasses, USGS BCR-2G and BHVO-2G basaltic glasses were measured for calibration of the major elements (i.e., Na, Mg, Al, Si, K, Ca, Ti, Mn, and Fe) in the basaltic Laki tephra. Most major element data for the four volcanic glass samples, and those for ZrO 2 in Lipari obsidian and Sheep Track tephra deviate <10% from the preferred values obtained by electron-beam analysis techniques complied in Kuehn et al. (2011). P 2 O 5 values range from 74% to 110% of the preferred values in Kuehn et al. (2011), and are consistent within uncertainties. BaO values are 7 24% of the preferred values of Kuehn et al. (2011), but are similar to those obtained using other trace element analytical techniques, such as X-ray fluorescence and ICP atomic emission spectroscopy. Our trace element data are generally consistent with those obtained in previous studies. However, heterogeneously distributed microcrysts in the glass materials may affect some elements such as B, P, and Cr. The analytical data for the INTAV samples obtained using femtosecond LA-ICP-MS and the NIST SRM and USGS standard glasses demonstrate that this approach is a viable alternative to electron-beam microanalysis techniques for the determination of major elements in rhyolitic and basaltic glasses. Keywords: volcanic glass, tephra, LA-ICP-MS, element abundance, femtosecond laser INTRODUCTION The geochemical characterization of tephras for correlation purposes has primarily focused on major and minor element compositions of individual volcanic glass shards determined by electron-beam analytical techniques, such as electron probe microanalysis (EPMA) and scanning electron microscopy coupled with energy dispersive X-ray spectroscopy (SEM-EDS) (e.g., Westgate and Gorton, 1981; Kuehn et al., 2011; Lowe, 2011). The abundances of trace elements such as the rare earth elements (REE) have also been used for identification of *Corresponding author ( ojigibito@icloud.com) Present address: Geochemical Research Center, Graduate School of Science, The University of Tokyo, Hongo, Bunkyo-ku, Tokyo , Japan. Copyright 2016 by The Geochemical Society of Japan. tephras (e.g., Eastwood et al., 1998, 1999). Instrumental neutron activation analysis (INAA) has been widely used as a quantitative analytical method for a wide range of trace elements in volcanic glass samples (e.g., Fukuoka et al., 1980; Westgate and Gorton, 1981; Fukuoka 1991, 1993, 1994). Inductively coupled plasma-atomic emission spectroscopy (ICP-AES) (e.g., Tamura et al., 2008) and solution nebulization-icp-mass spectrometry (SN- ICP-MS) (e.g., Pearce et al., 1999) have also been used for determination of trace element abundances in volcanic glass. However, these analytical techniques require separation of significant amounts of glass shards from tephra samples. For example, ICP-AES requires ~0.5 g of purified glass for each analysis and SN-ICP-MS requires g of glass to be digested for analysis (Pearce et al., 1999). The laser ablation-icp-ms (LA-ICP-MS) technique has the advantage that it can analyze individual glass shards without incorporation of coexisting mineral grains 403

2 Table 1. ICP MS instrument and laser ablation analytical conditions Thermo Fisher Scientific icap Qc Operating Conditions Operation mode Standard mode (no collision gas was used) RF power 1550 W Cool gas (Ar) 14 L/min Auxiliary gas (Ar) 0.8 L/min Carrier gas through sample chamber (He) 0.54 L/min Conditioning gas mixed to carrier gas (Ar) 0.83 L/min ThO + /Th + (oxide ratio) % Analyzed mass numbers 7 Li, 9 Be, 11 B, 23 Na, 24 Mg, 27 Al, 29 Si, 31 P, 39 K, 43 Ca, 45 Sc, 49 Ti, 51 V, 52 Cr, 55 Mn, 57 Fe, 59 Co, 60 Ni, 63 Cu, 66 Zn, 71 Ga, 73 Ge, 75 As, 85 Rb, 88 Sr, 89 Y, 90 Zr, 93 Nb, 95 Mo, 107 Ag, 111 Cd, 115 In, 118 Sn, 121 Sb, 133 Cs, 137 Ba, 139 La, 140 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, 182 W, 205 Tl, 208 Pb, 209 Bi, 232 Th, 238 U Analysis mode TRA (time resolved analysis) mode Dwell time s/element, 2 s/time slice ESI NWR-Femto Operating Conditions Fluence J/cm 2 Repetition rate Beam diameter Duration of laser ablation 5 Hz 15 mm 20 s/pit (e.g., plagioclase, hornblende, pyroxene, and quartz) in tephra samples. The LA-ICP-MS method has been utilized for trace elements analyses (e.g., REE, Th, and U) of glass shards in tephra samples (e.g., Pearce et al., 2008, 2011; Westgate et al., 2008, 2011). Secondary ion mass spectrometry (SIMS) has similar advantages to LA-ICP- MS, and is now also used for the in situ analysis of volcanic glass shards (e.g., Begét and Keskinen, 2003; Charlier and Wilson, 2010). The spatial resolution of SIMS is better (5 10 mm) than that of LA-ICP-MS, which is typically >10 mm (Lowe, 2011). However, SIMS is a less rapid and more complex analytical technique than LA-ICP-MS. At present, the major element compositions of volcanic glass samples are commonly determined using electron-beam analytical techniques, and trace element concentrations are analyzed using a range of other analytical techniques. However, the analytical site of the major element determination using the electron-beam technique is not necessarily the same as that of trace element measurements using other techniques such as LA-ICP-MS and SIMS. As such, there may not be a one-to-one correspondence between the major element composition and trace element concentration. Microcrysts within volcanic glasses can compromise not only the determined major element composition, but also the trace element analysis. Therefore, it is preferable to analyze the concentration of both major and trace elements simultaneously at each analytical site, which makes it easier to identify outliers caused by microcrysts. The simultaneous analysis of major and trace elements has a further advantage in that it is simpler and faster compared with a combination of electron-beam analytical techniques followed by the use of another analytical technique for trace element analysis. In this paper, we present analyses of a total of 58 major and trace elements in volcanic glass samples distributed by the INternational focus group on Tephrochronology And Volcanism (INTAV), as obtained by femtosecond LA-ICP-MS. The shorter pulse duration of the femtosecond laser can be expected to minimize elemental fractionation induced by thermal effects as compared with a nanosecond laser (Jochum et al., 2014). Maruyama et al. (2016) concluded that elemental data obtained for volcanic glass samples using femtosecond LA-ICP-MS is generally comparable with those of previous studies, and that it is adequate for successful identification and correlation of tephra samples. We further explore whether this technique can replace electron-beam analytical techniques for the major element determinations of volcanic glass shards, in addition to trace element analysis. SAMPLES AND ANALYTICAL METHODS INTAV volcanic glass samples Three homogeneous natural volcanic glasses were mounted in the INTAV sample mount for inter-laboratory comparisons by tephrochronologists (Kuehn et al., 2011): (1) rhyolitic obsidian ID3506 from Lipari Island, Italy (Kuehn et al., 2009); (2) phonolitic Sheep Track tephra from Mt. Edziza, British Columbia, Canada; and (3) microcryst-bearing basaltic tephra from the 1783 A.D. eruption of Laki volcano in Iceland. In addition, rhyolitic Old Crow tephra from Alaska, USA, was also mounted. Kuehn et al. (2011) briefly summarized the preparation and nature of the INTAV reference glass samples. The 404 S. Maruyama et al.

3 Fig. 1. Optical images of the INTAV volcanic glass samples after femtosecond laser ablation, taken with a Leica DVM5000 HD digital microscope: (a) Lipari obsidian ID3506; (b) Sheep Track; (c) Laki; (d) Old Crow. The white arrows indicate laser ablation pits, and black arrows on a glass shard of Old Crow tephra indicate parts that fractured during laser ablation. Using the LA- ICP-MS method, micron-sized mineral grains of plagioclase and pyroxene in the glass shards of the Laki tephra can be avoided during analysis. Px = pyroxene; Plg = plagioclase. volatile contents (i.e., H 2 O/OH, Cl, F, and SO 2 ) of the glasses, which are difficult to analyze by SN- and LA- ICP-MS, are presented in Kuehn et al. (2011). The INTAV mount was gently polished on lapping sheets (#8000 and #10000; 3M Japan Limited) to remove the carbon coating added for the electron-beam microanalysis previously carried out by Suzuki et al. (2014). The carbon coating was removed to minimize contaminants on the surface of the mount during laser ablation. After this, the mount was washed in ultrapure water in an ultrasonic bath for ca. 10 min. Analytical methods LA-ICP-MS analyses were performed using a Thermo Fisher Scientific icap Qc quadrupole ICP-MS instrument coupled to an ESI NWR-Femto laser ablation system at Kyoto University, Kyoto, Japan. The analytical conditions are summarized in Table 1. The wavelength and duration of the laser pulse of the NWR-Femto are 260 nm and 600 fs, respectively. The duration of the laser ablation was limited to 20 s so as not to produce a second signal peak by longer ablation, which would be caused by the smaller incidence angle of the laser resulting in a larger reflection rate and total reflection of the incident laser. This is known as optical waveguide phenomena and can produce different elemental fractionation features. During LA-ICP-MS analysis, the optical waveguide phenomena results in a higher ablation rate at a higher aspect ratio (i.e., a depth/diameter) in a laser ablation pit during the latter stages of ablation and, as such, may produce a second signal peak. The typical depths of the ablation pits were measured using a Keyence VK-X three-dimensional laser scanning microscope, and are mm for a 20 s ablation. Figure 1 shows optical images of the volcanic glasses after laser ablation, taken with a Leica DVM5000 HD digital microscope. The laser ablation pits in the four INTAV volcanic glasses differ from each other (Fig. 1). Dark-colored ablation debris surrounds the pits in the Laki basaltic tephra (Fig. 1c), whereas this is less evident or not noticeable for the rhyolitic glass samples (i.e., Lipari obsidian and Old Crow tephra) (Figs. 1a and d). No debris is visible around the pits in the Sheep Track tephra (Fig. 1b). Prior to and at the end of each set of volcanic glass analyses, two silicate glasses distributed by the National Determination of 58 elements in the INTAV volcanic glasses using fsla-icp-ms 405

4 Table 2. Major element data (wt.%) in four INTAV volcanic glass samples using the LA ICP MS method. The preferred mean values from electron-beam analytical methods (i.e., EPMA and SEM-EDS) compiled by Kuehn et al. (2011) are shown for comparison. n a SiO 2 TiO 2 ZrO 2 Al 2 O 3 FeO BaO MnO MgO CaO Na 2 O K 2 O P 2 O 5 Oxides total Trace elements Total Lipari obsidian ID3506 (Sample 1) LA-ICP-MS mean b s (LA-ICP-MS) c < RSD (%) Form of calibration line æ Y=aX+b Y=aX Y=aX Y=aX+b Y=aX Y=aX+b Y=aX+b Y=aX Y=aX Y=aX+b Y=aX Preferred e-beam mean of Kuehn et al. (2011) s (e-beam) d Number of contributions (e-beam) LA-ICP-MS/e-beam e s Other methods f XRF (Kuehn et al., 2011) s ICP-AES (Kuehn et al., 2011) SN-ICP-MS (Kuehn et al., 2011) s LA-ICP-MS (Kuehn et al., 2009) s Sheep Track tephra (Sample 2) LA-ICP-MS mean s (LA-ICP-MS) RSD (%) Form of calibration line æ Y=aX+b Y=aX Y=aX Y=aX+b Y=aX Y=aX+b Y=aX+b Y=aX Y=aX Y=aX+b Y=aX Preferred e-beam mean of Kuehn et al. (2011) s (e-beam) Number of Contributions (e-beam) LA-ICP-MS/e-beam s Other methods XRF (Kuehn et al., 2011) ICP-AES (Kuehn et al., 2011) Laki 1783 A.D. tephra (Sample 3) LA-ICP-MS mean s (LA-ICP-MS) RSD (%) Form of calibration line æ Y=aX+b Y=aX+b Y=aX+b Y=aX Y=aX+b Y=aX+b Y=aX+b Y=aX+b Y=aX+b Y=aX+b 406 S. Maruyama et al.

5 Preferred e-beam mean of Kuehn et al. (2011) s (e-beam) Number of Contributions (e-beam) g LA-ICP-MS/e-beam s Old Crow tephra LA-ICP-MS mean s (LA-ICP-MS) RSD (%) Form of calibration line æ Y=aX+b Y=aX Y=aX+b Y=aX+b Y=aX+b Y=aX Y=aX Y=aX+b Y=aX Preferred e-beam mean of Kuehn et al. (2011) s (e-beam) Number of Contributions (e-beam) LA-ICP-MS/e-beam s Other methods XRF (Kuehn et al., 2011) ICP-AES (Kuehn et al., 2011) SN-ICP-MS (Preece et al., 2011) s INAA (Kaufman et al., 2001) INAA (Preece et al., 2000) INAA (Preece SIMS and Keskinen, 2003) s LA-ICP-MS (Westgate et al., 1994): UT s LA-ICP-MS (Westgate et al., 1994): UT s LA-ICP-MS (Westgate et al., 1994): UT s LA-ICP-MS (Westgate et al., 1994): UT s LA-ICP-MS (Pearce et al., 2004): UT s LA-ICP-MS (Pearce et al., 2004): UT s LA-ICP-MS (Pearce et al., 2004): UT s Number of decimal places of the analytical values of each oxide component obtained in this study is set to the same as the electron-beam mean value compiled by Kuehn et al. (2011). RSD = relative standard deviation (%). a Total number of analyses of each glass sample. b 7 out of 25 analytical values of P obtained on Sample 1 were below the detection limit (~26 mg/g; Table 4). c The standard deviation (1s) of the analytical values obtained from the measurements on each volcanic glass sample. d The standard deviation (1s) of the analytical results of each volcanic glass sample obtained from the contributors without considering error propagation. e The ratio of the values obtained in this study to the electron-beam preferred mean values compiled by Kuehn et al. (2011). The mean values of the ratios are shown in Fig. 2. f The analytical data obtained by other methods summarized in Kuehn et al. (2011) for comparison. g There is no value from electron-beam analysis methods. The value obtained in this study is shown in Table 4 in the form of the element concentration (mg/g). Determination of 58 elements in the INTAV volcanic glasses using fsla-icp-ms 407

6 Table 3. Concentrations of trace elements (mg/g) in four INTAV volcanic glass samples Lipari obsidian ID3506 (Sample 1) Sheep Track tephra (Sample 2) Concentration 1s RSD (%) 1s mean n(bdl) a Concentration 1s RSD (%) 1s mean n(bdl) Li Be B Sc V Cr b Co Ni Cu Zn Ga Ge As Rb Sr Y Zr Nb Mo Ag Cd In Sn Sb Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Tl Pb Bi Th U Institute of Standards and Technology (NIST), NIST SRM 610 and 612, were repeatedly analyzed as reference materials. In this study, the reference values proposed by Jochum et al. (2011) were used for NIST SRM 610 and 612. Calibration of the major elements in the basaltic Laki tephra sample used two basaltic glasses (BCR-2G and BHVO-2G) provided by the U.S. Geological Survey (USGS). This was carried out because the major elements, such as FeO, MgO, and CaO, in basaltic glass are very different to those in a rhyolitic glass. The reference values for the two USGS basaltic glasses were taken from the GeoREM database ( In addition to the measurement of 58 isotopes (Table 1) in the NIST SRM reference glasses, 408 S. Maruyama et al.

7 Table 3. (continued) Laki 1783 A.D. tephra (Sample 3) Old Crow tephra Concentration 1s RSD (%) 1s mean n(bdl) Concentration 1s RSD (%) 1s mean n(bdl) Li Be B Sc V Cr b Co Ni Cu Zn Ga Ge As Rb Sr Y Zr Nb Mo Ag Cd In Sn Sb Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Tl Pb Bi Th U The form of the calibration line for each trace element is Y = ax. RSD = relative standard deviation (%). The total number of analyses of each glass sample is shown in Table 2. a Number of analytical values below detection limit as summarized in Table 4. No entry for an element means that all the analytical values were above detection limits. The analytical values below detection limits were also used for the calculations. b One significant outlier (585 mg/g Cr) for Sample 3 was excluded from the calculations. Determination of 58 elements in the INTAV volcanic glasses using fsla-icp-ms 409

8 Table 4. Detection limits and RSD values obtained from measurement of NIST SRM reference glasses and four INTAV volcanic glass samples Session Lipari obsidian and Laki tephra Sheep Track and Old Crow tephras Detection limit RSD (%) Detection limit RSD (%) 410 S. Maruyama et al. (mg/g) NIST 610 NIST 612 Lipari Laki (mg/g) NIST 610 NIST 612 Sheep Track Old Crow Li Be B Na Mg Al Si æ æ æ æ æ æ æ æ P K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Rb Sr Y Zr Nb Mo Ag Cd In Sn Sb Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Tl Pb Bi Th < < U The RSD value for each element was obtained from the ratios of the signal intensities to Si.

9 isotopes of 10 isotopes of major elements (i.e., Na, Mg, Al, Si, P, K, Ca, Ti, Mn, and Fe) were measured in the USGS reference glasses. As such, both the NIST SRM and USGS reference glasses were used to calibrate the major (USGS) and trace (NIST SRM) elements in the Laki tephra. The two NIST SRM reference glasses were used for calibration of all elements in the three rhyoliticphonolitic volcanic glasses. We consider that the two NIST SRM reference glasses are sufficient for calibration of trace element concentrations in both rhyolitic and basaltic glass samples. The analyses comprised measurements of 58 isotopes (Table 1), which were obtained by integrating 18 cycles of measurements across the isotopic mass range. The integrated 18-cycle background signals just prior to ablation were then subtracted from the integrated values obtained during ablation. The background-corrected value of each isotope was then converted to an elemental value using the appropriate isotopic abundance, and the ratio of each element to that of silicon (i.e., the internal standard) was calibrated against the reference glasses. In this study, a signal stabilizer in the gas-flow path between the laser ablation unit and ICP-MS was used for smoothing the signal intensity (Tunheng and Hirata, 2004). Therefore, the period of signal integration (18 cycles corresponding to ~36 s) was longer than the duration of laser ablation (20 s). The forms of the calibration lines for the major elements (Na, Mg, Al, P, K, Ca, Ti, Mn, and Fe) and two trace elements (Zr and Ba) are provided in Table 2. These calibrations for rhyolitic volcanic glass samples have been examined through repeated trials of data processing by Maruyama et al. (2016), and for basaltic glass in this study. During LA-ICP-MS analysis, some polyatomic interferences overlap the isotopes of the major elements. For example, 24 Mg and 39 K can have interferences from 12 C 2 and 38 Ar 1 H, respectively. 57 Fe can have interferences from polyatomic species such as 40 Ar 16 O 1 H, 40 Ca 16 O 1 H, and 28 Si 29 Si. When using a calibration for rhyoliticphonolitic volcanic glasses of the form of Y = ax (a = slope) for major elements, the calculated values of MgO, FeO, and K 2 O are 2 3%, 11 21%, and 4 11% lower, respectively, than those obtained by a calibration in the form of Y = ax + b (b = intercept). To reduce the effects of interference, the intercept is important for the calibrations and, in particular, for Mg, K, and Fe, which are more strongly affected by interference. In the analyses of the three rhyolitic-phonolitic volcanic glasses, the calibration lines used for Al, Ca, and Na are of the form of Y = ax, because the slopes can be negative or vary widely if calibrations of the form Y= ax + b are used. A calibration of the form Y = ax + b was used for the other major elements (Table 2). Calibrations applied to the basaltic Laki volcanic glass sample using the form Y = ax result in all major element contents deviating 2 21% from those calibrated using the form Y = ax + b and, as such, the latter calibration was used. This is due to the different concentrations of these elements in the NIST SRM and USGS basaltic glasses. A calibration of the form Y = ax was adopted for P, Zr, Ba (Table 2), and all the other trace elements (Table 3) for the four volcanic glass samples in the INTAV mount. Trace element concentrations calculated from calibrations in the form of Y = ax + b can sometimes yield negative values. This can affect the results for other major and trace elements, because of the data processing methodology adopted in this study, whereby the total abundance of all elements was set to wt.% by normalizing to the SiO 2 content. Elements that were not analyzed (e.g., S, Se, Hg, and halogens) and H 2 O were assumed to occur at negligible concentrations in the volcanic glass shards. Following Uemoto (2010), detection limits in this study were defined as 3 standard deviations (3s) of the gas-blank intensities. The detection limits of elements in two analytical sessions (Lipari obsidian and Laki tephra; Sheep Track and Old Crow tephras) were calculated from measurements of the NIST SRM reference glasses. Elemental analyses below detection limits were not excluded from data calculations (e.g., average concentrations), in order to avoid introducing statistical bias. The detection limits of 58 elements obtained over two analytical sessions are summarized in Table 4. The relative standard deviation (RSD) values of data for the NIST SRM reference and INTAV volcanic glass samples ratioed to Si are also summarized in Table 4. The RSD values for NIST SRM 610 are <15%, whereas those of NIST SRM 612 tend to be higher than this, particularly for lighter elements. These RSD values for the INTAV samples are almost same as those obtained for element concentrations (Tables 2 and 3). RESULTS AND DISCUSSION Comparison of major element data with values from electron-beam methods The major element data for the four INTAV volcanic glass samples obtained in this study are summarized in Table 2. Also shown are the preferred mean values from electron-beam analytical techniques acquired on a total of 27 instruments in 9 countries (Kuehn et al., 2011) and data from X-ray fluorescence (XRF) and ICP-AES analysis by Kuehn et al. (2011). Figure 2 shows the deviations of the major element compositions obtained in this study from the preferred mean values (Table 2) compiled by Kuehn et al. (2011). Most major element data for the four volcanic glass samples deviate by <10% (Table 2; Fig. 2). Moreover, considering the standard deviations (1s), our data compare Determination of 58 elements in the INTAV volcanic glasses using fsla-icp-ms 411

10 Fig. 2. Major element concentrations obtained in this study normalized to the preferred mean reference values from electronbeam microanalyses compiled by Kuehn et al. (2011) (Table 2). The mean values and standard deviations (1s; in parentheses) of each major element are shown below the symbol. The standard deviations (1s) of the mean values were obtained from the analyses of all four volcanic glass samples. The standard deviations (1s) of the individual glass samples (Table 2) and those of the mean values are shown by the thin gray and thick black bars, respectively. The gray zones represent the ranges of the standard deviations (1s) from the preferred electron-beam mean values. The comments annotated by some larger error bars indicate the INTAV samples to which the error bars belong. S1 = sample 1 (Lipari obsidian); S2 = sample 2 (Sheep Track tephra); S3 = sample 3 (Laki tephra); and OCt = Old Crow tephra. well to the preferred values of Kuehn et al. (2011) (Table 2; Fig. 2). Na 2 O and K 2 O data obtained in this study also deviate by <10% from the preferred values. (Table 2; Fig. 2). This indicates that the electron-beam data compiled by Kuehn et al. (2011) were not significantly affected by migration of alkali elements induced by a combination of charging and heating within the sample by incident electrons (Hayward, 2012). P 2 O 5 contents of Lipari obsidian and Sheep Track and Laki tephras in this study deviate from the preferred values <10% (Table 2; Fig. 2). In contrast, P 2 O 5 contents of the Old Crow tephra deviate more significantly (~74%; Table 2) from the preferred values of Kuehn et al. (2011). However, our data and the preferred values are within analytical uncertainty of each other. In general, our major element data also deviate <10% from those obtained by XRF and ICP-AES presented by Kuehn et al. (2011) (Table 2). However, the MgO value for Lipari obsidian obtained in this study is 35% and 58% lower than those from XRF and ICP-AES analysis, respectively. Moreover, the FeO values for Sheep track tephra obtained in this study are 12 13% lower than those from XRF and ICP-AES analysis, and the MgO value is 44% lower than that from ICP-AES analysis. The Na 2 O and K 2 O contents of Old Crow tephra obtained in this study are 10 12% higher than those by ICP- AES. The P 2 O 5 contents of Lipari obsidian, Sheep Track tephra, and Old Crow tephra obtained by XRF are 0.011, 0.045, and wt.%, respectively (Table 2; Kuehn et al., 2011), which are consistent with those of this study considering the uncertainties. P 2 O 5 contents of Sheep Track and Old Crow tephras obtained by ICP-AES are 0.03 and wt.%, respectively (Table 2; Kuehn et al., 2011), and they are also consistent with those of this study. The slight differences in values obtained by XRF and ICP- AES may be due to sample purification (i.e., the incorporation of other mineral grains with glass shards during sample preparation) and/or intrinsic chemical heterogeneity of the glass shards. ZrO 2 contents of the Lipari obsidian and Sheep Track tephra obtained in this study (0.023 and wt.%, respectively; Table 2) are almost identical to the preferred values from the electron-beam analyses (Kuehn et al., 2011). ZrO 2 contents for the Lipari obsidian obtained by XRF, ICP-AES, LA-ICP-MS, and SN-ICP-MS methods are summarized in Table 2 ( wt.%; Kuehn et al., 2009, 2011) and are consistent with those obtained in this study. The ZrO 2 concentrations of Sheep Track tephra obtained by XRF and ICP-AES (0.152 and wt.%, respectively; Kuehn et al., 2011) are almost identical those 412 S. Maruyama et al.

11 of this study (Table 2). Preferred ZrO 2 values for the Laki and Old Crow tephras obtained by electron-beam methods have not been established (Kuehn et al., 2011), as there were too few ZrO 2 analyses for these samples to assign preferred values (S. Kuehn, pers. comm., 2015). However, values for the Old Crow tephra obtained by other analytical techniques, including XRF, ICP-AES, LA-ICP-MS, SN-ICP- MS, INAA, and SIMS, have previously been published ( wt.%; Westgate et al., 1994; Preece et al., 2000, 2011; Kaufman et al., 2001; Begét and Keskinen, 2003; Pearce et al., 2004; Kuehn et al., 2011) (Table 2). The Zr content of Old Crow tephra obtained in this study is wt.% (Table 3), corresponding to wt.% ZrO 2. In summary, the ZrO 2 contents for the three tephras obtained in this study are similar to those from both electron-beam analytical techniques (Kuehn et al., 2011) and other methods. In contrast, the BaO values obtained in this study are much lower than the preferred values from electron-beam analytical techniques. In this study, the BaO value for Lipari obsidian is wt.% (Table 2), which is only 7% of the preferred value presented by Kuehn et al. (2011). Data for the Sheep Track and Laki tephras obtained in this study are ~10% and ~24%, respectively, of the preferred values in Kuehn et al. (2011). However, the BaO contents of Lipari obsidian and Sheep Track tephra obtained by other analytical techniques are wt.% (Kuehn et al., 2009, 2011) and wt.% (Kuehn et al., 2011), respectively. These values are almost identical to those of our study (Table 2). From the BaO values obtained by electron-beam analytical techniques, those of Lipari obsidian and Sheep Track tephra obtained by Lab 5 (0.007 and wt.%, respectively) are relatively similar to those of this study, and determined by other analytical techniques (Kuehn et al., 2011). The Ba concentration of Old Crow tephra obtained in this study (0.091 wt.% Ba; Table 3) can be converted to wt.% BaO, which is consistent with the BaO data ( wt.%) obtained by other analytical techniques, including XRF and ICP-AES (Kuehn et al., 2011), SN-ICP-MS (Preece et al., 2011), LA-ICP-MS (Westgate et al., 1994; Pearce et al., 2004), INAA (Preece et al., 1999, 2000, 2011; Kaufman et al., 2001), and SIMS (Begét and Keskinen, 2003) (Table 2). BaO data for the Laki tephra obtained by other analytical techniques was not presented by Kuehn et al. (2011), as no sufficiently pure glass separates for XRF or ICP-MS analysis were obtained (S. Kuehn, pers. comm., 2015). It is apparent that BaO contents determined by electron-beam analysis are considerably larger than those from other analytical techniques. This may be due to the X-ray line of Ti overlapping with that of Ba, resulting in erroneously high Ba concentrations (S. Kuehn, pers. comm., 2015). Trace elements in the INTAV volcanic glass samples Trace element data for the four INTAV volcanic glasses obtained in this study are summarized in Table 3. The RSD of the trace element data are mostly <100% (Table 3). Data for low concentration elements such as Be, In, and Bi, and some transition metals such as Cr, Co, Ni, Cu, Ag, and Cd, tend to exhibit higher RSD values. Lipari obsidian is regarded as an important primary or secondary standard material for EPMA (Kuehn et al., 2009) and in the comparison of archaeological obsidian artifacts in the Mediterranean area (e.g., Yamada, 2013). Therefore, the Lipari obsidian has previously been studied in detail. Trace element data for the Lipari obsidian obtained in previous studies are summarized in Table 5 and a comparison of these with our data shown in Fig. 3a. Kuehn et al. (2009) presented results for 31 trace elements (e.g., Sc, Cr, Ni, REE, Pb, Th, and U) in the Lipari obsidian ID3506 determined by XRF, SN-ICP-MS, and LA-ICP-MS techniques. Six of 24 trace elements obtained by LA-ICP-MS (Kuehn et al., 2009) deviate beyond analytical uncertainties from those of this study. The SN-ICP- MS data presented by Kuehn et al. (2009) are slightly higher than those of this study (Fig. 3a). Fifteen of 25 trace elements from our study are 8 28% lower than those determined by SN-ICP-MS, apart from Nb (~12% higher than that from SN-ICP-MS), and do not agree within analytical uncertainties (1s). Compared with the XRF results presented by Kuehn et al. (2009), 8 out of 18 trace elements (e.g., Sc, Pb, and Th) are different from those of this study. For example, our Sc and Th values are 587% and 69% of the XRF values (Fig. 3a). The results of our study compared with those from LA-ICP-MS methods (Kuehn et al., 2009) show that the data obtained using bulk analysis techniques (i.e., XRF and SN-ICP-MS) tend to be different from those obtained using LA-ICP-MS and, in particular, for Sc. A mm-sized fragment of Lipari obsidian can be readily obtained and, as such, sample purification is straightforward. Hence, these differences may be due to large-scale spatial heterogeneity of trace elements within Lipari obsidian ID3506, which is not readily detected by LA-ICP-MS analysis (i.e., a mm-sized analytical technique). In addition, some of the differences may reflect the range of analytical techniques used. Some previous studies have analyzed trace element abundances in other Lipari obsidian samples using LA- ICP-MS (Barca et al., 2007) and XRF methods (Francaviglia, 1999; De Francesco et al., 2008). A comparison between these and our data is shown in Fig. 3b. Considering our data, only 2 out of 29 trace elements (Sc and Nb) differ beyond uncertainties (1s) from those obtained using LA-ICP-MS by Barca et al. (2007). In contrast, data determined by XRF methods from both Francaviglia (1999) and De Francesco et al. (2008) are slightly different to those of this study and also those of Determination of 58 elements in the INTAV volcanic glasses using fsla-icp-ms 413

12 Table 5. Trace element data (mg/g) for the Lipari obsidian and Old Crow tephra obtained by previous studies Lipari obsidian ID3506 (Sample 1) Lipari obsidian other than ID3506 XRF SN-ICP-MS LA-ICP-MS XRF LA-ICP-MS a Kuehn et al. (2009) Kuehn et al. (2009) Kuehn et al. (2009) De Francesco et al. (2008) Francaviglia (1999) Barca et al. (2007) (mg/g) mean 1s mean 1s mean 1s mean 1s mean mean 1s n unknown b 8 Sc V Cr Co 3 0 Ni Cu 11 1 Zn Ga Rb Sr Y Zr Nb Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta Pb Th U Kuehn et al. (2009) (Fig. 3b). Our data for Nb, Rb, Zr, La, and Ce are % of those obtained by De Francesco et al. (2008), and Cr and V are 44% and 40% lower, respectively, than the data of De Francesco et al. (2008). The data of V, Cr, Rb, Zr, Nb, La, and Ce of De Francesco et al. (2008) are similar to those of this study and within analytical uncertainties (1s) (Tables 3 and 5). However, Sr and Ba values are 32% and 83% higher, respectively, than those of this study. Moreover, the Ni and Co concentrations obtained by De Francesco et al. (2008) are 20% and 13%, respectively, of those of this study. Rb, Y, and Zr data from this study are 8 37% lower than those obtained by Francaviglia (1999), and Nb is 40% higher. The Y, Zr, and Nb data differ beyond the analytical uncertainties (1s) of our study and that (±10%) of Francaviglia (1999). Only Sr concentration obtained in this study is almost identical to that obtained by Francaviglia (1999) (~14 mg/g; Tables 3 and 5). These differences may be due to multiple sources of Lipari obsidian (S. Kuehn, pers. comm., 2015), which may substantiate the hypothesis that large-scale trace element heterogeneity characterizes the Lipari obsidian samples, rather than the differences being analytical artifacts. The Sc concentration in Lipari obsidian obtained in this study is several times higher than reported values (Tables 3 and 5; Figs. 3a and b). The Old Crow tephra Sc data are also times higher than previously reported data (Tables 3 and 5; Figs. 3c and d). These differences may be due to heterogeneous distribution of Sc in volcanic glasses, given the relatively high RSD values for Sc in 414 S. Maruyama et al.

13 Table 5. (continued) Old Crow tephra SN-ICP-MS Westgate et al. (1994) Kaufman et al. (2001) Pearce et al. (2004) Preece et al. (2011) c UT613 UT1434b UT815b UT1434 UT815 (mg/g) mean 1s mean mean mean mean mean 1s n Sc V Cr Co Ni Cu Zn Ga 15 1 Rb Sr Y Zr Nb Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta Pb 18 3 Th U the INTAV volcanic glass samples (20 100%; Table 3). Alternatively, an uncorrected interference from 29 Si 16 O on 45 Sc in these glasses with high SiO 2 and low Sc may be responsible (Y. Kon, pers. comm., 2015). The Old Crow tephra has been the most extensively analyzed INTAV volcanic glass sample for trace elements, including by LA-ICP-MS, SN-ICP-MS, INAA, and SIMS methods (Table 5). A comparison between our data and those obtained in previous studies is shown in Figs. 3c e. Taking into account analytical uncertainties, data for more than 80% and 70% of the trace elements obtained in this study are consistent with those from previous LA- ICP-MS (Westgate et al., 1994; Pearce et al., 2004) and SN-ICP-MS (Westgate et al., 1994; Kaufman et al., 2001; Pearce et al., 2004; Preece et al., 2011) studies, respectively. Sc, Zn, and U concentrations determined this study are 23%, 20%, and 7% higher, respectively, than those from INAA (Preece et al., 2000), whereas values for 14 other trace elements (e.g., Cs, Ba, and some lanthanides) are 3 30% lower than those from INAA. Fifteen of 16 trace element analyses by INAA from Westgate et al. (1994) are similar to those of this study considering the analytical uncertainties (1s). Only the Ce value from this study is slightly lower (~8%) than the INAA data (Westgate et al., 1994). The Rb, Sr, Y, Zr, Nb, and Ba data from this study are 6 27% lower than those obtained by SIMS (Begét and Keskinen, 2003) (Fig. 3c). However, Ce, Th, and U data are 6 38% higher than the SIMS data. Nb, Ce, and U concentrations are within analytical uncertainties (1s) in all studies. As shown in Figs. 3c and d, Determination of 58 elements in the INTAV volcanic glasses using fsla-icp-ms 415

14 Table 5. (continued) Old Crow tephra LA-ICP-MS Westgate et al. (1994) UT613 (1991) d UT613 (1992) UT502 (1991) UT502 (1992) (mg/g) mean 1s mean 1s mean 1s mean 1s n Sc V Cr Co Ni Cu Zn Ga Rb Sr Y Zr Nb Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta Pb Th U Sc concentrations obtained in this study tend to be higher than those from previous studies (Preece et al., 2000, 2011). Concentrations of B, P, and Cr in the INTAV volcanic glasses In contrast to the Lipari obsidian and Old Crow tephra, there are few previous studies of the trace element concentrations in the Sheep Track and Laki tephras. Kuehn et al. (2011) noted that analyses of Sheep Track tephra sample have been carried out by ICP-AES and XRF methods. However, the detailed analytical results for this were not presented. Brounce et al. (2012) analyzed Li and B concentrations in tephra matrix glasses from the Lakagígar (Laki Fissure) eruption using SIMS. The concentrations of Li and B reported by Brounce et al. (2012) are and mg/g, respectively. The Li concentration determined in our study (Table 3) falls within the range reported by Brounce et al. (2012), whereas that of B (Table 3) is different. The B concentrations of Laki volcanic glass reported by Brounce et al. (2012) are below the B detection limits in this study (3.4 mg/g; Table 3). The concentrations of B in the INTAV volcanic glasses, apart from the Lipari obsidian, are considerably higher than those of the other trace elements, and their standard deviations are quite large (Table 3). The RSD of B concentrations in the Lipari obsidian is 10%, whereas those of the Sheep Track, Laki, and Old Crow tephras are 48%, 207%, and 49%, respectively (Table 3). The RSD values of B/Si ratios for the NIST SRM 610 and 612 reference glasses during the analytical session of the Lipari obsid- 416 S. Maruyama et al.