Revisiting the electron microprobe method of spinel-olivine-orthopyroxene oxybarometry applied to spinel peridotitesk
Fred A.Davis, ElizabethCottrell, Suzanne K.Birner, Jessica M.Warren, Oscar G.Lopez
DOI: 10.2138/am-2017-5823 Published on February 2017, First Published on February 01, 2017
Fred A. Davis
National Museum of Natural History, Smithsonian Institution, Washington, D.C. 20560, U.S.A.
Department of Earth and Environmental Sciences, University of Minnesota Duluth, Duluth, Minnesota 55812, U.S.A.
Uncorrected electron microprobe analyses of Wood spinels from sessions S1–S3. Sample-average uncorrected Fe3+/ΣFe ratios determined by EPMA in sessions S1–S3 (Table 2) plotted against Fe3+/ΣFe ratios determined by Mössbauer spectroscopy. Vertical bars show the range of compositions for a given sample across all grains measured (Supplementary1 Table S1), indicating the degree of intergranular heterogeneity exhibited by a sample. Circles represent samples chosen for the correction set, and triangles represent samples chosen for the validation set. All other Wood spinels are represented with diamonds.
Compositional range of spinels included in the correction and validation sets. Fe3+/ΣFe ratios by Mössbauer are from Wood and Virgo (1989), Bryndzia and Wood (1990), and Ionov and Wood (1992). MgO and Cr# of the correction set and validation set are EPMA measurements from this study (Table 2), while values for the correction standards used by Wood and Virgo (1989) are as reported in that study. The correction set used in this study spans a similar range of Cr# and MgO (b) as the correction standards used by Wood and Virgo (1989), and Fe3+/ΣFe ratios span a larger range (a). Taking all of these data together, Fe3+/ΣFe ratio is not correlated with Cr# (r2 = 0.003) or MgO (not shown, r2 < 0.001), and MgO and Cr# are highly correlated (r2 = 0.93).
Examples of the W&V89 correction applied to spinels from several independent analytical sessions. The first column (a, c, and e) shows uncorrected Fe3+/ΣFe ratios by EPMA of the correction set spinels measured at the start of each session and at the end of each session, along with validation set spinels analyzed in between. These are plotted against their published Fe3+/ΣFe ratios measured by Mössbauer. The second column (b, d, and f) shows the same measurements after correction of the EPMA data using the Wood and Virgo (1989) method. Uncorrected EPMA analyses from session A1 plot around the 1:1 line (a), and the W&V89 correction causes only imperceptible changes to the corrected Fe3+/ΣFe ratios (b). Uncorrected EPMA analyses from session B3 (c and d) are offset from the 1:1 line and display a relatively high degree of scatter around the trend with Mössbauer data (c). The W&V89 correction decreases scatter in the data and shifts it upward so that the corrected data lie on the 1:1 line (d). Uncorrected EPMA analyses from session B4 (e and f) are offset from the 1:1 line but are relatively tightly clustered along a linear trend with the Mössbauer data (e). The W&V89 correction shifts Fe3+/ΣFe ratios onto the 1:1 line (f).
Relationship between Cr# and ΔFe3+/ΣFeMöss-EPMA in uncorrected analyses of the correction set spinels. Cr# and ΔFe3+/ΣFeMöss-EPMA are the measured parameters that contribute directly to the W&V89 correction and the same analytical sessions are shown as in Figure 3. In session A1 (a), Cr# and ΔFe3+/ΣFeMöss-EPMA are uncorrelated (r2 = 0.01) and ΔFe3+/ΣFeMöss-EPMA is near zero, so the W&V89 correction makes negligible adjustments to the Fe3+/ΣFe ratio, as expected given that the uncorrected data already overlapped the Mössbauer values. In session B3 (b), Cr# and ΔFe3+/ΣFeMöss-EPMA are correlated (r2 = 0.81), with slope and intercept both significantly different from zero. The W&V89 correction shifts Fe3+/ΣFe ratios of all samples upward and Cr-poor spinels are adjusted more than Cr-rich spinels. In session B4 (c), Cr# and ΔFe3+/ΣFeMöss-EPMA are poorly correlated (r2 = 0.10) with slope near zero but an intercept significantly different from zero; consequently, the W&V89 correction shifts all Fe3+/ΣFe ratios upward by a nearly constant correction factor.
Literature compilation of spinel Fe3+/ΣFe ratios measured by Mössbauer spectroscopy and calculated from EPMA. Uncorrected spinel Fe3+/ΣFe ratios calculated from EPMA analyses of natural peridotite- and basalt-hosted spinels plotted against Fe3+/ΣFe ratios of the same spinels analyzed by Mössbauer spectroscopy (a). Uncorrected spinel Fe3+/ΣFe ratios by EPMA are biased to low Fe3+/ΣFe, with a mean ΔFe3+/ΣFeMöss-EPMA of 0.022 ± 0.049 (1σ). After correction by the W&V89 method, Fe3+/ΣFe ratios deviate less from the 1:1 line and are more evenly distributed around it (b), with a mean ΔFe3+/ΣFeMöss-EPMA of −0.007 ± 0.021 (1σ).
Mean Fe3+/ΣFe ratios of the validation set spinels measured by Mössbauer spectroscopy and calculated from EPMA. Mean uncorrected (a) and corrected (b) Fe3+/ΣFe ratios by EPMA (Table 4) were calculated by taking the unweighted average of the mean Fe3+/ΣFe ratios of all analytical sessions (A1–A4 and B1–B4; Supplementary1 Table S2). Error bars are 2 st.dev.
Relationship between analytical precision of spinel Fe3+/ΣFe ratios and total concentration of Fe for spinels from the validation set, Hawaiian xenoliths and Tonga. Magnitude of 1 st.dev. in corrected Fe3+/ΣFe ratios measured across all sessions (A1–A4 and B1–B4, Supplementary1 Table S2) for validation set and Hawaiian spinels as a function of the inverse of the multi-session average total Fe concentration on a 3 cation basis (a). In black is the best fit line through the origin (r2 = 0.57). Tonga sample BMRG08-98-2-2 was not included in this fit because it does not plot near the global trend in Cr#-MgO (Fig. 8). The equation for this line is given in the text (Eq. 3), and we use it to calculate precision in our corrected measurements of Fe3+/ΣFe ratios of unknown spinel samples. Deviations of each session average (Supplementary1 Table S2) from their respective multisession means (Table 4) plotted as a function of total Fe concentration on a 3 cation basis (b). The 1σ error envelope is calculated using Equation 3.
Relationship between MgO concentration and Cr# of natural peridotite spinels. Samples are separated by tectonic setting: abyssal peridotites (n = 743) from the compilation of Warren (2016): (Prinz et al. 1976; Hamlyn and Bonatti 1980; Dick and Bullen 1984; Michael and Bonatti 1985; Shibata and Thompson 1986; Dick 1989; Bryndzia and Wood 1990; Johnson et al. 1990; Juteau et al. 1990; Komor et al. 1990; Bonatti et al. 1992, 1993; Cannat et al. 1992; Johnson and Dick 1992; Snow 1993; Constantin et al. 1995; Arai and Matsukage 1996; Dick and Natland 1996; Ghose et al. 1996; Jaroslow et al. 1996; Niida 1997; Ross and Elthon 1997; Stephens 1997; Hellebrand et al. 2002a, 2002b; Brunelli et al. 2003; Hellebrand and Snow 2003; Seyler et al. 2003, 2007; Coogan et al. 2004; Workman and Hart 2005; Morishita et al. 2007; Cipriani et al. 2009; Warren et al. 2009; Brunelli and Seyler 2010; Dick et al. 2010; Warren and Shimizu 2010; Zhou and Dick 2013; Lassiter et al. 2014; Mallick et al. 2014; D’Errico et al. 2016), continental xenoliths not associated with subduction (n = 154): (Wood and Virgo 1989; Ionov and Wood 1992; Woodland et al. 1992), and supra-subduction zone for xenoliths and seafloor drilled samples from subduction-related settings (n = 85): (Wood and Virgo 1989; Canil et al. 1990; Luhr and Aranda-Gómez 1997; Parkinson and Pearce 1998). MgO and Cr# are correlated in the global data set (solid line, r2 = 0.82, slope = −15.0 ± 0.2, intercept = 21.46 ± 0.07, 1σ). The slope defined by the correction set used in this study (dashed line, r2 = 0.94, n = 7, slope = −11.6 ± 1.4, intercept = 21.5 ± 0.4, 1σ) is shallower, as is the line defined by the Wood spinels (not shown, r2 = 0.92, n = 32, slope = −12.7 ± 0.7, intercept = 21.80 ± 0.17, 1σ). Also shown are Hawaiian xenoliths and Tonga peridotite BMRG08-98-2-2 from this study.
Effect of activity of magnetite in spinel on the calculation of relative
relative to the quartz-fayalite-magnetite buffer (ΔQFM, Frost 1991 calibration) using all input parameters from sample 114923-57 at 1038 °C and 1.5 GPa and varying the value of
while holding Mg#Ol and
constant (a). The dashed lines show ±1σ error on the corrected EPMA measurement of spinel Fe3+/ΣFe ratio calculated using Equation 3. Dotted lines show ±1σ error on the uncorrected EMP measurement of spinel Fe3+/ΣFe ratio assuming a twofold increase in uncertainty for uncorrected measurements (see text). The increased uncertainty in
at low activities of magnetite has been demonstrated previously by Ballhaus et al. (1991) and Parkinson and Arculus (1999). The dependence of
(ΔQFM) on Fe3+/ΣFe ratio rather than activity of magnetite (b).
and activity of magnetite in each of the four Hawaiian xenoliths and comparison between Wood (1991) and Ballhaus et al. (1991) formulations of the spinel-olivine-orthopyroxene oxybarometer. Log
(ΔQFM; a) and
(b), calculated from corrected Fe3+/ΣFe ratios, for each session in which a given spinel was analyzed (Supplementary1 Table S4 and S5). The olivine and orthopyroxene compositions in Table 3 were used for all
calculations. Uncertainty in
(ΔQFM) includes contributions from analytical uncertainty on each phase (a). Uncertainty in
was determined as described in the Supplementary1 material. Uncertainty in corrected Fe3+/ΣFe ratios was calculated using Equation 3.
relative to the QFM buffer (Frost 1991), calculated for each measurement of the four Hawaiian spinel lherzolite xenoliths (c). Relative
on the x-axis was calculated using the Equation 4 and the MELTS Supplemental Calculator (Sack and Ghiorso 1991a, 1991b) to calculate
on the y-axis was calculated following the methodology of Ballhaus et al. (1991). Error using the Ballhaus et al. (1991) method was estimated by propagating through the Ballhaus et al. (1991) oxybarometer our estimates of uncertainty in spinel Fe3+/ΣFe ratio and olivine Mg#.