Elsevier

Chemical Geology

Volume 505, 5 February 2019, Pages 48-56
Chemical Geology

Low analytical totals in EPMA of hydrous silicate glass due to sub-surface charging: Obtaining accurate volatiles by difference

https://doi.org/10.1016/j.chemgeo.2018.11.015Get rights and content

Highlights

  • Literature data overestimates volatiles in silicate glass using volatiles by difference method (VBD).

  • Silicate glass with >2 wt% volatiles are overestimated by ~1 wt% on average.

  • Sub-surface charging decreases X-ray intensity and analytical totals, increasing VBD.

  • Using glass standards to calibrate VBD achieves accuracy of ±0.1 wt%.

Abstract

The major and minor element chemistry of silicate glass is commonly measured using electron probe micro-analysis (EPMA). The volatile content (H2O ± CO2) can, additionally, be quantified using “volatiles by difference” (VBD), but a review of literature data shows that this method consistently overestimates the volatile content. We propose that sub-surface charging during EPMA reduces analytical totals, consequently elevating VBD. Sub-surface charging produces an internal electric field due to trapped implanted electrons, resulting in fewer X-rays being generated and their depth of generation being shallower. The maximum electric field strength required to produce the observed overestimation of VBD is calculated to be ~10−1 V·nm−1. Crystals are often used as standards for glass analysis but, as amorphous materials have more defects in the band gap, glasses can trap more electrons resulting in greater amounts of sub-surface charging. As this is not included in matrix corrections, it causes errors for glass analyses, but not for crystal analyses. By calibrating VBD using hydrous glass standards, the effect of charging can be incorporated, and volatile contents can be determined to an accuracy of ±0.1 wt%, compared to overestimation by ~1 wt% using conventional VBD methods.

Introduction

Electron probe micro-analysis (EPMA) is a critical technique for analysing the composition of silicate glass in volcanology and petrology, such as in melt inclusions and interstitial glass (e.g., Faure and Schiano, 2005). Major and minor element concentration changes, which can be quantified directly using EPMA, provide information on the diversity of magmas (e.g., primary magma composition and mixing events prior to eruption) and their pre-eruptive crystallisation history (e.g., Kent, 2008; Michael and Cornell, 1998). Glass composition can be used in combination with mineral chemistry to test for equilibrium conditions (e.g., Fe-Mg exchange between melt and olivine, Roeder and Emslie, 1970) and estimate magma pressures and temperatures (e.g., olivine-, feldspar-, and pyroxene-melt thermobarometry, see Putirka, 2008, for a review). However, the concentration of the key volatile components (H2O and CO2) that have a profound effect on the physical properties of melts (density and viscosity, e.g., Giordano and Dingwell, 2003; Ochs and Lange, 1999), phase relations (e.g., Feig et al., 2006), degassing and eruptive style (e.g., Métrich and Wallace, 2009), cannot be easily and directly determined by EPMA. This shortcoming limits significantly the utility of EPMA in understanding volcanic processes.

One approach to this issue is to estimate the H2O + CO2 content of silicate glass by EPMA using the indirect “volatiles by difference” (VBD) method, whereby the discrepancy between the analytical total for measurable (major and minor) elements and 100 wt% provides an estimate for the total volatile content (Blundy and Cashman, 2008; Devine et al., 1995; Humphreys et al., 2006; King et al., 2002; Nash, 1992). Many trace elements are not analysed by EPMA and if they occur in high abundances will lead to an underestimation of the total. Typically, individual major elements are measured to ~1 % relative error, which results in a ±0.5–0.7 wt% error on VBD, corresponding to a combination of the errors on individual elements (Devine et al., 1995; Humphreys et al., 2006). The volatile component by VBD cannot be separated into H2O and CO2 but, as H2O is an order of magnitude more soluble in silicate melts than CO2, most of the VBD is H2O. The VBD method has been used widely to quantify the volatile content of experimental samples (e.g., Botcharnikov et al., 2008; Di Carlo et al., 2006; Erdmann and Koepke, 2016) and natural samples such as melt inclusions (e.g., Holtz et al., 2004; Métrich et al., 2004; Rutherford and Devine, 1996; Sommer, 1977).

There are a variety of techniques that can directly and precisely analyse H2O and CO2 in silicate glass, such as SIMS, FTIR and Raman (e.g., Hauri et al., 2002; Newman et al., 1986; Thomas, 2000). For comparison, EPMA has higher spatial resolution than SIMS and FTIR (~5 μm diameter using EPMA compared to ~15 μm for SIMS or ~100 μm for FTIR) and is more widely accessible (and less expensive) than SIMS. Also, EPMA does not suffer from problems due to fluorescence or the presence of nanolites, which can effect quantification using Raman (e.g., Di Genova et al., 2017b). Therefore, EPMA is often used to estimate H2O when other techniques are unavailable and, uniquely, provides the complete major and minor element glass chemistry in a single analysis.

A review of literature data (n = 524, see Supplementary material for complete dataset) of VBD compared to “measured volatile content” (H2O, and CO2 where available) is summarised in Table 1 and Fig. 1. In these studies, H2O concentration is measured using FTIR, SIMS, Karl-Fischer titration, or assumed in accordance with experimental conditions (e.g. solubility), whilst CO2 concentration is measured using SIMS or FTIR (Table 1). Errors are not shown but are typically <10 % relative for measured volatile contents and < 0.7 wt% for VBD. Most (n = 287) of the data is for measured volatile contents <2 wt%, with slightly fewer (n = 226) analyses of volatile-rich glasses with 2–6 wt%. There are very few data (n = 11) for glass with >7 wt% measured volatile content. If VBD and measured volatile content agreed, the data would be evenly distributed around the 1-to-1 trend (Fig. 1a), with equal number of analyses under- and overestimating the measured volatile content (Fig. 1c). Instead, most of the data lie above the 1-to-1 trend (Fig. 1a), with more analyses (>50 %) overestimating the measured volatile content (Fig. 1c). This indicates a systematic error, which is not necessarily obvious in small datasets. Including all data (n = 524), the volatile content is overestimated in 63.5 % of analyses (mean overestimation 0.41 wt% with one standard deviation, 1σ, 1.16 wt%). For data with measured volatile contents <2 wt% (n = 287), VBD overestimates the volatile content in 54.7 % of analyses (mean overestimation 0.08 wt%, 1σ 0.72 wt%) (Fig. 1b). For measured volatiles >2 wt% (n = 237), overestimation occurs in 74.3 % of analyses (mean overestimation 0.81 wt%, 1σ 1.43 wt%). Either EPMA VBD overestimates the true volatile content, or techniques such as SIMS, FTIR or Karl-Fischer titration underestimate the true volatile content. As a variety of different techniques are used to quantify the measured volatile content in these literature data, it seems unlikely that all these techniques would underestimate the true volatile content to the same degree. Hence, it is considered more likely that VBD consistently overestimates the true volatile content.

Evidently, VBD is accurate at low volatile contents (<2 wt%), but consistently overestimates the volatile content in volatile-rich glass (>2 wt%) by nearly 1 wt%. Such large discrepancies would have significant impact on the calculated physical and chemical properties of the melt and, in turn, its behaviour before and during volcanic eruptions. For instance, a 1 wt% overestimation in H2O concentration could change the calculated entrapment pressure of water-saturated melt inclusions by up to ~50 MPa, equivalent to ~2 km depth change (Newman and Lowenstern, 2002). Similarly, Di Genova et al. (2013) calculate that the viscosity difference between 2.5 and 4.0 wt% dissolved H2O is approximately an order of magnitude (102.4 to 103.3 Pa·s at 1023 K). Therefore, it is important to understand the cause of the overestimation of VBD and to develop a method to improve the accuracy of VBD measurements.

Section snippets

Volatiles by difference using electron probe micro-analysis

EPMA uses the intensity of characteristic X-rays, generated by bombarding a sample with an electron beam, to measure its composition. Typically, Kα X-ray lines are used for quantification of elements with atomic number < 30, as they have the highest intensity of X-rays emitted from a specific atom. Kα X-rays are generated by an incident electron ejecting an electron from the innermost shell (K shell) of a target atom, which is replaced by an electron from the shell above (L shell), emitting the

Methods

We modelled the effect of sub-surface charge on glass analysis using the Monte Carlo simulation program Win X-ray (Demers and Gauvin, 2004; Gauvin et al., 2006), which incorporates the charge density model proposed by Cazaux (1996). We then use the results from Win X-ray to calculate the VBD.

Results

Fig. 2 gives the φ(ρZ) curves of the Kα lines of interest, in order of increasing X-ray energy from upper left to lower right. Values of the φ(ρZ) function of the emitted curve [φ(ρZ)e] are always less than the generated curve [φ(ρZ)g], as some X-rays are always absorbed. At shallow depths in the sample, φ(ρZ)g for Fmax = 0.00 and 0.20 V·nm−1 are comparable, but at greater depths, the intensity of generated X-rays falls off more quickly for Fmax = 0.20 V·nm−1. For Si, Al, Mg and Na calculated

Effect of Fmax

Modelling results confirm that more low energy X-rays (<1 keV) and fewer high energy X-rays (>1 keV) are emitted when an electric field is present. This is because, although fewer X-rays are generated overall, they are generated at shallower depths reducing X-ray absorption which most affects strongly absorbed, rather than weakly absorbed, X-rays. Moreover, the deceleration of electrons will affect high energy X-rays more than low energy X-rays as the overvoltage (ratio of the accelerating

Conclusions

Sub-surface charging is an important process to consider during EPMA of insulating materials, especially hydrous silicate glass, due to its effect on quantitative analysis (Cazaux, 1996). Sub-surface charging causes element migration and redox changes during analysis (e.g., Hughes et al., 2018; Humphreys et al., 2006; Zhang et al., 2018). Our Win X-ray modelling shows that sub-surface charging can also have a measurable effect on X-ray generation and emission, resulting in low analytical totals

Acknowledgements

We would like to thank David Neave (Leibniz Universität) for the R script used to calculate glass density. ECH is supported by a NERC GW4+ DTP studentship from the Natural Environment Research Council, United Kingdom (NE/L002434/1) and is thankful for the support and additional funding from CASE partner GNS Science, New Zealand. GK acknowledges support from the New Zealand Strategic Science Investment Fund, New Zealand. We would like to thank Michael Stock and an anonymous reviewer for their

References (91)

  • F. Gaillard et al.

    The effect of water and fO2 on the ferric–ferrous ratio of silicic melts

    Chem. Geol.

    (2001)
  • O. Gedeon et al.

    Fast migration of alkali ions in glass irradiated by electrons

    J. Non-Cryst. Solids

    (1999)
  • E.H. Hauri et al.

    SIMS analysis of volatiles in silicate glasses

    Chem. Geol.

    (2002)
  • O. Jbara et al.

    A new experimental approach for characterizing the internal trapped charge and electric field build up in ground-coated insulators during their e irradiation

    Nucl. Instruments Methods Phys. Res. Sect. B Beam Interact. With Mater. Atoms

    (2002)
  • N. Métrich et al.

    2001 flank eruption of the alkali- and volatile-rich primitive basalt responsible for Mount Etna's evolution in the last three decades

    Earth Planet. Sci. Lett.

    (2004)
  • A. Namiki et al.

    Transition between fragmentation and permeable outgassing of low viscosity magmas

    J. Volcanol. Geotherm. Res.

    (2008)
  • V.B. Naumov

    Rhyolitic melts in eastern Transbaikalia and the North Caucasus: chemical composition, volatiles, and admixture elements (from data of study of melt inclusions in minerals)

    Russ. Geol. Geophys.

    (2011)
  • S. Newman et al.

    VolatileCalc: a silicate melt–H2O–CO2 solution model written in Visual Basic for excel

    Comput. Geosci.

    (2002)
  • A.R.L. Nichols et al.

    Is the Iceland hot spot also wet? Evidence from the water contents of undegassed submarine and subglacial pillow basalts

    Earth Planet. Sci. Lett.

    (2002)
  • O. Shorttle et al.

    Fe-XANES analyses of Reykjanes Ridge basalts: implications for oceanic crust's role in the solid Earth oxygen cycle

    Earth Planet. Sci. Lett.

    (2015)
  • J.S. Turner et al.

    Convection and mixing in magma chambers

    Earth-Sci. Rev.

    (1986)
  • J.D. Webster et al.

    Determining pre-eruptive compositions of late Paleozoic magma from kaolinized volcanic ashes: analysis of glass inclusions in quartz microphenocrysts from tonsteins

    Geochim. Cosmochim. Acta

    (1995)
  • J.D. Webster et al.

    Volatile and lithophile trace-element geochemistry of Mexican tin rhyolite magmas deduced from melt inclusions

    Geochim. Cosmochim. Acta

    (1996)
  • M. Wilke et al.

    The oxidation state of iron in silicic melt at 500 MPa water pressure

    Chem. Geol.

    (2002)
  • Y. Zhang et al.

    New calibration of infrared measurement of dissolved water in rhyolitic glasses

    Geochim. Cosmochim. Acta

    (1997)
  • G. Bastin et al.

    Electron probe quantification

    (1991)
  • H. Benhayoune et al.

    Some experimental evidence for the Φ(0) variation in EPMA of insulating materials

    X-Ray Spectrom.

    (1996)
  • J.D. Blundy et al.

    Petrologic reconstruction of magmatic system variables and processes

    Rev. Mineral. Geochem.

    (2008)
  • C. Bonnelle

    Charge trapping in dielectrics

    Microsc. Microanal.

    (2004)
  • R.E. Botcharnikov et al.

    Phase relations and liquid lines of descent in hydrous ferrobasalt: implications for the Skaergaard Intrusion and Columbia River Flood Basalts

    J. Petrol.

    (2008)
  • Y. Bottinga et al.

    Densities of liquid silicate systems calculated from partial molar volumes of oxide components

    Am. J. Sci.

    (1970)
  • J. Cazaux

    Electron probe microanalysis of insulating materials: quantification problems and some possible solutions

    X-Ray Spectrom.

    (1996)
  • A. Cherifi et al.

    The validation of the thermal step method

    IEEE Trans. Electr. Insul.

    (1992)
  • H. Demers et al.

    X-ray microanalysis of a coated nonconductive specimen: Monte Carlo simulation

    Microsc. Microanal.

    (2004)
  • J.D. Devine et al.

    Comparison of microanalytical methods for estimating H2O contents of silicic volcanic glasses

    Am. Mineral.

    (1995)
  • I. Di Carlo et al.

    Experimental crystallization of a high-K arc basalt: the golden pumice, Stromboli Volcano (Italy)

    J. Petrol.

    (2006)
  • D. Di Genova et al.

    Models for the estimation of Fe3+/Fetot ratio in terrestrial and extraterrestrial alkali- and iron-rich silicate glasses using Raman spectroscopy

    Am. Mineral.

    (2016)
  • D. Di Genova et al.

    A compositional tipping point governing the mobilization and eruption style of rhyolitic magma

    Nature

    (2017)
  • J.J. Donovan et al.

    An improved mean atomic number background correction for quantitative microanalysis

    Microsc. Microanal.

    (1996)
  • J.J. Donovan et al.

    Water by EPMA - New developments

    Microsc. Microanal.

    (2008)
  • D.L. Drew et al.

    Initiation of large-volume silicic centers in the Yellowstone hotspot track: insights from H2O- and F-rich quartz-hosted rhyolitic melt inclusions in the Arbon Valley Tuff of the Snake River Plain

    Contrib. Mineral. Petrol.

    (2016)
  • M. Erdmann et al.

    Silica-rich lavas in the oceanic crust: experimental evidence for fractional crystallization under low water activity

    Contrib. Mineral. Petrol.

    (2016)
  • S. Fakhfakh et al.

    An experimental approach for dynamic investigation of the trapping properties of glass-ceramic under electron beam irradiation from a scanning electron microscope

    Eur. Phys. J. Appl. Phys.

    (2003)
  • S. Fakhfakh et al.

    Analysis of electrical charging and discharging kinetics of different glasses under electron irradiation in a scanning electron microscope

    J. Appl. Phys.

    (2010)
  • S.T. Feig et al.

    Effect of water on tholeiitic basalt phase equilibria: an experimental study under oxidizing conditions

    Contrib. Mineral. Petrol.

    (2006)
  • Cited by (36)

    • Micro-Raman water calibration in ultrapotassic silicate glasses: Application to phono-tephrites and K-foidites of Colli Albani Volcanic District (Central Italy)

      2022, Chemical Geology
      Citation Excerpt :

      From chemical analyses, the water contents of glasses were estimated according to the by difference method (Devine et al., 1995; Humphreys et al., 2006). We are aware that this method might not be fully reliable given the estimation is affected by i) element concentrations not measured by EMPA and ii) surface charge effects (Hughes et al., 2019) leading to an overestimation of H2O concentration. For this reason we performed an external calibration using the dataset of Di Genova et al. (2014) for which both chemical analyses and water contents measured through KFT are available (Online Resource 1).

    • Uncommon K-foiditic magmas: The case study of Tufo del Palatino (Colli Albani Volcanic District, Italy)

      2021, Lithos
      Citation Excerpt :

      From chemical analyses, the H2O ± CO2 contents of glasses were estimated according to the by-difference method (Devine et al., 1995; Humphreys et al., 2006). We are aware that this method is not fully reliable given the estimation is affected by element concentrations not measured by EMPA and surface charge effects (Hughes et al., 2019) leading to an overestimation of H2O concentration. However, the obtained H2O values are close to those determined by micro-Raman spectroscopy measurements (Di Genova et al., 2017; Le Losq et al., 2012) on the same glasses.

    View all citing articles on Scopus
    View full text