Quantitative compositional mapping of mineral phases by electron

probe micro-analyser

Pierre Lanari1*, AliceVho1, Thomas Bovay1, Laura Airaghi2 & Stephen Centrella3

1
Institute of Geological Sciences, University of Bern, Baltzestrasse 1+3, 3012 Bern,
Switzerland
2
Univ. Grenoble Alpes, CNRS, ISTerre, F-38000 Grenoble, France
3
Institut für Mineralogie, University of Münster, D-48149 Münster, Germany
* corresponding author (pierre.lanari@geo.unibe.ch)

Accepted for publication in

Geological Society of London Special Publication:
“Metamorphic Geology: Microscale to Mountain Belts”
Editors: Ferrero, S., Lanari, P., Goncalves, P., Grosch, E.G.

Abstract: Compositional mapping has greatly impacted the mineralogical and

petrological studies over the past half-century while the electron probe micro-analyser
became increasingly used. Many technical and analytical developments have
benefited from the synergies of physicists and geologists and they have greatly
contributed to the success of this analytical technique. Large-area compositional
mapping has become routine practice in many laboratories worldwide improving our
ability to measure the compositional variability of minerals in natural geological
samples and reducing the operator bias as to where to locate single spot analyses. This
chapter aims to provide an overview of the existing quantitative techniques for the
evaluation of the rock and mineral compositions and to present various examples of
applications. A new advanced method for compositional map standardization that
relies on internal standards and accurately corrects the X-ray intensities for continuum
background is also presented. This technique has been implemented into the computer
software XMAPTOOLS. The improved workflow defines the appropriate practice of
accurate standardization and provides data-reporting standards to help improving the
petrological interpretations.

Supplementary material: is available at https://doi.org/xxxx’.

Introduction
Quantitative compositional mapping by electron probe micro-analyser
(EPMA) is increasingly being applied in Earth’s sciences, as the burgeoning literature
reported in this chapter attests. This technique has proved to be very useful and
powerful to image small scale compositional heterogeneities in geological materials,
especially in mineral phases. The characterization of the compositional variability in
minerals of a given sample significantly benefits mineralogical and petrological
studies and allows the processes controlling the formation and transformation of
magmatic and metamorphic rocks to be investigated.
EPMA instrumental and software solutions have significantly evolved over the
past half century (de Chambost 2011), breaking down the barriers to routine effective
measurements of reliable X-ray maps for minerals at a resolution of a micrometre.
Any X-ray map of elemental distribution is semi-quantitative in essence because it is
built by collecting characteristic X-rays emitted by the elements of the specimen
excited under a finely focused electron beam. X-ray maps are raw data and they are
corrected neither for background nor for matrix effects such as the mean atomic
number, absorption and fluorescence (ZAF) effects. Similar to single spot analyses, a
calibration stage called ‘analytical standardization’ is thus also required for X-ray
maps to derive fully quantitative data of element concentrations.
Modern EPMA instruments are equipped with beam current stabilization
systems and mapping is classically done over a time period ranging from few hours to
few days. In this chapter, we will mostly deal with cases involving quantitative
compositional maps measured by EPMA most of them using wavelength-dispersive
spectrometers (WDS). Applications with other instruments such as scanning electron
microscope (SEM) equipped with energy-dispersive spectrometers (EDS) are very
popular in the mining industry (Gottlieb et al. 2000) and can also been used in a
quantitative way (Seddio 2015; Ortolano et al. submitted). However, despite technical
improvements, the EDS still suffers from a much lower analytical precision than
WDS. Phase composition maps (PCMs) can also be generated from high-resolution
backscatter electron images (Willis et al. 2017).
This chapter has a two-fold goal. The first is to present a summary of the
technical and computational advances made over the past half-century that provided
the foundation of the more recent developments. This review is partly based on case
studies where quantitative compositional mapping has provided considerable benefits
in investigation of specific rock-forming processes. Secondly, we introduce an
advanced standardization function that implement a correction for the X-ray
continuum (background) attributable to the X-ray bremsstrahlung (literally “braking
radiations”) effect. This standardization procedure is implemented into an improved
workflow for the computer software XMAPTOOLS (Lanari et al. 2014c available via
http://www.xmaptools.com ).

Quantitative compositional mapping: an historical perspective

The point-by-point investigation of a surface by analysis of the characteristic
X-ray emission was initiated by Castaing (1952) who built the first practical micro-
beam instrument called electron probe micro-analyser. Only a few years later, this
technique was expanded to acquire two-dimensional qualitative images of element
distribution using the ‘flying spot X-ray’ method (Cosslett & Duncumb 1956). This
analogue operational mode became rapidly popular in the community because it
enabled the detection of element distributions at a microscopic scale by analysing the
areal density of flashes generated by X-rays of a specified energy range on a cathode
X-ray tube. Nevertheless, this method suffered from several limitations (e.g. Newbury
et al. 1990b) and has remained in most of the cases purely qualitative. To our
knowledge, only a few studies managed to derive fully quantitative contour maps
using the flying spot X-ray method (Heinrich 1962; Jansen & Slaughter 1982). In the
most recent study by Jansen & Slaughter (1982), this achievement was possible only
after several correction steps and a standardization based on the ZAF correction. This
correction included simplifications such as composite calibration factors and
smoothing of the X-ray intensities to eliminate the variations due to statistical errors,
small-scale sample tomography and defocusing (Mayr & Angeli 1985; Marinenko et
al. 1989).
 An alternative strategy applied by Tracy et al. (1976) consisted of measuring
several high-resolution spot analyses within individually zoned grains and then
extracting contour maps of chemical composition. For the first time, garnet
compositional zoning – already recognized along one-dimensional transects (Atherton
& Edmunds 1966; Hollister 1966) – appeared as two-dimensional quantitative images.
More important than the quantification of element distribution patterns in two
dimensions was the resulting interpretation that the equilibrium conditions (here
pressure and temperature) changed during garnet growth. The first quantitative maps
of trace element concentrations by EPMA (Spear & Kohn 1996; Cossio et al. 2002)
and LA-ICP-MS (Raimondo et al. 2017) were also acquired using garnet and then
used to quantify (Kohn & Spear 2000) and model (Lanari et al. 2017) the amount of
garnet resorption during cooling and exhumation. Clearly, the contouring technique of
Tracy et al. (1976) was a major step forward in our understanding of garnet zoning
systematics by EPMA. However it has remained time consuming and limited to very
low spatial resolution.
Instrumental and computational improvements in the early 90’s have paved the
way to the direct acquisition of digital intensity maps in which a semi-quantitative
microanalysis is carried out at every beam location in a digitally controlled scan
producing numerical X-ray matrices (Marinenko et al. 1989; Newbury et al. 1990a, b;
Launeau et al. 1994). Traditionally the maps were acquired in electric beam-tracking
mode with a moving beam on a stationary sample. Focusing issues were successfully
solved during the analysis or by applying post-processing corrections (Newbury et al.
1990b and references therein). The development of a high-precision computer-
controlled x-y-z stage with an absolute accuracy <1 µm over the sample position was
a precondition for the acquisition of maps in a mechanical beam tracking mode. The
stage was set for the next major steps forward in the quantitative compositional
mapping of mineral phases and assemblages. However, the petrologic community did
not immediately grasp the fundamental advantages of this technique due to the initial
lack of adapted computer tools. A new generation of instruments combined with the
development of new software solutions on personal computers clearly favoured the
emergence of robust quantitative mapping techniques at the dawn of the 21st century
(Kohn & Spear 2000; Clarke et al. 2001; De Andrade et al. 2006).
The fundamental contribution of computer hardware and software
developments in this endeavour deserves to be pointed out. Any X-ray map measured
at a specific X-ray emission line typically contains > 100,000 pixels (up to several
millions). For each element, the X-ray map is a matrix made of X and Y points
forming a grid of pixels. The matrices, between eight and twenty depending on the
number of elements, are organized along a third dimension. A project file thus
contains several millions of measurements and this extensive geochemical dataset
need to be carefully evaluated and investigated along several successive steps of data
processing (see below). In the past, the multi-channel compositional classification
could take a few hours for instance in the 90’s (Launeau et al. 1994) whereas it is now
generally performed within a few seconds or minutes, using the automated clustering
approach implemented in the software XMAPTOOLS (Lanari et al. 2014c). The
appearance of such user-friendly software solutions for classification and
standardization has increased the popularity of this technique (Cossio & Borghi 1998;
Goncalves et al. 2005; Tinkham & Ghent 2005; Lanari et al. 2014c; Ortolano et al.
2014; Yasumoto et al. submitted) and fostered the development of thermodynamic
models based on local bulk compositions (see Lanari & Engi 2017 for a review).
Precision and accuracy of quantitative compositional mapping by
EPMA
An example of quantitative mapping by EPMA with a typical analytical setup
is presented in the following to show some advantages of this approach compared to
single-spot analyses. A map of 600 × 520 pixels was generated at the Institute of
Geological Sciences of the University of Bern using a JEOL 8200 superprobe
instrument, based on two passes (i.e. two map acquisitions). Accelerating voltage was
fixed at 15 keV, specimen current at 100 nA, the beam and step sizes at 1µm and the
dwell time at 160 ms. This setup enabled to measure nine elements by WDS and
seven elements by EDS. The total measurement time for mapping was approximately
30 h corresponding to 15 h per scan (Fig. 1). Note that shorter dwell time may be used
to save measurement time but it affects the analytical precision proportionally (see
below). In this example, the total measurement time can be reduced to 15 h using
dwell time of 80 ms and 7.5 h using dwell time of 40 ms.
The investigated sample is a garnet-bearing metasediment from Val Malone
(Southern Sesia Zone, Italian Western Alps) recording blueschist to eclogite facies
conditions during the Alpine orogeny (Pognante 1989). To track possible drift in the
X-ray production during the time window of the map acquisition, the average and
standard deviation (2σ) of the aluminium Kα X-ray counts of every single line scan
(corresponding to a single column in the image and an acquisition time of 90 seconds)
in the map were plotted against the measurement time (Fig. 1d). Aluminium is
expected to be constant in the alpine garnet in the absence of ferric iron (labelled Grt2
in Fig. 1a). It is interesting to note that no significant time-dependent intensity drift
was observed during this first pass and thus, the corresponding X-ray maps can be
directly transformed into maps of element concentrations using one of the
standardization procedures available in the literature (e.g. Cossio & Borghi 1998;
Clarke et al. 2001; De Andrade et al. 2006). Acquisition times longer than 30 h may
involve a time-dependent drift in the X-ray production caused by small variations of
the beam energy in the specimen. In this case, the resulting drift needs to be corrected
prior to standardization by flattening the intensity of a phase having a homogeneous
composition. Software tools for intensity drift corrections are presented below.
For the current example, the X-ray maps were standardized using spot
analyses as internal standards (Fig. 1c) and the software XMAPTOOLS 2.4.1 (Lanari et
al. 2014c). The calibrated map contains 312,000 pixels. After the exclusion of the
‘mixed’ pixels occurring at the grain boundaries, ~290,000 spatially resolved fully
quantitative analyses were obtained. The average analytical precision for a single
garnet pixel composition ranges between 2.8 % for Si and 20 % for Mn (2σ, see Tab.
1). One of the main advantages of the quantitative mapping approach compared to
traditional single-spot analyses is that the composition of several pixels of
homogenous material can be averaged, significantly increasing the analytical
precision and the possibility to detect slight compositional zoning. An averaging over
a 10 × 10 µm2 square window as the one plotted in Figure 1a reduces for example the
mean analytical uncertainty of garnet composition to 0.28 % for Si and 2 % for Mn.
Statistic tools are therefore needed to integrate pixel information along several spatial
and compositional dimensions to identify and quantify small amounts of
compositional zonation in the analysed sample (Cossio & Borghi 1998; Tinkham &
Ghent 2005; Lanari et al. 2014c). The example presented above shows that
quantitative X-ray mapping represents an extremely precise analytical technique,
opening new prospects for accurate studies of various geological materials.
Fig. 1. Mapping example of a HP mylonite from the Southern Sesia Zone (Western
Alps, Italy) and stability test of the mapping conditions in homogeneous material
during analysis. (a) Semi-quantitative Al Kα X-ray map. White square (bottom left
part of the garnet): example of integration 10 × 10 pixels window (see Tab. 1 and text
for details). The Al content in garnet core (Grt1) is lower due to the presence of Fe3+
substituting with Al. (b) Mask image of the mapped sample. White: pixels used for
the stability test. The phase boundaries were removed using the BRC correction of
XMAPTOOLS to avoid mixed pixels and secondary fluorescence effects. Ion probe
spot analyses (SIMS spots in panel a) were manually removed as the topography of
the crater affected the measured X-ray intensity because of variable absorption
thickness. (c) Calibration curve (dashed line) for garnet Grt2 using the advanced
standardization procedure described in this study. The error bars represent the
uncertainty on individual pixel composition determined from the counting statistics
(at 2σ) using a Poisson law. (d) Test for time-dependent drift in Al Kα X-ray intensity
of Grt2 with time. Al is assumed to be constant in Grt2 in the absence of significant
variations of XFe3+ (< 0.01 %). The red spots represent the mean value of each pixel
column (measured in a time range of ~ 90 seconds) and the vertical bars represent the
relative standard deviation (at 2σ). No time-related drift in the measured intensity is
observed over 15 hours of acquisition for this map.
Quantitative compositional mapping as a tool to track compositional
changes and investigate rock-forming processes
In the past fifteen years, the quantitative compositional mapping technique has
been intensively applied to metamorphic and magmatic rocks in order to investigate
specific petrological processes. In the following sections we provide a short review of
a few key studies that emphasize the advantages of this technique.

Modal abundances and local bulk compositions

The quantitative mapping technique can be greatly helpful to determine modal
abundances of mineral phases and thus the trace element distribution among minerals
(Martin et al. 2013). Several multi-channel compositional classification algorithms are
available in the literature (Wilson & MacRae 2005) and they perform generally well,
although the selected algorithm can significantly affect the predicted phase
proportions (e.g. Maloy & Treiman 2007) and the local bulk composition estimates
(Carpenter et al. 2014). For mineralogical and industrial applications, the chemical
classification can also be associated with automated phase determination (Pownceby
et al. 2007). In most cases, modal abundances are retrieved from semi-quantitative
maps only, acquired either by EPMA or SEM, with relatively short dwell time (< 10
ms) that reduce the measurement time to a few hours while investigating larger areas.
The standardization of X-ray maps is however a significant step forward in the
effectiveness of these methods and is recommended as the knowledge of the absolute
concentrations of each pixel offers several additional advantages (see below).
Samples with complex textures and heterogeneities can be investigated locally,
provided that quantitative maps are available (Lanari et al. 2013; Carpenter et al.
2014; Riel & Lanari 2015). Ebel et al. (2016), for instance, analysed both the size and
the distribution of chondrules and refractory inclusions in chondrites. Because the
maps were calibrated, they were able to extract the major element compositions of
those objects. In another study, quantitative maps were used to determine the bulk
composition of a single basaltic clast (with a size of 1 × 1.2 mm2) in a lunar impact-
melt breccia (Mészaros et al. 2016). Local bulk composition derived from
compositional maps have also been used for mass balance computations in
pseudomorphic reactions (Centrella et al. 2015; Tedeschi et al. 2017) and
thermodynamic modelling of local equilibria (Marmo et al. 2002; Abu-Alam et al.
2014; Riel & Lanari 2015; Lanari & Engi 2017; Lanari et al. 2017; Tedeschi et al.
2017; Centrella et al. 2018). As noted by Lanari & Engi (2017), accurate
determination of local bulk compositions requires a density correction in order to
convert sampled area into weight fraction of a phase. Otherwise, significant errors are
introduced in the final local bulk composition especially for elements sequestrated in
dense minerals. The example of Mészaros et al. (2016) is used in the following to
show how significant this error can be. Uncorrected and corrected maps of FeO are
shown in Figure 2 as well as the relative difference in the local bulk rock
compositions estimated with several methods. Large discrepancies are observed if the
local bulk composition is approximated using single spot analyses and modal
abundances obtained from semi-quantitative maps (red curves in Fig. 2c). The
discrepancy is smaller if quantitative maps are used since the compositional
variability of the mineral phases such as pyroxene can be assessed. However, a
density correction is required to obtain an accurate estimate of the local bulk
composition (Lanari & Engi 2017). If the correction is not applied differences up to 5-
10 % are observed for Al2O3, FeO and MgO (Fig. 2). The weight-per-pixel of FeO
and MgO in pyroxene is higher in the density corrected map because the density
contribution of pyroxene to the average density of the domain is high (~110 %). The
weight-per-pixel of Al2O3 in plagioclase is lower in the density corrected map because
plagioclase has a lower relative density (~85 %).

Fig. 2. Local bulk composition of a single basaltic clast in a lunar impact-melt breccia
(modified from Mészaros et al. 2016) and comparison between density uncorrected
and density corrected maps. (a) Density corrected map (DCM) of FeO (in wt%); (b)
density uncorrected map of FeO. Mineral abbreviations are from Whitney & Evans
(2010), v is vesicle. (c) Relative differences (in %) of the local bulk compositions
obtained with several methods to a reference composition obtained from the density
corrected map using XMAPTOOLS. Density values of 2.33, 3.46 and 2.27 g.cm-3 were
used for cristobalite, pyroxene and plagioclase respectively. Method 1: manual – the
compositions were obtained from the modal abundances determined using semi-
quantitative maps and the average composition of spot analyses (shown in red, each
curve representing a different set of spot analyses). Method 2: uncorrected map – the
composition is obtained from the density uncorrected map in XMAPTOOLS. The
dashed lines show the domain that was used to generate the local bulk compositions.
The same domain was used to determine the modal abundances for the manual
method.
 Fig. 3. Pressure map and microstructural control on the degree of muscovite re-
equilibration. (a) Pressure map of sample to13-4 from Airaghi et al. (2017a)
calculated using the calibration of Dubacq et al. (2010) at 525 °C. (b) Mineral sketch
of the mapped area showing the microstructural domains associated to cleavage 1 (S1)
and 2 (S2). (c) Pressure vs Si (apfu) diagram. (d) Pressure vs Na (apfu) diagram. The
three groups are based on the equilibrium conditions obtained in (a). The group 1
contains all the muscovite pixels with pressure ranging between 15 and 12 kbar, the
group 2a between 12 and 9 kbar and the group 2b between 9 and 4 kbar. (e) Mineral
sketch of the mapped area showing the fraction of each group in the two
microstructural domains S1 and S2. (f) Map showing the spatial distribution of the
pixels corresponding to each group.

Thermobarometry from partially re-equilibrated minerals

Partial re-equilibration in metamorphic rocks allows mineral relicts to be
preserved through one or sometimes more than one metamorphic cycles (Manzotti &
Ballèvre 2013). These relics are important archives for petrologists as they reflect
changes in equilibrium conditions and can be used to retrieve individual P-T stages or
segments of the P-T paths. Single spot analyses by EPMA are traditionally used to
calculate activities of end-members in solid solution and thus to model mineral
reactions and mineral stability (see Spear et al. 2016 for a review). Garnet is one of
the most famous thermobarometers and numerous calibrations have been derived by
the community to evaluate the P-T conditions from garnet compositions either
through the equilibrium with another Fe-Mg phase or by using isochemical phase
diagrams (Baxter et al. 2017). However, one of the key aspects of garnet
thermobarometry is an understanding of the processes that control the local mineral
composition during growth (Kohn 2014). Compositional maps may help to clarify
whether the observed compositional zoning derives from continuous or discontinuous
reactions involving equilibrium or transport controlled growth (Spear & Daniel 2001;
Yang & Rivers 2001; Hirsch et al. 2003; Angiboust et al. 2014; Ague & Axler 2016;
Lanari & Engi 2017). Empirical and semi-empirical thermometers and barometers are
also available for a large spectrum of magmatic and metamorphic mineral phases (for
variable bulk rock compositions) and can be easily applied to derive temperature or
pressure maps (De Andrade et al. 2006; Lanari et al. 2014c; Lanari et al. 2014a;
Trincal et al. 2015). The identification of the compositional variability, not only
through the mineral assemblage, but also within a single grain has proven to be a great
assistance for thermobarometric studies (Marmo et al. 2002; Lanari et al. 2013; Abu-
Alam et al. 2014; Loury et al. 2016). More challenging are the high-variance
assemblages involving phyllosilicates forming at lower metamorphic conditions.
Originally, their investigation has required a multi-equilibrium approach that relies on
complex solid solution models (Vidal & Parra 2000; Vidal et al. 2001; Parra et al.
2002; Parra et al. 2005; Vidal et al. 2006; Dubacq et al. 2010). These techniques had
significant successes when applied to compositional maps as the pressure and
temperature conditions of formation can be put into a micro textural context (Vidal et
al. 2006; Ganne et al. 2012; Lanari et al. 2014b; Trincal et al. 2015; Scheffer et al.
2016). It becomes therefore possible to apply multi-equilibrium thermobarometry to
specific mineral phases that are observed in textural equilibrium. The assimilation of
the textural equilibrium to the thermodynamic equilibrium can lead to
misinterpretation of the textural and compositional relationship (see below).
The investigation of the re-equilibration processes in local mineral assemblage
require a forward thermodynamic model (Lanari & Engi 2017). Yet compositional
maps have shown that the phyllosilicates in the mineral matrix of metapelite
preferentially re-equilibrate in zones of high strain, while they are preserved in zones
of low strain such as microfold hinges (Abd Elmola et al. 2017; Airaghi et al. 2017b;
Airaghi et al. 2017a). The compositional variability observed within the different
microstructural positions and the quantification of the modal abundance of each
compositional group have been used to demonstrate that the phyllosilicates partially
re-equilibrated during prograde metamorphism through pseudomorphic replacement.
This process is mostly controlled by the presence of a metamorphic fluid in the
intergranular medium. In a detailed study of muscovite composition in amphibolite
facies metapelite, Airaghi et al. (2017a) retrieved the P-T conditions of different
muscovite compositional groups observed in different microstructural domains. The
example of Airaghi et al. (2017a) is used in the following to show how detailed
investigation can be carried out using the quantitative mapping approach. A pressure
map (Fig. 3a) has been calculated from the compositional maps of muscovite at a
fixed temperature of 525°C using the method of Dubacq et al. (2010) and the program
CHLPHGEQUI 1.5 (Lanari 2012). The pressure condition of each pixel is plotted
against P- and T-dependent element such as Si (Fig. 3b) and Na (Fig. 3c) in atom per
formulae unit (apfu). Muscovite was classified into three groups: the relics of the HP
stage (group 1 in Fig. 3), the phengite re-equilibrated between the pressure peak and
the temperature peak (group 2a in Fig. 3), and the muscovite re-equilibrated at the
temperature peak (group 2b in Fig. 3). The groups 2a and 2b correspond to the
compositional group msB of Airaghi et al. (2017a). Here the distinction is based on
the equilibrium conditions rather than on compositional criteria. The main advantage
of using compositional maps and chemical diagrams is the ability to depict the spatial
distribution of any group of pixels within a given range of composition or equilibrium
conditions (Lanari et al. 2014c). The spatial distribution of the three groups of
muscovite is shown in Figure 3f. Their relative distribution in both S1 and S2
cleavages can be quantified (Fig. 3e). These results are in line with those of Airaghi et
al. (2017a) showing that the metamorphic conditions retrieved for the muscovite
groups in different micro-structural positions do not reflect the P-T conditions of the
microstructure-forming stages. Rather they document successive fluid-assisted re-
equilibration events. Matrix minerals can continue to partially re-equilibrate during
prograde metamorphism once the deformation ceased. Similar replacement textures
have been also documented in other phases such as chlorite (Lanari et al. 2012),
biotite (Airaghi et al. 2017a) and garnet (Martin et al. 2011; Lanari et al. 2017) using
quantitative compositional mapping.

Petrochronology
The technique of quantitative compositional mapping is extremely helpful in
petrochronological studies (see Engi et al. 2017) to link metamorphic stages and
reactions to ages retrieved from major and accessory minerals (Williams et al. 2017).
In favourable cases, an age map can be constructed and reveals the continuous spatial
distribution of ages (Goncalves et al. 2005). Otherwise, quantitative compositional
mapping allows the most appropriate spots for dating to be identified. For example,
quantitative compositional mapping of allanite may reveal the existence of cores and
rims of different composition that, if large enough, could be dated separately (Burn
2016; Engi 2017). The mapping of the compositional heterogeneities in white mica
can highlight the most homogeneous grains for the 40Ar/39Ar dating. In addition, it
provides a strong petrological basis for interpreting the variability of in-situ Ar
40
Ar/39Ar dates (Airaghi et al. 2017b). The spatial resolution is often lower than the
characteristic size of the compositional variations in mica (Cossette et al. 2015;
Berger et al. 2017; Laurent 2017).
To conclude this short overview we can state that quantitative mapping is
therefore becoming an essential tool to provide a comprehensive image of the
petrological, thermobarometric and geochronological evolution of metamorphic rocks.

Standardization techniques: advantages and pitfalls

The extraction of elemental concentration values from X-ray intensities
requires matrix and other corrections. The X-ray data need to be standardized post-
collection either relative to similar data collected on natural and synthetic standards
using an empirical correction scheme (Jansen & Slaughter 1982; Newbury et al.
1990a, b; Cossio & Borghi 1998; Clarke et al. 2001; Chouinard & Donovan 2015) or
against internal standards assuming no matrix effects (De Andrade et al. 2006). Those
techniques are separately described in the following.
Castaing’s first approximation to quantitative analysis and matrix effects
As first noted by Castaing (1952), the primary generated X-ray intensities are
proportional to the respective weight concentration of the corresponding elements in
the specimen. In absence of significant matrix differences between unknown and
standard materials, this assumption can be expressed as:
𝐶!!"# 𝐼!!"#
≈ !"# (1)
𝐶!!"# 𝐼!
where the terms 𝐶!!"# and 𝐶!!"# are the composition expressed in weight concentration
of the element i in the unknown and the standard and the terms 𝐼!!"# and 𝐼!!"#
represent the net corresponding X-ray intensities corrected for continuum background
(see below) and any possible other issues such as peak overlap or drift. In the case of
significant physical and/or chemical differences between the unknown and the
standard (i.e. matrix effects), the Equation (1) becomes:
𝐶!!"# 𝐼!!"#
= 𝑘! !"# (2)
𝐶!!"# 𝐼!
where k represents a correction coefficient expressing the non-linear matrix effects. In
the specialized literature these effects are divided into atomic number (Zi), X-ray
absorption (Ai) and X-ray fluorescence (Fi) effects. The correction must be applied
separately for each element present in both the analysed specimen and in a given
standard reference material.

Bence-Albee empirical correction

The empirical procedure of Bence & Albee (1968) is based upon known
binary experimental data and it involves less computation time than the ZAF
correction described in the following. It assumes a simple hyperbolic calibration curve
between the weight concentrations and the net intensities of a given oxide in binary
system (Ziebold & Ogilvie 1964). The calibration curve is described in terms of a
single conversion parameter known as the α-factor
1 − 𝐼! 1 − 𝐶!
= 𝛼!" (3)
𝐼! 𝐶!
with 𝛼!" the α-factor for a binary between elements A and B; IA and Ca the net
intensity and mass concentration, respectively. This approach has been generalized by
Bence & Albee (1968) to more complicated oxide systems of n components using a
linear combination of α-factors such that:
𝐶! 𝑘! 𝛼!! + 𝑘! 𝛼!! + ⋯ + 𝑘! 𝛼!!
= (4)
𝐼! 𝑘! + 𝑘! + ⋯ + 𝑘!
where 𝛼!! is the α-factor for the n1 binary used to determine the concentration of
element n in a binary between n and 1.
Clarke et al. (2001) applied this procedure to standardize X-ray maps into map
of oxide mass concentrations. Several tests have indeed shown that the Bence &
Albee (1968) procedure yields results comparable to those obtained with the ZAF
method (Goldstein et al. 2003) while reducing significantly the computation time for
corrections (Clarke et al. 2001). The accuracy of this method mostly depends on the
quality of the α-factor estimates, as well as the choice of homogeneous and well-
characterized standard materials. Most of the correction factors for oxides and
silicates are well constrained and updates including small improvements are regularly
published (Albee & Ray 1970; Love & Scott 1978; Armstrong 1984; Kato 2005). The
software XRMAPANAL (Tinkham & Ghent 2005) uses the Bence-Albee algorithm to
standardize X-ray maps and provides several tools to display maps, compositional
graphs and to perform advanced statistical analyses.

ZAF and ϕ(ρz) corrections

The ZAF matrix correction was the first generalized algebraic procedure. This
standardization method is based on a more rigorous physical model taking into
account the atomic number effects, the absorption and fluorescence in the specimen.
The ratio of concentration in unknown and standard of an element i is given by
𝐶!!"# 𝐼!!"#
= 𝑍𝐴𝐹 ! !"# (5)
𝐶!!"# 𝐼!
where 𝑍𝐴𝐹 ! is the ZAF correction coefficient. Typical values of the ZAF coefficient
for metals are reported in several text books (e.g. Goldstein et al. 2003; Reed 2005).
The direct calculation of absorption in the original ZAF correction scheme
was lately improved by introducing an empirical expression of ϕ(ρz) to correct for
absorption (Packwood & Brown 1981). In the mid-80’s several ϕ(ρz) algorithms (e.g.
Riveros et al. 1992) including more accurate sets of mass absorption coefficients were
successively developed: PROZA (Bastin et al. 1986), citiZAF (Armstrong 1988), PAP
(Pouchou & Pichoir 1991), XPhi (Merlet 1994). Some of them are still used in
modern EPMA instruments. It is crucial for the ZAF or ϕ(ρz) corrections to measure
all the major and minor elements present in the specimen to ensure that all the
possible matrix effects are taken into account.
Both ZAF and ϕ(ρz) correction algorithms have been applied to X-ray maps in
order to generate maps of oxide mass concentrations (Jansen & Slaughter 1982;
Cossio & Borghi 1998; Prêt et al. 2010) and this option is currently available in the
software PETROMAP (Cossio & Borghi 1998) and in CAMECA’s commercial
software provided with the SX100. There are two main limits of this technique. First,
it is necessary to perform an accurate background correction to the X-ray maps prior
to standardization. This correction requires either the acquisition of upper and lower
peak background maps or the use of models to predict the theoretical background
values (e.g. Tinkham & Ghent 2005). The MAN algorithm for instance allows the
background value to be predicted from the mean atomic number of the specimen
(Donovan & Tingle 2003; Chouinard & Donovan 2015; Wark & Donovan 2017),
significantly reducing the total acquisition time. The second limitation is the large
relative uncertainty in the intensity of each pixel (see Tab. 1). This uncertainty is
propagated through the ZAF factors. To overcome this issue, Jansen & Slaughter
(1982) applied a preliminary ZAF correction to a group of pixels of the same mineral
phase in order to derive an estimate of the ZAF correction factors yielding for this
phase. These factors are then used to correct compositionally similar pixels. This
option is not available to our knowledge in any commercial software or freeware
solution.

Internal standardization using high-precision spot analyses

Of primary importance in routine EPMA analyses, are the analytical precision
and accuracy of spot analyses. The quality of standardization in spot analysis can be
quickly evaluated using either a reference material with known concentration, or
stoichiometric constraints on unknown mineral phases. These tests are routinely
applied before each analytical session and ensure the high quality of the data
produced. The internal standardization procedure of X-ray maps takes advantage of
the excellent quality of spot analyses. The goal is to calibrate the X-ray maps of every
mineral phase using high precision spot analyses of the same mineral phase (Mayr &
Angeli 1985; De Andrade et al. 2006). In this case, there are no matrix effects
between the unknowns (X-ray maps) and the standards (spot analyses):
!!!"# !"# ∗
𝐶!!"# = 𝐼 + 𝐼!!"#$ (6)
!!!"# !
∗
with 𝐼!!"# the X-ray intensity of the unknown uncorrected for background and 𝐼!!"#$
the intensity of background. This approach results in a strong dependence of the
accuracy of the compositional maps upon the accuracy of the spot analyses selected
for the standardization. The internal standardization procedure has been implemented
in a MATLAB©-based computer program (De Andrade et al. 2006), in the software
solutions XMAPTOOLS (Lanari et al. 2014c) and QNTMAP (Yasumoto et al.
submitted).

Spatial and chemical resolution and possible issues

Spatial resolution
In compositional mapping, the spatial resolution is determined by the spacing
of spot measurements and the X-ray excitation volume of the electron beam. It is
recommended to use a beam size smaller than the pixel size to reduce overlapping
(see XMAPTOOLS’ user guide for examples). For a given spacing between two pixels,
an increasing beam size will increase the fraction of ‘mixed’ pixels (i.e. mixed phases
see below).

Chemical resolution
The chemical resolution of an individual pixel depends on the dwell time, the
accelerating voltage and the beam current. The relative precision of any measurement
can be estimated using counting statistics (the generation of characteristic X-rays is a
Poisson process) from the total number of X-rays collected by the detector. Examples
of analytical precisions for different elements are given in Table 1.
As shown in the introduction, as soon as several pixels are taken into account,
the relative uncertainty and detection limits decrease relative to the values obtained by
single pixel counting statistics. Averaging of pixels of unzoned material or of a single
growth zone virtually increases the counting time thus reducing the relative
uncertainty (Tab. 1). This effect applies to any map, as the human eye is an
outstanding integrator of visual information as the human brain is able to detect
gradients even in noisy signals. If a mineral phase extends over a substantial lateral
range of pixels, it might be possible to discern compositional zoning below the
detection limit despite the noise that results from statistical fluctuations in the count at
a single pixel (Newbury et al. 1990a; Friel & Lyman 2006).

Issues
Radiation damage. Some elements, and particularly light elements, are
affected by degradation induced by local heating effects from electron beam exposure
over time. Either an increase or decrease in intensity could be observed because of
radiation damage. A decrease in intensity may reflect a loss of mass by volatilization.
The low-energy X-rays of volatile elements undergo a strong self-absorption in the
specimen that increases if the beam energy is increased (Goldstein et al. 2003).
Therefore, light elements (Na. K, Ca) have to be measured with the lowest beam
energy possible, during the first pass of the beam over the mapped area
Fig. 4. Examples of apparent compositional zoning in quartz caused by secondary
fluorescence effects. The scale bars show number of recorded X-ray counts. Mineral
abbreviations are from Whitney & Evans (2010). Mixed pixels have been removed
using the BRC correction (see text). The phases shown in white do not play any role
for secondary fluorescence effects. (a) Map of a micaschist sample from The
Briançonnais Zone (Chaberton Area) in the Western Alps (Verly 2014) showing the
secondary fluorescence of Fe in quartz at the contact with chlorite (light grey) but not
with phengite (dark grey). White: plagioclase and rutile. (b) Map of a mafic eclogite
from the Stak massif in NW Himalaya (Lanari et al. 2013; Lanari et al. 2014c). An
apparent zoning in Ti is observed in clinopyroxene (area 1) at the contact with rutile,
caused by secondary fluorescence of Ti, while a ‘real’ compositional zoning of Ti is
present in clinopyroxene and correlated with the variations in other major element
concentrations (area 2). White: garnet, amphibole, plagioclase and Fe-oxide. (c) Map
of a schist belonging to the TGU (Theodul Gletscher Unit) in the Zermatt area.
Secondary fluorescence of Ca observed in quartz at the contact with garnet (dark
grey) and plagioclase (light grey). White: rutile, titanite, apatite, phengite, paragonite,
pyrite, zoisite and epidote. (d) Secondary fluorescence of Fe in quartz at the contact
with garnet (light grey) and chlorite (dark grey) in another schist from the TGU.
White: paragonite, phengite, albite, pyrite, chlorite, zoisite, rutile, titanite and apatite.

Secondary fluorescence effects. A secondary fluoresce effect occurs when

the characteristic X-rays of a given element generate a secondary generation of
characteristic X-ray of another element. Because X-rays penetrate into matter farther
than electrons, the interacting volume of X-ray-induced fluorescence is generally
greater (up to 99% larger as proposed by Goldstein et al. 2003). This volume may
include more than one phase, and the induced X-rays originate from different
contributions. In X-ray images, secondary fluorescence effects can occur near phase
boundaries (e.g. Fig. 4) or melt inclusions (Chouinard et al. 2014). Compositional
mapping is a powerful tool to detect potential effects of secondary fluorescence that
would not be seen otherwise. Figure 4 shows a few examples involving an apparent
compositional zoning that is caused by secondary fluorescence effects. Secondary
fluorescence can affect several elements and is commonly associated with the
presence of anorthite (Ca), chlorite (Fe), epidote (Fe), garnet (Ca) or rutile (Ti). One
of the best candidates in which to observe secondary fluorescence effects is quartz
(Fig. 4a,c-d). The X-ray intensity caused by secondary fluorescence in a surrounding
mineral decreases from the source toward the inner part with a distance up to 100 µm
at 15 keV, 100 nA and for counting time < 300 ms (Fig. 4). Particular attention must
be paid when comparing differences of minor to trace elements concentrations in
phases occurring both as inclusions in a porphyroblasts and within the mineral matrix.
In the example shown in Figure 4d, quartz exhibits higher count rates of Ca in the
grains trapped as inclusions in garnet compared to those in the matrix. Note that a
slight secondary effect is also observed in the pressure shadows of garnet where
chlorite is present (light grey in Fig. 4d). If a mineral with low Ca, Fe and Ti
concentrations displays higher concentrations in a porphyroblast showing high Ca, Fe
and Ti contents, one might suspect a secondary fluorescence effect. In quartz
inclusions FeO can be overestimated by 0.25 to 0.65 wt% (Fig. 4d).
Thermal instability of the instrument. Several parts of the EPMA
instrument are sensitive to temperature changes. The gas commonly used as ionization
medium to produce a secondary cascade of electrons amplifying the signal in the
proportional counter, must have a constant pressure over time in order to ensure the
reproducibility in counting rates between the standards and the unknown. The
spectrometer crystals are also affected by thermal expansion, yielding to a change in
the Bragg angle and a consequent shift in the peak position (Jenkins & De Vries 1982;
Delgado-Aparicio et al. 2013).
Beam current stability. For any kind of quantitative analyses, the quality of
the data depends also strongly on beam current stability. The adjustment of the
balance between beam accelerating voltage and beam current allows the analytical
setup between spatial resolution and sensitivity to be optimized. Higher beam energies
results in higher count rates, but with a significant loss of spatial resolution due to the
greater interaction volume (e.g. Hombourger & Outrequin 2013 for field emission
EPMA). A variation in the beam energy during mapping causes an apparent variation
of the number of characteristic X-rays produced in the specimen. This issue is visible
if the element intensity shows a linear evolution with time of in a homogeneous
mineral phase.
Spectrometer focus. In a fully focused WD spectrometer, the electron beam,
the specimen, the analysing crystal and the detector are all located on constrained
geometry to satisfy the Bragg condition for the desired wavelength. This geometry
brings all the X-rays originating from the source to be focused at the same point on
the detector, and maximizes the collected radiation (Goldstein et al. 2003). During
WDS mapping with fixed stage and scanning beam, the beam may be scanned off the
point of optimum focus and the X-ray intensity decreases as a function of the
deflection (Newbury et al. 1990b). An example showing how spectrometer defocusing
during mapping can affect the X-ray counting rate is presented below in the section
Corrections
Surface irregularities. The presence of topography at the specimen surface
affects the measurements during X-ray mapping. The beam converges on the plan of
optimum focus, but it broadens above and below it. When the specimen surface is not
flat, different beam volumes may analyse different specimen portions. These features
will not appear in focus and will cause a decrease of intensity as discussed above. The
local topography affects the angle of incidence between the beam and the specimen
surface, and hence the number and trajectories of measured BSE and SE of the
backscattered electrons depends on the local orientation. The deviation from the ideal
angle results in a backscattered electron distribution distorted with the peak and so a
decrease in intensity. A complex surface also affects the local thickness of sample
along the X-ray path toward the detector. This effect is known as the geometric
absorption effect (Goldstein et al. 2003).

Quantitative mapping procedure

The procedure for obtaining compositional maps of elemental concentrations of a
micrometre- to millimetre-sized domain involves several steps:
1. Define the problem: what do we want to learn from this map? As for
traditional spot analyses, this decision will define the analysis procedures (see
Goldstein et al. 2003 for a complete description). For compositional mapping,
the analytical procedure used to investigate the compositional zoning of a
millimetre-sized porphyroblast (high chemical resolution but low ability to
identify small phases) is for example different from the one that will be used
to estimate a local bulk composition (higher ability to identify small phases
but lower chemical resolution).
2. Sample characterization using the optical microscope or SEM – including
BSE and EDS qualitative analyses – to determine the mineral phases to be
investigated. It is critical to know prior each mapping session the size of the
smallest object to be analysed. The spatial resolution of the map,
corresponding to the step size, must be at least five to ten times smaller to
identify possible compositional zoning within single grains. It is important to
distinguish between ‘pure’ pixels recorded away from grain boundaries and
‘mixed’ pixels recording mixing information (Launeau et al. 1994). Only pure
pixels can be used directly to measure the compositional variability of a given
mineral phase. It is important to minimize as much as possible the presence of
mixed pixels in the maps by using small beam size (< 1µm for the map
acquisition), as mixed pixels are the main source of misclassification (Maloy
& Treiman 2007). Small spot sizes also reduce the need of density correction
if several phases are measured in a single pixel (Ebel et al. 2016).
3. Select the sample and the appropriate area to be investigated. The surface of
the mapped area together with the step-size and the number of elements
measured by WDS determines the total measurement time to be added to the
time for the acquisition of point analyses necessary for the map
standardization.
4. Measure the spot analyses that will be used as internal standards for X-ray
map standardization. This step requires the calibration of the spot analyses
using either the ZAF or ϕ(ρz) corrections and appropriate standard materials.
We recommend measurement of at least one horizontal and one vertical
transect starting and ending at grain boundaries that can be easily recognized
 on the final X-ray maps. This will permit to detect and correct potential shifts
in the position of the spot analyses relative to the map (see below). The
distance between two spot analyses along a given transect must be a factor of
the pixel size (step-size) to avoid additional uncertainties introduced during
the projection. Additional spot analyses can be defined using BSE images. We
recommend measuring between 10 and 20 spot analyses per phases to be
standardized. This is particularly critical for mineral phases exhibiting
compositional zoning. Phases with known composition can be standardized
using a fixed composition (e.g. 100 wt% of SiO2 for quartz).
5. Check the quality of the spot analyses. As discussed above, the accuracy of the
standardized maps depends mostly on the accuracy of the spot analyses to be
used as internal standards. It is thus critical to evaluate the quality of the spot
analyses using the concentrations expressed in oxide weight percentage and
the structural formulas.
6. Collect the X-ray maps using WDS or/and EDS and the TOPO, SEI and/or
BSE images.
7. Export the X-ray maps as matrixes containing raw X-ray counts. It is not
recommended to use pseudo-coloured images exported by the instrument
software.

An advanced standardization procedure implemented in

XMAPTOOLS’ workflow

A description of the different steps of data processing is given in the following

sections.

Multi-channel compositional classification

The multi-channel compositional classification process allows the pixels of X-
ray maps to be categorized according to chemical information contained in the
element X-ray intensities. Several algorithms are available in the literature (Wilson &
MacRae 2005) and implemented in the available software solutions (Cossio & Borghi
1998; Clarke et al. 2001; Tinkham & Ghent 2005; Ortolano et al. 2014; Yasumoto et
al. submitted). In XMAPTOOLS the automated classification procedure is based upon
the k-means algorithm and requires the manual sampling of a representing pixel for
each previously identified phase (Lanari et al. 2014c). The function generates a mask
file containing the distribution of each mineral phase defined by the user. A mask file
can also be generated manually by selecting different groups of pixels out of binary or
a ternary chemical plots or by combining mask files from different classifications.
Once a mask file is generated, it may be merged with another mask file, or the group
may be further split by manual selection on chemical plots.

Corrections
If needed, corrections to the raw X-ray maps can be applied right after the
multi-channel compositional classification. Corrections include dead time correction
for WDS, intensity drift or topography corrections as well as the adjustment of the
map and/or standard positions.
Dead time correction (WDS). During EMPA analyses performed with WDS,
each pulse arriving from the detector is being processed by the electronics. In the
meanwhile, the detector remains completely blind to any incoming X-ray photon for a
time interval defined as ‘dead time’. A dead time correction has therefore to be
applied to the effective analytical counting time. In XMAPTOOLS, the dead time
correction is applied to X-ray maps measured by WDS using the following
relationships:
!!
𝐼!∗ = (7)
!!! × !"!! × !!

where 𝐼! and 𝐼!∗ are the measured and dead time corrected count rates of element i (in
counts per second) and 𝜏 the characteristic dead time of the detector in seconds.
Sample surface topography. Sample topography such as holes or irregular
surfaces may have a significant effect on the produced X-ray intensities. The local
topography introduces variable absorption path-length in different directions, so that
the intensity emitted varies according to the spectrometer position (Newbury et al.
1990a). A TOPO map can be measured in JEOL EPMA instruments using the Solid
State BSE detector. This detector consists of two opposite segments: one located at
the top of the image field and one located at the bottom. The topographic image is
constructed from the difference between the BSE signals returned by the two
detectors. The topography image corresponds to a light illumination at oblique
incidence and suppresses the atomic number contrast component of the BSE signal
(Kässens & Reimer 1996). In the XMAPTOOLS software, the topography correction
can be applied using the TRC module (TRC is for TOPO-related correction) when a
correlation is observed between the X-ray intensities of an element in a given phase
and the intensity of the TOPO map. The magnitude of the correction depends on the
spectrometer used, the element and the mineral phase analysed as the absorption
changes with the position of the spectrometer and the density of the target material.
Time-dependent drift. A time-dependent drift in the X-ray production related
to small variations of the beam current at the specimen surface may be observed for
acquisition periods exceeding 24 h (Fig. 5). There are three major causes for time-
dependent drift of the X-ray intensities to occur. First, the probe current can drift
during the analysis (e.g. Cossio & Borghi 1998). Second, the vacuum conditions in
the gun or the specimen chamber can change with time, causing more (or less)
interactions between the electron beam and the gas particles thus decreasing (or
increasing) the specimen energy and the production rate of characteristic X-rays. For
instance an abrupt increase of the pressure in the gun chamber may cause a sharp
decrease in the measured X-ray intensities. The third cause of time-dependent drift
may be related to beam-defocusing issues during scanning on a non-flat surface.
Defocusing indeed affects the geometry and size of the interaction volume thus
affecting the specimen energy density where the characteristic X-rays are produced.
The resulting drift might be corrected prior to standardization regardless of the cause
of the variation as soon as the time-dependency can be retrieved. For the correction to
be performed, a phase equally distributed within the mapped area and showing a
homogeneous composition in a given element must be identified. The Intensity Drift
Correction (IDC) tool has been implemented in XMAPTOOLS 2.4.1 for this purpose. It
enables detection of time-dependent drifts and applies any correction function defined
by the user. An example of time-dependent drift observed in a map acquired over ~90
h is presented in Figure 5. The investigated sample is the same as the one described in
the introduction (garnet-bearing metasediment from the Southern Sesia Zone in the
Western Alps). The Si Kα X-ray map (measured during the first pass) and Al Kα X-
ray map (measured during the second pass) of alpine garnet Grt2 were used to track
possible time-dependent drift and to define a correction function (Fig. 5). In this
example, the observed time-dependent drift exhibits a constant slope over the whole
acquisition time corresponding to an average drift in the X-ray intensities of ~0.02 %
per hour (~1 % for each scan of ~45 h). The similarity of the drift in the two scans
suggests that it reflects a progressive defocusing of the beam caused by a slight
inclination of the sample. This example shows that the beam stabilization system of
the JEOL 8200 superprobe used in this study performs well compared to the probe
current drift of 0.36 % per hour reported in Cossio & Borghi (1998). However, other
maps have shown significant time-dependent drift caused by either vacuum failure in
the gun chamber or electron beam current drift. In this case it is vitally important to
detect such a drift and to apply the appropriate correction to the X-ray maps prior
standardization.

Fig. 5. Example of time-dependent drift for a total measurement time of 90 hours on

the sample described in Figure 1. (a, b) Evolution of Si Kα X-ray intensity and Al Kα
X-ray intensities in Grt2 with time. Red spots: mean values of each pixel column
(corresponding to time interval of ~ 160 seconds). Vertical bars: standard deviation
(at 2σ). The observed time-dependent drift is fitted using linear function (blue dashed
line). It is interesting to note that the slope of this function remains constant during the
whole acquisition time (two passes). (c) Mask image of the mapped area showing in
white the pixels used for the stability test. The phase boundaries were removed using
the BRC correction. (d, e) Matrixes of correction factors (in %) used to correct the X-
ray maps of Si Kα and Al Kα respectively. In both cases, the drift correction is lower
than 1%.
Fig. 6. Positions of spot analyses (internal standard) for the map shown in Figure 5
are tested and corrected using the SPC module of XMAPTOOLS. (a) Case 1: X-ray
intensity signal of the Al Kα map pixels corresponding to the original position of the
spot analyses obtained from the EPMA coordinates. (b) Al2O3 mass concentration of
the spot analyses (internal standards). This trend is used as a reference to evaluate the
quality of the projection. The red arrows mark the pixels showing a poor match with
(a). (c) X-ray intensity signal of the Al Kα map pixels corresponding to the corrected
position of the spot analyses. (d) Synthetic map of the quality of the projection for
Case 1. The best position of the spot analyses (internal standard) on the map is
calculated as described in the text. In this example, the correlation coefficients were
calculated for a 21 × 21 pixels window centred on the original position (white star).
The original position is shifted from the optimal position and may be corrected by
moving the internal standard position on the map. (e) Synthetic map of the quality of
the projection for Case 2. After the correction, the position (black star) corresponds
with the best match. It is interesting to notice that even a shift of 2 µm can
significantly affect the quality of the match and thus of the standardization.

Mixed pixels and BRC correction. Mixed pixels are commonly observed at
the boundary between two phases, depending on the map resolution and beam size
(see XMAPTOOLS’ user guide for examples). Resulting localized features do not
represent authentic mineral zoning. Mixed pixels can be filtered using the BRC
correction (for border-removing correction) available in XMAPTOOLS. This correction
removes the pixels located between the different masks of the selected mask file. The
correction may be applied for different sizes of the mixing zone depending on the map
resolution. An alternative strategy is to evaluate the proportion of phases in each
mixed pixel using a distribution-based cluster analysis (Yasumoto et al. submitted).
Fig. 7. Advanced standardization technique based on internal standards. (a) The
calibration curve for a given element in a mineral phase is defined using the ranges in
oxide mass concentration (ΔC) and in the X-ray intensity (ΔI). The intercept values
define the background correction to be applied to the phase of interest. (b) Internal
standardization of homogeneous phase (case 1) or of an element below the detection
limit in mapping conditions (case 2). (c) Evolution of the difference between
standardizations with and without background correction with the oxide mass
concentration. (d) Matrix effects in a single mineral (here garnet) causing changes in
the slope of the calibration curve.

Maps and standards position. Before performing the analytical

standardization, the position of the maps and the spot analyses must be tested and, if
necessary, modified based on statistical criteria. It is critical to guaranty that the
positions of the spot analyses used as internal standard are not shifted, i.e. that the
analysed volume is the same in both analyses. In order to detect such a shift, the
standard data (spot analyses, in wt%) can be compared with the intensity data (in
counts) of the corresponding pixels on the map, as shown in Figure 6. An algorithm
that detects the optimal position of the maps and the spot analyses is available in
XMAPTOOLS. A map of the correlation coefficients between the standard and the
intensity data for different X and Y positions is calculated for each element. All maps
are then combined to produce a synthetic map evaluating the overall quality of the
projection. Examples of good and bad position of standards are shown in Figure 6. In
the first case (Case 1 in Fig. 6a,b), the comparison reveals a shift in the position of the
spot analyses, corrected using the Standard Position Correction (SPC) tool by shifting
vertically the positions of the spot analyses by of 2 µm. In the second case, the quality
of the projection was recalculated after correction (Fig. 6d,e). The match between
standard values (in wt%, Fig. 6c) and the pixel intensity (in counts, Fig. 4D) is greatly
improved. The projection in Case 2 plots in the optimal quality field of the resulting
synthetic map. Considering that a shift of few pixels can affect the quality of the
match (i.e. in the example in Fig. 4, a shift of 2 µm decreases the quality of the
projection by ~ 20%), the standard position correction is crucial to obtain a reliable
standardization

Fig. 8. Example of standardization

of garnet for Al, Mn and Mg. (a-
c) Compositional diagrams
showing the compositional ranges
observed in the garnet pixels of
the X-ray maps (histograms), the
internal standards (white dots)
with 2σ uncertainty and the
calibration curves. The continuous
lines are the calibration curves
assuming no background, the
dashed lines are the calibration
curves of the advanced
standardization method involving
a pseudo-background correction.
(d-g) Calibrated maps of MnO
and MgO without background
correction (d, f) and with
background correction (e, g). The
colour scales is identical for both
images of the same element. (h-i)
Difference maps in weight
percentage.
Advanced procedure for internal standardization
In the XMAPTOOLS software, the analytical standardization is performed using
high-precision spot analyses as internal standards (Lanari et al. 2014c) to obtain
numerical concentrations. To be accurate, the X-ray maps need to be corrected for
background (see Equation 6) prior to standardization. The background correction is
usually applied by using background values of the spot analyses (De Andrade et al.
2006). However it is not applicable if the map and spot analyses are acquired with
different spectrometer configurations. For high intensity-to-background ratios, where
!!
> 20, the background correction is not applied. The spot analyses are indeed
!!∗ !!!
already corrected for background and the map background effects on the calibration
curve are negligible. On the contrary, for elements with low intensity-to-background
ratios, the background significantly affects the slope of the calibration curve and a
background correction is required (see below). The acquisition of lower and upper
background maps would triple the measurement time. Hence, an advanced
standardization and correction strategy has been implemented to overcome this
problem and is described in the following. This calibration does not require any
background measurements and is thus extremely powerful at low count rates where
the background correction is critical and not always accurately predicted by MAN
models (see Fig. 1 in Wark & Donovan 2017).
The advanced standardization technique implemented in XMAPTOOLS 2.4.1
uses the variability commonly observed in the mass concentrations of the standards
(ΔC in Fig. 7a) to fit the slope of the calibration curve and to approximate the
corresponding background (Equation 6, see the star in Fig. 7a). For this
approximation to be accurate, the spot analyses need to capture the majority of the
compositional variability of the (zoned) minerals. This pseudo-background correction
is not applied to homogeneous phases where ΔC is small (case 1 in Fig. 7b). For trace
elements, the compositional variability can be only captured by the spot analyses
(case 2 in Fig. 7d), showing that the measured element is below the detection limit for
the mapping conditions (Lanari et al. 2014c). As previously mentioned, the difference
between a calibration using a background correction and a calibration assuming a
fixed background value of zero decreases with increasing intensity-to-background
ratios (Fig 7c). It is also important to notice that this advanced technique can only be
applied in absence of significant matrix effects in the mineral phase. Matrix effects
generally occur if a strong compositional zoning is observed between the core and the
rim of a dense mineral such as garnet. In this case, it may be necessary to apply
several distinct standardizations one for each garnet composition (Fig. 7d). The matrix
differences affect the slopes of the calibration curves (e.g. Lanari et al. 2014c) by
underestimating the background value (see the continuous line in Fig. 7d).
An example of garnet standardization is given in Figure 8. The investigated
sample is a garnet, kyanite and staurolite bearing metasediment from the Central Alps
(Todd & Engi 1997). A millimetre size garnet crystal has been compositionally
mapped at the Institute of Geological Sciences of the University of Bern using a JEOL
8200 superprobe instrument. Accelerating voltage was fixed at 15 keV, specimen
current at 100 nA, the beam and step sizes at 6 µm and the dwell time at 100 ms. The
compositions of garnet core (MnO > 1 wt%) and rims (MnO < 0.7 wt%) are reported
in Table 2. The standardization of aluminium does not require any background
correction (Fig. 8a), as the intensity-to-background ratio is typically higher than 60 for
almandine-rich garnet. The intensity-to-background ratios are much smaller for Mn
and Mg (~3 for both cases) and thus a background correction is required. For both
elements the background values have been approximated using the technique
presented above (see results in Figs. 8b,c). In this example the spot analyses captured
a large range of the observed compositional variability in Mn and Mg. The absence of
background correction (continuous lines in Figs. 8b,c) significantly affect the
standardize maps with relative differences up to 17 % in MnO and 27 % in MgO (Figs.
8d-g).

Fig. 9. Examples of standardization of garnet and phengite for elements where

background corrections are required. The blue rectangles show the reference
background values extracted from the background map (see text). The red spots show
outliers automatically rejected during the advanced standardization. (a) Mg in garnet.
(b) Mn in garnet. (c) Mg in phengite. (d) Na in phengite. (e) Ti in phengite. The
symbols are defined in the figure caption of Figure 6.

To evaluate the reliability of this advanced standardization technique and

especially the validity of the background predictions, a phengite, chlorite garnet-
bearing metasediment from the Zermatt-Saas Zone has been compositionally mapped
at the Institute of Geological Sciences of the University of Bern using a JEOL 8200
superprobe instrument with six separate acquisitions of the same map area, two for the
peak measurements and four for the off-peak background measurements. Background
maps have been calculated assuming a linear background distribution between the
lower and upper background values. The measured background values have been
compared with the predicted ones for garnet (Mg, Mn, see Fig. 9a,b) and phengite
(Mg, Na, Ti see Fig. 9c-e). The predicted and measured background values are in line
within 2σ uncertainty for all the elements above detection limits for the mapping
conditions and with low intensity-to-background ratios (see Fig. 10). The measured
background values for Mn, Mg, Ca and Ti were also compared with the X-ray
intensities measured in quartz. This test shows that the background maps were
correctly measured off-peak and that the background values reflect only the
contribution of the X-ray bremsstrahlung. To conclude, the advanced standardization
technique provided in XMAPTOOLS can successful correct the X-ray maps for
background during the standardization.

Local bulk compositions, structural formulas and P-T maps

The standardized maps can be either merged to generate mass concentration
images and extract local bulk compositions (Lanari & Engi 2017) or treated separately
to compute map of elemental distributions in number of atoms per formula units or
maps of P-T conditions (De Andrade et al. 2006; Vidal et al. 2006; Lanari et al.
2014c).

Fig. 10. Comparison of the measured and predicted backgrounds for various elements
and mineral phases. The predicted background values for quartz were obtained from
the mean intensities in large quartz grain avoiding secondary fluorescence effects.
Standards for data reporting

A sufficient amount of information must be included in publications (1) to

enable the independent replication of the analytical measurements and (2) to
demonstrate the validity of the proposed interpretations, including the uncertainty
estimates (Potts 2012). As already proposed by Horstwood et al. (2016) for LA-ICP-
MS U-(Th-)Pb data, comprehensive details about both data acquisition and data
processing of EPMA quantitative maps must be provided. They include:
• a description of the applied corrections (dead time, topography, drift) and the
used calibration method along with the analytical details for the acquisition of
both map and spot analyses. In supplementary material S1, we provide a table
template illustrating the recommended minimum analytical information types
to be submitted for publication.
• the calibrated oxide maps (in .txt file format) included in the electronic
appendix. An example is provided in supplementary material S2.
• a file listing the calibration parameters (calibration method, standards and
pixel compositions, background, slope of the calibration curve) for each phase
along with the calibrated oxide maps. This file can contain for example the
standardization reports generated by XMAPTOOLS during the calibration of
each map. An example is provided in supplementary material S3.
• a table with the average compositions of each compositional group including a
relative uncertainty on the structural formula and end-member proportions
Lanari et al. (2017). A typical example of garnet core and rim compositions
including the analytical uncertainties estimated from a Monte-Carlo simulation
is shown in Table 2. Such values can be obtained by using the export function
available in the ‘Quanti’ workspace of XMAPTOOLS.

Conclusions and future directions

This chapter shows a few examples demonstrating that quantitative
compositional mapping by EPMA is a powerful tool for many mineralogical and
petrological studies, but it requires the application of a rigorous data reduction scheme
to obtain accurate mineral compositions.
The advanced standardization function implemented in the software
XMAPTOOLS allows a pseudo-background correction for minor and trace elements
without the acquisition of background maps or estimation of background values using
MAN models. A background correction is required for an accurate calibration of
minor and trace elements; similar pseudo-background corrections are applied in other
programs using the internal standardization technique (Ortolano et al. submitted;
Yasumoto et al. submitted).
In the future, quantitative mapping may experience a significant jump in both
spatial and analytical resolutions, though the development of spectrometers with
higher resolution and lower detection limits, the acquisition of high-spatial-resolution
maps using field emission electron guns or with quantitative X-ray tomography to
characterize the elemental distribution in three dimensions. Beyond the technical
improvements, EPMA quantitative mapping has a great potential to be combined with
other mapping techniques such as maps of crystallographic orientation (Centrella et al.
2018) or trace elements distribution in minerals obtained by LA-ICP-MS mapping
(Paul et al. 2012; Paul et al. 2014; Raimondo et al. 2017).
Acknowledgments

The authors would like to sincerely thank the hidden contributions of several
colleagues that provided stimulating ideas and criticism leading to the recent
improvements of XMAPTOOLS: Amaury Pourteau, Tom Raimondo, Chloé Loury,
Marco Burn, Céline Martin, Nicolas Riel, Christophe Scheffer, Aude Verly, Joerg
Hermann, Martin Engi and Daniela Rubatto. Positive and constructive reviews by
Mike Williams and Gaetano Ortolano are also acknowledged, as is Philippe
Goncalves for efficient editorial handling.

References

Abd Elmola, A., Charpentier, D., Buatier, M., Lanari, P. & Monié, P. 2017. Textural-chemical changes
and deformation conditions registered by phyllosilicates in a fault zone (Pic de Port Vieux thrust,
Pyrenees). Applied Clay Science, 144, 88-103, http://doi.org/10.1016/j.clay.2017.05.008.

Abu-Alam, T.S., Hassan, M., Stüwe, K., Meyer, S.E. & Passchier, C.W. 2014. Multistage Tectonism
and Metamorphism During Gondwana Collision: Baladiyah Complex, Saudi Arabia. Journal of
Petrology, 55, 1941-1964, http://doi.org/10.1093/petrology/egu046.

Ague, J.J. & Axler, J.A. 2016. Interface coupled dissolution-reprecipitation in garnet from subducted
granulites and ultrahigh-pressure rocks revealed by phosphorous, sodium, and titanium zonation.
American Mineralogist, 101, 1696-1699, http://doi.org/10.2138/am-2016-5707.

Airaghi, L., Lanari, P., de Sigoyer, J. & Guillot, S. 2017a. Microstructural vs compositional
preservation and pseudomorphic replacement of muscovite in deformed metapelites from the Longmen
Shan (Sichuan, China). Lithos, 282-283, 262-280.

Airaghi, L., de Sigoyer, J., Lanari, P., Vidal, O., Sautter, B., Guillot, S., Xu, X. & Tan, X. 2017b. Total
vertical offset of the Beichuan fault (Longmen Shan, Sichuan, China) deduced from metamorphic
minerals. Journal of Asian Earth Sciences, in press.

Albee, A.L. & Ray, L. 1970. Correction factors for electron probe microanalysis of silicates, oxides,
carbonates, phosphates, and sulfates. Analytical Chemistry, 42, 1408-1414,
http://doi.org/10.1021/ac60294a030.

Angiboust, S., Pettke, T., De Hoog, J.C.M., Caron, B. & Oncken, O. 2014. Channelized Fluid Flow and
Eclogite-facies Metasomatism along the Subduction Shear Zone. Journal of Petrology, 55, 883-916,
http://doi.org/10.1093/petrology/egu010.

Armstrong, J.T. 1984. Quantitative analysis of silicate and oxide minerals: a reevaluation of ZAF
corrections and proposal for new Bence-Albee coeficients. In: Romig, A.D. & Goldstein, J.I. (eds)
Microbeam Analysis. San Francisco Press, San Francisco, 208-212.

Armstrong, J.T. 1988. Quantitative analysis of silicate and oxide minerals: comparison of Monte Carlo,
ZAF and Phi-Rho-Z procedures. In: Newbury, D.E. (ed) Microbeam analysis. San Francisco Press, San
Francisco, 239-246.

Atherton, M.P. & Edmunds, W.M. 1966. An electron microprobe study of some zoned garnets from
metamorphic rocks. Earth and Planetary Science Letters, 1, 185-193, http://doi.org/10.1016/0012-
821X(66)90066-5.
Bastin, G.F., Heijligers, H.J.M. & van Loo, F.J.J. 1986. A Further Improvement in the Gaussian ϕ(ϱ)
Approach for Matrix Correction in Quantitative Electron Probe Microanalysis. Scanning, 8, 45-67,
http://doi.org/10.1002/sca.4950080204.

Baxter, E.F., Caddick, M.J. & Dragovic, B. 2017. Garnet: A Rock-Forming Mineral Petrochronometer.
Reviews in Mineralogy and Geochemistry, 83, 469-533, http://doi.org/10.2138/rmg.2017.83.15.

Bence, A.E. & Albee, A.L. 1968. Empirical Correction Factors for the Electron Microanalysis of
Silicates and Oxides. The Journal of Geology, 76, 382-403.

Berger, A., Wehrens, P., Lanari, P., Zwingmann, H. & Herweg, M. 2017. Microstructures, mineral
chemistry and geochronology of white micas along a retrograde evolution: An example from the Aar
massif (Central Alps, Switzerland). Tectonophysics, 721, 179-195.

Burn, M. 2016. LA-ICP-QMS Th-U/Pb allanite dating: methods and applications. PhD thesis,
University of Bern.

Carpenter, P.K., Jolliff, B.L. & Donovan, J.J. 2014. Compositional Mapping by EPMA and µXRF.
Microscopy and Microanalysis, 20, 710-711, http://doi.org/10.1017/S1431927614005273.

Castaing, R. 1952. Application des sondes électroniques à une méthode d'analyse ponctuelle chimique
et cristallographique. University of Paris.

Centrella, S., Austrheim, H. & Putnis, A. 2015. Coupled mass transfer through a fluid phase and
volume preservation during the hydration of granulite: An example from the Bergen Arcs, Norway.
Lithos, 236–237, 245-255, http://doi.org/10.1016/j.lithos.2015.09.010.

Centrella, S., Putnis, A., Lanari, P. & Austrheim, H. 2018. Textural and chemical evolution of
pyroxene during hydration and deformation: a consequence of retrograde metamorphism. Lithos,
http://doi.org/10.1016/j.lithos.2017.11.002.

Chouinard, J. & Donovan, J. 2015. Quantitative Elemental Mapping with Electron Microprobe and
Automated Data Analysis. Microscopy and Microanalysis, 21, 2193-2194,
http://doi.org/10.1017/S1431927615011745.

Chouinard, J., Donovan, J., Aster, E. & Wallace, P. 2014. Quantitative Mapping of and Secondary
Fluorescence Effects in Olivine Hosted Melt Inclusions. Microscopy and Microanalysis, 20, 750-751,
http://doi.org/10.1017/S1431927614005479.

Clarke, G.L., Daczko, N.R. & Nockolds, C. 2001. A method for applying matrix corrections to X-ray
intensity maps using the Bence–Albee algorithm and Matlab. Journal of Metamorphic Geology, 19,
635-644, http://doi.org/10.1046/j.0263-4929.2001.00336.x.

Cossette, É., Schneider, D.A., Warren, C.J. & Grasemann, B. 2015. Lithological, rheological, and fluid
infiltration control on 40Ar/39Ar ages in polydeformed rocks from the West Cycladic detachment
system, Greece. Lithosphere, 7, 189-205, http://doi.org/10.1130/L416.1.

Cossio, R. & Borghi, A. 1998. PETROMAP: MS-DOS software package for quantitative processing of
X-ray maps of zoned minerals. Computers & Geosciences, 24, 805-814, http://doi.org/10.1016/S0098-
3004(98)00057-0.

Cossio, R., Borghi, A. & Ruffini, R. 2002. Quantitative Modal Determination of Geological Samples
Based on X-ray Multielemental Map Acquisition. Microscopy and Microanalysis, 8, 139-149,
http://doi.org/10.1017/S1431927601020062.

Cosslett, V.E. & Duncumb, P. 1956. Micro-analysis by a Flying-Spot X-Ray Method. Nature, 177,
1172-1173.
De Andrade, V., Vidal, O., Lewin, E., O'Brien, P. & Agard, P. 2006. Quantification of electron
microprobe compositional maps of rock thin sections: an optimized method and examples. Journal of
Metamorphic Geology, 24, 655-668, http://doi.org/10.1111/j.1525-1314.2006.00660.x.

de Chambost, E. 2011. A History of Cameca (1954–2009). Advances in Imaging and Electron Physics,
167, 1-119, http://doi.org/http://dx.doi.org/10.1016/B978-0-12-385985-3.00001-8.

Delgado-Aparicio, L., Bitter, M., Podpaly, Y., Rice, J., Burke, W., Rio, M.S.d., Beiersdorfer, P., Bell,
R., et al. 2013. Effects of thermal expansion of the crystal lattice on x-ray crystal spectrometers used
for fusion research. Plasma Physics and Controlled Fusion, 55, 125011.

Donovan, J.J. & Tingle, T.N. 2003. An Improved Mean Atomic Number Background Correction for
Quantitative Microanalysis. Microscopy and Microanalysis, 2, 1-7,
http://doi.org/10.1017/S1431927696210013.

Dubacq, B., Vidal, O. & De Andrade, V. 2010. Dehydration of dioctahedral aluminous phyllosilicates:
thermodynamic modelling and implications for thermobarometric estimates. Contributions to
Mineralogy and Petrology, 159, 159, http://doi.org/10.1007/s00410-009-0421-6.

Ebel, D.S., Brunner, C., Konrad, K., Leftwich, K., Erb, I., Lu, M., Rodriguez, H., Crapster-Pregont,
E.J., et al. 2016. Abundance, major element composition and size of components and matrix in CV, CO
and Acfer 094 chondrites. Geochimica et Cosmochimica Acta, 172, 322-356,
http://doi.org/10.1016/j.gca.2015.10.007.

Engi, M. 2017. Petrochronology Based on REE-Minerals: Monazite, Allanite, Xenotime, Apatite.

Reviews in Mineralogy and Geochemistry, 83, 365-418, http://doi.org/10.2138/rmg.2017.83.12.

Engi, M., Lanari, P. & Kohn, M.J. 2017. Significant Ages—An Introduction to Petrochronology.
Reviews in Mineralogy and Geochemistry, 83, 1-12.

Friel, J.J. & Lyman, C.E. 2006. Tutorial Review: X-ray Mapping in Electron-Beam Instruments.
Microscopy and Microanalysis, 12, 2-25, http://doi.org/10.1017/S1431927606060211.

Ganne, J., De Andrade, V., Weinberg, R.F., Vidal, O., Dubacq, B., Kagambega, N., Naba, S., Baratoux,
L., et al. 2012. Modern-style plate subduction preserved in the Palaeoproterozoic West African craton.
Nature Geosci, 5, 60-65, http://doi.org/http://www.nature.com/ngeo/journal/v5/n1/abs/ngeo1321.html -
supplementary-information.

Goldstein, J.I., Newbury, D.E., Echlin, P., Joy, D.C., Lyman, C.E., Lifshin, E., Sawyer, L. & Michael,
J.R. 2003. Scanning Electron Microscopy and X-ray Microanalysis. third edition ed. Springer, New
York.

Goncalves, P., Williams, M.L. & Jercinovic, M.J. 2005. Electron-microprobe age mapping of monazite.
American Mineralogist, 90, 578-585.

Gottlieb, P., Wilkie, G., Sutherland, D., Ho-Tun, E., Suthers, S., Perera, K., Jenkins, B., Spencer, S., et
al. 2000. Using quantitative electron microscopy for process mineralogy applications. JOM, 52, 24-25,
http://doi.org/10.1007/s11837-000-0126-9.

Heinrich, K.F.J. 1962. Concentration Mapping Device for the Scanning Electron Probe Microanalyzer.
Review of Scientific Instruments, 33, 884-884, http://doi.org/10.1063/1.1718013.

Hirsch, D.M., Prior, D.J. & Carlson, W.D. 2003. An overgrowth model to explain multiple, dispersed
high-Mn regions in the cores of garnet porphyroblasts. American Mineralogist, 88, 131-141,
http://doi.org/10.2138/am-2003-0116.

Hollister, L.S. 1966. Garnet zoning: an interpretation based on the Rayleigh fractionation model.
Science, 154, 1647-1651.
Hombourger, C. & Outrequin, M. 2013. Quantitative Analysis and High-Resolution X-ray Mapping
with a Field Emission Electron Microprobe. Microscopy Today, 21, 10-15,
http://doi.org/10.1017/S1551929513000515.

Horstwood, M.S.A., Košler, J., Gehrels, G., Jackson, S.E., McLean, N.M., Paton, C., Pearson, N.J.,
Sircombe, K., et al. 2016. Community-Derived Standards for LA-ICP-MS U-(Th-)Pb Geochronology –
Uncertainty Propagation, Age Interpretation and Data Reporting. Geostandards and Geoanalytical
Research, 40, 311-332, http://doi.org/10.1111/j.1751-908X.2016.00379.x.

Jansen, W. & Slaughter, M. 1982. Elemental mapping of minerals by electron microprobe. American
Mineralogist, 67, 521-533.

Jenkins, B. & De Vries, J.L. 1982. Practical X-Ray Spectrometry. Springer US.

Kässens, M. & Reimer, L. 1996. Contrast effects using a two-detector system in low-voltage scanning
electron microscopy. Journal of Microscopy, 181, 277-285, http://doi.org/10.1046/j.1365-
2818.1996.116396.x.

Kato, T. 2005. New Accurate Bence-Albee α-Factors for Oxides and Silicates Calculated from the PAP
Correction Procedure. Geostandards and Geoanalytical Research, 29, 83-94,
http://doi.org/10.1111/j.1751-908X.2005.tb00657.x.

Kohn, M.J. 2014. 4.7 - Geochemical Zoning in Metamorphic Minerals A2 - Holland, Heinrich D. In:
Turekian, K.K. (ed) Treatise on Geochemistry (Second Edition). Elsevier, Oxford, 249-280.

Kohn, M.J. & Spear, F.S. 2000. Retrograde net transfer reaction insurance for pressure-temperature
estimates. Geology, 28, 1127-1130.

Lanari, P. 2012. Micro-cartographie P-T-e dans les roches métamorphiques. Applications aux Alpes et
à l’Himalaya. Doctorate PhD, Grenoble.

Lanari, P. & Engi, M. 2017. Local bulk composition effects on mineral assemblages. Reviews in
Mineralogy and Geochemistry, 83, 55-102.

Lanari, P., Wagner, T. & Vidal, O. 2014a. A thermodynamic model for di-trioctahedral chlorite from
experimental and natural data in the system MgO–FeO–Al2O3–SiO2–H2O: applications to P–T sections
and geothermometry. Contr. Mineral. Petrol., 167-968, http://doi.org/10.1007/s00410-014-0968-8.

Lanari, P., Giuntoli, F., Loury, C., Burn, M. & Engi, M. 2017. An inverse modeling approach to obtain
P-T conditions of metamorphic stages involving garnet growth and resorption. European Journal of
Mineralogy, 29, 181-199.

Lanari, P., Guillot, S., Schwartz, S., Vidal, O., Tricart, P., Riel, N. & Beyssac, O. 2012. Diachronous
evolution of the alpine continental subduction wedge: evidence from P-T estimates in the Briançonnais
Zone houillere (France – Western Alps). Journal of Geodynamics, 56-57, 39-54.

Lanari, P., Riel, N., Guillot, S., Vidal, O., Schwartz, S., Pêcher, A. & Hattori, K.H. 2013. Deciphering
high-pressure metamorphism in collisional context using microprobe mapping methods: Application to
the Stak eclogitic massif (northwest Himalaya). Geology, 41, 111-114, http://doi.org/10.1130/g33523.1.

Lanari, P., Rolland, Y., Schwartz, S., Vidal, O., Guillot, S., Tricart, P. & Dumont, T. 2014b. P–T–t
estimation of deformation in low-grade quartz-feldspar-bearing rocks using thermodynamic modelling
and 40Ar/39Ar dating techniques: example of the Plan-de-Phasy shear zone unit (Briançonnais Zone,
Western Alps). Terra Nova, 26, 130-138, http://doi.org/10.1111/ter.12079.

Lanari, P., Vidal, O., De Andrade, V., Dubacq, B., Lewin, E., Grosch, E.G. & Schwartz, S. 2014c.
XMapTools: A MATLAB©-based program for electron microprobe X-ray image processing and
geothermobarometry. Comp. Geosci., 62, 227-240, http://doi.org/10.1016/j.cageo.2013.08.010.
Launeau, P., Cruden, A.R. & Bouchez, J.-L. 1994. Mineral recognition in digital images of rocks; a
new approach using multichannel classification. The Canadian Mineralogist, 32, 919-933.

Laurent, B. 2017. Localisation de la déformation au sein de zones de cisaillement haute-pression

basse-température et enregistrement isotopique 40Ar/39Ar. PhD PhD thesis, Université d'Orléans.

Loury, C., Rolland, Y., Cenki-Tok, B., Lanari, P. & Guillot, S. 2016. Late Paleozoic evolution of the
South Tien Shan: Insights from P–T estimates and allanite geochronology on retrogressed eclogites
(Chatkal range, Kyrgyzstan). Journal of Geodynamics, 96, 62-80,
http://doi.org/10.1016/j.jog.2015.06.005.

Love, G. & Scott, V.D. 1978. Evaluation of a new correction procedure for quantitative electron probe
microanalysis. Journal of Physics D: Applied Physics, 11, 1369.

Maloy, A.K. & Treiman, A.H. 2007. Evaluation of image classification routines for determining modal
mineralogy of rocks from X-ray maps. American Mineralogist, 92, 1781-1788,
http://doi.org/10.2138/am.2007.2477.

Manzotti, P. & Ballèvre, M. 2013. Multistage garnet in high-pressure metasediments: Alpine

overgrowths on Variscan detrital grains. Geology, 41, 1151-1154, http://doi.org/10.1130/g34741.1.

Marinenko, R.B., Myklebust, R.L., Bright, D.S. & Newbury, D.E. 1989. Defocus modelling correction
for wavelength dispersive digital compositional mapping with the electron microprobe. Journal of
Microscopy, 155, 183-198, http://doi.org/10.1111/j.1365-2818.1989.tb02881.x.

Marmo, B.A., Clarke, G.L. & Powell, R. 2002. Fractionation of bulk rock composition due to
porphyroblast growth: effects on eclogite facies mineral equilibria, Pam Peninsula, New Caledonia.
Journal of Metamorphic Geology, 20, 151-165, http://doi.org/10.1046/j.0263-4929.2001.00346.x.

Martin, C., Debaille, V., Lanari, P., Goderis, S., Vandendael, I., Vanhaecke, F., Vidal, O. & Claeys, P.
2013. REE and Hf distribution among mineral phases in the CV–CK clan: A way to explain present-
day Hf isotopic variations in chondrites. Geochimica et Cosmochimica Acta, 120, 496-513,
http://doi.org/10.1016/j.gca.2013.07.006.

Martin, L.A.J., Ballèvre, M., Boulvais, P., Halfpenny, A., Vanderhaeghe, O., Duchêne, S. & Deloule, E.
2011. Garnet re-equilibration by coupled dissolution–reprecipitation: evidence from textural, major
element and oxygen isotope zoning of 'cloudy' garnet. Jour Metam Geol, 29, 213-231,
http://doi.org/10.1111/j.1525-1314.2010.00912.x.

Mayr, M. & Angeli, J. 1985. Concentration mapping—a software package for the quantitative
determination of two-dimensional elemental distribution by an electron probe microanalyser. X-Ray
Spectrometry, 14, 89-98, http://doi.org/10.1002/xrs.1300140210.

Merlet, C. 1994. An accurate computer correction program for quantitative electron probe
microanalysis. Microchimica Acta, 114, 363-376, http://doi.org/10.1007/bf01244563.

Mészaros, M., Hofmann, B.A., Lanari, P., Korotev, R.L., Gnos, E., Greber, N.D., Greenwood, R.C.,
Jull, A.J.T., et al. 2016. Petrology and geochemistry of feldspathic impact-melt breccia Abar al’ Uj 012,
the first lunar meteorite from Saudi Arabia. Meteoritics & Planetary Science, 51, 1830-1848.

Newbury, D.E., Fiori, C.E., Marinenko, R.B., Myklebust, R.L., Swyt, C.R. & Bright, D.S. 1990a.
Compositional mapping with the electron probe microanalyzer: Part II. Analytical Chemistry, 62,
1245A-1254A, http://doi.org/10.1021/ac00223a001.

Newbury, D.E., Fiori, C.E., Marinenko, R.B., Myklebust, R.L., Swyt, C.R. & Bright, D.S. 1990b.
Compositional mapping with the electron probe microanalyzer: part I. Analytical Chemistry, 62,
1159A-1166A, http://doi.org/10.1021/ac00221a002.
Ortolano, G., Zappalà, L. & Mazzoleni, P. 2014. X-Ray Map Analyser: A new ArcGIS® based tool for
the quantitative statistical data handling of X-ray maps (Geo- and material-science applications).
Computers & Geosciences, 72, 49-64, http://doi.org/10.1016/j.cageo.2014.07.006.

Ortolano, G., Visalli, R., Godard, G. & Cirrincione, R. submitted. Quantitative X-Ray Map Analyser
(Q-XRMA): A new GIS-based statistical approach for Mineral Image Analysis. Computers &
Geosciences.

Packwood, R.H. & Brown, J.D. 1981. A Gaussian expression to describe ϕ(ρz) curves for quantitative
electron probe microanalysis. X-Ray Spectrometry, 10, 138-146, http://doi.org/10.1002/xrs.1300100311.

Parra, T., Vidal, O. & Agard, P. 2002. A thermodynamic model for Fe-Mg dioctahedral K white micas
using data from phase equilibrium experiments and natural pelitic assemblages. Contrib. Mineral.
Petrol., 143, 706-732.

Parra, T., Vidal, O. & Theye, T. 2005. Experimental data on the Tschermak substitution in Fe-chlorite.
American Mineralogist, 90, 359-370, http://doi.org/10.2138/am.2005.1556.

Paul, B., Woodhead, J.D., Paton, C., Hergt, J.M., Hellstrom, J. & Norris, C.A. 2014. Towards a
Method for Quantitative LA-ICP-MS Imaging of Multi-Phase Assemblages: Mineral Identification and
Analysis Correction Procedures. Geostandards and Geoanalytical Research, 38, 253-263,
http://doi.org/10.1111/j.1751-908X.2014.00270.x.

Paul, B., Paton, C., Norris, A., Woodhead, J., Hellstrom, J., Hergt, J. & Greig, A. 2012. CellSpace: A
module for creating spatially registered laser ablation images within the Iolite freeware environment.
Journal of Analytical Atomic Spectrometry, 27, 700-706, http://doi.org/10.1039/C2JA10383D.

Pognante, U. 1989. Lawsonite, blueschist and eclogite formation in the southern Sesia zone (Western
Alps, Italy). Eur. J. Mineral., 1, 89–104.

Potts, P.J. 2012. A Proposal for the Publication of Geochemical Data in the Scientific Literature.
Geostandards and Geoanalytical Research, 36, 225-230, http://doi.org/10.1111/j.1751-
908X.2011.00121.x.

Pouchou, J.-L. & Pichoir, F. 1991. Quantitative Analysis of Homogeneous or Stratified Microvolumes
Applying the Model “PAP”. In: Heinrich, K.F.J. & Newbury, D.E. (eds) Electron Probe Quantitation.
Springer US, Boston, MA, 31-75.

Pownceby, M.I., MacRae, C.M. & Wilson, N.C. 2007. Mineral characterisation by EPMA mapping.
Minerals Engineering, 20, 444-451, http://doi.org/10.1016/j.mineng.2006.10.014.

Prêt, D., Sammartino, S., Beaufort, D., Meunier, A., Fialin, M. & Michot, L.J. 2010. A new method for
quantitative petrography based on image processing of chemical element maps: Part I. Mineral
mapping applied to compacted bentonites. American Mineralogist, 95, 1379-1388,
http://doi.org/10.2138/am.2010.3431.

Raimondo, T., Payne, J., Wade, B., Lanari, P., Clark, C. & Hand, M. 2017. Trace element mapping by
LA-ICP-MS: assessing geochemical mobility in garnet. Contributions to Mineralogy and Petrology,
172, 17, http://doi.org/10.1007/s00410-017-1339-z.

Reed, S.J.B. 2005. Electron Microprobe Analysis and Scanning Electron Microscopy in Geology. 2nd
edition ed. Cambridge University Press, Cambridge.

Riel, N. & Lanari, P. 2015. Techniques, méthodes et outils pour la quantification du métamorphisme.
Géochroniques, 136, 53-60.

Riveros, J.A., Castellano, G.E. & Trincavelli, J.C. 1992. Comparison of Φ (ρz) Curve Models in
EPMA. In: Boekestein, A. & Pavićević, M.K. (eds) Electron Microbeam Analysis. Springer Vienna,
Vienna, 99-105.
Scheffer, C., Vanderhaeghe, O., Lanari, P., Tarantola, A., Ponthus, L., Photiades, A. & France, L. 2016.
Syn- to post-orogenic exhumation of metamorphic nappes: Structure and thermobarometry of the
western Attic-Cycladic metamorphic complex (Lavrion, Greece). Journal of Geodynamics, 96, 174-
193, http://doi.org/10.1016/j.jog.2015.08.005.

Seddio, S.M. 2015. Standards-Based Quantitative EDS Mapping. Microscopy and Microanalysis, 21,
2093-2094, http://doi.org/10.1017/S1431927615011241.

Spear, F.S. & Kohn, M.J. 1996. Trace element zoning in garnet as a monitor of crustal melting.
Geology, 24, 1099-1102.

Spear, F.S. & Daniel, C.G. 2001. Diffusion control of garnet growth, Harpswell Neck, Maine, USA.
Journal of Metamorphic Geology, 19, 179-195.

Spear, F.S., Pattison, D.R.M. & Cheney, J.T. 2016. The metamorphosis of metamorphic petrology.
Geological Society of America Special Papers, 523, http://doi.org/10.1130/2016.2523(02).

Tedeschi, M., Lanari, P., Rubatto, D., Pedrosa-Soares, A., Hermann, J., Dussin, I., Pinheiro, M.A.,
Bouvier, A.S., et al. 2017. Reconstruction of multiple P-T-t stages from retrogressed mafic rocks:
Subduction versus collision in the Southern Brasília orogen (SE Brazil). Lithos, 294-295, 283-303,
http://doi.org/10.1016/j.lithos.2017.09.025.

Tinkham, D.K. & Ghent, E.D. 2005. XRMapAnal: A program for analysis of quantitative X-ray maps.
American Mineralogist, 90, 737-744, http://doi.org/10.2138/am.2005.1483.

Todd, C.S. & Engi, M. 1997. Metamorphic field gradients in the Central Alps. Journal of Metamorphic
Geology, 15, 513-530.

Tracy, R.J., Robinson, P. & Thompson, A.B. 1976. Garnet composition and zoning in the
determination of temperature and pressure of metamorphism, central Massachusetts. American
Mineralogist, 61, 762-775.

Trincal, V., Lanari, P., Buatier, M., Lacroix, B., Charpentier, D., Labaume, P. & Muñoz, M. 2015.
Temperature micro-mapping in oscillatory-zoned chlorite: Application to study of a green-schist facies
fault zone in the Pyrenean Axial Zone (Spain). American Mineralogist, 100, 2468-2483,
http://doi.org/10.2138/am-2015-5217.

Verly, A. 2014. Évolution tectonique-métamorphique de la zone Briançonnaise à la latitude de

Briançon. Master 1 Master 1 thesis, University of Grenoble.

Vidal, O. & Parra, T. 2000. Exhumation paths of high-pressure metapelites obtained from local
equilibria for chlorite–phengite assemblages. Geological Journal, 35, 139-161,
http://doi.org/10.1002/gj.856.

Vidal, O., Parra, T. & Trotet, F. 2001. A Thermodynamic Model for Fe-Mg Aluminous Chlorite Using
Data from Phase Equilibrium Experiments and Natural Pelitic Assemblages in the 100° to 600°c, 1 to
25 kb Range. American Journal of Science, 301, 557-592, http://doi.org/10.2475/ajs.301.6.557.

Vidal, O., De Andrade, V., Lewin, E., Munoz, M., Parra, T. & Pascarelli, S. 2006. P–T-deformation-
Fe3+/Fe2+ mapping at the thin section scale and comparison with XANES mapping: application to a
garnet-bearing metapelite from the Sambagawa metamorphic belt (Japan). Journal of Metamorphic
Geology, 24, 669-683, http://doi.org/10.1111/j.1525-1314.2006.00661.x.

Wark, D. & Donovan, J. 2017. Improving EPMA Analyses Using Mean Atomic Number Backgrounds.
Microscopy and Microanalysis, 17, 580-581, http://doi.org/10.1017/S1431927611003771.

Whitney, D.L. & Evans, B.W. 2010. Abbreviations for names of rock-forming minerals. American
Mineralogist, 95, 185-187, http://doi.org/10.2138/am.2010.3371.
Williams, M.L., Jercinovic, M.J., Mahan, K.H. & Dumond, G. 2017. Electron Microprobe
Petrochronology. Reviews in Mineralogy and Geochemistry, 83, 153-182,
http://doi.org/10.2138/rmg.2017.83.5.

Willis, K.V., Srogi, L., Lutz, T., Monson, F.C. & Pollock, M. 2017. Phase Composition Maps integrate
mineral compositions with rock textures from the micro-meter to the thin section scale. Computers &
Geosciences, 109, 162-177, http://doi.org/10.1016/j.cageo.2017.08.009.

Wilson, N.C. & MacRae, C.M. 2005. An Automated Hybrid Clustering Technique Applied to Spectral
Data Sets. Microscopy and Microanalysis, 11, 434-435, http://doi.org/10.1017/S1431927605501168.

Yang, P. & Rivers, T. 2001. Chromium and manganese zoning in pelitic garnet and kyanite: Spiral,
overprint, and oscillatory (?) zoning patterns and the role of growth rate. Journal of Metamorphic
Geology, 19, 455-474, http://doi.org/10.1046/j.0263-4929.2001.00323.x.

Yasumoto, A., Yoshida, K., Kuwatani, T., Nakamura, D., Svojtka, M. & Hirajima, T. submitted. Fast
and precise quantitative electron probe chemical mapping technique and its application to ultrahigh-
pressure eclogite from the Moldanubian Zone of the Bohemian Massif (Nové Dvory, Czech Republic).
American Mineralogist.

Ziebold, T.O. & Ogilvie, R.E. 1964. An Empirical Method for Electron Microanalysis. Analytical
Chemistry, 36, 322-327, http://doi.org/10.1021/ac60208a024.
Table 1. Uncertainties on raw map data acquired by EPMA. The precision at 2σ-level
(in %) was estimated on the average intensity of the garnet external rim pixels. Oxide
weight % values are derived from quantitative spot analyses on the same garnet zone.

10×10 20×20
1 pixel
Oxide X-ray pixels pixels
Element uncertainty
wt% intensity uncertainty uncertainty
(% 2σ)
(% 2σ) (% 2σ)
Si 37.99 5040 2.8172 0.2817 0.1409
Al 20.88 2370 4.1082 0.4108 0.2054
Ca 12.23 3800 3.2444 0.3244 0.1622
Fe 27.02 2950 3.6823 0.3682 0.1841
Mg 0.99 100 20.0000 2.0000 1.0000
Mn 0.77 370 10.3975 1.0398 0.5199
Ti 0.10 30 36.5148 3.6515 1.8257

Table 2. Average compositions and standard deviation of garnet.

Garnet core (n = 825) Garnet rim (n = 1282)

... Average St. Dev. Average St. Dev.
SiO2 37.54 1.092 37.804 1.061
TiO2 0.076 0.056 0.07 0.052
Al2O3 20.751 0.736 21.022 0.736
FeO 32.78 1.076 31.823 1.062
Fe2O3 0 0 0 0
MnO 1.11 0.095 0.59 0.083
MgO 2.107 0.292 2.873 0.344
CaO 4.917 0.264 5.086 0.283
Na2O 0.035 0.047 0.036 0.05
K 2O 0.052 0.032 0.055 0.05
Structural formula (12 anhydrous oxygen basis)
Si 3.031 0.055 3.03 0.054
Al 1.975 0.062 1.986 0.062
Mg 0.254 0.035 0.343 0.041
2+
Fe 2.215 0.071 2.135 0.069
Mn 0.005 0.007 0.006 0.008
Ca 0.426 0.023 0.437 0.025
XAlm 0.746 0.012 0.722 0.014
XSps 0.026 0.002 0.014 0.002
XPrp 0.085 0.011 0.116 0.013
XGrs 0.143 0.008 0.148 0.008