Major to trace element imaging and analysis of iron age glasses using stage scanning in the analytical dual beam microscope (tandem)

Abstract

Dark and clear silicate glasses formed during an iron age vitrification event ≈ 1500 years ago at the Broborg hillfort near Uppsala, Sweden have been analyzed using a scanning electron microscope equipped with a micro-X-ray fluorescence (μXRF) spectrometer. Correlated µXRF and electron beam-induced energy dispersive spectrometry (EDS) X-ray maps were collected via stage-scanning at constant velocity. This coupled procedure represents a new approach for the cultural heritage community to conduct analytical studies of archaeometric specimens composed of metal, ceramic, or mixed inorganic/organic materials, where major and trace element compositions are registered in space for areas up to the centimeter-length scale at micrometer-scale resolution. Overview images were used to select areas for EDS beam scan maps correlated with multispectral cathodoluminescence (CL) imaging and co-located quantitative EDS and μXRF point analysis. Fe, Ca, Mg, Ti, P, Mn, Zr, Zn, and Y are enriched in the dark glass, while Si, Al, K, Na, Ba, Sr, Rb, and Ga are enriched in the clear glass. Unmelted material is comprised predominately of quartz (SiO₂) along with trace apatite (Ca₅(PO₄)₃[Cl,OH]) and zircon (ZrSiO₄). Multivariate statistical analysis was used to measure the area fractions of high variance components while lower variance components represented phase mixtures. Differences between calculated melt viscosities for the glass compositions are consistent with field and laboratory observations. Coupled large area EDS and μXRF imaging shows significant promise for informed selection of higher spatial resolution and higher sensitivity follow-up studies, e.g., those performed using synchrotron analysis.

Graphical Abstract

Introduction

Studies combining or comparing results from the scanning electron microscope (SEM) and atmospheric handheld X-ray fluorescence spectrometry (XRF) obtained ex situ are not uncommon in the literature [1,2,3,4,5,6], however XRF systems embedded in electron beam instrumentation represent more rare occurrence [7, 8]. Advances in the field of polycapillary X-ray optics [9,10,11,12] enabled the implementation of truly integrated microbeam XRF (μXRF) into the SEM [13,14,15]. The principal advantages of the new method in combining the two technologies into one platform includes: high fidelity positioning of the XRF beam via electron beam imaging and coupling energy dispersive spectrometry (EDS) obtained by electron beam excitation complete with low energy X-ray information (e.g., C, N, and O), to higher sensitivity XRF data for moderate to high energy X-ray K lines (e.g., Zn, As and Rb, Sr, Nb). Co-locating the two X-ray spectra that contain: (1) light element information used to identify organics, biological material and metallic corrosion, along with (2) trace element information from heavy elements useful for constraining object provenance, offers a cultural heritage analyst an efficient approach to the compositional documentation of complex artifacts.

Recently, the value of coupled quantitative EDS-μXRF point analysis to heritage studies has been demonstrated in determining the geological source locality for several highly polished obsidian tablets fabricated during the late Postclassic period in Central Mexico [16]. Analogously, in this effort, we utilize co-located quantitative EDS-μXRF analysis to examine late iron age glassy material that resulted from a complex vitrification event at the Broborg hillfort in Sweden. The Broborg hillfort is an ≈ 1500-year-old fortified archeological site situated approximately 20 km southeast of Uppsala, Sweden (17.9515°E, 59.7556°N) where portions of the fort’s stone walls have been reinforced via vitrification [17, 18].

The question of whether hillfort vitrification was the result of hostile interaction, or was intentionally performed by its builders has been part of a long running debate. However in the case of Broborg, the strategic location along a Migration Period waterway, the box-like ventilation structures found within the walls, the evidence for the use of charcoal rather than wood for fuel, and the fact that vitrification covers nearly the entire circumference of the inner rampart, all support the notion that vitrification was used as a deliberate construction technique [19, 20]. The vitrified components at Broborg have been previously analyzed in terms of their socio-historical context [17], as well as pre-Viking, Scandinavian materials’ technology development [21, 22], and represents among the earliest known vitrification efforts in Sweden [23].

In this paper, we extend EDS-μXRF methodology in The Analytical Dual Beam Microscope (TAnDeM, hereinafter tandem) by introducing a new archaeometric methodology that is applicable to the study of a wide range of object composition, using X-ray imagery produced by sequential EDS and μXRF mapping to make informed selection of areas for quantitative analysis. Such maps are acquired by continuous stage scanning of the artifact under stationary electron and photon microbeams and result in X-ray maps that cover centimeters squared in area at micrometer scale resolution. This approach allows for a more complete elemental budget, including major, minor, and trace element images to be evaluated prior to detailed follow-up studies. Post stage-scanning techniques used in this study include cathodoluminescence (CL) imaging that can be sensitive to defects and/or trace impurities [24, 25], and additional electron imaging to co-locate quantitative EDS and μXRF point analyses to determine element ratios potentially yielding information regarding an object’s provenance [16].

Materials and methods

The specimen examined labeled BB1b was collected at Broborg hillfort, near Uppsala, Sweden, during a 1980’s excavation by Peter Kresten and co-workers [20]. The sample was later transferred to the U.S. DOE Office of River Protection, and subsequently to Pacific Northwest National Laboratory (PNNL, USA) in 2016 as part of an evaluation of silicate glass stability [22]. The hand specimen is a white and light tan colored granitic gneiss and was taken from a vitrified portion of the hillfort. A centimeter-sized fragment of the specimen was cut from the main mass, vacuum impregnated with epoxy, and polished using a water-free procedure at PNNL and contains two varieties of glasses (dark and clear) as observed using light microscopy (Fig. 1), and represents a single specimen of a complex vitrification process.

Fig. 1

Microscopy of Broborg hillfort fragment BB1b with contrast features highlighted. A BSE image montage. B High dynamic range visible light image where a dark glass coating is shown that envelopes much of the specimen

An Hitachi S3700N scanning electron microscope (SEM) equipped with a Bruker XFlash 6|60 energy dispersive spectrometer (EDS) and a Gatan ChromaCL2 multispectral cathodoluminescence (CL) imaging system was used for electron beam imaging and analysis at the Smithsonian Institution. The SEM chamber was modified to include a Bruker XTrace X-ray Fluorescence (XRF) source coupled to polycapillary optics that yields an X-ray spot size of ≈ 33 μm (measured at Cu K), hence the term µXRF is used to describe the method. A further modification was made to the SEM by way of the addition of a piezo stick–slip substage capable of 5 cm travel in both X and Y axes (Fig. 2). The substage, aka Bruker Rapid Stage, addition allows one to raster specimens in the SEM under the static XRF microbeam, or an undeflected electron beam, to produce X-ray images. The substage can collect mm- to cm-sized X-ray images induced by the electron beam beyond the field of view obtainable by beam deflection without the need for constructing a montage of image tiles. While substage speeds of 20 mm s⁻¹ are achievable in atmosphere, velocities were limited to ≈ 4 mm s⁻¹ owing to limited heat dissipation in vacuum. Given the additional XRF capabilities to the traditional SEM platform, the coupled electron and µXRF beam technique has been dubbed “The Analytical Dual Beam Microscope” (tandem).

Fig. 2

Piezo stick–slip substage and specimen holder assembly coupled to the SEM stage

SEM–EDS conditions for imaging and analysis included an electron beam accelerating voltage of 15 kV and 1 nA of current. Quantitative EDS point analyses were collected under high vacuum for 60 live seconds using a pulse processing time constant of 470 ns while rastering the electron beam over a 40 µm × 40 µm area to mitigate the impact of time dependent X-ray intensities. Faraday cup measurements on the specimen holder collected before and after each analysis indicated variability was less than ± 1% relative to the set beam current. Bruker Esprit ver. 2.2 software was used for standards-based quantification of spectra using a phi-rho-Z matrix correction procedure. Primary elemental intensities were collected from commercially obtained materials through SPI Supplies, including: wollastonite (CaSiO₃) for Ca and Si (Jensen Quarry, Crestmore, CA, USA), synthetic TiO₂ for Ti (Atomergic Chemetals Corp.), orthoclase (KAlSi₃O₈) for K and Al (Madagascar), hematite (Fe₂O₃) for Fe (Elba, Italy), synthetic MgO for Mg (Atomergic Chemetals Corp.), rhodenite (MnSiO₃) for Mn (B.C., Canada), albite (NaAlSi₃O₈) for Na (Amelia, VA, USA), and synthetic CeP₅O₁₄ for P, while well-characterized materials from the Smithsonian collections were used to evaluate measurement quality relative to published values (Additional file 1: Table S1) [26, 27].

Micro-XRF spectra were collected using a Rh source operated at 50 kV and 600 µA. The XRF spot was aligned to the center of the electron beam raster using secondary electron (SE) imaging of an uncoated light microscopy glass coverslip at 3 kV, where the X-ray spot suppresses the yield of SE [14]. The working distance for this lower energy beam differs from that for 15 kV used for EDS analysis by 200 µm. Given the X-ray optical sensitivity to height differences, imaging and analysis of coupled EDS and µXRF was performed at 9.8 mm instead of at the standard analytical position of 10 mm. Point analyses and images were collected using unfiltered radiation, and spectra were processed using fundamental parameters (FP) quantification implemented in Bruker Esprit ver. 2.1 [28]. A single data set with a 100 µm thick Al source filter was used to evaluate signal to noise ratio differences relative to unfiltered spectra. X-ray emission depths were computed for the dark and clear glass compositions via Monte Carlo simulation for EDS using DTSA II [29] and Bruker XRF Check [30]. Stage scans for EDS and µXRF X-ray imagery were collected at 2 µm pixel⁻¹ and 16 µm pixel⁻¹ using constant velocities of 4.1 mm s⁻¹ and 0.08 mm s⁻¹, respectively.

In order to extract meaningful chemical information from the large size of the EDS hyperspectral dataset, a multivariate statistical analysis (MVSA) treatment was utilized. This type of analysis is commonly employed for hyperspectral datasets when the number of phases is large and/or when many of the phases contain similar elements [31,32,33,34,35,36]. In such cases, simple elemental maps extracted from the dataset are often difficult to interpret, as many regions of the sample will show similar intensities for each element. By contrast, the MVSA analysis allows the data to be ‘demixed’ so that the individual phases can be spatially isolated even when they share some of the same elemental signatures. The goal of this type of MVSA treatment is to decompose the three-dimensional hyperspectral dataset (N_rows × N_cols × N_channels) into a smaller number of easily interpretable components which describe the statistically significant features in the data. It is also useful for finding minority phases which might otherwise be lost in the analysis of a large dataset [37]. Additional information regarding the detailed MVSA procedures can be found in Additional file 1: Note (S-1).

The MVSA results were also used to estimate the area fraction of the specimen represented by the various components. To do so, each component image was first segmented using the Otsu method [38], providing the total number of pixels associated with each component. Finally, the fraction of the total number of specimen pixels represented by each component was calculated while ignoring the embedding epoxy and a low energy X-ray absorption artifact.

Results

EDS stage-scan imaging

EDS stage maps of major and minor elements reveal contrast between the dark glass enriched in Fe, Mg, and Ca relative to the clear glass regions enriched in Al and K (Fig. 3). An aggregate of isolated Fe-bearing spinels ([Fe²⁺,Mg]Al₂O₄) differs in morphology from large regions of irregularly shaped quartz (SiO₂) grains that represents the dominant unmelted crystalline residue (Fig. 3A, B). The dispersed Ca-rich feldspar (CaAl₂Si₂O₈-NaAlSi₃O₈) is of uncertain origin within the clear glass and is distinct from the feldspar relicts related to the protolith and represents a non-equilibrium composition [39] (Fig. 3C), while trace apatite (Ca₅(PO₄)₃[Cl,OH]) grains are likely unmelted crystalline remnants. BSE images showing the clear and dark glasses along with inhomogeneous crystalline zones are depicted in Fig. 4 while correlated elemental maps are shown in Additional file 1: Figure S1. MVSA processing of the EDS stage scan data yields six component images and spectra. The highest variance components, quartz and clear glass, are well-resolved while dark glass is comprised of multiple components, including: (1) feldspar quench crystals in dark glass convolved with Ca-rich feldspars, (2) dark glass, (3) spinel + dark glass, and (4) apatite + dark glass (Figs. 5, 6). The elemental complexity of the dark glass is represented by the four mixed components that account for the lowest variance in the system above the noise level (Fig. 6).

Fig. 3

Stage scan EDS images with highlighted extracted spectra. A Mg (red) and Fe (green) X-ray image overlay with dark glass and spinel + glass extracted spectra. B Al (red), Si (blue), and Fe (green) X-ray image overlay with quartz and clear glass spectra. C Ca (red), K (blue), and Fe (green) X-ray image overlay with apatite + mixed glass and Ca-rich plagioclase + clear glass extracted spectra

Fig. 4

BSE imagery at differing length scales. A Overview image of BB1b. B BSE montage of area highlighted by white rectangle in A. C Higher resolution image of clear glass-rich zone with heterogeneous regions highlighted by white polygons. White arrows highlight residual quartz (qtz) and zircon (zr) (see Additional file 1: Figure S1 for Si an Zr elemental maps). D Higher resolution image of dark glass-rich with white polygons outlining microcrystalline zones. Note: White- and black-filled ellipses represent the size of the µXRF beam for reference within dashed lined rectangles in C and D respectively

Fig. 5

Chemical phase output from multivariate statistical analysis of EDS stage scan data set. A Composite overlay of six component images that represent quartz (red), clear glass (green), Ca feldspar + dark glass (blue), dark glass (dark cyan), Fe-spinel + dark glass (magenta), apatite + dark glass (yellow). B Component spectra with colors conserved from A with thresholded area fractions expressed as a percentage

Fig. 6

Multivariate statistical analysis of EDS stage-scan data. A Eigenvalue versus component number with non-noise components labeled. B Component images with associated component number labels are each scaled individually and shown using a thermal color map to aid interpretation. Horizontal field width = 7.5 mm

µXRF stage-scan imaging

The overview XRF imaging data set contains an array of elemental information spanning major to minor elements, and a selection of trace elements. This large dynamic range comes at the expense of spatial resolution relative to electron beam-induced data given the tens of micrometer lateral size of the µXRF spot for laboratory-based systems [14]. These differences are illustrated in a side-by-side comparison of registered Ca K line images for EDS and µXRF showing high Ca contrast in three regions of BB1b including a complex particle zone that is part of a follow-up study and is outside the scope of this paper (Fig. 7).

Fig. 7

Stage scans depicting the Ca distribution in Broborg hillfort fragment where the pixel intensity is represented by gray level. A EDS image showing higher spatial resolution. B µXRF image showing large dynamic range contrast. Orange ellipse represents the residual apatite in mixed glass, dashed orange ellipse represents Ca-feldspar zone (milky inclusion-Fig. 1B), and light green ellipse represents the complex particle zone

With respect to depth resolution, the physics of X-ray attenuation govern the XRF information depth for a matrix composition [28]. Using a figure of merit of 90% of emitted X-rays, the computed emission depths are divided between major and minor elements and trace elements in the Broborg glasses (Additional file 1: Figure S2). For low to moderate energy X-ray lines, i.e., Na, Mg, Al, Si, and P, the XRF depth resolutions are estimated to be < 2.4 µm and are similar to those estimated for EDS emission depths obtained by a Monte Carlo simulation (< 1.4 µm). However, for more energetic major and minor element K lines (e.g., K, Ca, Ti, and Fe), the differences between XRF and EDS depth resolutions diverge nonlinearly (Additional file 1: Figure S2A). The nonlinear increase in information depth for trace elements continues for Cr, Mn, Cu, Zn, Ga, Rb, Sr, Y, Zr, and Nb, with increasing K line X-ray energies (Additional file 1: Figure S2B). The highest atomic number trace element examined in this study is Ba, and its emission depth is similar to that for low Z first row transition metals (e.g., Ti) given its concentration is measured using the L line X-ray at lower energy.

Micro-XRF stage scan images reveal that Zn, Zr, and Ti, are enriched in the dark glass while Ba, Si, and Rb are enriched in the clear glass (Fig. 8A–C). The distribution of Ti depicts a thin film of dark glass along the uppermost margin of the specimen coating both quartz and clear glass (Fig. 8A). Zr hotspots appear throughout both glasses highlighting the presence of surficial and buried zircon (ZrSiO₄) crystals (Fig. 8B). The sum of X-rays with energies less than the Rh K line source offers a view of the specimen that may resemble a BSE image, and as such may be useful as a navigational tool for ex situ XRF imaging studies, albeit with poorer contrast between low mean atomic number phases (Fig. 8D). The sensitivity of XRF can be demonstrated by spectral resolution of a trace Ba in the presence of a major to minor concentrations of Ti, while high energy trace element X-rays are significantly less convolved (Fig. 8E–G). To determine the practical limits of trace element imaging a second µXRF stage scan was conducted using a 100 µm Al filter. Comparative trace Zn images show that the low sensitivity (i.e., peak/background) can be improved by a factor of three using a filtered Rh source allowing for the ability to distinguish low concentration phases (< 190 ppm) that differ by 100 ppm (Fig. 9A–C). A consequence of using a filtered X-ray source is the elimination of low energy diffraction artifacts, which in turn allows for the detection of trace Ti, along with suppression of higher energy diffraction in the X-ray spectrum (Fig. 9D).

Fig. 8

Stage scan µXRF images, extracted spectra, and deconvolved spectral features of major, minor, and trace elements. A Ba (red), Si (blue), and Ti (green) X-ray image overlay. Small white arrows depict dark melt coating the specimen exterior, and larger black arrows highlight Ba- and Ti–rich regions of clear and dark glasses respectively. B Si (blue) and Zr (yellow) X-ray image overlay showing Zr hotspots. C Si (green), Zn (red), and Rb (blue) X-ray image overlay where black arrow highlights a diffraction artifact (see Fig. 9). D image of the sum of all X-rays with energies between 1 and 18 keV. E extracted spectra from Ti–rich and Ba-rich regions highlighted in A. F Deconvolved trace Ba signal in the presence of significantly greater Ti counts. G Deconvolved peaks for trace elements with high energy X-ray lines

Fig. 9

source filtering, and spectral features. A Unfiltered Zn image. B Filtered Zn image using 100 µm of Al. C Signal/noise enhancement of Zn and Ga K line peaks after background removal. D Artifacts in spectra owing to Bragg diffraction of quartz crystals highlighted by white ellipses in A and B are labeled d1 and d2 at lower and higher energy respectively. Horizontal field width = 7.5 mm

XRF trace element imaging 

Targeted imaging and analysis

Imaging and analysis of the Broborg hillfort glasses that followed overview stage-scans included: (1) higher spatial resolution EDS beam scanning with correlated multispectral CL imaging, and (2) quantitative EDS and µXRF point analysis. At higher spatial resolution the interfacial region between a quartz grain and the dark glass reveals the dark glass is comprised of an intergrowth of plagioclase and FeMgCa-rich quench crystals, likely Ca-pyroxene, within a residual glass bordering a smooth quartz interface (Fig. 10A, B). CL imaging in the visible, or red–green–blue (RGB), is a function of intrinsic luminescence coupled with trace impurity activated luminescence in natural phases [40, 41] and distinguishes these phases as quartz (red), plagioclase (violet), FeMgCa dendrites (non-luminescent) likely augite [(Ca)(Mg,Fe,Al)(Si,Al)₂O₆], and residual glass (light blue and green) (Fig. 10C, D).

Fig. 10

Beam scan images of quartz-dark glass interface. A electron image of interface depicting the EDS image (white rectangle) and CL image (dashed white square) regions. B EDS X-ray overlay of Mg (red), Si (green), Al (blue), and Fe (yellow) highlighted in A. C RGB CL image highlighted in A showing CL distinguishing two glasses, blue in the matrix and green in the boundary layer with quartz. D RGB CL image of adjacent interface zone showing intergrowth of blue-violet plagioclase blades and non-luminescent FeMg dendrites. Abbreviations for labeled phases include plagioclase (pl), quartz (qz), glass (gl), and boundary layer glass (gl_b)

Information derived from above stage-scans was used to identify homogeneous regions of the dark and clear glasses (Additional file 1: Figure S3) for co-located EDS and µXRF quantitative point analysis (Table 1; Additional file 1: Tables S2–S5). Major and minor elements presented in Table 1 represent standards-based EDS and FP-based µXRF quantification with a comparison of results obtained for well-characterized materials presented in Additional file 1: Table S1. A graphical comparison of major, minor, and trace element compositions for the dark and clear glasses is shown in Fig. 11 where nine elements are enriched in the dark glass and eight are enriched in the clear glass. However, the degree of enrichment in the dark glass for Ti, Fe, Mg, Ca, P, Mn, Zn, Y, and Zr is significantly greater than that for Si, Al, Na, K, Ga, Rb, Sr, and Ba in the clear glass (Fig. 11). A multi-element ratio plot, comprised of major, minor, and trace elements shows normalized mean dark and clear glass compositions, where larger ions are enriched in the clear glass, and conversely, smaller ions are enriched in the dark glass (Fig. 12A). The same plot is used to compare the dark glass to bulk rock compositions for amphibolites (a rock comprised of amphibole, plagioclase feldspar, quartz, and biotite) found at Broborg [21] (Fig. 12B). While the multi-element ratio patterns are not entirely dissimilar, the dark glass is enriched relative to the bulk amphibolites for all elements apart from Ba. The coupled data from co-located EDS and µXRF quantified analyses allows one to probe such elemental relationships for a large dynamic range of elemental concentrations using the tandem analysis method.

Table 1 Quantitative EDS and µXRF point analyses for the dark and clear glasses

Fig. 11

Elemental enrichment factors and 1 σ propagated uncertainties based upon the ratio of mean compositions in Table 1

Fig. 12

taken from Rudnick and Gao [48, 49] and X-axis element order generally, but not exclusively, in order of decreasing ionic radius

Normalized multi-element diagram, or spidergram, where specimen compositions are shown relative to the Earth’s bulk continental crust. A Mean dark and clear glass compositions in BB1b are shown in red nested- and blue cross haired-filled circles. B Patterns for six representative amphibolites from [21, 42] shown in black and gray lines enriched at Broborg hillfort relative to the boulder lithologies found in the surrounding area [20] and the mean dark glass composition measured here. Composition of the bulk continental crust

The mean glass composition for clear and dark glasses was also used to compute anhydrous viscosities for Fe-bearing silicate melts  [43]. Temperatures used for the calculation were taken as the mean of the range reported from Fe–Mg exchange between olivine and dark melt (1048 °C), as well as the mean of liquidus temperatures for the clear melt (1068 °C) as previously determined by Kresten et al. [20] (Fig. 13). Both temperatures result in an estimate of the dark melt being less viscous than the clear melt by a factor 74 × during the hillfort vitrification event, and this viscosity difference is relatively insensitive to a broad range of possible melt temperatures. However, whether the dark and clear melts were simultaneously molten cannot be determined from our data.

Fig. 13

Computed viscosities based upon dark and clear glass compositions as a function of temperature. Temperature values taken from Kreston et al. [20] for the estimates of the clear melt liquidus and FeMg exchange between olivine and melt for the dark melt. Note: The range of liquidus temperatures for the clear melt is smaller than the size of the blue-filled circle. Viscosity uncertainties are ± 10% for 95% confidence limits [41] and the gray dashed lines represent isothermal ties for the two temperatures

Discussion

Broborg hillfort glasses

The quantitative analyses presented here represent the most complete effort to characterize chemical differences between the dark and clear glasses found at Broborg hillfort with respect to an increased dynamic range of element concentrations. While the compositions reported here by no means represent the range of glass chemistry found at Broborg, they offer insight into conditions for a single portion of the hillfort rampart. Major and minor element patterns compare favorably to earlier electron beam studies with some notable differences: (1) dark glass compositions reported previously are enriched in MgO and CaO and depleted in SiO₂, Na₂O, and K₂O relative to the dark glass in BB1b [18, 21], and (2) the composition of a selected clear glass composition is within uncertainty of the mean clear glass value reported here for SiO₂, CaO, TiO₂ and outside of 1 σ values for Fe₂O₃, K₂O, and P₂O₅ [18].

The large difference in computed anhydrous viscosities between the melts, suggesting approximately a 74-fold decrease in resistance to flow for the dark melt relative to the clear melt estimated for specimen BB1b, is consistent with our laboratory observation of a thin film of dark glass coating the stone exterior. The computed viscosity for the dark melt at 1048 °C of 4.5 Pa·s is equivalent to that for pine honey with 21% moisture at 25 °C of 4.8 Pa·s [44] and is thus in good agreement with field observations at Broborg where dark melts are noted to coat rocks and additionally form melt “drips” compared with clear glasses found localized in situ in the vitrified zone of the rampart [17, 20, 21].

The mean clear glass composition presented here is in good agreement with the major and minor elements for the experimentally derived clear glass compositions obtained by partial melting of granitoid rocks from the hillfort [21], and supports the earlier suggestion that clear glasses from Broborg generally result from nonequilibrium partial melting of the feldspathic component of granitic gneiss [18, 20]. The mean dark glass composition we report is generally consistent with, but not an exact compositional match for the experimental glasses derived from partial melting of amphibolite [21]. The field association noted for amphibolite on the rampart with dark glass at Broborg has been noted as unequivocal, and the dark stones are found in anomalous proportions relative to the abundant granitoid boulders surrounding the hillfort [18, 20]. Interestingly, our multielement plot shows a similar pattern between bulk amphibolites from Broborg and the dark glass from BB1b (Fig. 12 B) with the exception of depleted Ba values which can be related to retention of plagioclase feldspar during partial melting [45]. As such, these new element patterns support the earlier field, analytical, and experimental studies that link the dark glass to the incongruent melting of amphibolite as the most likely source to form the vitreous cement in the hillforts’ rampart [18, 20, 21].

Value of correlated stage-scan imaging

Registered imagery from EDS and µXRF stage scanning collected in tandem offers a new method to evaluate specimens with microstructural features that span millimeters to centimeters in scale. These combined hyperspectral data sets cover a large fraction of the energy spectrum from soft X-rays (e.g., B, C, O) to high energy K lines (e.g., Zr, Nb, Mo) with the realization that high energy X-rays may be generated from tens to hundreds of times deeper within specimens relative to EDS emission depths. Such complete microchemical characterization enables an elemental context to be established for follow-up studies using a variety of ancillary methods to probe materials in the SEM for defects, crystallographic orientation, and atomic resolution studies using CL, electron backscattered diffraction (EBSD), and 3D focused ion beam sectioning and lift out for transmission electron microscopy (3D FIB/TEM).

Archaeometric studies of metal, ceramic, and mixed materials that demand nondestructive or minimally destructive analysis will benefit from this coupled approach, particularly those that require larger examination areas; for example, durability of ancient concrete [46, 47]. Another important role tandem analysis can fill is screening specimens proposed for synchrotron studies. Given the high value of time on beam lines at light source facilities, advanced knowledge of a material obtained from laboratory µXRF and EDS can be invaluable for optimizing measurements performed with high brightness sources, especially for synchrotron mapping studies.

Conclusions

Correlated EDS and µXRF stage scan imaging using the tandem system for a Broborg vitrified hillfort specimen reveals an assemblage dominated by FeMgCa-enriched dark glass, SiAlK-clear glass, quartz, spinel, Ca-feldspar, apatite, and zircon. MVSA output obtained from the EDS hyperspectral data allows for the determination of area fractions for chemical phases over a large area covering tens of square millimeters at micrometer-scale resolution. Dark and clear glasses compositions define distinct populations and differences in their computed viscosities at vitrification temperatures align well with laboratory and field observations.

While used to illustrate an overview analysis of late iron age glasses at a vitrified hillfort, the coupled imaging and analysis technique described here is broadly applicable to a variety of cultural heritage materials and enables routine minor and trace element data collected via µXRF that complements traditional major element EDS output. This new capability for the SEM platform allows for a workflow that includes a nondestructive microchemical overview via stage scanning prior to detailed studies performed either in situ, or ex situ, for example at synchrotron facilities.