*3.2. Archaeometric Analyses: Defining the Analytical Approach Based on Material Features*

Once the selection of materials to be analysed has been completed, the setup of the analytical protocol may vary according to the features of the materials themselves. An indepth discussion of the potential and limits of all possible analytical techniques applicable to the study of archaeological glass would be beyond the scope of this article. Therefore, in the following discussion, reference will be made to the most commonly used analytical techniques for archaeometric studies on glass to foster data comparability.

The first criterion influencing the choice of analytical protocol is whether the materials to be analysed are naturally or intentionally coloured glasses. The archaeometric approach to the study of nonintentionally coloured glass (which means without the intentional addition of colouring and opacifying agents to the base glass) is simpler than deeply coloured (and often opaque) glass. The latter is, in fact, characterized by greater heterogeneity in terms of microstructure and compositional features, due to the addition of specific raw materials aimed at imparting the desired colour and degree of opacity. The setup of a more articulated analytical approach is needed based on the interrelation of different investigation techniques to achieve an in-depth characterization of colouring and opacifying agents. When dealing with naturally coloured glass, the archaeometric approach is based on the use of fewer analytical techniques, the primary objective being the determination of the chemical composition of the base glass and, if needed, the provenance of the materials used as vitrifying agents. Figures 1 and 2 show block diagrams of the two different analytical approaches applicable to the study of naturally and intentionally coloured glass; it is always advisable to carry out an observation and documentation of the samples in optical microscopy, preliminary to the filing of the fragments/objects under study.

**Figure 1.** Block diagram of the proposed analytical approach for naturally coloured archaeological glass.

**Figure 2.** Block diagram of the proposed analytical approach for coloured archaeological glass.

When archaeometric analyses have to be performed on intentionally coloured glasses, it is highly recommended to start with an in-depth study of the colouring (and opacifying) phases. As more extensively discussed elsewhere [8], since deeply coloured glasses show highly heterogeneous microstructures and microtextures, a thorough characterisation of colouring and opacifying phases should be carried out before the bulk composition of the vitreous matrix to avoid misunderstandings in data processing and interpretation.

A combined Standard Colour System Chart (such as NCS Index or PANTONE) and Visible Reflectance Spectroscopy (VIS–RS) approach is here proposed as a starting point. It can support, in fact, an objective definition of the chromatic hues and shades of deeply coloured glasses avoiding any subjective nomenclature. The example par excellence is represented, in this case, by a particular category of glass-based artefacts: mosaic tesserae. For these small cubes of coloured glass, a chronotypological study is unfeasible; therefore, colours and opacity are the only macrofeatures that can be used to methodically select among copious assemblages of samples to be studied. NCS coordinates can be used to effectively separate the tesserae into preliminary chromatic macrocategories (i.e., yellow, green, blue, red, and black). These groups can be defined based on NCS coordinates and, more precisely, by taking the second part of the NCS-notation into account, which describes the hue by means of a numerical code. For example, a tessera with NCS-notation S 2030-G70Y will be described as of a yellow colour and code G70Y indicating a colour shade described as a yellow (Y) with 70% resemblance to yellow and a 30% resemblance to green (G). It is possible, in this way, to avoid the use of definitions such as "greenish yellow" or "yellowish green", eliminating a first important degree of subjectivity in the definition of colour. For NCS evaluations to be accurate, it is important that observations are made in controlled repeatable daylight conditions avoiding any artificial source of light.

After preliminary NCS-aided discrimination between chromatic macrocategories, further data on optical properties (L\*a\*b\* numerical coordinates and the reflectance for each wavelength in the visible spectrum) can be collected by VIS–RS. Figure 3 shows a comparison between reflectance curves acquired on tesserae belonging to Red, Yellow, Green, and Blue chromatic macrocategories identified by taking NCS-notations into account. Reflectance curves of tesserae belonging to the NCS-Red macrocategory displayed a very flat behaviour in the wavelength range between 400 and 580 nm, followed by an increase in reflectance intensity for the wavelengths above 580 nm. NCS-Blue and NCS-Green tesserae showed bell-shaped reflectance curves; the reflectance peak was located in the region between 440 and 540 nm for the blue tesserae, while, for the green tesserae, it was slightly shifted between 470 and 540 nm. Last, for the NCS-Yellow tesserae, reflectance curves were characterised by an increase in reflectance intensity for the wavelengths above 560 nm. Though further research is needed, the potentiality of VIS–RS seems to go far beyond the description of colours by means of reflectance curves and L\*a\*b\* numerical coordinates. In particular, the shapes of the reflectance curves in the visible spectrum and

the percentages of reflectance can deliver preliminary qualitative information relating to the colouring and opacifying agents [6,20,21].

**Figure 3.** Comparison between reflectance curves acquired by VIS–RS on NCS-Red, Yellow, Green, and Blue tesserae.

After preliminary OM documentation, sampling, embedding, and polishing, SEM– EDS should be performed on deeply coloured glass. Back-scattered electrons (BSE) signals allow detecting and documenting of the different morphologies of the crystals precipitated into the glassy matrix, with EDS spot measurements also ascertaining their elemental composition. SEM–EDS is undoubtedly suitable to carry out high magnification morphological inspection of the inclusions dispersed in glassy matrix, as well as providing a qualitative and semiquantitative analysis of their elemental composition. However, to provide a more in-depth characterisation of these inclusions, necessary to identify raw materials responsible for the colour and opacity of the tesserae, SEM–EDS inspection needs to be integrated with other analytical techniques.

Especially if coloured and opaque, archaeological glasses are strongly heterogeneous materials; it is, thus, quite challenging to define what is the most suitable analytical technique for providing a full characterisation of the inclusions. It would, maybe, be more correct to claim that there is not only one. The choice of the technique/s is highly dependent upon the nature of the inclusions that we want to investigate. An integration of at least one molecular (such as Raman Microscopy) and one mineralogical analysis (X-ray Powder Diffraction—XRPD, micro-XRD or SEM equipped with an Electron Backscattered Diffraction detector—EBSD, if available) is recommended, offering an appropriate compromise to achieve a thorough characterisation of the inclusions responsible for the colour and opacity of glass.

The characterisation of lead–tin–antimony-based compounds found in ancient glasses as colouring and opacifying agent can be taken as an example. As discussed elsewhere in more detail [8], lead–tin–antimonate crystals have been attested to in several assemblages of opaque coloured glasses [22–25]. The most credited hypothesis is that the presence of lead–antimonate inclusions doped with tin could be related to the use of tin-rich metallurgical scraps [26,27]; therefore, an exact characterisation of these inclusions could represent the starting point for providing insights into the identification of a possible area of origin of the raw materials used as colouring and opacifying phases. SEM–BSE images (Figure 4a) provide information on the morphology of the inclusions, showing the occurrence of micrometric anhedral crystals in the vitreous matrix, frequently clustered together; EDS spot measurements carried out on the crystals gave preliminary information on the elemental composition of the crystals, demonstrating that they were mainly made of antimony and lead, although tin could also be detected (Figure 4b). To achieve an exact characterisation of the compound, Raman microscopy was directly performed on the inclusions (Figure 4c). Acquired spectra showed the typical features assigned to lead antimonate doped with tin: in addition to the shifted Pb–O lattice mode at 140 cm−1, a peak at about 450 cm−1, an increase in the band at about 330 cm<sup>−</sup>1, and a collapsed band at 510 cm−<sup>1</sup> were observed, indicative of the partial replacement of the Sb+5 species by a larger Sn4+ cation [28,29].

**Figure 4.** Example of lead–tin–antimony-based inclusions detected in a yellow mosaic glass tessera: (**a**) SEM–BSE image; (**b**) EDS spectrum; (**c**) Raman spectrum with bands at 140, 332, 450, and 510 cm−<sup>1</sup> suggesting lead antimonate doped with tin [28,29].

In the above example, Raman microscopy can be considered as the most suitable analytical technique to gain data on the composition of the lead–tin–antimony-based compounds used to impart yellow colour and opacity to the glass. There are, however, cases in which the use of this analytical technique may not be informative. An example is represented by the opaque red coloured glass containing metallic copper inclusions. SEM–BSE images (Figure 5a) showed, at high magnification, nanometric rounded particles exclusively made of copper as EDS spot analysis demonstrate (Figure 5b). The nanometric size of these inclusions did not allow Raman microscopy analysis, where the magnifications of the most commonly accessible instruments hardly exceeds 1000×. Diffractometric analysis (Figure 5c) can, in this case, be decisive, allowing one to distinguish the presence of metallic copper within the glass matrix. It should, of course, be emphasized that, unless instruments such as microdiffractometers and/or BSDE detectors interfaced with SEM are available, diffractometric analysis usually requires powdering the sample and, therefore, is destructive.

Once the investigation of colouring and opacifying phases was completed, we analyzed the bulk chemistry. To investigate the base glass, both in terms of compositional recipes and provenance of raw materials, a combination of Electron Probe Micro Analysis (EPMA) and Laser Ablation–Inductively Coupled Plasma–Mass Spectrometry (LA–ICP– MS) analyses was proposed, together with specific data processing. Major and minor oxides, aimed at identifying both the fluxing agent and the "recipes" used in the glassmaking process, can be determined by EPMA, while LA–ICP–MS analysis needs to be carried out for measuring trace elements and, thus, drawing inferences on the provenance of the sands used as vitrifying agents. Several analytical techniques can be employed for the quantification of major, minor, and trace elements in archaeological and historical glasses, such as Wavelength Dispersive X-ray Fluorescence spectrometry (WDXRF), Inductively Coupled Plasma–Optical Emission Spectrometry (ICP–OES), Ion Beam Analysis (IBA), and Neutron Activation Analysis (NAA) [30]. However, when dealing with opaque coloured

glasses the prime advantage of combining EPMA and LA–ICP–MS for quantification of major to trace elements is the possibility of performing both analyses on the same mounted and polished samples where the study of colourants and opacifiers had previously been carried out. When dealing with archaeological and historical glasses, it is important to identify "recipes" that can be linked to primary production furnaces and, therefore, provide information on geographical areas of production and provenance [3]. However, for deeply coloured glass, a further degree of difficulty is due to the "contamination" of the compositional data of the matrix with materials added as colouring and opacifying agents.

**Figure 5.** Example of nanometric copper-based inclusions detected in a red mosaic glass tessera: (**a**) SEM–BSE image; (**b**) EDS spectrum; (**c**) X-ray diffraction pattern showing the presence of metallic copper (Card No. standardreference4-0836).

To investigate the base glass, an integration between EPMA and LA–ICP–MS can work as a suitable compromise. Especially in the last few years, EPMA has been gradually being replaced by LA–ICP–MS for the determination of minor and major oxides, calculated by difference given a known oxide (generally SiO2). Recent research shows, indeed, that close correspondence is generally observed between the data achieved by EPMA and "new generation" LA–ICP–MS equipment when analysing colourless or naturally coloured glasses [31,32]. LA–ICP–MS can perform major to trace element analysis of almost all elements within a sample during a single run, due to specific quantification protocols such as Internal Standard Independent (ISI) and Sum Normalization (SN) methods [33]. Although the potentialities of this technique are significant, its application to the study of deeply coloured opaque glass still needs to be thoroughly explored. In particular, it should be noticed that the most commonly used quantification method (the Sum Normalization) assumes that glass is almost exclusively comprised of oxides in known oxidation states and that the sum of the concentration of all oxides should equal 100% [33]. The first statement is quite difficult to verify when dealing with deeply coloured and opaque glasses such as tesserae, when the addition of several compounds is responsible for the colour shades and the opacity. Recent research also states that EPMA analysis of major and minor elements has several important advantages compared with other techniques, such as LA–ICP–MS as

well as Secondary Ion Mass Spectrometry (SIMS): higher spatial resolution (from one to several μm), a well-established matrix correction procedure, and a lower cost [34]. Given the current state of knowledge, further studies should, therefore, be conducted on the specific application of LA–ICP–MS for the quantification of not only trace, but also major and minor elements in opaque deeply coloured glass before a complete dismission of EPMA for quantitative analysis of major and minor elements can be proposed. When dealing with opaque deeply coloured glasses, EPMA data can, however, be affected by the addition of materials acting as colourants and opacifiers. To better compare the composition of the base glass with the categories reported in the literature for naturally coloured glass, EPMA data should, thus, be adequately processed in order to, if not completely neutralise, at least minimize this effect of contamination. Following the method proposed by Robert Brill [35], reduced composition can be obtained by subtracting the oxides of elements presumably due to additives from the total sum of all those measured and by normalising the remaining data. However, the main concern associated with Brill's method remains the following: when the recalculation is carried out, how can we be sure not to incur arbitrary subtractions? Previously performed in-depth characterisation of colouring and opacifying phases can provide considerable help and a treasurable guide in preventing any subjective subtractions, as we do know what materials are responsible for the hue and opacity of the samples under study and this information can guide us in the recalculation of data.
