This work relied on close visual stereomicroscopy of the surface of the painting and air-path x-ray fluorescence analysis (XRF) analysis. A small number of cross section samples and paint scrapings were obtained for more detailed characterization of the pigments. All the samples described here were taken from the bottom edge of the panel at or close to the site in the Madonna’s robe indicated by a black dot in Fig. 1. The samples were analyzed using scanning electron microscopy with energy dispersive x-ray analysis (SEM-EDX), polarized light microscopy (PLM), microRaman spectroscopy and electron backscatter diffraction (EBSD). EBSD was crucial to identificaton of the phases. Electrons can be diffracted by atoms vibrating in lattice planes of crystalline compounds. These diffracted electrons form cones which due to the low angle of scattering appear as lines on a detector. The lines from different planes of atoms form Kikuchi patterns which can be related to crystal structures without need for a specific orientation of the crystals. The Kikuchi patterns can be compared to patterns from standard and vouchered samples. Pattern formation is a result of electron diffraction at the sample surface down to 20–30 nm. For pattern acquisition, a sample is tilted at high angle, typically at 70°, reducing spatial resolution of the analysis, however excellent patterns can be obtained from crystallites as small as 10 nm, therefore EBSD can be very useful for identification of pigment particles.
The cross-section samples were extraordinarily sensitive to water used during the initial grinding and polishing process of preparing the cross sections—much more than typical, perhaps due to the ground having a low concentration of gypsum in glue. This sensitivity meant that the surface could not be polished flat. Argon ion milling was used to try to smooth the surface, but this was not successful. The difficulty of the sample work up meant that the analyses had to be conducted on rough surfaces, nevertheless we were able to obtain interesting results and novel discoveries.
From surface examination of the painting the lining of the Virgin’s mantle looks as if it had been painted using a glaze of azurite over a layer of yellow paint. However, cross sections show that in the painting there are passages where the yellow is on top of the blue and others where the colors are mixed (Fig. 2).
XRF confirmed the general palette of inorganic pigments by inference from elemental compositon as lead white, yellow iron earths, green earth, azurite, lead tin yellow, and vermilion. Spectra from the blue drapery contained peaks for lead, mercury, and a very small peak for the Kβ line of arsenic. This was tentatively thought to be due to orpiment, a pigment that has been found in Giotto’s works, but a close surface examination did not reveal any of the typical yellow tablet-shaped particles of orpiment.
Lead tin yellow type II
Two yellow pigments were found: iron earth and lead tin yellow type II, as expected for a fourteenth century painting. Lead–tin yellow type II is a term used to cover a range of pigments that are opaque yellow glasses or glassy materials [3]. The formula is given as PbSn1-xSixO3 where x ≥ 0. The crystallographic group of the endmember PbSnO3 is cubic. The backcatter electon (BSE) image of a single particle at higher magnification is shown in Fig. 3a. This image and EDX spectra obtained at diffierent spots within the pigment particle (Fig. 3b) show that the partcle is composed of a Ca–Pb-Si matrix holding densely packed crystallites of lead tin oxide, lead silicate and lead tin silicate—the grey levels giving a clue to the diversity of the composition of the particles. The Kikuchi pattern from EBSD of one particle of the predominant light grey level matches the pattern of PbSnO3 [4]. A slightly darker phase contains silicon, calcium and lead, but no tin. None of the phases contain iron, zinc or potassium—elements that have been found in the pigment used by later artists. This indicates that a simple recipe was used for the production of this colorant.
Azurite
The blue paint used for the Virgin’s mantle is not the deep blue of lapis lazuli or ultramarine, rather it is colored predominantly by azurite. Transmitted light microscopy of a small scraping of the paint from under the brown trim on the mantle shows that the particle size range of the pigment used in this painting is wide: while most particles are small, some are 35 µm long, giving the paint a saturated, intense color. Only a small proportion of white pigments was added to the blue paint. The EBSD pattern confirmed the identity of the blue mineral (Fig. 4a, b). The EDX spectrum (Fig. 4c) of the particle labelled 1 in Fig. 2a shows that the azurite used in this work is very pure, and does not contain any detectable transition element impurities or substitutions, such as zinc or manganese, which are sometimes found in azurite or malachite [5, 6]. The cross sections do not contain red particles which are often observed in azurite paint, which have been shown to be naturally occuring iron oxide impurities [7].
In addition to azurite, there are some smaller green and green–blue particles in the paint. Some of these contain only copper, determined using SEM-EDX, and have a grey level in the BSE image (Fig. 5a) very similar to that of the azurite. These are likely malachite, which is often geologically associated with azurite and no further analysis was undertaken. Other particles, as already noted, have a more complex compostition and are discussed next.
Mixite
The most interesting finding was the identification of a Cu-Bi arsenate mineral mixed into the paint. Several smaller green–blue particles in the cross sections and in dispersed samples from the blue drapery are brighter than azurite in the BSE image which implies they have a higher average atomic number. Two green–blue particles in the center of the section in Fig. 2A have a crystal habit that is approximately columnar hexagonal. These particles can be seen to extend lengthwise into the the sample. They are small with a diameter of c. 8 µm and an observable length of c. 20 µm. Their appearance in BSE is very similar to that published of one sample of the mineral mixite which has bundles of acicular crystals [8]. A map of the distribution of the elements copper, bismith, arsenic and calcium is shown in Fig. 5b. Bismuth and arsenic are located in these particles and qualitative elemental analysis indicates they are ternary metal oxides of copper, arsenic and bismuth. A representative spectrum of the particles labeled 2 in Fig. 2 is presented in Fig. 5c. The green–blue particles contain varying proportions of bismuth, higher in some and not detectable in others; however, the sample was too rough for reliable quantification of the phases [9] therefore, microRaman and EBSD were employed for further characterization.
The microRaman spectrum obtained from one of the green–blue particles at the top right edge of the sample illustrated in Fig. 2B is shown in Fig. 6. The sloping background is due to fluorescence from the mineral, a common problem for green/blue compounds which can affect the ability to measure Raman emission. For this sample, peaks occur at 853, 825, 472, 457(sh), and 420 cm−1. These can be assigned to the symmetric (υs) and asymmetric (υas) stretches and bending modes of the orthoarsenate group (AsO4
3−) and hydroxyorthoarsenates (HAsO4
2− and H2AsO4
−) in a distorted tetrahedral symmetry, while the absence of bands at 867–870 cm−1 may indicate the absence of H2AsO4
− moeity [10, 11]. These Raman shifts are similar to those measured by Frost et al. for two samples of mixite–one of which from the Boss Tweed mine in Montana has υ2 851.8 and 830.8 cm−1 and υ4 (bending mode) 475.4 and 473.7 cm−1, values that are comparable to the mineral in Giotto’s paint [8].
Since the Raman spectrum obtained from the green–blue pigment in the cross section could be due to a number of arsenate-containing minerals [12], we obtained an EBSD pattern to confirm the identification. The rough surface of the cross sections (Fig. 7) was the first challenge to acquiring good diffraction patterns; additionally some damage to the cross section due to heating by laser radiation during Raman spectroscopy had occurred on the top surface of the particles. However, a diffraction pattern was obtained from the rounded side of one of the particles (Fig. 8). The pattern matches the phase BiCu6(OH)6(AsO4)3(H2O)3 which has hexagonal-low (6/m) symmetry, consistent with the apparent crystal habit of the particles observable in the cross sections (see Fig. 5a) [13]. The mixite solution overlaid in Fig. 8b corresponds extremely well to the collected pattern. Furthermore, since no other Bi-Cu-As-O crystal structures in the ICSD database matches the experimental pattern, we are confident in the identification of the particles in Giotto’s paint as the mineral mixite.