3.1. Portable instrumentation
3.1.1. Elemental analysis – portable X-ray Fluorescence (pXRF)
In Table 1 the results of in situ pXRF analyses are reported by color (the term greyish encompasses a variety of dull green to blue shades, present in the leaves and the swirl under the central figure of the vault (Figure 3a) and the festoon, Sphinx and temple on the North wall (Figure 3b)). PXRF was used to obtain qualitative information on the key elements associated with traditional pigments in use during Roman times, mainly based on Fe, Cu, Pb (in bold in Table 1). Some relevant trace elements, such as for example As and Sn, were as well considered for obtaining further information on the manufacturing and sourcing of pigments.
XRF results for all the studied spots (Table 1) show the intense signal of Ca, as expected. The presence of Ba should also be highlighted on the northern wall, especially in the Sphinx (points 7, 12, 13), festoon decoration (points 4 and 5), and on the upper part of the temple (points 6 and 11).
As it concerns key elements, all the studied points show variable quantities of Fe, and, for many of them, also of Pb (Figure 4a). The highest Pb contents in red areas belong to points 22 and 24, on the vault, and to point 20 on the East wall, prosecution of the vault. On the other hand, Pb-rich yellows (12 and 13) and orange (17) are all found on the northern wall. Relatively high Pb is also found in point 10, again on the vault, which is a greyish-blue.
Fe-pigments are yellow to dark red (ochres), and green (glauconite and celadonite, earths). Higher Fe contents are observed in red and yellow areas, with no wall-specific trends.
Cu appears in blue and greyish-blue pigmented areas only (Figure 4b), which could suggest the use of Egyptian blue or azurite.
No evident groupings appear in the Pb vs. Fe and Cu vs. Fe diagrams if Pb and Cu, respectively, are low (Figure 4c and 4d). The dark red points 18 and 19 measured on the column of the temple (North wall) show both low Pb and low Fe: as can be seen from Figure 3b, a thick yellowish-brown patina affects the area and could have hampered the collection of the XRF spectra.
The two blue spots 1 and 2, both on the southern wall, and a greyish-blue element on the vault (point 10) clearly show a considerable amount of Cu (Figure 5a). No Cu is found in the other greyish areas, excluding therefore the use of Cu-based pigments, either the abovementioned blue, or green ones (malachite, copper sulphates, etc.). Copper, moreover, positively correlates with Sn for points 2 and 10, while point 1 shows no Sn, which is possibly linked to different manufacturing techniques (36,37) (Figure 5a). The same two blue points, however, have fairly similar, and relatively high, Cr content, which has been reported for Egyptian blue (30). The use of two different pigment batches for these two points seems suspect, but due to the lack of a more extensive sampling, no further conclusions can be made on the choice of a more or less high-quality Egyptian blue. Cr also correlates with Fe (only point 7 is excluded as Cr is below detection limit, R2 = 0.9264) for the greyish samples (Figure 5b). In these latter, there is also a positive correlation between K and Fe, suggesting the use of green earths, although some fluctuations are evident (R2 = 0.8681, Figure 5c).
As already highlighted by Figure 4, yellows, orange and reds contain either Fe or Pb, or mixtures of the two. The orange point 17 is Pb-based. For the Fe-rich yellows and reds, it appears that the latter have been prepared with an As-rich ochre, whereas yellow is As-free (Figure 6a), similarly to what observed in House of Marcus Lucretius and House of Gilded Cupids in Pompeii (38). Mimetite (Pb5(AsO4)3Cl) and vanadinite (Pb5(VO4)3Cl), containing As and V respectively, have been confirmed in lead ore deposits but also in iron-rich soils (ochres) (39,40).
Pb-rich pigments show specific elemental trends with respect to their color: for example, Sn and Fe positively correlate in the Pb-rich reds, with an opposite trend to what seen in yellow and orange (Figure 6b). Also As and Pb, and Sn and Pb appear positively correlated, with some distinctions (Figure 6c and 6d respectively): in the first case, a correlation appears with no strict link to color (points 13, 12, 22), while in the second, the three shades appear isolated from each other. (41) report Sn as a marker for Spanish red lead obtained as a silver smelting by-product.
3.1.2. Molecular spectroscopy – Raman
The in situ Raman analyses (Table 2) often show calcite (CaCO3), characterized by peaks at 158, 282, 714 and 1086 cm-1 (42) and better visible in the spectrum of a white area (Figure 7e).
Together with the recurring symmetric Edge filter noise, the occurrence of two phenomena further hindered the revelation of the pigments’ Raman spectra: high fluorescence due to the conservation state of the surfaces, and the arising of very intense luminescence bands.
The latter appear in the spectra of greyish, yellow and orange spots acquired on the North wall and on another green area (Figure 7a-d): they seem to characterize earth-based pigments. The bands are centered at about 1430 cm-1, corresponding to ca. 884 nm considering the exciting laser at 785 nm. The 880 nm line could be due to the presence of Nd3+ in calcite, apatite Ca5(PO4)3 anhydrite (CaSO4) or albite (NaAlSi3O8) (43), all compounds possibly present in, or over, the pigmented material. In these spectra (Figure 7b-d) the signature of a calcium oxalate phase (weddellite) is visible in the signals at 188 and 909 cm-1 (44): it could be due to biological activity or to the presence of organic binders (45,46).
The spectrum in Figure 7f was acquired on a red spot and shows, apart calcite signature, a weak signal referable to red ochres (228 cm-1) (47).
3.1.2. Molecular spectroscopy – Diffuse Reflectance Infrared Fourier Transform (DRIFT) spectroscopy
The DRIFT analyses were carried out in situ on large fragments to be reintegrated and provided by the restorers, including white, yellow, red, and black colors. For blue, greyish, dark red, and red of the North wall, the tiny fragments employed for the other laboratory investigations were analyzed. All the DRIFT spectra obtained are shown in Figures 8 and 9 and the wavenumber positions are listed in Table 2.
The spectrum relative to the white area (Figure 8c) clearly shows the features of calcite: 4270 cm-1 (3ν3(CO3)2-), 2515+2590 cm-1 doublet (ν1 + ν3 and/or 2ν2 + ν4 (CO3)2-), 1795 cm-1 (ν1 + ν4 (CO3)2-), 1400 cm-1 ca. strong absorption (ν3 (CO3)2-) inverted by the reststrahlen effect, 880 cm-1 (ν2) distorted in a derivative shape (48–50). The same features are more or less evident in the other DRIFT spectra, especially in that of the black spot (Figure 8a), and are highlighted with a dotted red line. The band at 2400 cm-1 visible in almost all the DRIFT spectra could be attributed to (CO3)2- ν1 + ν3 mode but in relation to cerussite (PbCO3) rather than calcite: thus, even if at lower wavenumber with respect to literature (2410 cm-1 (49)) it may lead to think that lead carbonate is often associated to calcium carbonate. However, the proximity of CO2 signature may distort the spectrum in this region.
In Figure 8e, the blue pigment’s spectrum corresponds to Egyptian Blue: it shows the two typical inverted bands at 1060 and 1006 cm-1 and the shoulder at 1250 cm-1 assigned to a double silicate of cuprorivaite (51). On another blue spot (Figure 8f) the features of gypsum (CaSO4·2H2O) are revealed: 2230 cm-1 (2ν3 (SO4)2-; ν2 + νL H2O), 2120 cm-1 ca. (ν1 + ν3 (SO4)2-) and the reststrahlen inverted band at 1140 cm-1 (ν3 (SO4)2-) (49,50).
The interpretation of the spectrum of the greyish sample (Figure 8d) is challenging because it suggests a mixture of pigments. The presence of kaolinite (Al2(Si2O5)(OH)4), that can be associated to earths, can be inferred by its fundamental Si-O stretching mode included between 1010 and 1100 cm‑1 (49), here at about 1065 cm-1 (52); also, the inverted band at about 1550 cm-1 can be found in all the spectra related to red color in Figure 9.
The DRIFT spectra acquired on the yellow spots can invariably be represented by that in Figure 8b. The features are mostly those of calcite to whose spectrum they are almost superimposable. Nevertheless, the kaolin characteristic features at 4524 (combination band ν + d OH) and 3690 cm-1 (OH stretching) can suggest the usage of yellow ochre (49).
The spectra of three different red spots, the one analyzed in situ (Figure 9c), red from the North wall (Figure 9b) and dark red (Figure 9a) have all an ochre basis: spectra a) and b) in Figure 9 show the typical shoulder of pure hematite at 1109 cm-1 (51), spectrum c) displays the already mentioned kaolin band at 4525 cm-1 (49), and all the three of them exhibit an inverted band ranging from 1030 to 1065 that can be attributed to the silicate matrix. The spectra of the red areas too show the peak at 2400 cm-1 probably attributable to cerussite; in addition, especially pronounced in spectrum b, 1740 and inverted 1415 cm-1 bands are also visible: these could be respectively assigned to carbonate ion ν1 + ν4 and ν3 modes typical of cerussite or hydrocerussite (Pb3(CO3)2(OH)2).
In almost all the spectra, but more evident in Figures 8a (blue), 8b (yellow) and 9b (red from the North wall) signals are present in the regions of the CH stretching, around 2900 cm-1 and in the amide I and amide II ones, respectively at about 1690 and 1570 cm-1 (49,50,53), leading to infer the presence of a proteinaceous material.
3.2. Laboratory instrumentation
3.2.1. Molecular spectroscopy – micro-Raman
The laboratory Raman investigations were carried out both with the 532 and with the 785 nm lasers: as expected, the former gave the best results for white, black and blue samples, the latter for the yellow, red and earth-tones (see Table 2 for all the results).
On the white sample, apart from calcite signature, which was found also in other colors spectra (see Figure 11a, c), the presence of gypsum and cerussite was revealed by their principal signals at 1006 and 1050 cm-1 respectively (Figure 10b) (42).
The Raman analysis performed on the blue sample allowed to obtain a spectrum only when focusing on a blue particle (Figure 10a’): the result is compatible with Egyptian Blue associated to gypsum (Figure 10a) (42).
Raman spectroscopy allowed the identification of a carbon-based pigment with the two typical bands D and G at about 1375 and 1587 cm-1 respectively. Comparing the obtained spectrum with those of reference black pigments, a similarity was highlighted with bituminous materials such as asphaltum, Van Dyck Brown and Cassel Earth (54).
The red pigment analyzed (Figure 11a) confirmed the presence of the principal signals of hematite (a-Fe2O3, main constituent of the red ochres) at 292, 405, 613 cm-1 (47). Once again, a very intense luminescence band appeared in relation to earth-based pigments. It is centered at 1370 cm-1, corresponding to ca. 880 nm, therefore the same consideration made in paragraph 3.1.2 can be done.
The dark red sample spectrum in Figure 11c, instead, is differentiated for the presence of a peak at 674 cm‑1. This band is sometimes associated to hematite in peculiar conditions, for example it is related to structural disorder as a consequence of Al-for-Fe substitution (55) or due to hematite recrystallization with high temperature (56). Notwithstanding, its presence as the only main peak apart those related to calcite, make magnetite (Fe3O4) the most probable attribution (47,57), which would account for the darker color of the area.
The Raman spectrum of the yellow sample displays peaks at 250, 300, 390 and 558 cm-1 (Figure 11b), attributed to the mineral goethite (a-FeOOH), one of the principal constituents of yellow ochres (47). Both for goethite and for hematite, the main broad band centered at about 1310 and 1300 cm‑1 respectively must have been hidden by the very intense luminescence signal.
3.2.2. Molecular spectroscopy – Fourier Transform Infrared Attenuated Total Reflectance (FTIR-ATR)
In all the FTIR-ATR acquired spectra (Figures 12, 13, Table 2) the strong signals related to calcite are evident: it is clear that even though scraping a very small quantity of surface pigment, the carbonate substrate cannot be avoided and is present in the analyzed powder. Its signals are highlighted in Figure 12 with red dotted lines and are positioned at 712 (n4 (CO3)2-), 844, 871 (n2 (CO3)2-), 1085 (n1 (CO3)2-) and 1397 (n3 (CO3)2-) cm-1 (58,59) and mainly characterize the spectrum of the white sample (Figure 12b). A broad band between 900 and 1200 cm-1 is also visible in almost all the spectra, and it appears much lower in the carbon-based (see 3.2.1) black one (Figure 12a). This band could be attributable to the presence of silicates, which are well-known to absorb in this region: they could be part of the pigments or be traces of dust/soil withstanding the delicate brush cleaning of the surface of the samples.
The ATR analyses confirm that the blue samples are constituted by Egyptian Blue (Figure 12h), a silicate compound displaying bands in the region of asymmetrical Si-O-Si stretching, between 1000 and 1280 cm-1 (60). Its peaks (Table 2, (61)) overlap with those of gypsum (Table 2 (62)) in the spectra of Figures 12i and 13i. Furthermore, quartz is present, as its typical infrared absorbance doublet is visible in both spectra at 776 and 798 cm‑1, together with the peak at 470 cm-1 (63).
Due to the extremely small size of the greyish sample (Figure 12g), the presence of green earths cannot be confirmed, as their principal peak centered between 950 and 970 cm-1 (64) could not be detected. Nevertheless, kaolinite, which is present in ochres and earths, is recognizable by the bands at 437 and 530 cm-1 (65,66), even if the spectrum ‑acquired on a too small quantity of powder- is too noisy to discern the typical signals in the high wavenumber region (Figure 13g). The signals at 668, 1005 and 1165 cm-1 could suggest the addition of a small quantity of Egyptian Blue. Also, the presence of quartz is ascertained due to the presence of the above-mentioned doublet.
In the yellow (Figg. 12f, 13f) and red (Figg. 12d, 13d) samples, the spectra of ochres can be clearly recognized: the distinction between red and yellow ones could be tricky since their infrared spectra contain many similarities showing the same absorbance peaks due to kaolinite presence: outer hydroxyl ions around 3690, 3665 and 3650 cm-1 as well as inner hydroxyl ions at 3620 cm-1 (65) are present both in the yellow and in the red sample spectra (Figure 13f and e, respectively). Kaolinite also displays peaks at 479 (Si-O), 536 (Si‑O‑Al), 938 and 914 (Al-O-H) 1009 (Si-O-Al) and 1032 (Si-O-Si) cm-1 (65,66). At the same time, the peaks of pure ferric oxide can be found between 400 and 750 cm-1 (65).
The dark red sample, whose spectrum is shown in Figures 12c and 13c, only exhibits two broad bands at 466 and 565 cm-1, the latter feature is consistent with the presence of magnetite (67).
The spectra in Figures 12d and 12e, instead, were acquired on the same red fragment coming from the North wall. Apart the already-mentioned ochre-based composition, both spectra show signals around 606 and 630-640, 984, 1070 and 1163 cm-1, typical of a sulphate phase, in particular the positions fit best with baryte (Ba2SO4) (68). Spectrum e) exhibits also an absorption band at about 680 cm-1, that, according to (69) is characteristic of lead white and helps its distinction from chalk. This could have been present in the mixture of pigments, or, according to (70), lead carboxylates and lead carbonate could have been formed on red lead based paints due to the interaction with an egg-based binding medium. Also, lead carbonate appears to be the most stable phase of Pb-pigments degradation in atmospheric conditions (71).
The band at about 1650 cm-1, present also in the white and blue ones, falls in the amide I region of proteins (70,72,73), on the other hand, the amide II region lacks its band at around 1540 cm-1, which could be included in the broadened and asymmetric band of carbonates (compare for example Figg. 12a and 12b).
Looking at the high wavenumber region, indeed, the above-mentioned red (Figure 13d, e) and white (Figure 13b) samples’ infrared spectra show a broadened calcite-related signal at 2870 cm-1, probably including a shoulder at 2850 cm-1, and peaks around 2920 and 3290 cm-1. These signals are referred to the CH stretching of proteins (73) and are typical of egg white (70). There is no trace, instead, of lipid-connected (70) or beeswax infrared absorbance bands (74).