Raman Spectroscopic Analysis of Ancient Egyptian Pigments拉曼光谱研究古代埃及颜料

RAMAN SPECTROSCOPIC ANALYSIS OF ANCIENT EGYPTIAN PIGMENTS* A. ROSALIE DAVID Keeper of Egyptology, The Manchester Museum, University of Manchester, Oxford Road, Manchester M13 9PL, UK H. G. M. EDWARDS† and D. W. FARWELL Department of Chemical and Forensic Sciences, University of Bradford, Bradford BD7 1DP, UK and D. L. A. DE FARIA Instituto de Quimica, Universidade de Sao Paulo, Av. Prof. Lineu Prestes 748, CP 26077, 05513-970 Sao Paulo SP, Brazil The application of FT-Raman spectroscopy and visible Raman microscopy to the non- destructive analysis of pigment specimens excavated from Tell el Amarna by Flinders Petrie in the 1890s has provided information about the chemical composition of the materials used by XVIIIth Dynasty artists in the New Kingdom at the time of King Akhenaten, c. 1340 BC. Comparison of the Raman spectra of the samples labelled ‘red and yellow ochre’ with documented, archival material from geological collections provided a clear indication of the materials used in the iron(III) oxide/hydroxide system, including a-hematite, goethite, maghemite, magnetite and lepidocrocite. The yellow–orange specimen labelled ‘realgar’ proved to be a mixture of realgar and pararealgar; since the specimen had been sheltered from light since its excavation, this could indicate that the ancient Egyptian artists recognized the colour variation and may have used this to effect in their decorations. A specimen of yellow ochre contained goethite, a-FeO.OH, with particles of crystalline, highly ordered graphite; in contrast, the red ochre specimens contained amorphous carbon particles. KEYWORDS: EGYPT, MIDDLE KINGDOM, TELL EL AMARNA, PIGMENTS, RAMAN SPECTROSCOPY, NON-DESTRUCTIVE ANALYSIS, REALGAR, OCHRES INTRODUCTION Raman spectroscopy has been applied with demonstrable success to the study of pigments in historiated ancient manuscripts (Best et al. 1992, 1993, 1995; Clark et al. 1995, 1997, 1998; Burgio et al. 1997a), medieval cantorals (Edwards et al. 1999c), polychrome statuary (Edwards et al. 2000) and wall-paintings (Coupry et al. 1994; Russ et al. 1995; Edwards et al. 1997, 1998, 1999a–c; Edwards 1998; Edwards and Rull Perez 1999; Smith et al. 1999). A wealth of information about ancient technology used in the preparation of pigments and their subsequent interactions with substratal material has emerged from these studies, including mineral instability due to the presence of neighbouring pigments and changes in their molecular structures with time. A major advantage of Raman spectroscopic techniques over infrared absorption is the simplicity of presentation of complex artefacts to the spectrometer in the Archaeometry 43, 4 (2001) 461–473. Printed in Great Britain * Received 26 October 2000; accepted 8 May 2001. † To whom correspondence should be addressed. q University of Oxford, 2001 former, which requires little or no sample preparation. The wavenumber coverage of the Raman spectra, typically 50–3500 cm–1 in one instrument, also provides a unique means of analysis for inorganic minerals (especially heavy metal salts, oxides or sulphides) and organic dyes and binders (Vandenabeele et al. 2000), which may occur together in works of art. Although organic dyes and colorants were recognized as ‘fugitive’ by ancient artists, nevertheless they are still found in manuscripts, especially those that have been preserved in libraries. The situation is different in the case of prehistoric rock art, since the colour palette is strictly limited (Russ et al. 1995; Smith et al. 1999) to naturally occurring metal oxides and charcoal or bone white (calcined ivory or bone). However, the exposure of these paintings to environmental climatic changes such as temperature and humidity, and colonization by algae, fungi and lichens, creates new problems for conservators, which Raman spectroscopy is addressing (Seaward and Edwards 1995, 1997; Edwards and Rull Perez 1999). Tomb paintings, wall-paintings and frescoes in public buildings that are visited by members of the public provide other biodeteriorative scenarios for archaeologists, art historians and conservators. Although colonization by biologically active organisms was thought to occur slowly over long periods of time, recent reports indicate that destructive invasion of paintings and their substratal rock or plaster is now proceeding at alarming rates, often fuelled by hydrocarbon emissions and pollutants in urban environments. Our earlier Raman studies (Seaward and Edwards 1995) on the colonization of the Zuccari 16th-century frescoes in the Palazzo Farnese in Caprarola, Italy, by Dirina massiliensis forma sorediata, which was first recognized in 1986 and is now responsible for the permanent disfigurement of over 80% of the painted surfaces, have indicated the complexity of the biodeteriorative process. Molecular spectral identification of key biomarkers such as whewellite, weddelite, carotenoids, polyphe- nolic acids and lichen pigments has assisted in the recognition of biodeteriorative erosion of frescoes and wall-paintings elsewhere, so providing an early warning device for curators of these works of art. Although the ancient Egyptian artists clearly used an extensive palette, as shown by tomb- paintings, papyri, decorative ceramics and coloured funerary artefacts, there have been few Raman spectroscopic studies. Clark and Gibbs (1997) have reported a non-destructive Raman study of coloured ancient Egyptian faience fragments from Amarna (in the Petrie Collection at University College London), and Coupry et al. (1994) and Coupry (1998) have reported Raman spectroscopic studies of blue pigments from the Egyptian New Kingdom. In the present study, we have taken the opportunity to analyse non-destructively, for the first time, using Raman spectroscopy the pigments excavated from Tell el Amarna by Sir William Flinders Petrie in 1891–2, which are now housed in the Manchester Museum. EXPERIMENTAL Samples The pigment samples studied in the present work were excavated by Sir William Flinders Petrie at Tell el Amarna in Middle Egypt in 1891–2 (Petrie 1894). The site consists of the remains of a royal city which was constructed, occupied and then deserted within the very short time of 20 years or so during the reign of Amenhotep IV (Akhenaten), c. 1350–1334 BC, in the XVIIIth Dynasty. Akhenaten is now considered as a pioneer of a beautiful culture, whose art forms were quite distinct from those that had preceded it. The ‘Amarna culture’, as it became known, came into conflict with the orthodox priesthood, who saw it as a heresy and supervised 462 A. R. David et al. the destruction of the city on the pharaoh’s death in 1334 BC (Sanson 1972; Phillips 1998). Samples from the Amarna cultural period excavations can hence be realistically dated to a very narrow period of occupation. In the present study, some coloured pigments from Petrie’s excavations at Amarna (Drawer 1985) were deposited in the Manchester Museum of Egyptology in 1896 and have not been examined since. Their rarity and special significance therefore lend themselves to non-destructive Raman spectroscopic analytical examination for the first time. Details of the specimens presented for Raman spectroscopic analysis are given in Table 1; the samples are believed to be mineral pigments of colours yellow, red, red–brown, green, orange, orange–yellow, blue and yellow–brown, from the remains of artists’ palettes used in decoration of this vibrant artistic period of Dynastic Egyptian history. In some cases, the samples were found adhering to fragments of stone mortars, which presumably were used to produce the fine pigment powders required for the art work. The Raman spectroscopic technique did not necessitate the detachment of the minerals from their associated artefacts; nor did it involve the pulverizing or grinding of archival material. The mineralogical descriptions of the specimens in Table 1 are based only upon the initial examination prior to cataloguing for the Museum collection; it is stressed that chemical analysis has not been carried out on these specimens hitherto. Raman spectroscopy Raman spectroscopy is a vibrational spectroscopic technique that depends on the analysis of the wavenumbers of scattered laser radiation. The pattern of the scattered radiation wavenumbers is characteristic of the molecular composition of the material being studied, and in this respect the Raman spectrum is complementary to the infrared. Since infrared and Raman spectros- copy arise from different physical processes, however, the information acquired from Raman spectra is not identical to that obtainable from the infrared spectra in band appearance, intensity and often wavenumbers. The application of near-infrared laser excitation for the recording of Raman spectra provides a useful means of minimizing the fluorescence that can occur along with the Raman spectra of organic materials, and which effectively swamps the weaker Raman signals. A particular advantage of Raman spectroscopy in the characterization of pigments is the wealth of information about geological and synthetic minerals and biological materials which can be gathered from the same spectrum over the extended wavenumber range offered by the Raman technique in comparison with the infrared. This has a special application for the non-destructive analysis of sensitive archaeological materials and artefacts. In addition, the 463Raman spectroscopic analysis of ancient Egyptian pigments Table 1 Samples from Tell el Amarna studied in the present work Sample code Colour Classification Number of specimens studied Uncat NCR Ore Yellow Yellow ochre 1 NCR Ore Red–brown – 2 – Green Malachite (?) 1 7458 Orange Red ochre 1 5664 Orange–yellow Realgar 1 2478 Blue Frit 1 1964 Blue Fragment 1 1129 Yellow–brown Ochre (glass fruit colour) 1 weak response of the Raman technique to water means that desiccation of specimens need not be undertaken which, in conjunction with the versatile sampling arrangements that require no mechanical or chemical pre-treatment of specimen surfaces, means that there are a host of applications in the field of archaeological science. FT-Raman spectroscopy Fourier-transform Raman spectra were obtained using a Bruker IFS 66 infrared spectrometer with an FRA 106 Raman module attachment and dedicated Raman microscope. Excitation was effected using 1064 nm radiation from a Nd3‡/YAG laser operating with a maximum power of 750 mW, although laser powers at the sample of about 100–200 mW were typical of those used here. Spectra were recorded at 4 cm–1 spectral resolution over the wavenumber range, 100–3500 cm–1, to check for any organic components, such as resins, which would be indicated by n(CH) scattering near 3000 cm–1 and d(CH2) modes near 1400 cm –1. Co-accumulation of spectral scans was undertaken to provide improvement in signal-to-noise ratios; typically, about 100–200 scans were made, representing a total accumulation time of , 3 min for each specimen, to provide good-quality Raman spectra. Wavenumbers are correct to better than 61 cm–1 for sharp bands. For macroscopic sampling, each specimen was mounted directly and vertically in the sample illuminator; fragile samples could be studied using a 908 illuminator attachment and horizontal mounting. In each case, the sample ‘footprint’ studied was about 100 mm diameter. For microscopic studies, samples were held horizontally on a microscope stage, and a ‘footprint’ of about 8 mm was obtained using a 100 · objective lens. Visible Raman microscopy The visible Raman measurements were performed using a Renishaw microscope system fitted with a Peltier-cooled CCD detector (Wright, 600 · 400 pixels) and with an Olympus metallurgical microscope. The spectra were excited with the 632.8 nm line of a He–Ne laser (Spectra Physics, model 127) which was focused on a c. 1–2 mm (micrometre) spot, with an average spectral resolution of about 2–4 cm–1 over the wavenumber range of 100–1800 cm–1. Laser powers were kept as low as possible (c. 0.7 mW) to avoid sample degradation. Co-accumulation of spectral scans over 5–50 s was effected to provide good spectral signal- to-noise ratios. It is important to note that in all cases no prior mechanical or chemical treatment of the specimens was necessary; hence, the mineral pigments could be returned to their museum archive in exactly the same condition as they left it. RESULTS AND DISCUSSION For the purpose of discussion, the results from the red, orange and yellow pigments and from the blue and green samples (shown in Table 1) will be presented separately. Red, orange and yellow samples These samples had been archived (Table 1) as ochres (uncat NCR,7458), realgar (5664) or had no identification at all (NCR Ore, 1129). As part of our Raman spectroscopic studies designed to provide a database of pigments used in Antiquity, we have included work on the iron(III) oxide/hydroxide system that forms the basis of ochre pigment coloration (de Faria et al. 1997; 464 A. R. David et al. Clark and Curri 1998; Bersani et al. 1999). In prehistoric cave-paintings, the limited palette available to the artists usually consisted of red, yellow, white and black colours, which are normally comprised of red and yellow ochres, gypsum or calcite, carbon or pyrolusite and manganese(IV) oxide (Coupry et al. 1994; Russ et al. 1995; Smith et al. 1999). The colours of the ochres are of particular relevance in the current work, since admixture of raw pigments with sands or clays and heating to various temperatures was a technology that was available even in prehistoric times. Some of the compounds in the iron(III) oxide/hydroxide system of relevance to the present study include minerals such as hematite (a-Fe2O3), goethite (a-FeO.OH), maghemite (g-Fe2O3), magnetite (Fe3O4) and lepidocrocite (g-FeO.OH). In the construction of our Raman spectro- scopic database, mineral samples from the geological collections of the National History Museum, London, were analysed non-destructively. Key papers on the Raman spectra of iron oxides and oxyhydroxides have recently appeared in the literature (de Faria et al. 1997; Smith et al. 1999; Bikiaris et al. 2000). Much careful interpretation of the pigments in this system has been undertaken with regard to their mineralogical classifications. The visible Raman spectra of some of these system components using 632.8 nm excitation are illustrated in the stack plot in Figure 1, and the key wavenumbers that can be used to characterize these materials are given in Table 2. Having established the database for these minerals, it is appropriate to apply the analysis to the Amarna ochre samples, bearing in mind that small wavenumber shifts due to environmental effects can be expected between the pure mineral and the same mineral found in admixture with clays and sand. Raman spectra of the ochre samples were obtained using 633 nm and 1064 nm excitation, but the spectra obtained at 633 nm were stronger. A stack plot of the FT-Raman spectra of the Amarna ochre specimens is given in Figure 2. Comparison of the Raman band wavenumbers of the Amarna ochres in Figure 2 (Table 3) with the iron(III) oxide/hydroxide standards in Figure 1 (Table 2) provides the following conclusions: (i) Sample 7458: red ochre. This can clearly be assigned to a-hematite, but the presence of the features at 462 cm–1 and 252 cm–1 in some spectra (not shown here) is indicative of a-quartz, probably from admixture with sand. Sand was normally used as an aid to fine particle production through grinding (Best et al. 1995). (ii) NCR Ore: red–brown ochre. This may also be assigned to a-hematite but, unlike sample 465Raman spectroscopic analysis of ancient Egyptian pigments Figure 1 An FT-Raman spectral stack plot of minerals in the iron(III) oxide/hydroxide system: a, goethite; b, lepidocrocite; c, hematite; d, maghemite; e, magnetite. 633 nm excitation, wavenumber range 150–1700 cm–1, 30 s spectral accumulation, 2 cm–1 spectral resolution. 7458, there is no indication of a-quartz present in any of the specimen regions sampled. However, at larger wavenumber shifts this sample gave bands at 1043, 1280, 1330, 1453, 1669, 2937, 2978 and 3016 cm–1. These are assignable to an organic resin that contains CO, alphatic CH2 and CˆC groups. A possible candidate for this is a terpene such as abietic acid or resinous gums such as colophony, dammar and pinene. It is suggested that this specimen has possibly been prepared for some special use—perhaps involving decoration of a sarcophagus or wooden item, where a resin binding medium would assist in the adhesion of the pigment to the substrate. Alternatively, the admixture of resin and hematite produces a brown pigment, which could have been the colour desired by the ancient artists. Small black particles in the specimen were identified as carbon from characteristic bands in the Raman spectrum. (iii) Uncat NCR Ore: yellow ochre. This is very inhomogeneous. Two characteristic spectra were obtained from different regions of the specimen; one, of distinctly red particulates, matches an assignment to a-hematite, but the other contains a large quantity of sand, as shown by the Raman bands at 252 and 464 cm–1, and has a yellow colour that fits better with goethite, a hydrated iron(III) oxide/hydroxide. On heating, goethite produces hematite (de Faria et al. 1997; Bersani et al. 1999). Hence, the Uncat NCR Ore (yellow ochre) specimen is very different from the similarly catalogued NCR Ore (red–brown ochre) specimen described above and they were probably designed for different artistic purposes, in that the Uncat NCR Ore (yellow ochre) has been subjected to a simple preparation technology involving heating, whereas the NCR Ore (red–brown ochre) involves mixing with a resin binder, which may have involved a heating or melting process. (iv) In both the NCR Ore (red–brown ochre) and Uncat NCR Ore (yellow ochre) specimens, small black particles were identified as carbon from the Raman spectra recorded. In Figure 3, a Raman spectrum typical of the NCR Ore (red–brown ochre) is seen to be amorphous carbon with the characteristic twin peaks at 1580 and 1340 cm–1, whereas the carbon particles in the 466 A. R. David et al. Figure 2 An FT-Raman spectral stack plot of archival ochre pigment specimens from the Amarna excavations: a, yellow-brown ochre (1129); b, yellow ochre (Uncat NCR Ore); c, red ochre (7458); d, red ochre (Uncat NCR Ore); e, red-brown ochre (NCR Ore). 1064 nm excitation, Wavenumber range 200–800 cm–1, 4 cm–1 spectral resolution, 1000 scans spectral accumulation. Uncat NCR Ore (yellow ochre) specimen exhibit just the 1580 cm–1 band, which is also much narrower in spectral bandwidth than the more typical carbon spectra. This result, which indicates that the carbon particles in the Uncat NCR Ore (yellow ochre) arise from a highly crystalline, ordered graphite, whereas those in the NCR Ore (red–brown ochre) specimen arise from an amorphous carbon, indicates that the two specimens probably originate from different sources. Alternatively, there is a possibility that the NCR Ore (red–brown ochre) red pigment has been made from the yellow specimen. It is clear that the yellow goethite in the Uncat NCR Ore (yellow ochre) has not been subjected to thermal or mechanical processing, which would have destroyed the highly oriented graphitic carbon particles. Hence the presence of red hematite particles within this specimen can be attributed to natural geological contamination and not to residual processing. (v) Sample 1129: yellow–brown ochre. This is believed in the cataloguing to be the origin of ‘glass fruit’ colours found in the Amarna excavations. The Raman spectrum of t