Physiochemical Characterisation of Pigments from Theran Wall Paintings

The goals of the study were the identification of the exact nature of the pigments, especially blue and green, and the technical features of paint application (binding medium, thickness, adherence to plaster) which would provide evidence about the painting technique (fresco or secco). The results showed that all the pigments, except Egyptian Blue, were naturally occurring, while the predominant blue ones were amphiboles (mainly riebeckite). It seems, finally, more likely that fresco was used for the majority of the cases, while secco is also apparent.

INTRODUCTION

Approximately 130 years after the first excavations in Thera, conducted by Fouqué, Alafouzos and Nomikos, and 30 years after the beginning of systematic excavations at Akrotiri, a number of questions concerning the techniques employed in the construction of the wall paintings still remain open. There have been many multidisciplinary studies conducted during the past years, concerning all aspects related to the Akrotiri wall paintings, including examination of raw materials and painting techniques.

Detailed study of the wall paintings over the years has revealed a plethora of artistic and technical achievements of the ancient artists and obviously there is still more to come, since systematic and coordinated work has only recently started.

The present paper deals mainly with technical aspects of the wall paintings, summarising already existing knowledge and adding more information on the nature and provenance of pigments and raw marerials, as well as the painting techniques employed.

Most of the publications on the various technological aspects of the Theran wall paintings appeared in the second half of the 1970s. The first paper ever published was by Noll et al. (1975), who argued that the buon fresco technique (with lime as binder) was used and no organic binders were detectable in the pigments. Another interesting point was the non-existence of green pigments. A year later the first systematic work on the nature of the blue pigments appeared (Philippakis et al. 1976), identifying three types of pigment: a) a group consisting of Egyptian Blue, with the predominant element being copper. Other elements detected were tin, lead and arsenic, and it was suggested that they might be related to the nature of Egyptian Blue; in other words, that the source of the copper may have bronze filings, slag from a bronze-smelting furnace or bronze corrosion products; b) a second type consisting of glaucophane, with the prominent element being iron and complete absence of copper. Nickel should be associated with the presence of glaucophane; c) a combination of types a and b, with iron being the predominant element and the presence of various amounts of copper. An interesting conclusion was that the glaucophane-rich pigment had so far been detected only in samples from Akrotiri and Knossos, and the authors hypothesised that the technology and use of this material as a pigment was confined primarily to the artists of Thera, and thus forgotten or lost after the eruption of the volcano.

The work of Profi et al. (1977) added information on the red and related colours: the red, pink, orange and yellow pigments were identified as ochres containing hematite and/or iron hydroxides (goethite and limonite); kaolinite was also detected. As far as brown was concerned, two types were detected: the first one was identified as ochre, like the previous pigments; the second one was composed of a layer of ochre over a layer of Egyptian Blue.

During the Congress Thera and the Aegean World II, two papers on the technology of wall paintings were presented. One by Philippakis (1978) with results similar to the ones presented above, and a second by Asimenos (1978) with observations on the painting techniques. According to him, the painting method was basically at secco, with some cases of mixed technique. The criteria for this hypothesis were determined as follows: 1) on to a primer, in the form of a thick coating of lime plaster, a second thinner layer of finer plaster was applied; 2) the preliminary sketches were done in paint or by incision (in the latter case, no smooth cuts were observed under the microscope, suggesting that the plaster was already dry); 3) parts of many compositions are over-painted; 4) many of the pigments, especially the blue, are flaking, an indication that the adhesion of the pigment to the wall was destroyed when the bonding medium decayed. Televantou (1994) came to the same conclusions about the technique of the wall paintings of Akrotiri and especially of the West House. It was suggested that the secco method and not the fresco was the painting technique employed, based on the following observations: a) the lack of pigment penetration in the plaster; b) the uniform layer of the plaster applied on the wall, and the string lines which have left their imprint on it; c) the incision for the planning of the composition; d) various details painted over other, larger single-colour surfaces.

Past physicochemical analyses and technological observations concerning the plasters and the pigments of the Akrotiri paintings have given valuable but fragmented information about the nature of the raw materials and the technique of the wall paintings. However there is no consensus about the painting technique employed, that is whether it was at fresco (true/buon fresco) or at secco.

As already mentioned, this paper is an attempt, using recent methods of analysis, to elucidate the nature of the pigments, the form in which they were applied to the plaster and the actual painting procedure. We inevitably focus our attention on the nature of one pigment which was difficult to prepare and to apply: the blue. Finally there is another important question that deserves an answer: why is there no record of green colour in the Akrotiri wall paintings or in most of the contemporary frescoes from other Aegean sites?

SAMPLES

All of the samples come from wall paintings belonging to the final phase of the site. The majority come from one building, Xeste 3 (basically rooms 3 and 15, but also from the staircase and room 5, as well as from the area between Xeste 3 and the House of the Benches).

These particular samples were chosen in order to investigate any possible differences in the pigments and their application techniques between rooms of the same building.

In addition, in order to obtain a picture concerning use of the same pigment in different buildings, a few samples from other building complexes were also selected: West House, Building B, areas of Sector A, Sectors B-G, Building Complex D (Porch of the triangular square, area of Pillar 17 and D17).

At the same time we conducted a chemical analysis of raw materials - lime plaster and red pigment - found, in a jar and in a cup respectively, in the West House.

The selected samples covered a wide range of colours and hues and they were in a variety of states of preservation. The thickness of almost all the samples varied from 7-12 mm., with two exceptions of 3 and 18 mm. (AKR 26 and AKR 50 respectively). Most of the samples had been fixed with mobylite. To the naked eye only a single layer of white plaster was evident, containing various impurities. The posterior surface was irregular without imprints of the underlying surface, while the front painted surface was smooth and flat.

EXPERIMENTAL TECHNIQUES

In order to identify the nature of the pigments, various analytical methods were applied, such as X-ray diffraction (XRD), electron microprobe, scanning electron microscopy coupled to an energy dispersive analyser (SEM-EDAX) and optical microscopy. These techniques have been used sporadically in the past in wall painting studies, but this is the first time that they have been applied simultaneously to material from Thera, in an attempt to obtain the maximum amount of information.

For the identification of the crystalline phases of the samples, X-ray diffractograms were taken directly on the surface of the sample. With this technique the surface of the sample was not destroyed, and detection was much better than with the conventional technique used in previous work which required scraping of the pigment. A small sample 5 x 5 mm. was enough to give a full diffractogram on a computerised Siemens X-ray diffractometer using copper radiation. The identification of the crystalline phases was made by computer software, using the PDF database of the Joint Committee for Powder Diffraction Standards.

The polished sections of the samples, containing a cross-section of the plaster and the pigment, were prepared and studied by optical and electron microscopy.

All samples were examined by optical microscopy in order to study the colour, nature and thickness of the pigment layer and its adherence to the plaster. Then the same samples were examined by SEM and electron microprobe with energy dispersive spectrometer, which enables chemical analysis of single pigment grains up to 2 μm.. Based on the analytical results, the mineral formula of the amphibole-grains and the other minerals were calculated by computer software (Perdikatsis 1986). The amphiboles were identified according to the nomenclature of Leake (1978), using the computer programme MINPET (1996) for rock and mineral classification.

RESULTS 

PLASTER

The base of all the wall paintings was lime plaster. XRD and microscopic examination showed that the plaster consisted mainly of calcite with small amounts of quartz and pumice. After removal of the carbonates with 10% acetic acid, the non carbonate and pumice part of the plaster was determined to be 46-50%. In all cases two generations of calcite were distinguished: primary calcite grains up to 100 μm and very fine lime material with grain size less than 5 μm. That means that the plaster material was primarily a mixture of calcite fragments and slaked lime (Ca(OH)₂) which was then transformed to secondary calcite during drying.

In many cases a thin layer of fine plaster was found between the pigment and the main plaster. They looked like a lime wash applied as a final layer before painting.

RED PIGMENT

A total number of 16 samples containing 'red' pigment were studied, as can be seen in Table 1. The term 'red' includes colours varying from dark red to pink as well as dark hues of orange.

Microprobe analysis showed that Fe was the major element in all samples. However, the overall analysis of the paint layer was dominated by Ca, followed by Si and Fe, along with smaller amounts of Al, K, and Mg. These results were confirmed by XRD analysis where, as can be seen in Table 2, the principal minerals were hematite and calcite accompanied by variable amounts of clay minerals (kaolinite, illite), feldspars and hornblende. It is thus obvious that the main colourant was hematite, varying in concentration and in aggregate size (Figs. 2c, 2d). The concentration of each of the main constituents was responsible for the intensity and the hue of the resulting colour. This is the reason for the large variety of hues observed in the red samples. Of the calcite present in the samples studied, part was primary and the rest secondary, formed from lime during drying of the painting. The ratio of hematite to inert minerals determined the hue and the intensity of the colour.

The thickness of the paint layers varied from 10 (AKR 30) up to 160 μm (AKR 24), with the majority being around 100 μm. Thickness of paint layers, like composition, was also responsible for the intensity of the colour. For example, AKR 35 is pink because it is only 20 μm thick on a white background, despite the fact that its composition is identical to that of AKR 34, which is a dark red but 5 times thicker.

Grain size of hematite in the samples studied was usually <5 μm. However, in the paint layers it very often formed aggregates whose diameter ranged up to 150 μm (Fig. 1e). The aggregates also contain clay minerals and form centres of very intense colour, easily distinguished under the optical microscope. The formation of the hematite-rich agglomerates results in the uneven distribution of Fe in the paint, and as a result partial chemical analysis of the paint may be misleading for the characterisation of the pigment.

In one case, AKR 35, an almost pure calcitic fine layer was observed between the plaster and the pigment. This is an indication of lime wash applied to the surface prior to the application of the red paint.

In most cases some significant penetration of the red pigment into the plaster (or lime wash) was observed. This was detected by optical microscopy as a thin layer, of about 20-40 μm, of intermediate red/white colour.

YELLOW

This also includes a few examples of light orange colours which were closer to yellow than red.

The major element found in these pigments was Fe, as in the previous case. Again the predominant element was Ca followed by Si. XRD analysis showed that hematite was the main colourant, but goethite and limonite (amorphous goethite) were also found in some of the samples (Table 2). Primary and secondary (through transformation of the lime to calcium carbonate) calcite was present, along with the same accessory minerals as in the red pigments. Grain size of goethite and hematite was <5 μm, but in many cases aggregates of 20-100 μm were formed, creating areas of intense colour. Clay minerals were evenly distributed within the pigment matrix in submicroscopical (<1 μm) grain size.

Paint layer thickness varied from 20 μm (AKR 18, yellow/pink) up to 200 μm (AKR 13, yellow). In most cases a transitional zone between the pigment and the plaster was observed due to the penetration of iron oxides. This zone was about one third of the thickness of the main paint layer.

In summary, the red, brown, orange, yellow, pink and intermediate hues can be characterised as typical ochres. Ochre is a natural mineral mixture of iron oxides, mainly hematite and goethite or limonite, with calcite and silicates like quartz, kaolinite, illite and other clay minerals. The amount of hematite and goethite in the ochre determines the intensity, the brightness and the nature of the pigment, from yellow ranging to dark red and brown. In Fig. 2 a characteristic example of dark red (2c) and yellow-mustard (2e) with their corresponding optical microscopy cross-section (2d and 2f respectively) can be seen.

BLUE

Three types of blue pigment were recognised according to the microscopical, XRD, SEM and microprobe analysis.

I. Egyptian Blue (e.g. AKR 40, AKR 43) (Figs. 1d, 1e, 1f).

II. Amphibole accompanied by chlorite and talc (e.g. AKR 41, AKR 42) (Fig. 1a).

III. Egyptian Blue with amphibole accompanied by chlorite and talc (AKR 45, AKR 46) (Figs. 1b, 2a, 2b).

The macroscopic colour of the samples varied according to the presence and the ratio of amphiboles and Egyptian Blue. Samples which consisted mainly of Egyptian Blue exhibited a marine blue colour, samples consisting of Egyptian Blue and amphiboles had a less bright colour, while samples containing only amphiboles exhibited a dull grey-blue colour. Intermediate colour hues were also achieved by amphibole-Egyptian Blue mixtures in varied proportions.

The Egyptian Blue particles varied in size from 20-100 μm, with the larger ones being deeper. By contrast, the common amphibole particle size was about 30-200 μm, because amphiboles in fine grain size virtually lose their colour.

A quite noticeable feature of the blue samples was the descaling of the pigment layer, which in some cases was severe. This loose connection to the plaster, probably due to the relatively coarse nature of the blue pigment, is also apparent at Knossos (Cameron et al. 1978).

This was also found in some samples, associated with amphiboles and chlorites. Its colour is light to dark green and depends on the iron content. In the analysed samples the FeO content was 3-5%.

The colour of chlorite is grass-green, olive-green or yellowish green. However, talc and chlorite do not have any overall significant colouring property. They are both accessory minerals to amphiboles in metamorphic rocks.

AMPHIBOLES

Amphiboles are silicates with a structure consisting of tetrahedral (SiO₄) double chains, forming [Si₄O₁₁]_n chains and strips of edge-sharing octahedra.

The amphiboles are the most complex of the rock-forming minerals, exhibiting considerable chemical and structural variation, and occurring in a wide variety of rocks.

According to the nomenclature of Leake (1978), the standard amphibole formula is:

A_0-1B₂C₅^viT^iv₈O₂₂(OH)₂, where T^iv₈ (tetrahedral sites) = Si_6-8Al_0-2C^vi₅(octahedral sites) = excess Al from T^iv₈ plus Fe³⁺, Mg, then Fe²⁺ and then Mn, B₂ = excess Fe²⁺, Mn, Mg from C^vi₅, then Mg then Ca, then Na, A_0-1 = excess Na from B₂, then all K.

According to the (BCa+ BNa) and BNa values, four main amphibole groups are defined as (Fig. 3a):

B = (Fe²⁺, Mg, Mn)₂ : Iron-magnesium-manganese (Fe-Mg-Mn) Amphiboles

B = Ca₂ : Calcic Amphiboles

B = CaNa : Sodic-calcic Amphiboles

B = Na₂ : Alkali Amphiboles

The definition of the alkali-amphiboles for (Ana+AK) < 0.5 is given according to Fe³⁺/(Fe³⁺+CAl) versus Mg/(CMg+Fe²⁺) (Fig. 3b).

According to Fig. 2a, the analysed amphiboles belong mainly to alkali- and sodic-calcic-amphiboles.

According to Fig. 3b, the alkali-amphiboles are characterised as magnesio-riebeckite (Samples 41, 42, 45, 52, 56), and the sodic-calcic (Fig. 3c) as winchite (Samples 41,42,44,45,56).

A few amphibole analyses belong to a calcic group (samples 42, 45, 56) and are characterised as tremolites (Fig. 3d).

The most common amphibole in the analysed samples is riebeckite followed by winchite; only in a few cases was tremolite found.

As regards the colour of the amphiboles, it is known that sodic amphiboles are usually blue to black. The intensity and hue of blue colour depends on the Mg, Fe-content. Pure Fe-riebeckite is black, but Mg-riebeckite is dark blue to blue (Anthony et al. 1995).

The colour of winchite is blue to bluish violet and, as in riebeckite, the Mg, Fe-content determines the colour. Finally, tremolite has a grey colour.

Glaucophane has not been determined. Although glaucophane is known as blue amphibole, its colour is grey ranging to lavender-blue. Especially in powder form it is always grey.

Hornblende belongs to the calcic amphiboles, with Ana + AK <0.5; Ti <0.5, usually of black colour, and is very common in volcanic rocks.

BLACK

No black pigment minerals were found. The black colour is probably a form of carbon and thus beyond the detection capability of the methods employed. The black particles are associated with other minerals, scattered in a matrix like calcite, quartz, kaolinite and illite. In some cases feldspar and hornblende grains were also observed.

DISCUSSION

Apart from the Egyptian Blue, all pigments used in Akrotiri came from naturally occurring sources. This explains the presence of particular minerals in the paint layer. More specifically, the presence of hornblende in the red, brown and yellow pigments indicates a local provenance for the ochres, because hornblende is known to be a common mineral in volcanic rocks and their weathering products. The use of local material was also evident in a grey-blue sample (AKR 53) with the presence of typical minerals for volcanic materials, like hornblende, zeolites and augite.

The provenance of the blue amphiboles should be assigned to areas of metamorphic rocks of low temperature and high pressure, such as the Cycladic islands of Syros and Siphnos. However, the analysed sodic amphiboles of these islands are glaucophane and not riebeckite. The suggestion of Cameron et al. (1978) that blue amphiboles were imported from Crete is questionable.

Two of the samples studied (AKR 14 and 50) contained a pigment macroscopically characterised as green. Although the borderline between blue and green is not clear, SEM and microprobe analysis did not reveal any of the known typical green minerals. They both contained amphiboles as colourant, probably hornblende, which can have an oily green colour, but by no means can it produce the colour which these two samples exhibited. The only explanation for the green is that it resulted from painting amphibole blue on top of yellow ochre. This is supported by the fact that in both samples there is co-existence of the green with yellow and brown (Table 1).

One of the red ochre samples analysed, AKR 31, found in a pot, consisted almost exclusively of hematite and was calcite free. This, combined with the fact that secondary calcite was always detected in the pigments, indicates that the ochres were ground before use to fine powder and then mixed with lime in variable proportions, according to the intensity of the required colour. Egyptian Blue, however, was found in relatively large grains (up to 60 μm) because when it is ground to fine powder it loses its colouring power (Noll et al. 1975). The same applies to the 'blue' amphiboles, which were found in even larger grains. These, in small grain sizes, turn to grey or become almost colourless.

In the paint layer there is co-existence of both primary and secondary calcite. Primary calcite is usually coarse, of rhombohedral shape, and must therefore have been crushed and then added to the pigment. Secondary calcite was very fine (<5 μm) and without shape, and was therefore formed by the transformation of lime to calcite during drying. It could also have been formed during burial, either by deposition and penetration from the environment or from bleaching from the underlying plaster. However, if that was the case, layers of secondary calcite would have been detected on top of the paint layer, which occurred only in one case (AKR 63). The use of lime as a binder does not automatically exclude the possible use of organic ones, but these cannot be detected (Noll et al. 1975) since they have decayed with time or burned from exposure to relatively high temperatures during the destruction of the site.

Penetration of the paint into the plaster was observed in some of the samples, as an underlying zone of a few tenths of microns, of lighter colour than the main pigment. It was more apparent in yellow than in red pigments due to small grain size and easier dilution of goethite-limonite (compared to hematite) in basic environments (as the pH of wet plaster). It was not observed in blue pigments, because of the large grain size and insolubility of Egyptian Blue and amphiboles (silicates). In general, the penetration is in inverse relationship to the pigment grain size. This difference in grain size was also responsible for the flaking of some of the blue samples, while all red and yellow still preserved good contact with the plaster. This was also observed in Knossian wall paintings by Cameron et al. (1978).

As far as the preparation of the wall is concerned, in most of the samples examined a single layer of lime plaster was observed. The existence of a final coating (lime wash or slip?) is not invariable, and consequently, not indicative of the technique (at fresco or at secco?) which may be supposed to have been used. In some cases, the presence of a red line of 1-3 μm, 200 μm under the paint layer (samples AKR 47 and AKR 49), marking the limits between the main plaster body and the final coating, could not be related to the presence of a specific pigment in the overlying paint layer, or to the application of a specific painting technique.

From the analytical results in Table 2 it is apparent that there are no significant variations among pigments of the same colour from different rooms of one building. The same applies to pigments from the walls of different buildings.

Comparing the present results with similar work conducted on material from Knossos (Cameron et al. 1978) and Ayia Irini (Majewski and Reich 1973; Cummer and Schofield 1984, 148), it seems that, as far as the preparation of the plaster and the nature of the pigments is concerned, in general the same materials and techniques were used in the contemporary sites of the Aegean. The only considerable difference is the blue at Ayia Irini which was composed of a layer of Egyptian Blue over a layer of carbon black (Majewski and Reich 1973). If this is the case, then the absence of blue amphiboles in the same area is also notable.

However, the situation is not the same as regards the painting techniques applied in wall paintings at the various sites. The main problem here is that there are no objective criteria for the characterisation of the painting technique and observations are therefore interpreted in a subjective manner, sometimes leading to contradictory conclusions. The most analytical discussion of this subject is that of Cameron et al. (1978), who concluded that the Knossian wall paintings were executed principally by the buon fresco technique and only some details by the secco.

It is known that with the fresco technique the crucial point is to paint on a surface that first dries out so that chemical bonding of the pigment is achieved through the slow transformation of lime to calcite (Levidis 1994). According to Cameron et al. (1978), the absence of considerable amounts of fillers in the plaster, which would accelerate drying and hardening of the surface, is an indication that there was a preference for slow drying in order to apply the pigment as fresco. This accords with the present analyses of plaster samples which were of similar composition to the Knossian ones. However, contra the above, Televantou (1994) has suggested that the application of the final coating of the plaster all at once indicates that drying rate was insignificant because painting was done in secco.

A second argument that would point towards fresco according to Cameron et al. (1978) is that blue colours that are usually difficult to fix in a buon fresco medium, are, as at Knossos, with some exceptions very well bonded, probably due to the chemical relation of the binder to the painting surface.

The macroscopic lack of penetration of pigments in the plaster was interpreted by Televantou (1994) as an indication of secco. In the samples studied, considerable penetration was detected microscopically in the red and especially in the yellow pigments, as has already been explained above. However, this does not necessarily constitute a criterion, because penetration can take place due to environmental conditions after burial; and, furthermore, penetration should be more evident in secco when the underlying surface is dry, due to osmotic pressure.

Other criteria set out by Cameron et al. (1978) were either not met, or were based on macroscopic observations and therefore could not be evaluated by the methods used in this study. For example, there was no evidence for the dragging up of soft plaster by the paint brush, as was the case at Knossos. Cameron et al. (1978) also suggested that, with the fresco technique, polishing the surface after painting should enrich it in calcite, a feature which was not detected (except in one case) in our samples.

Finally, information on the techniques proposed for other contemporary Aegean wall paintings studied can be summarised as follows. At Ayia lrini, after macroscopic and scientific analysis, it was suggested that the majority of the pigments were applied in buon fresco technique, except blue and black which were applied on a dry surface (Majewski and Reich 1973). For Phylakopi there are only visual observations leading to the conclusion that the method used was a combination of fresco and tempera (secco), with the outlines made in fresco and the rest on dried plaster (Bosanquet 1904, 79). Observations made by the excavators on wall paintings from the latest phase of Ialysos (LMIB-LMII) have led to the conclusion that the base was executed in fresco and the motifs in tempera (Monaco 1941, 69-71).

To conclude, it can be maintained that the majority of the pigments of the Theran wall paintings were applied in fresco technique, but the secco technique was also used.

-------------------------------------------------------------

--------------------------------------------------------