An investigation, using synchrotron radiation and other techniques, of the composition of San rock art paints and excavated pigments from Maqonqu shelter, and comparative paint data from three other sites in KwaZulu-Natal, South Africa.

Abstract

This study aimed to: 1) characterise the individual San parietal art rock art paint colours; 2) relate paint compositions to erosion susceptibility; 3) determine if paint pigments can be related to pigment samples excavated from a Shelter deposit, and/or a variety of field samples; and 4) determine if paint samples from geographically distinct sites can be distinguished on their composition. A combination of mineralogical (X-ray diffraction (XRD), synchrotron micro-XRD (μ-XRD)) and chemical (energy dispersive X-ray micro-analysis (EDX), X-ray fluorescence spectrometry (XRF), and synchrotron micro-XRF (μ-XRF)) analytical techniques were used. Maqonqo Shelter (MQ), 35 km south-east of Dundee, KwaZulu-Natal, South Africa, was the primary study site chosen as it contained both a large number of paintings, as well as a large deposit. Thirty paint (of various colours) and 3 blank wall samples were collected using Silver Mylar tape and analysed using a combination of EDX, μ-XRD and μ-XRF techniques. Sixty two large (> 2.5 g) ‘ochre’ pieces were selected from the upper three layers of the deposit and analysed using XRD, XRF and EDX. A further 63 small pieces (< 2.5 g) were analysed using μ-XRD and μ-XRF techniques. To compare the MQ paint samples with potential source materials, three distinct sample sets were collected. The first included samples of the Shelter wall and surface rocks located near the painted panel (analysed by XRD, XRF and EDX). A second sample set of 17 samples was collected from the surrounding landscape (± 3 km radius of MQ; analysed by XRD and XRF). Their selection was based on ease of accessibility, degree of pulverulence, and perceived Fe content i.e., red and/or yellow colouration. No white sources were found. A third set of 11 samples (obtained from six sites, analysed using XRD and XRF) was collected within ± 50 km distance of the Shelter. Their selection was based on old mining survey reports that detailed the location of Fe ore outcrops. Paint samples from three additional shelters i.e., Fergies Cave (FC), Giants Castle Game Reserve, central Drakensberg; Sheltered Vale (SV), Mount Currie District, south-western KwaZulu-Natal; and Twagwa Shelter (TW), Izingolweni District, southern KwaZulu-Natal, were collected to compare paint composition over distance. Site selection was determined according to the following criteria: 1) the shelters had to reside a significant distance away from the primary site so as to minimise any possible interaction that might have existed between the authors of the respective artworks (each site is at least 100 km distant from the other); 2) each had to be located upon a distinct geological formation so that external influences from different regions, and their possible affects on the paint samples, could be noted; and 3) the climatic regimes of each of the shelters should be relatively distinct. Fifteen paint and nine blank wall samples were collected from the three shelters (three each of red, white and blank samples; analysed using EDX, μ-XRD and μ-XRF), with the exception that no white samples were collected from FC. In total, 673 EDX, 212 μ-XRD, 378 μ-XRF, 98 XRD, 98 XRF and 6 ICP-MS traces were produced and analysed. Due to the extremely heterogeneous nature of the paint samples at the microii scale, the more generalised EDX reduced window scans were used as the basis of the paint samples’ characterisation, with the data obtained from the more precise μ-XRD and μ-XRF techniques providing additional supportive information. Irrespective of colour, almost all of the MQ paint samples had elevated Ca contents that tended to increase in the order of black < orange £ red and yellow < pink < white. The predominant Ca-based mineral was gypsum, although Ca-oxalates, whewellite and weddellite, were also present. The blank samples collected from MQ also had high gypsum content, but no Ca-oxalate. It is thus proposed that the Ca-oxalates formed after the painting event and were derived from the original paint constituents. The white pigments consisted of gypsum (dominant), anhydrite, bassanite and whewellite, or a combination thereof. Whewellite increased within increasing paint depth, while gypsum showed the reverse trend. This indicates that, whilst both gypsum and whewellite were originally present within the original paint pigment, additional gypsum has been added via secondary evaporite deposition. Although initially considered to be sourced along with the gypsum, another potential whewellite source is organic additives. The most likely source for the white pigments would be precipitates found on sandstone walls of shelters near MQ. Of more immediate importance, however, is that the pigments, being gypsum based, are water-soluble and thus susceptible to erosion. Most of the orange paints had an elevated Al content and contained gibbsite, suggesting bauxitic material associated with locally sourced dolerite within the Ecca Series within KwaZulu-Natal (as evidenced by their respective Ti levels). Two samples were so similar that it is likely that the same pigment was utilised in the creation of both images. Two samples did not contain high Al contents, however, indicating that they were probably sourced from the soft, ochreous material found within local Fe nodules. A consistent combination of goethite and haematite, together with a low Al and elevated Ti content, indicate that the yellow and red samples were probably sourced from Fe nodules found locally, the red samples differing from the yellow pigments primarily in their higher haematite content. A low Si and relatively low Fe content discounts red sands/clays and Fe-ores as sources of the red pigments. The red samples were ‘thinner’ than the other samples with quartz contents comparable to those of the blank samples. The thin nature of the red paints, the erratic distribution of whewellite upon the paint surfaces, the dominance of gypsum and, to a certain extent quartz, all strongly suggest that the red paints are at least partly absorbed into the surface of the Shelter wall. This, together with the strong staining ability of haematite, is probably the most important reason that the red pigments have outlasted images painted in other colours. It may also account for the high degree of variability found within the red paint dataset, though age differences between the sampled images could also be a contributing factor. The single dark red paint sample, except for an elevated Mn content, was very similar in many ways to the red paint samples analysed. The only readily available pigment source identified that had both low Al and high Fe and Mn contents, was plinthite. The pink samples represented the ‘middleiii ground’ between the red and white paints, suggesting that this colour was the result of a blending of the two. The black paint sample had the highest recorded Fe content of the entire paint dataset. A high Mn and relatively low Al content suggest that a soft inner core of an Fe nodule was used in its manufacture. The presence of maghemite and a dark colouration strongly suggest that the manufacture also involved calcination. The initial distinction between the paint and excavated samples was that the former all exhibited elevated Ca and S values due to the deposition of secondary evaporite minerals. Even when taking these additional deposits into account, however, the two datasets still remained distinct indicating that the excavated materials sampled were not utilised in the manufacture of the MQ paints. A potential exception concerned the orange paint samples, which were similar in composition to both doleritic samples from deeper excavated layers and the local (weathered doleritic samples) and distant (bauxite samples) field samples. Whilst weathered dolerite/bauxitic material was clearly the source of the orange pigments, a more detailed investigation is needed to find a precise location. No other relationships between the paint pigments and the excavated pigments and field samples were established. A comparison of the blank samples from all four study sites showed that the techniques used could distinguish between different sites despite sampling the smallest and, relatively speaking, poorest quality samples. The FC blank samples had elevated C and Ca contents (associated with Caoxalates). The conditions within this Shelter favour the formation of weddellite and whewellite, the former not typically found at the other three sites. In addition, low K, Si and Al contents (often associated with sandstone matrix minerals) indicate that the surface of the relatively dense, compact Cave sandstone is more resistant to physical erosion compared to the other sites, and/or FC shelter experiences a high amount of secondary deposition, with the result that a majority of the samples are composed of evaporite minerals. The SV samples were composed primarily of the evaporite-type minerals, with only minor sandstone ‘contamination’ indicated by quartz and kaolinite. The quartz content, whilst not always high, was present in most of the samples analysed, possibly indicating a greater amount of more uniform surface erosion (relative to the other sites). The TW blank samples were distinct from the other shelters’ as they contained no Ca-based minerals but did contain the very rare mineral schlossmacherite. A comparison of the paint colours also revealed differences between the different shelters. Whilst the white samples from SV and MQ are dominated by whewellite and gypsum (minerals probably present within the pigments when they were applied), the presence of quartz, sanidine and apatite in the SV samples indicated a degree of shelter wall ‘contamination’, with anhydrite, bassanite and glushinskite suggesting climatic variations that favoured various evaporite depositional regimes. The TW white paint contained minimal secondary deposited minerals common in the other shelters. The one mineral that is dominant within the TW samples is minamiite. As this mineral was not identified in any of the blank samples, it is likely that this mineral originates from the original pigment source. The TW white paints also contained 10 to 40 times more Zn than those recorded for any of the other paint samples. This was possibly present within the structure of greigite. The red SV samples could be distinguished from MQ red samples by the presence of wall ‘contaminants’ in a manner similar to that described for the white samples. The TW samples indicate a change in pigment source and/or manner of paint manufacturing technique, for these red samples contained minamiite. This mineral is white and thus its selection could not have been based on colour but rather it must represent a paint additive. With the exception of only one sample from TW, no goethite was found within any of the red samples collected from the three additional sites indicating a different haematite source to that of MQ. An interesting facet of this study, although not directly addressed, concerns what the results do not show with respect to the compositional nature of the pigments analysed. Most texts available today list a number of pigment sources stated to have been utilised in the manufacture of the San parietal rock art. This study has shown that very few of these potential sources were utilised within the four shelters investigated. In addition, this study has also highlighted the presence of minerals about which little is known, yet which appear to be commonly associated with parietal rock art.