Visual inspection and low level microscopy
The banner was examined by eye and with low level magnification. Through this process it was possible to determine a broad categorisation of materials, the banner’s construction and painting techniques. Close study of the textile-paint interface indicated evidence of a possible preparatory layer applied directly to the silk (Fig. 2a). Figure 2b shows a transparent yellow toned coating/varnish layer over the silver-coloured paint to create a gold colour that appears to overlap onto the silk. These preparatory layers, paint layer and the coating/varnish have considerably stiffened the silk.
The condition of the textile, paint and paint-textile interfaces including presence of soiling, creasing, splits and abrasion were documented based on visual and low-level microscopy. The differences in flexibilities between the painted and unpainted areas were evident. Paint deformation resulting from previous storage around the narrow banner pole and areas of paint loss associated with flexing and creasing are evident (Fig. 2c, d). Some of these distortions may be the result of different flexibilities of the materials but also different responses to environmental conditions between the painted and unpainted areas. This damage is typical for banners [6].
Under low level magnification micro-cracking of the paint surface and paint losses (Fig. 2c, d) were visible, particularly on the areas with several layers of paint.
Light microscopy
Due to previous substantial damage it was possible to be able to take cross-sections that comprise the whole layer build-up of the preparatory, paint and varnish layers on both sides of the silk support. Fibre identification, longitudinal and cross section, was carried out on the banner and the fringe. The banner was identified as silk and the fringe was a mixture of cotton and a man-made fibre. The man-made fibre was confirmed by FTIR as viscose rayon. Polarised light microscopy from P1 (Fig. 3) shows a layer build up starting with a white preparation/priming layer of ca. 50 µm that seems to have impregnated the silk support on both sides. This is followed by one thin further paint layer on each side of the banner (ca. 10 µm). A similar white ground has been found on many mid nineteenth to early twentieth century processional painted banners, most of them on a silk support [2, 7]. In P4 (Fig. 4) there is a silver coloured paint layer which was identified as aluminium (see SEM-EDX results) (ca. 10 µm), which in some places of the cross-section is covered by a transparent yellow layer, discernible in UV, applied to make the silver coloured paint appear golden. Figure 5 shows a loose sample again of the white ground layer with a red paint layer which appears glossy due to a coating of a resin based varnish (see FTIR results).
Protein staining was carried out on embedded cross sections following a method detailed by Schäfer [8]. The metal chelate stain, SYPRO® Ruby was used as it is readily reversible, does not modify the protein and is compatible with mass spectrometry. Figure 6 shows an embedded sample from P3. It can be seen that the silk, a large part of the sample, stains as it is a protein. The method was developed to identify and localise the presence of protein in historical paint cross sections. Here it also functions to indicate the protein fibre component, silk. There is also evidence for a protein type material within the areas where the metallic paint has been used. However, it is not possible to identify from this test alone which type of protein it is. An analysis technique such as gas chromatography with mass spectrometry would be required to do this [9].
SEM and EDX
Figure 7 shows a scanning electron microscopy (SEM) image of a loose sample taken from P1. A cross section sample taken from P1 was examined on both sides using SEM-EDX. Analysis of the back of the banner detected lead and traces of aluminium, silicon, potassium, sodium and calcium. The lead is likely to be attributed to lead white commonly used as a ground layer in banners made by Tutill [2]. Analysis of the metallic paint (P2) showed the presence of aluminium which is frequently used in metallic paints to give the appearance of silver.
Ion-milling to produce high quality smooth samples has been used in materials analyses for a considerable time [10] and was investigated here. Ion-milling involves hitting the sample surface with accelerated submicron ion particles while it is mounted on a rotating table inside a vacuum chamber. Its first reported use in a cultural heritage application was published in 2006 by Boon and Asahina [11]. Subsequently work by Prati et el. [12, 13] has shown the advantages of ion-milling embedded paint samples. Figure 8a, b shows an embedded sample before and after ion-milling, demonstrating the high structural definition of the different materials and distinctive layers not routinely seen with traditional polishing techniques. Figure 9 shows ion-milled samples (a and b) taken from P1 and P4 respectively indicate that different preparation techniques have been used on the same banner i.e. Fig. 9a shows the inorganic materials from the ground layer spread though out the silk fabric while Fig. 9b shows a preparatory organic layer on the silk fabric preventing the ground paint from having contact with the silk. This use of different preparation techniques on different parts of the same banner has not been reported before.
FTIR-ATR
All the FTIR analysis was done using the external ATR arm and thus is considered macro (bulk) analysis. A large range of materials was identified.
Samples from the fabric and fringe were analysed with standards from a fabric supplier (Whaleys, Bradford, UK) used as reference material. FTIR-ATR confirmed the use of silk for the fabric and a mixture of viscose rayon and cotton for the fringe. Figure 10 shows the spectrum obtained from side 1 of a sample taken from P4 compared with a spectrum of new silk. The amide I and amide II bands at 1620 and 1514 cm−1 respectively, are associated with the fibroin component of the silk. Other fibroin related bands at 1442 and 1164 cm−1 arise from the vibrations of the side chains of alanine and tyrosine in the silk [14]. FTIR gave a clear molecule signature for silk.
Figure 11 shows the spectra obtained from the two thread types taken from the fringe on the bottom of the banner. One thread type has been identified as cotton and the other as viscose rayon. The FTIR spectra of these materials are very similar. However, a peak around 1100 cm−1 associated with the glucose ring is present in cotton but missing in the viscose rayon.
Figure 12 shows the spectrum of sample P1 compared with spectra of linseed oil (aged), calcium carbonate, lead white and silk. It is not possible to identify any of the materials with 100 % certainty as often broad bands mask the presence of narrow bands that would appear within their area. However, bands at wavenumbers specific to linseed oil, silk, calcium carbonate, and lead white were detected. Linseed oil has very strong CH2 asymmetric stretching at 2920 cm−1 and strong CH2 stretching 2850 cm−1. The C=O stretching is seen around 1740 cm−1 and C–O ester stretching at around 1160 cm−1. The broadening of the C=O peak at 1740 cm−1 is indicative of the hydrolytic degradation of triglycerides which leads to the formation of the free fatty acids which is seen here [15]. Additionally the peak at around 1706 cm−1 may be associated with fatty acids formed by hydrolysis of glycerol esters [16]. Lead white has a strong band at around 1370 cm−1 assigned to the CO
−23
and a band around 680 cm−1 both seen on the banner spectrum. Calcium carbonate also exhibits a broad band around 1392 cm−1 but additionally at 872 cm−1 which is seen here. Silk is also seen at 1618 cm−1, 1531 cm−1 associated with the amides I and II.
The banner exhibited areas of an opalescent bloom previously reported to be present on marching banners and which could be attributed to the formation of metal soaps [17, 18]. The simultaneous presence of free fatty acids in the binders and metals from pigments make the formation of soaps predictable [19]. However micro-pitting of the paint surface or mould are also known to cause bloom. The spectrum of a sample from P4 showed two strong sharp bands between 1500 and 1600 cm−1 one at around 1586 and another at 1538 cm−1 as a result of COO- asymmetric stretching associated with calcium soaps [20, 21]. The CO
−23
of CaCO3 creates a broad band at around 1400 cm−1 which masked the symmetric bands of the calcium soaps. However by performing first derivative calculations, the overlapping peaks could be seen at around 1472, 1433 and 1420 cm−1 which are all absorbance bands present in calcium soaps. The formation of the soaps is also confirmed by the reduced presence of the C=O ester and acid stretching at 1700–1750 cm−1 associated with the linseed oil components. Despite the presence of lead white detected on most of the samples from the banner the formation of lead soaps, on the samples analysed, was not detected as the doublets associated with lead soaps occur at lower frequencies 1540–1500 cm−1 [22] and these were not seen on the spectrum. However, this may be due to the use of macro FTIR and it may be that the use of micro FTIR may have found these soaps to be present. Therefore it cannot be deduced from these results that they were not present.
A sample from the red translucent paint at P4 (see Fig. 5 for light microscopy image) appeared to contain a resinous material and produced a spectrum indicating that it was possibly shellac, as a broad doublet type C=O band around the lower regions of 1700 cm−1 was observed as well as a band at 1636 cm−1 associated with the C=C bonding. Identifying differences in spectra between aged drying oil and resin proved difficult as the C=O band broadens to lower frequencies in oil due to ageing. However, the vinyl band at 1636 cm−1 present in shellac but not oil and a strong band at 1160 cm−1 in oils which is weaker in shellac can help differentiate. Additionally the C=O doublet seen in shellac resin is not seen in drying oil. Figure 13 shows the banner spectrum compared with shellac (supplied A.F.Suter, London). For confirmation of shellac the detection of erythrolaccin (a yellow dye present in shellac) could be achieved using high performance liquid chromatography.
Attempts to identify the possible unknown preparatory layer which was visible by eye and on the ion milled samples (Fig. 9) were explored. Spectra were compared with a spectrum of latex rubber known to have been used on painted banners [7] and rabbit skin glue commonly used on paintings. Neither of these was detected on this banner using FTIR. The FTIR analysis of the reconstructions showed that some combinations of materials are very difficult to detect. The spectrum of the combination of silk fabric coated with a 20 % (aq) latex rubber solution is shown in Fig. 14a. Here it is possible to see the absorbance bands associated with silk but the latex rubber bands are not easily identified despite their occurrence at different areas on the spectrum. Figure 14b shows the spectrum of the silk with 10 % (aq) rabbit skin glue. As would be expected, as both are proteins, it is not possible to identify the presence or absence of rabbit skin glue. This problem with using this type of analytical methodology to analyse multi-layered samples has been noted [23]. Often characteristic absorptions of some materials are masked by stronger bands of others. Therefore we cannot conclude from the FTIR results that a preparatory layer is not present only that it cannot be detected in this case.
Fourier transform infrared spectroscopy microscopy was not carried out here. This technique is able to achieve high spatial resolution and gains in sensitivity in detecting trace materials and it may have yielded information on the specific location of materials in the unembedded samples. However, the problems associated with bands from resin interference in embedded samples often prevents identification of oils resins etc. as the absorbance bands associated with these appear in the same parts of the spectrum. Much research is being done to prepare successful cross sections in materials which have less detrimental effects on the analysis [24, 25].
Raman
Work concentrated on optimising micro-Raman analysis conditions, by using different excitation wavelengths, exposure times and power settings to detect the differing materials within the samples. Additionally care was taken to choose settings which did not damage the sample through overheating. Analysis at 488 nm yielded poor results for the embedded samples as all spectra were badly affected by fluorescence. A loose sample from P1 did produce a spectrum but required a high power setting of 50 %, which could potentially damage the sample. Using lower power settings yielded no information. In contrast using an excitation wavelength of 785 nm, exposure time of 10 s and power setting of 0.1 % on an embedded sample from P1 produced a good spectrum from the dark area (top paint layer). The peak at 544 cm−1 has been identified as red lead by comparison to a minimum standard, a red lead mineral (supplied by National Museums Scotland Collection), Fig. 15. The peak at 1086 cm−1 visible on the spectrum of the embedded sample is indicative of calcite (calcium carbonate) which is often used as pigment filler [26, 27].
Using 785 nm and 0.1 % power produced good spectra for lead white in both embedded and loose samples. Figure 16a, b show spectra of lead white. Figure 16a shows the spectra as they were collected and Fig. 16b shows them with baseline smoothing and the wavenumber range reduced. The detection of lead white occurs at around 1050 cm−1 (when compared to the spectrum produced from the lead white mineral cerussite, RRUFF database) in the samples P1 embedded, P1 loose sample and P3 loose sample. Additionally a good spectrum for the identification of lead white was achieved using the Renishaw system at 325 nm for a loose sample from P1.
The use of Raman to identify the silk in embedded and loose cross sections proved to be difficult. Figure 17 shows the spectrum of embedded silk at the 785 nm excitation wavelength showing it was difficult to obtain a good quality spectrum for silk for an embedded cross section due to the extent of fluorescence from the resin block as well as the fibre. However, with baseline correction and truncating of the sample spectrum two peaks were identifiable when compared to the spectrum obtained from a loose silk sample (new), one being associated with silk 1665 cm−1 but the other (1588 cm−1) most likely due to the embedding resin (see Fig. 17). This problem was noted by Macdonald et al. [2] when analysing cross sections of banners with Raman.
Problems with intense fluorescence signals from organic materials and dyed fabrics have been documented [28, 29]. These signals often mask the signals for the lesser reacting molecules. The use of confocal Raman microscopy has been reported as a way to prevent this [29] and Lorenzetti et al. [30] reported on its usefulness to determine the dyes on historic dyed cotton yarns by using a confocal condition 2–3 µm lower than the fibre surface. However Macdonald et al. [31] reported that the results often included output from both above and below the point of focus and so these contributed to the Raman response albeit to varying degrees. Thus the use of confocal microscopy may not be ideal as it is not always certain which area in the sample yields the results. When using Raman to analyse samples from painted textiles the presence of the textile such as silk, as well as the typical organic oils and resins associated with pigments, mean that fluorescence signals will be strong from such samples so care has to be taken when processing these results.
In summary a total of four different wavelengths were used in the analysis, ranging from 325 nm in the ultra-violet range, 488 and 523 nm in the visible range and 785 nm in the infrared region. The analysis of cross sections, either loose or embedded, was most successful using an excitation wavelength of 785 nm. Relatively strong signals were obtained using short scan times typically of around 10 s. Using the excitation wavelength 325 nm reduced fluorescence but failed to detect pigments, excepting lead white, or silk with any degree of certainty. Similarly Burrafato et al. [28] compared the use of three excitation wavelengths 531.5, 632.8 and 780 nm for the identification of pigments both dry and in egg tempera, casein tempera, oil and fresco. They reported that the presence of other materials often made detection difficult or impossible e.g. cinnabar in oil could only be detected using 780 nm. The use of Fourier Transform (FT) Raman at 1064 nm has been reported to be useful for the determination of organic materials such as dyes as it does not suffer from fluorescence interference. The signal from the sample is weakened at this wavelength [32] however the use of a sensitive single element such as indium gallium arsenide or liquid nitrogen-cooled germanium plus an interferometer converting the Raman signal greatly improves the output. Perhaps this wavelength’s use, due to reduced fluorescence, would have helped in identification of some materials here.