Analysis of synthetic dyes in an embroidery of Emile Bernard (circa 1892) M R van Bommel* and A M Wallert Netherlands Institute for Cultural Heritage PO Box 76709 1070 KA Amsterdam The Netherlands E-mail: Maarten.van.Bommel@icn.nl Web site: www.icn.nl I Vanden Berghe and J Wouters Royal Institute for Cultural Heritage Jubelpark 1 B-1000 Brussels Belgium E-mail: Ina.vandenberghe@kikirpa.be; Jan.Wouters@kikirpa.be Web site: http://www.kikirpa.be J Barnett Regina Textilia Historical textiles research and consultancy Oude Looierstraat 65-67 1016 VH Amsterdam The Netherlands E-mail: reginatextilia@ision.nl R Boitelle Van Gogh Museum PO Box 75366 1070 AJ Amsterdam The Netherlands E-mail: boitelle@vangoghmuseum.nl Web site: www.vangoghmuseum.nl. *Author to whom correspondence should be addressed Abstract Research focused on early synthetic dyes is presented. A selection of 65 dyestuffs, produced in the period 1850–1900, had been investigated previously. Both non-destructive and destructive analytical techniques were evaluated for dyestuff identification. To establish the effectiveness of the selected dyes and the suitability of the analytical techniques, well-dated objects have to be examined. Therefore, an embroidery designed by Emile Bernard (dated 1892) was studied; the results are presented in this paper. In 27 samples taken from the object, 13 different synthetic dyes and 3 natural dyes were detected, often in mixtures. Only two dyes could not be identified. The blue, mauve, purple and green colours were all severely faded whereas the red, pink and yellow colours turned out to be more stable. Keywords synthetic dyes, high-performance liquid chromatography, 3D fluorescence, embroidery, Emile Bernard Introduction Since the introduction of modern analytical techniques, progressively more attention has been paid to the analysis of natural dyes of historical or art objects. In contrast, the analysis of synthetic dyestuffs is a somewhat unexplored area. Since the invention of mauve by W H Perkin, in 1856, natural dyestuffs have been rapidly replaced by synthetics. By 1900, approximately 400 new dyestuffs had been developed (Lehne 1893) although they were not all produced commercially. Recently a selection of 65 synthetic dyes covering all dye classes available in the period 1850–1900 was investigated (Van Bommel 2004). For analysis, threedimensional (3D)fluorescence spectrometry with a fibre optic system was used in a first phase. This nondestructive technique, showing the interaction between emission wavelength, excitation wavelength and emission intensity, can be used for natural and synthetic dye identification. Subsequently, high-performance liquid chromatography (HPLC) was used in combination with photo diode array detection (PDA). With PDA detection, ultraviolet-visual (UV-vis) absorption spectra from 200 to 700 nm were measured from all the components separated with HPLC. The combination of these spectra and their chromatographic behaviour is a powerful tool for identification of dyestuffs. Two different HPLC procedures were evaluated. First, the HPLC procedure developed by Wouters (1992), using a gradient of water, methanol and phosphoric acid, for the identification of natural dyestuffs was applied to synthetic dyes. Most synthetic dyestuffs can be analysed using this system and the basic dyes in particular show excellent results. On the contrary, the results of the acid dyes were less satisfying because of the bad chromatographic properties under these conditions. Therefore, a second HPLC procedure was tested using a gradient of water, methanol and tetrabutyl ammonium hydroxide (TBA). TBA acts as counter ion to the negatively charged acid dyes, resulting in improved chromatographic behaviour and in a significant improvement of detection limits. Some of the selected dyestuffs were identified on numerous objects (Whiting 1978, Rabe et al. 1990, Wouters 1992) but it is unknown if the selection covers all the important synthetic dyestuffs used in the second half of the 19th century. To gain more knowledge of these dyestuffs and to demonstrate the suitability of the techniques, accurately dated objects need to be analysed. The present research focuses on embroideries designed by Emile Bernard. Five of these embroideries, part of the Bonger Collection, are now kept at the Vincent van Gogh Museum, Amsterdam, The Netherlands. As part of the Preservation of Cultural Patrimony Act, in 1996 the Dutch State acquired the collection of Andries Bonger, which is administered by the Van Gogh Museum. Andries Bonger (1861–1934) was a merchant who lived in Paris from 1879 where he struck up a friendship with Theo van Gogh, Vincent van Gogh’s brother. Theo introduced Bonger to the work of French avant-garde artists. After returning to The Netherlands in 1892, Bonger began collecting art himself, in particular the works of two friends, Odilon Redon (1840–1916) and Emile Bernard (1868–1941). The first embroidery selected for examination is Bernard’s Two Breton Women, circa 1892; wool on cotton, trimmed with a braided border and stretched on a strainer (78 cm × 61 cm). The work (Figures 1 and 2) is now extremely faded and is dominated by a greenish grey hue, whereas the unexposed parts of the textile still show a wide range of very bright colours. Owing to its date and the extreme fading, early synthetic dyes were expected to be present. Because long yarn end are present at the back of the object, samples for dyestuff analysis could be easily taken without disturbing the integrity of the object. The experimental details of the techniques are described in a subsequent paper (Van Bommel et al. 2005). Results and discussion Red and pink yarns The results of the analysis of the red and pink yarns are presented in Table 1. The red yarns were dyed with synthetic colourants: cochineal red A (CI-16255)1, ponceau RR (CI-16150) and ponceau 3RO (CI-16050). Ponceau 3RO is present in a small amount. These colourants belong to the class of acid red dyestuffs. The ratio of the mixture of the three dyes in samples 1 and 2, determined with HPLC, indicates a small difference in concentration: the concentration of cochineal red A in sample 2 was about 10 per cent higher. There is no colour difference visible at the back of the object and, as the samples were taken from the red dress, colour differences were not expected. However, on the front of the object, which has been exposed to light (Figure 3), sample 2 is faded to a more orange colour whereas the other part is still red. Although analysis of samples from the front of the object was impossible, one can still conclude that cochineal red A is less light-fast compared with the other dyestuffs in this area. Sample 3 consists only of ponceau RR, which means that a different dye was used. This area shows yet another red colour compared with the area where sample 1 and 2 were taken. A small amount of naphthol yellow S was found as well, but this could be a cross-contamination; that is, some fibres from another location could have contaminated the sample. The pink samples were dyed with carminic acid (CI-75470) in a rather low concentration, which explains the pink colour. Carminic acid indicates the use of a red insect dye from a natural origin. Considering the date of the object, cochineal (Dactylopius coccus L) is most likely as synthetic carminic acid was not developed until 1991 (Pietro 1998). Cochineal can be more precisely identified by means of minor components. However, because of the low concentration, no Figure 1. Front of embroidery Two Breton Women, designed by Emile Bernard, circa 1892 Figure 2. Back of embroidery Two Breton Women, designed by Emile Bernard, circa 1892 Table 1. Analytical results of the red and pink yarns Figure 3. Detail of discolouration of the red dress of the woman on the left. The number on the white arrows corresponds with the sample number in Table 1. The samples are taken from the back minor components were found in samples 4 and 6. In sample 5, minor components were found, among them dcII, which proved that cochineal was used (Wouters and Verhecken 1989). The pink colours were slightly faded whereas the fading of the red areas dyed with synthetic dyes was much stronger. 3D fluorescence made it possible to recognize the presence of a combination of several acid red colourants on the red samples and to detect the presence of cochineal on both pink samples. Although the acid dyes were readily identified with the use of the TBA gradient, the results of the phosphoric acid were less easy to interpret owing to the bad chromatographic properties under the conditions used. Carminic acid, a natural dye, was identified with phosphoric acid gradient. Some acid red dyes together with carminic acid were found in sample 4 with the use of the phosphoric acid gradient. Because these red acid dyes were not found in this sample with the TBA gradient, it is believed that this is caused by sample carry-over from the previous analysis. Yellow and green yarns Both synthetic and natural dyes were found in the yellow and green samples (Table 2). In sample 7, naphthol yellow S (CI-10316), Martius yellow (CI- 10315) and picric acid (CI-10305) was found. Because the concentration of Martius yellow and picric acid was very low, these are considered to be byproducts. Sample 8 was dyed with the yellow picric acid, which was originally derived from blue indigo (Indigofera tinctoria L) by the treatment of concentrated nitric acid (Christie 2001). After 1841, a procedure was developed to produce picric acid from phenol. In sample 9, both synthetic and natural dyestuffs were found. Naphthol yellow S and picric acid were found in an approximately equal ratio. The identification of quercetin, kampferol and rhamnetin indicates the use of the natural dye buckthorn or Persian berries (Rhamnus species). However, the presence of buckthorn was not confirmed by 3D fluorescence spectrometry, where auramine (CI-41000) was clearly identified as main dye. The different analytical results cannot be explained, a larger sample has to be analysed as the concentration dyestuffs found with HPLC was rather low. A mixture of natural and synthetic dyes was also found in sample 10 where picric acid, luteolin and apigenin, indicating the use of weld (Reseda luteola L) were found. In this sample, picric acid was present in abundance. The combination of natural and synthetic dyes is unexpected, since these are mordant and acid dyes respectively. Using two different dyeing procedures makes the preparation of the textiles complicated. The fading of these yellow areas is not severe. The pale colours (samples 9 and 10) look more faded than the bright yellow in samples 7 and 8. The two green samples were both dyed with a mixture of indigo carmine (CI- 73015) and picric acid. Indigo carmine is prepared by treating indigo with concentrated sulphuric acid; as a result, mono-sulphonic and di-sulphonic indigotin are formed (Christie 2001). Mixing the blue and yellow dye resulted in the green colour observed. Dyeing green by mixing yellow and blue dyestuffs was not uncommon. As there are no stable green natural dyes, dyeing green has been done with yellow and blue dyes for centuries. When synthetic dyestuffs were developed, one was able to dye green with a single dye, but the result of Table 2. Analytical results of the yellow and green yarns The two green samples were both dyed with a mixture of indigo carmine (CI- 73015) and picric acid. Indigo carmine is prepared by treating indigo with concentrated sulphuric acid; as a result, mono-sulphonic and di-sulphonic indigotin are formed (Christie 2001). Mixing the blue and yellow dye resulted in the green colour observed. Dyeing green by mixing yellow and blue dyestuffs was not uncommon. As there are no stable green natural dyes, dyeing green has been done with yellow and blue dyes for centuries. When synthetic dyestuffs were developed, one was able to dye green with a single dye, but the result of these samples indicates that mixing dyes was done as well. The ratio between the two dyes in both samples is approximately 10 per cent indigo carmine and 80 per cent picric acid. However, the concentration of dyestuffs found in the dark green sample 12 was about nine times that in green sample 13, which seems to be much more faded on the front of the object. With the exception of sample 9, analytical results of the techniques applied were consistent. Out of the 3D contour plot from the yellow samples 7 and 8, a yellow colourant from the type of picric acid, Martius yellow and/or naphthol yellow S was detected and in the green samples indigo carmine as well. However, with this technique no distinction can be made between these synthetic yellow dyes. It is not always possible to differentiate even indigo carmine from them as they all are emitting in the same area. Not only interference between the colourants but also of the yarns itself emits in that part of the contour plot. Hence, HPLC needs to be applied for more definite interpretation of the 3D fluorescence results. Blue and turquoise textiles All blue and turquoise textiles were dyed with synthetic dyes (Table 3). Sample 11 was divided into two separate samples because the textile fibres themselves had differing textures. However, both samples were dyed with indigo carmine: two indigo carmine components were found. As indigo carmine is an acid dye, the response with the phosphoric acid gradient is very low. The ratio between these two components differs in these samples, which could indicate the use of a different dye bath. Furthermore, a small amount of an unknown red component was found in both samples. This colourant has a spectrum comparable with that of indirubin, a stereoisomer of indigotin (Schweppe 1992). Therefore, this component could be sulphonated indirubin, but positive identification is not possible due to the lack of reference material. The presence of an unknown red colourant in the presence of indigo carmine is often also recognizable on the 3D contour plot. Indigo carmine was also identified in four of the six blue samples. However, in these samples other blue or violet components were also found such as fuchsine (CI-42510) and methyl violet (CI-42535). These dyes belong to the basic violet dyes and are mixtures of triphenyl methanes, that is, three phenyl groups attached to one central carbon atom. Generally, basic dyes gave better results with the phosphoric acid gradient. The phenyl groups can have different side groups. Fuchsine consists of four components that are well separated with slightly different spectra. Methyl violet is a mixture of tetra-, penta- and hexamethylated pararosaniline. The hexa-methylated form is also known as crystal violet (CI-42555). See Figure 4 for the chromatogram of sample 14. In addition to the four fuchsine components, the three methyl violet components were detected. The combination of the three components indicates the use of methyl violet. The ratio between the blue indigo carmine and the violet fuchsine and methyl violet cannot be calculated at the moment because they were identified with different HPLC systems. Quantification can be done with the use of accurately weighted standards of the colourant, but this is beyond the scope of this paper. The constitution of the dyestuffs in sample 15.1 was very complex. Besides indigo carmine, fuchsine and methyl violet, numerous colourants were found with spectra similar to fuchsine and methyl violet but eluting at different retention times. This means that these components are closely related and could be by-products. These components were not found in the other samples. Since the amount of sample and dye is approximately the same in all blue samples, these by-products indicate another dye bath. Diamond green G (CI42040) and diamond green B (CI-42000) were found as well. The concentration of the green components was much higher than the concentration of fuchsine and methyl violet. The ratio between indigo carmine and diamond green cannot be calculated as the analytical procedure for both dyes is different. From the same area, a second sample was taken which was faded even at the back of the object. Sample 15.2 was dyed with a mixture of indigo carmine and diamond green G and B. Basic violets, such as fuchsine and methyl violet were not present, it could be that these were completely faded. The fourth blue sample where indigo carmine was found, sample 17, was dyed with colourants with spectra similar as methyl violet. However, methyl violet was not found in this sample. These colourants remain unidentified but belong to the class of basic violets. Two blue samples, 16 and 18, were dyed with water blue IN (CI-42780). Despite the fact that this is an acid dye, the result obtained with the phosphoric acid gradient was significantly improved compared with the result of the analysis with TBA gradient because this dyestuff contains basic groups as well. The colour difference can be explained by the concentration of the dyestuff present. In sample 16, the concentration dyestuff was about eight times higher than in sample 18, which was pale blue. Table 3. Analytical results of the turquoise and blue yarns Figure 4. HPLC chromatogram of sample 14. The dye consists of a mixture of four components, indicating fuchsine and three components that point to the use of methyl violet Figure 5. Detail of discolouration of the blue area of the tree between the two women. A sample taken from the back is shown with the area at the front to illustrate the discolouration With 3D-fluorescence, indigo carmine was detected on both turquoise samples (11.1 and 11.2), while the blue samples 14, 16, 17 and 18 showed an identical contour plot from which Water blue IN could clearly be identified. On blue samples 15.1 and 15.2, a combination of Diamond green B or G and indigo carmine were detected. Taking into account that the identification of indigo carmine is impossible in the presence of Water blue IN because of interference of the last one, and that minor colourants risk being undetectable in the 3D-contour plot, these results are very close to what is found with the chromatographic techniques with the exception of samples 14 and 17. In sample 14, indigo carmine, fuchsine and methyl violet were identified with HPLC, and in sample 17 indigo carmine and an unknown basic violet, but clearly not water blue IN. This inconsistency could be because dyes were mixed with closely related spectra. All the turquoise and blue samples were seriously faded to a greyish-green colour (Figure 5). The area where sample 15 was taken was a little bit more greenish compared with other areas, presumably because of the presence of diamond green. Miscellaneous coloured yarns The results of the mauve, purple, brown and cream samples are presented in Table 4. The two mauve samples were dyed with fuchsine; the colour difference observed is caused by the fact that the darker mauve contains a double concentration of fuchsine. Fuchsine was also found in the purple sample, together with methyl violet. Both purple and mauve areas were seriously faded. As the purple area was more faded than the two mauve areas, one can conclude that methyl violet is less light-fast. In the brown and light brown yarns, samples 22 and 23 respectively, picric acid and indigo carmine were found. The amount of indigo carmine in relation to the amount of picric acid in these samples is approximately twice as low as in the green samples 12 and 13. For that reason, the picric acid predominates, resulting in the colours observed. Furthermore, the concentration of dyestuff in the brown sample was higher then in the light brown sample, resulting in a darker colour. Finally, two samples were analysed which were almost colourless. Dyestuffs were found, but at a very low concentration. As a result, it was not possible to identify the red dyestuff found in sample 25, but it is different from sample 24, where cochineal was found. The red dyestuff in sample 25 is most likely of anthraquinone type dye; whether it has a natural or a synthetic origin remains unknown. The 3D-fluorescence analysis revealed the same main synthetic dyes as discussed above, except for the almost completely faded samples 24 and 25 where it was no longer possible to detect any colourant. Conclusion 3D fluorescence spectrometry has proved to be very useful as a first scanning technique for synthetic dye identification before destructive techniques are applied. However, as the interpretation of a 3D contour plot is not always reliable because of interference between the colourants and the wool fibre, and as minor colourants risk staying undetected, it should be applied complementary to chromatographic techniques. The use of the phosphoric acid gradient showed excellent results for natural and basic dyes, whereas the acid dyes showed much better results with the TBA gradient. As many different dyes were found, the combination of both HPLC techniques and 3D fluorescence spectrometry is preferred when sample size allows. Most yarns were dyed with synthetic dyes, although some natural dyes were found as well. The colours observed are consistent with the dyes found. Mixtures and variations in concentration were applied to achieve a specific colour or shade. It is interesting that different dyeing processes, such as acid, basic and mordant dyeing were used on a single yarn to obtain the final colour. The embroidery is severely faded and particularly the balance between the different colours is lost since dark colours, such as blue, purple and green are much more faded then the yellow, pink and red colours. Generally, the fading is stronger when the concentration dyestuff is lower. Based on the results of this examination, it can be concluded that the original appearance was very different from its present one. Notes 1 Colour index (CI) numbers of the synthetic dyes identified are derived from the Colour Index 1971, 3rd edition, Society of Dyers and Colourists, Bradford, England. Table 4. Analytical results of the miscellaneous coloured yarns References Christie, R M (ed.), 2001, Colour Chemistry, Cambridge, RSC Paperbacks, 3. Lehne, G, 1893, Tabellarische Uebersicht über die Künstlichen Organischen Farbstoffe, Hermann, Berlin, Germany, V. Pietro, A, 1998, ‘Synthesis of carminic acid, the colourant principle of cochineal’, Journal of the Chemical Society, Perkin Transactions 1, 575–582. Rabe, J G, Bischof, M and Fischer, C-H, 1990, ‘Natürliche und synthetische farbstoffe in Teppichen und Flachgeweben’, Restauro 3, 189–194. Schweppe, H, (ed.), 1992, Handbuch der Naturfarbstoffe, Landsberg, Germany, Ecomed, 282–318. van Bommel, M R, Wallert, A M, vanden Berghe, I and Wouters, J, 2005, ‘3Dfluorescence and HPLC analysis of early synthetic dyes’, Analytical Letters, submitted. Whiting, M, 1978, ‘The identification of dyes in old oriental textiles’ in Preprints of the 5th Triennial Meeting of the ICOM Committee for Conservation, Zagreb, 2 September 1979. Wouters, J and Verhecken, A, 1989, ‘The coccid insect dyes: HPLC and computerized diode-array analysis of dyed yarns’, Studies in Conservation 34, 189–200. Wouters, J and de Craecker, G, 1992, ‘Analysis of dyes and technique of a 20th-century Iranian carpet’, Journal of Dyes in History and Archaeology 11, 38–47. Wouters, J and Rosario-Chiniros, N, 1992, ‘Dye analysis of pre-Columbian Peruvian textiles with high-performance liquid chromatography and diode-array detection’, Journal of the American Institute for Conservation 31, 237–255.