Introduction

Industrial archaeology and the analysis of specimens manufactured some centuries ago are of great interest for understanding the changes that have taken place in industry [1]. At the same time, these changes reflect in some way the transformation experienced by society, especially since the industrial revolution. These kinds of artifacts have been underestimated sometimes by many scientists as a target of analysis and research projects. This is the case with wallpapers and the wallpaper industry. It must be taken into account that even objects of industrial origin that covered basic functions such as decoration, by current standards, should be considered as artworks, at the same level as engravings, lithographs, etc. The analysis of wallpapers or other industrial artifacts can provide very valuable information about how manufacturers introduced the newly available products and techniques into their manufacturing processes.

Scientific analysis of art and decorative artifacts can provide many surprises leading to the discovery of forgeries or repaintings. In the case of the most modern pigments, there is a period of time between the year in which they were discovered and the year in which they were accepted as pigments, or in which they started to be produced in factories and thus, started to be in widespread use. Since the 18th century this information has usually been recorded. In this sense, the use of a pigment during the transition period of time would constitute an early use of it. In contrast, the use of a pigment some years later after other one replaced it would constitute a late use of it.

In this work, a complete analysis of a wallpaper from the 19th century is reported, in which an unexpected pigment was determined. The analysis was performed by combining the information obtained by non-invasive micro-Raman spectroscopy and micro-XRF spectroscopy in a completely non-invasive way.

Raman spectroscopy has been successfully used in the study of many different kinds of artworks, such as medieval wall paintings [2, 3], painting on canvas [4, 5], glass and ceramics [6, 7], prehistoric art [8], and architectural paint analysis [9] among other many recent applications. Regarding studies of artworks on paper, Raman spectroscopy has become a very powerful tool. In the most recent studies, Raman spectroscopy has been used in the analysis of pigments and materials present in postage stamps [10], manuscripts [11–13], antique Bibles [14], lithographs [15], etc., but only a few recent chemical studies of wallpapers using Raman spectroscopy can be found in literature [1, 16, 17]. Other works have been published on wallpaper but on restoration processes [18] and historical aspects [19, 20]. The work by D. Wise et al. [21] must be highlighted, which explored the application of Raman spectroscopy to the studies of works on paper and discussed the benefits that a detailed knowledge of inks and pigments can have for both conservators and curators.

With regard to the use of XRF in the analysis of artworks, it is possible to find in literature some interesting recent applications. For example, XRF has been applied in the studies of pottery, ceramics and raw materials used in this kind of artworks [22–24], in the identification of forgeries [25], in the studies of the pigments present in mural paintings [26], coloured engravings [27], metallic objects [28], and historical aspects of pigment productions [29, 30]. However, only one paper has been found in the literature regarding the study of wallpapers by this analytical technique [31]. In the work by Bicchieri et al. [32] the foxing phenomenon of the paper support was studied by XRF, allowing the determination of the metal chips and ions responsible for this degradation process in a non-destructive and easy way. Several recent reviews can be found in the literature summarising the most important applications of XRF to the service of artwork studies, conservation and restoration [33, 34], proving the suitability of this technique for non-destructive analysis. However, some authors [35] argue that XRF technique is not always non-destructive, especially on materials such as pigments and in particular on organic compounds such as cellulose.

It should be pointed out that artworks are almost never analysed only by XRF technique, that is, this kind of study is complemented with other atomic techniques (XRD, PIXE, etc.) or other molecular techniques (Raman, FTIR, ATR, etc.). For instance, some authors have recommended the use of XRF and Raman spectroscopy together in the fields of artwork, archaeological and forensic studies [36, 37] as a good symbiosis between atomic and molecular analyses.

Following that recommendation, the work presented in this paper was designed in order to show the advantages and disadvantages of combining the analytical information given by both techniques. We decided to proceed on a specimen of wallpaper due to our experience with this kind of artwork.

Instrumentation

The current state of the art in modern analytical instrumentation led us to select two portable commercial micro-systems: Raman and XRF. We chose this alternative, thinking of future works on paper, items of invaluable significance, as well as about other kinds of artwork.

The Raman analysis was performed using a Renishaw Raman portable fibre optic micro-probe with a 785 nm laser and a CCD detector. In order to avoid thermal decomposition of the samples, laser powers between 50 and 5 mW were used (powers at the source), with integration times between 100 and 300 s and one scan per spectrum. The micro-probe was coupled to a 20× objective that allows one to focus on a simple grain. As the smallest vibration on the microprobe would lead to loss of focus and diminution in the signal-to-noise ratio, the probe was set on a special stage. The micro-probe incorporates a micro-TV camera allowing the user to focus on the sample and view it magnified. With this system it is even possible to distinguish between two different grains in a mixture of pigments, as will be seen later in the analysis of the green colours.

In the case of X-ray fluorescence spectroscopy (XRF), a Röntec ArtTax portable micro-XRF equipment was used. The instrument consists of an X-ray tube with Mo anode working at a maximum voltage/current of 50 kV/0.6 mA and using a special Xflash detector (5 mm2). The X-rays are collimated by a Tantalum collimator with a diameter of 0.65 mm. The measuring head of the equipment implements a CCD camera that allows one to focus on the sample using a motorised XYZ positioning unit controlled by the computer.

As the Raman and XRF analysis carried out in this study does not require any sampling, there was no restriction on the number of recorded spectra (mapping versus sampling). In order to be sure of the spectra of a given colour shade, several analyses in different areas of the same shade were made. Thus, the spectra shown in the figures below are really representative of the different colour shades observed in the wallpaper.

Specimen

The wallpaper analysed represents a hunt scene. According to the literature [38], this wallpaper was manufactured by Jacquemart and Bernard (Paris) using a drawing by Carle Vernet (1758–1836); it is entitled as Chasse de Compiègne. It is divided into 25 pieces: 1–5 Départ pour la Chasse devant les grilles du Château de Compiègne. Calèche de la Reine de Naples, 6–11 Poursuite du Cerf, 12–16 Passage de la Rivière, 17–20 La Curée and 21–25 Déjeuner sur l′herbe. Au Fond, les Ruines du Château de Pierrefonds. It is considered one of the most important and beautiful wallpapers ever manufactured. A piece of the whole wallpaper can be seen in Fig. 1. The work is of artisan manufacture, created on hand-crafted paper, superimposing the different colours until the whole drawing has been completed using the block-printing technique (xylography). The wallpaper presents a rich variety of colours and shades: oranges, black, white, blues, greens, yellows, pinks, reds, greys and browns.

Fig. 1
figure 1

A view of a piece of the wallpaper after restoration

Full size image

The first edition of the work is from 1812, with re-editions after 1815. Until 1815 the horsemen’s jackets were painted in red, but when the Napoleonic Empire collapsed in 1814, red jackets were banned for political reasons, and manufacturers had to paint them in blue. The wallpaper analysed in this work presents the jackets in red, thus, it belongs to the first edition. This wallpaper can be admired in Paris, Vienna, Italy, the Victoria and Albert Museum (UK) and in the Basque Country (Spain).

In our country, the wallpaper was found in two places, a room of a house-palace that belonged to the Gastañaduy family located in the town of Ezkoriatza (60 km east from Bilbao, northern Spain) and also in a room at the Torre de los Varona, a tower-palace located in the town of Villanañe (50 km south-west of Bilbao), which has belonged to the same family for 1,300 years. The house palace of Eskoriatza was in a bad state of conservation during the second half of the 20th century, with some rooms directly exposed to the ambient conditions. The whole restoration of the palace started at the end of the 1990s, including some cleaning and conservation of the wallpaper of this work [39].

Results

The results obtained are presented below arranged by colour. In the case of Raman analysis, all spectra have been compared with spectra collected from standards with the same equipment and in the same condition, which are grouped in a large database [40]. The elements present in each XRF spectra were defined by working with ArtTaxCtrl software provided with the ArtTax system.

Oranges and reds

Micro-XRF analysis of the orange colours (Fig. 2) showed the presence of Fe, Hg, Pb, Ca and S, and thus, HgS, Fe2O3, CaCO3 and Pb3O4 could be present. Table 1 summarises the relationships (peak area) among all determined elements in the orange shades. The Raman spectrum of the very light orange (Fig. 3) shows bands at 550, 391 and 316 cm−1 that are consistent with the presence of red lead (Pb3O4), bands at 344, 287 and 254 cm−1 that are due to the presence of vermilion (HgS) and a weak peak centred at 1085 cm−1 that is due to calcium carbonate (CaCO3). The spectrum of the light orange colour shows the same bands as very light orange, with the band of the calcium carbonate very weak due to the low concentration of this pigment in this shade. The spectrum of the medium orange (not shown in the figure), and the spectrum of the dark shade show clear peaks at 344, 287 and 254 cm−1, due to vermilion, and a very weak peak at 550 cm−1 that is due to the presence of a little amount of red lead. However no iron pigment was detected with Raman spectroscopy, probably due to the low quantity present in the mixture as well as due to overlapping effect with the peaks of red lead and vermilion, but it is clear that a little amount of red iron oxide would be present in these two dark shades, as determined by the XRF analysis.

Fig. 2
figure 2

XRF spectrum of orange colour

Full size image
Table 1 Relationships among all elements in orange and red shades
Full size table
Fig. 3
figure 3

Raman spectra (1,250–200 cm–1) of very light orange (i), light orange (ii) and dark orange (iii)

Full size image

The XRF and Raman spectra provide us additional information on the role of each pigment in the different shades. From the data in Table 1, the proportion Pb/Hg is almost eight times higher in the light shade than in the dark one, whereas the proportion of Hg/Fe is four times higher in the medium shade than in the dark one. The proportion Pb/Hg in medium and dark shades is almost the same, but according to the ratio Hg/Fe and Pb/Fe, the dark shade was obtained by adding iron (Fe2O3). In the light shades there are not appreciable amounts of iron.

With regard to calcium, the highest amount of this element is in the very light shades (see Pb/Ca and Hg/Ca ratios), if we compare with the ratios in dark shades, where the amount of calcium is very low. For light and very light shades, the proportion Pb/Hg is very similar, but the very light shade seems to be obtained by adding 10 times more calcium carbonate than in the light shade, according to the Hg/Ca and Pb/Ca ratios.

XRF spectra of the red colours present the same elements as the orange colours but in different proportions, as can be observed in Table 1. The relationship Hg/Pb is approximately the same for all shades, whereas the relationship Hg/Fe and thus Pb/Fe is approximately the same for the light and medium red colours but is almost four times less than for the dark shade. In contrast, the light shade has a higher amount of calcium than the medium and dark shades, with the ratios Pb/Ca and Hg/Ca being almost the same. Thus, light red was obtained by adding calcium (CaCO3) to medium shade, whereas dark red was obtained by adding iron (Fe2O3) to the medium shade.

In the Raman spectra (not shown) of the red colours (horsemen’s jackets), only those peaks of the vermilion pigment (344, 287 and 254 cm−1) can be seen, as well as the main peak of minium (very weak band at 550 cm−1), but it is clear according to XRF analysis that a certain amount of calcium carbonate and red iron oxide is present in the other red shades. XRF analysis showed that the orange colours have a higher proportion of Pb than red colours, that is, three to seven times more lead depending on the shade, whereas red colours have a higher proportion of Hg: three times more Hg depending on the shade.

The mixture of vermilion with minium has a rich historical background according to the literature. As well, there is historical confusion and controversy over the use of the term minium [41]. It seems that both pigments have been used since antiquity mixed in different proportions for different reasons (economical factors).

Pinks

Two different shades of pink colour are present in the wallpaper. According to their XRF analysis, calcium and mercury were found. Semi-quantitatively, XRF analysis showed that in both shades there is almost the same amount of calcium (the light shade has a little more calcium than the dark one), whereas the dark shade has twice the XRF signal of mercury than the light one. According to Raman spectra (not shown), the pink colour is a mixture of calcium carbonate (peak at 1085 cm−1) and mercury vermilion (peaks at 344 and 254 cm−1). The lighter shade presents a more intense peak at 1085 cm−1 than the darker shade. The results obtained by Raman spectroscopy corroborated the data obtained with the XRF system.

Blues

Many areas of the wallpaper are painted in several shades of blues including parts of the ground and the forest. With regard to the XRF analysis of the blue colours, in all shades there are significant amounts of calcium, iron and in the very light blue colour there is a high signal of lead, and thus, Prussian blue, CaCO3 and lead white could be present. The dark shade is the colour with a higher concentration of iron, and less calcium. For instance, the proportion Ca/Fe is seven times higher in the light shades than in the dark one. Light and very light shades present, semi-quantitatively speaking, the same amount of calcium and iron, but the very light shade has a significantly higher signal of lead than the light shade. This metal probably comes from lead white. However, no Raman record of this pigment has been obtained, probably due to its low concentration in the colour. Medium and dark blues have almost no signal of lead.

The Raman spectra of four different shades of blue always show bands at 2153, 2116, 2091, 949, 597, 531, 507 and 471 cm−1 that belong to Prussian blue (Fe4[Fe(CN)6]3), whereas the peak at 1085 cm−1 is due to calcium carbonate. The feature at 275 cm−1 would be a contribution of both Prussian blue and calcium carbonate. It seems that different proportions of calcium carbonate and Prussian blue were mixed to obtain all shades as XRF analysis has determined, except the very light one, where lead white was also added.

Yellows

XRF analysis of yellow shades provides a curious result. The dark shade presents more signal of calcium than the light one, but the ratio of the XRF signal for calcium to iron (Ca/Fe) is approximately 3 for the light shade and 2 for the dark one, which would explain the difference between the two shades. As calcium and iron were determined by XRF, yellow iron oxide and CaCO3 could be present in this colour.

Although the Raman spectra of the two yellow shades present a high background fluorescence, it is possible to determine that the yellow pigment used is a yellow iron oxide (Fe2O3.xH2O) according to the peaks at 397 and 300 cm−1 that are present in the two spectra. The spectra of both shades have another weak feature at 1085 cm−1 that is due to the presence of calcium carbonate. In addition, the spectrum of dark yellow presents two broad bands at 1600 and 1315 cm−1 that are consistent with the presence of carbon black (amorphous C). Unfortunately, XRF cannot identify amorphous carbon detected by Raman spectroscopy in the dark yellow colour, which according to Raman analysis is a mixture of yellow iron oxide, calcium carbonate and carbon black.

White

The wallpaper presents areas that are painted in white, such as dresses, the water of the river, and some horses. XRF analysis of white colour shows the presence of calcium and lead, coming from calcium carbonate and lead white, as Raman spectroscopy corroborates. All the white areas analysed present the same Raman spectrum, where it is possible to distinguish two bands centred at 1085 and 1050 cm−1. These two peaks are due to the stretching mode of carbonate anion in calcium carbonate (CaCO3) and lead white [2PbCO3.Pb(OH)2], respectively.

Black and greys

XRF analysis of the black colour presents a little signal of calcium. This fact could be strange or a surprise, but according to the Raman spectrum of this colour (Fig. 4, i), the pigment present in the sample is carbon black (broad bands centred at ca. 1600 and 1315 cm−1), and shows a very weak band at 962 cm−1, due to the presence the calcium phosphate [Ca3(PO4)2]. This compound is usually related to the presence of carbon black of animal origin, thus, the carbon black is in fact bone black. It is usually difficult to distinguish between carbon black of animal origin and vegetal origin because the signal of calcium phosphate is very weak, and it is not always possible to see the peak at 962 cm−1. In this study, however, it was detected using a portable Raman system with a 20× long-distance objetive in the microprobe.

Fig. 4
figure 4

Raman spectra (2,000–500 cm–1) of black (i), dark grey (ii) and light grey (iii)

Full size image

XRF analysis of greys shows the presence of a strong signal of calcium, probably coming from the contribution of calcium carbonate (mainly) and calcium phosphate. The Raman spectra of the grey colours (Fig. 4, ii and iii) show the bands of carbon black and a peak at 1085 cm−1 that is due to the presence of calcium carbonate. The light grey colour seems to have more calcium carbonate than the black grey one according to the intensity of the peak at 1085 cm−1. The Raman spectra of grey shades show a very weak band at 962 cm−1, due to the presence of calcium phosphate [Ca3(PO4)2].

Another two bluish greys can be found in the forest. XRF analysis of these shades of bluish greys shows the presence of iron and calcium, coming probably from Prussian blue and calcium carbonate. This hypothesis was corroborated by the results of Raman spectroscopy measurements. The Raman spectra of both shades (not shown) present the characteristic bands of Prussian blue (Fe4[Fe(CN)6]3), carbon black and calcium carbonate. The proportion Fe/Ca is approximately the same in both shades, thus, carbon black should be present in different proportions to obtain both shades. Unfortunately, XRF cannot detect the presence of carbon black.

Browns

In the three shades of brown found in the ground, there are calcium and iron coming from calcium carbonate and iron oxide. The light shade has a higher amount of calcium than the medium and dark shades, whereas the dark shade has a higher amount of iron, with the ratio of the XRF signal for calcium to iron (Ca/Fe) approximately 2 for the light shade and 1 for the medium and dark shades. Carbon black would be the pigment added to obtain the dark shade from the medium one, whereas calcium carbonate would have been added to the medium shade to obtain the light one.

In the three shades of the brown colour from the horse that were analysed, the same results were found as were obtained for the browns in the ground. The light shade presents a higher signal of calcium whereas the dark shade presents a higher signal of iron, with the ratio of XRF signal for calcium to iron (Ca/Fe) 0.10 for the light shade and 0.05 for the dark and medium shades. As in the case of the browns from the ground, carbon black would be the pigment added to obtain the dark shade, whereas calcium (calcium carbonate) would have been added to the medium shade to obtain the light one.

According to the Raman spectrum of the brown colours from a horse, there is evidence of the presence of a burnt sienna-type pigment (Fe2O3, haematite mainly), with peaks at 406 and 294 cm−1 [42]. In all the shades, calcium carbonate was found (1085 cm−1). In addition, the dark shade has carbon black (the two characteristic broad bands centred at 1600 and 1315 cm−1). The same fact is found for the brown shades from the ground, but instead of a burnt sienna-type pigment, a raw sienna-type pigment (Fe2O3.xH2O, goethite mainly) was used, showing bands at 397 and 300 cm−1 [42].

The relationship (Ca/Fe) can be used to explain why brown colours from the horse (Ca/Fe 0.10–0.05) are darker than brown colours from the ground (Ca/Fe 3–2). In addition, in the browns from the forest, burnt sienna was used, which is deeper in colour than the raw sienna used in the brown shades from the ground. XRF analysis does not distinguish between goethite and haematite, but due to XRF analysis it is possible to see the different proportions of earths and calcium carbonate used to prepare the different shades of brown.

Greens

According to the XRF spectrum of the very light green colour (Fig. 5), the presence of a high signal of lead, together with tin and copper, is clear. Raman spectrum of this shade of green colour (Fig. 6 i) shows bands at 975 and 323 cm−1. Through the micro-TV camera, it could be verified that this colour was in fact a mixture of a green-blue pigment and a yellow pigment. With regard to the green-blue pigment, there was evidence of the presence of a basic copper sulphate, Raman band at 975 cm−1, and the presence of Cu and S in the XRF analysis. It must be pointed out that the S peak in XRF may overlap with Pb K alpha or may even be a consequence of the presence of the binder. The nature of this pigment will be discussed later.

Fig. 5
figure 5

XRF spectrum of light green colour from the grass

Full size image
Fig. 6
figure 6

Raman spectra (2,500–250 cm–1) of the very light green from the grass (i), light green from the grass (blue grain) (ii) and light green from the grass (yellow grain) (iii)

Full size image

In terms of identifying the yellow pigment (band at 323 cm−1), none of the most common ones from the beginning of the 19th century fit the spectrum. The presence of tin determined by XRF suggests the presence of a tin yellow pigment, that is, lead-tin yellow type I (Pb2SnO4) and/or type II (PbSn1-xSixO3). At first these two pigments were dismissed, because according to the literature they were not used in the years in which the wallpaper was manufactured [43]. Surprisingly, the Raman spectrum of the lead-tin yellow type II (PbSn1-xSixO3) matched up with the spectrum of the yellow pigment found in the very light green colour [40].

In order to identify the green-blue pigment mixed with lead-tin yellow type II, several spectra on the green-blue grains were collected using the 50× objective. Figure 7 shows one of the recorded spectra. Recently Hayez et al. [44] presented a summary table with the characteristic wave number of Raman spectra obtained by several authors referencing basic copper sulphates used as pigments and/or found in artworks, such as antlerite [Cu3(SO4)(OH)4], brochantite [Cu4(SO4)(OH)6], posnjakite [Cu4(SO4)(OH)6·H2O], langite [Cu4(SO4)(OH)6·2H2O] and chalcanthite [CuSO4·5H2O]. Posnjakite, langite and chalcanthite are mainly blue in colour, whereas, antlerite and brochantite are mainly green.

Fig. 7
figure 7

Raman spectra (2,500–200 cm–1) of the green-blue grains of the pigment (Br brochantite and An antlerite) found in the very light green from the grass

Full size image

Taking into account the proposed assignments suggested in the work by Hayez et al., we assumed that our compound was in fact brochantite with a certain small amount of antlerite. These compounds not only match the position of the most intense Raman bands, but also their relative intensities in the other minor bands. On the one hand, the bands centred at 1069, 986 (strong shoulder), 636 (shoulder), 343 (shoulder), 296, 261 and 239 cm−1 could correspond to antlerite. On the other hand, bands at 976, 611, 398 and 313 cm−1 are consistent with the presence of brochantite. Finally, features at 601, 502, 479, 448 and 412 cm−1 would signify a contribution of both antlerite and brochantite. In the case of the peaks at 601 and 412 cm−1, the contribution of antlerite is higher than brochantite, whereas in the peak at 479 cm−1 the contribution of brochantite would be more important than antlerite. In addition, the spectra were compared with freshly synthetized brochantite (see below) and a sample of mineral brochantite.

With regard to the light green from the grass, XRF analysis shows that calcium and iron are present, thus, calcium carbonate, Prussian blue and yellow iron oxide could have been used in this shade, as found in the Raman analysis. Under the Raman system TV camera, two different types of grains can be distinguished in the light green from the grass. On the one hand, the Raman spectrum of the blue grain (Fig. 6 ii) shows peaks at 2153, 2116, 2091, 597, 531 cm−1 that belong to Prussian blue, and at 1085 and 710 cm−1 that are due to the presence of calcium carbonate. There is a feature at 275 cm−1 that is a contribution of both Prussian blue and calcium carbonate. On the other hand, the spectrum of the yellow grain (Fig. 6 iii) is consistent with the presence of a yellow iron oxide (peaks at 548, 479 shoulder, 397 and 300 shoulder cm−1).

In addition, XRF analysis shows the presence of copper, that is, it is possible that copper sulphate would have been also used in the green. This pigment was not detected by Raman spectroscopy, probably due to a poor Raman scattering and its low concentration in the sample. Thus, this green colour would be a mixture of Prussian blue, yellow iron oxide and calcium carbonate and the copper sulphate pigment.

In the forest, two different shades of greyish green can be observed. XRF analysis of these colours shows the presence of calcium, iron and copper. Both shades have the same quantity of calcium, but the dark shade has a little more iron than light one. Copper has been detected in these colours, thus, it is possible that the copper sulphate could be present. However any Raman spectra of this compound would have been able to be recorded. The light shade has a little more quantity of copper than the dark one.

The Raman spectrum of the dark greyish green from the forest (Fig. 8 i) shows bands at 2153, 2116, 2091, 597 and 531 cm−1 that are due to Prussian blue; two broad bands centred at 1600 and 1315 cm−1 that belong to carbon black; and a weak peak at 1085 cm−1 that is due to calcium carbonate. The characteristic band of phosphate at 962 cm−1 is also seen in the spectra, thus, the carbon black is again of animal origin (bone black).

Fig. 8
figure 8

Raman spectra (2,500–250 cm–1) of the dark greyish green from the forest (i) and very light greyish green from the forest (ii)

Full size image

In contrast, the Raman spectrum of the very light greyish green (Fig. 8 ii) has the bands of Prussian blue (very weak), carbon black, calcium carbonate and yellow iron oxide (at 397 cm−1). The absence of bands from a yellow pigment in the dark greyish green could be due to its low concentration in the colour. Probably, the yellow pigment used in the dark shade would be the same as in the light shade of greyish green.

Brochantite/antlerite

The presence of brochantite/antlerite in the sample requires a special explanation. According to the literature, basic copper sulphates have been found in artworks from the 15th and 16th centuries [45]. These kinds of compounds are very rare as natural minerals in Europe, thus, some authors state that they were probably obtained artificially [46]. However, there is a lack of recipes corresponding to the synthesis of these pigments before the 17th century [47]. In fact, the earliest recipe for posnjakite was given in a manual for the preparation of paints in the late 18th or early 19th century [45]. This last fact is very significant because the wallpaper is dated to the beginning of the 19th century, thus, these compounds would be available in high quantity by industrial synthesis.

As mentioned above, brochantite was synthesized in the laboratory to obtain a standard spectrum of this compound. The first intention was to obtain posnjakite using a recipe from the literature [46]. After the synthesis, Raman analysis of the compound was performed, but instead of recording the Raman spectrum of posnjakite, the spectrum of brochantite was always obtained. It must be mentioned that more than 40 Raman spectra were collected. Further XRD and thermogravimetric analysis confirmed that brochantite was the only mineral phase present. According to Marani et al. [48], the aging of the fresh basic salt in its mother liquor induces the complete conversion of posnjakite to the more stable brochantite, with posnjakite as a kinetic precursor of brochantite.

We performed a chemical reaction simulation using the chemical equilibria and database information on stability constants for all the basic copper sulphate minerals included in the software MEDUSA (Make Equilibrium Diagrams Using Sophisticated Algorithms) [49]. Figure 9 shows the distribution diagram of the copper-containing species for a theoretical situation where sodium bicarbonate (expressed as [OH–]tot) is added to a 0.05 M copper sulphate solution. As seen, when the stoichiometric amount of bicarbonate is added (75 mM), all the copper is in the form of a solid brochantite phase. This confirms as well the stability of brochantite versus posnjakite.

Fig. 9
figure 9

Fraction diagram of the synthesis of brochantite, when bicarbonate is added to a copper sulphate solution

Full size image

Synthesis was achieved by adding powdered sodium bicarbonate to a solution of copper sulphate with constant magnetic stirring. During the reaction, CO2 is generated because the initial pH of copper solution is around 4, thus, incompatible with bicarbonate, which has a thermodynamic acidic constant (pka) of ca. 6.3. When a stoichiometric amount of bicarbonate was added to form brochantite, the final pH of the solution was ca. 6. The solid phase was kept under stirring for 2 days in order to attain equilibrium. In this way, brochantite was the only solid phase formed.

As mentioned above, some amount of antlerite was found in some of the Raman spectra from the very light green of the sample. From the thermodynamic point of view it is possible to form antlerite from brochantite. The wallpaper was found in a very bad state of conservation, with high levels of humidity and with presence of fungi and other microorganisms. As well, the house had been uninhabited and without a roof for years. Moreover, close to the house there is a heavy industry (steel makers) that has been dumping SOx into the atmosphere for the last 120 years. The combination of this pollutant gas and humidity could have promoted a partial decay of brochantite to antlerite according to the reaction:

$${\text{3Cu}}_{{\text{4}}} {\left( {{\text{SO}}_{{\text{4}}} } \right)}{\left( {{\text{OH}}} \right)}_{{\text{6}}} + {\text{H}}_{{\text{2}}} {\text{SO}}_{{\text{4}}} \to {\text{4Cu}}_{{\text{3}}} {\left( {{\text{SO}}_{{\text{4}}} } \right)}{\left( {{\text{OH}}} \right)}_{{\text{4}}} + {\text{2H}}_{{\text{2}}} {\text{O}}$$

We have simulated this reaction again using the software MEDUSA [49]. If SOx is dissolved in water (humidity on the wallpaper) and it is oxidised to SO3, sulphuric acid would develop, reacting with brochantite and producing antlerite. This can be seen in Fig. 10 where sulphuric acid is added to an equivalent amount of 0.05 M brochantite. As seen, brochantite is transformed into antlerite when sulfuric acid is added. For example, if a paint layer of brochantite of 1 cm2 and 20 μm thick is considered, and we assume that 100% is brochantite (this is not true of course, it will be less because the paint layer is a mixture of the pigment and the binder), taking into account the density of the pigment (3.97 g/cm3), only 0.12 mg of H2SO4 would be necessary to transform 20% of brochantite into antlerite. The current pollution levels in the area show ca. 8 μg/m3 of SOx, thus, a reasonable enough amount to promote the decay of brochantite into antlerite.

Fig. 10
figure 10

Fraction diagram of the transformation of brochantite into antlerite when H2SO4 is present

Full size image

Finally, it can not be stated if brochantite was used deliberately, or if the manufacturer wanted to synthesize posnjakite but he obtained brochantite. Probably, Jacquemart did not take care about this.

Conclusions

The analysis is valid with regard to chronology because all pigments determined were available at the time in which the work was created (1812). Nevertheless, the presence of the lead-tin yellow type II can be considered a late use of it if, and in contradiction to what Hermann Kühn [43] says, this pigment was used for painting after the middle of the 18th century.

According to Hermann Kühn [43], it seems that lead-tin oxides started to be used in ceramic glazes before 1300, however, almost all artworks in which lead-tin yellow has been identified were created between about 1300 and 1750. In addition, it is believed that the term giallorino (Cennino) or massicot (used in northern Europe) present in many treatises and manuscripts refers in many cases to lead-tin yellow and not to lead oxide (PbO). There is misinformation that has lead to the erroneous idea that PbO was widely used in oil painting. In many cases those identifications have come from microchemical tests for Pb, thus, a positive result of PbO was inferred. These identifications have not been supported by more advanced analytical instruments, and so, many of them are probably wrong.

According to the current state of knowledge, authors report an out of time use of lead-tin yellow type II in a wallpaper from 1812 by Jacquemart. It must be pointed out that this pigment has been detected in a semi-industrial manufactured decorative product, thus, lead-tin yellow type II was used as a common raw material by Jacquemart. However, there is much reason to believe that the use of lead-tin yellow type II may have been an accident or used simply as a trial in a single run of printing. It would be very interesting to analyse other artworks by this company, or the same wallpaper from other locations in order to clarify this controversy.

The analysis of greens is usually difficult because in many cases unstable pigments are involved, such as copper acetates, copper sulphates and carbonates, copper arsenates, etc., whereas in other cases complex mixtures of green, blue and yellow pigments are used. Surprisingly, it was difficult to find a green colour in any of the wallpapers analysed by the authors that was composed only of a single green pigment. For instance, the green colours present in this artwork show complex mixtures of pigments: lead-tin yellow type II and brochantite/antlerite on the one hand and Prussian blue, yellow iron oxide, carbon black and calcium carbonate on the other hand.

The difficulty of determining the presence of basic copper sulphates has been shown. These types of compounds are chemically very close and can be found together. In addition, it is always problematic to state whether the determined material is the original pigment used or a degradation product. However, with the help of chemical equilibria and database information on stability constants, it is possible to simulate the chemical system in which all these compounds are involved, resulting in a reliable explanation of the experimental data obtained, in this case, by Raman spectroscopy and XRF.

Recently, authors have analysed a wallpaper named Les Monuments de Paris manufactured by Dufour (Paris) in 1812 [17]. In that wallpaper some of the pigments found in Chasse de Compiègne such as Prussian blue, CaCO3, HgS, red lead, lead white, carbon black (but of vegetable origin), raw and burnt sienna and yellow iron oxide were also detected. However, in Les Monuments de Paris other different pigments were found, such as red iron oxide, Scheele’s green and gypsum. These differences could be considered a consequence of the competition between the two most famous wallpaper manufacturers in Paris, at the beginning of the 19th century, in the search for beauty colour palettes.

The complementarity of non-invasive and non-destructive micro-XRF and micro-Raman analysis has been demonstrated in the study of the pigments present in an artwork on paper support, and all colours have been successfully determined. Sometimes, when a specimen is analysed by Raman spectroscopy it is difficult to ensure that all pigments and materials present in the artwork have been identified. In many cases, analysis is made difficult by the low scattering properties of some of these materials, as well as the low concentration (the material is below the limit of detection), the overlapping of bands and peaks or the inherent fluorescence phenomena of Raman technique even when infrared lasers are employed. This problem can be partially overcome if an earlier elemental analysis is performed.

The methodology that has been followed in this study consists of a screening of the element present in the different parts and colours of the specimen using a portable non-invasive XRF system. In this way, it is possible to determine what kind of materials and pigments are present (iron pigments, cobalt pigments, calcium pigments, etc). Afterwards, the specimen is analysed by Raman spectroscopy in order to know the molecular composition of the pigments and materials, that is, how the elements determined by XRF are linked among them. If the elemental composition is known, it is possible to create a tentative list of pigments that could be present. For example, if in a green colour, arsenic, iron and copper are determined, the operator knows that iron yellow oxide, Prussian blue, Scheele’s green, emerald green, azurite, malachite, verdigris or other copper green pigments could be present, whereas for example, no cobalt, lead or manganese pigment would be present. This is very useful for interpreting the Raman spectra because the assignment of Raman signals is more reliable, and we ensure that all elements are related to a pigment or material, thus, all pigments are determined in an elemental and molecular way.

As an example of this complementarity, the identification of some of the pigments determined ambiguously by Raman spectroscopy should be highlighted, such as the presence of the lead-tin yellow type (II) in the light green colour that was corroborated by XRF with the determination of tin in this shade.

The presence of twelve pigments in the manufacture explains the beautiful chromatic richness that the wallpaper shows (see Table 2). The variety of shades of greens and red-oranges present in the artwork, in which different mixtures of different pigments have been found should be mentioned. In addition, the use of both burnt and raw sienna mixed with black carbon of animal origin (bone black) or calcium carbonate to obtain many different shades of browns must be emphasized. Another important feature of this analysis consists is use of XRF analysis to show semi-quantitatively how the manufacturer created the mixtures of different pigments to obtain different shades of the same colour.

Table 2 List of the pigments determined
Full size table