1 Introduction

The use of Chinese wallpapers to decorate the walls of the so-called Chinese rooms in castles, palaces or wealthy manor houses in Europe became very popular during the second half of the eighteenth century [1]. They represented an expensive, rare novelty and reflected the status and the personal taste of the owners of the houses [2, 3]. The Portuguese imported the first Chinese pictures in the sixteenth century, followed by the Dutch in the seventeenth century [3]; since then, the western taste for Chinese graphic art has spread in Europe. This new fashionable trend led to the production of wallpapers imitating the Chinese motifs by European craftsmen at the beginning of the eighteenth century. It was only then – as suggested by Emile de Bruijn [4] – that Chinese painting workshops responded to this western demand, by producing wallpapers specifically for export to Europe [5, 6]. In contrast to other goods imported from China, though, wallpapers remained a small-scale private trade; therefore, Chinese wallpapers kept their identity of rare, expensive and luxury products up to the present day [5, 6].

To date, few works have been published on scientific studies on wallpapers. Castro and collaborators [7,8,9,10,11] focused on the use of non-destructive spectroscopic techniques – specifically, Raman and XRF spectroscopy – to identify pigments, binders, and investigate on degradation issues on some nineteenth century European wallpapers. In some cases, the analyses have been performed on fallen fragments [7, 8, 10], thus excluding the need for sampling. As the analyses were completely non-destructive, the fragments could be consequently restored and reintegrated in the wallpapers. In other cases, no sample nor fragment was collected, resulting in a completely non-destructive and non-invasive approach [9, 11]. Literature on technical studies on Chinese wallpapers is even more scarce. In their work, Mairinger et al. [12] report an example of deterioration of paper due to the presence of copper pigments on a fragment of Chinese wallpaper. Pessanha and collaborators [2] carried out a multi-analytical study on a hand-painted Chinese wallpaper from the eighteenth century belonging to a private Portuguese collection. They identified the palette and the main fibres used for the production of the paper sheets, by means of a combination of non-destructive techniques (namely XRF, confocal Raman spectroscopy and optical microscopy). Most of the other studies published on this topic so far, though, have mainly focused on conservation strategies [5, 13,14,15].

During the past few decades, the combined use of in situ non-invasive techniques has been successfully assessed to analyse paintings on light, fragile, precious support such as paper and parchment. Techniques such as Raman spectroscopy, X-ray fluorescence (XRF), fibre optic reflectance spectroscopy (FORS), hyperspectral imaging (HSI) – by means of portable equipments – have proved their effectiveness in the analysis of pigment palettes in books, illuminated manuscripts [16,17,18,19,20,21], East Asian prints and paintings [22,23,24,25,26,27,28,29,30].

In this framework, this present work aims to study a set of hand-painted Chinese wallpapers belonging to the National Museum of Ancient Art in Lisbon (Museu Nacional de Arte Antiga – MNAA), by means of in situ non-invasive techniques. More specifically, the combined use of EDXRF, Vis–NIR FORS, and technical photography – namely ultraviolet fluorescence, visible and infrared photography – allowed to investigate the painting technique, to define the use of pigments and provide a preliminary conservation condition survey.

The results presented in this work represent a first step towards the investigation and characterisation of the constituent materials and degradation products of the wallpaper panels belonging to the Museum, in the context of a broader research project which will also involve micro-destructive analyses. The outcomes of this project will provide useful data in view of the restoration intervention and the future integration in the permanent exhibition of the Museum; moreover, it will contribute to explore trade and cultural exchanges between Europe and China, through the study of what is considered a unique and hybrid product of global trades and connections.

Furthermore, although nowadays many Chinese wallpapers can still be found in situ, both in Portugal and in other European countries (mainly England Ireland and Netherlands), it is rare to find them in museums, where usually only single panels are displayed [15]. Therefore, the study and the consequential exposition of the wallpapers represents an extraordinary occasion for the National Museum of Ancient Art to display a collection unique in its kind in the whole country.

2 Materials and methods

2.1 The wallpaper panels

The set of wallpapers belonging to the National Museum of Ancient Art in Lisbon is comprised of six paper panels from the eighteenth century, donated to the Museum in 1949. Before their donation to the Museum, the panels were probably hanging on the walls of a Chinese room in a Portuguese palace; the documents related to the donation of the wallpapers to the Museum, though, do not mention any information about the origin of the set and its arrival to Europe. The height of the panels is approximately 345 m, with widths varying from 0.68 to 1.20 m for a total linear width of 524 m; the size of the set is then consistent with the values reported in literature [1, 6, 15].

The panels represent the process of production of porcelain, from the collection of the raw materials to the actual making and selling of objects (vessels, plates, vases); they fit then into the category of wallpapers decorated with scenes depicting manufacturing activities. There were, indeed, three main themes represented in Chinese wallpapers: nature motifs, everyday life scenes (such as agricultural and manufacture activities), and a third category depicting combinations of human figures and flowering plants [1, 5, 31]. Figure 1a shows one of the panels of the set (panel number 3).

Fig. 1
figure 1

Full image of Panel n.3 (345 × 120 cm ca.) (a) and magnified images of one of the human figures showing details of the drawing and painting technique (b, c). d IR photography of the area depicted in b

Full size image

Currently, the six panels are glued on big sheets of brown paper (see for example Fig. 1a at the top, bottom and left edge of the panel; Fig. 3a–d, where the brown paper shows below the sheets of wallpaper). There is no record about their donation to the Museum, nor about their condition at the time; it is likely, though, that the panels have been glued on the sheets of brown paper when they have been removed from their original location to be donated to the Museum. Once they became part of the collection of MNAA, they have been kept in the Museum store in horizontal position. For the in situ campaign, the panels have been moved to an exhibition area of the Museum, where they have been placed on top of thin sheets of expanded polyethylene on the floor. In this way, the panels have always been kept in the same horizontal position, minimizing the stress; the setup for the analyses was defined accordingly.

2.2 In-situ non invasive techniques

2.2.1 Technical photography

Due to the dimensions of the panels, the setup for the acquisition was determined prior to the in situ campaign in the National Museum of Ancient Art. More specifically, for each panel, a specific number of quadrants was determined according to their width: four quadrants for the narrowest ones (panels 1 and 6), and eight for the largest ones (panels 2–5). For each quadrant, ultraviolet fluorescence (UVF), visible (Vis) and infrared (IR) photography has been performed. The number of quadrants was defined so that all the corresponding photographic acquisitions had overlapping areas – more precisely, around 40–50% of the total area of each quadrant.

A Nikon® D3400 camera equipped with a Si-CCD sensor and an AF-P DX Nikkor 1:3.5–5.6G zoom lens by Nikon was used for Vis and UVF photography. Two Labino Standard MPXL 35W UV lights were used for the UVF acquisitions. IR photography was carried out with a Nikon D3100 camera (modified to remove the internal IR-blocking filter) equipped with a Si-CCD sensor. A X-Nite 1000 nm filter (30 mm Diameter × 2 mm Thick) by LDP LLC—MaxMax was applied on the objective to block the radiation with wavelength shorter than 1000 nm. The camera was equipped with the same objective used for UVF and Vis photography. An X-Rite, ColorChecker Passport Photo 2 was included in each acquisition for the correction of white in post-processing.

The acquired photos have been post-processed in Adobe Lightroom for the calibration of white and the corrections related to optical aberrations, and subsequently merged with Adobe Photoshop (Fig. 1a).

2.2.2 Energy dispersive X-ray fluorescence spectroscopy (EDXRF)

X-ray fluorescence analyses were performed by means of a Tracer III-SD Fluorescence handheld Energy Dispersive X-Ray Fluorescence (h-EDXRF) spectrometer from BRUKER, equipped with a 10 mm2 XFlash® SDD, a peltier cooled detector with a typical resolution of 145 eV at 100,000 cps and a Rh target. The instrumental parameters were as follows: 40 kV and 11 μA, without filter, acquisition time of 30 s, and a spot size of 12 mm2 (3 mm × 4 mm). All the spectra were acquired using the S1PXRF software and subsequently treated with the ARTAX software.

2.2.3 Fibre optic reflectance spectroscopy (FORS)

FORS analysis was performed using a portable spectrometer B&W TEK i-Spec® 25 equipped with a Si PDA/InGaAs/Extended InGaAs detectors for measurements across the Vis/NIR range from 345 to 2500 nm. More precisely, the measurements were performed in the ranges of 350–1000 nm, 900–1700 nm, and 1550–2500 nm with an approximate resolution of, respectively, 4.0 nm, 4.5 nm and 15.0 nm (FWHM). The device is equipped with a handheld reflectance probe, with an integrated 5W tungsten halogen source (model TRP5). The fibre optic bundle is made of one common end with seven optical fibres (whose diameter is 400 µm) in a “6 around 1” configuration, terminated with a probe tip that is inserted in the dedicated spot on the handheld probe. The latter was kept perpendicularly upon the surface of the panels, at a working distance of ca. 5 mm.

The following parameters were used for the acquisition of the data: 67 ms, 25 accumulations (first channel of acquisition); 335 μs, 59 accumulations (second channel of acquisition); 330 μs, 100 accumulations (third channel of acquisition). The spectrometer was calibrated against a B&W TEK diffuse reflectance standard. For each selected point – corresponding to the points where h-EDXRF analyses were performed—three spectra were acquired, and the average spectra were calculated. All the spectra were acquired using the iSpec® 4 software and subsequently processed with MATLAB® R2022a.

3 Results and discussions

3.1 Painting technique

As already reported in Sect. 2.1, the wallpapers represent the production of porcelain; different scenes are depicted, including human figures, nature motifs, landscapes, and architectural elements.

Most of the drawings consist of dark ink outlines and colours applied within them, either in homogeneous layers or as thin washes (Fig. 1b–d); this reflects the technique used in the production of Chinese wallpapers, as reported in literature [1, 15]. In some cases – such as the clothes of the human figures (Fig. 1b–c) – thin washes of different colours have been applied over a homogeneous layer of white pigment. The so-called boneless, that is, drawings lacking the ink outline [32], are also present (Fig. 2b).

Fig. 2
figure 2

Magnified images of some of the drawings highlighting some details of the painting technique. a close-up of a human figure and details of one of the architectural structures. b examples of the boneless drawings (the umbrella and part of a mountain in the back). c, d visible and infrared photos of a tree, showing the presence of a stylized drawing below the detailed leaves and branches

Full size image

Small brushes have been used to add small details – such as the lips and the eyes of the human figures (Fig. 2a) – or, instead, to realise the main features of figures and motifs.

It is the case, for example, of bushes and trees; the comparison between IR and visible photography (Fig. 2c, d) suggests that a stylized shape of the foliage was first realised and painted, and only subsequently the details of the branches were added to make them more realistic. Finally, the presence of straight lines – such as those realised for the depiction of architectural elements (Fig. 2a) – suggests the use of specific tools.

3.2 Conservation condition survey

The panels have been removed from their original location and glued on top of a layer of thick brownish paper. This process seems to have been carried out without a defined protocol or the necessary precautions. In some cases, for example, pieces and fragments of paper that were falling from the panels have been re-attached: some of them have been glued on top of others, some others have been re-attached in the wrong place or with an incorrect orientation. This lead to create a sort of «patchwork» texture, where the surface is uneven and multi-layered because of the incorrect arrangement of the re-attached fragments (Fig. 3c).

Fig. 3
figure 3

Magnified images of some areas highlighting some issues related to the state of conservation. a, b visible and UV fluorescence images of an area showing traces of glue on top of the surface. c the “patchwork” texture. d holes probably indicating the possible presence of pests or small rodents. e, f signs of the use of cutting and pointed tools

Full size image

Furthermore, a glue/adhesive seems to have been applied not only between the original panels and the underlying paper, but also on top of the panels (Fig. 3b), in areas corresponding to the cuts and the rips which were probably caused during the tearing process. The light bright fluorescence identified through the observation of the UV images could suggest the use of an animal/collagen-based adhesive, as opposed for example to starch paste [33, 34]. Clear signs of the use of pointed and/or cutting tools are also present (Fig. 3e, f).

The brownish, extended areas (Fig. 1a) on all the panels are the evidence of the poor condition of the paper support, which looks very fragile and brittle.

In addition to the brownish areas and the general brittleness, the panels show smalls holes and losses characterised by a more rounded shape and different edges: the paper in this case does not look cracked/fractured but slightly tore or “eaten” (Fig. 3d).

This condition seems to be the consequence of the direct consequence of the action of some pests or small rodents. The future analysis regarding the state of conservation of the paper sheets will also be aimed at understanding whether the possible presence of animals can be linked to the irregular brownish staining affecting the panels.

3.3 Paper support

In order to evaluate the influence of the paper support on the final results, both EDXRF and Vis–NIR-FORS analyses were performed on each panel.

The presence of Rh, Ni and Cu in the EDXRF spectra is due to the nature of the equipment, while Ar is related to the presence of air. The presence of S in the EDXRF spectra (Fig. 4a) could suggest that alum was used for the sizing process of the paper support [24]. During the preparation of the substrate, in fact, the sheets were usually coated with a white or coloured ground, and then sized with animal glue and alum [5, 6, 31]. The presence of Fe Ca, K and Ti in XRF analysis on paper used as a support for painting has already been reported in a study on Chinese export watercolours [24]; in particular, Ti could be related to a difference in the use of filling or sizing material as compared to the traditional Chinese method. The wallpaper panels under study date to the eighteenth century, but the artificial form of titanium dioxide (TiO2) has only been synthesized in Europe at the end of the nineteenth century and used as a white pigment from the beginning of the twentieth century [35]. At the same time, the discussion on the use of TiO2 in the mineral form of anatase in ancient Chinese art is controversial. The use of native anatase as white pigment to decorate ancient (ca. 4300–2800 BC) Chinese pottery is reported in the study of Zuo et al. [36], but a later study affirms that the amount of anatase component detected in those manufacts is not sufficient to infer that the mineral was sourced to be used as a white pigment [37]. It must be noted also that no information is present about the process the wallpapers underwent to be removed from the original place and subsequently arranged to be donated to the Museum. As described in Sects. 2.1 and 3.2, at present they are glued on big sheets of brown paper, and extra glue seems to be present on the surface of the panels. Therefore, the presence of Ti could also be related to the use of modern or contemporary materials in the phase of the donation of the wallpapers to the Museum. Further analyses – such a Raman spectroscopy and μXRD – are needed, in order to clarify and investigate more in-depth the reason for the presence of Ti in the EDXRF spectra of the paper support.

Fig. 4
figure 4

Representative EDXRF (a) and Vis–NIR-FORS (b) spectra of the paper support

Full size image

The Vis–NIR-FORS spectra of the support (Fig. 4b) show the typical features due to the nature of the paper [28, 38]. More specifically, the range 1440–1600 nm is characterised by first overtones of the O–H stretching vibrations in the cellulose and hemicellulose structure (and lignin, to a smaller extent). The absorption at around 1790 nm is due to the first overtone of the C–H stretching vibration of cellulose, while the range 1916–1942 nm (and around 1980 nm) is where combinations of O–H asymmetric stretching and O–H deformation vibration of H2O occur [38]. The absorption band around 2092 nm is related to combinations of O–H and C–H deformation bands and O–H stretching bands in the cellulose structure. Finally, absorptions around 2270–2272 nm are due to combination bands of stretching and deformation vibrations of O–H, C–O, C–H and C–H2 groups within the structure of cellulose and hemicellulose [38].

The presence of these spectral features can lead to an additional challenge during the interpretation of the spectra, especially for those pigments whose spectral features fall in the same range of those of the paper support (see for example paragraphs 3.7 and 3.9).

3.4 Red and pink hues

The analyses on the red areas identified two main groups, the first one corresponding to the architectural elements, such as windows and roofs, and the second one corresponding to some of the clothes of the human figures.

EDXRF spectra referring to the first group show the main presence of Hg, in addition to the elements related to the paper support (Fig. 5a), from which it is possible to infer the use of cinnabar/vermilion (HgS). More precisely, the name cinnabar refers to the natural crystalline form of mercury sulphide, while vermilion is the standard term used to identify the synthetized mercuric sulphide [39]. The presence of Pb can be ascribed to the white details realised in the surrounding areas. FORS spectra (Fig. 5b) show a sigmoid curve, with an inflection point at 608 nm. The reported value represents the average of the values of the inflection points of all the curves, calculated by performing the first derivative. The results confirm the use of cinnabar/vermilion: the position of the inflection point for HgS reported in literature is in fact at 580–605 nm [19, 21, 40]. The red colour of HgS is caused by electronic transitions from the lower energy valence band to the higher energy conduction band; the valence-to-conduction gap is responsible for the sharp slope centred around the inflection point. The fact that the position of the inflection point can vary within a range can be explained by different factors, such as the presence of impurities in the semiconductor, the nature of the support or substrate, the different particle size, or the concentration of the pigment in the paint layers [21, 26]. In addition to this, the presence of mixtures of pigments often affects the clear identification of the spectral features of the different constituents [21, 26]. Therefore, bathochromic or hypsochromic shifts – that is, respectively, shifts towards longer or shorter wavelengths – in the position of the inflection points can occur [21].

Fig. 5
figure 5

Representative EDXRF (a) and Vis–NIR-FORS (b) spectra of the red area reported in the circle in a

Full size image

With regard to the red areas of the clothes, the high amount of Pb in the EDXRF spectra (Fig. 6a, b) suggests that the white layer below the red washes is made of lead white (2PbCO3·Pb(OH)2). The presence of Hg in only some of the spectra allows to discriminate between two groups: the first one characterised by a mixture of red dye (not detected by EDXRF for its organic origin) and HgS, and a second one where the HgS is not present.

Fig. 6
figure 6

Representative EDXRF (a, b) and Vis–NIR-FORS (c) spectra of the red areas characterized by the presence of dye. The curves named a and b in (c) correspond, respectively, to the areas reported in the circles in a and b

Full size image

Vis–NIR-FORS spectra are presented as apparent absorbance – Log (1/R) – to enhance the absorption features of red dyes [21, 41, 42].

The results show two absorption sub-bands, located, respectively, at 525 nm and 565 nm in the case of the first group, and at 529 and 570 for the second one (Fig. 6c). The presence of these sub-bands suggests that anthraquinone-based red dyes have been used: the mechanism of absorption of light is related to transitions among delocalised molecular orbitals – more specifically, n → π* transitions of the carbonyl groups in the anthraquinone structure [21, 41, 43]. Dyes of animal origin, such as cochineal and lac dye, show bands at 520–525 and 550–565 nm [21, 41, 43, 44]. Lac dye and cochineal are, in fact, the most likely used red dyes in Chinese paintings. Historic Chinese painting treatises do not mention laking processes during the preparation of the dye; in other words, the dye was not precipitated onto an insoluble substrate to create a compound which is more stable as opposed to the raw dye [28], but directly mixed with animal glue before its application [45]. Recent studies [28, 46] report that FORS analyses on red dyes on Asian paintings – despite not having undergone laking processes – show the same spectral features as those cited in literature for red lakes [21, 41, 42, 44]. Moreover, Villafana and collaborators [28] compared the results obtained from the application of the same dye extracts on both sized and unsized paper. The spectra for dyes on unsized paper did not show identifiable features, as opposed to those on sized paper. The results obtained in this study – in agreement with those mentioned on Asian paintings [28, 46] – are consistent with those cited in literature for red lakes, as reported above. Thus, it is reasonable to think that – despite the absence of a proper laking process as it was known in the West – the sizing or the paint preparation on Chinese paintings have some influence on the stability of the unlaked dyes.

In general, FORS does not allow to distinguish between cochineal and lac dye [21, 28]. In the studies performed on Chinese paintings belonging to the Freer Gallery of Art and Arthur M. Sackler Gallery collections, the position of the sub-bands for lac dye is consistently shifted 7–8 nm (towards longer wavelength) from those in cochineal; these results were subsequently confirmed by HPLC analyses [43]. As reported in Table 1, in this work, the difference in the position of the sub-bands identifies two groups, which also differ for the presence/absence of HgS. According to Bisulca and Winter [45], vermilion/cinnabar was added to shift the colour to red. The two hues are indeed slightly different (Fig. 6a, b, the areas in the circles), with the one characterised by the only presence of red dye resulting in a colder purplish tone. Commonly, HgS was added to lac dye; no instance has been found of HgS mixed with cochineal in the Freer and Sackler collections. In this work, though, where cinnabar/vermilion has been identified, the two absorption sub-bands are closer to those of cochineal. Further analyses – such as Raman and HPLC/MS – are needed, to investigate more in-depth the nature and the use of the dyes on the panels under study.

Table 1 Average values (in bold) and range of the positions of the two sub-bands in the two different groups (a and b)
Full size table

Finally, regarding the pink hues, EDXRF spectra show the presence of Pb and Hg suggesting a mixture of lead white and cinnabar/vermilion; Vis–NIR-FORS spectra of the pink hues are characterized by an inflection point at 598 nm, thus confirming the presence of HgS. Since the present study is part of a broader ongoing project, some later analyses by means of portable remote Raman have been performed in collaboration with the Raman Spectroscopy Research Group from Ghent University; more specifically, the results show the typical Raman features of cinnabar/vermilion, lead white and gypsum (CaSO4 ·2H2O) [47]. As a final remark on the use of HgS, the techniques used in this study did not allow to distinguish between the two forms cinnabar and vermilion, therefore the term “cinnabar/vermilion” has been used when mentioning this pigment. Furthermore, the Chinese language often does not differentiate between the two words – as opposed to English – therefore information about which form was preferred to the other is not easily found in the historical Chinese sources [30]. Nevertheless, the presence of HgS in this study is significant, as it has been used both for the red hues – whether alone or in mixtures with the anthraquinone-based dye of animal origin – and the pink hues, in combination with lead white and gypsum.

3.5 Blue hues

The EDXRF spectra for the blue hues are very similar to those related to the paper support, except for the presence of Pb related to the white layer below in the case of the clothes of the human figures – as it is for the previous cases (Fig. 7a, area 1). The absence of Cu allows to exclude the use of azurite, which was one of the blue pigments used for Chinese paintings and wallpapers [2, 32, 48]. Furthermore, most of the spectra show a slightly higher content of Fe as compared to the spectra of the paper support, which could suggest the presence of Prussian Blue (Fe4[Fe(CN)6]3).

Fig. 7
figure 7

Representative EDXRF (a) and Vis–NIR-FORS (b) spectra of the blue areas. The dark blue spectrum in (a) corresponds to the area in circle n.1, while the light blue spectrum corresponds to the area in circle n.2

Full size image

The FORS spectrum (Fig. 7b) shows a strong absorption band around 660 nm and a sharp increase towards the infrared region with an inflection point around 705 nm. Furthermore, a relative reflectance maximum around 485 nm is present. These values agree with the spectral features of the blue dye indigo [21, 28, 49, 50]; the absorption of light in this case is determined by delocalised molecular orbitals [21] in the structure of indigotin, which is the main component of the dye.

Mixtures of indigo and Prussian blue were widely used in Asian paintings [25, 26] and specifically in Chinese paintings on paper and silk [32]. Despite being a pigment synthesized in Europe at the beginning of the eighteenth century, in fact, Prussian blue became available for Chinese (and Asian in general) artists since that time [32], as a consequence of the trades between East and West. When mixtures of pigments are involved, the identification through FORS becomes more challenging; previous studies report the realization of colour charts with mixtures of indigo and Prussian blue in different proportions, to have reference spectra to be compared with real artworks [25, 51]. In this specific case study, the results obtained from the FORS analyses on the wallpaper panels seems to suggest the main presence of indigo; the presence of Fe in the EDXRF spectra could indicate instead that Prussian blue has been used, whether in mixture or alone. This could perhaps also explain the fact that the blue areas appear dark in the IR photography (the details of the trousers in Fig. 1d, and the windows of the building in Fig. 2d); Prussian blue in fact absorbs the infrared radiation, appearing dark, while indigo transmits the infrared radiation, appearing transparent [28, 52]. Nevertheless, these hypotheses need to be further explored by means of more thorough and complementary analyses – e.g. FTIR and Raman spectroscopies – in order to determine the nature and composition of the blue hues.

3.6 Purple hues

The EDXRF spectra do not show diagnostic elements other than Pb which is related to the white layer below, suggesting an organic origin of the purple colourant (Fig. 8a).

Fig. 8
figure 8

Representative EDXRF (a) and Vis–NIR-FORS (b) spectra of the purple area reported in the circle in a

Full size image

From the FORS spectra, a mixture of a red dye of animal origin (probably lac dye or cochineal) and indigo can be inferred: all the curves present in fact two absorption sub-bands at 530 nm and 572 nm and a stronger absorption band at 665 nm (Fig. 8b). The reported values represent the average values of the positions of the bands in all the spectra collected (Table 2).

Table 2 Average values (in bold) and range of the positions of the absorption bands of the purple areas
Full size table

Mixtures of indigo and a red dye have been widely reported in previous studies on western and Chinese paintings [21, 45]. In particular, the purple colour in Chinese paintings was obtained mixing indigo with lac dye, rather than cochineal; the latter was more used in mixtures with Prussian blue, and mainly during the nineteenth and twentieth centuries [45]. The positions of the two sub-bands towards longer wavelengths could suggest the use of lac dye rather than cochineal.

3.7 Green hues

The very high amount of Cu shown in the EDXRF spectra (Fig. 9a) suggest the use of a copper-based green. The most widely copper-based green pigment used in Asian paintings was malachite (CuCO3·Cu(OH)2) [2], but the use of emerald green and copper chlorides and trihydroxychlorides (such as atacamite and paratacamite) has also been proved [2, 30]. According to the EDXRF spectra, the absence of As leads to exclude the presence of emerald green, while the absence of Cl leads to exclude the use of copper chlorides. Another copper-based green pigment is Verdigris, but its use in Chinese paintings has not been reported; the use of the term “verdigris” has been found in a technical manual, but scientific investigations then suggested it was atacamite [53]. In the light of this, malachite is likely to have been used for the green hues of the wallpaper panels under study, even though another possible option could be a mixture of azurite (2CuCO3·Cu(OH)2) and a yellow organic dye [30].

Fig. 9
figure 9

Representative EDXRF (a) and Vis–NIR-FORS (b) spectra of the green areas. The dark green spectra in a and b correspond to the area in circle n.1 (group trees and bushes), while the light green spectra in a and b correspond to the area in circle n.2 (group windows and vases). The grey curve in b represents the paper support, and it is shown for comparison

Full size image

As for the FORS spectra, the attention was focused on checking whether the spectral features typical of malachite or azurite could be identified, thus confirming the hypothesis suggested by the EDXRF results. Azurite shows characteristic absorption bands at about 1495, 2285 and 2350 nm, specifically related to the presence of hydroxyl and carbonate groups in the structure of the pigment. These features have not been detected in the FORS spectra of the green hues, which allowed to rule out the possibility of a mixture of azurite and a yellow dye. Malachite is a basic copper carbonate and usually shows specific absorption bands due to hydroxyl groups and carbonate ions in the 2200–2350 nm region [16]. More specifically, the absorptions at around 2215/2220 nm and 2270/2275 nm are due to the second overtone of the CO32− stretch, while those around 2357/2360 nm and 2400 nm are due to combination of O–H bending and stretching [16, 19, 54, 55]. From the FORS spectra (Fig. 9b), two groups can be distinguished: a first one, corresponding to windows and vases (Fig. 9a2) and a second one, corresponding to trees and bushes (Fig. 9a1). In the spectra referring to the first group (the light green curve in Fig. 9b), the features of malachite appear as weak absorptions, and in the specific case of the absorption at 2270 nm, it overlaps with those related to the paper support (namely, combination bands of stretching and deformation vibrations of O–H, C–O, C–H and C–H2 groups within the structure of cellulose and hemicellulose [38]). Nevertheless, they lead to think that malachite is present, while the same cannot be stated for the group corresponding to bushes and trees, where the features do not seem to appear (the dark green curve in Fig. 9b). One of the hypotheses to explain this fact could be that a layer of malachite was applied first, and washes of a dye – or a mixture of dyes – were subsequently added, making the identification of the features of malachite more challenging. Instances of this technique have already been described. In the work of Kogou and collaborators [24], the authors report the use of a mixture of gamboge and indigo on top of a malachite layer in green areas corresponding to leaves. Indigo-based green leaves have also been identified in one of the paintings belonging to the Freer and Sackler collection [32]. Furthermore, the FORS spectra referring to the second group show an absorption band in the visible range, specifically at around 660 nm, which leads to confirm the presence of indigo. Finally, the comparison between visible and IR photography of an area depicting a tree seems to corroborate the hypothesis of overlapping layers: as described in paragraph 3.1 and shown in Fig. 2c, d, the IR photo shows the presence of a stylized foliage below the detailed branches shown in the visible photo.

3.8 Yellow and brown hues

EDXRF and FORS spectra for the yellow areas do not show significant element or features that allow for the identification of a specific pigment or dye. EDXRF shows a prevalent presence of Pb, related to the white layer below. The absence of Fe and As leads to exclude, respectively, the presence of yellow ochres (FeO(OH)) and orpiment (As2S3), suggesting that an organic dye has likely been used. Yellow dyes were frequently used in Asian paintings; gamboge was one of the most common, but it has proved to be difficult to identify [24, 30]. In the study conducted on the Freer and Sackler collection, wherever gamboge is present it has been identified only after FTIR analyses were conducted [30]. The discrimination among the different yellow dyes through FORS is, in fact, very challenging, as their main absorption feature is a band at around 400–450 nm [49]. Furthermore, in this specific case, the yellow dye does not show fluorescence (Fig. 3b, the trousers of the human figure at the top-centre) which could infer the presence of gamboge [56].

Regarding the brown areas, EDXRF spectra show the main presence of Fe, suggesting the presence of Fe-based pigments such as ochres and earth pigments. The typical features of these pigments are not evidently shown in the FORS spectra; therefore, in this case, the technique cannot confirm the results obtained through the EDXRF analysis nor provide additional information.

3.9 White and black hues

The EDXRF spectra of all the white areas – vests, porcelain objects, some of the carnations – show a high and prevalent presence of Pb, from which it can be inferred the use of lead white (2PbCO3·Pb(OH)2). On the contrary, the FORS spectra do not allow to distinguish clearly between the spectral features of the paper support (as, for example, the absorptions in the range 1440–1600 nm and the absorption around 1980 – see 3.3) and those of lead white. This pigment usually shows, in fact, an absorption band at around 1445 nm – due to the hydroxyl groups’ first overtone of the stretching mode vibrations – and three bands related to the carbonate ion at around 2000 nm, 2322 nm and 2136 nm, due to combinations and overtones of the symmetric and asymmetric stretching mode vibrations [57].

Regarding the black areas, the EDXRF spectra do not suggest the use of specific materials, as they show the same elements as those of the paper support (Fig. 4a). Considering that black corresponds to total absorption, FORS spectra of black pigments or inks do not show specific features that allow for the identification of their origin [21]. In the case of Chinese black inks, the main components are soot (either wood soot or lampblack), animal glue and additives [58,59,60]; therefore, further analyses – such as Py-GC/MS – are needed to better understand the nature of the black painted areas in the wallpaper panels under study.

4 Conclusions

The combined use of technical photography and spectroscopic techniques proved to be successful for the non-invasive in situ investigation of the set of Chinese wallpaper panels belonging to MNAA, allowing for a preliminary identification of the palette, of the painting technique and for the detection of issues related with the state of conservation. More specifically, EDXRF and Vis–NIR FORS showed their effectiveness when used together, as they are complementary in the identification of inorganic pigments and organic dyes. The results obtained through technical photography integrated the data obtained by EDXRF and FORS, strengthening the results from another point of view.

Limitations regarding the identification of some of the pigments and dyes have occurred. It is the case, for example, of the yellow dye and the black ink, for which the techniques involved in the study were not able to define their origin; more in-depth analyses are needed to guarantee a proper identification. Additional investigations will be performed also to confirm and clarify some of the results obtained so far, such as those regarding the composition of the green and blue hues, the use of the anthraquinone-based red dyes, and the possible presence of western modern materials. Moreover, specific analyses will be carried out to investigate the degradation processes and the alteration that affect the paper support, to give a better understanding of the questions arisen about its state of conservation. In order to address all these questions, complementary micro-destructive analyses on collected samples will be carried out, by means of Raman, FTIR, μXRD, SEM–EDS, HPLC/MS and Py-GC/MS.

Despite some limitations and the need for more in-depth analyses, this study showed the validity of an in situ non-invasive approach in the definition of a methodology not only for the investigation of the wallpaper panels belonging to MNAA, but for the study of Chinese wallpapers in general, providing useful information to address conservation and historical research questions.