Introduction

Joan Miró (Barcelona 1893–Palma de Mallorca 1983) is one of the great artists of the twentieth century. The originality of Miró’s works in art forms ranging from painting, paper and sculpture to monumental ceramic murals, painted textiles and even theatre set designs, strongly influenced the artistic practice of the twentieth century. Miró’s most prolific and experimental production was in Mallorca, from 1956 until his death in 1983. For the first time, he had extensive studio spaces (known as Taller Sert and Son Boter) that allowed him to work on large-format artworks, and on many works simultaneously. The Mallorca studios have been preserved as the artist left them at his death, including over a hundred tubes of oil paints, cans of synthetic paints, palettes, and other materials. These offer rich resources for the study of Miró’s painting materials and techniques. The investigation of the materials left in the artist’s studios, complemented with the study of micro-samples taken from his artworks, provide useful information for understanding Miró’s materials. It also helps to comprehend changes they have undergone after ageing, which may also contribute to the interpretation and conservation of the artworks. Other studies of artists’ studio materials have shown the importance of these kinds of collections [1,2,3,4,5].

The Fundació Miró Mallorca, built around Miró’s studios, also holds an extensive archive and a large collection of artworks that were left there after Miró’s death [6]. By examining the artist's works, conservators at the Fundació noted that some passages of yellow paint have discoloured and, in some cases, have become powdery, resulting in significant visual changes. As an example, Femme dans la rue (1973) appears in a 1978 photograph with Miró in his Taller Sert studio (Fig. 1a). When compared to its current condition, the original vibrant yellow appears to have faded (Fig. 1b). A total of 26 artworks in the collection present these phenomena, all painted in the 1970s. This deterioration is also observed in the paints on some of the artist’s palettes preserved in the studios.

Fig. 1
figure 1

Photograph of Joan Miró in his studio Taller Sert in 1978 (© Jean Marie del Moral) (a), showing the original condition of Femme dans la rue (1973) (b), now showing discolouration of the original vibrant yellow

Full size image

During this study, the degraded paints were identified as cadmium yellow (CdY), a colour favoured by Miró throughout his career. In a letter written to his friend Enric Ricart during his formative years, Miró wrote: “Instead of chrome, I recommend … cadmium lemon, it is splendid… Chrome, next to cadmium lemon, is a dull and lightless colour.” [7]. CdY has been characterised in two of his works from the 1930s [8, 9] and all the orange and yellow paints found in his studios in Mallorca are cadmium-based [10].

Cadmium yellow is a synthetic pigment based on cadmium sulfide (CdS), which began to be commercialised for artistic purposes from 1840 [11]. Lighter CdY paints, usually containing zinc, correspond to a later date of pigment production (from 1920) [11]. Cadmium yellow was quickly adopted by many artists including Van Gogh, Picasso, Seurat, Matisse and Munch [11]. It appears that cadmium sulfide produced in the late nineteenth and early twentieth centuries was poorly lightfast, resulting in discolouration in many works of art after only 20 years [11]. The deterioration of CdY-based paints has been reported in paintings including The Scream by Edvard Munch [12, 13], as well as in other artworks by Van Gogh [14, 15], Matisse [16,17,18,19], James Ensor [20] and Picasso [21, 22]. Through these studies and further analyses on model paints, the degradation pathway of CdY—from the yellow cadmium sulfide (CdS) to the white cadmium sulfates (CdSO4/CdSO4·H2O)—was established, and the possible factors that have a key role in CdS paints degradation were identified [23]. Studies on model paints have demonstrated that the degradation of CdY is triggered by light and/or high relative humidity (≥ 95%) and that zinc cadmium sulfide (Cd1−xZnxS) solid solutions are more susceptible to photo-oxidation than pure CdS. Moreover, residues of the synthesis (e.g., chlorine/chloride species [13, 18]) and early imperfect synthesis processes [15, 19, 24] may play a role in CdS paint stability. Most of the reported studies are related to paintings dating from the late nineteenth and early twentieth centuries, while the artworks by Miró were painted in the 1970s. Therefore, the collection of Fundació Miró Mallorca represents a unique case study for examining the deterioration of this paint at a later stage of pigment production history [11].

In this work, nine samples of yellow paint from both artworks and studio materials were analysed. The aims of this study are to (i) identify the composition of the yellow paints and understand the characteristics of the CdS pigment present therein, (ii) compare the composition of artist studio materials with the artworks, and (iii) propose factors that could have triggered the degradation of the CdY paints.

Materials and methods

Samples selection and preparation

Under visible light, most of the 26 artworks in the collection that present CdY deterioration show loss of the intensity of the yellow colour, as well as chalking: a form of deterioration where the paint becomes crumbly and powdery. In some artworks, a whitish crust was observed on the superficial paint layer—hiding a more intense yellow layer underneath—and some small shiny crystal-like efflorescence was visible with the naked eye. In other paintings, where the paint layer was thinly applied, the yellow had become nearly transparent (see Additional file 1: Fig. S1).

Paint microsamples were taken from three representative artworks from the collection of the Fundació Miró Mallorca, which show deterioration of the yellow oil paint (see Fig. 2). Microsamples of yellow paint were also collected from three paint tubes from different brands (Lucien Lefebvre-Foinet, Mir and an unknown one), a cup used for mixing paint, and two palettes (Additional file 1: Fig. S2). These paint tubes and objects were found in Taller Sert and Son Boter studios. Table 1 presents the full description of the samples and their provenance.

Fig. 2
figure 2

Sampled artworks by Joan Miró and macro images of the sampled yellows (marked with white dotted rectangles and arrows in the artworks). a Untitled (FPJM-00029), 1978, oil on canvas, 92 × 72,5 cm (sample A2). b Untitled (FPJM-00054.1), 1973, Oil, acrylic and charcoal on canvas, 270 × 355 cm (sample A7). c Untitled (FPJM-00053), 1974, Oil, acrylic and chalk on canvas, 270.5 × 355 cm (sample A9)

Full size image
Table 1 Samples name, image of the object or of the point on the artwork from which the sample was taken, description of the sampled artwork/object and of the region where the samples were taken
Full size table

All artworks had been displayed in the artist's studio Taller Sert from the time they were painted in the 1970s until they were moved to the new museum building, completed in 1992, and later replaced with replicas. It is worth noting that the studio Taller Sert has an uncontrolled environment with intense natural light through window glass. Recent datalogger readings show temperatures reaching over 30 °C in the summer and levels of relative humidity (RH) over 80% in the winter. The artworks in the studio were exposed to these conditions for around two decades.

The paint tubes and the palettes were kept in the studios Taller Sert and Son Boter until the time of the analyses except for one of the palettes (from which sample A4 was taken). This palette was donated by the Miró family to the Fundació in 1986 and was displayed inside an enclosed case (made of glass and with a wooden board covered with fabric inside) until 2018, when it was removed from the display case and placed in the Taller Sert.

When the amount of the collected material was sufficient, part of it was analysed without embedding it in resin. In the other cases, where the available material was limited, samples were prepared as cross-sections embedded in epoxy resin (ResinPro®) followed by dry hand polishing with abrasive sheets (Micro-Mesh®).

Multi-analytical approach

Samples were analysed using a multi-analytical approach that included investigation using elemental, structural, molecular and luminescence techniques, as described below.

Scanning electron microscopy with energy dispersive X-ray analysis (SEM–EDX)

Elemental analyses were performed on raw samples A1–A6 with a scanning electron microscope (SEM, Hitachi S-3400N, used at 15 kV), equipped with a Bruker AXS XFlash 4010 EDS system. Elemental mapping of cross-sections of samples A7, A9, A12 was performed using a Zeiss EVO LS15 SEM–EDX used at 20 kV. The samples were carbon-coated prior to EDX mapping to improve surface conductivity.

X-ray diffraction (XRD)

Identification of crystalline phases was performed using X-ray diffraction analysis on unmounted samples (A1–A6) using a Bruker D8 Advance diffractometer with monochromatised CuKα radiation (1.54 Å) at 40 kV and 40 mA.

Synchrotron Radiation (SR) µXRD mapping on micro cross-sections (samples A4, A7, A9 and A12) and paint tube fragment (sample A5) was performed at the ID13 beamline of the European Synchrotron Radiation Facility (ESRF, Grenoble), benefitting from the Historical Materials BAG [25]. The μXRD branch was used to perform crystalline phase mapping using a 2.5 × 2.5 μm2 beam with an energy of 13 keV. Two-dimensional (2D) diffraction patterns were collected in transmission mode at every pixel of 2D maps and converted into 1D diffractograms by azimuthal integration, using the Jupyter Notebooks based on the PyFAI software package [26].

In both cases, the identification of the crystalline phases was performed using Match! and QualX software, while maps analysis was carried out with PyMca software [27].

Micro X-ray fluorescence (µXRF) and X-ray absorption near edge structure (µXANES)

High lateral resolution information on the distribution and speciation of S of samples A7, A9, A5, A4 and A12 was obtained by performing hyperspectral 2D μXRF mapping and μXANES spectroscopy analysis in XRF mode at S K-edge at the scanning X-ray microscope end-station hosted at the beamline ID21 of the ESRF (Grenoble, France) [28, 29]. The energy calibration was performed using reference powders of CaSO4⋅2H2O. The incident beam was focused with Kirkpatrick-Baez mirrors down to a beam of 0.5 × 0.6 μm2 (h × v). Hyperspectral 2D μXRF maps were recorded by scanning the energy across the S K-edge employing the following conditions: from 2.46 to 2.4675 keV, 2.5 eV step; from 2.468 to 2.485 keV, 0.5 eV step; from 2.49 to 2.53 keV, 2 eV step. An additional map employing a monochromatic primary beam of fixed energy at 3.573 keV was acquired, to excite Cd at its L3-edge. The software PyMca was used to fit the XRF spectra and to separate the contribution of different elements. S−II and SVI chemical state maps are obtained by extracting the maps obtained at the energy of 2.473 keV and 2.482 keV to favour the excitation of the S-II and SVI species respectively and following the procedure described in literature [20]. XANES spectra from selected regions of interests (ROIs) of hyperspectral 2D μXRF maps were also extracted and compared.

Fourier-transformed infrared spectroscopy (FTIR)

FTIR analyses were performed on unmounted samples A1-A6 in transmission mode with a Bruker Tensor 27 system in the spectral range 4000 cm−1 to 400 cm−1, with 3 cm−1 of resolution and accumulating 32 scans.

Micro Attenuated Total Reflectance (μ-ATR) FTIR was carried out on points of the cross-sections using a Thermo Nicolet iN10 MX spectrometer equipped with an MCT/A detector cooled with liquid Nitrogen and a Germanium crystal, with a real aperture of 25 × 25 μm2. Data were collected in the range 4000–750 cm−1, at a spectral resolution of 4 cm−1 and with 256 average scans. Spectra are displayed without correction for ATR.

Micro photoluminescence (µPL) measurements

The PL properties of the microsamples were probed with a time-gated photoluminescence microscopy system, fully described elsewhere [30]. The system is based on a pulsed Q-switching laser source (λem = 355 nm, pulse duration = 1 ns, repetition rate = 100 Hz, FTSS 355–50, Crylas GmbH) and time-gated detector (C9546-03, Hamamatsu Photonics and R6, Qimaging). The image detector is coupled to an epi-fluorescence microscope equipped with a 50× (NA = 0.5) refractive objective, providing imaging from a circular field of view 300 μm in diameter with a spatial resolution of 0.6 μm. By using a time-gated approach and coupling the detection with a hyperspectral system, it is possible to reconstruct time-gated hyperspectral images of the PL emission at different timescales.

Raman spectroscopy

Raman spectra were acquired using an inVia Qontor confocal Raman microscope (Renishaw plc, Wotton-under-Edge, UK). The spectra were recorded using the 785 nm laser excitation, a 1200 lines/mm dispersive grating, and 50× long-distance magnification with 1 cm−1 spectral resolution and 1 µm spatial resolution. The spectra were acquired using the WiRE 5.5 software. The acquisitions varied between 5 and 20 s, 2 to 118 mW power, 5 and 20 accumulations.

Py/GC–MS and GC–MS

According to a procedure described in [2, 31,32,33], Py/GC–MS analysis was conducted on micro fragments of the samples (approximately 0.30 µg) for a preliminary screening regarding organic composition. After ascertaining the presence of lipid-bound paint, it was decided to perform a quantitative analysis of the paint samples via GC–MS to establish the exact nature of the identified lipid fraction and to obtain the molar ratios of the most significant fatty acids found within the samples. The procedure used is detailed in the literature [34,35,36,37] and has already been successfully applied on modern and contemporary oil paintings. The molar ratios among the most important fatty acids considered are: Azelaic to Palmitic (A/P), Oleic to Stearic (O/S) and Palmitic to Stearic (P/S).

Results

Results of samples characterisation are summarised in Table 2.

Table 2 Results of the multi-analytical characterisation performed on the samples
Full size table

Elemental composition and crystalline phases

Elemental analyses revealed that the samples from artworks, palettes and Lucien Lefebvre-Foinet paint tube (sample A5) are similar in elemental composition, containing as main elements: Cd, S and a low quantity of Zn (Additional file 1: Figs. S3 and S4). In samples A3 and A6, only Cd and S were detected, together with small peaks for Mg, Al and Si. These elements may be indicative of the presence of additives such as kaolin and magnesium carbonate (as confirmed by FTIR analysis). Al and Si traces were also found in sample A9, while sample A12 presents small traces of Cl.

X-ray diffraction analyses showed that samples from all of the artworks, palettes and paint tube A5 are composed of poorly crystalline CdS or Cd1−xZnxS in mixed crystalline form (hexagonal and cubic). Due to the broad diffraction peaks (Fig. 3), it was not possible to establish the amount of Zn present by estimating the shift of the diffraction peaks with respect to the refence CdS pattern [38]. Instead, paint tubes A3 and A6 show the presence of crystalline hexagonal CdS (Fig. 3b). Moreover, most of the samples from artworks or palettes analysed through SR µXRD (A7, A9, A12), show the presence of cadmium sulfate hydrate (3CdSO4·8H2O) at the paint surface or within the paint layer (Additional file 1: Fig. S5), related to paint degradation.

Fig. 3
figure 3

a SR μXRD patterns of samples A4, A5, A7, A9 and A12 recorded at beamline ID13 of ESRF employing a beam with an energy of 13 keV. The broad peak between 8 and 14°, indicated with an asterisk, is related to the embedding resin, while other sharp peaks in samples A7, A9 and A12 can be ascribed to cadmium sulfate hydrate. Peaks indicated with squares in sample A5 are related to the support tape. b XRD patterns of samples A1, A2, A4, A3 and A6 recorded employing monochromatised CuKα radiation (1.54 Å)

Full size image

Molecular composition

FTIR spectra from the paint fragments and cross-sections (Additional file 1: Fig. S6) show the typical peaks associated with fatty acids at ca. 2920, 2850 [v(CH)], and 1710–1730 [ν(CO)] cm−1, suggesting a lipid-containing binding medium [39]. In most of the samples, IR bands at 1560–1530 cm−1 were also detected, ascribed to the presence of metal carboxylates (Cd and Zn ones [40, 41]), formed as a consequence of the reaction of free fatty acids with metal cations or intentionally added to the paint tube formulation as additives [42, 43]. Magnesium carbonate (ca 1450, 1415, 875, 820 cm−1) was also detected in Mir paint tubes (samples A3 and A6). This compound was used in the formulation of twentieth century oil paints tubes as an additive [42,43,44,45].

Cadmium sulfate in hydrated form was detected in all the samples from artworks and palettes. The sulfate was particularly evident in all the samples from artworks (broad band in 1000–1100 cm−1 range and below 650 cm−1) and from palettes A1 and A12. Interestingly, sulfates were also found in samples from the Lucien Lefebvre-Foinet paint tube (sample A5). Cd and/or Zn oxalates were also found in samples A1 and A2, through the absorption bands at ca 1620, 1320 and 820 cm−1. These compounds can form due to the reaction between Zn2+ and Cd2+ ions of the pigment and oxalate ions resulting from the oxidation of the oil binder [46, 47].

The presence of cadmium sulfate hydrate was also confirmed using Raman spectroscopy (Additional file 1: Fig. S7), which detected the bands at 476, 983 and 1006 cm−1, characteristic of the SO4 bending mode, and the oscillation of the free and bound (SO4)2− ion respectively [48, 49]. The Raman spectra from the samples investigated also indicated bands related to cadmium sulfide and its multi-phonon resonant scattering. More specifically, the bands at 238, 305, and 633 cm−1are ascribed to the transverse optical (TO) and longitudinal optical (LO and 2LO) modes of cadmium sulfide respectively, while the bands at 218 and 346 cm−1 are ascribed to the multi-phonon modes [50, 51]. Previous studies have suggested that the increase of zinc content in the Cd1−xZnxS solid solutions is likely to be associated with the disappearance of the multi-phonon modes and the shift of the LO mode band to lower wavenumbers [46]. The presence of bands related to the multi-phonon modes in the Raman spectra from the samples investigated might be due to the presence of only a small proportion of zinc in the Cd1−xZnxS solid solution, as suggested by elemental analysis. Bands related to a drying oil present as binding medium were detected at 869, 1063 and 1080 cm−1 and, in sample A5, which was examined in the range 180 to 3000 cm−1, additional bands for oil at 1304, 1441, 1657, 1740, 2853, and 2909 cm−1 could also be detected [52].

The distribution of sulfides/sulfates in the paint cross-section was achieved through SR μXRF chemical state maps of samples A4, A7, A9 and A12 (Fig. 4 and Additional file 1: Fig. S8). In the bulk of all the samples sulfides were detected, which can be associated to the preserved pigment (Cd1−xZnxS). In the samples A7, A9 and A12, a layer of sulfates was found on the surface, while in sample A4 sulfides have been mainly detected in the whole paint layer, with a localised agglomerate of sulfates that well corresponds to the elemental distribution of Zn. The localisation of sulfates at the paint surface confirms their presence as degradation products of CdS, due to its interaction with the environment [23]. The agglomerates can be instead related to the formation of metal soaps [13].

Fig. 4
figure 4

False colour RGB composite image generated following SR µXRF maps of Cd/Zn/S and S−II/SVI [step size (h × v), 1 × 1 µm2; exp. time, 10 ms/pixel] and average XANES spectra recorded at the S K-edge of selected ROIs belonging to sample A4 and A7. It is noted that the detection limit for Zn L-lines in the current experimental conditions is very low

Full size image

The organic composition of the paints analysed using GC–MS, showed that all the samples from paintings and studio materials are bound in drying oils. All chromatograms obtained after transesterification of the paint fragments and subsequent GC–MS analysis showed the presence of the typical fatty acids characteristic of siccative oils: saturated mono-fatty acids, from C8 to C26 (palmitic and stearic being the most abundant); saturated di-fatty acids (azelaic, suberic and sebacic acids in particular), as the most abundant oxidation products; unsaturated fatty acids (mainly oleic acid); glycerol (and glycerol derivatives); oxo-, hydroxy- and metoxy- octadecanoic acids as by-oxidation products of unsaturated fatty acids.

Based on the ratios between palmitic to stearic acid (P/S) and the presence/absence of peculiar fatty acids or bio-markers [34, 37, 43], samples from artworks were found to contain different kind of drying oils. Specifically, sample A2 presented the typical chromatographic profile of a cold-pressed linseed oil-based paint, with a P/S = 1.87. Sample A7, on the other hand, was characterised by the presence of specific markers related to stand-linseed oil, such as 7-(o-pentylphenyl)-heptanoic acid and 9-(o-propylphenyl)-nonanoic acid [53]. Safflower oil was likely identified in sample A9 due to the P/S = 2.20 and a significant presence of behenic and arachidic acids [39]. This diversity of siccative oils used reflects the complexity of the composition of twentieth century manufactured artists’ oil paints, as reported in the specific literature [34, 37, 43, 54]. Similarly, in the samples taken from the artist's oil paint tubes, the presence of a binder consisting of linseed oil was detected in A3 and A6, while safflower or sunflower oil was detected in sample A5. Sample A3 also contain an alkyd-based resin, whose typical components were identified by GC–MS. This could be related to an oil-modified alkyd resin or maybe due to a contamination.

The complexity of the commercial paints analysed was determined also by the detection of several organic additives. For example, in sample A3 traces of beeswax were detected, likely used as a paint stabiliser. In sample A5, a Pinaceae resin (i.e., colophony) was found, probably added as a thickener. Lastly, sample A6 was found to contain metal stearates added as dispersing agents [39].

Linseed oil and a Pinaceae resin were found in the samples taken from the artist's palettes. The resin may be present as a commercial additive or added by the artist for the purpose of thickening the paint on the palette [34]. Different drying rates were registered, according to the Azelaic to Palmitic (A/P) and Oleic to Stearic (O/S) molar ratios, which may be related to both the different thickness of the samples and their exposure to oxygen [55, 56]. In addition, as is well known, the presence of metal cations can strongly influence the drying of oil paints [57]. It is significant to note that in correspondence with the identification of Zn by elemental analysis, there was also a higher abundance of oleic acid, which can be quantified by the O/S molar ratio, which was already reported in literature as typical of twentieth century Zn-containing white paint [58, 59].

Photoluminescence properties

Under UV illumination, many of the paintings present a bright pink/orange luminescence in the deteriorated CdY areas (Fig. 5). This peculiar and intense emission has been observed also in other paintings with degraded CdY paints [20, 60]. Its origin has been related to surface crystal defects in reactive cadmium yellow paints, possibly manufacturing using nanocrystalline or poorly crystalline CdS pigments [21, 22].

Fig. 5
figure 5

Deteriorated yellow pigment in artworks Untitled (FPJM-00135) (a) and Untitled (FPJM-00034) (b) with yellow paints details (white dotted rectangles) under visible light and UV light, showing a peculiar pink/orange luminescence

Full size image

To characterise the luminescent properties of the paints, all samples were analysed with time-gated hyperspectral micro-imaging and results are summarised in Table 2. Paint tube samples containing crystalline CdS (samples A3 and A6) have a sharp nanosecond emission (peaked at 515 and 517 nm, respectively) and a microsecond emission spectrally broad and peaked above 850 nm. These two emission bands are related to the near band edge (NBE) and deep trap states (DTS) emissions of pure CdS, respectively [50, 61]. Instead, poorly crystalline CdY paint tube (sample A5) shows a weak NBE emission peaked at 485 nm, and a broad microsecond emission peaked above 750 nm, ascribed to Cd1−xZnxS pigment (with x ~ 0.16) accordingly to the position of the NBE emission [50, 61].

On their turn, most of the degraded samples from artworks and palettes present an emission, occurring at the microsecond timescale, heterogeneously distributed along the paint stratigraphy (Additional file 1: Fig. S9). Indeed, as illustrated in Fig. 6a for sample A7, the paint surface displays an emission much more intense than the one from the bulk, suggesting that it is linked to degradation at the paint surface. The strong intensity of the emission at the sample surface, which optically diffuses in the surrounding paint layers, prevents a quantitative comparison of the spectral and lifetime behaviour from the innermost parts of the paint. Despite this, it is worth noting that the emission at the paint surface, peaked at around 650–700 nm, occurs at shorter wavelengths than the emission observed in the poorly crystalline tube sample A5 (Fig. 6b). Indeed, the features of this intense microsecond emission resembles the orangish emission observed in other artworks with degraded CdY [21] and can be ascribed to DTS emission of cadmium-based paints with a high density of crystal defects. It is worth noting that, due to the absence of NBE emission, it was not possible to exactly establish the possible Zn content on the basis of the NBE shift [50]. Nonetheless, the low percentage of Zn detected through elemental analysis does not account for the shift observed in the trap state emission.

Fig. 6
figure 6

a Cross section of sample A7 under visible and UV light. b Time-gated photoluminescence microsecond emission spectra of samples A5 and A7 (surface)

Full size image

Summary of the multi-analytical characterisation

The multi-analytical approach adopted here has allowed us to provide an interpretation of each samples analysed, summarised in Table 3. Samples from artworks, palettes and from the paint tube by Lefebvre-Foinet are made of poorly crystalline Cd1−xZnxS, with a low percentage of zinc, as demonstrated by the combined results from elemental analysis, XRD and Raman. The presence of Zn has been associated with the cadmium pigment in solid solution Cd1−xZnxS, rather than to the widely diffused white pigment zinc white (ZnO), since this compound was not detected by any of the techniques used. Contrastingly, Mir brand paints (samples A3 and A6), were found to be composed of crystalline CdS paint with the presence of MgCO3 as an extender (as highlighted by EDX analysis and FTIR).

Table 3 Interpretation of the results achieved through the multi-analytical characterisation performed on the samples
Full size table

All pigments are dispersed in an oleic matrix (FTIR, Raman and GC–MS) with different kind of organic additives (GC–MS). Samples from artworks, palettes and from the paint tube by Lefebvre-Foinet also contain degradation products and compounds due to oil/pigment interaction as cadmium sulfate hydrate (FTIR, SR µXRD and Raman), carboxylates and oxalates (FTIR), while samples A3 and A6 show only the presence of additives such as magnesium carbonate (FTIR, EDX) and small presence of oxalates (sample A6).

Discussion

Based on similarity in chemical composition and the crystalline structure, it is reasonable to assume that Miró used Lefebvre-Foinet paint tubes in the artworks that have deteriorated. Until now, it had been assumed that the deterioration was caused by Miró’s use of a cheap or poor-quality yellow paints [62]. However, Lucien Lefebvre-Foinet was a Parisian manufacture of high-quality paints used by many renowned artists, including Jean Paul Riopelle, Piet Mondrian, Henri Matisse and Alberto Giacometti [63]. More than a hundred oil tubes from this brand are conserved in Miró’s studios, five of which are of Cadmium Yellow Lemon No.1, suggesting that this was one of Miró's favourite brands. The low crystallinity of the pigment may be related to the method used in its synthesis [24], which has enhanced the reactivity of the resulting paints. However, no information is available on the synthesis methods used by the manufacturer. An inscription on some of the boxes containing the paints suggests that they could have been purchased in 1966. To date, the degradation of CdY paint has always been related to the use of imperfect pigments synthetised during the early production period of CdY (between the middle of the nineteenth century and the beginning of the twentieth century) [11, 13, 15]. Pigment produced after this period and after issuing the Cd1−xZnxS patent were considered more stable. This research instead clearly shows that reactive pigments were also produced in later times.

In the early twentieth century, Cd1−xZnxS-containing paints were manufactured to obtain lighter cadmium yellow shades. The presence of zinc in CdY paints of artworks and palettes may have further influenced the stability of the paint. In fact, previous studies have shown that the stability of the alteration products of CdY paints depends on the Cd/Zn stoichiometry [23], demonstrating that artificially aged Cd1−xZnxS samples degrade more than samples composed of CdS, when aged in the same environmental conditions. However, it has also been shown that the only presence of zinc due to the use of Cd1−xZnxS does not induce the paint degradation, since the ageing conditions strongly influence CdY paint stability (especially light and high levels of relative humidity [23]). Therefore, it is reasonable to suggest that the presence of zinc, the poorly crystalline (hence reactive) pigment and the environmental conditions may have all contributed to the degradation of CdY paints.

The strong influence of environmental conditions in promoting the degradation of cadmium yellow paints is demonstrated by comparing the two palettes studied. The yellow in one of the palettes (sample A4) is still vibrant and non-degraded, while the yellow in the second palette (sample A12) is visibly degraded, presenting powdery paint and significant discolouration of the vibrant yellow. Despite this, the chemical characterisation of samples from the two palettes showed that they have similar chemical and crystalline properties. Both paints contain poorly crystalline Cd1−xZnxS together with carboxylates and sulfates, the latter less abundant in A4 sample and mainly concentrated in the area with Zn. The two palettes were kept in different environmental conditions: a factor that has strongly influenced the relative stability of the paint in each case. The palette containing the vibrant yellow paint was kept inside a display case with no direct sunlight from 1986 until 2018, when it was returned to the studio Taller Sert. Therefore, it was protected by the display case for 32 years, and not exposed to the more extreme fluctuations of relative humidity and intense light levels in the Taller Sert to which the second palette was exposed. These findings are in line with the conclusions of the study by Monico et al. [23] in which they state that the photooxidation of cadmium sulfide to cadmium sulfate is triggered by light and high levels of relative humidity; however, it should be noted that traces of hydrated cadmium sulfate were also detected in samples A4 and A5. This compound is known to have been used in the wet synthesis process of CdY and its presence in non-degraded paints may be due to unreacted starting reagents [11, 16].

Poorly crystalline samples showed weak luminescence emission from the layer beneath the surface in CdS samples, while the degraded surface of the same samples presented an intense emission, different from the typical emission of CdS. Based on paint composition (mainly CdS and 3CdSO4·8H2O), the emission can be ascribed to a high density of surface defects in the degraded poorly crystalline CdS, rather than to other emitting components. Moreover, the comparison between less degraded samples (sample A4) and the sample from a paint tube containing poorly crystalline CdS (sample A5) suggests that the emission properties of CdS-based paints are affected by degradation and that the presence of the intense and peculiar orangish emission in cadmium yellow paints indicates paint degradation.

Conclusions and future perspectives

In this study, the results of analytical characterisation of CdY paint samples from artworks by Joan Miró and from oil paint tubes and palettes kept in his studios has provided evidence of the degradation of CdY paints in his paintings from the 1970s. This expands on earlier research into CdY alteration found in works from 1880 to 1920 [11, 13, 15]. The study has shown that the main compounds formed due to degradation are cadmium sulfates and zinc/cadmium oxalates and carboxylates. The analyses identified the pigment as poorly crystalline Cd1−xZnxS with low percentage of Zn. Although the cause of the chemical change leading to deterioration are difficult to assess due to the presence of multiple factors, we hypothesise that the degradation can be related either to (i) the poorly crystalline structure of the pigment—possibly as a result of the manufacturing method, (ii) the presence of zinc, and (iii) the environmental conditions to which artworks have been exposed (intense natural light and RH levels likely to reach 95%). The comparison of the degradation of the paint from palettes stored and displayed under different conditions strongly suggests that the environmental conditions are a major factor contributing to the deterioration of cadmium yellow paint.

This study has implications for the conservation of artworks by Miró and other artists with similar problems of degradation of CdY paints. It is essential that the artworks are stored and exhibited under conditions with controlled light and relative humidity. Cadmium sulfates, the deterioration products that are concentrated on the paint surface, are soluble, so any surface cleaning intervention can lead to their removal. Regarding the application of protective coatings, varnishes or consolidants, more research is required to determine if they are beneficial or detrimental. In a work by Van Gogh, the application of a varnish over early deteriorating yellow paint has resulted in the formation of a brown layer of cadmium oxalates between the paint and the varnish [14]. Methylcellulose has been applied as a consolidant and to enhance the colour in some of Miró's works at the Fundació with satisfactory results, therefore its effectiveness as a consolidant for degraded yellows could be studied and compared to other matte consolidants. Since most of Miro’s late works are unvarnished, glazing with UV-filtering glass can also be an effective solution to protect artworks with fragile and degraded surfaces.

Further research could also include artificially ageing tests (at different ageing steps) on paints prepared from the various paint tubes manufactured by Lucien Lefebvre-Foinet kept in the studios. These studies will aim to (i) determine if the paint employed in the artworks and in the paint tubes present the same degradation and under which conditions; (ii) determine the extent of degradation as a function of the ageing time; (iii) establish what can happen with the removal of soluble degradation products as cadmium sulfates and further ageing and assess possible side effect of highly soluble compounds. A program of stepped artificial ageing of paints prepared from Lucien Lefebvre-Foinet tube paints (sample A5) could help to verify if the peculiar photoluminescence develops because of degradation, and at which stage of degradation is detectable. If conclusive, this could lead to a possible use as an early diagnostic tool. Comparative analysis of yellow paint in artworks in the collection of the Fundació that have been kept under the same conditions as the one studied, but still present vibrant yellow, would also yield information on the differences in the pigment composition which have resulted in more stable paint.