2D X-ray and FTIR micro-analysis of the degradation of cadmium yellow pigment in paintings of Henri Matisse I N V I T E D P A P E R 2D X-ray and FTIR micro-analysis of the degradation of cadmium yellow pigment in paintings of Henri Matisse E. Pouyet1,2 • M. Cotte1,3 • B. Fayard1 • M. Salomé1 • F. Meirer4 • A. Mehta5 • E. S. Uffelman6 • A. Hull7 • F. Vanmeert8 • J. Kieffer1 • M. Burghammer1 • K. Janssens8 • F. Sette1 • J. Mass9 Received: 20 March 2015 / Accepted: 13 May 2015 / Published online: 3 June 2015 � Springer-Verlag Berlin Heidelberg 2015 Abstract The chemical and physical alterations of cadmium yellow (CdS) paints in Henri Matisse’s The Joy of Life (1905–1906, The Barnes Foundation) have been recognized since 2006, when a survey by portable X-ray fluorescence identified this pigment in all altered regions of the monumental painting. This alteration is visible as fad- ing, discoloration, chalking, flaking, and spalling of several regions of light to medium yellow paint. Since that time, synchrotron radiation-based techniques including elemen- tal and spectroscopic imaging, as well as X-ray scattering have been employed to locate and identify the alteration products observed in this and related works by Henri Matisse. This information is necessary to formulate one or multiple mechanisms for degradation of Matisse’s paints from this period, and thus ensure proper environmental conditions for the storage and the display of his works. This paper focuses on 2D full-field X-ray Near Edge Structure imaging, 2D micro-X-ray Diffraction, X-ray Fluorescence, and Fourier Transform Infra-red imaging of the altered paint layers to address one of the long-standing questions about cadmium yellow alteration—the roles of cadmium carbonates and cadmium sulphates found in the altered paint layers. These compounds have often been assumed to be photo-oxidation products, but could also be residual starting reagents from an indirect wet process synthesis of CdS. The data presented here allow identifying and map- ping the location of cadmium carbonates, cadmium chlo- rides, cadmium oxalates, cadmium sulphates, and cadmium sulphides in thin sections of altered cadmium yellow paints from The Joy of Life and Matisse’s Flower Piece (1906, The Barnes Foundation). Distribution of various cadmium compounds confirms that cadmium carbonates and sul- phates are photo-degradation products in The Joy of Life, whereas in Flower Piece, cadmium carbonates appear to have been a [(partially) unreacted] starting reagent for the yellow paint, a role previously suggested in other altered yellow paints. Electronic supplementary material The online version of this article (doi:10.1007/s00339-015-9239-4) contains supplementary material, which is available to authorized users. & E. Pouyet emelinepouyet@gmail.com 1 European Synchrotron Radiation Facility, 6, rue Jules Horowitz, 38000 Grenoble, France 2 ARC-Nucléart - CEA/Grenoble, 17 rue des Martyrs, 38054 Grenoble Cedex 9, France 3 LAMS (Laboratoire d’Archéologie Moléculaire et Structurale) UMR-8220, 3 rue Galilée, 94200 Ivry-sur-Seine, France 4 Inorganic Chemistry and Catalysis, Debye Institute for Nanomaterials Science, Utrecht University, Universiteitsweg 99, 3584 CG Utrecht, The Netherlands 5 Stanford Synchrotron radiation Lightsource, SLAC National Accelerator Laboratory, 2575 Sand Hill Rd., Menlo Park, CA, USA 6 Department of Chemistry and Biochemistry, Washington and Lee University, Lexington, VA 24450, USA 7 Department of Chemistry and Biochemistry, University of Delaware, Newark, DE, USA 8 AXES Research Group, Department of Chemistry, University of Antwerp, Groenenborgerlaan 171, 2020 Antwerp, Belgium 9 Scientific Research and Analysis Laboratory, Conservation Department, Winterthur Museum, Winterthur, DE 19735, USA 123 Appl. Phys. A (2015) 121:967–980 DOI 10.1007/s00339-015-9239-4 http://dx.doi.org/10.1007/s00339-015-9239-4 http://crossmark.crossref.org/dialog/?doi=10.1007/s00339-015-9239-4&domain=pdf http://crossmark.crossref.org/dialog/?doi=10.1007/s00339-015-9239-4&domain=pdf 1 Introduction With the rapid rise of chemical industry during the end of the nineteenth century and the beginning of the twentieth century, numerous new inorganic and organic pigments were developed and introduced as alternatives to well- established traditional pigments, often outclassing them thanks to their colour intensity, purity, cost, and covering power. However, in Impressionist and early Modernist paintings, various synthetic inorganic pigments have started to undergo chemical and physical degradation phenomena ranging from fading and colour shifts to spal- ling and flaking. Several examples of the discoloration of synthetic yellow pigments from the turn of the twentieth century (e.g. zinc yellow (K2O�4ZnCrO4�3H2O) [1], chrome yellow (PbCrO4) [2–6], and cadmium yellow (CdS) [7, 8]) have been recently reported. In particular, despite having a good reputation regarding permanency, physical manifestations of photo-oxidative degradation of yellow cadmium sulphide (CdS) pigments have been observed over the past decade in works by Pablo Picasso, Vincent Van Gogh, Georges Seurat, Henri Matisse, Ferdinand Leger, Edvard Munch, and James En- sor [7–10]. This degradation appears in many different ways; from paint chalking, i.e. drying and crumbling, to fading, flaking, spalling, and in its most advanced cases to the formation of a thick (20–50 lm) ivory to tan alteration crusts covering the original yellow paint. The first systematic study of CdS pigment degradation was performed in 2005 by Leone et al. [9] on 12 paintings dating from 1887–1923. In this study, based on X-ray diffraction (XRD) and scanning electron microscopy com- bined with an energy-dispersive spectrometry (SEM–EDS), the presence of different oxidized Cd species, including cadmium carbonate (CdCO3), cadmium hydroxide (Cd(OH)2), and cadmium sulphate (CdSO4) was observed at the surface of paintings; these materials were identified as the main products of the degradation process. Combining these observations with the results from artificially degraded mock-ups and supplementary time-of-flight-secondary ion mass spectrometry (ToF-SIMS) analyses of the surface of the painting samples, a preliminary degradation mechanism was suggested. The photo-oxidation of the CdS pigment was proposed to generate CdO, CdSO4, and SO2 gas, which in high relative humidity environments, convert to H2SO4 re- sulting in acid hydrolysis of the paint binding medium. In 2009 and 2012, Van der Snickt et al. [7, 8] extended the study of CdS pigment degradation to paintings by James Ensor and Vincent van Gogh by using synchrotron radiation (SR)-based techniques. In the case of Still Life with Cabbage by James Ensor [7], the degradation of yellow CdS was related to the photo-oxidation of the cadmium sulphide to (hydrated) cadmium sulphate (CdSO4�nH2O). Repeated hydration and drying of the paint surface resulted in the formation of a thin (a few microns thick) layer of semi-transparent whitish globules of CdSO4�nH2O on the surface of the CdS-based paint. Another notable finding of this study was the identification of (NH4)2�Cd2(SO4)3 in the pigment layer, thought to be related to a previous aggressive cleaning treatment with dilute ammonia. Two supplementary degradation products were identified in the case of the Flowers in a blue vase by Vincent Van Gogh [8]: cadmium oxalate (CdC2O4) and lead sulphate (PbSO4). The presence of an apparently unoriginal varnish, likely applied after the initial dete- rioration of the CdS yellow, was the possible source of C2O4 2- and Pb 2? ions. These ions could have then reacted with Cd 2? and SO4 2- ions, produced during the initial photo- oxidation of CdS, leading to the formation of a thin layer of CdC2O4 on the top of the painting surface and the pre- cipitation of PbSO4 in the varnish layer. Simultaneously, Mass et al. [11], from the Conservation Department of the Winterthur Museum, initiated a study to characterize visible changes observed in cadmium yellow areas of The Barnes Foundation’s The Joy of Life (Henri Matisse, 1905–1906, Barnes Foundation 719, Fig. 1a), in particular for preservation purposes. Studies were per- formed at different synchrotron facilities: at ID21 (ESRF), at BL4-3 (SSRL) [11], and at the IRENI (Synchrotron Radiation Center, SRC) [12, 13]. By combining micro- Fourier transform infrared (lFTIR) spectroscopy with mi- cro-X-ray fluorescence (lXRF) and micro-X-ray near edge structure (lXANES) spectroscopy (at the Cd LIII-edge and at the S K-edge in scanning mode), spectra were collected over dozens of points. Degradation products were identi- fied in the altered cadmium yellow area from the darkened yellow foliage at the upper left of the painting (three samples: S111, S112, and S113 in Fig. 1b), the yellow fruit in the tree with an ivory-coloured alteration crust at the upper right, (one sample: S115 in Fig. 1b), and in the faded yellow field beneath the central reclining figures (two samples: S117 and S5 in Fig. 1b). CdCO3 was present in high concentrations in all altered regions, supporting the hypothesis that it is more likely a poorly soluble photo- degradation product than a filler or residual starting reagent. In the unaltered yellow paint, the identification of CdS and CdCl2�nH2O suggested that chloride was intro- duced as the starting reagent for the synthesis of the cad- mium yellow. CdSO4�nH2O was also found to be enriched in the off-white alteration layers of the samples studied; as a more soluble species, it was also found elsewhere in the cross sections. Alteration crusts identified had no remain- ing CdS, all of which had degraded into a mixture of cadmium sulphates, carbonates, and oxalates. 968 E. Pouyet et al. 123 These examples illustrate several advantages of com- bining synchrotron radiation-based X-ray techniques, such as lXRF, lXANES, and lXRD to elucidate the degrada- tion processes of paint containing CdS yellow pigment. First, combined techniques allow the identification of mi- nor and major components even when present as amor- phous/nanocrystalline materials (as in the case of CdS) or in a crystallized form. The sensitivity of these SR-tech- niques is critical for the identification of photo-alteration products in paintings, allowing for preventive conservation measures (such as closer management of light levels and relative humidity) to be implemented when degradation is observed. Secondly, the micrometric resolution of these probes is essential to reveal the presence of the degradation/alteration layer on the painting’s surface, which is typically only a few microns thick, and to establish the stratigraphy of the alteration/degradation products. The identification of CdCO3 and CdSO4- nH2O alone does not constitute a conclusive proof for a photo-oxidation process. Both compounds are known to be reagents in the wet and/or dry synthesis process of CdS. Thus, their identification as either unreacted starting reagents, side products of the original syn- thesis, or as degradation compounds relies mostly on their location and distribution in regards to the original paint layer. Mass et al. [11] demonstrated the interest of using a scanning probe to accurately localize and identify the stratigraphy of various species in order to decide whether a particular species can be categorized as unreacted starting reagents or as degradation compounds. However, the recent use of 2D full-field XANES (FF-XANES), combined with 2D lFTIR, lXRF and lXRD measure- ments when speciation through XANES is ambiguous, has allowed faster and more detailed identification and dis- tribution of various chemical compounds, even in highly complex and heterogeneous sections. The new approach allows for a deeper elucidation of the degradation phe- nomena under way, leading to more specific preservation recommendations, and the ability to identify degradation at an earlier stage. In particular, the full-field analysis appears to be an ideal way to extend this study over an entire 2D region and to eliminate some of the ambiguities that are inherently present when only a point-by-point analysis is performed. Fig. 1 a Henri Matisse, French, 1869–1954 The Joy of Life, between October 1905 and March 1906, oil on canvas, 69� 9 94� in. (176.5 9 240.7 cm), oil on canvas, The Barnes Foundation, BF719; c upper left zoom showing tan-brown alteration crusts on the yellow foliage and on the yellow fruit in the tree at the upper right, and zoom on the faded region below the central reclining figures, sampling locations for this study are represented by black cross; b Henri Matisse, French, 1869–1954 Flower Piece, 1906, 21 7/8 9 18� in. (55.6 9 46.4 cm), oil on canvas, The Barnes Foundation, BF205; d altered (brown) and non-altered (yellow) regions of yellow paint to the right of the pitcher, sampling locations are represented by black cross 2D X-ray and FTIR micro-analysis of the degradation of cadmium yellow pigment in paintings of… 969 123 2 Experimental 2.1 Paintings and samples location In order to complete the identification and localization of the original paint components and the degradation mate- rials, two micrometric fragments S111 and S5 were sampled from The Joy of Life painting (Fig. 1a), in the altered cadmium yellow paint from the darkened yellow foliage at the upper left of the work, and in the faded yellow field beneath the central reclining figures, re- spectively (Fig. 1b). Two other fragments (named BF205-darkened and BF205-undarkened) were sampled from the Flower Piece painting (Fig. 1c), from the dry and cracked darkened re- gion below the right side of the pitcher, and the intact yellow region above this darkened area (Fig. 1d), respec- tively. This sampling strategy was devised to compare degradation phenomena in two of Matisse’s artworks showing two different stages of photo-degradation and to obtain information on the cadmium yellow pigment syn- thesis in the paints used by Matisse. 2.2 Sample preparation and mounting In previous work [11], X-ray analyses were carried out on the surfaces of polished cross sections of painting fragments embedded in polyester resin, the most classical approach to prepare transversal sections from paint fragments. One unpublished result obtained following this sample prepa- ration method is detailed below for sample S5. The ex- amination of samples from historical paintings using combined elemental, molecular, and structural methods benefits from the preparation of thin sections [14]. Ideally, hyperspectral analyses are performed on a unique section. However, the sample preparation requirements (thickness, embedding media, etc.) differ as a function of the tech- niques involved [15]. Accordingly, in the present study, two adapted thin sections were prepared: one for combined XANES and XRD analyses, and another one for FTIR analyses. For X-ray-based techniques, the sample was first embedded using synthetic resin (Historesin, Leica) and microtomed to obtain a section thinner than 20 lm. The section was then sandwiched between two ultralene foils (4 lm thick, from Spex Sample Prep) to provide me- chanical stability during data acquisition. For lFTIR ana- lysis, micro-compression was favoured because it prevents spectral contamination from the embedding material [12, 13]. Considering that lXRF results are not affected by any of these forms of sample preparation (compressed sample between diamond windows, cross section, and thin section from embedded fragment), the resulting elemental maps allow the combination of results from X-ray and infrared spectroscopies together with the visible observations. 2.3 Analytical methods and data processing 2.3.1 Micro-X-ray fluorescence and chemical maps in scanning mode lXRF data were collected at the X-ray micro-spectroscopy beamline ID21 at the European Synchrotron Radiation Facility (ESRF) [16]. Monochromatic radiation was ex- tracted from an undulator coupled with a fixed-exit double- crystal monochromator equipped with a Si(111) crystal and focused using a Kirkpatrick-Baez mirror pair to a typical beam size of 0.2 lm ver. 9 0.6 lm hor. with a flux rang- ing from 10 9 to 10 10 ph/s and an energy ranging from 2 to 9 keV. The samples were mounted in a vertical plane, at an angle of 60� with respect to the beam, and raster scanned in the beam using a combination of stepper motors and piezo actuators. The XRF data were collected by a single channel solid state detector. This allowed the distribution of the different elements present in the paint cross sections, in particular of the Cd- and S-containing compounds, to be mapped at sub-micrometre resolution. In a few cases, the elemental composition maps were complemented by sul- phur speciation maps. Two maps were recorded setting the energy of the incoming X-rays to 2.4728 and 2.4825 keV, respectively, i.e. to energies where the absorption and consecutively the XRF from sulphides (S 2- ) or sulphates (S 6? ) is enhanced. By assuming a binary composition consisting of CdS and CdSO4�nH2O and determining the absorption of these two compounds at these two specific energies, it is possible to derive the concentration maps of CdS and CdSO4�nH2O, as detailed elsewhere [7, 17]. The XRF data were batch fitted using the PyMca software package [18]. 2.3.2 Full-field XANES In FF-XANES measurements, a stack of X-ray radiographs of a thin sample is acquired, while tuning the X-ray energy across the absorption edge of the element of interest [19]. This novel technique allows the acquisitions of millions of XANES spectra in a very short time (less than 1 h per acquisition) and makes possible the acquisition of full XANES spectra at the micrometre scale over millimetric 2D regions. FF-XANES data were collected at ID21 in the full-field configuration, across the S K-edge, Cl K-edge, and Cd LIII- edge. The beam size of 1.5 9 1.5 mm 2 was defined using slits; the spatial structures of the beam were smoothed using an X-ray decoheror (rotating graphite foil, 125 lm 970 E. Pouyet et al. 123 thick). Samples were mounted vertically, perpendicular to the beam, and radiographs were acquired with a detection ensemble comprising a scintillator, a magnifying objective, and a CCD camera. It resulted in images with a pixel size of 300 9 300 nm 2 and a field of view of 315 9 360 lm2. For data analysis, the 10 6 XANES spectra acquired in full-field mode were fitted as a linear combination (LC) of references, using the TXM-Wizard software [20]. The primary set of Cd reference compounds (spectra presented in Fig. 2) was composed of: cadmium sulphate (CdSO4, 99.99 % metals basis), hydrated cadmium sulphate (CdSO4�nH2O, 99.999 % metals basis), cadmium sulphide (CdS, 99.995 % metals basis), cadmium nitrate tetrahy- drate (Cd(NO3)2�4H2O, [99.999 % metals basis), cadmi- um oxide (CdO, [99.99 % metals basis), hydrated cadmium chloride (CdCl2�nH2O, [99.99 % metals basis), cadmium carbonate (CdCO3, [99.999 % metals basis) (all purchased from Sigma Aldrich) and cadmium oxalate (synthesized at Washington and Lee University). The final choice of references for the least squares fitting (LSLC) varies from one sample to another, with regards to com- plementary lXRD and lFTIR results. In the case of FF- XANES measurements across the S K-edge, the low en- ergy of incoming photons led to a strong absorption by the sample. Even the thinnest section obtained by microtome was insufficiently thin to prevent over-absorption effects, drastically reducing the white line intensity of sulphur species, distorting spectral features, and making LSLC analysis untenable. For these data sets, the selection of Region of Interest (ROI) characteristics of the different sulphur species present was preferred as a qualitative approach. At the Cd LIII-edge and Cl K-edge, the transmission obtained on a 10 lm section was sufficient to allow LSLC treatment. One of the most important assets of LSLC is its quantitative character, since it provides information on the relative contribution of each reference present in the mix- ture. However, this strategy presents an important draw- back, as it necessitates an a priori knowledge of the phases constituting the sample. As for single XANES LC fitting, the method also relies on: (1) an accurate energy calibra- tion of references and data sets; (2) a reliable choice of references (guided for example by complementary analyses by methods such as XRD, FTIR, and Raman); and (3) a reference set as close as possible to the actual composition of the sample itself to minimize differences in the back- ground shape and in the pre- and post-edge regions be- tween references and data sets. These constraints and the massive data set size moti- vated the use of another strategy for data analyses in the case of sample S111, namely principal component analyses (PCA) and subsequent k-means clustering. Pixels with similar XANES spectra were pooled in reduced PC space (in the score plot), effectively segmenting the image based on the variance in the recorded XANES features into a pre-defined number of regions (k areas) consisting of pixels with a similar XANES signature. PCA and clustering were performed using the TXM-Wizard soft- ware [21]. 3530 3580 3630 N or m al iz ed a bs or pt io n Energy (eV) CdSO4·nH2O CdSO4 Cd(NO3)2·4H2O Cd(OH)2 CdC2O4 CdCl2·nH2O CdCl2 CdO CdS CdCO3 Fig. 2 Cd LIII-edge XANES spectra acquired in transmission mode using FF-XANES technique on various cadmium references: cadmi- um sulphate (CdSO4, 99.99 % metals basis), hydrated cadmium sulphate (CdSO4�nH2O, 99.999 % metals basis), cadmium sulphide (CdS, 99.995 % metals basis), cadmium nitrate tetrahydrate (Cd(NO3)2�4H2O, [99.999 % metals basis), cadmium chloride (CdCl2, [99.99 % metals basis), hydrated cadmium chloride (CdCl2- nH2O, [99.99 % metals basis), cadmium oxide (CdO, [99.99 % metals basis), cadmium carbonate (CdCO3, [99.999 % metals basis) (all purchased from Sigma Aldrich), and cadmium oxalates (synthe- sized at Washington and Lee University) 2D X-ray and FTIR micro-analysis of the degradation of cadmium yellow pigment in paintings of… 971 123 2.3.3 Micro-X-ray diffraction Some samples (BF205 darkened and undarkened) were further studied by lXRD at the ID13 (ESRF) and P06 (PETRA III) beamlines. At ID13, the X-ray beam energy of 12.9 keV was selected by means of a Si(111) double- crystal monochromator. The beam was focused with a Kirkpatrick-Baez mirror optic down to 2.5 9 1.5 lm2 (hor. 9 ver.). Diffraction signals were recorded in trans- mission geometry with a 2 k 9 2 k ESRF FReLoN de- tector (50.0 (h) 9 49.3 (v) lm2 pixel size). At P06, a hard-X-ray micro- and nanoprobe beamline at the PETRA III storage ring (DESY, Hamburg, Germany), the X-ray beam energy of 21 keV was selected by means of a Si(111) double-crystal monochromator [22]. The beam was focused with a Kirkpatrick-Baez mirror optic down to 0.6 9 0.8 lm2 (hor. 9 ver.). Diffraction signals were recorded in transmission geometry with a PILATUS 300 K area detector. Data were unwrapped using both XRDUA [23] and PyFAI [24]. The Match! and EVA packages were used as well during the preliminary phase identification. 2D maps of compounds were then generated with PyMca by using ROIs (each ROI corresponding to a particular diffraction angle range). These ranges were chosen based on the peak intensity and the absence of overlaps with other phases (defined in Online Resource (1)). 2.3.4 Micro-FTIR analyses Organic compounds are known to play an important role in the degradation mechanisms of paint layers (e.g. through hydrolysis of the drying oil binder); moreover, some of the alteration products are organometallic compounds. Thus, complementary lFTIR analyses in the mid-IR domain were performed at ID21. The FTIR spectromicroscope is based on a commercial instrument and is composed of a Thermo Nicolet Nexus infrared bench associated with an infrared Thermo Continuum microscope [25]. The infrared beam was emitted from a short straight section (containing fo- cusing electron lenses) upstream of a bending magnet of the ESRF ring. The edge radiation was collected, collimated, and transferred to the spectrometer and microscope using a set of 12 mirrors. In the microscope, a 932 Schwarzschild objective was used in confocal mode; an aperture defined the size of the spot illuminating the sample. The signal was detected using a liquid N2-cooled single element 50 lm MCT detector. In this configuration, the beam size was 8 9 8 lm2 and FTIR spectra were acquired in transmission mode using the diamond compression technique. The OMNIC and PyMca packages were used for data analysis. 3 Results 3.1 The Joy of Life (1905–1906): study of the degradation process 3.1.1 S5 sample: faded yellow field beneath the central reclining figures Following the above strategy, analyses were performed on two thin sections from the same initial S5 fragment, which had previously been analysed in scanning mode as a cross section [11]. lFTIR was performed on a fragment prepared with the micro-compression cell (Fig. 3a). Chemical maps (record- ed with a step size of 6 9 6 lm2) are presented in Fig. 3d. FF-XANES was acquired on an embedded thin section of 10 lm (Fig. 3f), at the Cd LIII-edge and the Cl K-edge (pixel size 0.6 9 0.6 lm2). Cd LIII-spectra were fitted with the following set of references: CdS, CdCO3, CdCl2�nH2O, and CdSO4�nH2O based on the previous lXANES results [11]. Results of LSLC fitting are presented in Fig. 3i. lXRF maps were acquired at 7.2 keV (step size: 1 9 1 lm2) on the two sections, allowing, together with the optical observations, the correlation of results obtained with the three techniques. The elemental maps obtained on the pressed sample and the thin embedded section are shown in Fig. 3c, h, respectively. Additional speciation maps obtained in lXRF mode at the S K-edge from the same sample but prepared as a cross section, (pixel size: 1 9 1 lm2) are presented in Fig. 4. The visible images and chemical maps reveal a complex mixture and stratigraphy of Cd-based compounds in the predominately unaltered yellow region and the alteration white crust (Fig. 3b, g). In the yellow internal region, Cd is highly concentrated (lXRF, Fig. 3h) and mainly present as CdS (confirmed by XANES at Cd LIII-edge, Fig. 3i and Online Resource (2)). XANES at the Cd LIII-edge allows the identification of other cadmium-based species as well: carbonates (con- firmed by lFTIR, Fig. 3d), sulphates (confirmed by lFTIR, Fig. 3d, and chemical maps at S K-edge, Fig. 4), and chlorides, CdCl2 (confirmed by lXRF, Fig. 3c and h, and XANES at the Cl K-edge, Online Resource(3)). For sul- phates and chlorides, improved data fitting was obtained when spectra of hydrated references were employed. The imaging capability combined with micrometric resolution achieved with the full-field microscope reveals that CdS is heterogeneously present in the yellow layer and is intermixed with CdCO3 and CdSO4�nH2O (Fig. 3i). Cadmium chlorides have been previously identified in several regions of this painting (sample S115: yellow fruit in the tree at the upper right and sample S112:darkened 972 E. Pouyet et al. 123 upper left corner) and are thought to be the residual starting reagent from the wet process synthesis of the CdS used for this painting [11]. This conclusion is also supported by our present findings where cadmium chlorides are found in the internal yellow region, rather than accumulated in the al- teration white crust (which would be the case if introduced as a contaminant from the environment). Concerning the white altered layer, chemical mapping highlights a complex stratigraphy mainly composed of CdCO3 (lFTIR, Fig. 3d, e point1, and XANES at Cd LIII- edge, Fig. 3i and Online Resource(2)). In the upper part of the white layer, both sulphates and oxalates are also identified (lFTIR, Fig. 3d, XANES at the Cd LIII-edge, Fig. 3i and Online Resource (2), and chemical maps at the C2O4CO32- SO42- Remainder CdCO3 CdS O-H = Carbonates Sulfates Cl, Cd Acrylic polymer Carbonates, Cd Zein Oxalates Carbonates Sulfates Cl, Cd S Cl Cd S Cl Cd CdS, CdCO3, CdSO4·nH2O, CdCl2·nH2O, Cd, K, S, Cl CdCO3 Cd, K, Cl CdC2O4 CdSO4·nH2O Cd, S, Ca, Si 50 μm 50 μm (b)(a) (c) (g) (d) (f) (h) (i) O C-O-C = O C-N (e) 1 3 2 Point 1 Point 2 Point 3 900140019002400290034003900 A bs or ba nc e (a rb it . u ni ts ) Wavenumber (cm-1) Fig. 3 Combination of SR lFTIR, lXRF, and FF-XANES for the study of a fragment from The Joy of Life (S5). Schematic views of results obtained with b lFTIR and lXRF; g FF- XANES at the Cd LIII-edge and lXRF; c–e results from the compressed frag- ment displayed in a Visible light microscope image; c lXRF elemental maps of Cd, S, and Cl (step size: 1 9 1 lm2). d lFTIR maps of esters (1710–1750 cm -1 ), amides (1616–1700 cm -1 ), hydroxyls (3320–3450 cm -1 ), sulphates (1022–1206 cm -1 ), oxalates (1306–1327 cm -1 ), and carbonates (1342–1535 cm -1 ) (step size: 6 9 6 lm2); e IR spectra of carbonates (point 1), poly (vinyl acetate) polymer (point 2), and zein (point 3)-rich area; f–i results from the 10-lm-thick section displayed in f Visible light microscope image. h lXRF elemental maps of Cd, S and Cl (step size: 1 9 1 lm2); i Speciation maps (FF-XANES) of CdCO3, CdS as well as the sum of the remainder Cd components used in LSLC fitting (CdCl2�nH2O and CdSO4�nH2O) (pixel size: 0.3 9 0.3 lm2) 2D X-ray and FTIR micro-analysis of the degradation of cadmium yellow pigment in paintings of… 973 123 S K-edge, Fig. 4). Since the oxalates, carbonates, and sulphates are all colourless, it explains the chemical cause of the fading observed in the altered region. The cadmium carbonates are enriched in the white su- perficial layer (Fig. 3d, i) where CdS is completely absent (Figs. 3i, Fig. 4). They can also be observed in smaller quantities in the yellow paint layer. This distribution con- firms similar results found in the altered yellow fruits from the same work, suggesting that in the off-white alteration layer, the cadmium carbonate is a photo-degradation pro- duct, even though it has been suggested as a residue of the CdS synthesis in other systems [10, 11]. The presence of CdC2O4 is limited to the uppermost alteration layer (Fig. 3d, i and Online Resource (2)). Also observed in the darkened foliage in the upper left corner of The Joy of Life [11] and a painting by Van Gogh [8], this product is identified as a degradation product, derived from either varnish residues on the painting (residues of a par- tially removed natural resin varnish have been observed in several paint cross sectional samples removed from the piece) or from the oil binder. Such degradation could be the result of a cleaning treatment, natural ageing, or the breakdown of the binder during the photo-degradation process. As seen in earlier examples of altered cadmium sul- phide-containing paints, cadmium sulphates are distributed throughout the paint layer [11] (Fig. 3d, i), see, for ex- ample, the data from sample S115. The distribution of sulphide and sulphate species (Fig. 4), obtained in lXRF mode on cross sections [7]), shows that sulphates (pre- sumably cadmium sulphate) are dispersed in the original paint layer, but are also enriched on the surface of the off- white alteration crust. Cadmium sulphide is virtually absent in this upper layer, consistent with the absence of a yellow colour in this region, whereas it is intact beneath the alteration crust. These data are again consistent with cad- mium sulphates being photo-degradation products rather than residues of the CdS synthesis in The Joy of Life paints. lFTIR also revealed the presence of an organic com- pound made up of a poly(vinyl acetate) polymer (Fig. 3e, point 2) characterized by a strong peak in the C–O stretching absorption region (1300–900 cm -1 ). This pro- duct is likely a restoration/consolidation material used to reduce flaking. Flaking and spalling were particularly problematic in this region of the painting, and several campaigns of consolidation have been carried out, in par- ticular to allow the work to travel in 1992. In a Cd-free area of about 50 lm (Fig. 3c), the lFTIR spectra exhibit a pe- culiar feature, characteristic of amides. Comparison with databases suggests the presence of zein, characterized by a band at 3286 cm -1 from amide A (Fig. 3e point3). Zein is the major storage protein of maize and was proposed as a possible base material for polymer applications in the early twentieth century [26]; here, it is again interpreted as a restoration treatment—the painting is known to have been stabilized with a glue lining in the early twentieth century, prior to the polymer-based consolidation treatments. 3.1.2 S111 sample: darkened yellow foliage Another discoloured area of The Joy of Life (1905–1906), from the darkened upper left corner, was also sampled: sample S111. Since this fragment was very small, it was impossible to employ more than one sample preparation strategy. Priority was given to XANES analysis at the Cd and S edges, and thus, a 10-lm-thick thin section was prepared. Contrary to the previous fragment, no white al- teration zone was observed (Fig. 5a). The white thick layer visible in the optical image relates to the lead white-rich ground layer identified by lXRF analyses as a mixture of sulphide sulphate 50 μm 0,00 0,50 1,00 1,50 2,00 2,50 3,00 3,50 2460 2470 2480 2490 2500 2510 2520 N or m al iz ed a bs or p� on Energy (eV) CdSO4·nH2O CdS (b)(a) (c) 2.4728 2.4728 Fig. 4 lXRF maps of sulphur, in The Joy of Life (BF 719), sample S5 from below the central reclining figures. a UV picture of S5 cross section; b XRF maps were acquired at 2.4728, 2.4825, and 2.5189 keV highlighting the relative distribution of sulphides (red) and sulphates (green), respectively; c Reference XANES spectra at the S K-edge of CdS and CdSO4�nH2O. 974 E. Pouyet et al. 123 Ba- and S-containing coarse grains (previously identified as barium sulphates, Fig. 5b), dispersed in a Pb white- based matrix (lead white, Fig. 5b). The degradation seems to be limited to a few micrometres on the uppermost part of the Cd yellow paint and is related to the browning of the original yellow pigment. The section has been analysed using FF-XANES at the S, Cl K-edge, and Cd LIII-edges. Supplementary lXRF maps were also acquired at a primary energy of 3.7 keV. After exposing the sample to these multiple XANES and XRF acquisitions, the embedding resin started to lose its mechanical strength, thus preventing further investigation using other techniques such as FTIR or XRD. At the S K-edge, over-absorption issues prevented reliable LSLC fitting, limiting data analysis to the localization of both sulphide and sulphate species based on ROI calculations (as described in Online Resource (4); results not shown). At the Cl K-edge, the main information obtained was the identification of CdCl2�nH2O as the single chlorine-con- taining compound present. Both results were used to con- firm the results obtained at the Cd LIII-edge. [At the Cd LIII-edge, some pixels suffered from over-absorption and were set to zero (and therefore not taken into account during data fitting).] LSLC, using similar references to those used for the S5 sample, presented a large difference between experimental data and fit results, suggesting the presence of species not covered by the reference data set. In order to identify the relevant cadmium spectral basis set, principal component analysis was used. PCA followed by k-means clustering using the PC basis identified four main clusters (Fig. 5c): clusters 1, 3, 5, and 6 (cluster 2 and 4 are related to low absorbing area with noisy pixels and were not retained for further analysis). From each cluster, the average XANES of the cluster was extracted and fitted with the complete list of references presented above. Mixtures of CdS, CdSO4, Cluster 6Cluster 5Cluster 2Cluster 3 Cluster 4Cluster 1 Cd K Pb S S Cl(b)(a) (c) 40 μm 40 μm 40 μm 40 μm 0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8 3535 3585 3635 N or m al iz ed a bs or pt io n Energy (eV) XANES of cluster 1 XANES of cluster 3 XANES of cluster 5 XANES of cluster 6 Fig. 5 a Visible light microscope images of S111 10 lm-thick thin section; b lXRF elemental maps of Cl, K, S, Cd, and Pb acquired at 3.7 keV(step size: 0.7 9 0.7 lm2); c Results of k-means clustering combined with PCA analyses of FF-XANES acquisition at Cd LIII- edge (step size: 0.7 9 0.7 lm2) 2D X-ray and FTIR micro-analysis of the degradation of cadmium yellow pigment in paintings of… 975 123 CdSO4�nH2O, CdCO3, and CdCl2�nH2O were able to fit the XANES for the four clusters adequately. The full spectral stack was then refitted using the spectra for the clusters identified above, leading to a significantly improved fit. Cluster 1 is representative of most of the sample. Its content of CdCO3 is higher than that of any other cluster; it also contains traces of CdS and CdCl2�nH2O. On the sur- face of this cluster is cluster 3, which displays a similar composition but with added CdSO4�nH2O, again in sig- nificantly lower concentration than CdCO3. Below the surface of the left corner (yellow), cluster 5 reveals an area where the CdS content increases but still contains a relatively high amount of CdCO3. Consequently, even in the CdS-rich yellow region, the CdCO3 is found in high amounts, suggesting that a large fraction of the CdCO3 present is possibly unreacted starting reagent [11]. How- ever, further investigation is needed because the cadmium yellow hue in this region of the painting was intended to be a dark yellow/orange (based on comparison with a 1905 study of The Joy of Life at The San Francisco Museum of Modern Art and the Statens Museum for Kunst, Copen- hagen), and CdCO3 has typically been observed at high concentrations in the paler shades of cadmium yellow. Another explanation for the CdCO3-rich interior of this yellow paint is the relatively thin paint layer being studied here compared to samples S115 [11] or S5, which represent more advanced states of alteration where discrete alteration crusts are visible. The entire yellow paint layer in this case may be in the process of converting into an alteration zone. On the surface of this area is cluster 6 that is highly concentrated in Cd (lXRF), present mainly as CdS, and CdCO3. The high concentration of CdS in this region of the sample may be due to the morphology of the brushstroke observed (Fig. 5a). The top of the sample would be more exposed to direct ambient light, whereas the texture/im- pasto of the paint that causes yellow paint to appear on both the side and top of this sample may have protected the side from direct exposure. Contrary to S5, this sample is missing the thick white alteration crust, maybe explaining the lack of a distinct layering structure for the alteration products. However, the distribution and amount of sulphates species suggest ad- vanced photo-degradation. The degradation mechanism appears to be a two-step process; in the first step, sulphides are directly photo-oxidized to sulphates, and in the second stage, sulphur-based species, in particular cadmium sul- phide, are completely lost, replaced by cadmium carbonate, as seen for cluster 1. The presence of some CdSO4 in cluster 5 suggests that the degradation process is still not complete. However, the visible colour of the degraded sample (ochre brown) is not directly linked here to the identification of Cd-based compounds. Consequently, supplementary lXRD analyses and organic phase analyses are now mandatory to fully understand the photo-degradation mechanism in this sample. 3.2 Flower Piece: degradation and synthesis processes study 3.2.1 Darkened BF205 sample Another micro-sample of cadmium yellow paint from the painting Flower Piece (1906), demonstrating photo- degradation, was taken from the dry, cracked, and darkened region below the right side of the painting’s pitcher. The sample was too brittle to allow sectioning without adding strengthening material, consequently the surface of the resin block was first covered with a sulphur-free tape in- suring the integrity of the 10 lm section during slicing with the microtome (Fig. 6a). The optical stratigraphy is very similar to that observed in sample S5, with an off- white layer at the surface of the painting. A residual ground layer was also identified on the left corner of the section. Four main experiments were carried out on this section in order to identify the degraded materials present: lXRF sulphur valence maps (step size 2 9 2 lm2); results are presented in Fig. 6b. lXRF at 7.3 keV (step size 2 9 2 lm2); elemental mapping results are illustrated in Fig. 6c. FF-XANES at Cd LIII-edge (pixel size 0.7 9 0.7 lm2); results of LSLC fitting are presented in Fig. 6d. lXRD 12.9 keV (ID13, step size 2 9 2 lm2); phase maps based on the integrated diffraction intensity over regions of interest are shown in Fig. 6e. Based on these four techniques, the painting’s ground layer, paint layer, and alteration layer could be chemically characterized. lXRD (Fig. 6e) combined with lXRF (Fig. 6c) allowed for the identification of sphalerite (ZnS), barite (Ba0.99Sr0.01(SO4)), anglesite (PbSO4), and hydrocerussite (Pb3(CO3)2(OH)2) in the ground layer area. Two different ground layers were more accurately identified: the first one (innermost) contains mainly hydrocerussite with sphalerite and barite on which a very thin second layer containing hydrocerussite and barite with grains of anglesite is ap- plied. The presence of sphalerite and barite together sug- gests the use of lithopone, a co-precipitate of BaSO4 and ZnS introduced in 1874. The first layer may be a com- mercial ground, while the second has probably been pre- pared and applied by Matisse himself. The anglesite was also seen by l-FTIR in The Joy of Life and may result from the interaction of the unstable CdS with the lead white in the ground. By combining FF-XANES at ID21 at the Cd LIII-edge with lXRD and lXRF at sulphide and sulphate absorption- specific energies, the composition of the painting layer is 976 E. Pouyet et al. 123 defined mainly as a mixture of CdS, CdCO3, and CdSO4�nH2O. The white degraded area is mainly composed of CdCO3 with a small amount of CdC2O4 and CdSO4�nH2O, iden- tified by FF-XANES and lXRD, similar to what was ob- served for samples S5 and S115 [11]. A small increase in the CdC2O4 and CdSO4�nH2O content is observed at the surface of this area in FF-XANES data; however, no clear stratigraphy inside the degraded area was observed from the lXRD results. The high concentration of CdCO3 in the off-white al- teration crust of the paint layer may suggest that this compound as well as CdSO4�nH2O and CdC2O4 are the products of degradation processes. However, the presence of CdCO3 in a high amount in the paint layer as well suggests that at least some of it may have been initially present as a filler or residual starting reagent for this work. Though the former possibility is less likely since CdCO3, used as cadmium white, would have been an expensive filler [10]. 3.2.2 Undarkened BF205 sample Cadmium carbonate (CdCO3), as illustrated above, has been identified in the altered cadmium yellow (CdS) paints found in Impressionist, early modernist, and post-Impres- sionist works. As CdCO3 is highly insoluble (Ksp of 1.0 9 10 -12 ), when it is formed solely as a result of photo- alteration, it is mostly confined to the location where it is formed, i.e. at the surface of the paint layer. However, when an unclear stratigraphy is present and CdCO3 is distributed throughout the paint layer, conclusions about its origin in the paint layer are equally unclear. In cadmium yellow paint in works such as Edvard Munch’s The Scream (c. 1910, The Munch Museum, Oslo), the hypothesis has recently been proposed that CdCO3 was used in the indirect wet process synthesis of CdS (for example, through reac- tion of CdCO3 with Na2S) [10, 11]. To address the question of the origin of CdCO3, a flake of (to the naked eye) undegraded pale cadmium yellow paint was removed from Henri Matisse’s Flower Piece so that the distribution of CdCO3 could be studied, both as a function of depth in the paint layer and in individual pig- ment particles. The visible fluorescence of the ultraviolet- illuminated paint cross section removed from Flower Piece (Fig. 7a) shows that in the top half of this sample, the cadmium yellow is dispersed in a zinc white base and in- dividual cadmium sulphide particles are visible thanks to their orange fluorescence in the ultraviolet. A section of 15 lm thickness was prepared from a visually non-de- graded paint fragment from the lemon yellow area (Fig. 7b). Three main measurement types were carried out on this section to identify possible residual starting reagents: 300 μm Greenockite (CdS) Otavite (CdCO3) Hydrocerussite (Pb3(CO3)2(OH)2) Barite (B0.99Sr0.01(SO4) Cadmium oxalates (CdC2O4) Cadmium sulfates hydrated (CdSO4·nH2O)CdS CdCO3 CdSO4·nH2O CdC2O4 Cl Pb S Ba Cd Cr μ-XRD (2×2μm² ) μ-XRF (2×2μm²) Full-field XANES LC results at Cd L3-edge (0.7×0.7μm²) Sulfur valence map (2×2μm²) min max min max (a) (b) (d) (e) (c) sulphide sulphate Fig. 6 a Optical image of BF205 darkened sample prepared as 10-lm-thick thin section (bottom); b Sulphide and sulphate distribu- tion from sulphur valence maps in lXRF (step size: 1 9 1.2 lm2); c Elemental mapping results from fit of XRF map (step size: 2 9 2 lm2); d LSLC fitting results of full-field stack acquired at Cd LIII-edge on the thin section using CdS, CdCO3, CdCl2�nH2, and CdSO4�nH2O (pixel size: 0.7 9 0.7 lm2); e lXRD maps of sphalerite (ZnS), barite (Ba0.99Sr0.01(SO4)), anglesite (PbSO4), hydrocerussite (Pb3(CO3)2(OH)2), CdS, cadmium oxalate (CdC2O4), cadmium sulphate hydrated (CdSO4�nH2O), and otavite (CdCO3) (step size: 2 9 2 lm2) 2D X-ray and FTIR micro-analysis of the degradation of cadmium yellow pigment in paintings of… 977 123 lXRD at 21 keV (P06, step size 2 9 2 lm2), results pre- sented in Fig. 8b; FF-XANES at the S K-edge (ID21, pixel size 0.7 9 0.7 lm2), analysed as ROI integration (Online Resource (4)); and FF-XANES at the Cd LIII-edge (ID21, pixel size 0.7 9 0.7 lm2), analysed by LSLC fitting and presented in Fig. 8c. The Cd concentration map derived from FF-XANES at the Cd LIII-edge reveals that cadmium-containing particles show up as ‘‘hot spots’’ evenly dispersed throughout the zinc white base. The presence of CdCO3 grains is estab- lished by FF-XANES at the Cd LIII-edge and by lXRD (Fig. 8b, c). The XRD measurements show CdS as a dif- fuse halo in the vicinity of the larger CdCO3 grains. The Cd LIII FF-XANES results suggest that cadmium sulphide and sulphate form rings/shells surrounding the CdCO3 grains. This observation is supported by sulphide and sulphate maps acquired at S K-edge in FF-XANES mode (Online Resource (4)). However, the larger beam size and the relative insensitivity of XRD to poorly crystalline material did not allow confirming the sulphate/sulphide/carbonate morphology suggested by the FF-XANES measurements by diffraction imaging. Even though XRD cannot confirm the core/shell mor- phology of the CdCO3 particles surrounded by CdS or CdSO4, it clearly shows that in this sample, CdCO3 is co- localized with CdS instead of forming a surface degrada- tion layer. The co-localized core and shell morphology suggests that a large fraction of CdCO3 in this sample is residual Fig. 7 a Lemon cadmium paint cross section from Flower Piece (BF 205) showing cadmium yellow particles dispersed in a zinc white matrix (405 nm illumination), b Optical image of the lemon-hued cadmium paint prepared as thin section, 15 lm thick 0 0.2 0.4 0.6 0.8 1 1.2 1.4 3520 3570 3620 N or m al iz ed a bs or pt io n Energy (eV) 0 0.2 0.4 0.6 0.8 1 1.2 1.4 3520 3570 3620 N o rm al iz ed a b so rp tio n Energy (eV) Full-field XANES LC results at Cd-L3edge (0.7x0.7μm²) μ-XRD maps (2x2μm²)15 μm thin-sec�on CdSO4.nH2O CdS CdCO3 CdS CdCO3 25 μm min max 0 0.2 0.4 0.6 0.8 1 1.2 1.4 3520 3570 3620 N or m al iz ed a bs or pt io n Energy (eV) XANES spectrum of CdS rich area XANES spectrum of CdSO4·nH2O rich area XANES spectrum of CdCO3 rich area tlusertifSLCLtlusertifSLCLtlusertifSLCL (a) (b) (c) Fig. 8 lXRD and FF-XANES acquisition on the thin section presented in a; b lXRD maps of CdS and otavite (CdCO3) (step size: 2 9 2 lm2), and l-XRD Q-patterns acquired on CdS-rich grain; c CdS, CdCO3, and CdSO4�nH2O maps obtained by FF-XANES at the Cd L-III edge, and examples of LSLC fitting results of single pixel XANES spectra acquired in CdCO3, CdS, and CdSO4�nH2O-rich areas 978 E. Pouyet et al. 123 starting reagent. Conversion to CdS was incomplete; the large and poorly soluble CdCO3 particles were trapped during the precipitation reaction of CdS, coated by a thin layer of nanocrystalline CdS. This example might be the first evidence confirming the precipitation process hy- pothesized by Plahter and Topalova-Casadiego [10] for the role of cadmium carbonate identified in the yellow paints of the Munch Museum version of The Scream. To fully confirm their theory, a similar investigation should be performed on a cross section from a relatively non-de- graded region to ascertain whether or not this same pattern of precipitation (CdS coating CdCO3 starting material) is also present in yellow paints of the Munch Museum ver- sion of The Scream. Further analyses of this cross section, as well as analyses of replicate cadmium yellow pigments synthesized from CdCO3 following early twentieth century recipes, should also be considered at better resolutions. 4 Conclusion The composition of sub-millimetre fragments of cadmium yellow paints from The Joy of Life and Flower Piece as a function of depth using SR-lFTIR, lXRF, lXRD, and FF- XANES imaging confirmed that CdCO3, CdC2O4, and CdSO4�nH2O are degradation products rather than paint fillers, and, in the case of The Joy of Life, that these compounds are not residual synthesis reagents either. These colourless compounds are responsible for the observed fading of the cadmium yellow paint in The Joy of Life. Despite the high solubility that causes cadmium sul- phate to migrate through the paint layers in many of the samples studied, cadmium sulphate is also identified as a photo-degradation product. In the case of sample S5 (from The Joy of Life) and sample BF205-darkened (from Flower Piece), an oxidation of the original CdS pigment, probably induced by UV–visible irradiation and uncontrolled rela- tive humidity levels is the initiation point and basis of the observed fading and discoloration. Based on previous work [8, 11], the oxidation of CdS to CdSO4�nH2O at or just below the paint surface explains the formation of the CdSO4�nH2O compound which, as it is highly water sol- uble, may then diffuse through the paint layer. The surface enrichment of CdSO4�nH2O in the case of S5 suggests that while CdSO4�nH2O has the potential solubility to migrate through the paint layer, surface accumulation resulting from photo-degradation can also be observed. As suggested previously [11], this compound can initiate the subsequent stages of degradation by reacting with CO2 to form CdCO3. Acid hydrolysis of the organic binder or the varnish which leads to spalling and cracking of the paint layer allows for further photo-oxidation of the newly exposed CdS in the micro-cracks. The very high insolubility (Ksp of 1.0 9 10 -12 ) of CdCO3 explains its presence as the dominant end product of a series of degradation reactions. A tertiary degradation process involving the further breakdown of cadmium oxalate into cadmium carbonate cannot be excluded. The presence and the location of CdC2O4 are for the moment not fully understood, although TOF-SIMS data from this same painting (see this volume) show that long-chain fatty acids are depleted in the regions of CdS alteration, suggesting that the acid hydrolysis of the binding medium is a possible mechanism for the formation of oxalate anions. Alternatively, the presence of varnish, as observed in the case of the S5 sample, may have partially dissolved CdSO4 and freed Cd 2? ions, which in turn may then have precipitated out with C2O4 2- instead of SO4 2- , explaining the formation of a cadmium oxalate film [8]. The pale brownish appearance of the S111 sample ap- pears to be organic in nature as no evidence for the for- mation of a dark brown inorganic photo-degradation product was established. Soil and other fine particulate may lodge in the interstices of the crumbling paint surface, formed or enhanced during photo-oxidation, giving rise to the darkening observed [11]. Acid hydrolysis of the drying oil paint binder may also result in the formation of chro- mophores contributing to the darkened appearance. In ei- ther case, further analyses of organic degradation products and trace analyses of other particles embedded in the paint are required to fully understand the observed darkening. Both the intact lemon cadmium paint and photo-oxi- dized cadmium yellow paint from Flower Piece were ex- amined by UV and VIS photo-microscopy, XRD, and FF- XANES. Investigation of undarkened lemon cadmium paint from Flower Piece reveals that a large amount of cadmium carbonate is present; however, it is not accumu- lated at the paint surface but co-located with individual CdS yellow pigment particles suspended in a zinc white base. Co-location of CdS and CdCO3, possibly in a core/ shell morphology suggests that the CdS pigment used by Henri Matisse to paint this work may have contained a large fraction of unreacted starting reagents, leftovers of an incomplete synthesis. This finding provides the first phy- sical evidence supporting the theory of Plahter et al. that cadmium carbonate-rich cadmium yellow paints in the first decade of the twentieth century were likely prepared by the indirect wet process method, and are not invariably indicative of photo-alteration. Consequently, in the case of photo-oxidized paint, the high ratio of CdCO3 compounds present in yellow paint may be related to this starting reagent, whereas the white crust may combine both CdCO3 as a starting reagent and CdCO3 as second degradation product of photo-alteration, as it is found in the presence of CdSO4�nH2O and CdC2O4. In final conclusion, analytical methods with a high sensitivity for chemical speciation and ability to map the 2D X-ray and FTIR micro-analysis of the degradation of cadmium yellow pigment in paintings of… 979 123 distribution of various compounds at sub-micron resolution are essential for deciphering the synthesis and degradation pathways of pigments in hierarchically complex objects such as paintings. The access to both molecular and structural mapping capabilities by combining FF-XANES, lXRF, lXRD, and lFTIR in 2D is opening new possi- bilities for the rigorous study of degradation products and mechanisms in paint layers, and more generally for the micro-analysis of artworks for the benefit of these works’ long-term preservation. Acknowledgments This work is supported by the Andrew W. Mellon Foundation, the Barnes Foundation, the Lenfest Foundation, and the National Science Foundation DMR 0415838. The European Synchrotron Radiation Facility are acknowledged for providing beamtime. Barbara Buckley of the Barnes Foundation and Unn Plahter of the University of Oslo are thanked for their many helpful discussions. References 1. L. Zanella et al., The darkening of zinc yellow: XANES spe- ciation of chromium in artist’s paints after light and chemical exposures. J. Anal. At. Spectrom. 26(5), 1090–1097 (2011) 2. L. Monico et al., Degradation process of lead chromate in paintings by Vincent van Gogh studied by means of synchrotron X-ray spectromicroscopy and related methods. 2. Original paint layer samples. Anal. Chem. 83(4), 1224–1231 (2011) 3. L. Monico et al., Degradation process of lead chromate in paintings by Vincent van Gogh studied by means of synchrotron X-ray spectromicroscopy and related methods. 1. Artificially aged model samples. Anal. Chem. 83(4), 1214–1223 (2011) 4. L. Monico et al., Degradation process of lead chromate in paintings by Vincent van Gogh studied by means of spectromi- croscopic methods. 3. Synthesis, characterization, and detection of different crystal forms of the chrome yellow pigment. Anal. Chem. 85(2), 851–859 (2012) 5. L. Monico et al., Degradation process of lead chromate in paintings by Vincent van Gogh studied by means of spectromi- croscopic methods. 4. Artificial aging of model samples of co- precipitates of lead chromate and lead sulfate. Anal. Chem. 85(2), 860–867 (2013) 6. L. Monico et al., Degradation process of lead chromate in paintings by Vincent van Gogh studied by means of spectromi- croscopic methods. Part 5. Effects of nonoriginal surface coatings into the nature and distribution of chromium and sulfur species in chrome yellow paints. Anal. Chem. 86(21), 10804–10811 (2014) 7. G. Van der Snickt et al., Characterization of a degraded cadmium yellow (CdS) pigment in an oil painting by means of synchrotron radiation based X-ray techniques. Anal. Chem. 81(7), 2600–2610 (2009) 8. G. Van der Snickt et al., Combined use of synchrotron radiation based micro-X-ray fluorescence, micro-x-ray diffraction, micro- x-ray absorption near-edge, and micro-fourier transform infrared spectroscopies for revealing an alternative degradation pathway of the pigment cadmium yellow in a painting by Van Gogh. Anal. Chem. 84(23), 10221–10228 (2012) 9. B. Leone et al., The deterioration of cadmium sulphide yellow artists’ pigments, in Triennial meeting (14th), The Hague, 12–16 September 2005: preprints, 2005, ed. by James & James 10. U. Plahter, B. Topalova-Casadiego, The Scream by Edvard Munch: painting techniques and colouring materials, in Studying Old Master Paintings Technology and Practice, ed. by M. Spring. The National Gallery Technical Bulletin 30th Anniversary Con- ference Postprints (London: Archetype, 2011), pp. 244–252 11. J.L. Mass et al., The photodegradation of cadmium yellow paints in Henri Matisse’s Le Bonheur de vivre (1905–1906). Appl. Phys. A 111(1), 59–68 (2013) 12. C.S. Patterson et al., Synchrotron-based imaging FTIR spec- troscopy in the evaluation of painting cross-sections. e-Preserv. Sci. 10, 1–9 (2013) 13. J. Mass et al., SR-FTIR imaging of the altered cadmium sulfide yellow paints in Henri Matisse’s Le Bonheur de vivre (1905–6)— examination of visually distinct degradation regions. Analyst 138(20), 6032–6043 (2013) 14. E. Pouyet et al., Thin-sections of painting fragments: opportuni- ties for combined synchrotron-based micro-spectroscopic tech- niques. Herit. Sci. 3(1), 3 (2015) 15. E. Pouyet et al., Preparation of thin-sections of painting frag- ments: classical and innovative strategies. Anal. Chim. Acta 822, 51–59 (2014) 16. M. Salomé et al., The ID21 scanning X-ray microscope at ESRF. J. Phys. Conf. Ser. 425(18), 182004 (2013) 17. M. Cotte et al., Blackening of Pompeian Cinnabar paintings: X-ray microspectroscopy analysis. Anal. Chem. 78(21), 7484–7492 (2006) 18. V.A. Sole et al., A multiplatform code for the analysis of energy- dispersive X-ray fluorescence spectra. Spectrochim. Acta Part B At. Spectrosc 62(1), 63–68 (2007) 19. B. Fayard et al., The new ID21 XANES full-field end-station at ESRF. J. Phys. Conf. Ser. 425(19), 192001 (2013) 20. Y. Liu et al., TXM-Wizard: a program for advanced data col- lection and evaluation in full-field transmission X-ray mi- croscopy. J. Synchrotron Radiat. 19(2), 281–287 (2012) 21. U. Boesenberg et al., Mesoscale phase distribution in single particles of LiFePO4 following lithium deintercalation. Chem. Mater. 25(9), 1664–1672 (2013) 22. C.G. Schroer et al., Hard X-ray nanoprobe at beamline P06 at PETRA III. Nucl. Instrum. Methods Phys. Res. Sect. A 616(2–3), 93–97 (2010) 23. W. De Nolf, F. Vanmeert, K. Janssens, XRDUA: crystalline phase distribution maps by two-dimensional scanning and to- mographic (micro) X-ray powder diffraction. J. Appl. Crystallo- gr. 47(3), 1107–1117 (2014) 24. J. Kieffer, D. Karkoulis, PyFAI, a versatile library for azimuthal regrouping, in Journal of physics: conference series (IOP Pub- lishing, 2013) 25. J. Susini et al., Why infrared spectromicroscopy at the ESRF? ESRF Newsl. 41, 24–25 (2005) 26. D. Saviello et al., Synchrotron-based FTIR microspectroscopy for the mapping of photo-oxidation and additives in acrylonitrile– butadiene–styrene model samples and historical objects. Anal. Chim. Acta 843, 59–72 (2014) 980 E. Pouyet et al. 123 2D X-ray and FTIR micro-analysis of the degradation of cadmium yellow pigment in paintings of Henri Matisse Abstract Introduction Experimental Paintings and samples location Sample preparation and mounting Analytical methods and data processing Micro-X-ray fluorescence and chemical maps in scanning mode Full-field XANES Micro-X-ray diffraction Micro-FTIR analyses Results The Joy of Life (1905--1906): study of the degradation process S5 sample: faded yellow field beneath the central reclining figures S111 sample: darkened yellow foliage Flower Piece: degradation and synthesis processes study Darkened BF205 sample Undarkened BF205 sample Conclusion Acknowledgments References