Kandinsky’s fragile art: a multidisciplinary investigation of four early reverse glass paintings (1911–1914) by Wassily Kandinsky Steger et al. Herit Sci (2019) 7:27 https://doi.org/10.1186/s40494-019-0268-8 R E S E A R C H A R T I C L E Kandinsky’s fragile art: a multidisciplinary investigation of four early reverse glass paintings (1911–1914) by Wassily Kandinsky Simon Steger1* , Diana Oesterle2,3, Simone Bretz4, Lisa Frenzel5, Heike Stege6, Iris Winkelmeyer5, Oliver Hahn1,7 and Gisela Geiger2 Abstract This work highlights the rediscovery of the technique of reverse glass painting by the artists of the “Blaue Reiter” col- lective in the early 20th-century and focusses particularly on the role of Wassily Kandinsky (1866–1944). Kandinsky created more than 70 reverse paintings on glass and showed several of them in exhibitions together with paintings on canvas and cardboard, implying a coequal importance of these techniques. Four of his early (1911–1914) reverse glass paintings (Auferstehung, Allerheiligen II, Rudern, Apokalyptischer Reiter II) were selected for investigation and their iconography, painting techniques and painting materials were examined. Two paintings were executed on so-called cathedral glass, revealing a “hammered surface”, whereas Kandinsky used a corrugated glass panel for Rudern. A multi-analytical, non-invasive approach [X-ray fluorescence (XRF), diffuse reflectance infrared Fourier transform spec- troscopy (DRIFTS), VIS spectroscopy (VIS), Raman spectroscopy] was taken to identify the pigments and classify the binding media. The results reveal a broad palette of materials. Several pigments like lead white, zinc white, strontium yellow, Prussian blue, viridian, cadmium yellow, ultramarine blue, cinnabar and carbon black were found in most of the four paintings. The use of the rare synthetic organic pigments PR60 and PB52 is discussed. In two works of art, cadmium carbonate is associated with cadmium yellow. The identification of aluminium foil along with tin foils in Rudern indicates an early use of this material for reverse glass paintings. Keywords: Kandinsky, Reverse glass painting, Non-invasive analysis, Pigment identification, DRIFTS, Synthetic organic pigment, Cadmium carbonate, Aluminium foil © The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/ publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Introduction In the past, the technique of reverse glass paintings was often considered to be part of the stained glass genre; however, in contrast to stained glass, reverse glass paint- ings are viewed in reflected light and their creation does not involve a firing step. The paint layers are, compared with paintings on canvas, applied in reverse succession, starting with the front most layer and ending with the backing layer. Little art-historical research has been done on this technique and especially its importance for 20th- century art is poorly understood. The mass production of folkloristic reverse glass paintings in the 19th-century created a rather undervalued image of this technique in art history. A very limited number of scientific studies on reverse glass paintings have been published. Several publications deal with 14th to 19th-century reverse glass paintings including sampling of the objects [1–4]. Trans- portation and sampling of paintings is often restricted due to the fragility of the glass support, hence non-inva- sive methods have been carried out to collect information on the materials [5, 6]. Recently, non-invasive analyses of 20th-century reverse glass paintings has been published [7–9]. A publication on Chinese reverse glass paintings from the 19th-century discusses a possible influence of these paintings on European 20th-century reverse glass painting from the “Blaue Reiter” collective [10]. Open Access *Correspondence: simon.steger@bam.de 1 Federal Institute for Materials Research and Testing (BAM), Division 4.5, Unter den Eichen 44-46, Berlin, Germany Full list of author information is available at the end of the article http://orcid.org/0000-0002-3805-2784 http://creativecommons.org/licenses/by/4.0/ http://creativecommons.org/publicdomain/zero/1.0/ http://creativecommons.org/publicdomain/zero/1.0/ http://crossmark.crossref.org/dialog/?doi=10.1186/s40494-019-0268-8&domain=pdf Page 2 of 17Steger et al. Herit Sci (2019) 7:27 This paper will highlight the rediscovery of the tech- nique of reverse glass painting by the artists of the “Blaue Reiter” collective in the early 20th-century and in par- ticular will shed light on the role of Wassily Kandinsky (1866–1944). Four of his early reverse glass paintings (Auferstehung, 1911, Allerheiligen II 1911, Rudern, c. 1912, Apokalyptischer Reiter II, 1914) were selected and their iconography, painting technique and painting mate- rials were examined. The main objectives of this study are: (1) to work out the starting date of modern reverse glass paintings (2) to discuss the importance of this tech- nique for Kandinsky’s oeuvre, (3) to examine the paint- ing technique, (4) to highlight the influence of folkloristic 19th-century reverse glass paintings from Bavaria, and (5) to conduct non-invasive, in  situ spectroscopic analy- ses to identify the colourants and classify the binders. Art historical context Folkloristic art was an important source of inspiration for Wassily Kandinsky, Gabriele Münter (1877–1962) and their colleagues from the “Blaue Reiter” collective. In September/October 1908, Kandinsky and Münter together with their friends Marianne von Werefkin (1860–1938) and Alexej Jawlensky (1864–1941), spent their first summer stay in Murnau near Munich. The four artists came back next year in the summer of 1909, enjoy- ing the natural surroundings and the quiet life in this small village. Until 1914, Kandinsky and Münter lived not only in Munich, but also spent several months every year in Murnau, where Münter bought a small villa. Murnau, Seehausen and Uffing am Staffelsee were the main sites of folkloristic reverse glass painting tradition, a technique that started to disappear at the turn of the century. Generally, Münter states that this technique was new to their entire group and that they first came in touch with it in Murnau. Moreover, it was Jawlensky who introduced them to the local master brewer Johann Krötz and to the artist Heinrich Rambold (1872–1955) [11, 12]. The private collection of Krötz included numerous traditional Bavarian reverse glass paintings, which were important sources of inspiration for the artists. Kandin- sky, Münter and Jawlensky started to build up their own collection of European and non-European reverse glass paintings, which they used to decorate their flats [13]. Heinrich Rambold was the last artist in the region who created folkloristic reverse glass paintings as souvenirs for tourists. He focussed primarily on 19th-century tem- plates of religious motifs including saints, votive pictures and titular saints in various sizes, but he also created his own designs in an expressive colour style [14]. Münter was the first one of the four to start with this technique by learning from Rambold and copying some of his paintings [11, 12]. She also inspired Kandinsky to try this technique, and they spent evenings creating reverse paintings on glass with their own style and motifs. Tradi- tional 19th-century reverse glass paintings reveal charac- teristic features like two-dimensional areas of unbroken colour, simplification of the forms, reduction of the col- ouration and dominance of the line. The combination of the graphic contour and pictorial elements was a con- temporary achievement of non-academic pictorial art. The graphic enclosing of coloured areas also played an important role in the works of Paul Gauguin and of the members of the “Nabis” and “Fauves” groups. Jawlensky and Werefkin in particular, studied these French paint- ers and discussed their observations with Kandinsky and Münter in Murnau. The four artists became enthusiastic about the intense and luminous paintings by Georges Braque, André Derain and Henri Matisse [15]. The same luminosity of colours can be found in the traditional Bavarian reverse glass paintings. Hence, for the “Blaue Reiter” artists, the “primitive” originality of the folklor- istic reverse glass paintings corresponds to the modern pictorial principles of the avant-garde movements. To achieve a look of simplicity, nativeness and originality, the “Fauves” and “Brücke” artists searched for oriental and exotic artworks; however, the “Blauer Reiter” col- lective found these desired features in the folkloristic art of the surrounding rural area. There were no academic rules and guidelines for the technique of painting on the reverse side of a glass panel, so it became an important source of inspiration, that supported the striving for clar- ity of expression and simplification of composition. This led Münter and Kandinsky to consciously adapt their art to the pictorial principles of the folkloristic art. Wassily Kandinsky—reverse glass painter Kandinsky must have thought highly about this tech- nique, as he created more than 70 reverse glass paint- ings in his career and allowed three of them to be shown in the first “Blaue Reiter” exhibition at the Moderne Galerie Thannhauser in Munich in 1911/12 [11]. His fascination and intense focus on the technical and stylis- tic opportunities of this technique are especially visible during his stays in Murnau and when he was finalizing his search for a fundamental, novel pictorial conception in 1911. Kandinsky also states in a letter to Franz Marc that he hardly knows any work that is more delightful than creating reverse glass paintings—unfortunately, these paintings are so fragile [16]. He continued with this technique during his stays in Moscow (1915–1921) and Paris (1933–1944). Besides Gabriele Münter and Hein- rich Campendonk (1889–1957), Kandinsky created the largest number of reverse glass paintings among the art- ists associated with the “Blaue Reiter” collective, which Page 3 of 17Steger et al. Herit Sci (2019) 7:27 further indicates that this genre plays a substantial role in his oeuvre. His first glass painting dates from 1909, when, in con- trast to Münter, he already starts with his own sketches (e.g. Mit Gelbem Pferd, 1909 and Abendmahl, 1909/10). After these first attempts, Kandinsky started to use typi- cal stylistic elements like closed contours with dark lines. He painted motifs like saints, apocalyptic scenarios or All Saints’ Day, which are in close relation to his early works (Fig.  1). Several pictorial topics and motifs were first painted on glass, and later executed on canvas or paper or as woodcut. For example, Kandinsky created two anal- ogies of the reverse glass painting Allerheiligen I (1911) as a coloured woodcut and an oil painting on cardboard. Religious topics and motifs are especially present in his artworks from 1911 until 1914, where they express a hid- den flow of encrypted symbols [17]. The reproduction of glass paintings with other techniques led to a loss of the aesthetic properties of the glass substrate. The interplay of the material iconography and the painted objects for reverse glass paintings has hardly been described in art historical research. However, Kandinsky used structured glass panels, like cathedral glass or ornamental glass for several works, that reveal an experimental character [18]. Three of them (Rudern, Mit Kneifer, Mit Reiter, all dated 1912) are listed in the catalogue raisonné and are explic- itly described as paintings “on corrugated glass” [19]. Wackernagel states that Kandinsky wanted to distance himself from the folkloristic tradition of reverse glass painting by using structured or corrugated glass panels [20]. Kandinsky integrated the corrugation of the panel in Rudern (Fig.  2) as an independent stimulus and compli- cated the exploration of the subjects by integrating metal foils as collage-like objects [8]. The left half of the paint- ing shows a boat that, as indicated by thick black brush strokes, is lifted by a big wave. The dimly visible people hold six oars that extend radially in pairs out of the boat. The entire glass panel is filled with colourful areas that do not depict specific objects. The combined use of struc- tured glass and metal foils hinders a clear identification of the graphic elements, which indicates that the veiling of the motifs was done solely externally with technical media [8]. Hence, Kandinsky’s paintings on structured glass panels seem to avoid a straightforward recognizabil- ity of the motifs. This can also be observed in his abstract canvas paintings, which he created in parallel from 1911 onwards. The corrugated glass structure of Rudern is not only an attractive material property, but also directly implies the representation of water and waves. The com- bined observation of the reflections from the metal foils and the glass surface can be interpreted as light reflec- tions on the water. In 1910, Kandinsky created a watercol- our painting that shows several similarities with Rudern: six black lines (the oars) with double arcs above, the con- tinuous red line painted from the right to the left, and Fig. 1 Photographs of the framed paintings: a Auferstehung, 1911 (GMS 112), b Allerheiligen II, 1911 (GMS 122), c Rudern, c. 1912 (GMS 108), d Apokalyptischer Reiter II, 1914 (GMS 106); ©Städtische Galerie im Lenbachhaus und Kunstbau München Page 4 of 17Steger et al. Herit Sci (2019) 7:27 the red, gently contoured area in the lower right part of the painting. The side-inverted illustration of the oblique oar lines indicates that Kandinsky used this watercolour painting as a draft for Rudern and for the canvas paint- ing Improvisation 26 (Rudern) (Fig.  2d). The objects in the watercolour painting can still be read clearly, whereas the other two paintings show an advanced mystification of motifs. Kandinsky created Rudern and Improvisa- tion 26 (Rudern) in a close context in 1912. The painting Improvisation 26 (Rudern) yields details of the drawing, the watercolour painting and the reverse glass painting; however, the illustration of the content becomes less clear as the motifs and forms further dissolve and as the colours gain greater importance, in direct contrast to the disappearing forms [8]. The painting appears more dif- ferentiated and the border between clear recognizability and mystification of the motifs remains fluent. The state- ments of Zweite [18] imply that the reverse glass painting Rudern was also a preliminary study for Improvisation 26 (Rudern); but unlike his drafts, drawings and sketches, Kandinsky showed reverse glass paintings together with paintings on canvas in his exhibitions (e.g. gallery “Der Sturm”) implying a coequal importance of the two techniques. Description of the paintings Four reverse glass paintings (Auferstehung, Allerheiligen II, Rudern, Apokalyptische Reiter II) by Wassily Kandin- sky (Fig.  1), housed in the Städtischen Galerie im Len- bachhaus und Kunstbau in Munich, were examined during the research project “Hinterglasmalerei als Tech- nik der Klassischen Moderne 1905–1955” (2015–2019). They were created between 1911 and 1914 in Murnau and stored in Kandinsky and Münter’s villa after Kan- dinsky had to leave Germany in 1914. Kandinsky and Münter used Allerheiligen II and other reverse glass paintings to decorate their dining room as shown in a photograph taken by Münter in 1913 (Fig. 3). Kandinsky’s reverse glass paintings remained in Murnau until 1957, when Münter donated them among other artworks to Fig. 2 Photographs of Rudern, c. 1912 (GMS 108), Städtische Galerie im Lenbachhaus und Kunstbau München (a–c): a unframed front side, b reverse side in transmitted light, c reverse side, along with a photograph of the canvas painting Improvisation 26 (Rudern), 1912 (GMS 66) (d). ©Städtische Galerie im Lenbachhaus und Kunstbau München (a, d); ©Simone Bretz (b, c) Page 5 of 17Steger et al. Herit Sci (2019) 7:27 the Städtischen Galerie im Lenbachhaus und Kunstbau in Munich. One of the four paintings (Allerheiligen II, 1911) is signed with the monogram “K”. Münter wrote with a pen- cil “Kandinsky/Gemalt im August 1911/in Murnau“on the backing cardboard of Auferstehung (1911) and “Kan- dinsky (Glasbild) Rudern” on the backing layer of Rudern (c. 1912). Apokalyptischer Reiter II (Fig.  4) dates from 1914 and was the last reverse glass painting that Kand- insky created in Murnau. Moreover, as already discussed in the introduction to Rudern, Kandinsky created several paintings with the same content on different supports, like canvas, paper, cardboard and glass. In 1911, Kandin- sky executed Allerheiligen II as a watercolour, canvas and as a reverse glass painting (Fig. 5). A glass fragment of the preliminary study for Apokalyptischer Reiter II from 1914 has been preserved, showing roughly painted lines on the upper part of the painting. The analogue watercolour painting dates from 6. July 1914 and shows the same cen- tral motifs in a mirror image, but the oval cartouche and the four illustrations of animals on the edges are missing. Painting technique The major artistic challenge of painting on the reverse side of a glass panel may be the reverse succession of the paint layers. Kandinsky had to apply the frontmost layer (i.e. the most visible layer) first and the background layer last. He started his paintings with black contours and lines, followed by the detailed painting of internal areas and the application of the background. The four paintings reveal multi-layered paint systems (2–3 layers) showing variably thick paint layers. Kan- dinsky used a variety of brushes and stippled the paint or swiftly coloured large areas. He painted wet-in-wet or applied the paint as dots side by side or one upon another, indicating an optical mixture of colours. The typical structure of an inhomogeneous application of col- our indicates the use of a bristle brush in several areas of Rudern (Fig. 2b, c), whereas the white area of the painting seems to be stippled. Compared with Rudern, the appli- cation of paint layers in Auferstehung (Fig.  6) was more Fig. 3 The dining room of the Murnau house; the wall is decorated with reverse glass paintings by Kandinsky. Photo: Gabriele Münter (1913), Gabriele Münter- und Johannes Eichner-Stiftung, Munich (Inv.-Nr. 2235)/©VG Bild-Kunst, Bonn 2019 Fig. 4 Photographs of Apokalyptischer Reiter II, 1914 (GMS 106), Städtische Galerie im Lenbachhaus und Kunstbau München: a unframed front side, b reverse side in transmitted light, c reverse side. ©Städtische Galerie im Lenbachhaus und Kunstbau Munich (a); ©Simone Bretz (b, c) Page 6 of 17Steger et al. Herit Sci (2019) 7:27 Fig. 5 Photographs of Allerheiligen II, 1911 (GMS 122), Städtische Galerie im Lenbachhaus und Kunstbau München (a–c); a unframed front side, b reverse side in transmitted light, c reverse side, along with a photograph of the watercolour painting Allerheiligen II, 1911 (GMS 616), Städtische Galerie im Lenbachhaus und Kunstbau München (d). ©Städtische Galerie im Lenbachhaus und Kunstbau München (a, d); ©Simone Bretz (b, c) Fig. 6 Photographs of Auferstehung, 1911 (GMS 112), Städtische Galerie im Lenbachhaus und Kunstbau München; a unframed front side, b reverse side in transmitted light, c reverse side. ©Städtische Galerie im Lenbachhaus und Kunstbau München (a); ©Simone Bretz (b, c) Page 7 of 17Steger et al. Herit Sci (2019) 7:27 intense and homogeneous. One needs to consider that these paintings make their impressions solely from the front side, hence the gloss, brush structure, and pastosity of the reverse side were not of artistic importance. Kandinsky often used cardboards painted black as supports for his reverse glass paintings as they further intensify the depth of colours [21]. Such black-painted cardboard was used in Auferstehung, Allerheiligen II and Rudern. Black-painted supports (paper or wooden veneer) are also known from traditional Bavarian and Swiss reverse glass paintings from the 18th and 19th- century respectively [22]. 19th-century Bavarian reverse glass paintings may also have been Kandinsky’s source of inspiration to decorate the frames of his paintings (Fig.  1). The folkloristic paintings from Oberammer- gau, in particular, have black frames painted with flower motifs. Kandinsky used a commercial wooden frame for Auferstehung, which he decorated with green, blue and violet paints (Fig. 1). He chose industrially gilded frames for the other three paintings, which he partially over- painted with lustre paint (Fig. 1). Glass technique For paintings on canvas, the final application of varnish creates a certain depth of colours and protects the upper- most paint layers. The glass panel of reverse glass paint- ings, however, itself behaves as a varnish, which gives rise to the painting’s great luminosity and protects the paint layers from the front. Kandinsky used glass panels of variable sizes in landscape and portrait format. The glass panels of the studied paintings show the following dimen- sions: 21.9 × 11.4 × 0.28 (Auferstehung), 31.1 × 47.8 × 0.41 (Allerheiligen II), 21.5 × 25.7 × 0.44 (Rudern) and 30.3 × 21.2 × 0.25 cm (Apokalyptischer Reiter II). The sur- face structure of the glass panel plays an important role for the final visual appearance of a reverse painting on glass. Generally, artists prefer flat glass panels, because light reflections of an uneven surface distract the observ- er’s attention from the painted motifs. Kandinsky created most of his reverse glass paintings on flat panels, but ten paintings on structured glass are known from his Mur- nau period between 1911 and 1913. He used commercial products, that were sold for different applications (e.g. windows). Generally, there were two different types of structured glass panels: ornamental glass (3–4 mm thick) that shows a corrugated surface and the thinner cathe- dral glass (2–3  mm) that reveals a “hammered” surface structure (Fig. 7a). Both types were produced in the roll- ing process, which was invented in 1847 [23]. Rolled plate glass was made by thinning molten glass between two rollers and then placing it on the casting table. The sur- face structure of ornamental glass was first engraved into the table in which the molten glass was then cast. This procedure was expensive, so the process was adapted in 1884; now the structures were engraved in an extra pair of rollers, which pressed into the already thinned glass [23]. The rolling process for cathedral glass included a higher temperature and a faster rolling procedure, and the thinned glass was drawn out on a water-cooled cast- ing table. The quick cooling led to a quick contraction of the glass surface, yielding a “hammered” appearance of the surface. The other, rolled side of the glass panel was not affected by this procedure and stayed rather smooth. Kandinsky did not always choose the same side to paint onto. In Auferstehung, the structured surface is the front side (Fig.  7a), whereas for Allerheiligen II, the smoother surface is the front side (i.e. he painted on the structured side). In Rudern, Kandinsky painted on the corrugated side of the ornamental glass (Fig. 2c). Fig. 7 Photographs of a Auferstehung (front side) and b Allerheiligen II (reverse side) in glancing light, showing the structured surface of the cathedral glass. ©Simone Bretz Page 8 of 17Steger et al. Herit Sci (2019) 7:27 Methods of the material analyses X‑ray fluorescence (XRF) The handheld spectrometer Tracer III-SD (Bruker AXS Microanalysis GmbH) was fixed on a tripod perpendic- ularly to the sample (sample to spectrometer distance ~ 1  mm, spot size ~ 10  mm). The instrument consists of an electrothermally cooled Xflash SDD detector (energy resolution = 150  eV for Mn Kα radiation) and an X-ray tube equipped with a rhodium anode. The excitation parameters were set to 40 kV, 15 μA and 20 s (acquisition time). VIS spectroscopy (VIS) The spectrophotometer SPM 100 (Gretag-Imaging AG, Regensdorf, Switzerland), measures the reflection of vis- ible light (from 380 to 730 nm) with a spectral resolution of 10 nm. The surface of the sample is illuminated for half a second, using a 2 W bulb (spot size 4 mm). The reflec- tance spectra are normalized, and the first derivative is plotted for better peak comparison. Diffuse reflectance infrared fourier transform spectroscopy (DRIFTS) Diffuse reflectance spectra were recorded, using a 4100 Exoscan FTIR spectrometer (Agilent) fixed on a tripod perpendicularly to the sample (sample to spectrometer distance ~ 1  mm, spot size ~ 10  mm). In this configura- tion, the reflected signal is collected inside an imaginary cone of 45° around the emission beam, which supports the detection of diffuse reflected light. The instrument is equipped with a ZnSe beam splitter, a Michelson inter- ferometer and a thermoelectrically cooled dTGS detec- tor. For every spectrum, 500 scans were recorded in the mid IR range (650–4000  cm−1) with a spectral resolu- tion of 4  cm−1. A gold reference cap was used for back- ground calibration. The spectrum intensity was defined as pseudo-absorbance A′ = log (1/R). The Thermo Sci- entific™ OMNIC™ Specta software (Version 9.7, Madi- son, WI, USA.) was used for comparison with internal databases. Raman spectroscopy Raman measurements were performed with an i-Raman®Plus spectrometer (B&W Tek Inc.) equipped with a handheld fibre optic probe and a CCD detec- tor. The probe was connected to a microscope head (BAC151B, B&W Tek Inc.) with an Olympus 50× objec- tive and was fixed on a motorized xyz stage of a tripod. After focusing through the microscope, the maximum laser power of the 785 nm diode laser was ~ 160 mW. The recorded spectra range from 100 to 3300  cm−1 (spectral resolution 4  cm−1) with varying acquisition parameters of 1–200 s and 2–10% laser power. The holographic grat- ing was fixed at 1200 lines/mm. The Thermo Scientific™ OMNIC™ Specta software (Version 9.7, Madison, WI, USA.) was used for baseline correction and for compari- son with internal databases. Analytical results The reverse side of Auferstehung (1911) (Fig.  6c) is dominated by yellow, green and bluish colours. These areas partly cover the orange and red areas, that are vis- ible from the front side. Zinc white (ZnO) and smaller amounts of lead white (2PbCO3·Pb(OH)2) were present in all coloured areas, whereas the white uppermost layer consists only of zinc white. A small amount of Gyp- sum (CaSO4·2H2O) was detected and other fillers were absent. Prussian blue (Fe4[Fe(CN)6]3) was used solely for the blue and dark blue painted areas. The dominant green hues of the reverse side comprise mainly mixtures of Prussian blue with various yellows. Strontium, Naples and chrome yellow were admixed to achieve greenish yel- low to bluish green hues. Strontium yellow (SrCrO4) and chrome yellow (PbCrO4) were also individually used for the yellow and dark yellow areas. Both pigments yield fingerprinting Raman spectra. Strontium yellow shows typical CrO4 bending modes at 338, 349, 372, 400 and 431  cm−1 and intense CrO4 stretching bands at 865 (νs), 892 and 914  cm−1 (νas) (Fig.  8e) [24, 25]. The intense Raman band at 838  cm−1 (CrO4 stretching) along with several weaker ones at 344, 358, 375 and 399 cm−1 (CrO4 bending) (Fig. 8f ) properly identify chrome yellow in the greenish yellow area (Fig.  6c, spot 2) [26]. Characteristic bands at 278, 535 (FeC stretching), 2094 and 2154  cm−1 (CN stretching) prove the presence of Prussian blue [27, 28]. An additional weak spectral feature at 251 cm−1 orig- inates from an underlying layer and can be ascribed to cinnabar (HgS). Intense Sb peaks in the XRF spectrum of the bluish green areas (e.g. Fig. 6c, spot 1) hint at Naples yellow (Pb2Sb2O7). Due to extensive fluorescence, it was not possible to record a usable Raman spectrum from these areas. However, Naples yellow yields one major band in the mid IR range, so it could be identified using DRIFTS (Fig.  8a). The prominent inverted and broad band (reststrahlen band) shows a minimum at 675 cm−1, whereas the other IR active modes are out of range for this device [29, 30]. The DRIFT spectrum also reveals an intense band at 2098 cm−1 originating from Prussian blue (CN asymmetric stretching) [31]. Despite this variety of mixed greens, Kandinsky also used a green pigment. The XRF spectrum of spot 4 (Fig. 6c) is dominated by Cr and Sr signals, suggesting the presence of viridian (hydrated chromium oxide) and strontium yellow. Both pigments were detected in the DRIFT spectrum (Fig. 8b). Viridian yields a typical weak band at 1064 cm−1 and two spectral Page 9 of 17Steger et al. Herit Sci (2019) 7:27 features at 1255 and 1287  cm−1 originating from chro- mium borate, a by-product from the production pro- cess of the pigment [32]. Strontium yellow shows intense bands at 854, 876 and 933  cm−1. Additionally, several lead white bands also appear in the spectrum: 1046 (CO3 symmetric stretching), 3538  cm−1 (OH stretching) and an inverted band with a minimum at 1392  cm−1 (CO3 antisymmetric stretching) [7]. The presence of cinnabar and brown earth pigments in the red–orange and brown areas was confirmed by XRF respectively. All measured areas yield typical bands for drying oil. An example is given in Fig.  8a, showing bands at 1468 (CH2 scissor- ing) and 1754 (C=O stretching) cm−1 and inverted spec- tral features with minima at 2848 and 2915  cm−1 (CH2 stretching) [7, 8, 33]. The palette of Allerheiligen II (1911) (Fig. 5a–c) is sum- marized in Table  1. Zinc white and, in minor amounts, lead white were found in all colours. Gypsum was iden- tified only in the bluish green areas, whereas other fill- ers (e.g. barium sulfate, chalk) are again absent. The thin contour lines and black areas were painted with carbon black. Prussian blue and ultramarine blue (Na7Al6Si6O24S3) were individually used for the blue areas, and both were mixed with cinnabar to create violet and pinkish red hues (Fig.  5c, spots 2 & 5). The Raman spectrum of the bluish violet area (Fig.  8g) shows an intense ultramarine blue band at 546 cm−1 (stretching of the S3 − radical ion) and two bands at 252 and 342  cm−1 originating from HgS stretching vibrations of cinnabar [34–36]. Raman measurements of a pinkish violet area (Fig.  8h) also yielded typical cinnabar bands at 249 and 338  cm−1 along with bands of Prussian blue at 338, 534, 2092 and 2153  cm−1. Kandinsky mixed Prussian blue in various ratios with strontium yellow and minor amounts of cadmium yellow to obtain bluish green and green hues (Fig.  5c, spot 3). Both major pigments can be clearly identified by their DRIFT spectra (Fig.  8c), as Prussian blue yields an intense reststrahlen band with a minimum at 2084 cm−1 (CN asymmetric stretching) and strontium yellow reveals its strong IR bands at 856, 878 and 931  cm−1 [31]. The yellow hues were created with a mixture of strontium yellow with minor amounts of cad- mium yellow. The brownish to dark yellow areas (Fig. 5c, spot 4) yielded intense Cd, S, Zn and Pb signals in the XRF spectrum, suggesting the presence of zinc white, lead white and cadmium yellow (CdS). The DRIFT spec- trum (Fig.  9c) of the same area shows a sharp inverted band with a minimum at 855  cm−1 (CO3 out of plane bending) and a weak combination band at 2472  cm−1, which can be assigned to cadmium carbonate (CdCO3) [37, 38]. A detailed interpretation is given in the section below (cf. “Analytical results” section). Drying oil was classified in most DRIFT spectra as binding media, yield- ing bands at 1470, 1752  cm−1 and two spectral features in the 2840–2930 cm−1 range (Fig. 8c) [7]. Only the spec- trum of the orange area gives a hint at a different binder (Fig. 8d), showing a broad band at 1560 cm−1, which can be ascribed to the amide II of a proteinaceous compound Fig. 8 DRIFTS (a–d) and Raman spectra (e–h) of Auferstehung (a, b; e, f ) and Allerheiligen II (c, d; g, h) along with detailed photos of the measured areas; offset was applied for better comparison. DRIFT spectra originate from the following spots: a) the green area of Fig. 4c (spot 1), b) the green area of Fig. 4c (spot 4); c) the bluish green area of Fig. 5c (spot 3), and d) the orange area of Fig. 5c (spot 1); RSB = reststrahlen band. Raman spectra originate from the following spots: e) the yellow area of Fig. 4c (spot 3), f ) the greenish yellow area of Fig. 4c (spot 2); g) the violet area of Fig. 5c (spot 2), and h) the pinkish red area of Fig. 5c (spot 5). × = measured spots Page 10 of 17Steger et al. Herit Sci (2019) 7:27 (e.g. egg white, casein), along with intense signals (1740, 2858, 2935 cm−1) of lipids (oil, egg yolk) [7]. The spectral feature at 1464 cm−1 may be a combination of the amide III band from the protein and the CH2 scissoring vibra- tion of the fatty compound. Additionally, a weak band of an oxalate (1322 cm−1) [39] and one of an unknown sub- stance (3398 cm−1) are visible. The reverse side of Rudern (c. 1912) is partly covered with aluminium and tin foils (glossy silvery appearance and whitish matt, respectively) (Fig. 2c) showing differ- ent surface structures (Fig.  10b). Zinc white dominates among the white pigments, while lead white appears in small amounts. Gypsum was found only in the brown painted area, associated with iron oxides. The use of barium sulfate (BaSO4) is restricted mainly to some greenish areas that contain synthetic organic pigments (SOP). Kandinsky used pure carbon black for the thick contour lines and bone black (C + Ca5(PO4)3(OH)) for the black painted area. Ultramarine blue was identified in the blue areas by its intense Raman band at 546 cm−1 (Fig.  11f ). The light greenish blue hues (Fig.  2c, spot 1) originate from anthraquinone lake, PB52, which is partly mixed with Prussian blue. PB52 yields a char- acteristic Raman spectrum (Fig.  11e) with an intense doublet at 1303 and 1358  cm−1 and weaker bands at 483, 712 and 1251  cm−1 [40]. Kandinsky used four dif- ferent mixtures for the green and bluish green areas. Table 1 Comparison of the analytical results of the four paintings SOP synthetic organic pigment a XRF b Raman c DRIFTS d VIS e colour not in the painting Auferstehung (Jüngstes Gericht), 1911 (23 spots tested) Allerheiligen II, 1911 (34 spots tested) Rudern, c. 1912 (27 spots tested) Apokalyptischer Reiter II, 1914 (32 spots tested) White Zinc whitea, lead whitea,c, gypsumc Zinc whitea, lead whitea,c, gypsumc Zinc whitea, lead whitea, gypsumc, barium sulfatea,c Zinc whitea, lead whitea, barium sulfatea,b Black –e Carbon blackb Carbon blackb, bone blacka Carbon black? Red Cinnabara,b – PR83c Cinnabara,d, PR60b Orange – Cinnabar + cadmium yellowa,d – – Yellow Chrome yellowa,b, strontium yellowa,b,c, Naples yellowa,c Strontium yellow + cadmium yellowa,b,c,d, cadmium yellowa,c Strontium yellowa,b,c,d, cadmium yellowa,d, yellow SOPc Cadmium yellowa,d, cadmium yel- low + strontium yellowa,c,d Green Viridian + strontium yellowa,c, Prussian blue + chrome yellowa.c, Prussian blue + stron- tium yellowc, Prussian blue + Naples yellowa,b,c Prussian blue + strontium yel- low + cadmium yellowa,c Viridian + strontium yellowa,d, Prussian blue + strontium yellow + cadmium yellowa,b,c, Prussian blue + yellow SOPc, PB52 + yellow SOPc Viridian + strontium yellowa,c,d, emerald green + strontium yellowa,c, viridian + emerald green + strontium yellowa,c Blue Prussian bluea,b,c Prussian bluea,bc,d, ultramarine bluea,b,c,d Prussian blueb,c, ultramarine blueb,c,PB52b,c Prussian bluea,b,c,d, ultramarine bluea,b,d, cobalt bluea Violet Prussian blue + cinnabara,b,c Prussian blue + cinnabara,b,c, ultramarine blue + cinnabara,b,c Ultramarine blue + PR83a,c Prussian blue + PR60a,b,c, cobalt blue + PR60a,b Brown Brown earthsa – Brown earthsa,c,d – Metal – – Aluminium foila, tin foila – Binder Drying oilc Drying oilc, proteinaceous binder + lipidsc Drying oilc Drying oilc Fig. 9 DRIFT spectra of yellow areas of a) Apokalyptischer Reiter II (Fig. 6c, spot 4), b) Rudern (Fig. 3c, spot 2) and c) Allerheiligen II (Fig. 5c, spot 4) along with a cadmium carbonate powder spectrum (d). RSB = reststrahlen band. × = measured spots Page 11 of 17Steger et al. Herit Sci (2019) 7:27 Both viridian and Prussian blue were mixed with stron- tium yellow to obtain bluish green hues. Mixtures of an unknown yellow SOP with either Prussian blue or PB52 were found in the light green areas (e.g. Fig.  2c, spot 4). The Raman spectra of these areas were covered by extensive fluorescence, whereas the DRIFT spectrum reveals several bands at 758, 876, 996, 1160, 1235, 1269, 1341 and 1466 cm−1 that can be ascribed to the yellow SOP (Fig. 11b). Please note that the ~ 1000–1150 cm−1 range is covered by the structured reststrahlen band (minimum at 1064  cm−1; SO4 asymmetric stretching) of barium sulfate [41]. Due to this extensive band over- lapping, an exact identification of the SOP was not suc- cessful. Additionally, the strong IR absorption band of Prussian blue is visible at 2098 cm−1. Strontium yellow and cadmium yellow were identified in the yellow and dark yellow areas respectively. The DRIFT spectrum of the cadmium yellow, as in Allerheiligen II, shows characteristic signals of cadmium carbonate (sharp, inverted band with a minimum at 855  cm−1) (Fig.  9b). An alizarin-based red was found in the red areas, yield- ing typical bands at 846, 1190, 1276, 1298, 1351, 1366 and 1472  cm−1 in the DRIFT spectrum (Fig.  11a) [7, 9]. It needs to be emphasised that a clear discrimina- tion between PR83 (synthetic alizarin) and natural madder lake is not possible with this method, but we would rather assume PR83 in the 1910s than its natu- ral analogue. A mixture of PR83 and ultramarine blue Fig. 10 Detailed photographs of the metal foils of Rudern: a front side, b reverse side. ©Simone Bretz Fig. 11 DRIFTS (a-d) and Raman spectra (e–h) of Rudern (a, b; e, f ) and Apokalyptischer Reiter II (c, d; g, h) along with detailed photos of the measured areas; offset was applied for better comparison. DRIFT spectra originate from the following spots: a) the red area of Fig. 3c (spot 5), b) the green area of Fig. 3c (spot 4), c) the bluish green area of Fig. 6c (spot 3) and d) the green area of Fig. 6c (spot 5); RSB = reststrahlen band. Raman spectra originate from the following spots: e) the greenish blue area of Fig. 3c (spot 1), f ) the blue area of Fig. 3c (spot 3), g) the pinkish red area of Fig. 6c (spot 2) and h) the blue area of Fig. 6c (spot 1). Reference spectra (grey lines): cf. [40]; × = measured spots Page 12 of 17Steger et al. Herit Sci (2019) 7:27 was identified in the bluish violet area. Drying oil was classified as the binding media; however, a correct clas- sification was sometimes hampered by extensive band overlapping in the DRIFT spectra. An unaffected spec- trum is given in Fig. 9b, showing intense bands at 1466 and 1753 cm−1 and inverted bands with minima at 2849 and 2916 cm−1, which classify drying oil properly [7, 8, 33]. The painting Apokalyptischer Reiter II (1914) reveals several unpainted areas (i.e. the glass substrate is visi- ble) and thin paint layers (Fig. 4a–c). XRF measurement of the glass shows high amounts of calcium, potassium and silicon and weak peaks of manganese, iron, stron- tium and lead. Zinc white and small amounts of lead white were reported from every spot. Barium sulfate was found only in the red areas, associated with a SOP. Kandinsky used ultramarine blue and Prussian blue (Raman bands at 275, 543, 2094, 2154; Fig. 11h) for the blue areas. The presence of significant Co peaks in the XRF spectra of the greenish blue (body of the horse) and lilac areas (the oval form around the inner scene) (Fig.  4c) suggest the presence of cobalt blue (Co-metal oxides, variable composition). Viridian was found in most green areas, where it is mixed with strontium yel- low. Viridian yields typical IR bands at 1062, 1259 and 1291  cm−1 (Fig.  11d), of which the latter two indicate chromium borate, a by-product of the production pro- cess [32]. Strontium yellow shows characteristic bands at 858, 878 and 935 cm−1 in the DRIFT spectrum. XRF measurements of the bluish green areas (Fig.  4, spot 3) yield particularly high values for copper and arsenic, suggesting the presence of a Cu–As green. Emerald green (3Cu(AsO2)2·Cu(CH3COO)2) was identified by means of DRIFTS, showing typical bands at 780, 1024 (CH bending), 1466 (COO symmetric stretching) and 1560 (COO antisymmetric stretching) in the spectrum (Fig. 11c) [42]. Moreover, a combination of both greens can be observed in some areas. Kandinsky used cad- mium yellow, sometimes mixed with minor amounts of strontium yellow, for the yellow hues. Cadmium carbonate was not present in any DRIFT spectrum of these areas (Fig.  9a). Cinnabar and the red SOP PR60 were found in the red and pinkish red areas (Fig.  4c spot 2) respectively. PR60 yielded a high number of intense Raman bands, which are in good agreement with the reference spectrum (Fig.  11g) [40]. PR60 was mixed with cobalt blue to create the lilac hue of the oval form around the inner scene. Drying oil was clas- sified in all spectra, and a typical spectrum is given in Fig.  9a. Bands at 1466, 1747, 2856 and 2943  cm−1 can be ascribed to the oily binder, whereas the small spec- tral features at 1321 and 1365 cm−1 originate from oxa- lates [39]. Discussion Pigments A direct comparison of the pigments used in the paint- ings is given in Table  1. The results reveal specific dif- ferences in the palettes, but the following observations are valid for all the paintings: (1) zinc white is dominant among the white pigments, whereas lead white appears only in minor amounts and fillers were scarcely detected; (2) Prussian blue and ultramarine blue were mainly used for the blue areas and both were mixed with various reds to obtain violet hues, whereas only Prussian blue was mixed with various yellows for green hues; (3) Kandinsky particularly used strontium yellow and cadmium yellow for yellow areas; (4) violet pigments are absent; here these hues are always mixtures of blues and reds; and (5) vir- idian mixed with strontium yellow is the most common green in the paintings. Kandinsky used up to four differ- ent mixtures for green hues in one painting, but only one mixed green is reported from Allerheiligen II, where he varied the ratio of Prussian blue and strontium yellow (with minor amounts of cadmium yellow) to obtain green to bluish green hues (Fig. 5a–c). Kandinsky’s choices of materials evolved with his paint- ing style throughout his career, as several technical exam- inations suggest [e.g. 20, 43–46]. Unpublished reports by the Doerner Institut and several technical studies [44– 46] of paintings between 1901 and 1913 allow a com- parison of the used pigments with the presented results of our study. Generally, the previously reported results are consistent with our observations: zinc white is the dominant white pigment; Prussian blue and cadmium yellow are the most common blue and yellow pigments, respectively; cinnabar, Naples yellow, strontium yel- low, chrome yellow and yellow ochre were also found in some paintings; viridian was mostly used for green areas, and violet pigments are completely absent. Addition- ally, unpublished results of Kandinsky’s Munich palette from 1910/11 reveal the presence of cobalt blue, viridian, cinnabar, cadmium yellow, chrome yellow, bone black, zinc white and a synthetic red lake (unpublished report, Doerner Institut). Certain differences need to be empha- sized when comparing them with the results of the four reverse glass paintings. Several proofs of PR3 in paintings from 1910 and of cadmium red in a painting of 1913 indi- cate an early use of these pigments by Kandinsky [44, 46], but both reds are absent in our study and in the Munich palette. The identification of emerald green in Apokalyp- tischer Reiter II proves that Kandinsky must have used this pigment also in his Munich period and not only in his later paintings [43]. McMilian et  al. investigated a cardboard paint- ing (1930) and four canvas paintings (1938–1941) and explored Kandinsky’s use of Ripolin colours, a French Page 13 of 17Steger et al. Herit Sci (2019) 7:27 brand of oleoresinous enamel paint [43]. The results show that mainly emerald green and viridian, sometimes mixed with cadmium yellow, Prussian blue or cerulean blue, were used for the green hues in these paintings. Cadmium yellow appears in four paintings, whereas strontium yellow is completely absent. Prussian blue, ultramarine blue and cobalt-based blues were identified as blue pigments. Moreover, the authors reported cobalt violets (Co-arsenate in one painting, Co-phosphate in three paintings) from the violet areas. Some organic yel- low and red pigments were found in two paintings, but a closer identification was not possible with the methods applied. Microscopic investigation of several paint sam- ples confirmed that Kandinsky mixed tube paints with dry pigment powders and other tube paints before apply- ing them [43]. Generally, the palette reveals significant differences from the pigments identified in our study: (1) strontium yellow is absent; (2) cobalt violets were used instead of mixing red and blue paints; and (3) titanium white (production started 1909/10 [47]) is present in three paintings along with zinc white, lead white, barium sulfate and chalk. The use of synthetic organic pigments in two paint- ings needs to be further emphasised. PR83 (synthetic alizarin; CI 58000:1) and anthraquinone lake PB52 (CI 63000) were reported in Rudern. PR83 is a common SOP and can be considered as part of the standard palette of artists at that time. It has often been reported in various artworks and 20th-century reverse glass paintings [e.g. 7, 9, 48]. PB52 is formed by combining an acidic (Alizarin Saphirol B, CI 63010) and a basic dye (Methylene Blue, CI 52015) [49] and was previously identified in a product palette for artists’ colours made by Farbenfabriken vorm. Friedr. Bayer & Co (1924) [50]. No further information on this pigment is available, and to the best of our knowl- edge only one proof of it in an artwork has been reported so far: Stege et al. found PB52 in the painting Masken auf der Strasse (1910) by Ernst L. Kirchner [51]. Its unspec- tacular greenish blue hue and its limited use in the paint- ing could indicate that this colour was not so important for Kandinsky and that the pigment could be a cheap replacement product for a more expensive inorganic blue. This situation is well known for cinnabar, which was often replaced by PR3 at that time. Moreover, the naph- thalene sulfonic acid pigment lake PR60 (CI 16105) was identified in Apokalyptischer Reiter II. It was mentioned as a pigment among the Mussini resin-oil colours (H. Schmincke & Co. GmbH & Co. KG), where it was used at least until 1922 [48]. It was also found in two Eilido colour charts (Pelikan/Wagner) from 1912 and 1926/38, which further indicates its use as artists’ colours [52]. PR60 was reported in several paintings by Ernst L. Kirch- ner between 1913 and 1926 and in a lithographic ink of a printed poster from 1919 [51, 53]. Kandinsky painted large areas of Apokalyptischer Reiter II with PR60, reveal- ing a characteristic, intense scarlet red hue (Fig.  4) that is unattainable with inorganic colourants. Hence, in con- trast to PB52 (Rudern), Kandinsky very probably used PR60 intentionally, and it played a major role in the com- position of Apokalyptischer Reiter II. Another important aspect of reverse glass paintings is the use of metal foils (cf. Rudern; Fig.  10), as they enhance the gloss and create a glittering effect when the painting is viewed in reflected light. Metal pigments (e.g. fine-grained bronze powder) were also reported in other early paintings by Kandinsky [20]. Moreover, such metal powders and a tin foil were found in some reverse glass paintings by Heinrich Campendonk, who joined the “Blaue Reiter” collective in 1911 [7, 54]. Tin foils were common packaging materials for cigarettes and choco- late bars in the 19th and early 20th centuries [55]. His- torical references also recommend tin foils as a backing layer for reverse glass paintings [55, 56]. Kandinsky not only used tin foils, but also, in Rudern, glossy and silvery aluminium foils. The first commercial aluminium foil was produced in Switzerland in 1910 and started to replace its more expensive tin counterpart [57]. Its first use in Switzerland and Germany was for wrapping chocolate bars [57]. The presence of cadmium carbonate in two paintings (Allerheiligen II, Rudern) was proven by DRIFTS (Fig. 9). The measured area of Rudern does not seem to show any colour change, whereas the area of Allerheiligen II reveals a significant browning (Fig.  5c, spot 4). Cadmium car- bonate was reported in early 20th-century paintings by Henri Matisse, James Ensor, Pablo Picasso and Edvard Munch and is often accompanied by other Cd phases like sulfates, oxalates and chlorides [58–64]. Synchrotron- based spectroscopic methods applied on the microscale have allowed researchers to reveal the distribution of various cadmium compounds and confirm that cadmium carbonate can occur as a photo-degradation product of cadmium yellow and as a remnant of the production pro- cess [55, 56]. It was not only used as a starting agent for cadmium yellow synthesis [47, 65] but was also added as a lightener to the pigment powder [66]. However, the source of CdCO3 in the study presented here cannot be determined, and further investigations are needed to answer this question. Please note that cadmium car- bonate is absent in Apokalyptischer Reiter II (Fig.  9a), although Kandinsky used cadmium yellow for the yel- low areas. Reasons for this observation cannot be proven with the data available, and the following hypotheses need further investigations to be validated: (1) the paint- ings are all dated in a narrow timespan of 1911–14, which may be too short to see differences in alteration of the Page 14 of 17Steger et al. Herit Sci (2019) 7:27 same substance (i.e. when the preservation conditions have been equal for all paintings); (2) Kandinsky worked with painting materials from different brands and com- panies throughout his career, hence it is likely that he did not used the same cadmium yellow in 1911 and 1914; (3) it needs also to be considered that one brand may also include the same pigment in different pigment qualities; and (4) other paintings (canvas, watercolour) from the same time should be further investigated to reveal if cad- mium carbonate and/or degraded cadmium yellow occur to a larger extent. Binding media Barnett [67] described the painting technique of the investigated works of art as tempera (Auferstehung), tempera and china ink (Apokalyptischer Reiter II) and tempera and oil (Allerheiligen II; Rudern). It cannot be determined if this classification is based on inspections of the front or of the reverse side. Our visual inspection of the reverse side of the paintings suggests oil-based bind- ers for Auferstehung, Rudern and Apokalyptischer Reiter II rather than tempera, indicated by the slightly pastose, glossy and wet-in-wet application of the paint. China ink can be found in Allerheiligen II, Rudern and Apokalyp- tischer Reiter II and shows the typical craquelure pattern. Allerheiligen II displays differences in the visual appear- ance of the painted areas ranging from glossy to matt (Fig.  7). This suggests the use of tempera and oil-based paints and confirms the description of Barnett [67]. Kandinsky mentions that he divides his artworks between oil and watercolour paintings, although he states that he also used different materials (tempera, gouache, watercolour, oil) in both categories [68 and references therein]. Personal notes on the binding media systems of his reverse glass paintings are not known, but Roethel [69] mentions that Kandinsky was not yet using glass as a special kind of surface, as the character of brush strokes is in no way different from that used on canvas. Tem- pera paints can be roughly described as water-thinnable paints (e.g. egg) or as a system with a continuous aqueous phase (e.g. oil in a water emulsion of egg yolk and linseed oil), but the exact definition may vary distinctively for dif- ferent artists and periods [68]. Around 1900, Kandinsky conducted many experiments with tempera, as several handwritten recipes prove. For all mixtures marked “tem- pera”, he used egg yolk and various water-soluble (e.g. casein, gum arabic) and -insoluble components (e.g. mas- tic, wax, copal, Canada balsam) [70]. Several of his his- torical tube paints (tempera and oil paints) are preserved, but many of them cannot be exactly dated [71]. In Ger- many, the earliest reference to commercially available, tempera-based tube paints dates from the 1870s [68]. Analysis of the binders of nine paints from Kandinsky’s Munich palette (1910/11) identified beeswax, drying oil, egg yolk and resins as major and polysaccharides (e.g. gum arabic) as minor components, but all the paints are mixtures with at least two binders [70]. The results sug- gest that Kandinsky used tube paints, which he modified sometimes by adding other components (mainly bees- wax in turpentine) to adjust the texture, drying time and gloss of the paint [70]. The results of the four paintings in our study classify drying oil as the main binder. Only the DRIFT spectrum of the orange area of Allerheiligen II gave a hint at the presence of proteins as the water- soluble part of a tempera. The matte appearance (Fig. 7b) further suggests a different binding media system for that specific orange area. Generally, the structure (i.e. brush stroke, pastosity) and gloss of the painted surface do not play an important role for reverse glass paintings. That might be why Kandinsky used mainly oil-based colours for these paintings. Furthermore, oily binders provide the best adhesion properties for the smooth, non-porous glass surface. It is likely that Kandinsky preferred the oily binder to create more durable paintings rather than cre- ating complex paint surfaces with different appearances using tempera mixtures. The four paintings show a good state of preservation, indicating a skilful use of materials. We want to emphasize that DRIFTS may not be sen- sitive enough to detect small amounts of other binders, which could be also present in the other paintings. Please note that a mixture of one-part egg yolk and one-part oil (i.e. maximum oil content for egg tempera) would yield a dry paint with ~ 90% lipids and only ~ 10% proteins after evaporation of the water [72]. Furthermore, DRIFTS is restricted to the surface and cannot probe underlying paint layers. Advanced analyses (e.g. GC–MS), including micro-samples, are needed to get more precise informa- tion on the binders used in these paintings. Therefore, although it cannot be excluded that some of the studied paints are tempera paints, it seems rather likely that they are indeed oil paints. Conclusion The results show clearly the great importance of this technique in Kandinsky’s oeuvre, as he not only used it in Murnau between 1909 and 1914, but also contin- ued with it later in Moscow and Paris. Kandinsky pre- sented several of his reverse glass paintings in exhibitions together with paintings on canvas and cardboard (e.g. “Der Blaue Reiter” exhibition, 1911), implying a coequal importance of these techniques [11]. He found inspira- tion in folkloristic Bavarian reverse glass paintings from the 19th- century and adapted his paintings with their characteristic features like black-painted backboards and painted frames or their stylistic features like two- dimensional areas of unbroken colour, simplification of Page 15 of 17Steger et al. Herit Sci (2019) 7:27 the forms, reduction of the colouration and dominance of the line. The use of corrugated glass and cathedral glass, however, could be evidence that Kandinsky wanted to distance himself from the folkloristic tradition of reverse glass painting [20]. The presented analytical results of the four reverse glass paintings reveal a broad palette of materials. Kandinsky painted not only with inorganic pigments, but also with synthetic organic pigments. For the scarlet red pigment PR60 in Rudern, a deliberate choice of this specific pigment seems likely, despite that time’s ongoing debate on the fastness and stability of coal tar pigments in general. The application of metal foils is a specific feature of reverse glass paintings. The aluminium and tin foils in Rudern, which were both used originally as wrapping material, create a glittering effect that can be interpreted as light reflections on the water. Further research on Kandinsky’s reverse glass paintings from different periods is needed to compare the palettes and may define certain mixtures as specific for this technique. Moreover, advanced analyses of binding media (e.g. GC– MS), including micro-sampling, would give precise infor- mation also on the minor binding media components. Authors’ contributions SS conducted the measurements and was responsible for most of the writing of the manuscript. SS, HS and OH interpreted the data. DO and GG examined the iconography and the art historical context. SB, LF and IW investigated the painting and glass technique and made the photographs available. GG managed the research project. All authors contributed to the discussion of the results and reviewed the manuscript. All authors read and approved the final manuscript. Author details 1 Federal Institute for Materials Research and Testing (BAM), Division 4.5, Unter den Eichen 44-46, Berlin, Germany. 2 Museum Penzberg-Sammlung Campendonk, Am Museum 1, Penzberg, Germany. 3 Institute of Art History, LMU Munich, Zentnerstraße 31, Munich, Germany. 4 Conservator for Reverse Paintings on Glass, Garmisch-Partenkirchen, Germany. 5 Städtische Galerie im Lenbachhaus und Kunstbau München, Luisenstraße 33, Munich, Germany. 6 Doerner Institut, Bayerische Staatsgemäldesammlungen, Barer Str. 29, Munich, Germany. 7 Centre for the Study of Manuscript Cultures, University of Hamburg, Warburgstraße 26, Hamburg, Germany. Acknowledgements The authors would like to thank Patrick Dietemann for helpful comments on the binding media section and Olivier Bonnerot for fruitful discussions. Competing interests The authors declare that they have no competing interests. Availability of data and materials The datasets used and/or analysed in this study are available from the cor- responding author on reasonable request. Funding The project “Hinterglasmalerei als Technik der Klassischen Moderne 1905– 1955“is funded by the Volkswagen-Stiftung, Hannover “Forschung in Museen” reference 89921. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. Received: 16 March 2019 Accepted: 25 April 2019 References 1. Bretz S, Baumer U, Stege H, von Miller J, von Kerssenbrock-Krosig D. A German house altar from the sixteenth-century: conservation and research of reverse paintings on glass. Stud Conserv. 2009;53(4):209–24. 2. Baumer U, Dietemann P, Koller J. Identification of resinous materials on 16th and 17th-century reverse-glass objects by gas chromatography/ mass spectrometry. Int J Mass Spectrom. 2009;284:131–41. 3. Baumer U, Dietemann P. Identification and differentiation of dragon’s blood in works of art using gas chromatography/mass spectrometry. Anal Bioanal Chem. 2010;397:1363–76. 4. Baumer U, Fiedler I, Bretz S, Ranz HJ, Dietemann P. Decorative reverse- painted glass objects fromthe fourteenth to twentieth centuries: an overview of the binding media. Stud Conserv. 2012;57(sup1):9–18. https ://doi.org/10.1179/20470 58412 Y.00000 00034 . 5. Hahn O, Bretz S, Hagnau C, Ranz HJ, Wolff T. Pigments, dyes, and black enamel—the colorants of reverse paintings on glass. Archaeol Anthrop Sci. 2009;1:263–71. 6. Mollica Nardo V, Renda V, Anastasio G, Caponetti E, Saladino ML, Vasi CS, Ponterio RC. A combination of portable non-invasive techniques to study on reverse glass paintings at Mistretta museum. Microchem J. 2019;146:640–4. 7. Steger S, Stege H, Bretz S, Hahn O. Capabilities and limitations of hand- held diffuse reflectance infrared fourier transform spectroscopy (DRIFTS) for the analysis of colourants and binders in 20th-century reverse paint- ings on glass. Spectrochim Acta A. 2018;195:103–12. 8. Steger S, Bretz S, Stege H, Hahn O. Methodological approach for in situ spectroscopic analysis of modern reverse paintings on glass: a case study of Kreuzabnahme (1914/15)—an outstanding example by Carlo Mense. Eur Phys J Plus. 2019. https ://doi.org/10.1140/epjp/i2019 -12549 -6. 9. Steger S, Stege H, Bretz S, Hahn O. A complementary spectroscopic approach for the non-invasive in situ identification of synthetic organic pigments in modern reverse paintings on glass (1913–1946). J Cult Herit. 2019. https ://doi.org/10.1016/j.culhe r.2019.01.011. 10. Steger S, Oesterle D, Mayer R, Hahn O, Bretz S, Geiger G. First insights into Chinese reverse glass paintings gained by non-invasive spectroscopic analysis—tracing a cultural dialogue. Archaeol Anthrop Sci. 2019. https :// doi.org/10.1007/s1252 0-019-00799 -3. 11. Roethel HK. Gabriele Münter: 1877–1962. Ausstellung vom 13. Okt. bis 2. Dez. 1962, Städt. Galerie im Lenbachhaus, München. [exhibition cata- logue]. Munich: Städt. Galerie im Lenbachhaus; 1962 (in German). 12. Hoberg A. Gabriele Münter in München und Murnau 1901–1912. In: Hoberg A, Friedel H, editors. Gabriele Münter 1877–1962. Retrospektive. Munich: Prestel; 1992. p. 27–46 (in German). 13. Gockerell N. Hinterglasbilder, Schnitzereien und Holzspielzeug von Gabri- ele Münter gesammelt, kopiert und in ihren Werken dargestellt. Munich: Prestel; 2000 (in German). 14. Dütsch I. “…daß die Glasbilder nicht aussterben…” Neue Erkenntnisse zur Geschichte der Murnauer Hinterglasmalerei In: Bayerisches Jahrbuch für Volkskunde. Munich; 2002. p. 81–102. (in German). 15. Vail KPB. Die Werke. In: Friedel H, editor. Kandinsky—Absolut. Abstrakt. Munich: Prestel; 2008. p. 50 (in German). 16. Der Salmen B. Der Almanach „Der Blaue Reiter“. Bilder und Bildwerke in Originalen. Murnau: Schloßmuseum Murnau; 1998 (in German). 17. Hoberg A. assily Kandinsky—Absolut. Abstrakt. Konkret. In: Friedel H, edi- tor. Kandinsky—Absolut. Abstrakt. Munich: Prestel; 2008. p. 190–225 (in German). 18. Zweite A. Kandinsky und München, Begegnungen und Wandlungen 1896–1914. Munich: Prestel; 1982. 19. Roethel HK, Benjamin JK. Kandinsky. Werkverzeichnis der Ölgemälde: 1900–1915, vol. 1. Munich: C.H. Beck Verlag; 1982 (in German). 20. Wackernagel RH. “Bei ‚Öl‘auch Aquarell…, bei ‚Aquarell’auch Öl usw.”, Zu Kandinskys Ateliers und seinen Maltechniken. In: Friedel H, editor. Das Bunte Leben. Wassily Kandinsky im Lenbachhaus. Cologne: DuMont; 1995. p. 547–67 (in German). 21. Winkelmeyer I. El cosmos de color de Kandinsky Tecnologia artistica de los años 1896 a 1914. In: Fernández Félix M, editor. Kandinsky, pequeños https://doi.org/10.1179/2047058412Y.0000000034 https://doi.org/10.1179/2047058412Y.0000000034 https://doi.org/10.1140/epjp/i2019-12549-6 https://doi.org/10.1016/j.culher.2019.01.011 https://doi.org/10.1007/s12520-019-00799-3 https://doi.org/10.1007/s12520-019-00799-3 Page 16 of 17Steger et al. Herit Sci (2019) 7:27 mundos. Instituto Nacional de Bellas Artes/Museo del Palacio de Bellas Artes: México; 2018. p. 174–99 (in Spanish). 22. Bretz S. Hinterglasmalerei, die Farben leuchten so klar und rein. Maltech- nik, Geschichte, Restaurierung. Munich: Klinkhardt & Biermann; 2013 (in German). 23. Cable M. The development of flat glass manufacturing processes. Trans Newcomen Soc. 2004;74:19–43. 24. Otero V, Campos MF, Pinto JV, Vilarigues M, Carlyle L, Melo MJ. Barium, zinc and strontium yellows in late 19th- early 20th-century oil paintings. Herit Sci. 2017;5:46. 25. Hakeem MA, Jackson DE, Hamlin JJ, Errandonea D, Proctor JE, Bettinelli M. High pressure raman, optical absorption, and resistivity study of SrCrO4. Inorg Chem. 2018;57:7550–7. 26. Monico L, Janssens K, Hendriks E, Brunetti BG, Miliani C. Raman study of different crystalline forms of PbCrO4 and PbCr1-xSxO4 solid solutions for the noninvasive identification of chrome yellows in paintings: a focus on works by Vincent van Gogh. J Raman Spectrosc. 2014;45:1034–45. 27. Barsan MM, Butler IS, Fitzpatrick J, Gilson DFR. High pressure studies of the micro-raman spectra of iron cyanide complexes: prussian blue (Fe4[Fe(CN)6]3), potassium ferricyanide (K3[Fe(CN)6]), and sodium nitro- prusside (Na2[Fe(CN)5(NO)]·2H2O). J Raman Spectrosc. 2011;42:1820–4. 28. Samain L, Grandjean F, Long GJ, Martinetto P, Bordet P, Strivay D. Rela- tionship between the synthesis of prussian blue pigments, their color, physical properties, and their behavior in paint layers. J Phys Chem C. 2013;117:9693–712. 29. Silva CE, Siva LP, Edwards HGM, De Oliveira LFC. Diffuse reflection FTIR spectral database of dyes and pigments. Anal Bioanal Chem. 2006;386:2183–9. 30. Vetter W, Schreiner M. Characterization of pigment-binding media sys- tems comparison of non-invasive in situ reflection FTIR with transmission FTIR microscopy. ePS. 2011;8:10–22. 31. Miliani C, Rosi F, Borgia I, Benedetti P, Brunetti BG, Sgamellotti A. Fiber-optic fourier transform mid-infrared reflectance spectroscopy: a suitable technique for in situ studies of mural paintings. Appl Spectrosc. 2007;61(3):293–9. 32. Zumbuehl S, Scherrer NC, Berger A, Eggenberger U. Early viridian pig- ment composition characterization of a (hydrated) chromium oxide borate pigment. Stud Conserv. 2009;54:149–59. 33. Invernizzi C, Daveri A, Vagnini M, Malagodi M. Non-invasive identification of organic materials in historical stringed musical instruments by reflec- tion infrared spectroscopy: a methodological approach. Anal Bioanal Chem. 2017;409:3281–8. 34. Frost RL, Martens WN, Kloprogge T. Raman spectroscopic study of cin- nabar (HgS), realgar (As4S4), and orpiment (As2S3) at 298 and 77K. Neues JB Miner Monat. 2002;10:469–80. 35. Desnica V, Furic K, Schreiner M. Multianalytical characterization of a variety of ultramarine pigments. ePS. 2004;1:15–21. 36. Caggiani MC, Cosentino A, Mangone A. Pigments checker version 3.0, a handy set for conservation scientists: a free online Raman spectra data- base. Microchem J. 2016;129:123–32. 37. Ashoka S, Nagaraju G, Thipperudraiah KV, Chandrappa GT. Controlled synthesis of cadmium carbonate nanowires, nanoribbons, nanorings and sphere like architectures via hydrothermal method. Mater Res Bull. 2010;45:1736–40. 38. Ashoka S, Chithaiah P, Chandrappa GT. Studies on the synthesis of CdCO3 nanowires and porous CdO powder. Mater Lett. 2010;64:173–6. 39. Monico L, Rosi F, Miliani C, Daveri A, Brunetti BG. Non-invasive iden- tification of metal-oxalate complexes on polychrome artwork sur- faces by reflection mid-infrared spectroscopy. Spectrochim Acta A. 2013;116:270–80. 40. Fremout W, Saverwyns S. Identification of synthetic organic pigments: the role of a comprehensive digital Raman spectral library. J Raman Spectrosc. 2012;43:1536–44. 41. Miliani C, Rosi F, Daveri F, Brunetti BG. Reflection infrared spectroscopy for the non-invasive in situ study of artists’ pigments. Appl Phys A Mater. 2012;106:295–307. 42. Buti D, Rosi F, Brunetti BG, Miliani C. In-situ identification of copper-based green pigments. Anal Bioanal Chem. 2013;405:2699–711. 43. McMillan G, Casadio F, Fiedler I, Sorano-Stedman V. An investigation into Kandinsky’s use of Ripolin in his paintings after 1930. J Am Inst Conserv. 2013;52(4):258–77. 44. Steele E, McMillan G, Khandekar N, Mysak E. Side by side: the technical investigation of sketch I for painting with white border and painting with white border. In: Smithgall E, editor. Kandinsky and the harmony of silence: painting with white border. Washington D.C: Yale University Press; 2011. p. 106–21. 45. McMillan G, Kowalski V. Kandinsky’s materials and techniques: a prelimi- nary study of five paintings. In: Bashkoff T, editor. Kandinsky (exhibition catalogue). New York: Solomon R. Guggenheim Foundation; 2009. p. 121–3. 46. Zumbühl S, Gross M. Alexej von Jawlensky, Wassily Kandinsky – Eine Künstlerfreundschaft aus materialtechnologischer Sicht, Reihe Bild und Wissenschaft – Forschungsbeiträge zu Leben und Werk Alexej von Jawlenskys, vol. 3. Locarno: Alexej von Jawlensky Archiv; 2009. p. 256–63. 47. Mayer R. The artist’s handbook of materials and techniques. New York: Viking Press; 1991. 48. Fux A. Zur Verwendung synthetisch organischer Pigmente in der Kunst des 20. Jahrhunderts. MA-Thesis, Hochschule der Künste Bern, 2013. (in German). 49. Venkataraman K. The Chemistry of Synthetic Dyes, vol. 5. London: Aca- demic Press; 1971. 50. Schäning A. Synthetische organische Farbmittel aus einer technolo- gischen Materialsammlung des 19./20. Jahrhunderts: Identifizierung, Klassifizierung und ihre Verwendung sowie Akzeptanz in (Künstler)Farben Anfang des 20. Jahrhunderts. Ph.D.-thesis, Akademie der Bildenden Künste Wien; 2010. (in German). 51. Stege H, Richter M, Steuer C. Indanthrenblau, Helioechtrot and Pig- mentscharlach—identification of synthetic organic pigments in paint- ings of Ernst Ludwig Kirchner using Raman microscopy. ZKK J Art Technol Conserv. 2013;27(1):30–42. 52. Skowranek H, Stege H, Krekel C, Steuer C. Eilido colours. Sources relating to the introduction of coal-tar colours and their controversial reception in the early 20th-century. In: Eyb-Green S, Townsend JH, Pilz K, Kroustallis S, van Leeuwen I, editors. Sources on art technology: back to basics. London: Archetype Publications; 2016. p. 34–42. 53. Centeno S, Buisan VL, Ropret P. Raman study of synthetic organic pigments and dyes in early lithographic inks (1890–1920). J Raman Spec- trosc. 2006;37:1111–8. 54. Geiger G, Bretz S. Heinrich Campendonk, Die Hinterglasbilder. Cologne: Wienand; 2017 (in German). 55. Freytag O. Hinterglasmalerei: Ihre künstlerische Eigenart und Arbeits- weise in Vergangenheit und Gegenwart. Ravensburg: Otto Maier Verlag; 1937 (in German). 56. Wessels E. Die Hinterglasmalerei: Anleitungen zur Herstellung von Malereien hinter ober unter Glas, sowie Glasmalerei-Imitation, Glas- Vergoldung und dergleichen. Esslingen: Neff; 1913 (in German). 57. Skrabec QR. Aluminum in America: a history. Jefferson: McFarland; 2016. 58. Van der Snickt G, Dik J, Cotte M, Janssens K, Jaroszewicz J, De Nolf W, Groenewegen J, Van der Loef L. Characterization of a degraded cadmium yellow (CdS) pigment in an oil painting by means of synchrotron radia- tion based X-ray techniques. Anal Chem. 2009;81(7):2600–10. 59. Topalova-Casadiego B, Plahter U. The Scream by Edvard Munch: painting techniques and colouring materials. In: Spring M editor. Studying old master paintings technology and practice The national gallery technical bulletin 30th anniversary Conference postprints. London: Archetype; 2011. p. 244–252. 60. Pouyet E, Cotte M, Fayard B, Salomé M, Meirer F, Mehta A, Uffelman ES, Hull A, Vanmeert F, Kieffer J, Burghammer M, Janssens K, Sette F, Mass J. 2D X-ray and FTIR micro-analysis of the degradation of cadmium yellow pigment in paintings of Henri Matisse. Appl Phys A Mater. 2015;121:967–80. 61. Mass JL, Opila R, Buckley B, Cotte M, Church J, Mehta A. The photodegra- dation of cadmium yellow paints in Henri Matisse’s Le Bonheur de vivre (1905–1906). Appl Phys A Mater. 2013;111:59–68. 62. Mass J, Sedlmair J, Schmidt Patterson C, Carson D, Buckley B, Hirschmugl C. SR-FTIR imaging of the altered cadmium sulphide yellow paints in Henri Matisse’s Le Bonheur de vivre (1905–6)—examination of visually distinct degradation regions. Analyst. 2013;138(20):6032–43. 63. Miliani C, Monico L, Melo MJ, Fantacci S, Angelin EM, Romani A, Janssens K. Photochemistry of artists dyes and pigments: towards better under- standing and prevention of colour change in works of art. Angew Chem Int Ed. 2018;57:7324–34. Page 17 of 17Steger et al. Herit Sci (2019) 7:27 64. Comelli D, MacLennan D, Ghirardello M, Phenix A, Schmidt Patterson C, Khanjian H, Gross M, Valentini G, Trentelman K, Nevin A. Degradation of cadmium yellow paint: new evidence from photoluminescence studies of trap States in Picasso’s Femme (Époque des “Demoiselles d’Avignon”). Anal Chem. 2019;1:1. https ://doi.org/10.1021/acs.analc hem.8b049 14. 65. Ghirardello M, Mosca S, Marti-Rujas J, Nardo L, Burnstock A, Nevin A, et al. Time-resolved photoluminescence microscopy combined with X-ray analyses and raman spectroscopy sheds light on the imperfect synthesis of historical cadmium pigments. Anal Chem. 2018;90:10771–9. 66. Fiedler I, Bayard MA. Cadmium Yellows, Oranges and Reds. In: Feller RL, editor. Artists pigments: a handbook of their history and characteristics, vol. 1. London: Archetype Publications; 1986. p. 65–108. 67. Barnett VE. Katalog. In: Friedel H, editor. Das Bunte Leben. Wassily Kandin- sky im Lenbachhaus. Cologne: DuMont; 1995. p. 570–653 (in German). 68. Neugebauer W. Von Böcklin bis Kandinsky Kunsttechnologische Forschungen zur Temperamalerei In München zwischen 1850 und 1914. Berlin: Pro Business digital; 2015 (in German). 69. Roethel HK. Vasily kandinsky: painting on glass (Hinterglasmalerei). New York: Solomon R. Guggenheim Foundation; 1966. 70. Wackernagel RH. “Ich werde die Leute… in Öl und Tempera beschwin- deln,…”. Neues zur Maltechnik Wassily Kandinskys. ZKK Zeitschrift für Kunsttechnologie und Konservierung. 1997;11(1):97–128. (with English translation, pp. 129–144). (in German). 71. Dagron B. Inventaire du matériel de l’atelier du peintre Vassily Kandinsky. Techne, la science au service de l’histoire de l’art et des civilisations. 1998;8:64–76 (in French). 72. Dietemann P, Neugebauer W, Lutz L, Beil C, Fiedler I, Baumer U. A colloidal description of tempera and oil paints, based on a case study of Arnold Böcklin’s painting Villa am Meer II (1865). ePS. 2014;11:29–46. https://doi.org/10.1021/acs.analchem.8b04914 Kandinsky’s fragile art: a multidisciplinary investigation of four early reverse glass paintings (1911–1914) by Wassily Kandinsky Abstract Introduction Art historical context Wassily Kandinsky—reverse glass painter Description of the paintings Painting technique Glass technique Methods of the material analyses X-ray fluorescence (XRF) VIS spectroscopy (VIS) Diffuse reflectance infrared fourier transform spectroscopy (DRIFTS) Raman spectroscopy Analytical results Discussion Pigments Binding media Conclusion Authors’ contributions References