MA-XRF scanning
The resulting elemental distribution maps made it possible to distinguish many new features in the painting. Figure 1 presents the visible image of the painting together with distribution maps of selected elements (As, Co, Ni, Pb, Hg, Fe, Ca). The calcium (Ca–K) map, for instance, helps visualize the bone black remains of the original hat—bone black is a calcium phosphate-based black pigment (Fig. 1h). Since the last treatment of the painting in the early 1990s, there has been much debate about the authenticity of the hat [20]. The cobalt (Co–K) map together with the Ni/As/Bi/K maps point to the use of excessive amounts of (now discolored) smalt in the background paint (Fig. 1b–d). Smalt is a blue pigment made of finely ground potassium glass that is colored blue by the addition of cobalt ore. Arsenic, iron, nickel and bismuth, are introduced with the cobalt ore and are associated with its geological source [21]. Unfortunately, smalt is not a stable pigment in oil media and the background now has a monochrome, translucent brownish color, interspersed with vague dark passages. The presence of smalt, however, in such large quantities, mixed with yellow lakes, bone black and earth pigments, suggests its color was originally different, possibly more greenish and more nuanced. Smalt is also found in parts of Isaac’s cloak, and in Rebecca’s jewelry, her rings and pearls (or glass beads) in her hair. Interestingly, the arsenic (As–K) map reveals areas of high intensity in Isaac’s sleeve and Rebecca’s dress that are not related with smalt, which also contains arsenic (Fig. 1b). The Co/Ni/Bi/K signals are low in these areas. The lead (Pb–L) and mercury (Hg–L) distribution maps demonstrate that these elements are also present in high amounts/concentrations: lead in Isaac’s sleeve, and mercury and lead in Rebecca’s dress (Fig. 1e, f). XRF analysis of arsenic in the presence of lead and mercury presents some challenges. In particular, the most intense arsenic spectral line is the Kα peak at 10.56 keV, an emission line that overlaps with the lead Lα at 10.54 keV. The arsenic Kβ peak at 11.73 keV, on the other hand, does not interfere with the Pb–L lines but shows overlap with the mercury Lβ peak at 11.82 keV. Nonetheless, the As–K, Pb–L and Hg–L maps shown in Fig. 1 display clear and different signal distributions, suggesting that, for this painting, the peak fitting algorithm of the PymCa software was relatively successful in separating the different elemental contributions to the spectral peaks. However, care must still be taken when interpreting mapping results with high concentrations of both Pb and Hg, as is the case in the red dress.
Figure 2 shows visible images of Isaac’s sleeve with corresponding elemental distribution maps of As, Pb, Sn and Fe, which can be related to specific color areas or underlayers. Rembrandt has suggested light and space in a highly sophisticated manner, with the brilliant yellow sleeve as a highpoint. Rembrandt achieved this effect by making use of bright yellow tones, combined with strongly, raised texture, which brings the sleeve literally to the foreground and enhances its brilliance. The uneven surface texture reflects the real daylight, thus reinforcing the effect of light [22]. Under the yellowish and orangey brown brushstrokes of the sleeve there is in fact an underlayer of pure lead white paint that was used to build up the thick impasto, as can be seen in the lead (Pb–L) distribution map (Fig. 2e). The painting has a brownish gray so-called ‘quartz’ or clay ground rich in ground sand and clay minerals [23]. Therefore, the lead in the Pb–L distribution map is not from the ground, but originates from the paint layers. Groen confirmed the presence of a lead white underlayer in a paint cross-section from Isaac’s sleeve, where the compact white layer can be seen underneath a transparent brown paint [7]. Apart from thick daubs and blobs of paint, Rembrandt also used a palette knife to create surface relief. The patterns of square ridges of paint testify to the use of palette knife, which Rembrandt started to use for this purpose only in the 1650s [5]. Careful comparison of the As map with the visible image reveals that the arsenic high intensity areas correspond with the brownish half-shadow and shadow areas of the sleeve, and not with the bright yellow lit areas of the sleeve (Fig. 2d). The latter show strong signals for both lead (visible in Pb–M and Pb–L maps, but only Pb–L is shown here) and tin (Sn–L) (Fig. 2f). This indicates the use of lead–tin yellow pigment, a lead–tin oxide (Pb2SnO4), in the final paint layer. The relatively smooth orangey brushstrokes in the half-shadows, on the other hand, are rich in iron, which points to the use of an earth pigment (Fig. 2c). The As/Pb/Fe maps (Fig. 1) show that the sleeve was originally slightly larger, but is covered by Isaac’s black belt in the final composition. The belt is visible in the Ca distribution map, which suggests the pigment is bone black (Fig. 1h).
The mercury (Hg–L) map shows high concentrations of this element almost exclusively in the red dress of Rebecca (Fig. 1f). Hg is associated with vermilion, a mercury sulfide (HgS). The warm red tones of Rebecca’s dress are also reflected in her hands and sleeves, as well as those of Isaac. These red brushstrokes of the light reflections show up in the Hg map, but with much lower intensity than the dress itself. The map also visualizes an initial broader expanse of the dress at the lower left in the first lay-out of the composition, now covered by Isaac’s clothing. Some tin is detected in the lighter/orangey red passages, in the center of the dress, indicating the addition of lead–tin yellow to the vermilion paint (tin map not shown). The blobs of lead white (under) paint used to build up the impasto are clear to see in the lead (Pb–L) map. Here again, the uneven relief/texture of the paint enhances the brilliance of this passage. The dark red paint areas of the dress reveal a higher signal for potassium, a good marker for the alum substrate of red lake pigment [6] (potassium map not shown).
The As map indicates that the entire skirt of Rebecca’s red dress is rich in As that is not associated with cobalt (smalt), with slightly higher concentrations in the dark red areas than in the light red areas (Fig. 1b). The areas of high intensity in the As map in the sleeve and dress indicate the use of an arsenic-containing pigment, such as orpiment (As2S3) or realgar (As4S4), but its use does not seem related to the light-colored areas or final highlights as one might expect. Based on comparison of the XRF maps with the painting, we conclude that the arsenic pigment is used in the mid- and shadow tones, mixed with other pigments, and/or in underlayers. Subsequent analysis using p-XRD and p-Raman (“p-XRD and p-Raman spot analyses”) and cross-section analysis (“Cross-section analyses”) shed further light on the type of arsenic pigment, and how it was applied.
p-XRD and p-Raman spot analyses
Portable XRD and Raman analyses were undertaken to obtain structural information about the arsenic pigment in a non-invasive manner. Measurement spots were selected based on information from the MA-XRF elemental distribution maps. We collected two XRD data sets and nine Raman spectra from different As-containing paint areas of the yellow sleeve of Isaac, as well as a few Raman spectra from As-rich spots of the red dress of Rebecca. Before collecting the XRD data, new XRF spot measurements were made at the same spot using the same instrument (see “Experimental”) to confirm the presence of As. At the first analysis spot of the yellow sleeve, Fe, As, Pb and Sn were detected. The XRD and Raman spectra both show peak patterns characteristic of lead–tin yellow type 1 (data not shown). XRD also identifies hydrocerrusite (Pb3(CO3)2(OH)2), as well as palmierite (K2Pb(SO4)2). Palmierite is a common degradation product in Old Master paintings, often associated with degraded smalt, lakes or ultramarine in combination with lead white that is present in the same paint layer or an adjacent paint layer [8]. The diffraction data show no indication for the presence of an arsenic compound, suggesting that it is either present below the detection limit or in a non-crystalline form. Similarly, Fe, As, Pb and Sn were detected at the second analysis spot of the yellow sleeve, a thick orangey daub. The XRD data are very comparable to that of the first spot, and also show diffraction patterns of lead–tin yellow type 1, hydrocerrusite and palmierite (Fig. 3a). The Raman spectrum, however, does not contain peaks for lead–tin yellow, but instead reveals a large broad peak at around 340 cm−1, that corresponds to spectra of amorphous arsenic sulfide glass g-AsxSx (Fig. 3b) [18]. The Raman spectra of other As-rich analysis spots of the yellow sleeve show a similar feature at around 340 cm−1 (spectra not shown). Although XRD and Raman were done in the same area, XRD has a much larger spot size (2 mm) compared to Raman (45 μm), and a deeper penetration. This may explain why lead–tin yellow was picked up with XRD, but not with Raman. Raman analysis of the red dress of Rebecca was less successful. The spectra only showed the presence of vermilion, and gave no indication of an As-S species in the As-rich passages as determined by MA-XRF (Fig. 3c).
Cross-section analyses
The archives of the RCE (Amsterdam, Netherlands) have 15 paint cross-sections of The Jewish Bride taken by Karin Groen during the treatment of the painting in the early 1990s. The sample forms contain detailed descriptions of the build-up and composition of the paint layers as well as schematic drawings of the samples. They also include the results of SEM-EDX analyses of the cross-sections performed at the DSM Laboratories (Geleen, Netherlands) at the time. There is no mention of the presence of an arsenic-containing pigment in the paint layers. During re-examination of the cross-sections with the light microscope, however, we observed unusual, almost perfectly round bright-yellow particles with a diameter of 2–5 μm in two cross-sections (Figs. 4, 5). These particles looked very familiar to those we encountered several years before when examining paint cross-sections from Rembrandt (workshop?), Man in a Red Cap, c. 1660 (Museum Boijmans van Beuningen, Rotterdam) (Fig. 6) [24]. In bright field, the particles characteristically exhibit a dark cross in the middle caused by internal light reflections. EDX detects exclusively the elements As and S with an atomic ratio of 2:3 in the particles, which corresponds with orpiment (As2S3). The particles are interpreted as a purified form of artificial orpiment glass obtained by dry processing, a sublimation reaction (further discussed in “Discussion”). They are isotropic and X-ray-amorphous, which corresponds with the Raman results from the previous section (“p-XRD and p-Raman spot analyses”) and explains why portable XRD did not pick up a diffraction pattern [24]. The particles are not a degradation product of the natural form [25]. In fact, this artificial form of orpiment is more stable than the natural product, which is very vulnerable to degradation when exposed to light [18].
Figure 4 presents the light microscopic and SEM-EDX analyses of the cross-section from Isaac’s yellow sleeve, visible under the black paint of the belt (sample 40/17), a later revision by the artist. The black paint layer (layer 3) contains bone black, with minor additions of yellow and red organic lake pigments. Underneath the black paint layer is the yellow–brown paint of the sleeve, which appears as two layers in the UV image (layer 2a and 2b). The yellow–brown paint (layer 2) is a rich mixture of pigments, in which we identified lead–tin yellow, lakes, a little earth, a single particle of smalt and some black pigment. Several bright yellow, ball-shaped particles of artificial orpiment, varying in diameter between 2 and 5 μm, are visible throughout the layer. They exhibit medium-gray contrast in the backscattered electron (BSE) image. The middle yellow arrow in the BSE image points to a conglomerate of three of these particles. The high intensity areas in the arsenic EDX distribution map correspond to areas with the bright yellow ball-shaped particles. These areas are also rich in sulfur, as shown by the sulfur map. Apart from the orpiment, sulfur is also associated with the lake pigments, which explains its distribution/presence throughout all paint layers. No other arsenic-containing particles or traces of arsenic were found in the paint, apart from the spherical particles. The cross-section is incomplete as the quartz ground is not present, and the bottom part of the yellow–brown paint layer shows what appears to be remnants of the lead white underlayer (layer 1), as described earlier (“MA-XRF scanning”).
In the cross-section from Rebecca’s red dress (sample 40/08), bright yellow ball-shaped particles of orpiment can be noticed in a thin orangey brown underpaint or undermodeling (layer 2) (Fig. 5). Like the previous cross-section, no other arsenic-containing particles or traces of arsenic were found in this layer, apart from the ball-shaped particles. This layer further contains lead–tin yellow, vermilion, significant amounts of lake, and a little earth. The orangey brown underpaint is applied over a thick blackish sketch layer (layer 1), and further worked up with two opaque red paint layers consisting of mostly vermilion (layers 3, 4) and a thick red glaze (layer 5). Interestingly, Groen identified the addition of gum here to thicken the glaze [7].