Synchrotron-Based X-ray Spectromicroscopy Used ... - ACS Publications

Aug 11, 2007 - Palais du Louvre, 14 Quai F. Mitterrand, 75001 Paris, France. Gru1newald is a famous German painter of the 16th century, whose celebrit...
0 downloads 0 Views 601KB Size
Anal. Chem. 2007, 79, 6988-6994

Synchrotron-Based X-ray Spectromicroscopy Used for the Study of an Atypical Micrometric Pigment in 16th Century Paintings M. Cotte,*†,‡ E. Welcomme,‡ V. A. Sole´,† M. Salome´,† M. Menu,‡ Ph. Walter,‡ and J. Susini†

European Synchrotron Radiation Facility, BP 220, 38043 Grenoble Cedex, France, and C2RMF, UMR171-CNRS, Palais du Louvre, 14 Quai F. Mitterrand, 75001 Paris, France

Gru1 newald is a famous German painter of the 16th century, whose celebrity is associated with his unique skill in handling colors. This article presents the analysis of materials used to render a metallic aspect in the Isenhein Altarpiece and the Basel’s Crucifixion. Such samples are challenging objects for microanalysis due to both chemical and physical complexity. Their study by synchrotron-based X-ray microscopy techniques was made possible thanks to recent developments carried out at the ID21 beam line (European Synchrotron Radiation Facility, ESRF). A submicron X-ray fluorescence probe revealed the main presence of lead, sulfur, antimony, and calcium. The fluorescence-line interferences (in particular K-lines of sulfur with M-lines of lead, and K-lines of calcium with L-lines of antimony) were resolved with the fitting program, PyMCA. 2D-mapping highlighted the presence of micrometer grains of sulfur and antimony into a lead matrix. XANES measurements were performed at both the sulfur K-edge and the antimony L-edge to refine information from an atomic to a molecular level. Beam stability was a key point in this study to selectively probe micrometer pigment grains, dispersed in the lead matrix. They confirm that the grains are made of stibnite (antimony sulfide), a very atypical pigment. Chemical mapping of sulfides is perfectly correlated with antimony mapping and provides a clear visualization of the stibnite pigments, in addition to their identification. Besides its artistic relevancy, this work aims at illustrating developments of synchrotron X-ray microprobe methods for the chemical characterization and observation of complex and micrometer-scale materials.

CHEMICAL ANALYSES OF ANCIENT PAINTINGS: CLASSICAL AND EMERGING METHODS Questions tackled when performing chemical analyses on ancient paintings usually fall into two main classes: past and future. Past-related studies are intended to reveal the artist technique and knowledge through the identification of material * To whom correspondence should be addressed. Email: marine.cotte@ culture.gouv.fr. Phone: +33 1 40 20 57 59. Fax: +33 1 40 20 68 56. † European Synchrotron Radiation Facility. ‡ C2RMF.

6988 Analytical Chemistry, Vol. 79, No. 18, September 15, 2007

components, with the possible underlying question of authentication. Future-related works deal with problems of conservation and restoration of the artifacts through the better understanding of alteration phenomena. Both contexts share common technical challenges which, to a large extent, are linked to the complexity of the materials: (i) Imaging is essential, since paintings are quite always multilayer arrangements, with the gluing layer, ground layer, priming layer, colored layers, glazes, varnishes, pentimenti, and restoration layers or superficial pollution and degradation as well. The discriminative study of each and all the strata is much more relevant than an averaged measurement. (ii) The lateral resolution is another tricky issue, for two main reasons intimately tied to the previous point. First, sampling on a work-of-art must be as less invasive as possible and implies removal of a minimum of matter. Hence, methods enabling analyses of fragments of less than 0.001 mm3 must be favored. Second, the layer thickness is usually in the range of a few to a hundred micrometers with a pigment grain of variable size down to sub-micrometer. Such small grains are responsible for the optical properties of the pigment (light diffraction and correlated optical effects). These considerations imply the use of microprobes. Even working on transversal cross-sections, the penetration depth has to be taken into account. Indeed, it must be sufficiently low to avoid an in-depth average, in particular, for the discriminative study of a mixture of grains dispersed into a matrix. (iii) The chemical sensitivity is essential in regard to the chemical complexity of the materials. Paintings are hybrid materials, a mixture of inorganic salts and organic binders, crystallized and amorphous phases, major and minor elements. A single technique is generally not sufficient to describe the whole composition of the paintings, and a multimodal approach is often crucial. A wide variety of instruments is commonly used to get insight into ancient paintings. Most of them are classical laboratory techniques, probing samples at different levels (atomic, molecular, structural), at different scales (from millimeters to nanometers), and with different sensitivities (major to trace elements). Table S-I in Supporting Information summarizes the imaging techniques most commonly used for the characterization of paintings, as well as some of their main characteristics. Although laboratory instruments become more and more powerful and remain the prime equipment for the study of cultural heritage objects, two recent evolutions can be underlined. On the one hand, the necessity of reaching sites (excavation sites, museums, 10.1021/ac0708386 CCC: $37.00

© 2007 American Chemical Society Published on Web 08/11/2007

monuments, etc.) fostered the development of portable instruments enabling in situ analyses.1,2 On the other hand, specific studies require higher levels of performance and are only possible with large instruments such as static secondary ion mass spectrometry,3-5 or in large-scale facilities (synchrotron and neutron6) which provide brighter and smaller analyzing microprobes. For instance, synchrotron radiation based-techniques are more and more exploited for the analysis of cultural heritage objects as exemplified by the growing number of dedicated workshops, conferences, and schools.7 Most of the experiments take benefit from the high flux of the source, which enables “rapid” acquisitions of data, with a good signal-to-noise ratio.8-10 A vivid example of the comparison of diffraction patterns obtained both with an X-ray generator and on a synchrotron instrument is given by Martinetto et al.8 Another asset of the synchrotron source is the energy tunability. It is essential for X-ray absorption spectroscopy.11-13 It is also at the root of synchrotron K-edge imaging, which consists in acquiring an X-ray transmission radiography above and below the K-edge of an element of interest. It was for example applied to image lead and barium in Manet’s paintings.14,15 Third, the spatial resolution is also a unique dimension offered by synchrotron instruments. For example, one of the main added-values of synchrotron FTIR microscopy over classical FTIR microscopy is the possibility to easily reduce the spot size from ∼20 × 20 µm2 to ∼5 × 5µm2 for the same signalto-noise ratio.11,16,17 Similarly, X-ray microanalyses can now rely on a sub-micrometer probe enabling 2D and 3D imaging in various contrast modes (fluorescence, absorption, and diffraction). As an example, it was recently used to image the in-depth progression of blackening cinnabar in wall paintings.18 It is also fundamental for the success of confocal XRF microscopy, which was originally (1) Moioli, P.; Seccaroni, C. X-Ray Spectrom. 2000, 29, 48-52. (2) Miliani, C.; Rosi, F.; Burnstock, A.; Brunetti, B. G.; Sgamellotti, A. Submitted for publication. (3) Van Ham, R.; van Vaeck, L.; Adams, F.; Adriaens, A. Anal. Bioanal. Chem. 2005, 383, 991-997. (4) Keune, K.; Boon, J. J. Anal. Chem. 2004, 76, 1374-1385. (5) Keune, K.; Boon, J. J. Anal. Chem. 2005, 77, 4742-4750. (6) Laurenze-Landsberg, C.; Schmidt, C. Not., Neutroni Luce Sincrotrone 2006, 11, 24-27. (7) See for example the dedicated issues Appl. Phys. A: Dooryhe´e, E., Menu, M., Susini, J., Eds. Synchrotron Radiation in Art and Archaeology. Appl. Phys. A 2006, 83 (2). (8) Martinetto, P.; Anne, M.; Dooryhe´e, E.; Tsoucaris, G.; Walter, Ph. J. Phys. IV 2000, 10, 465-472. (9) Salvado, N.; Pradell, T.; Pantos, E.; Papiz, M. Z.; Molera, J.; Seco, M.; Vendrell-Saz, M. J. Synchrotron Rad. 2002, 9, 215-222. (10) Chianelli, R. R.; Perez De la Rosa, M.; Meitzner, G.; Siadati, M.; Berhault, G.; Mehta, A.; Pople, J.; Fuentes, S.; Alonzo-Nunez, G.; Polette, L. A. J. Synchrotron Rad. 2005, 12, 129-134. (11) Cotte, M.; Checroun, E.; Susini, J.; Walter, Ph. Appl. Phys. A, to be published. (12) Sanchez del Rio, M.; Sodo, A.; Eeckhout, S. G.; Neisius, T.; Martinetto, P.; Dooryhe´e, E.; Reyes-Valerio, C. Nucl. Instrum. Methods Phys. Res., Sect. B 2005, 238, 50-54. (13) Farges, F.; Chalmin, E.; Vignaud, C.; Pallot-Frossard, I.; Susini, J.; Bargar, J.; Brown, G. E., Jr; Menu, M. Phys. Scr. 2005, T115, 885-887. (14) Dik, J.; Krug, K.; den Leeuw, M.; Bravin, A. Z. Kunsttechnologie Konservierung 2005, 315-322. (15) Krug, K.; Dik, J.; den Leeuw, M.; Whitson, A.; Tortora, J.; Coan, P.; Nemoz, C.; Bravin, A. Appl. Phys. A 2006, 83, 247-251. (16) Cotte, M.; Walter, Ph.; Tsoucaris, G.; Dumas, P. Vib. Spectrosc. 2005, 38, 159-167. (17) Cotte, M.; Dumas, P.; Richard, G.; Breniaux, R.; Walter, Ph. Anal. Chim. Acta 2005, 553 (1-2), 105-110. (18) Cotte, M.; Susini, J.; Metrich, N.; Moscato, A.; Gratziu, C.; Bertagnini, A.; Pagano, M. Anal. Chem. 2006, 78, 7484-7492.

developed on the lab source,19,20 and is now migrating to synchrotron beam lines (Berliner Elektronenspeicherring-Gesellschaft fu¨r Synchrotronstrahlung (BESSY),21 Hamburger Synchrotronstrahlungslabor (HASYLAB),22,23 and European Synchrotron Radiation Facility (ESRF)24). The in-depth resolution is particularly appropriate for noninvasive investigation of multilayered paintings. This article is intended to illustrate the recent developments of microanalytical methods at the ID21 beam line at the ESRF. They are exemplified by the identification of an atypical pigment in Gru¨newald’s paintings. Namely, the methodological approach proposed here is the combination of µ-X-ray fluorescence (µ-XRF) at low energies and µ-X-ray absorption spectroscopy (µ-XANES) for elemental and chemical imaging at the sub-micrometer scale. AN ILLUSTRATIVE CASE Mathias Gru¨newald was a major painter of the German Renaissance (first part of the 16th century). There is, in his work, a vivid strength, which remained unparalleled among the works of his contemporaries. His masterpiece, the Isenhein Altarpiece, is a complicated structure, composed of nine large panels which are 3 m high and 1.65 m large. It was painted between 1512 and 1515-1516 and represents both pain and celebration scenes such as the Crucifixion, the Annunciation or the Resurrection. This workof-art has recently been the subject of an extensive analysis to evaluate the preservation state assessment of the work.25 An older painting, the Basel’s Crucifixion (1500-1508), was also studied, in order to probe a possible evolution in the pictorial technique of the painter. His personality is expressed both through his expressionist style as well as in his handling of colors and of their association. Highly contrasting areas of light and shadow and unusually intense and iridescent color are characteristic of his art. In this article, a particular interest is put on a gray pigment, with a metallic shine. Preliminary in situ macrofluorescence analyses performed with a portable instrument26 revealed the presence of antimony, lead, and possibly sulfur. Sulfur identification was difficult due to the high concentration of lead, whose M-lines emission strongly interferes with the K-lines of sulfur. Lead is a very common element in painting as it occurs in various pigments (such as lead white, minium, lead chromate, lead sulfate, lead antimonate, lead tin yellows), and also driers (such as lead oxide or lead acetate). Antimony is more atypical and its presence in painting is generally related to Naples yellow (Pb2Sb2O7). This lead antimony oxide was already used as a glass colorant and opacifier by Egyptians around 1600-1400 B.C. Its use in painting (19) Kumakhov, M. A. X-Ray Spectrom. 2000, 29, 343-348. (20) Fiorini, C.; Longoni A.; Bjeomikhov, A. IEEE Trans. Nucl. Sci. 2001, 48 (3), 268-271. (21) Kanngiesser, B.; Malzer, W.; Reiche, I. Nucl. Instrum. Methods B 2003, 211 (2) 259-264. (22) Proost, K.; Janssens, K.; Vincze, L.; Falkenberg, G.; Gao N.; Bly, P. HASYLAB Annu. Rep. 2002. (23) Janssens, K.; Proost, K.; Falkenberg, G. Spectrochim. Acta 2004, 559,1637164. (24) Vincze, L.; Vekemans, B.; Brenker, F. E.; Falkenberg, G.; Rickers, K.; Somogyi, A.; Kersten M.; Adams, F. Anal. Chem. 2004, 76, 6786-6791. (25) Menu, M.; Ezrati, J.-J.; Laval, E.; Page`s, S.; Principaud, A.; Rioux, J.-P; Walter, Ph.; Welcomme, E.; Nowik, W. In La Technique de Gru ¨ newald et de ses Contemporains, C2RMF e´d.; Muse´e de Colmar, Eu-Artech: Colmar, France, 2007. (26) Laval, E.; Page`s-Camagna, S.; Walter, Ph. In Mona Lisa : Inside the Painting; Abrams: New York, 2006; pp90-93.

Analytical Chemistry, Vol. 79, No. 18, September 15, 2007

6989

Figure 1. Left: the Resurrection, from the Isenheim Altarpiece, by Gru¨newald. Bottom right: detail on the sleeping guard. Top right: detail on the coat of mail, where the sample MGN1 was taken. Picture taken by E. Lambert. Copyright C2RMF.

is much more recent. It would start ca. 1630 and increased from 1750 to 1850.27,28 Two obvious facts are in contradiction with the presence of Naples yellow in Gru¨newald’s painting: the color of the pigment and the date of the painting. They suggest that antimony occurs in a different chemical environment. Therefore, the main question is to characterize the antimony pigments and in particular, to determine whether they are associated with lead and/or sulfur. As shown below, antimony is present as small grains of a few micrometers diameter. The grain size and the chemical complexity of this atypical pigment make its chemical analysis highly demanding. It represents an extreme case of the technical challenges expounded in the introduction. The analyses were conducted on three samples taken in regions where antimony had been identified thanks to portable X-ray fluorescence measurements. Samples MGN1 (no. 13035) and MGN3 (no. 13036) come from the guard’s coat of mail, in the Resurrection (Unterlinden Museum, Colmar) as shown in Figure 1, whereas sample G4 (no. 13877) comes from the soldier’s armor, in the Crucifixion (museum of Basel). Cross-sections were prepared by embedding samples in polyester resin polymerized by a peroxo organic catalyst, cut with a diamond saw, and eventually polished with a diamond paste. METHODOLOGICAL APPROACH Elemental Mapping. Method of Elemental Mapping. µ-X-ray fluorescence mapping was performed at the X-ray microscopy beam line, ID21, at the ESRF (Grenoble, France) (www.esrf.fr/ UsersAndScience/Experiments/Imaging/ID21).29 The scanning X-ray microscope is optimized for very low background and low (27) Wainwright, I. N. M.; Taylor, J. M.; Harley R. D. In Artists’ Pigments. A Handbook of Their History and Characteristics; Feller, R. E., Ed.; National Gallery of Art: Washington DC, 1986; vol. 1, pp 219-254. (28) Clark, R. J. H.; Cridland, L.; Kariuki, B. M.; Harris, K. D. M.; Withnall, R. J. Chem. Soc., Dalton Trans. 1995, 2577-2582. (29) Susini, J.; Salome´, M.; Fayard, B.; Ortega, R.; Kaulich, B. Surf. Rev. Lett. 2002, 9, 203- 211.

6990 Analytical Chemistry, Vol. 79, No. 18, September 15, 2007

Figure 2. (a) X-ray fluorescence spectrum, acquired at the Fe K-edge and fitted with PyMCA. (b) Raw integrated fluorescence map of sulfur (excited at 2.495 keV), (c-e) fitted maps of sulfur, lead, and silicon, respectively (excited at 2.55 keV). (f) RGB visualization of the silicon, sulfur, and lead fitted maps. Map size, 50 × 20 µm2; step size, 0.25 µm in both directions; sample, MGN1.

detection limits. The energy range goes from 2.1 to 7.2 keV. The energy beam is determined with a fixed-exit, double crystal Si (111) monochromator, located upstream of the microscope, with an energy resolution of ∆E/E ) 10-4. The beam size is reduced to ∼1 × 0.4 µm2 (horizontal × vertical) thanks to Fresnel zone plates by geometrical demagnification of the synchrotron source. At fixed energy, the beam remains fixed while the sample is raster scanned horizontally and vertically to obtain two-dimensional images. The microfluorescence signal is collected in the horizontal plane perpendicular to the incident beam direction by using a small-area (30 mm2), HpGe solid-state, energy-dispersive detector with an energy resolution of 135 eV at 6 keV. Possible flux variation in the incoming beam (a few percent over several hours) are corrected thanks to a normalization detector, inserted just upstream of the sample. The detection limit for elements ranging from Fe to P is ∼10 ppm. The instrument is operated under vacuum to minimize air absorption, which is significant for lightelement fluorescence lines, and to avoid scattering from air. A typical X-ray fluorescence spectrum, obtained at the Fe K-edge (7.2 keV), is given in Figure 2a. It clearly illustrates the main problem faced when treating data: the numerous interferences of the K-lines of the low-Z elements, with the L or M lines of higher-Z elements. In particular, Pb M-lines interfere with the K-lines of Si, S, and Cl (in the region 1.7-3.0 keV), while Sb L-lines interfere with the Ca K-lines (3.5-4.3 keV). A first approach to obtain the fluorescence image of each of the elements separately is by taking benefit from the energy tunability. For example, the distribution of sulfur can be obtained with an excitation beam of 2.495 keV: above S K-edge and below Pb M-edges. Then the energy is tuned above Pb M-edges and below Cl K-edge (e.g., E ∼ 2.7 keV) and the lead image can be derived by subtracting the sulfur map previously measured. Theoretically feasible, this method is rather complex, long, and

Figure 3. Visible light picture and fitted maps of Si, P, S, Cl, K, Ca, Cr, Mn, Fe, Sb, and Pb (excited at 7.2 keV) on the sample MGN3. Map size, 150 × 80 µm2; step size, 1 µm in both directions.

imperfectly reliable due to the fact that absorption coefficients strongly depend on the excitation energy. We adopted a different approach much more rapid and effective, which consists in advancing X-ray fluorescence fitting algorithms in order to accurately discriminate the different contributions of the fluorescence lines. As mentioned previously, one of the key issues was the precise fit of the M-lines series of lead, which are poorly tabulated. Experimental measurements on lead compounds permitted both completion and correction of standard databases. This work was carried out in the framework of the development of the PyMCA software by the BLISS (Beam Line Instrumentation Software Support) group of the ESRF. This software is freely downloadable and allows interactive as well as batch processing of large data sets, which is particularly well suited for X-ray imaging. It implements a Levenberg-Marquardt algorithm to fit the spectra with constraints on the fitting parameters (detector characteristics, detection geometry, matrix composition, excitation energy, etc.). A complete emission line series (i.e., M or L series) is fitted by taking into account theoretical intensity ratios and line emission energies. They are particularly helpful for analysis of data collected at low energies. A more detailed description of this code can be found elsewhere.30 Figure 2a presents the resulting fit (in red) of the average fluorescence spectrum (in black), with the contribution of the various elements. The fitting configuration thus obtained is applied to each pixel of 2D maps to calculate the different elemental maps through a batch treatment. In order to check the reliability of the software, two XRF maps were acquired on the same area: one below the Pb MV-edge (at 2.495 keV) and the other above the Pb MV-edge (2.55 keV). Both are above the S K-edge. The first one directly gives the sulfur map, without interference from lead (Figure 2b). The second one was fitted to separate the contribution of sulfur (Figure 2c) and of lead (Figure 2d). The high similarity of Figure 2b and Figure 2c, without any correlation to the Figure 2d, proves the accuracy of the calculations in PyMCA. Silicon signal is well separated from lead fluorescence as well (Figure 2e). Figure 2f highlights the anticorrelation between the distribution of sulfur and lead. Grains of sulfur and silicon are surrounded with a matrix of lead. (30) Sole´, V. A.; Papillon, E.; Cotte, M.; Walter, P.; Susini, J. Spectrochim. Acta Part B 2007, 62, 63-68.

Results of Elemental Mapping. Figure 3 is an overview of elemental distributions in the sample MGN3, obtained by an excitation energy of 7.2 keV. Four main layers can be distinguished, from deep to the surface. The deeper stratum contains mainly calcium, possibly calcium carbonate. Lead is the major constituent of the second white stratum. X-ray diffraction enabled for the more precise identification of lead white, a mixture of cerussite and hydrocerussite.25 Lead is also present in the third gray layer, in association with antimony and sulfur. The superficial layer is more complex, with a mixture of almost all the detectable elements and without antimony. In the three samples, MGN1, MGN3, and G4, antimony was specifically detected in an intermediate layer. More detailed maps were acquired to highlight these precise regions with a better resolution and with a lower excitation energy (at 4.707 keV, corresponding to the Sb LI-edge). Maps of Sb, S, and Pb are given for samples MGN3 and G4 in Figure 4. They show correlated Sb and S distributions, with the largest grains a few micrometers in diameter, embedded in a lead matrix. At this level, it seems that Sb and Pb are not chemically associated, whereas Sb and S seem to be linked. Additional X-ray absorption spectroscopy experiments were performed to pursue this point and are described below. Chemical Mapping by µ-XANES. Method of Chemical Mapping by µ-XANES. The principle of XANES (X-ray absorption near edge structure) is to probe the coordination environment of a given element by tuning the energy of the probing X-ray photons across the absorption edge of this element. The spectral features observed close to the absorption edge provide information about speciation (oxidation state, coordination geometry, etc.). XANES was performed at both the S K-edge (from 2.45 to 2.57 keV, with a step of 0.25 eV) and at the Sb LI-edge (from 4.67 to 4.770 keV, with a step of 0.2 eV). With consideration of the pigment grain size, beam stability was a challenging point. Indeed, while tuning the energy, one cannot prevent the intrinsic beam spot motion. This movement is mainly induced by the monochromator rotation and by the zone plate translation (which is applied to keep the focal length constant). The beam spot motion in the sample plane is about 1 µm (horizontal) × 2.5 µm (vertical) at the S K-edge and the Sb LI-edge, over an energy scan of 60 eV. This motion is highly prejudicial when working on heterogeneities of a few microAnalytical Chemistry, Vol. 79, No. 18, September 15, 2007

6991

Figure 4. Elemental and chemical mapping on MGN1 (map size, 42 × 20 µm2) and G4 (map size, 49 × 20 µm2). Step size, 0.25 µm in both directions. Total S, Pb, and Sb maps were acquired at 4.707 keV, sulfide S(-II) maps at 2.472 keV and sulfate S(VI) maps are derived from acquisitions at 2.472 and 2.482 keV. S(-II)/ : image analysis of sulfide maps, performed with Visilog software (Noesis), to visualize the different sulfide grains (see Figure 7 for statistical distributions of the grain sizes). The colors have no significance as they are randomly applied for better visibility.

Figure 5. Spot tracking exemplified at (a) the S K-edge and (b) the Sb LI-edge. XANES spectra acquired with (w) and without (wo) spot tracking on grains of antimony sulfides (sample G4).

meters, as in the present case. The solution developed at ID21 and called “spot tracking” consists in calibrating the beam motion and subsequently correcting the sample position for each energy.31 In this way, the zone excited with the beam is constantly the same. Figure 5 compares XANES spectra acquired at the S K-edge and the Sb LI-edge, on the same grain, with (w) or without (wo) spot tracking. The spectral quality is highly improved thanks to the spot tracking. An even more important issue is highlighted by spectra acquired at the S K-edge. With the spot tracking, the peak energy (2472 eV) and the shape is typical of sulfides (see below). Without the spot tracking, it presents two additional features, at 2.482 keV and above 2.5 keV. The first one is related to a sulfate group. The second one is characteristic of the Pb MV-edge (cf. reference spectrum of Pb2Sb2O7 in Figure 6). From such a spectrum, it would be assumed that the grain contains both lead and sulfur. In fact, the presence of lead edge in the spectrum is only due to the beam motion. At the beginning of the spectrum acquisition, the beam is centered into the grain of the gray pigment, whereas at high energy, it has moved outside of the grain (31) Salome´, M. et al., manuscript in preparation.

6992 Analytical Chemistry, Vol. 79, No. 18, September 15, 2007

Figure 6. X-ray absorption spectra at (a) the S K-edge and at (b) the Sb LI-edge. Model compounds (stibnite (Sb2S3), kermesite (Sb2S2O), anglesite (PbSO4), galena (PbS), and Naples yellow (Pb2Sb2O7)) and painting fragments (MGN1, MGN3, and G4).

and probes the lead white matrix. This example illustrates how essential the spot tracking correction can be to avoid misinterpretation. Results of Chemical Mapping by µ-XANES. µ-XANES spectra were acquired on a minimum of five grains for each sample, at the S and Sb edges. The average spectra are presented in Figure 6 and are compared to reference spectra obtained on pure standards containing lead, antimony, or sulfur: stibnite (Sb2S3), kermesite (Sb2S2O), galena (PbS), anglesite (PbSO4), and Naples yellow (Pb2Sb2O7). The peak position at the S K-edge (2472 eV) is characteristic of sulfide ions. With comparison with a wide set of references,18 the shape of the edge is in perfect agreement with the identification of an antimony sulfide and more precisely stibnite (Sb2S3). The Sb LI-edge position is rather sensitive to Sb oxidation state.32 Indeed, reduced Sb(III) species (stibnite and kermesite) exhibit a LI edge about 6 eV below the one of Naples yellow (Sb(V), according to the theoretical formulas 2PbO‚Sb2O5).33 By (32) Rockenberger, J.; zum Felde, U.; Tischer, M.; Tro¨ger, L.; Haase, M.; Weller, H. J. Chem. Phys. 2000, 112, 4296-4304. (33) Sandalinas, C.; Ruiz-Moreno, S. Stud. Conserv. 2004, 49, 41-52.

Figure 7. Size distributions of the grain surfaces, in the crosssections of the three samples, MGN1, MGN3, and G4, via the statistical treatment of the sulfide and antimony maps with Visilog over n grains for each sample, cf. Figure 4.

comparison, spectra obtained on samples are characteristics of a Sb(III) speciation. They agree with the identification of stibnite and clearly refute the hypothesis of a more oxidized compound such as Naples yellow. Furthermore, additional µ-X-ray diffraction confirmed the presence of stibnite in the three samples.25 Sulfur K-edge spectra acquired outside of the grains exhibit a different feature, with an edge at higher energy (around 2.482 keV), as the one in Figure 5a (wo). This signal is characteristics of highly oxidized sulfur, namely, sulfates. To image the relative distribution of sulfides and sulfates, the sulfur fluorescence was mapped exciting it at two energies: first at 2.472 keV, which specifically excites sulfides, second, at 2.482 keV, which excites both sulfides and sulfates. A correction was then performed to separate the contribution of sulfides from this latter map. The principle of this process is detailed elsewhere.18 The resulting sulfide, (S(-II)) and sulfate, (S(+VI)), maps are presented for samples MGN1 and G4 in Figure 4. They clearly demonstrate that sulfides are perfectly correlated with antimony. Sulfates were generally observed on top of the stibnite layer, forming a kind of veil in the superficial layer, with sometimes some additional small grains in deeper layers. They could have been originally introduced into the varnish or may result from superficial alteration. A statistical treatment performed on the sulfides and antimony maps enables a better appreciation of the size and geometry of the grains. Part of the results is presented in Figure 4 for samples MGN1 and G4. Almost each grain is identified and separated from others. Grains smaller than the beam size (