Article pubs.acs.org/ac
High-Definition X-ray Fluorescence Elemental Mapping of Paintings Daryl L. Howard,*,† Martin D. de Jonge,† Deborah Lau,‡ David Hay,‡ Michael Varcoe-Cocks,§ Chris G. Ryan,‡ Robin Kirkham,‡ Gareth Moorhead,‡ David Paterson,† and David Thurrowgood§ †
Australian Synchrotron, 800 Blackburn Road, Clayton, Victoria 3168, Australia The Commonwealth Scientific and Industrial Research Organisation, Clayton, Victoria, Australia § National Gallery of Victoria, Melbourne, Victoria, Australia ‡
S Supporting Information *
ABSTRACT: A historical self-portrait painted by Sir Arthur Streeton (1867−1943) has been studied with fast-scanning X-ray fluorescence microscopy using synchrotron radiation. One of the technique’s unique strengths is the ability to reveal metal distributions in the pigments of underlying brushstrokes, thus providing information critical to the interpretation of a painting. We have applied the nondestructive technique with the event-mode Maia X-ray detector, which has the capability to record elemental maps at megapixels per hour with the full X-ray fluorescence spectrum collected per pixel. The painting poses a difficult challenge to conventional X-ray analysis, because it was completely obscured with heavy brushstrokes of highly X-ray absorptive lead white paint (2PbCO3·Pb(OH)2) by the artist, making it an excellent candidate for the application of the synchrotron-based technique. The 25 megapixel elemental maps were successfully observed through the lead white paint across the 200 × 300 mm2 scan area. The sweeping brushstrokes of the lead white overpaint contributed significant detrimental structure to the elemental maps. A corrective procedure was devised to enhance the visualization of the elemental maps by using the elastic X-ray scatter as a proxy for the lead white overpaint. We foresee the technique applied to the most demanding of culturally significant artworks where conventional analytical methods are inadequate.
R
revealed a hidden portrait by van Gogh.22 Mobile scanning XRF systems utilizing X-ray tube sources show great promise for in situ analysis of artworks, which is important for objects that cannot be moved due to either fragility or insurance costs.23,24 Three-dimensional scanning is possible with the explicitly depth-sensitive technique of confocal XRF,25−27 which has been applied to a painting by David Teniers.28 XRF can be employed to investigate the speciation and chemical environment of elements with X-ray absorption nearedge structure (XANES), which, for example, was used to investigate the degradation process of lead chromate, such as that found in van Gogh’s Sunf lowers paintings.29,30 One of the major drawbacks of scanning XRF has been the relatively slow pixel acquisition rate, traditionally on the order of one per second; thereby the overall size or the spatial resolution of a scan has been restricted. With these limitations, the full composition of a painting has been difficult to obtain and information regarding attribution has greater uncertainty. Vincent van Gogh’s Patch of Grass is perhaps the best known example of XRF elemental mapping of a painting.22 It was scanned with a 0.5 mm2 beam size over a 17.5 × 17.5 cm2 area. The dwell time was 2 s per pixel, with a total scan time of
evealing and understanding the components and construction of painted layers in paintings is critical to conservation and art historical studies such as attribution, provenance, and discerning aspects of an artist’s technique. In recent years there have been rapid developments in a number of methods used for imaging artworks and historical samples, such as transmission radiography,1,2 infrared spectroscopy,3 and fluorescence and multispectral imaging4,5 to name a few. Analysis on both the microscopic and macroscopic scales can occur either at single points or across surface areas, and methods include X-ray diffraction,6−8 infrared (IR) and Raman spectroscopy,9−11 and emerging techniques using terahertz radiation,12 ultraviolet−visible radiation,13 and mass spectroscopic methods.14,15 X-ray fluorescence (XRF) spectroscopy has become an important nondestructive tool for the elemental analysis of artworks.16−18 An advantage of the XRF technique is its ability to reveal metal distributions in the pigments of underlying brushstrokes; thus, it has the capability to characterize a wide range of pigments. From this, correlations may be drawn about pigment use within an artist’s oeuvre and, potentially, inference of the dates surrounding pigment manufacture, introduction, and patterns of use. Single-point XRF spectroscopy is extensively used because of the relative ease of pigment discrimination,19 and scanning XRF,20,21 which provides twodimensional spatially resolved elemental distributions, has © 2012 American Chemical Society
Received: December 25, 2011 Accepted: March 7, 2012 Published: March 7, 2012 3278
dx.doi.org/10.1021/ac203462h | Anal. Chem. 2012, 84, 3278−3286
Analytical Chemistry
Article
painting. A secondary objective is to take advantage of the full XRF spectrum collected to correct the elemental maps from the absorption by the lead white surface layers, thus providing an important tool that reduces the level of assumption in the interpretation of the painting. Many analysis techniques for paintings rely on detailed examination of a limited number of discrete points on the artwork, and this information is then used to make estimations about the remainder of the work. These hindrances are overcome with the fast-scanning XRF technique presented, and it is becoming possible to have a nearly complete data set for the pigments of interest from the most analytically challenging of paintings.
approximately 2 days. To help overcome these time limitations, we have used the new Maia XRF detector system31−33 on an Xray microprobe. With its large solid angle and its capability to record megapixel images with short per-pixel dwell times, the choice of spatial pixel size and dwell time is largely dictated by experimental requirements, not limitations arising from low pixel acquisition rates. Paintings can present a unique challenge to X-ray analysis since many historical pigments such as lead white, lead−tin yellow, arsenic-containing greens, and mercury-containing vermilion contain heavy metals of high X-ray absorption. When used together, they are not able to be visually separated using X-radiography, for example, and in situations where there is extensive painted coverage, they provide blanket absorbance. In cases where the polychromatic output from tube-based X-ray sources is inadequate, synchrotron radiation may be employed instead. The advantages of synchrotron radiation are its narrow bandwidth, its brilliance, and the tunability of the source.22,28,34,35 These desirable properties often permit measurements which would otherwise be infeasible with conventional methods. Here we examine with synchrotron radiation one such challenging paintinga historical selfportrait by Sir Arthur Streeton which comprises both lead white ground layers and lead white overpainted surface layers. Arthur Streeton (1867−1943) is one of Australia’s most celebrated artists, renowned for capturing the nature and light of the Australian landscape. Streeton’s activities in portraiture were comparatively infrequent; thus, the fast-scanning XRF analysis of his abandoned self-portrait is a unique opportunity to discover more about the artist and his technique. The painting was chosen for its unusual construction that offered a challenging, yet definable experimental pathway for the XRF analysis. The artwork comes from a collection of studio materials including incomplete canvases, overpainted canvases, paint tubes, and other materials retained by the artist’s family following his death. In preparing the canvas for reuse, Streeton sweepingly applied lead white with a wide brush to obscure his abandoned portrait. Streeton painted over many earlier works later in his career, and his family oral history indicates that he prepared canvases for reuse by applying primer over older works which had not sold. He is believed to have been making use of conveniently available canvases and not undertaking deliberate destruction. The collection of the artist’s studio materials contains several examples of work in this intermediate state. Many of these canvases include a lead white pigmented layer as a constituent of the canvas priming layers. Included in the artist’s paint box are tubes of both lead- and zinc-based whites. Streeton has reprimed the surface using lead white oilbased paint. He is likely to have chosen lead white over zinc white (ZnO) for the overpaint due to its better covering power and opacity. A notable example from the period of interest is a still life painting, Crayf ish (1925, oil on canvas on composition board, National Gallery of Victoria (NGV) collection),36 in which a seemingly complete landscape has been detected by radiography below the final composition. The technical advancement of elemental mapping of paintings is the overall objective of this work. Here we present high-definition (large area, high spatial resolution) 25 megapixel elemental maps, with a 50 × 50 μm2 pixel size. The pixel size used is on the order of a brush hair diameter, thus permitting observation of the full detail of the artist’s craft. With such information, a high-definition elemental map may reveal both the finest brush work and the complete context of a
■
EXPERIMENTAL SECTION The scanning XRF mapping of the painting was conducted at the X-ray fluorescence microscopy beamline of the Australian Synchrotron.37 The painting measures 510 mm wide by 600 mm high. Due to the heavy brushstrokes of the lead white overpaint concealing the portrait, an incident beam energy below the lead L3 edge (∼13 keV) was used to avoid intense lead L shell fluorescence, which would swamp the detector with excess photons and also could mask the fluorescence of other elements of interest. Another consideration was to avoid lead Raman inelastic scattering,38 which is significant at energies close to the lead L edge and would mask mercury fluorescence. With these considerations, the painting was excited with a monochromatic beam energy of 12.6 keV. The beam was focused with a Kirkpatrick−Baez (KB) mirror pair to a size of approximately 10 × 10 μm2. The beam intensity was monitored with a N2-filled ion chamber upstream of the KB mirrors. An area of approximately 200 × 300 mm2 corresponding to the location of the head was raster-scanned at 16.4 mm s−1, providing a dwell time of approximately 3 ms per 50 × 50 μm2 pixel, and a 25 megapixel elemental map was recorded for a total scan time of 22.5 h. Several microscopic samples were removed from areas near the forehead and from the background, where existing damage was already present, to analyze the elemental content of the paint layers. The paint samples were embedded in polyester casting resin, polished to approximately 50 μm thickness, and mounted on a quartz slide. However, only one sample was found suitable for further analysis because of its orientation in the casting resin. The cross-section was oriented perpendicular to the painting’s surface so that the surface to ground layers were visible, and it was analyzed with an incident energy of 18 keV at an 8 ms dwell per pixel with both an X-ray spot size and a pixel size of 2 × 2 μm2. The higher incident energy was chosen to directly map lead and to possibly reveal more pigment metals with higher energy fluorescence such as selenium and strontium. The average time to scan a crosssection of 700 × 400 μm2 area was approximately 13 min. The cross-section was analyzed by means of an Oxford Inca electron microprobe equipped with a Si(Li) X-ray detector using a beam of 15 kV. The synchrotron-based X-ray fluorescence was acquired with the Maia detector, which consists of an array of 384 1 × 1 mm2 silicon detectors oriented axially in backscatter (180° to the incident beam) geometry and is designed to operate at a distance of only 10 mm from the sample, subtending an active solid angle of ∼1.3 sr.31,33,39 The painting was positioned an additional 7 mm downstream from the detector due to constraints imposed by the prototype painting mount and by the cooling and vacuum lines to the detector; thus, the full 3279
dx.doi.org/10.1021/ac203462h | Anal. Chem. 2012, 84, 3278−3286
Analytical Chemistry
Article
filter. Images were recorded with an ISO speed rating of 100 and a shutter speed of 0.25 s at aperture f/11. The light source was an Osram halogen bulb (650 W, 240 V, model 64540), which has a peak wavelength of approximately 850 nm. The IR transmission image was recorded with the painting placed between the light source and the camera, while the IR reflectance image was recorded with both the light source and camera incident to the front of the painting. The X-radiograph of the painting was acquired at 25 kV with a 4 mA current with a 40 s exposure on Agfa Structurix D7 film.
efficiency of the detector was not utilized. The paint sample cross-sections were scanned at the optimal distance to the detector. Maia uses a low-latency approach involving integration of the sample stage motion system with continuous fly scanning, thereby leading to zero readout overhead. The system can achieve dwell times down to 50 μs at event rates exceeding 10 M/s, so megapixel images are routine. The energy resolution of the detector in this study was 400 eV full width at halfmaximum for the Zn Kα line at 8.6 keV, having an average count rate of 1.5 × 106 photons s−1. By suppressing electronic noise contributors, the resolution of the present diode array design is expected to improve to 230−250 eV as observed in earlier prototypes. The adoption of silicon drift detector arrays in the medium term should improve resolution to about 150 eV, possibly along with improvements in count rate capacity. Full spectral XRF data are deconvoluted into elemental concentration maps using a matrix transform method called dynamic analysis40 using the GeoPIXE software suite.41 The modest energy resolution necessitates full spectral deconvolution, which is equivalent to a full least-squares fit to each pixel to resolve overlapping features, and this has greater sensitivity than simple region of interest (ROI) mapping methods from higher resolution detectors, due in part to its ability to correct for changing background levels.39 ROI mapping utilizes predefined “electronic windows” for a limited set of elements of interest. Full XRF spectra are not saved with ROI mapping, so data collection rates on the order of milliseconds per pixel can be achieved. ROI mapping has been used to great effect to reveal the hidden iron-based inks in the Archimedes Palimpsest, for example.42 However, ROI mapping requires in-depth prior knowledge of the elements within a specimen. Thus, the method has limited chances of identifying unsuspected elements, and if elements are subsequently found to be missing, data must be reacquired. Relative concentrations for the painting elemental maps were sufficient for the purposes of this study, and its maps have been displayed as the square root of the intensity to enhance the image contrast. A constant climate is the ideal situation for keeping works of art stable. The environmental conditions in the beamline’s experimental endstation were logged with NGV’s environmental monitoring equipment for several days before and during the experiment. Conditions were considered ideal for the storage of paintings with a temperature of 23 ± 0.5 °C and 55% ± 5% relative humidity over the monitoring time span. The prevention of beam damage to the artwork is an important issue when synchrotron radiation is used. A series of yet to be published studies were designed to evaluate the safety of the technique for application to artwork. A selection of Streeton’s unused historic paints were applied to canvas swatches, and these were irradiated with a dose 103 times greater than we would foresee ever occurring in practice. At no point was damage observed, including by IR spectroscopic studies of materials before and after exposure. The incident flux on the painting was 7 × 109 photons s−1. The relative absorption of IR radiation by materials used in paintings often makes IR imaging a useful technique for revealing underdrawings. The same properties can be used to image paintings using photography of transmitted IR. Infrared images were acquired with a Fujifilm ISPro digital camera, which is sensitive to 380−1000 nm radiation. The camera was equipped with a 60 mm/f 2.8 lens fitted with a Kodak 87C IR
■
RESULTS AND DISCUSSION A visible light image (detail) of the painting is shown in Figure 1a, which clearly shows the sweeping brush strokes of lead white Streeton used to obscure the self-portrait.
Figure 1. (a) Visible light image of the self-portrait (detail), (b) transmission X-radiograph of the whole painting, and (c) transmission infrared image (detail).
X-radiography and infrared imaging have complementary roles in the technical examination of artwork due to their different sensitivities to artists’ materials. For example, lead white strongly absorbs X-rays but weakly absorbs infrared radiation, while graphite and charcoal weakly absorb X-rays but strongly absorb infrared radiation. An X-radiograph of the complete painting is shown in Figure 1b. Given the large amount of lead white used over the surface and in the original ground layers, limited contrast is observed in regions around the collar, the ear, and the bridge of the nose. The transmission IR image is shown in Figure 1c (detail), which reveals the underdrawing with great detail. Larger size transmission and reflectance IR images are provided in the Supporting Information. The excellent visualization of the underlying image by infrared techniques made this painting a preferred candidate for our instrumental method and development work. Very few paintings have been imaged by XRF, and having 3280
dx.doi.org/10.1021/ac203462h | Anal. Chem. 2012, 84, 3278−3286
Analytical Chemistry
Article
several successful forms of imaging for the same painting enables assessment of the application of XRF to other paintings. An X-ray fluorescence spectrum obtained from an area of the painting with the Maia detector is shown in Figure 2. Several
Figure 2. Representative X-ray fluorescence spectrum of a 50 mm2 area of the painting near the chin collected by 1 of the 384 detector elements of the Maia detector array. Several major elemental fluorescence lines and the scattering components are labeled.
major elements contributing to the spectrum and the scattering components are identified. The low-energy cutoff of the detector is slightly below 4 keV. Given the 12.6 keV incident energy used and the low-energy cutoff of the Maia, we are in principle able to detect K shell fluorescence from calcium to arsenic and L shell fluorescence from tin to mercury. The lead white overpaint strongly attenuates fluorescence of
