www.nature.com/scientificreports
opeN
A Hidden portrait by edgar Degas
David thurrowgood1,†, David paterson2, Martin D. de Jonge2, Robin Kirkham3,
saul thurrowgood4 & Daryl L. Howard2
Received: 23 December 2015
Accepted: 22 June 2016
Published: 04 August 2016
the preservation and understanding of cultural heritage depends increasingly on in-depth
chemical studies. Rapid technological advances are forging connections between scientists and
arts communities, enabling revolutionary new techniques for non-invasive technical study of
culturally signiicant, highly prized artworks. We have applied a non-invasive, rapid, high deinition
X-ray luorescence (XRF) elemental mapping technique to a French Impressionist painting using a
synchrotron radiation source, and show how this technology can advance scholarly art interpretation
and preservation. We have obtained detailed technical understanding of a painting which could not be
resolved by conventional techniques. Here we show 31.6 megapixel scanning XRF derived elemental
maps and report a novel image processing methodology utilising these maps to produce a false colour
representation of a “hidden” portrait by edgar Degas. this work provides a cohesive methodology
for both imaging and understanding the chemical composition of artworks, and enables scholarly
understandings of cultural heritage, many of which have eluded conventional technologies. We
anticipate that the outcome from this work will encourage the reassessment of some of the world’s
great art treasures.
Preserving and interpreting the world’s material cultural heritage requires increasingly sophisticated understandings of chemical composition, environmental history, and deterioration mechanisms1. Knowledge and
understanding of historic materials has conventionally required the removal of samples which are subjected to
analytical techniques, and the process frequently alters or destroys the specimen. Even sub-millimetre sampling
“damage” to works of substantial cultural heritage can be unacceptable for highly valued objects. In art examination it is highly desirable that materials can be identiied without sampling, and without change to the material
being studied. Conventional analytical techniques have given inconclusive outcomes, in particular where the area
of interest is obscured by an upper layer2,3.
Concealed paintings, early compositions that have been hidden by subsequent work, are important insights
into artworks and artists. hey can reveal the evolution of an artist’s technique and can prove invaluable to the
attribution of a work4,5. Conventional X-radiography of paintings has been undertaken since 1896, and has been
heavily relied upon in the understanding of paintings6. X-ray absorption is mainly provided by the heavy metal
components of pigments used, and the technique provides minimal quantitative or speciic elemental identiication information. he interpretation of X-radiography images is a highly subjective process. In recent years considerable efort has been expended into developing large-area non-invasive examination techniques of artworks
and archaeometric study of objects to fulil a growing need to accurately understand the elemental and molecular
composition of artworks7–16. his new analytical information has become critical in attribution and degradation
studies and art historical assessments and is used to direct the practices of art conservators as they seek to implement new preservation strategies.
It has been demonstrated with the X-ray luorescence (XRF) technique that metallic elements from pigments
in an underpainting can be detected and resolved with suicient sensitivity to enable reconstruction of concealed paint layers2,4,5,7,10,17. he irst major synchrotron study, which revealed a woman’s head under the Van
Gogh painting Patch of Grass required extended examination time (~2 days, 2 second per pixel dwell time), and
produced modest resolution 0.5 mm over an area of only 175 × 175 mm2 2. his showed the power of scanning
XRF, but also highlighted what had been the traditional limitation of slow pixel acquisition rates, which oten
resulted in compromises to the overall scan size and/or spatial resolution. In recent years the development of
rapid scanning XRF methods8,19,20 with millisecond analysis times have dramatically improved data collection
1
national Gallery of Victoria, Melbourne, Victoria, Australia. 2Australian Synchrotron, clayton, Victoria, Australia.
The Commonwealth Scientiic and Industrial Research Organisation, Clayton, Victoria, Australia. 4Queensland Brain
Institute, University of Queensland, Brisbane, Queensland, Australia. †Present address: Queen Victoria Museum and
Art Gallery, Launceston, Tasmania, Australia. Correspondence and requests for materials should be addressed to D.T.
(email: David.Thurrowgood@launceston.tas.gov.au) or D.L.H. (email: Daryl.Howard@synchrotron.org.au)
3
Scientific RepoRts | 6:29594 | DOI: 10.1038/srep29594
1
www.nature.com/scientificreports/
Figure 1. Portrait of a Woman. Edgar Degas, French, 1834–1917, Portrait of a Woman (Portrait de Femme), c.
1876–80, oil on canvas, 46.3 × 38.2 cm, National Gallery of Victoria, Melbourne, Felton Bequest, 1937.
(a) Visible light image. he boxed region highlights the XRF scan area. (b) X-radiograph. he obscured portrait
is rotated 180 degrees relative to the upper portrait. he face and ear of the obscured sitter are the primary
source of contrast. (c) Relected infrared image (detail). A partial outline of the obscured sitter’s face is indicated
with a dotted line. he extensive use of highly infrared-absorbing black paint in the inal composition provides a
limited view of the underlying igure.
rates, enabling the potential to measure a signiicant portion of a painting at spatial resolutions on the order of the
size of a paint brush bristle10,21.
Here we demonstrate a non-invasive examination of a concealed painting, and deliver new art historical
observations, made possible with high-deinition scanning XRF analysis4,10,21. We illustrate this by investigating
a previously diicult to interpret22 portrait by Edgar Degas (1834–1917), one of the greatest French painters of
the 19th century and a founding member of Impressionism. We have examined his painting Portrait of a Woman
(Portrait de Femme, oil on canvas, 463 × 382 mm2, painted circa 1876–1880) from the collection of the National
Gallery of Victoria, Australia (Fig. 1a).
Results
Conventional imaging. Portrait of a Woman by Edgar Degas (Fig. 1a) has historically been known to have
a concealed igure, and the work has been criticised since at least 1922 for the gradually increasing outline of
the underpainting22. Degas painted directly on the underlying portrait with no intermediate ground paint layer
using exceptionally thin paint layers, thus little pigment is present to provide hiding power. he hiding power
of paint layers oten decreases as oil paintings age. he index of refraction of the natural oil medium has a tendency to increase over time, thus the diference between the pigment’s and the oil’s indices of refraction become
smaller, leading to less light scattering at the oil-pigment interface and therefore yielding lower opacity23. he
gradual increase in transparency of pigments such as emerald green (Cu(C2H3O2)2·3Cu(AsO2)2)24 and lead-based
pigments25 has been observed and studied, with metal soap formation considered to play a major role in the
process24,26.
he identity of the woman in the black dress and bonnet is currently unknown. In the visible light image
(Fig. 1a), it can be observed that the underlying portrait runs in the opposite orientation to the upper composition. he shoulders of the hidden portrait are the source of the diagonal lines radiating from the present sitter’s
bonnet to the top corners of the painting. he sitter’s face appears discoloured as the impression of the hidden
composition shows through.
An X-radiographic image (Fig. 1b) and a relected infrared image (Fig. 1c) of the painting represent the limits
of conventional practice for imaging the work27. he X-radiographic image indicates that the underlying portrait
is a young woman in three-quarter view. he main source of contrast is provided by the face and ear. Infrared
imaging is sensitive to carbon-based pigments and is oten used to reveal underdrawings which may consist of
graphite or charcoal for example28,29. In the relected infrared image, it is apparent that the black paint used in the
garment of the upper painting has high opacity to infrared radiation (Fig. 1c). his is indicative of a carbon-based
pigment such as carbon black, and its extensive use in the present portrait provides limited views to the underlying work. Portions of the underlying igure’s face are observed where it is overlapped by the present sitter’s
face. Overall the underpainting cannot be resolved as more than a faintly outlined female igure by conventional
techniques, and it has long been considered to be indecipherable, to the disappointment of academics studying
the artwork22.
High-definition XRF mapping. The use of the Maia 384 detector array30 at the X-ray Fluorescence
Microscopy (XFM) beamline31 has enabled the rapid collection of high-deinition synchrotron XRF data over
areas spanning tens of dm2 (Fig. 2). he 31.6 megapixel scanning XRF elemental maps obtained from Portrait of
Scientific RepoRts | 6:29594 | DOI: 10.1038/srep29594
2
www.nature.com/scientificreports/
Figure 2. Simpliied schematic of the synchrotron-based scanning X-ray luorescence microscope. A
monochromatic undulator-based X-ray source is focused by Kirkpatrick-Baez mirrors. he focused beam
passes through an aperture in the Maia detector onto the raster-scanned sample. he X-ray photon events are
detected in a backscatter geometry and analysed to produce elemental and scatter maps. (Edgar Degas, French,
1834–1917, Portrait of a Woman (Portrait de Femme), c. 1876–80, oil on canvas, 46.3 × 38.2 cm, National
Gallery of Victoria, Melbourne, Felton Bequest, 1937).
a Woman are presented in Fig. 3. Under the experimental conditions, relevant detectable elements range from
K-edge excitations of Z = 20–33 (Ca to As) and L-edge excitations of Z = 50–80 (Sn to Hg).
XRF imaging can be used to deduce pigment use based on the elements observed within the context of the
painting. However it cannot be used to unequivocally identify pigments. Pigment identiication can be further
supported when diferent elements are highly co-located. Co-located elements could indicate a pigment intrinsically containing diferent metals or a mixture of pigments used by the artist to achieve the desired colour. For
instance, Fe and Mn are co-located in the hidden sitter’s hair (Fig. 3a), strongly suggesting the use of the brown
pigment umber, a ine-grained rock consisting of manganese oxides and hydroxides (5–20% composition) with
iron oxyhydroxides (~45–70%)32. he Fe:Mn atomic ratio of 6:1 determined from the XRF data is consistent with
the composition of umber and its use in the hair.
he As and Cu maps suggest a possible headdress or adornment was attempted on the hidden sitter (Fig. 3a).
Cu and As are oten associated with green pigments belonging to the copper arsenite group32. A historic pigment commonly associated with the copper arsenites is Scheele’s green, which is considered to be a mixture of
several components of varying composition, (e.g., copper diarsenite (2CuO·As2O3·H2O), copper metaarsenite
(CuO·As2O3), copper arsenate (Cu(AsO2)2), etc.)32. In regions with high correlation of As and Cu, we ind the
atomic ratio of As:Cu as 2:1, consistent with the use of copper arsenite. Arsenic is also present in the hair and
bonnet of the upper composition; however these areas are relatively free of Cu, barring a region of repair, suggesting the presence of another arsenic pigment. Potential arsenic pigments are likely arsenic-sulphur based such as
realgar (As2S2, orange-red), pararealgar (As4S4, red-orange) or orpiment (As2S3, yellow to greenish-yellow), however the relatively low energy sulphur luorescence (~2.3 keV) is below the low energy limit of the luorescence
detector to aid this line of reasoning.
he hidden sitter’s face consists of several elements. he Zn map provides the best overview of the face and
showcases Degas’ brushwork (Fig. 3b). Here Zn would most likely be in the form of zinc white pigment (ZnO),
which came into widespread use ater 184532. Zinc appears to be the most thickly applied element detected on
the face, and it also present in the ear and hair of the hidden sitter. Similar to the Zn map, Co deines the hidden sitter’s face and ear, and it is also present in the hidden sitter’s garment. Cobalt is probably present as a blue
pigment, which is useful in deining lesh tones, with examples being cobalt blue (CoAl2O4) or smalt (Co-doped
alkali glass). Mercury is predominant in the facial area and would most likely correspond to the red pigment
vermilion (HgS), which would contribute to a pink lesh tone. It is primarily used on the lips, face and ears of the
hidden portrait. Cobalt is also present on the lips of the present portrait. Iron is also present in areas of the face
that are free of Mn, suggesting another pigment besides umber as was postulated for the hair. Hematite (Fe2O3)
and goethite (FeO(OH)) are plausible pigments for use in the creation of lesh tones as they have the ability to
generate red and yellow hues.
he background of the painting is deined by chromium, which is also present in some features of the hidden
sitter’s face such as the eyes. Based on the background colour, several possibilities exist such as chrome yellow
(PbCrO4, PbCrO4·xPbSO4 or PbCrO4·xPbO). Chrome yellow has a tendency to darken with time and its degradation process with respect to the works of Van Gogh has been studied in detail33–36. he Zn:Cr ratio varies over
the background, ranging from approximately 1:1 above the hidden sitter’s head to 5:1 on the let and right sides of
the painting. Based on the 1:1 Zn:Cr ratio, another possibility is the use of zinc yellow (K2O·4ZnCrO4·3H2O)32 or
the green pigment viridian (Cr2O3·2H2O) augmented with a Zn-containing pigment. Zinc yellow is also sensitive
to degradation, with George Seurat’s masterpiece A Sunday on La Grande Jatte displaying the efect37,38.
Scientific RepoRts | 6:29594 | DOI: 10.1038/srep29594
3
www.nature.com/scientificreports/
Figure 3. High-deinition 31.6 megapixel X-ray luorescence elemental maps of Portrait of a Woman.
(a) Eleven elemental maps providing an overview of the construction of the painting (426 × 267 mm2 scan).
he maps have been downsized by averaging over 4×4 pixels and displayed as the square root of the elemental
counts with the threshold value displayed in square brackets (e.g., Zn has a maximum display threshold of
100 counts, corresponding to 104 photons per pixel). (b) Detail of zinc map, also in square-root counts. he
ine brush work of the hidden sitter is clearly revealed. Image size shown is approximately 118×66 mm2 (~2.2
Mpixel). (Edgar Degas, French, 1834–1917, Portrait of a Woman (Portrait de femme) c. 1876–80, oil on canvas,
46.3 × 38.2 cm, National Gallery of Victoria, Melbourne, Felton Bequest, 1937).
he overall counts observed for Ni is low, suggesting Ni is present in low concentrations or in lower paint
layers where its luorescence would be more attenuated. At the cost of lower spatial resolution, an improved signal
to noise was achieved by averaging the map over 4 × 4 pixels, which is an advantage of high deinition mapping
when counting statistics are low. he Ni map shows a general Ni distribution throughout the painting, and it
appears that Ni is predominant in areas of the hidden sitter’s face. he rather uniform distribution throughout
the remainder of the painting may suggest that Ni is present in the ground layer. Based on the rather low counts
observed for Ni, it is unlikely a Ni-based pigment. Nickel is oten present as an impurity in many pigments, with
lead white or cobalt-based pigments being common examples39.
he hidden sitter’s garment forms an outline showing through the upper composition in the visible light image
of the painting (Fig. 1a), and it is found to consist primarily of Mn, Co and Hg. hese elements do not appear
to cover the whole of the hidden garment, thus it may also consist of, for example, low Z inorganic pigments,
carbon-based black or other organic dye-based pigments which are undetectable by XRF.
he black painted areas of the upper artwork have low concentrations of Ca (Fig. 3a). hus the black pigment used is unlikely to belong to the cokes family of black pigments, such as bone black32. Bone black contains
approximately 84 wt% Ca3(PO4)240, thus Ca luorescence would be readily observed from the painting surface if
bone black had been used. he upper portrait’s pigment composition is more likely to belong to the lame carbons
family of blacks, such as lamp black, with the carbon source possibly from a hydrocarbon precursor32.
he Ba distribution image is particularly useful in identifying the location of the upper portrait relative to the
earlier composition. Barium is routinely observed in art as barium sulphate (BaSO4) and was in common use for
preparation of commercial canvas grounds and used as a low-cost iller material or hue alteration in commercial
pigment mixtures, including zinc white and viridian32.
X-ray scatter maps.
he X-ray scatter maps (Fig. 4) provide further complementary information to the
elemental maps. he inelastic (Compton) scatter is sensitive to the lighter elements and thus enables imaging
of dense organic components such as the canvas (Fig. 4a)10. In contrast, the elastic (Rayleigh) scatter is more
Scientific RepoRts | 6:29594 | DOI: 10.1038/srep29594
4
www.nature.com/scientificreports/
Figure 4. X-ray scatter maps. (a) Inelastic and (b) elastic scatter. he boxed areas are shown inset for
greater detail, and reveal an approximate 21 × 13 mm2 puncture damage, previously restored by illing and
overpainting, and not readily evident by visible light examination. Maps displayed in a linear count scale.
sensitive to the heavier elements such as lead and mercury (Fig. 4b). Here the elastic scatter provides an image of
the ground layer and for example, the lead white paint brush stroke running across the hidden sitter’s forehead,
and is complementary to the inelastic scatter map. Earlier damage and restoration is clearly identiiable in the
scatter maps and is highlighted in Fig. 4. A negative image of the hidden sitter’s face is observed in both scatter
maps. We attribute this primarily to the heavy application of zinc-based paint in this area (Fig. 3), which would
attenuate the incident X-ray beam and then further attenuate (self-absorb) the scattered X-rays from the underlying ground layer and canvas. Overall these attenuation efects would yield lower sensitivity to materials below
the zinc layer.
Pb Raman Imaging. X-ray Raman scattering is an inelastic scattering of X-rays from core electrons. It is
normally a weak process, but can become considerably stronger through a resonance efect if the incident photon
energy is immediately below the absorption edge of a matrix material41. he incident beam excitation energy,
12.6 keV, was chosen ~0.4 keV below the Pb L3 edge to minimize the Raman scattering signal from the Pb-rich
pigments and ground layer. Another consideration in the choice of energy was to remain above the Hg L3 absorption edge at 12.284 keV and keep the incident energy high enough to minimise the inelastic scatter tail from
interfering with luorescence lines and thereby limiting sensitivity. We have previously used the 12.6 keV incident
beam energy to successfully image a painting wholly covered in lead white paint10. he Maia 384A detector’s low
energy sensitivity cutof is approximately 4 keV, so (surface) Pb detection via the low energy Pb M luorescence
lines (~2.3 keV) was not possible. In this work we found that Raman scattering could be detected and used for
imaging Pb.
To demonstrate the Raman scattering efect below the Pb L3 edge (13.035 keV), Fig. 5a shows logarithmic plots
of three spectra of a paint sample containing lead white (basic lead carbonate, 2PbCO3·Pb(OH)2) obtained at
12.6, 12.8 and 13.0 keV. Zn luorescence, originating from the canvas preparation, is also indicated in the plot. he
ratio of the integrated intensity of the most intense Raman scattering band to the elastic (Rayleigh) scatter IRaman/
Ielastic = 0.55 at 13.0 keV, and IRaman/Ielastic = 0.034 at 12.6 keV, illustrating the rapid intensity drop of the Raman
signal as a function of energy below the Pb L3 edge. he change in elastic scatter intensity was negligible over this
incident energy range.
Due to its relatively low intensity at 12.6 keV excitation energy, the Pb Raman scattering was best detected in
areas of the portrait containing surface applications of Pb-based paint, in particular the white brush strokes below
and to the right of the sitter’s face (Fig. 5b,c). A relatively low Pb Raman signal was observed for the Cr-containing
background of the painting, which may support that a non-Pb based Cr-containing pigment such as zinc yellow
or viridian was used rather than Pb-based chrome yellow (vide supra).
he Raman signal is likely not practical to use as a reliable imaging method for paintings given their highly
variable nature. However it does highlight that the choice of excitation energy is an important experimental consideration when working immediately below an absorption edge of any painting component.
Scientific RepoRts | 6:29594 | DOI: 10.1038/srep29594
5
www.nature.com/scientificreports/
Figure 5. Pb Raman imaging. (a) X-ray spectra obtained from a lead white-containing paint sample
demonstrating Raman scattering for three excitation energies below the Pb L3 absorption edge (13.035 keV).
he spectra have been ofset by factors of 10 for clarity. (b) Pb Raman scatter map (detail) indicates areas of
lead-based paint application to the painting. Map displayed in a linear count scale. (c) Detail of Portrait of a
Woman shows that the white brush strokes yield the strongest Pb Raman signal. he low intensity of the Raman
scatter at 12.6 keV excitation energy renders surface Pb most sensitive to detection, as scatter from below the
surface would be attenuated by the overlying high density paint layers. (Edgar Degas, French, 1834–1917,
Portrait of a Woman (Portrait de Femme), (c). 1876–80, oil on canvas, 46.3 × 38.2 cm, National Gallery of
Victoria, Melbourne, Felton Bequest, 1937).
Colour Reconstruction. Elemental maps enable false colour reconstruction of concealed artworks, which
provide insight to the colour palette of the artist. Previous researchers2,42 have attributed a false colour to their
elemental maps, and were able to create a plausible colour representation of the underpainting. For instance, with
the Van Gogh painting Patch of Grass the false colour efect was achieved by manually overlaying two elemental
maps with manually assigned colour and transparency2.
A false colour image of the underlying painting from Portrait of a Woman was made using a methodology for
layering multiple elemental maps. It was generated using custom-written sotware capable of merging the high
resolution, high dynamic range elemental images manually assigned with colours most likely associated with each
element (e.g., red for Hg, blue for Co). he colour is chosen based on published examples of the typical colour of a
pigment. Pigment colour is not a standardized value, and varies considerably in natural pigments43. he resulting
false colour image (Fig. 6) is a plausible representation of the artist’s work from the period44,45, and we have presented it to emphasise the underlying image.
the Hidden portrait.
Based on the observed XRF elemental maps, we propose that the revealed underpainting is a previously unknown portrait of the model Emma Dobigny. Dobigny, whose real name was Marie
Emma huilleux, modelled for Degas between 1869–1870 and is reported as a favourite model of Degas and other
French artists of the period44. We observe strong resemblance between the revealed underpainting and several of
Degas’ portraits of Emma Dobigny. Literature suggests that Degas had a special fondness for Emma Dobigny44
which may account for the otherwise uninished or unsatisfactory painting being retained by Degas.
Discussion
he high-deinition XRF maps have recorded the construction of the painting and its condition. Regions of
change, alteration or inconsistent pigment presence are immediately evident from the elemental and scatter maps.
Degas’ transformation of palette and technique is clearly documented with the years spanning the hidden portrait of Emma Dobigny (c. 1869) and the unknown sitter (c. 1876–1880). he more conventional, thicker image
Scientific RepoRts | 6:29594 | DOI: 10.1038/srep29594
6
www.nature.com/scientificreports/
Figure 6. he hidden portrait of Emma Dobigny. False colour reconstruction of Degas’ hidden portrait
(detail). he image was created from the X-ray luorescence microscopy elemental maps. (Edgar Degas, French,
1834–1917, Portrait of a Woman (Portrait de femme) c. 1876–80, oil on canvas, 46.3 × 38.2 cm, National
Gallery of Victoria, Melbourne, Felton Bequest, 1937).
layer of the underpainting is juxtaposed strikingly to the thinly applied later painting, and his change in palette
provides exceptional elemental contrast. hese maps open up new possibilities for detailing the condition of an
artwork at a point in time, and provide a method for non-invasively identifying artist technique in previously
unimaginable detail.
he high deinition scanning XRF technique allows the plausible attribution of pigments to regions, and is a
signiicant advance on conventional X-radiography which provides only pigment density for a relatively limited
number of (mostly heavy) elements. While pigment identiication relies on a subjective assessment of the pigments plausibly in use at a period in time, and the colours likely to be used in areas such as the face or hair, the
methodology provides dramatically improved probability of identifying the correct pigment when compared to
conventional non-invasive techniques.
We are not aware of any other current analytical technique that could have achieved such an imaging outcome
for this painting. he data generated by this study has provided a better understanding of the artist’s technique.
he 60 µm spatial resolution allows us to observe with conidence that a majority of the hidden sitter’s face has
been achieved as one action. However the disproportionate and blurred form of the ears is indicative of several
attempts to achieve the inal proportions and features. Degas is reported as having painted “pixie” like ears at
about this period46. By examining single elemental maps of the painting it is possible to observe such a “pixie” like
ear shape (e.g., Mn and Fe, Fig. 3) which appears to have been reworked to a more conventional form (e.g., Co and
Hg, Fig. 3). Careful study of the data reveals numerous intricacies of painting technique and brush stroke direction of the underpainting. It reveals stylistic information and elemental composition information that is unlikely
to be reproducible by persons attempting to copy a work, and the technique has strong potential for application
in authentication studies4,5.
Consideration has been given to the properties of synchrotron radiation, and the research group used visible
and chemical observation to look for radiation-induced change in preliminary experiments. Pigment binder
matrices were studied by Fourier Transform Infrared (FTIR) spectroscopy before and ater extended X-ray exposure at the XFM beamline, and spectroscopic changes were not detected. No evidence for any chemical or physical
change was observed for radiation doses 10,000 times that reported for this study, which is in accord with recent
indings by other research groups using intense radiation sources47,48.
his study has successfully demonstrated a virtual reconstruction of a hidden portrait by Edgar Degas and has
delivered a better understanding of his work and artistic practices. he authors propose that the unfolding technological developments for examining artwork using synchrotron radiation-based techniques will signiicantly
impact the ways cultural heritage is studied for authentication, preservation and scholarly purposes. We anticipate
that the high quality outcome presented here and the propagation of the rapid-scanning XRF detector technology used will further stimulate growing interest in the better understanding of our cultural assets. Parallel work
Scientific RepoRts | 6:29594 | DOI: 10.1038/srep29594
7
www.nature.com/scientificreports/
using portable XRF systems7 is demonstrating that a version of the technique is becoming viable (at substantially
reduced spatial resolution and increased data collection time) outside of a synchrotron facility, raising a strong
likelihood that precedents being set at synchrotron facilities will directly inluence emerging ield-based technologies. Until recently XRF large area scanning facilities were built in-house, and this had limited the technique’s
availability. With the introduction of commercial large scanning area instruments on the market49, the technique
has the potential to expand rapidly.
Methods
XRF imaging. he scanning XRF mapping of the painting Portrait of a Woman was performed at the X-ray
luorescence microscopy (XFM) beamline of the Australian Synchrotron31. he X-ray luorescence was acquired
with the Maia 384A detector array, which integrates the sample stage motion with continuous ly scanning,
leading to zero data readout overhead50,51. An incident excitation beam energy of 12.6 keV was used to circumvent intense luorescence from the Pb L absorption edges, which would originate primarily from the painting’s
Pb-based ground layer and thereby limit detection sensitivity to other elements in the pictorial paint layers. he
low-energy sensitivity of the detector is limited to approximately 4 keV, thus Pb-M luorescence (~2.3 keV) was
not detectable for example. he energy resolution of the detector is 375 eV at Mn Kα.
he artwork was itted to a custom manufactured cradle for scanning. he painting was placed approximately
13 mm from Maia detector rather than the optimal distance of 10 mm, since the painting was not perfectly lat.
he painting is shown mounted at the XFM beamline in Supplementary Material Fig. S1. A 426 × 267 mm2 area
was raster-scanned at 16.4 mm s−1, providing a dwell time of approximately 3.7 ms per 60 × 60 µm2 pixel and
yielded a 31.6 megapixel data set in 33 h. Given the 10 × 10 µm2 incident beam size used, the average time an area
of the painting was in the beam was 0.6 ms. he average incident lux on the painting was 1.5 × 109 photons s−1.
Full spectrum XRF data were deconvoluted into elemental maps using the dynamic analysis52 method implemented in the GeoPIXE sotware suite53. he highly complex nature of a painting with its varying pigments, paint
layers and thicknesses is inherently diicult to model with respect to the calculation of X-ray luorescence yields.
hus the elemental concentrations are considered semi-quantitative. We have shown the elemental maps as the
square root of the observed counts and adjusted the maximum intensity threshold to best represent the high
dynamic range data in a single image.
Raman scattering.
An oil-based lead white paint sample (basic lead carbonate, 2PbCO3·Pb(OH)2) was
applied to a commercially prepared canvas support. he XRF spectra of lead white was obtained at 12.6, 12.8 and
13.0 keV incident energies with the Maia detector, each with a dwell time of 25 s.
Infrared image. he relected infrared image was acquired with a Sony DSC-V1 digital camera (1/30 s exposure, f/2.8 aperture) with its infrared ilter removed. he light source was a 150 W Philips infrared heat lamp
(Infraphil type KL7500A/90) incident to front of the painting at 1 m distance.
False colour image reconstruction.
A false colour image was created by treating the elemental maps as
layers of a single image. Each elemental map was exported as a 32-bit TIFF image of 124 MB size. he custom sotware constructs an image layer for each elemental map by manual assignment of a single solid colour to the image
layer, and each pixel in the image layer is manually assigned an opacity that is proportional to the corresponding
measurement in the elemental map. For an elemental map, E, the corresponding layer opacity, L, is given by
L = (αE2.2)γ/2.2, where α is a transparency level and γ is a gamma correction value that acts like a contrast setting.
hese parameters help compress the high dynamic range measurements in the elemental maps down to the range
visible on computer screens. Finally, all layers are merged by alpha composition using the “over” operator54,
where the stack of layers are combined from the bottom layer up to the top. All parameters (layer stacking order,
layer colour, opacity) are manually selected for each image layer. he sotware enables processing of the full data
set to generate a composite image with the same resolution as the initial data in a processing pipeline fashion
that requires only a tiny fraction of the computer memory (compared to a similar operation performed in image
editing sotware) allowing merging of any number and any resolution of coloured elemental maps. Further information about the false colour image reconstruction is given as Supplementary Material and the parameters used
for the reconstruction are given in Table S1 and the coloured elemental maps used are presented in Fig. S2.
References
1. Pinna, D., Galeotti, M. & Mazzeo, R. Scientiic Examination for the Investigation of Paintings. A Handbook for Conservatorsrestorers. (Centro Di, 2009).
2. Dik, J. et al. Visualization of a Lost Painting by Vincent van Gogh Using Synchrotron Radiation Based X-ray Fluorescence Elemental
Mapping. Anal. Chem. 80, 6436–6442 (2008).
3. Lang, J. & Middleton, A. Radiography of Cultural Material 2edn, (Elsevier, 2005).
4. Alfeld, M. et al. Visualizing the 17th Century underpainting in Portrait of an Old Man by Rembrandt van Rijn using synchrotronbased scanning macro-XRF. Appl. Phys A. 111, 157–164 (2013).
5. Alfeld, M. et al. Scanning XRF investigation of a Flower Still Life and its underlying composition from the collection of the
Kröller–Müller Museum. Appl. Phys. A 111, 165–175 (2013).
6. Bridgman, C. F. he Amazing Patent on the Radiography of Paintings. Stud. Cons. 9, 135–139 (1964).
7. Alfeld, M. & Broekaert, J. A. C. Mobile depth proiling and sub-surface imaging techniques for historical paintings—A review.
Spectrochim. Acta Part B 88, 211–230 (2013).
8. Bergmann, U., Manning, P. L. & Wogelius, R. A. Chemical Mapping of Paleontological and Archeological Artifacts with Synchrotron
X-rays. Annu. Rev. Anal. Chem. 5, 361–389 (2012).
9. Hocquet, F. P. et al. Elemental 2D imaging of paintings with a mobile EDXRF system. Anal. Bioanal. Chem. 399, 3109–3116 (2011).
10. Howard, D. L. et al. High-Deinition X-ray Fluorescence Elemental Mapping of Paintings. Anal. Chem. 84, 3278–3286 (2012).
11. Janssens, K., Dik, J., Cotte, M. & Susini, J. Photon-Based Techniques for Nondestructive Subsurface Analysis of Painted Cultural
Heritage Artifacts. Acc. Chem. Res. 43, 814–825 (2010).
Scientific RepoRts | 6:29594 | DOI: 10.1038/srep29594
8
www.nature.com/scientificreports/
12. Liu, Y. et al. Imaging Mass Spectrometry with a Low-Temperature Plasma Probe for the Analysis of Works of Art. Angew. Chem. Int.
Ed. 49, 4435–4437 (2010).
13. Dafara, C., Pampaloni, E., Pezzati, L., Barucci, M. & Fontana, R. Scanning Multispectral IR Relectography SMIRR: An Advanced
Tool for Art Diagnostics. Acc. Chem. Res. 43, 847–856 (2010).
14. De Nolf, W., Dik, J., Van der Snickt, G., Wallert, A. & Janssens, K. High energy X-ray powder difraction for the imaging of (hidden)
paintings. J. Anal. At. Spectrom. 26, 910–916 (2011).
15. Dooley, K. A., Conover, D. M., Glinsman, L. D. & Delaney, J. K. Complementary Standof Chemical Imaging to Map and Identify
Artist Materials in an Early Italian Renaissance Panel Painting. Angew. Chem. Int. Ed. 53, 13775–13779 (2014).
16. Rosi, F. et al. Noninvasive Analysis of Paintings by Mid-infrared Hyperspectral Imaging. Angew. Chem. Int. Ed. 52, 5258–5261
(2013).
17. Woll, A. R. et al. he Unique History of he Armorer’s Shop: An Application of Confocal X-ray Fluorescence Microscopy. Stud.
Cons. 53, 93–109 (2008).
18. Medjoubi, K. et al. Development of fast, simultaneous and multi-technique scanning hard X-ray microscopy at Synchrotron Soleil.
J. Synch. Rad. 20, 293–299 (2013).
19. Ryan, C. G. et al. he New Maia Detector System: Methods For High Deinition Trace Element Imaging Of Natural Material. AIP
Conf. Proc. 1221, 9–17 (2010).
20. Vogt, S. & Lanzirotti, A. Trends in X-ray Fluorescence Microscopy. Synchrotron Radiat. News 26, 32–38 (2013).
21. Dredge, P. et al. Mapping Henry: Synchrotron-sourced X-ray luorescence mapping and ultra-high-deinition scanning of an early
Tudor portrait of Henry VIII. Appl. Phys. A 121, 789–800 (2015).
22. Gott, T. Edgar Degas’s Portrait of a Woman. A Re-assessment. Art Bull. Victoria 48, 84–87 (2008).
23. Brill, T. B. Light, Its Interaction with Art and Antiquities (Plenum, 1980).
24. Keune, K., Boon, J. J., Boitelle, R. & Shimadzu, Y. Degradation of Emerald green in oil paint and its contribution to the rapid change
in colour of the Descente des vaches (1834–1835) painted by héodore Rousseau. Stud. Cons. 58, 199–210 (2013).
25. Mazzeo, R. et al. Attenuated total relection micro FTIR characterisation of pigment–binder interaction in reconstructed paint ilms.
Anal. Bioanal. Chem. 392, 65–76 (2008).
26. Noble, P., van Loon, A. & Boon, J. J. Chemical changes in old master paintings II: darkening due to increased transparency as a result
of metal soap formation. ICOM-CC 14th Triennial Conference Preprints, he Hague, 12–16 September (James & James, 2005).
27. Gott, T. Edgar Degas’s Portrait of a Woman. (2008) Available at: http://www.ngv.vic.gov.au/essay/edgar-degass-portrait-of-a-woman/
(Accessed: 30th May 2016).
28. Kushel, D. A. Applications of Transmitted Infrared Radiation to the Examination of Artifacts. Stud. Cons. 30, 1–10 (1985).
29. Burmester, A. & Bayerer, F. Towards Improved Infrared Relectograms. Stud. Cons. 38, 145–154 (1993).
30. Ryan, C. G. et al. Maia X-ray luorescence imaging: Capturing detail in complex natural samples. J. Phys.: Conf. Series 499, 012002
(2014).
31. Paterson, D. et al. he X-ray Fluorescence Microscopy Beamline at the Australian Synchrotron. AIP Conf. Proc. 1365, 219–222
(2011).
32. Eastaugh, N., Walsh, V., Chaplin, T. & Siddall, R. Pigment Compendium: A Dictionary and Optical Microscopy of Historical Pigments.
(Elsevier, 2008).
33. Monico, L. et al. Degradation Process of Lead Chromate in Paintings by Vincent van Gogh Studied by Means of Synchrotron X-ray
Spectromicroscopy and Related Methods. 2. Original Paint Layer Samples. Anal. Chem. 83, 1224–1231 (2011).
34. Monico, L. et al. Full spectral XANES imaging using the Maia detector array as a new tool for the study of the alteration process of
chrome yellow pigments in paintings by Vincent van Gogh. J. Anal. At. Spectrom. 30, 613–626 (2015).
35. Monico, L. et al. Evidence for Degradation of the Chrome Yellows in Van Gogh’s Sunlowers: A Study Using Noninvasive In Situ
Methods and Synchrotron-Radiation-Based X-ray Techniques. Angew. Chem. Int. Ed. 54, 13923–13927 (2015).
36. Monico, L. et al. Degradation Process of Lead Chromate in Paintings by Vincent van Gogh Studied by Means of Spectromicroscopic
Methods. Part 5. Efects of Nonoriginal Surface Coatings into the Nature and Distribution of Chromium and Sulfur Species in
Chrome Yellow Paints. Anal. Chem. 86, 10804–10811 (2014).
37. Casadio, F. et al. Electron energy loss spectroscopy elucidates the elusive darkening of zinc potassium chromate in Georges Seurat’s
A Sunday on La Grande Jatte—1884. Anal. Bioanal. Chem. 399, 2909–2920 (2011).
38. Zanella, L. et al. he darkening of zinc yellow: XANES speciation of chromium in artist’s paints ater light and chemical exposures.
J. Anal. At. Spectrom. 26, 1090–1097 (2011).
39. Church, A. H. he Chemistry of Paints and Painting. (Seeley and Co., 1890).
40. Gettens, R. J. & Stout, G. L. Painting Materials: A Short Encyclopedia. (Dover, 1966).
41. Jaklevic, J. M., Giauque, R. D. & hompson, A. C. Resonant Raman scattering as a source of increased background in synchrotron
excited x-ray luorescence. Anal. Chem. 60, 482–484 (1988).
42. Bull, D., Krekeler, A., Alfeld, M., Dik, J. & Janssens, K. An intrusive portrait by Goya. Burlington Mag. 153, 668–673 (2011).
43. Bruni, S., Cariati, F., Casadio, F. & Toniolo, L. Spectrochemical characterization by micro-FTIR spectroscopy of blue pigments in
diferent polychrome works of art. Vib. Spectrosc. 20, 15–25 (1999).
44. Tinterow, G. & Loyrette, H. Origins of Impressionism. (Metropolitan Museum of Art, 1994).
45. Kendall, R. Degas By Himself (Chartwell, 2001).
46. Sidlauskas, S. Resisting Narrative: he Problem of Edgar Degas’s Interior. Art Bull. 75, 671–696 (1993).
47. Bertrand, L. et al. Mitigation strategies for radiation damage in the analysis of ancient materials. Trends Analyt. Chem. 66, 128–145
(2015).
48. Moini, M., Rollman, C. M. & Bertrand, L. Assessing the Impact of Synchrotron X-ray Irradiation on Proteinaceous Specimens at
Macro and Molecular Levels. Anal. Chem. 86, 9417–9422 (2014).
49. Alfeld, M. et al. A mobile instrument for in situ scanning macro-XRF investigation of historical paintings. J. Anal. At. Spectrom. 28,
760–767 (2013).
50. Kirkham, R. et al. he Maia Spectroscopy Detector System: Engineering for Integrated Pulse Capture, Low-Latency Scanning and
Real-Time Processing. AIP Conf. Proc. 1234, 240–243 (2010).
51. Ryan, C. G. et al. Elemental X-ray imaging using the Maia detector array: he beneits and challenges of large solid-angle. Nucl.
Instrum. Methods Phys. Res., A 619, 37–43 (2010).
52. Ryan, C. G. Quantitative Trace Element Imaging Using PIXE and the Nuclear Microprobe. Int. J. Imaging Syst. Technol. 11, 219–230
(2000).
53. Ryan, C. G. et al. Quantitative PIXE Microanalysis of Geological Material Using the CSIRO Proton Microprobe. Nucl. Instrum.
Methods Phys. Res., B. 47, 55–71 (1990).
54. Porter, T. & Duf, T. Compositing Digital Images. Comput. Graph. 18, 253–259 (1984).
Acknowledgements
The authors wish to thank Michael Varcoe-Cocks and John Payne of the National Gallery of Victoria for
their support. his research was undertaken at the X-ray luorescence microscopy beamline at the Australian
Synchrotron, Victoria, Australia.
Scientific RepoRts | 6:29594 | DOI: 10.1038/srep29594
9
www.nature.com/scientificreports/
Author Contributions
D.L.H., D.T., D.P., R.K. and M.D.d.J. performed the experiment. D.L.H. reduced the XRF data, took the infrared
photograph and prepared the igures. S.T. wrote the false colour image reconstruction code and created the false
colour images. D.L.H. and D.T. wrote the manuscript. All authors discussed the results.
Additional Information
Supplementary information accompanies this paper at http://www.nature.com/srep
Competing inancial interests: he authors declare no competing inancial interests.
How to cite this article: hurrowgood, D. et al. A Hidden Portrait by Edgar Degas. Sci. Rep. 6, 29594;
doi: 10.1038/srep29594 (2016).
his work is licensed under a Creative Commons Attribution 4.0 International License. he images
or other third party material in this article are included in the article’s Creative Commons license,
unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license,
users will need to obtain permission from the license holder to reproduce the material. To view a copy of this
license, visit http://creativecommons.org/licenses/by/4.0/
Scientific RepoRts | 6:29594 | DOI: 10.1038/srep29594
10