Research Article

Received: 2 December 2008 Revised: 10 March 2009 Accepted: 12 March 2009 Published online in Wiley Interscience: 19 May 2009

(www.interscience.com) DOI 10.1002/xrs.1181

PIXE analysis of a painting by Giorgio Vasari

N. Grassi,∗ P. Bonanni, C. Mazzotta, A. Migliori and P. A. Mandò
Among ion beam analysis (IBA) techniques, particle induced x-ray emission (PIXE) is the most widely used in the field of cultural
heritage, providing the elemental composition of materials in a non-invasive and non-destructive way. This paper reports on
the analysis by PIXE performed at one of the external beamlines of the 3 MV Tandetron accelerator of the LABEC laboratory
(Istituto Nazionale di Fisica Nucleare (INFN) – Florence) on a wood painting by Giorgio Vasari, belonging to the altar-piece
‘Pala Albergotti’ (church of Sts Flora and Lucilla, Arezzo). The measurements were taken in the framework of a campaign of
scientific investigations, promoted by the Soprintendenza of Arezzo, preliminary to the restoration of the whole altar-piece.
PIXE was used to identify the elemental composition of the original materials and to characterise the alteration layer visible
on the surface of the painting. Measurements on the same spot with proton beams of different energies (differential PIXE) also
allowed us to get information on the depth distribution of elements and on the structure of layers without the need of picking
up samples. Copyright
c 2009 John Wiley & Sons, Ltd.

Introduction differential PIXE, i.e. using proton beams with different energy on
the same spot.[1 – 4] Also, the lower depth probed by protons
Giorgio Vasari (Arezzo, 1511 – Florence, 1574) is one of the most can be helpful to reveal and characterise very thin surface
versatile artists of his time, probably better known as the author layers, as it is expected for the dark alteration on the painting
of the famous ‘Lives of the Most Excellent Painters, Sculptors, and surface.
Architects’, which is considered to be the ideological foundation
of art–historical writing. He was also an excellent architect
(he designed and realised, e.g. the Uffizi building and the
Vasarian Corridor in Florence) and a very good and productive
Experimental Setup and Measurement
painter. Conditions
The ‘Pala Albergotti’ is an altar-piece painted by Vasari in 1567,
Measurements were performed at the 3 MV Tandetron accelerator
presently kept in the church of Sts Flora and Lucilla in Arezzo. It
of the LABEC laboratory (Laboratorio di tecniche nucleari per i Beni
is a complex set of paintings on wood, composed by a central
Culturali, INFN-Florence), using the external collimated beamline
large panel and several lateral lunettes, included in a wooden
dedicated to cultural heritage applications.
frame.
Single-spot PIXE measurements were carried out with a
Owing to the poor conditions of the painting surface, the
2.9 MeV proton beam (energy on painting surface) on different
Soprintendenza of Arezzo recently promoted its restoration,
areas of the painting, as indicated in Fig. 1. Differential PIXE
which has been supported by preliminary scientific investiga-
sequences with proton energies ranging from 0.9 to 4.9 MeV
tions aimed at obtaining information on the original materials
were subsequently performed on selected spots. The beam size
and on the state of conservation of the artwork. Such prelim-
was defined by a 0.5 mm collimator. Few cm downstream the
inary investigations were coordinated by the restorers of the
collimator, a 8 µm thick Upilex window was used to extract
private consortium R.I.C.E.R : C.A and involved experts in dif-
the beam into atmosphere. The painting surface was kept
ferent scientific methodologies. Within this project, we were
at a distance of about 1 cm from the beam exit window.
asked to contribute with particle induced x-ray emission (PIXE)
measurements on the lateral painting representing Santa Lu- Beam current intensity varied between a few pA to 30 pA
cia (Figs 1 and 2). Here, besides the characterisation of the maximum (indirectly measured through the standard rotating
original pigments used by the artist, the main question was chopper[5,6] installed on the beamline). Measuring time was about
about the nature of the dark alteration of the surface in 100–300 s for each single run. Such measurement conditions
the central part of the painting. Such an alteration is absent guarantee extremely low beam charge densities delivered
on the side parts, which had been covered by the wooden to each area, thus ensuring no visible effect at all on the
frame, probably for centuries, and where colours maintain a painting.
brighter hue. The PIXE detection system was the standard used in our
The PIXE measurements were intended as a complement to the laboratory,[5] i.e. based on two silicon detectors, separately
elemental analysis previously carried out by x-ray fluorescence optimised for lower x-ray energies and medium-high x-ray
(XRF) using a portable spectrometer. It is well known that these
two techniques, very similar in their basic principle of operation,
∗
can provide information that is to some extent complementary. Correspondence to: N. Grassi, INFN-Sezione di Firenze, Via B. Rossi 1, 50019
In this particular case, PIXE was chosen to better investigate Sesto Fiorentino (Firenze), Italy. E-mail: [email protected]
the low-Z component (for which the sensitivity is in general Istituto Nazionale di Fisica Nucleare (INFN) – Sezione di Firenze and Dipar-
higher with respect to XRF) and for the possibility to obtain timento di Fisica dell’Università di Firenze, Via G. Sansone 1, 50019 Sesto
301

stratigraphic information in a non-invasive way by means of Fiorentino (Firenze), Italy

X-Ray Spectrom. 2009, 38, 301–307 Copyright

c 2009 John Wiley & Sons, Ltd.
 N. Grassi et al.

1. A silicon drift detector (SDD) by Ketek GmbH, with 150 eV

energy resolution at 5.9 keV, covering a solid angle of ∼4 msr.
The detector has a 8 µm thick Be window, the volume in front
of which is saturated with a continuous He flow (overall x-ray
transmission is still above 30% at 1 keV).
2. A Si(Li) detector by Ortec, with an energy resolution of 185 eV
at 5.9 keV, covering a much larger solid angle (∼100 msr)
to maximise efficiency for higher energy x-rays. A 400 µm
thick Mylar absorber was used to avoid the large count-rate
otherwise caused by the detection of low-energy x-rays.
The standard setup also includes a high-purity Germanium
detector to collect γ rays emitted from the target nuclei, i.e. for
particle induced gamma-ray emission (PIGE), which is routinely
exploited in our laboratory as a technique complementary to PIXE
(e.g. Ref. [7]). For the analysis of the painting by Vasari, however,
Figure 1. Areas of the painting ‘Santa Lucia’ (50 × 35 cm2 ) analysed with PIGE spectra did not add further significant information to PIXE
PIXE.
results.
Figure 2 shows the painting in front of the measurement setup
of the external beamline.

Results and Discussion

The analysed areas were selected mainly from the side parts, to
compare the composition among spots inside and those outside
the region formerly covered by the wooden frame. As can be
seen from Fig. 1, most of these spots are located in the area
of the background, whose main colour is green (with different
hue/darkness, depending on the region). The original colour can
actually be identified only from the lateral areas, because of the
strong darkening of the central region, clearly affected by some
surface alteration.

Areas in the green background

Figure 2. The painting by Vasari in front of the measurement setup at the From the analysis of the spots in the region with no alteration
external beamline at LABEC.
(formerly below the frame), we found that the green background
is mainly characterised by Cu and Pb (with the presence of Sn only
in the paler areas which show a yellow-green hue). Examples of
energies. This way, high detection efficiency is maintained from PIXE spectra (lower and higher x-ray energy detectors) collected
1 keV upwards, allowing for sensitive identification of elements with 2.9 MeV protons from two spots of different hue are given in
throughout the full detectable range (i.e. sodium included). In Fig. 3. Figure 4 summarises the results for all the green ‘clean’ spots
particular, we used: analysed. A higher abundance of Cu, anticorrelated to Pb and Sn,

Figure 3. PIXE spectra (lower and higher x-ray energy detectors) obtained with 2.9 MeV protons from two green spots of different hue. Counts are
normalised to particle fluence (in arbitrary units). The energy range 20 ÷ 30 keV is shown with expanded vertical axis to highlight the presence of Sn
302

K lines.

www.interscience.com/journal/xrs Copyright

c 2009 John Wiley & Sons, Ltd. X-Ray Spectrom. 2009, 38, 301–307
PIXE analysis of a Vasari painting

In order to characterise the dark alteration of the surface, el-

emental compositions of couples of spots close to each other,
one in the ‘clean’ area and the other in the ‘altered’ area, were
compared. Figure 5 shows an example of PIXE spectra obtained
with 2.9 MeV protons for one of these couples of spots. The same
Cu Kα intensity in the two spots indicates a relatively similar matrix
composition of the green paint layer, so that comparing x-ray
spectra is reasonably correct. The ‘altered’ area appears to be
associated to:

1. Increase in the x-ray peak intensity of some light elements

such as Mg, Al, Si, K, and Ca.
2. Decrease of the Cu L x-ray intensity (at equal Cu K lines
intensity).

Similar results were found in all the ‘altered’ areas analysed,

as can be seen from Fig. 6. Despite local variability (expected in
highly non-homogeneous objects such as paintings) in the altered
spots Al, Si and Fe show, in general, higher abundance, while Cu L
intensity is lower.
This evidence strongly suggests the presence of a surface
layer, whose inorganic fraction is mainly composed of light
elements, presumably deriving from alteration of the original
materials and/or from deposit of atmospheric particulate. It is
also plausible that a thin protective layer (darkened through
the centuries) is present and that a part of the detected
elements are inorganic traces naturally present in this material
(in addition to the contributions of pollution and/or alteration).
This surface layer strongly absorbs Cu L x-rays (energy lower
than 1 keV) from the underlying paint layer. Straightforward
calculations show that the thickness of the altered layer cannot
be more than few microns, since the Cu L x-rays are anyway still
detectable.
Differential PIXE helped us to confirm this hypothesis and added
further information. Figure 7 reports an example of differential PIXE
spectra (lower x-ray energy detector only) from one of the ‘altered’
green spots.
A qualitative comparison among spectra already gives us some
hints: it is apparent, for example, that the relative intensity of
the Si and Cu Kα peaks largely varies with beam energy. The
strong decrease of Cu with respect to Si at decreasing incident
beam energy cannot be explained only by the different trend
Figure 4. Green ‘clean’ spots: x-ray peak areas (normalised to particle of production cross-section with proton energies for the two
fluence) of Cu Kα , Pb Lα and Sn Kα lines. elements, while it can be accounted for if the Si is mostly from a
layer over the one containing Cu. If this is the case, at decreasing
beam energies (i.e. smaller beam penetration in the material) the
fraction of Cu investigated by the beam is, in fact, getting smaller
always correspond to a darker green colour. As a whole therefore, and smaller.
the use of a green Cu-based pigment can be deduced, mixed For a more soundly based deduction, the following procedure
to lead white and some lead-tin yellow for the paler yellowish can be used. Elemental concentrations are calculated from
areas. x-ray spectra taken at different beam energies, assuming a
Lighter elements (such as Al, Si, Ca and Fe) with relatively small homogeneous composition throughout the whole proton range.
abundance were also detected. While Al, Si, and Fe are probably If this assumption is correct, the same ratios for the concentrations
due to the deposit of atmospheric particulate, Ca (at least part of the main elements are obtained at different beam energies
of it) can be assumed to originate from the preparation layer because matrix effects, producing different integrated x-ray yields
underlying the paint layers (preparation is commonly based on in the different spectra, are properly taken into account by
calcium sulphate in wood paintings). The lower intensity of the the software. Instead, a concentration ratio that varies when
Ca x-ray peak in the paler spot might be thus explained by Ca calculated from the spectra at different beam energies proves
x-rays being more absorbed in the higher-density paint layer, that the assumption of a homogeneous composition throughout
owing to the presence of Sn and Pb. Besides, all these light the depth is not correct. The trend versus beam energy of
elements can also be present as impurities in the pigments or in the ‘apparent concentrations’ points out the distribution of
303

the binder. elements with depth in a much more reliable way – although

X-Ray Spectrom. 2009, 38, 301–307 Copyright

c 2009 John Wiley & Sons, Ltd. www.interscience.com/journal/xrs
 N. Grassi et al.

Figure 5. PIXE spectra (lower x-ray energy detector; counts normalised to particle fluence) from a ‘clean’ and an ‘altered’ area in the green background,
obtained with 2.9 MeV protons.

Figure 6. X-ray peak intensity for Al, Si, Fe and ratio between peak area of Cu L and Cu Kα x-rays for all the green spots analysed.

still qualitative – than the bare comparison of peak areas in the present also in an underlying layer with respect to Cu. In
different spectra. fact, their apparent concentration ratio to Cu strongly increases
Figures 8 and 9 display the results obtained with this pro- at higher beam energies, i.e. when the protons reach deeper
cedure from the differential PIXE analysis of a ‘clean’ spot layers. The contribution of Ca at deeper layers is most probably
and an ‘altered’ spot (the same as that of Fig. 7), respectively. associated to the calcium sulphate preparation layer (as it was
Apparent concentrations of the different elements (obtained
also guessed from the standard PIXE measurements, as seen
with the matrix calculation option of the software GUPIX[8,9] )
above), while the depth distribution of Pb suggests the presence
are plotted after normalisation to the Cu apparent concentra-
tion. Indeed, Cu can be considered as a ‘reference’ element, of an imprimitura made of lead white, which might be present
whose presence is acknowledged to be in the green paint between preparation and pigment layers. Both elements are
layer. anyway present in the surface layers, since the peaks of their x-rays
For the clean area (with no altered colours), the main appear also in the spectra taken at the lowest beam energies.
304

information from the trends in Fig. 8 is that Pb and Ca are Instead, the rather regular distributions for Al, Si and Fe apparently

www.interscience.com/journal/xrs Copyright

c 2009 John Wiley & Sons, Ltd. X-Ray Spectrom. 2009, 38, 301–307
PIXE analysis of a Vasari painting

Figure 8. Differential PIXE analysis of a ‘clean’ spot in the green background

area: ratio of the apparent concentrations with respect to Cu as a function
of the incident beam energy.

Figure 9. Differential PIXE analysis of an ‘altered’ spot in the green

background area: ratio of the apparent concentrations with respect to
Cu as a function of the incident beam energy.
Figure 7. Selection of PIXE spectra (lower energy x-ray detector) from a
differential PIXE sequence performed on one of the ‘altered’ green spots.
Counts are normalised to particle fluence (arbitrary units).
see Fig. 5) is therefore verified through direct measurements by
differential PIXE. Part of Pb is still detected from a layer below the
one containing Cu.
indicate their presence mainly as impurities in the pigments or
binder.
Examples of other areas
Additional information from differential PIXE can be obtained
about the thickness of the Cu-based layer in the analysed spot. Areas with different colours were also examined. Figure 10 shows,
From Fig. 8, it can be seen that the apparent concentrations do e.g. the PIXE spectra obtained with 2.9 MeV protons from two
not practically change for beam energies below 1.3 MeV, while spots on the blue chalice held by the Saint’s hand: spectra from a
they are very different at 1.8 MeV. Assuming that the underlying clean spot are compared to those of an altered spot. The presence
layers are not reached by protons of 1.3 MeV, by means of proton of Cu indicates the probable use of azurite as blue pigment;
range calculations an estimate of the order of 15–25 µm can be Pb can again be attributed to lead white (mixed to azurite, but
deduced for the local thickness of the green paint layer, in the probably also present in deeper layers, as in the green area).
hypothesis of a malachite pigment + a proper organic binder Also here the altered spot is characterised by an increase of peak
fraction. intensity for some light elements and a slight decrease in Cu L
Very different trends are obtained for the lighter elements in x-ray peak.
the altered spot (Fig. 9). Here, Al, Si, Ca and Fe show a major The incarnato areas are mainly characterised by Hg and Pb, as
apparent concentration ratio to Cu in the spectra taken at lower expected. These elements can be in fact reasonably attributed
beam energies, pointing out their higher concentration in the to the use of a mixture of the red pigment cinnabar (HgS) with
very surface layer, over the green paint layer containing Cu. The abundant lead white (this was a frequently used mixture for
presence over the green paint of such a surface layer (that we had incarnato areas). Figure 11 displays PIXE spectra collected with
305

already guessed from the comparison of 2.9 MeV PIXE spectra - the higher x-ray energy detector from two spots, the upper lip

X-Ray Spectrom. 2009, 38, 301–307 Copyright

c 2009 John Wiley & Sons, Ltd. www.interscience.com/journal/xrs
 N. Grassi et al.

Figure 10. PIXE spectra (lower and higher x-ray energy detectors; counts normalised to particle fluence) from a ‘clean’ and an ‘altered’ area on the blue
chalice, obtained with 2.9 MeV protons.

compatible with the compound CaSO4 , while the other detected

elements (Si, Fe, Cu, Pb), which are present in very small quantity,
can be interpreted as impurities or residuals from the lost overlying
paint layers.

Conclusions
PIXE and differential PIXE measurements allowed us to characterise
the materials and the layer structure of the painting analysed.
Differential PIXE, in particular, proved to be a powerful tool
to obtain stratigraphic information in a fully non-invasive way.
In particular, we were able to identify the presence of a
calcium sulphate preparation covered by a lead white based
Figure 11. PIXE spectra (higher x-ray energy detector only; counts imprimitura, to characterise the paint layers, and to characterise
normalised to particle fluence) obtained with 2.9 MeV protons from two the inorganic components of the surface layer in the altered
incarnato areas.
areas.

of the Saint and one of her fingernails. The darker hue of the Acknowledgements
lip is obtained with higher quantity of Hg and a lower quantity
This work was partially supported by the project ST@RT of Regione
of Pb.
Toscana.
Finally, direct measurements on the substrate, in areas where
The authors are very grateful to the Soprintendenza of Arezzo
the paint layers had been lost, confirmed the use of a preparation
(in particular to Paola Refice) and to Isabella Droandi, Paola Baldetti
based on calcium sulphate, which had been deduced in a and Marzia Benini of the consortium R.I.C.E.R : C.A. for giving us the
totally non-invasive way by differential PIXE. Figure 12 shows an opportunity to perform the measurements. Many thanks also to
example of PIXE spectra from the uncovered preparation (obtained the colleagues who investigated the painting with other scientific
with 2.9 MeV protons). Ca and S intensities are quantitatively techniques for the helpful discussions on the results.

Figure 12. PIXE spectra (lower and higher x-ray energy detectors) obtained with 3 MeV protons from an area on the preparation substrate, locally
306

uncovered due to pigment loss.

www.interscience.com/journal/xrs Copyright

c 2009 John Wiley & Sons, Ltd. X-Ray Spectrom. 2009, 38, 301–307
PIXE analysis of a Vasari painting

References [5] P. Del Carmine, F. Lucarelli, P. A. Mandò, A. Pecchioli, Nucl. Instrum.

Methods B. 1993, 75, 480.
[1] G. Demortier, J. L. Ruvalcaba-Sil, Nucl. Instrum. Methods B. 1996, 118, [6] M. Chiari, A. Migliori, P. A. Mandò, Nucl. Instrum. Methods B. 2002,
352. 188, 162.
[2] C. Neelmeijer, I. Brissaud, T. Calligaro, G. Demortier, A. Hautojärvi, [7] N. Grassi, A. Migliori, P. A. Mandò, H. Calvo del Castillo, Nucl. Instrum.
M. Mäder, L. Martinot, M. Schreiner, T. Tuurnala, G. Weber, X-Ray Methods B. 2004, 219-220, 48.
Spectrom. 2000, 29, 101. [8] J. A. Maxwell, J. L. Campbell, W. J. Teesdale, Nucl. Instrum. Methods B.
[3] O. Enguita, A. Climent-Font, G. Garcia, I. Montero, M. E. Fedi, 1989, 43, 218.
M. Chiari, F. Lucarelli, Nucl. Instrum. Methods B. 2002, 189, 328. [9] J. A. Maxwell, W. J. Teesdale, J. L. Campbell, Nucl. Instrum. Methods B.
[4] P. A. Mandò, M. E. Fedi, N. Grassi, A. Migliori, Nucl. Instrum. Methods 1995, 95, 407.
B. 2005, 239, 71.

307

X-Ray Spectrom. 2009, 38, 301–307 Copyright

c 2009 John Wiley & Sons, Ltd. www.interscience.com/journal/xrs