Professional Documents
Culture Documents
ABSTRACT
A novel hyperspectral imaging system (HI90, Bruker Optics), working in the mid-infrared range and recently developed
for the remote identification and mapping of hazardous compounds, has here been optimized for investigating painting
surfaces. The painting Sestante 10 (1982) by Alberto Burri has been spectrally and spatially investigated with the HI90
system revealing the distribution of inorganic materials constituting the artworks. In order to validate the results
obtainable by the imager for the pigment identification previous tests on laboratory models were performed. Yellow,
white and blue pigments painted with different binders (namely egg, alkyd, acrylic and vinyl) were investigated by the
HI90. Afterwards, the polychrome painting Sestante 10 was investigated focusing the attention on the inorganic material
distribution revealing the presence of different extenders (kaolin, BaSO4, CaSO4) mixed with the various silica-based
pigments present in the painting. The brightness temperature spectra collected by HI90 have also been compared to
single point reflection spectra acquired by a conventional portable FTIR spectrometer (Alpha-R by Bruker Optics)
highlighting the good spectral quality of the imaging system. This comparison permitted also to evaluate the spectral
response and the diagnostic strengths of the spectral range available by the HI90 imaging (1300-860 cm-1), validating the
reliability of the obtained chemical images. This study clearly highlights the high potential of the new hyperspectral
imaging system and opens up new perspectives in the current scientific interest devoted to the application of mapping
and imaging methods for the study of painting surfaces.
Keywords: mid-FTIR, pigments, hyperspectral imaging, non invasive methods, portable techniques, pigments
1. INTRODUCTION
Imaging techniques have been widely applied in the cultural heritage science offering an alternative view of the artwork
through the visualization of underlying or surface layers otherwise not perceptible to the naked eye. They were employed
to art objects since the 20s when E.W. Forbes for the first time experimented the X-ray radiography to see and image
through the paintings revealing inner structures, metal elements and materials with different compositions. Since then,
many improvements have been made in the imaging technologies evolving from the concept of basic imaging towards
the necessity of correlating the images with the material characterization achievable through the modern spectral, multi
and hyperspectral imaging systems.
At this regard, the research trend turned towards the exploitation of the different spectral ranges and the optimization of
the spectral resolution in terms of number of acquirable bands/spectra to increase the specificity of the imaging
technique. As an example, the use of a portable scanning macro X-ray fluorescence system recently allowed for imaging
not only the elemental pigment distribution but to reveal the artists modus operandi visualizing changes made during the
painting creation and in some cases discovering hidden paintings from later overpainting1. Molecular imaging
investigations in the visible (VIS: 400-750 nm) and near infrared (NIR: 750-2500 nm) ranges have been exploited for the
pigment characterization and localization investigating both the electronic levels and the combination and overtone
vibrational modes of the infrared range2. More recently, Delaney et al.3 demonstrated the effectiveness of a spectral
Optics for Arts, Architecture, and Archaeology IV, edited by Luca Pezzati,
Piotr Targowski, Proc. of SPIE Vol. 8790, 87900Q 2013 SPIE
CCC code: 0277-786X/13/$18 doi: 10.1117/12.2020477
2. EXPERIMENTAL
2.1 Mid-Infrared Hyperspectral Imager
The hyperspectral imager HI90 is based on the combination of a Michelson interferometer with a focal plane array (FPA)
MCT detector. The plane mirror interferometer is actively aligned. The maximum lateral resolution is 256x256 pixels
and cooling is performed by a Stirling cooler. The configuration applied in this study operates in the spectral range 1300-
860 cm-1. A spectral resolution of 4 cm-1 and a lateral resolution of 128x128 pixels were chosen. The hyperspectral
imager measures the emitted and reflected radiation of the painting from a distance of app. 1 m. The field of view of each
pixel is 0.69 mrad, investigating a total area of 8.8x8.8 cm2. The data acquisition time for 32 scans with these parameters
is app. 80 s. The radiometric calibration is automatically performed by an internal active calibration system. This enables
the calculation of brightness temperature (K) spectra. The measurements are performed illuminating the sample with an
infrared radiation source. The source is controlled via USB allowing the user to control the radiant power. To minimize
the exposure time of the analyzed painting, the source is equipped with an iris shutter which opens only during the data
acquisition. During the data acquisition time of c. 80s the measured heating of the painting surface was 4-5 C.
The user can choose between various forms of displaying the measurement results. Among others there is the display of
the measured brightness temperature at a chosen frequency or the difference between the brightness temperatures at two
different frequencies. Furthermore the correlation between the measured spectra and reference data and the results of the
automatic identification routines can be displayed. In this work, the hyperspectral images were processed using a linear
filter with a Gau kernel of 3x3 pixels with a width of 3 pixels14.
The HI90 has overall dimensions of 550x455x210 mm, and a mass of 33 kg. The device is mounted on a tripod and can
be positioned with a motorized and automatically controlled pan-tilt head. In figure 1 the HI90 system is reported
working on the Burris painting Sestante 10 during the in situ measurements at the Ex-Seccatoi del Tabacco (Perugia,
Italy).
Figure 1. Picture of the HI90 imaging device and Alpha-R portable spectrometer working in situ on the Burris painting Sestante 10 at
the Ex-Seccatori del Tabacco (Perugia, Italy).
2.2 Portable Alpha-R spectrometer
The portable infrared spectrophotometer ALPHA-R Bruker Optics is equipped with a Globar radiation source, a
modified Michelson interferometer to work in all spatial conditions (RockSolidTM) and a DLaTGS detector. The working
optical layout for reflection measurements is 22/22. Its weight is about 7 kg and its dimensions are 20x30x12 cm3. 200
interferograms in the spectral range 7500-375 cm-1 with a spectral resolution of 2 cm-1 were acquired. The sampling area
was 28 mm2. Reflection spectra were directly compared with the brightness temperature spectra collected by the HI90. A
background correction using a reference spectrum from a gold flat mirror was applied for representing the reflectance
profile (R) expressed in the graphs as pseudo absorbance, log (1/R).
For the preparation of the laboratory model a cotton canvas primed with calcium carbonate and an acrylic medium has
been purchased by Zecchi. Different pigments in combination with four organic binders (whole egg, acrylic resin -CTS,
vinyl resin- CTS and alkyd resin- Gamblin) have been painted in separate square sectors (ca. 2x2 cm2). Three different
types of yellow, white and blue pigments have been considered on the basis of their infrared vibrational properties and
their occurrence on work of arts. The following pigments have been applied: lead chromate (Sigma-Aldrich), yellow
ochre (Zecchi) and PY173 (Kremer) for the yellow sectors, barium white (Zecchi), gypsum (Carlo Erba) and lead white
(Sigma Aldrich) as white pigments and ultramarine blue (Zecchi), azurite (Zecchi) and phtalocyanine blue (Kremer) as
the blue ones.
3. RESULTS
3.1 Laboratory models
Figure 2a depicts the mid-FTIR spectra recorded by the portable mid-FTIR spectrometer Alpha-R from the yellow, white
and blue pigments painted with egg tempera in the laboratory model. The spectral range achievable by this device (7000-
400 cm-1) allows the visualization of both the fundamental, combination, and overtone modes of the investigated
pigments9. The non invasive identification of the painting materials by single point infrared reflection spectroscopy is
generally achieved by combining all these spectral data observable in the wide range4,10.
In figure 2a, the black rectangle highlights the spectral range accessible by the HI90 imaging device, namely 1300-860
cm-1. In this range, most of investigated pigments show a number of infrared bands, some of which are very intense as
a) yellow
y2
uh
y1
m
wightness t :emperatun
!.
(1 /R)
uopo: Tg
;.,,
a) b) c) d)
o 7 o 1.8 o i
g) h)
Y3c
Figure 3. a) Visible image of the laboratory model in correspondence of the yellow pigments (imaged area dimensions ca. 12x6 cm2).
False color representation of the difference between: b) the mean brightness temperature (K) between 1225 and 1098 cm-1 and the
mean brightness temperature (K) between 925 and 950 cm-1, c) the mean brightness temperature (K) between 1025 and 1040 cm-1 and
the mean brightness temperature (K) between 1045 and 1050 cm-1, d) the mean brightness temperature (K) between 1303 and 1306
cm-1 and the mean brightness temperature (K) between 1309 and 1312 cm-1, e) representation of the model, labels indicate the pigment
and binder combination. Brightness temperature profiles from f) lead chromate, g) ochre, h) PY173; (the rectangle highlights the
spectral region considered for calculating the mean brightness temperature difference).
Figure 4b-d reports the chemical images representing the distribution of the white pigments (namely barium sulfate,
gypsum and lead white) mixed with the different binders in comparison with the corresponding visible photo of the
laboratory model (figure 4a). For sake of clarity, the brightness temperature spectra for each sector are also presented
(figure 4 f-h). Regarding the detection of BaSO4, it resulted reliably distributed (figure 4b) when considering the
a) b) c) d)
e)
L6
g)
S
h)
0
w2a-7/
ro Wla vv-'
W1b W2b W3b
W24
W lb
Wia W21
Figure 4. a) Visible image of the laboratory model in correspondence of the white pigments (imaged area dimensions ca. 12 x 6 cm2).
False color representation of the difference between: b) the brightness temperature (K) at 1065 cm-1 and the mean brightness
temperature (K) between 1035 and 1110 cm-1, c) the mean brightness temperature (K) between 1045 and 1165 cm-1 and the mean
brightness temperature (K) between 1056 and 1070 cm-1, d) the brightness temperature (K) at 1360 cm-1 and the brightness
temperature (K) at 870 cm-1, e) representation of the model, labels indicate the pigment and binder combination. Brightness
temperature profiles from f) barium white, g) gypsum, h) lead white; (the rectangle highlights spectral region considered for
calculating the mean brightness temperature difference).
The case study related to blue pigments revealed positive results for the identification and localization of ultramarine
blue whilst not reliable data have been recorded for azurite and phtalocyanine blue. In figure 5a and b the visible and the
a) b C
MIR III
MEIN
MUM o 2
1200
wavenumber/cm-1
900
d) e)
b2a
B1a 1 B2:31 F31111
61b12b 63b1
I _J
Bic] 62c B3-c ,
-.MN I.-
b3d
b2d
63d
B1dir62d
l J 1200
wavenumber/cm-I
900 1200
wavenumber/cm-I
900
Figure 5. a) Visible image of the laboratory model in correspondence of the blue pigments (imaged area dimensions ca. 12 x 6 cm2)
false color representation of the difference between the brightness temperature (K) at 975 cm-1 and the brightness temperature (K) at
1050 cm-1, c) brightness temperature profiles from ultramarine blue (the rectangle highlights the limits considered for calculating the
mean brightness temperature difference), d) representation of the model, labels indicate the pigment and binder combination.
Brightness temperature profiles from e) azurite and f) phtalocyanine blue. Asterisks indicate the spectral features of the acrylic-b,
vinyl-c and alkyd-d binders.
The overall results have been confirmed by point measurements by the Alpha-R (figure 6g) which indicate clearly the
presence of kaolin in the blue, purple and green sectors (all of them having a brightness temperature profile with a band
at 1000cm-1, figure 6f). The point investigation confirms also the presence of gypsum in the purple whilst indicates the
presence of the calcium sulfate also in the dark green not highlighted by the imaging since the overlapping with the vinyl
binder.
Fe?,'";;!..
o
kaolin gypsum
' \ CaCO3 1
"VI
l'green ,' l
ff
.
1300 1200 1100 1000 900 3500 2800 2100 400 700
0.8 I
1 wavenumber./cm wavenumber/crri 1
Figure 6. a) Color image of a portion of the Burris painting Sestante 10, the black rectangle highlights the area investigated by the
HI90 system. Correlation images with reference spectra collected from: b) light green; c) green; d) blue; and e) purple areas, f)
brightness temperature profiles for each colored sector, g) infrared reflection spectra acquired by the Alpha-R system with the spectral
assignment from each colored sector (the rectangle highlight the portion of the range available by the HI90). Figure adapted from Ref
11.
4. CONCLUSIONS
The study here presented highlights the potential of the mid-FTIR range to create molecular image of the distribution of
pigments on painting surfaces through the use of a the novel hyperspectral imaging system HI90 based on a Michelson
interferometer equipped with a focal plane array MCT detector.
Preliminary tests on a laboratory model mimicking a modern art painting, demonstrated the feasibility of spectrally and
spatially discriminate among sulfate-, silicate-, and carbonate-based pigments also when worked with different binders.
Overall, the comparison among the brightness temperature spectra recorded by the HI90 and the reflection profiles
acquired from the same model by a portable point-analysis FTIR spectrometer demonstrated that data produced by the
imaging device are highly reliable in term of both position and profile of the infrared bands. Difficulties aroused only for
organic yellow, phtalocyanine blue and azurite, since the weak/absent infrared signals in the range accessible by the
HI90, did not allow for their positive identification. In some cases, false positive results were observed, mainly due to the
contribution of signals from the organic binders overlapping with the pigment bands.
All these shortcomings are related to the current limitation of the spectral range available (1300-860 cm-1) which was
selected for the optimal detection of gaseous compounds. On the basis of the results obtained during this feasibility study
and also considering the previous knowledge about the response of painting materials in the mid-infrared range in
reflection model, an extension of the current spectral range will be considered. The use of appropriate optics and detector
will enable to investigate wider and different spectral ranges.
REFERENCES
[1] De Nolf, W., Dik, J., Van Der Snickt, G., Wallert, A., Janssens, , K. High energy X-ray powder
diffraction for the imaging of (hidden) paintings, J. Anal. At. Spectrom. 26, 910-916 (2011).
[2] Delaney, J. K., Zeibel, J. Z., Thoury, M., Littleton, R., Palmer, M., Morales, K. M., de La Rie, E. R.,
Hoenigswald, A., Visible and Infrared Imaging Spectroscopy of Picasso's Harlequin Musician: Mapping
and Identification of Artist Materials in Situ, Appl. Spectrosc. 64, 584-594 (2010).
[3] Ricciardi, P., Delaney, J. K., Facini, M., Zeibel, J. G., Picollo, M., Lomax, S., Loew, M. Near Infrared
Reflectance Imaging Spectroscopy to Map Paint Binders In Situ on Illuminated Manuscripts, Angew.
Chem. Int. Ed. 51, 5607-5610 (2012).
[4] Miliani, C., Rosi, F., Brunetti, B. G, Sgamellotti, A.,In situ Non-invasive Study of Artworks: the
MOLAB Multi-technique Approach, Acc. Chem. Res. 43 (6), 728-738 (2010).
[5] Rosi, F., Miliani, C., Burnstock, A., Brunetti, B. G., Sgamellotti, A., Non-invasive in-situ investigations
versus micro-sampling: a comparative study on a Renoirs painting, App. Phys. A 89, 849-856 (2007).
[6] Miliani, C., Rosi, F., Borgia, I., Benedetti, P., Brunetti, B. G., Sgamellotti, A., Fiber-Optic Fourier
Transform Mid-Infrared Reflectance Spectroscopy: A Suitable Technique for in Situ Studies of Mural
Paintings, Appl. Spectrosc. 61(3), 293-299 (2007).
[7] Rosi, F., Daveri, A., Miliani, C., Verri, G., Benedetti, P., Piqu, F., Brunetti, B. G., Sgamellotti A.,Non-
invasive identification of organic materials in wall paintings by fiber optic reflectance infrared
spectroscopy: a statistical multivariate approach, Anal. Bioanal. Chem. 395 (7), 2097-2106 (2009).
[8] Miliani C., Domenici, D., Clementi, C., Presciutti, F., Rosi, F., Buti, D., Romani, A., Laurencich Minelli,
L., Sgamellotti, A., Colouring materials of pre Columbian codices: non invasive in situ spectroscopic
analysis of the Codex Cospi, J. Archaeol. Sci. 39, 672-679 (2012).
Proc. file.
of SPIE Vol. 8790 87900Q-10
The author has requested enhancement of the downloaded All in-text references underlined in blue are linked to publications on ResearchGate.