Characterisation of metal carboxylates by Raman and infrared spectroscopy in works of art

Article abstract of DOI:10.1002/jrs.4520

This work introduces the complementary use of μ-Raman and μ-Fourier transform infrared (IR) spectroscopy for the detection of specific carbon chains and cations for the identification of metal carboxylates within oil paint microsamples. Metal carboxylates (metal soaps) form naturally when free fatty acids react with metal cations and may also be found as additives or degradation products. Twenty-two metal carboxylates were synthesised, and their spectra assembled in a reference database. Metal salts of cations commonly present in oil paintings were used, including lead, zinc, calcium, cadmium, copper and manganese. The fatty acids selected were the saturated acids palmitic (C1 6:0) and stearic (C18:0) and the polyunsaturated oleic acid (C1 8:1). Azelaic acid (C9 diacid), a product resulting from autoxidation of polyunsaturated acids, was also included. Metal carboxylates were characterised by Raman and IR spectroscopy, and their structures were confirmed by X-ray diffraction. Raman and IR spectroscopy proved to be complementary techniques for a full identification of the metal carboxylates in complex aged paint. Raman enables the differentiation of the carbon chain length in the C-C stretching region (1120-1040 cm-1), and IR distinguishes the metal cation in the COO^- stretching absorption region (1650-1380cm^-1). Principal component analysis was applied to the spectra in order to facilitate a fast and accurate method to discriminate between the different metal carboxylates and to aide in their identification. Finally, spectra from case studies were successfully projected in the principal component analysis models built, enabling a higher confidence level for the identification of copper palmitate and copper azelate in two 19th-century Portuguese oil paintings.

Full text of DOI:10.1002/jrs.4520

Research article
Received: 30 December 2013
Revised: 5 May 2014
Accepted: 15 May 2014
Published online in Wiley Online Library
(wileyonlinelibrary.com) DOI 10.1002/jrs.4520
Characterisation of metal carboxylates by Raman
and infrared spectroscopy in works of art†
Vanessa Otero,a,b Diogo Sanches,a,b Cristina Montagner,a,b
Márcia Vilarigues,a,c Leslie Carlyle,a,b João A. Lopesd and Maria J. Meloa,b*
This work introduces the complementary use of μ-Raman and μ-Fourier transform infrared (IR) spectroscopy for the detection
of specific carbon chains and cations for the identification of metal carboxylates within oil paint microsamples. Metal
carboxylates (metal soaps) form naturally when free fatty acids react with metal cations and may also be found as additives
or degradation products. Twenty-two metal carboxylates were synthesised, and their spectra assembled in a reference
database. Metal salts of cations commonly present in oil paintings were used, including lead, zinc, calcium, cadmium, copper
and manganese. The fatty acids selected were the saturated acids palmitic (C16: 0) and stearic (C18: 0) and the polyunsaturated
oleic acid (C18: 1). Azelaic acid (C9 diacid), a product resulting from autoxidation of polyunsaturated acids, was also included.
Metal carboxylates were characterised by Raman and IR spectroscopy, and their structures were confirmed by X-ray diffraction.
Raman and IR spectroscopy proved to be complementary techniques for a full identification of the metal carboxylates in complex
aged paint. Raman enables the differentiation of the carbon chain length in the C–C stretching region (1120–1040cmꢀ1), and IR
distinguishes the metal cation in the COOꢀ stretching absorption region (1650–1380cmꢀ1). Principal component analysis was
applied to the spectra in order to facilitate a fast and accurate method to discriminate between the different metal carboxylates
and to aide in their identification. Finally, spectra from case studies were successfully projected in the principal component
analysis models built, enabling a higher confidence level for the identification of copper palmitate and copper azelate in two
19th-century Portuguese oil paintings. Copyright © 2014 John Wiley & Sons, Ltd.
Additional supporting information may be found in the online version of this article at the publisher’s web site.
Keywords: oil paintings; metal carboxylates; Raman microscopy: μ-Raman; μ-Fourier transform infrared spectroscopy: μ-FTIR; conservation
Background
Introduction
Oil paint systems
This paper explores the molecular characterisation of metal
carboxylates by Raman microscopy (μ-Raman) and micro-
Fourier transform infrared spectroscopy (μ-FTIR) with the aim
of furthering their characterisation in oil paintings. Principal
component analysis (PCA) is applied to facilitate the identifi-
A drying oil is a natural product mainly composed of triglycerides,
i.e. esters of glycerol linked to three long-chain carboxylic acids
(Scheme 1). The fatty acid portion of the drying oils generally
used in oil paints is mostly composed of C18 polyunsaturated
cation of this phenomenon in oil paint layers using Raman
acids: oleic acid (C18 : 1), linoleic (C18 : 2) and linolenic (C18 : 3);
and infrared (IR) spectra. The methodology being presented
here allows results from microsamples, a particular advantage
when dealing with works of art.
This work was undertaken as part of a study of the painting
*
Correspondence to: Maria J. Melo, Departamento de Conservação e Restauro,
Faculdade de Ciências e Tecnologia, Universidade Nova de Lisboa, Monte da
Caparica, Portugal.
materials and techniques of Tomás da Anunciação (1818–1879)
and Cristino da Silva (1829–1877), which has been carried out
in the framework of the interdisciplinary project ‘Crossing
Borders: History, materials and techniques of Portuguese painters
from 1850 to 1918 (Romanticism, Naturalism and Modernism)’.
The identification of the green colorant used in the trees and
foliage of Paisagem e animais (Fig. 1(A)) by Tomás da Anunciação
and Cinco artistas em Sintra (Fig. 1(B)) by Cristino da Silva was
challenging as there were no visible green pigment particles
nor did the green arise from a mixture of blue and yellow
pigments. While copper carboxylates were found using FTIR, it
was not possible to identify the specific carbon chain using the
published reference data based on pure materials alone.[1–5]
Therefore, this work was carried out in an attempt to fully
characterise the copper carboxylates in the two oil paintings, by
combining IR and Raman microspectroscopy with PCA.
E-mail: mjm@fct.unl.pt
†
This article is part of the special issue of the Journal of Raman Spectroscopy
entitled “Raman in Art and Archaeology 2013” edited by Polonca Ropret and
Juan Manuel Madariaga.
a Departamento de Conservação
e Restauro, Faculdade de Ciências e
Tecnologia, Universidade Nova de Lisboa, Monte da Caparica, Portugal
b REQUIMTE-CQFB, Faculdade de Ciências e Tecnologia, Universidade Nova de
Lisboa, Monte da Caparica, Portugal
c
Research Unit VICARTE: Vidro e Cerâmica para as Artes, Faculdade de Ciências
e Tecnologia, Universidade Nova de Lisboa, Monte da Caparica, Portugal
d REQUIMTE, Laboratório de Química Analítica e Fisico-Química, Departamento de
Ciências Químicas, Faculdade de Farmácia, Universidade do Porto, Porto, Portugal
J. Raman Spectrosc. (2014)
Copyright © 2014 John Wiley & Sons, Ltd.

V. Otero et al.
as capable of self-repair, preventing paint failure.[15,16] The di-
acids, such as azelaic acid (C9), may also contribute to the
stabilisation of the network because of their chain building ability
in three dimensions.[15,16] Mono and dicarboxylic acids that are
not incorporated in the polyanionic network may migrate to react
with metal cations, leading to the formation of an increased
proportion of metal carboxylates (metal soaps). In some cases, metal
soaps form aggregates within the paint film (predominantly in the
preparation layers). Their formation is still not fully understood, but
it appears that aggregation depends on the paint composition
(pigments, extenders and driers), the availability of free fatty acids,
the build-up of the layers, the environmental conditions (temperature,
relative humidity, etc.) and the conservation history of the paint-
ing.[1,10,15–27] The presence of metal soap aggregates in high propor-
tions can negatively influence the appearance of oil paintings.[17,26,28]
The identification of metal soaps in oil paintings has been
achieved mainly with the use of IR spectroscopy and mass
spectrometry.[1,10,15–23] IR has been hampered by the fact that in
an actual oil paint, pigments, such as lead white, and other addi-
tives can obscure the identification of the carbon chain by IR. The
metal ion most frequently detected is lead, related to the presence
of lead white and minium,[1,15–17,21,23,24,28] followed by zinc, copper
and calcium.[19,25,27,28] Concerning the type of fatty acid, palmitic,
stearic and azelaic acids were detected by spectrometry
techniques.[16,18,22] However, specific characterisation, including
the metal ion and the carbon chain, has only been reported in a
few cases.[1,21,23–25] In the case of published results using Raman
spectroscopy, the detection of a lead soap in a cross section from
an oil painting did not clearly identify the carbon chain.[15]
Spectral databases and characterisation of metal carboxylates
Figure 1. (A) Paisagem e animais (‘Landscape and animals’) (1851,
40.5× 51.5cm) by Tomás da Anunciação and (B) Cinco artistas em Sintra
(‘Five artists in Sintra’) (1855, 87 × 129cm) by João Cristino da Silva; oil on
canvas, National Museum of Contemporary Art, Museu do Chiado, Lisbon.
In their comprehensive study,[1] Robinet and Corbeil provide
X-ray diffraction (XRD), IR and Raman data, which characterise
their synthesised pure reference compounds of saturated
palmitic and stearic fatty acids and the unsaturated oleic and
linoleic fatty acids, as well as their copper, zinc and lead soaps.
Their discussion of IR and Raman results focused on highlighting
the spectral characteristics, which could be used to differentiate
between these pure metal carboxylates. In their case studies,
however, they were not able to use FTIR to identify the specific
metal carboxylate, this was only possible using XRD; furthermore,
they did not report on the application of Raman in their case
studies. More recently, Mazzeo et al. used attenuated total reflec-
tance-FTIR to investigate degradation products in reconstructions
of oil and egg paints. For this work, they synthesised metal car-
boxylates based on palmitate with a wide range of metal ions
and a small percentage of saturated fatty acids with 12–18
carbon atoms, e.g. palmitic (C16 : 0) and stearic (C18 : 0).[6,7]
Drying of the oil predominantly results from photoxidation and
autoxidation reactions. The reactivity of the double bonds of the
unsaturated fatty acids promotes the initiation of these
reactions.[4,7–12] This process may be catalysed by the presence
of metal ions from pigments and driers (e.g. lead oxide or ace-
tate).[13,14] The end result is a complex three-dimensional
polyanionic network based on the glycerol ester with a cross-
linked fraction of carboxylate anions stabilised by metal
cations.[7–10] The paint system has been recently characterised
Scheme 1. Hydrolysis of a triglyceride composed of two unsaturated acids, (a) linolenic and (b) linoleic, and one saturated acid, (c) stearic.
wileyonlinelibrary.com/journal/jrs
Copyright © 2014 John Wiley & Sons, Ltd.
J. Raman Spectrosc. (2014)

Characterisation of metal carboxylates by Raman and infrared spectroscopy in works of art
such as lead, zinc, copper, cadmium and manganese (II and III).[3]
They did not identify the specific carbon chain associated with
the metal carboxylates formed in their reconstructions. While
their syntheses were restricted to the palmitate, their work, as
well as that of Robinet and Corbeil, was extremely important for
the methodology we used to synthesise our carboxylates.
The work presented here broadens the experimental design of
the previous studies, providing characterisation by Raman and IR
spectroscopy of a comprehensive set of metal soaps from four
different carbon chains and six metal ions. Using this data, models
were developed based on PCA, which can specifically characterise
metal soaps within actual complex aged paint systems.
μ-FTIR. Infrared analyses were carried out with a Nicolet Nexus
spectrophotometer coupled to a Continuμm microscope (15× objec-
tive) with a MCT-A detector cooled by liquid nitrogen. Spectra were
obtained in transmission mode, between 4000 and 650 cmꢀ1, resolu-
tion of 8 cmꢀ1 and 256 scans. Samples were previously compressed
using a Thermo diamond anvil compression cell. Spectra are shown
here as acquired, without corrections or any further manipulations,
except for the occasional removal of the CO2 absorption at ~2300–
2400 cmꢀ1. For each metal carboxylate, three IR spectra were ac-
quired. A total of 66 spectra were considered for the PCA models.
X-ray powder diffraction patterns were recorded on a Rigaku,
model MiniFlex II benchtop X-ray diffractometer, using
monochromatised CuKα radiation at 30 kV and 15 mA settings in the
1° < 2θ < 40° range at a scanning speed of 1°/min. The sample area
of the sample holder is approximately 2cm2. XRD was used to confirm
the structures and purity of the metal carboxylates synthesised. The
diffractograms obtained were compared with Corbeil and Robinet ref-
erence data[29] and International Centre for Diffraction Data database.
Principal component analysis[30–32] was applied to the C–C
stretching region (1120–1040 cmꢀ1) of the Raman spectra and
to the COOꢀ stretching absorption region (1650–1380 cmꢀ1) of
the IR spectra. Manual baseline correction of the Raman spectra
for the case study samples was performed for the PCA. Raman
spectra were preprocessed using baseline weighted least squared
(second-order polynomial) and followed by standard normal
variate (SNV) and mean centring. The purpose of using SNV was
to ensure that all spectra have equal variance.
Experimental section
Materials and methods
Synthesis of metal carboxylates
All reagents and solvents were of analytical grade, and Millipore wa-
ter was used. The following metal salts were used: calcium chloride
dihydrate, manganese(II) chloride tetrahydrate, cadmium chloride
hydrate, zinc chloride, copper acetate monohydrate and lead nitrate.
The carboxylic acids used were palmitic, stearic, oleic and azelaic acid.
The synthesis of metal carboxylates was adapted from the
methods published in the study of Corbeil[1] and Mazzeo
et al.[3] Metal carboxylates were prepared by adding an aqueous
solution of the metal salt (0.4 M) to an aqueous solution of the
carboxylic acid (0.2 M) at 70 °C, with continuous stirring (for 20–
30 min after mixing). In some cases, sodium hydroxide was added
to facilitate precipitation. The precipitate was filtered and washed
with warm water, ethanol and ether. A total of 22 metal carboxylates
were synthesised. Manganese(II) and cadmium oleate did not precip-
itate in sufficient quantities to allow their characterisation by XRD.
All IR spectra were preprocessed with a Savitzky–Golay filter
with 15-point window size, second-order polynomial and first
derivative, followed by SNV and mean centring.
Results and discussion
Characterisation of fatty acids by Raman and IR spectroscopy
Actual paint samples
This work investigated three classes of fatty acids: the saturated
palmitic (C16 : 0) and stearic (C18 : 0) acids, the unsaturated oleic
(C18 : 1) acid and the dicarboxylic acid, azelaic (C9 di) acid (Fig. 2).
As shown in the literature, in particular by Robinet and Corbeil,[1]
these fatty acids show distinctive spectral features. The respective Ra-
man and IR spectra between 1800 and 650 cmꢀ1 are presented in
Fig. 2. The full IR spectra between 4000 and 650 cmꢀ1 of the fatty
acids may be seen in Fig. S1. Spectral assignments are listed from Ta-
bles S1 to S8 and were made in accordance with the literature.[1,5,33,34]
As it is well known,[1,18,35] a distinction between the saturated acids
is possible by IR because of the CH2 wagging (out-of-plane bending)
modes coupled with the carboxyl vibrations between 1320 and
1180 cmꢀ1. In this region, palmitic and stearic acid present seven
and eight bands, respectively (Table S5). As pointed out by Robinet
and Corbeil,[1] the Raman spectra are almost identical with the excep-
tion of a low-intensity band between 1110 and 1085 cmꢀ1 assigned
to C–C stretching vibration. For palmitic acid, this band is found at
1098 cmꢀ1, and for stearic acid, it is at 1104cmꢀ1 (Table S1).
Two microsamples were taken from the green paint in Paisagem e
animais (1851, by Tomás da Anunciação) and one from Cinco artistas
em Sintra (1855, by Cristino da Silva). Sampling was performed with
a microchisel from Ted Pella microtools under a Leica MZ16 stereo-
microscope (7.1× to 115× zoom range), equipped with a Leica ICD
digital camera and a fibre-optic light Leica system (Leica KI 1500
LCD). Two of the green samples, one from each painting, were
analysed without any further processing, and one from Paisagem
e animais, with an intermediate green layer, was prepared as a cross
section by embedding in Technovit 2000LC resin, dry polishing
with micromesh cloths and with a final polish using a 12 000 cloth.
Equipment and data acquisition
μ-Raman. Raman microscopy was carried out using a Labram
300 Jobin Yvon spectrometer, equipped with a HeNe laser
17 mW operating at 632.8 nm and a solid state laser operating
at 532 nm. Spectra were recorded as an extended scan. The laser
beam was focused with 50× and 100× Olympus objective lens.
The laser power at the surface of the samples was varied with
the aid of a set of neutral density filters (optical densities 0.3,
0.6 and 1). In particular, metal carboxylates were analysed using
the DPSS laser 532 nm with 20-s laser exposure time for 20 scans.
For each metal carboxylate, three Raman spectra were acquired.
In the figures spectra are shown as acquired. A total of 65 spectra
were considered for the PCA models.
As expected, the Raman and IR spectra of oleic acid are differ-
ent to those of the saturated acids. The IR spectrum presents the
olefinic C–H stretching absorption, around 3005 cmꢀ1, which is
likely to be of low diagnostic use in a painting as it will disappear
during the drying and ageing of the oil. In the Raman spectrum,
the C¼C stretching vibration appears around 1656 cmꢀ1
.
The spectral features of azelaic acid enable a straightforward differ-
entiation; contrary to the other fatty acids, and because of its shorter
hydrocarbon chain, the most intense band in the IR spectrum of
J. Raman Spectrosc. (2014)
Copyright © 2014 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/jrs

V. Otero et al.
Figure 2. (A) Raman and (B) infrared spectra between 1800–650 cmꢀ1 of (i) palmitic, (ii) stearic, (iii) oleic, and (iv) azelaic acid. The molecular structures
of the fatty acids are shown as insets.
azelaic acid is not caused by the C–H stretch but by the C¼O
The characteristic C¼O stretching band of the fatty acids around
1700 cmꢀ1 disappears, and the characteristic COOꢀ stretching
bands, asymmetric and symmetric, of the metal carboxylates
between 1600 and 1400 cmꢀ1 appears.[1] As this has been
thoroughly discussed in the work by Robinet and Corbeil, only
specific aspects concerning the discrimination between carbon
chain lengths and the metal ion, which allowed the construction
of the PCA models, will be discussed in the following sections.
stretching band around 1700 cmꢀ1. As for its Raman spectrum, the
C–C stretching profile has only two bands at 1098 and 1063 cmꢀ1
,
and the CH2 rocking vibration at 906 cmꢀ1 has a higher intensity com-
pared with the other fatty acids, probably because of the contribution
of O–H out-of-plane deformation. Also, because the azelaic acid is a di-
mer molecule with a centre of symmetry, the C¼O symmetric
stretching becomes Raman active and appears at 1645 cmꢀ1 [36]
.
In agreement with what is observed in previous work,[1] the
Raman spectral features of metal carboxylates are very similar to
their corresponding pure fatty acids. Differences are found in low-
intensity bands below 1000cmꢀ1, assigned to the CH2 rocking vi-
brations. In the region around 900 cmꢀ1, one can also expect bands
due to the O–H out-of-plane deformation. This could explain the
fact that the most important differences between fatty acids and
carboxylates are observed in this region of the Raman spectrum.
On the other hand, the IR spectral features of the metal carbox-
ylates are very distinctive from their corresponding fatty acids.
Characterisation of metal carboxylates by Raman spectroscopy
A summary of the main characteristic Raman bands of the metal
carboxylates synthesised are presented from Tables S1 to S4. Band
assignments were made in accordance with the literature.[1,5,34]
The first point to highlight from our investigation is that Raman char-
acteristic bands between the 1150 and 1000 cmꢀ1 region, assigned to
C–C stretching vibrations, allows differentiation between carbon chain
lengths both for fatty acids and their metal carboxylates (Figs. 3 and 4).
Figure 3. Raman spectra of metal palmitates and stearates.
wileyonlinelibrary.com/journal/jrs
Copyright © 2014 John Wiley & Sons, Ltd.
J. Raman Spectrosc. (2014)

Characterisation of metal carboxylates by Raman and infrared spectroscopy in works of art
Figure 4. Raman spectra of metal azelates and oleates.
As will be discussed in the PCA section, the most
successful discrimination of the fatty acids chain lengths
is based on the Raman spectral features found between
1120 and 1040 cmꢀ1. The C¼C stretching region between 1675
and 1640 cmꢀ1 allows the identification of unsaturated carbon
chains.
As previously noted, the Raman spectra of metal palmitates
and stearates are similar to their respective fatty acids (Tables
S1 and S2; Figs. 2(A) and 3). For metal stearates, the region
used for carbon chain length discrimination appears at
slightly higher wavenumbers than for the metal palmitates.[1]
Also, comparing the Raman spectra of lead palmitate and
stearate, it is possible to observe multiple low-intensity bands
between 1380 and 1180 cmꢀ1, which may be associated to
CH2 wagging modes, as observed in their IR spectra, discussed
in the following text. Lead palmitate and stearate present nine
and ten bands, respectively. This is not observed in the Raman
spectra for the other metal cations and, for this reason, was
not considered for PCA; also, these bands are difficult to observe
in actual oil paintings.
Differentiation of different metal cations coordinated to the
same chain length is also difficult because it depends on the
detection of a low-intensity band around 900–950 cmꢀ1 attrib-
uted to CH2 rocking vibrations (Table 1). Because this band rarely
Table 1. Characteristic Raman bands assigned to C–C stretching and CH2 rocking vibrations (cmꢀ1) for the metal carboxylates
Palmitates Stearates Oleates Azelates
Assignment
Calcium
Manganese(II)
Copper
Zinc
1130
m
w
m
w
m
w
m
w
m
w
m
w
m
w
m
w
m
w
m
w
m
w
m
w
1130
m
w
m
w
s
1120
1092
1064
959
w
m
ν(C–C)
1098
1062
931
1104
1062
937
1095
s
m-s
w
1062
959
w
m
ρ(CH2)
ν(C–C)
1130
1098
1064
935
1131
1105
1062
933
w
m
w
s
1097
1066
943
m
m
m
ρ(CH2)
ν(C–C)
1128
1098
1062
946
1128
1103
1062
947
1120
1085
1064
946
w
m
m
w
w
m
m
w
w
s
1091
1061
947
s
w
s
w
s
ρ(CH2)
ν(C–C)
1131
1101
1063
952
1131
1107
1063
952
1122
1097
1065
952
w
s
1103
1064
950
s
m
m
w
m
w
m
w
m
w
m
w
ρ(CH2)
ν(C–C)
Cadmium
Lead
1130
1101
1062
935
1131
1105
1063
932
1101
1064
927
s
m
m
ρ(CH2)
ν(C–C)
1130
1099
1063
925
1131
1105
1063
925
1122
1095
1062
937
w
m
m
m
1095
1064
942
s
w
m
ρ(CH2)
The bands that allow distinction between the different carbon chain lengths are highlighted in bold.
J. Raman Spectrosc. (2014)
Copyright © 2014 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/jrs

V. Otero et al.
appears in spectra from actual oil paintings, it was also not
considered for the PCA models.
Characterisation of metal carboxylates by infrared spectroscopy
In addition, in the region between 1500 and 1400 cmꢀ1
,
A summary of the main characteristic IR bands of the metal carbox-
ylates synthesised are presented in Tables S5–S8. Band assignment
was made in accordance with the literature.[1,3,10,17,18,35,37]
The CH stretching profiles of the metal carboxylates are similar
to their corresponding fatty acids (Fig. S1), and for this reason,
only the range 1800–650 cmꢀ1 is shown in Figs. 5 and 6.
there is a group of bands assigned to the CH2 bend, where
it is possible to observe small intensity changes for each
metal cation with the same carbon chain length, which again
are of low diagnostic use in actual oil paintings. As will be
discussed in the following text, under PCA results, differentia-
tion of the metal ions is more accurately achieved in the IR
spectral region between 1600 and 1400 cmꢀ1, which includes
the COOꢀ stretch and CH2 bend.
Interestingly, it was only possible to identify an M–O vibration
band in the Raman spectra for copper palmitate, stearate and
azelate at 623 cmꢀ1. For the other metal ions, no M–O vibration
band could be detected (Fig. 4).
The COOꢀ asymmetric and symmetric stretching bands between
1600 and 1400 cmꢀ1 are very characteristic for each metal cation
(Table 2 and Figs. 5 and 6). Also, in this region, differences can be
found in the CH2 bending vibrations; however, these do not allow
distinction between the different carbon chains. Consequently, as
already pointed out and contrary to what was observed in the Raman
spectra, IR features easily distinguish the metal cations. This was
Figure 5. Infrared spectra of metal palmitates and stearates.
Figure 6. Infrared spectra of metal azelates and oleates.
wileyonlinelibrary.com/journal/jrs
Copyright © 2014 John Wiley & Sons, Ltd.
J. Raman Spectrosc. (2014)

Characterisation of metal carboxylates by Raman and infrared spectroscopy in works of art
Table 2. Characteristic infrared bands assigned to COOꢀ stretching and CH2 bending vibrations (cmꢀ1) for the metal carboxylates
Palmitates
Stearates
Oleates
Azelates
Assignment
Copper
Calcium
1586
1468
1444
1421
1579
1540
1469
1434
s
1586
1468
1444
1421
1576
1540
1469
1434
s
1588
vs
w
1588
vs
νas(COOꢀ)
δ(CH2)
m
m
m
s
m
m
m
s
1464
1433
1574
1538
1468
w
s
1417
1578
1542
1466
1433
1418
1565
1434
1411
1544
s
νs(COOꢀ)
νas(COOꢀ)
vs
vs
s
m-s
m
s
s
m
m
m
δ(CH2)
m
sh
m
vs
sh
s
1431
w
νs(COOꢀ)
νas(COOꢀ)
δ(CH2)
Manganese(II)
Cadmium
Zinc
1564
1464
1419
1544
1463
1424
1539
s
1564
1466
1419
1544
1463
1424
1539
s
m
m
s
m
m
s
νs(COOꢀ)
νas(COOꢀ)
δ(CH2)
vs
m
m
s
m
m
s
1424
1556
1535
1464
s
νs(COOꢀ)
νas(COOꢀ)
1547
1527
1465
1454
1404
1504
1488
1471
1405
s
vs
vs
s
s
1464
m
1464
m
m
m
w
sh
s
δ(CH2)
1397
1538
1513
1461
m
sh
s
1397
1538
1513
1461
m
sh
s
1402
s
νs(COOꢀ)
νas(COOꢀ)
Lead
1513
vs
m
m
sh
m
δ(CH2)
1443
1406
sh
s
1418
m
1418
m
νs(COOꢀ)
The COOꢀ asymmetric stretching bands that allow distinction between the different metal cations are highlighted in bold.
confirmed also by the results obtained with PCA, which identified it
Table 3. Characteristic infrared bands assigned to COOꢀ asymmetric
stretching (cmꢀ1) for metal palmitates and atomic mass, electronega-
tivity (Pauling scale) and effective nuclear charge (Zeff; Slater’s rules)
for each of the elements studieda
as the best region for discriminating the different metal ions of a spe-
cific carbon chain, as it will discussed in more detail in the following
text.
In which concerns chain length, the same spectral features already
described for their fatty acids are found. For both azelaic acid and the
metal azelates, the most intense band of the IR spectrum is the COOꢀ
asymmetric stretching band between 1600 and 1500 cmꢀ1 (and not
the C–H stretch as for the other fatty acid chains; Table S8).
It is interesting to note that contrarily to zinc palmitate and
stearate, where the COOꢀ asymmetric stretching is a single band
at 1539 cmꢀ1, zinc azelate and oleate present a doublet at 1556/
1535 cmꢀ1 and at 1547/1527 cmꢀ1, respectively (Table 2). This
splitting may be due to a vibrational coupling involving the car-
boxylate groups coordinated to the zinc ion.[34,38,39]
Cu
Ca
Mn
Cd
Zn
Pb
νas(COOꢀ)/cmꢀ1
Atomic mass/u
Electronegativity
Zeff/nuclear
1586 1579 1564 1544 1539 1513
63.5
1.9
40.1
1.0
54.9
1.55
3.6
112.4
1.69
4.35
65.3
1.65
4.35
207.2
1.8
4.2
2.85
5.65
charge units
aData from http://nautilus.fis.uc.pt/st2.5/ and D. R. Lide (ed.), CRC
Handbook of Chemistry and Physics. 87th ed. USA: CRC Press.
2006 (consulted on the 20 March 2013).
The differences observed for the COOꢀ asymmetric stretch, for a
same fatty acid chain, are in the following order (values for metal
palmitates, Table 3): Ca2+ (1579 cmꢀ1) > Mn2+ (1564 cmꢀ1) > Cd2+
(1544 cmꢀ1) > Zn2+ (1539 cmꢀ1) > Pb2+ (1513 cmꢀ1). Copper ion
is not considered in this discussion, as contrary to the other metal
ions, it forms binuclear or polynuclear complexes.[2] Table 3 shows
that this trend is not directly correlated with ion mass but with
parameters such as electronegativity and effective nuclear charge.
We could expect that the higher the electronegativity (or Zeff), the
higher the bond strength and therefore higher absorption
frequencies for the COOꢀ stretch. The correlation observed is,
however, in the inverse order, i.e. the higher the electronegativity,
the lower the absorption frequency. This could be explained if we
considered that the bond character is changing, i.e. from a higher
degree of covalent bonding to an ionic bond. We may anticipate
that the lower the covalent character, the lower the frequency,
as observed from Ca2+ to Pb2+
.
Principal component analysis
The first assays performed using the full range of the Raman and
IR spectra were unable to discriminate metal ions and fatty acid
chains. Therefore, and based on the previous spectral analysis, a
PCA model was built considering the C–C stretching region
(1120–1040 cmꢀ1) of the Raman spectra to differentiate the
carbon chains of palmitates, stearates and azelates (Fig. S2). The
J. Raman Spectrosc. (2014)
Copyright © 2014 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/jrs

V. Otero et al.
oleate class was excluded from the PCA model because it is easily
identified by Raman due to the presence of the C¼C stretching
band in the region between 1675 and 1640 cmꢀ1. A second
PCA model using the COOꢀ stretching absorption region
(1650–1380 cmꢀ1) of the IR spectra allows discrimination of the
metal ions for a specific carbon chain (Fig. S3). The score plots
of the PCA models for palmitates, stearates, azelates and oleates
are shown in Fig. 7(B)–(E).
maximum located around 1605 cmꢀ1 (Fig. S5(A)). PC2 separates
lead and copper from the other metal cations, capturing
29.37% of the IR spectral variance. The loading of the PC2 shows
a broad band around 1535 cmꢀ1 corresponding to the inflection
point of the maximum band of the FTIR spectrum located at
1550 cmꢀ1. The loading of the PC2 in the negative range shows
two bands at 1489 and 1570 cmꢀ1. These bands correspond to
a global minimum of the spectral first derivative of lead and
copper, respectively (Fig. S5(B)). As may be seen in Fig. 7(C) and
(D), in addition to the metal palmitates under discussion, the
same discrimination is also observed for the metal stearates
and azelates. As noted earlier, this discrimination is probably
related to the type of coordination of the carboxylate group with
the metal cations and to the degree of the covalent versus ionic
bond character.
The model using the C–C stretching region (1120–1040 cmꢀ1
)
of the Raman spectra of metal carboxylates was able to discrimi-
nate between two classes of carboxylate anions: saturated and
dicarboxylate (Fig. 7(A)). The first principal component (PC1)
discriminates the saturated carbon chains (palmitates and
stearates) from the dicarboxylate azelates, capturing 68.04% of
the Raman spectra variance. The loading corresponding to PC1
(Fig. S4(A)) shows a band around 1094 cmꢀ1, which corresponds
to the most intense C–C stretching band of the azelate class
(Table 1). The differentiation between palmitates and stearates
is achieved using the third PC (PC3) capturing only 5.55% of the
Raman spectra variance. As reflected by the PC3 loading
(Fig. S4(B)), the presence of two specific and opposite bands at
1098 and 1107 cmꢀ1 allows discrimination between the two
saturated carbon chain types. As mentioned earlier, the low-
intensity C–C stretching band of stearates appears at slightly
higher wavenumbers than palmitates (Table 1).
Case studies
Two 19th-century oil paintings, Paisagem e animais (Fig. 1(A)) by
Tomás da Anunciação and Cinco artistas em Sintra (Fig. 1(B)) by
Cristino da Silva, have been thoroughly studied by a multi-analytical
approach.[40] In both paintings, the greens in the foliage were of
particular interest as it was not possible to see or identify any specific
green pigment particles in samples from these areas. The presence
of copper was detected in both paintings by micro-energy dispersive
X-ray fluorescence and scanning electron microscopy coupled with
an energy dispersive X-ray spectrometer. Arsenic is associated with
copper in the greens of the Cinco artistas em Sintra painting, which
may indicate that a copperacetoarsenite-based pigment was origi-
nally present, although no particles are currently evident. IR analysis
suggested the presence of a copper carboxylate due to the detection
Metal ions were discriminated applying PCA to the COOꢀ
stretching absorption region (1650–1380 cmꢀ1) of the IR spectra
for each specific carbon chain. The PCA model applied to the IR
spectra of metal palmitates is shown in Fig. 7(B). PC1 captures
56.19% of total spectral variance and discriminates the copper,
calcium and manganese(II) from the remaining ions. The loading
corresponding to this component shows that the discrimination
is possible because of the maximum of the first derivate of the
FTIR spectrum (centred around 1562 cmꢀ1). Lead, zinc and
cadmium are grouped because of the presence of a local
of an absorption asymmetric stretching band at 1586 cmꢀ1 [1,2]
.
The μ-sample from Cinco artistas em Sintra has a distinct Raman
spectrum (Fig. 8(b)), and the IR spectrum shows the copper carbox-
ylate absorption asymmetric stretching band at 1586 cmꢀ1 (Fig. 8
Figure 7. Score plots of the PCA model applied to (A) the C–C stretching region of the Raman spectra of the metal carboxylates; the COOꢀ stretching
absorption region of the metal (B) palmitates, (C) stearates, and (D) azelates and (E) oleates infrared spectra. ( ) Palmitates, ( ) stearates, ( ) azelates,
(
) case study Raman spectrum observed in Fig. 8(b), ( ) case study infrared spectrum observed in Fig. 8(d), and ( ) case study Raman spectrum
observed in Fig. 9(b).
wileyonlinelibrary.com/journal/jrs
Copyright © 2014 John Wiley & Sons, Ltd.
J. Raman Spectrosc. (2014)

Characterisation of metal carboxylates by Raman and infrared spectroscopy in works of art
Figure 8. (A) Raman spectra of (a) copper palmitate and (b) a green μ-sample taken from Cinco Artistas em Sintra oil painting of Cristino da Silva; (
)
chrome orange (phoenicochroite, Pb2CrO5). (B) Infrared spectra of (c) copper palmitate and (d) a green μ-sample from Cinco artistas em Sintra oil paint-
ing of Cristino da Silva. ( ) Kaolinite and ( ) barium sulfate.
(d)). Both Raman and IR spectra were projected onto the re-
spective PCA models. The projection of the spectrum onto
the Raman PCA model undoubtedly points to the presence
of palmitate (Fig. 7(A)). In this case, the model’s residual Q
statistic is low, giving high confidence to the PCA result.
The identification of the copper ion was also confirmed by
the IR PCA model (Fig. 7(B)).
Raman spectra collected from the slightly transparent green
areas observed in the cross section (Fig. S6) from Paisagem e
animais was compared with the Raman spectral database. This
suggested the presence of copper azelate (Fig. 9), also verified
by the projection onto the Raman PCA model (Fig. 7(A)).
However, the model’s residual Q statistic for this case study,
which is a measure of the amount of variation not captured by
the PCA model, is significant. It means that in the case study,
there is some variability in the Raman spectrum, which is not
accounted for in the PCA model. This decreases the confidence
level of the assignment. This nonmodelled variance is likely
related to the lower signal-to-noise ratio of these spectra when
compared with the model calibration samples.
As noted, particles of copper pigments commonly used in the
19th century, such as verdigris (copper acetate) or emerald green
(copperacetoarsenite),[2,41] were not detected in the paints. It is
possible that the copper carboxylates identified were intentionally
produced by mixing copper pigments with oil resulting in a green
glaze. van Eikema Hommes[42] discusses numerous historical
recipes for easel paintings to create green glazes using verdigris,
in which, the finely ground pigment particles are mixed with
pretreated oils.[42] It has been shown that fatty acids, such as
linoleic, react easily with copper (II) acetate, producing binuclear
or polynuclear complexes.[2,27] Another possibility, in the case of
the Cinco artistas em Sintra, where arsenic was identified, is that a
reaction took place over time between a copper containing
pigment with free carboxylic acids.[2,27] This could explain the forma-
tion of palmitate over the other fatty acids present in a drying oil.
Conclusions
Twenty-two metal carboxylates were successfully synthesised
and characterised by Raman and IR spectroscopy. These proved
to be complementary techniques: while Raman allows a clear
identification of the carbon chain length, IR enables the identifi-
cation of the metal cation coordinated to the carboxylate group.
PCA models were built with the reference spectra and used to
support the characterisation of microsamples from case studies.
Copper palmitate and copper azelate were detected in two
19th-century Portuguese oil paintings. To the best of our knowl-
edge, this was the first time μ-Raman, combined with μ-FTIR,
was used to provide an accurate characterisation of a metal soap
in an actual oil painting. In the future, we plan to introduce more
reference spectra into the PCA models and to integrate all Raman
and IR spectra into one PCA model, which will provide an
efficient tool to facilitate the full characterisation of metal soaps.
This will also contribute to a better understanding of the different
types of coordination that take place between the carboxylate
groups and the metal cations. Specific information such as this
could further our understanding of the mechanism of metal soap
formation within oil paint systems and the formation of metal
soap aggregates.
Figure 9. Raman spectra of (a) copper azelate and (b) a green area of a
cross-section from Paisagem
Anunciação (Fig. S6).
e animais oil painting of Tomás da
J. Raman Spectrosc. (2014)
Copyright © 2014 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/jrs

V. Otero et al.
[20] G. Osmond, K. Keune, J. J. Boon, AICCM Bull. 2005, 29, 37.
Acknowledgements
[21] J. J. Boon, K. Keune, J. Zucker, Microsc. Microanal. 2005, 11, 444.
[22] K. Keune, J. J. Boon, Stud. Conserv. 2007, 52, 161.
[23] M. Cotte, E. Checroun, J. Susini, P. Walter, Appl. Phys. A 2007,
89, 841.
[24] M. Spring, C. Ricci, D. Peggie, S. Kazarian, Anal. Bional. Chem. 2008,
392, 37.
[25] E. S. B. Ferreira, J. J. Boon, M. Stampanoni, F. Marone, Study of the
mechanism of formation of calcium soaps in an early 20th century
easel painting with correlative 2D and 3D microscopy, ICOM-CC Tri-
ennial 16th Conference, Lisbon, 2011.
[26] P. Noble, A. van Loon, J. J. Boon, Chemical changes in old master
paintings II: darkening due to increased transparency as a result of
metal soap formation processes, ICOM-CC Triennial 14th Conference,
The Hague, Holland, 2005.
This work has been supported by national funds through
Fundação para a Ciência e a Tecnologia (FCT) under the project
PTDC/EAT-EAT/113612/2009. We also thank FCT-MCTES for the
PhD grants of Vanessa Otero SFRH/BD/74574/2010, Cristina
Montagner SFRH/BD/66488/2009 and Diogo Sanches SFRH/BD/
65690/2009. We thank REQUIMTE for supporting the project
PEst-C/EQB/LA0006/2013. We would also like to thank the
director, Paulo Henriques, and the curator, Maria de Aires, of
the Museu Nacional de Arte Contemporânea – Museu do Chiado,
for their collaboration.
[27] K. Keune, J. J. Boon, R. Boitelle, Y. Shimadzu, Stud. Conserv. 2013, 58, 199.
[28] Y. Shimadzu, K. Keune, J. J. Boon, J. H. Townsend, K. J. van den Berg,
The effects of lead and zinc white saponification on surface appear-
ance of paint, ICOM-CC Triennial 15th Conference, New Delhi, 2008.
[29] M. Corbeil, L. Robinet, Powder Diffr. 2002, 17, 52.
[30] A. Nevin, I. Osticioli, D. Anglos, A. Burnstock, S. Cather, E. Castellucci,
Anal. Chem. 2007, 79, 6143.
[31] C. Miguel, J. A. Lopes, M. Clarke, M. J. Melo, Chemometr. Intell. Lab.
2012, 119, 32.
[32] P. Vandenabeele, L. Moens, Analyst 2003, 187, 193.
[33] J. De Gelder, K. De Gussem, P. Vandenabeele, L. Moens, J. Raman
Spectrosc. 2007, 38, 1133.
References
[1] L. Robinet, M. Corbeil, Stud. Conserv. 2003, 48, 23.
[2] M. Gunn, G. Chottard, E. Rivière, J. Girerd, J. Chottard, Stud. Conserv.
2002, 47, 12.
[3] R. Mazzeo, S. Prati, M. Quaranta, E. Joseph, E. Kendix, M. Galeotti,
M. Anal. Bional. Chem. 2008, 392, 65.
[4] R. J. Meilunas, J. G. Bentsen, A. Steinberg, Stud. Conserv. 1990, 35, 33.
[5] K. Nakamoto, Infrared and Raman Spectra of Inorganic and Coordination
Compouds Part B (6th edn), John Wiley & Sons, Hoboken, 2009.
[6] M. Lazzari, O. Chiantore, Polym. Degrad. Stabil. 1999, 65, 303.
[7] J. D. J. van den Berg, Analytical chemical studies on traditional lin-
seed oil paints. (MolArt; 6), University of Amsterdam, 2002.
[8] J. J. Boon, S. Peulvé, O. F. van den Brink, M. C. Duursma, D. Rainford,
Molecular aspects of mobile and stationary phases in ageing tem-
pera and oil paint films in Early Italian Paintings, Techniques and
Analysis Symposium, Maastricht, 1997.
[9] J. D. J. van den Berg, K. J. van der Berg, J. J. Boon, Chemical changes in cur-
ing and ageing oil paints, ICOM-CC Triennial 12th Conference, Lyon, 1999.
[10] J. van der Weerd, A. van Loon, J. J. Boon, Stud. Conserv. 2005, 50, 3.
[11] J. Mallégol, J. Gardette, J. Lemaire, J. Am. Oil Chem. Soc. 1999, 76, 967.
[12] M. Cotte, E. Checroun, J. Susini, P. Dumas, P. Tchoreloff, M. Besnard,
Ph. Walter, Talanta 2006, 70, 1136.
[34] F. A. Miller, Amides, Carboxylate Ion, and C–O Single Bonds in Course
Notes on the Interpretation of Infrared and Raman Spectra (Eds: D. W.
Mayo, F. A. Miller, R. W. Hannah), John Wiley & Sons, Hoboken, 2004
[35] D. Chapman, J. Chem. Soc., 1958, 152, 784.
[36] N. B. Colthup, L. H. Daly, S. E. Wiberley, Introduction to Infrared and
Raman Spectroscopy, Academic Press, Boston, 1990.
[37] R. W. Corkery, Artificial biomineralisation and metallic soaps, PhD
thesis in Applied Mathematics, Australian National University, 1998.
[38] M. A. Mesubi, J. Mol. Struct. 1982, 81, 61.
[39] K. S. Patel, J. A. Faniran, Spectrochim. Acta. Part A. 1975, 31, 123.
[40] D. Sanches, A. Ferraz, T. Vitorino, L. Carlyle, M. Vilarigues, R. Macedo,
M. J. Melo, A técnica e a cor do romantismo pelas mãos de Tomás
de Anunciação in IV Congresso de Historia da Arte Portuguesa,
Fundação Calouste Gulbenkian, Lisboa, 21 a 24 de Novembro de
2012, in press.
[41] L. Carlyle, The Artist’s Assistant: Oil Painting Instruction Manuals and
Handbooks in Britain 1800–1900 with Reference to Selected
Eighteenth-Century Sources, Archetype Publications, London, 2001.
[42] M. van Eikema Hommes, Changing Pictures: Discoloration in
15th–17th Century Oil Paintings, Archetype Publications, London,
2004.
[13] Z. Osawa, Polym. Degrad. Stabil. 1988, 20, 203.
[14] C. S. Tumosa, M. F. Mecklenburg, Rev. Conserv. 2005, 6, 39.
[15] J. J. Boon, in Reporting Highlights of the De Mayerne Programme (Eds:
J. J. Boon, E. S. B. Ferreira), Netherlands Organisation for Scientific
Research (NWO), The Hague, The Netherlands, 2006, pp. 21–32.
[16] J. J. Boon, F. G. Hoogland, K. Keune, Chemical processes in aged oil
paints affecting metal soap migration and aggregation, AIC paintings
specialty group postprints, Washington, 2007.
[17] J.J. Boon, J. van der Weerd, K. Keune, P. Noble, J. Wadum, Mechanical
and chemical changes in Old Master paintings: dissolution, metal
soap formation and remineralization processes in lead pigmented
ground/intermediate paint layers of 17th century paintings, ICOM-
CC Triennial 13th Conference, Rio de Janeiro, 2002.
Supporting information
[18] M. J. Plater, B. De Silva, T. Gelbrich, M. B. Hursthouse, C. L. Higgitt, D.
R. Saunders, Polyhedron 2003, 22, 3171.
[19] J. van der Weerd, M. Geldof, L. S. van der Loeff, R. M. A. Heeren, J. J.
Boon, Z. Kunsttechnol. Konserv. 2004, 17, 407.
Additional supporting information may be found in the online
version of this article at the publisher’s web site.
wileyonlinelibrary.com/journal/jrs
Copyright © 2014 John Wiley & Sons, Ltd.
J. Raman Spectrosc. (2014)