Natural products such as adhesives in oil paintings

Abstract The study and analysis of the materials employed in artistic paintings provide deeper knowledge about the history of the work of art, including restoration efforts made in the past, and the development of painting techniques through the centuries. Gas chromatography coupled to mass spectrometry is the main analytical technique employed, as it proved to be the most suitable technique for the analysis of complex mixtures, thanks to its combination of sensitivity, wide range of applicability and versatility. Further, μFT-IR technique has also been employed to get a preliminary screening of the samples taken from paintings. In this paper, the analytical protocol based on these two techniques has been applied for analysing natural terpenic resins; its performance has been tested on microsamples collected from paintings of valuable artistic interest.


Introduction
The identification of compounds employed in artistic paintings provides both art historians and restorers with valuable information for a deeper knowledge on the technique used by artists as well as for a correct planning of restoration projects. Through the knowledge of the substances used, it is possible to get important information about the history of the work of art, the development of painting techniques across the centuries, and also about old restoration efforts. In particular, the analysis of binding media is one of the most important goals of the analytical chemistry in conservation studies (Casoli & Palla 1994). Identification of binding media involves several analytical problems mainly related to very low available sample amount and the heterogeneity and very low content of organic components (Rella et al. 2006;Chiavari et al. 2008;Burger et al. 2009).
Identification of original varnishes is another serious problem that painting conservators have to face (de la Cruz-Cañizares et al. 2005;osete-Cortina & doménech-Carbó 2005;Echard et al. 2007; Thoury et al. 2007;Rogge 2014). Historically, paint varnishes have undergone changes in composition and application: Teopilus in the XI century and Cennino Cennini in the XV century described various recipes used by the artists in the preparation of paint varnishes (Colombini et al. 2000). Ageing of varnishes leads to yellowed products, which can obscure colours, craze, become very brittle, and show changes in solubility (dietemann et al. 2000;scalarone et al. 2002doménech-Carbó et al. 2006;Echard & Lavédrine 2008). Varnish represents the first layer that is removed from a painting during a restoration treatment. It is of fundamental importance for conservators to have information about the nature of the varnish, for instance its being the original one or not and the best solvent mixtures to be employed for the removal of old discoloured or otherwise degraded layers. Further, restoration treatments may also introduce resins in paintings, both on the front of the painting, as varnish or for regeneration purposes, and on its back, as wax/resin lining paste for repairing and strengthening the support (Van den Berg et al. 2000).
Both artists and restorers have largely used diterpenic resins in virtually all parts of paintings. Venice turpentine and strasbourg turpentine are very popular (Romero-noguera et al. 2008;Lattuati-derieux et al. 2010), whereas pine resins, first of all colophony, show the problem of becoming dark and brittle. nevertheless, sources suggest that pine colophony-drying oil mixtures frequently used (Gough & Mills 1974;Van den Berg et al. 2000). on the other hand, triterpenoid resins, which originate from several families of broad-leaved trees, proved to be most useful as picture varnishes (Van Aarssen et al. 1990;Papageorgiou et al. 1997;Rontani et al. 2015). Indeed, they present some advantages in comparison to diterpenic resins, for example they are usually less yellowing and more durable than many conifer resins and more readily solvent-soluble than the copals (osete- Cortina et al. 2004).
Following all these aspects, it is clear how the work of modern restorers is difficult but stimulating at the same time. They indeed need more information about the pigments, the media and the varnishes employed by artists in order to keep homogeneity with the original painting. Their work is made even harder by past restoration efforts, not always performed using a correct approach (Bocchini & Traldi 1998).
Analysis of binding media has always represented an important goal (Rella et al. 2006;daher et al. 2010;Gelzo et al. 2014) for analytical chemistry, applied to the field of human heritage conservation. These substances, mixed to pigments (like oxides, salts, etc.) and under the action of oxygen and sunlight, endure an alteration of original macromolecules giving origin to ageing process.
The study of molecular changes due to ageing processes is particularly important, because these transformations cause physical changes such as variation in solubility. The problem has to be taken into account when treatments to remove varnishes are applied (Cartoni et al. 2001(Cartoni et al. , 2003. The use of organic solvents or solvent mixtures to remove old, discoloured, or otherwise degraded, natural and synthetic resin varnishes from old masters paintings has a long tradition, probably at least 300 years (Caley 1990) and remains a standard cleaning method in use by conservators around the world (Ruhemann 1968).
Although empirical observation suggested that no damage was done by the use of organic solvents, no studies on the effects of the cleaning solvents, as now employed in the removal of varnish from old master painting, conducted under the conditions used in a conservation studio, have been published (White & Roy 1998).
Another important issue regards the markers: these compounds allow the precise identification of a resin in samples withdrawn from a painting. Indeed, to reveal the presence of a particular substance used by an artist and to plan a systematic study of ageing process, it is necessary to find out suitable markers. Analytical procedures were developed to analyse complex matrices like binding media and varnishes (Casas-Catalan et al. 2004;Piccirillo et al. 2005

Results and discussion
natural resins are secreted by a large variety of plants. They are largely composed by mono-, sesqui-, di-and tri-terpenoids. Interesting and very useful issue for classification and identification is that di-and tri-terpenoids are not found together in the same resin (Mills & White 1994). In particular, according to their composition, resins can be classified as diterpenic (having mainly three-ring compounds with C20 skeleton) or triterpenic (having mainly four rings compounds with C30 skeleton). Many compounds, both neutral and acidic, can be isolated from each resin. Acidic compounds prevail in diterpenoid resins, such as rosin (or colophony), Venice turpentine and sandarac (scalarone et al. 2003); neutral compounds are predominant in triterpenic resins, such as dammar, mastic, elemi gum (Mills & White 1977).

μFT-IR analyses of terpenic natural resins
Preliminarly, a μFT-IR analysis was performed for a more complete characterisation of the natural resins used in fine arts (Perez-Alonso et al. 2006;Katsibiri & Howe 2010;Prati et al. 2011;Matheson & McCollum 2014). For this, FT-IR spectra of the most commonly used terpenic resins were collected, in order to identify diagnostic bands. Micro FT-IR analyses were carried out by means of nicolet 510 P FT-IR spectrometer coupled to a nicolet IR microscope, mod, nicPlan, A Mercury-Cadmium-Telluride detector, cryogenically cooled, was used to examine the region between 4000 and 650 cm −1 . Each recorded spectrum is the sum of 64 scans collected at a resolution of 4 cm −1 .
Colophony and mastic resins were analysed both fresh and artificially aged, whereas dammar and elemi resins were just investigated (Russo & Avino 2012). In Table 1 of the supplementary Material, characteristic absorption bands for terpenic resins are reported (derrick 1989).
In such paper, indeed, an identification of the specific resin is made on the basis of significant shifts of the six absorption bands reported in the table. The authors did not look for this kind of identification because experimental conditions were different and a study on shifts due to ageing of the absorption bands of resinous material was needed. At this purpose, the authors would like to propose the employment of μFT-IR analyses on samples collected from paintings only as a screening test to detect the presence of oils or resinous material and to focus the GC-Ms analyses in this direction (Lucero-Gómez et al. 2014).
In Figure 1 of the supplementary Material, μFT-IR spectra of colophony (a) and mastic (b) resins, both fresh and artificially aged, are reported. The ageing of the resins examined was performed by exposing specimens on glass of the resins, previously dissolved in methylene chloride, to the action of uV radiation for 22 days and then leaving the resins to ultimate the ageing process for 18 years in the dark.
In all the cases examined, it is possible to identify the characteristic absorption bands in the range of wave numbers reported in Table 1 of the supplementary Material. Also the aged samples showed the diagnostic absorption bands considered.
These results obtained on standards allow to consider μFT-IR analysis as a powerful tool for a preliminary screening test of the material collected from a painting. since μFT-IR is a not destructive technique for the sample, we could use it before GC-Ms analysis of the samples taken from paintings, minimising their loss.
In Figure 2(a) of the supplementary Material, the total ion chromatogram (TIC) of colophony (or rosin, as this resin is also called) is reported. The analysed material is sold under the name of abietic acid. In fact, as it can be seen from the results obtained, it is a mixture of compounds characteristic of pinaceae resins. Moreover, only a trace of abietic acid was recovered. The main compound turned out to be the dehydroabietic acid, an oxidate derivative of abietic acid (Rontani et al. 2015).
several compounds were identified in order to characterise this resin. Among them, it was important to point out the species that survive ageing processes, in order to find out suitable markers that allow to recognise colophony in a sample taken from an old master painting. Thus, we analysed artificially aged colophony. The artificial ageing process was performed by exposing colophony specimens on glass, previously dissolved in methylene Figure 1. Mass spectra of peaks 12 (6-dehydroabietic acid methyl ester) and 13 (1-Phenanthrenecarboxylic acid, 1,2,3,4,4a,9-hexahydro-1,4a-dimethyl-7-(1-methylethyl)-9-oxo-, methyl ester). chloride, to the action of uV radiation for 22 days and then leaving the resin to ultimate the ageing process for 18 years in the dark.
The results obtained from the analysis of aged colophony are shown in Figure 2(b) of the supplementary Material. The comparison of the results of the 'fresh' and 'aged' resin allows the identification of two unmistakable markers: dehydroabietic acid and 7-oxo-dehydroabietic acid (Rontani et al. 2015). Although most of the compounds (labels 1-13) listed in Table 1 survived the ageing process, their abundance tends to decrease, so they are almost never found in samples taken from paintings.
Compounds 12 (6-dehydroabietic acid methyl ester) and 13 (1-phenanthrenecarboxylic acid, 1,2,3,4,4a,9-hexahydro-1,4a-dimethyl-7-(1-methylethyl)-9-oxo-methyl ester) showed mass spectra characterised by the same fragmentation patterns of dehydroabietic acid and 7-oxo-dehydroabietic acid, respectively. The identification of terpenic compounds was obtained by the comparison of mass spectra contained in the library of spectra nbs75K.l of the mass selective detector HP-5972. In some situations, the identification was obtained by the comparison of mass spectra available in literature (Chang et al. 1971;shiojima et al. 1992; Van der doelen 1999; Van den Berg et al. 2000). In other situations, the identification was achieved by the direct interpretation of the mass spectra, for instance, for the compounds 12 and 13. In fact, in this case, the same principal fragments were found showing a mass two units smaller (Figure 1). For this reason, the interpretation of these mass spectra suggests that compounds 12 and 13 should have the same molecular structure of dehydroabietic acid methyl ester (dHA) and 7-oxodehydroabietic acid methyl ester (7-oxo-dHA), respectively, with an extra double bound. The attribution of the molecular structures proposed for compounds 12 and 13, through the interpretation of the mass spectra are reported in Figure 2.
The significant presence of abietane-type compounds as abietic acid in fresh pine colophony is reported in literature (Mills & White 1994; Van den Berg. et al. 2000). However, in our samples methyl abietate was always present only in very small amounts. Moreover, in aged samples it is totally absent (van der doelen, van den scalarone et al. 2005;Pitthard et al. 2006;Azémard et al. 2014;dellaportas et al. 2014;Fife et al. 2015;Lee et al. 2015).
To clarify this problem, samples of resin just collected from trees of the genera pinus were analysed. In Figure 3(a) of the supplementary Material, the chromatogram of a sample of resin of Pinus marittimus is reported. As it can be seen in such figure and in Table 1 abietic acid is indeed one of the main components, together with its isomer neoabietic acid, followed by sandaracopimaric acid and dehydroabietic acid. no trace of 7-oxo-dehydroabietic acid was found.
since the so-called colophony is the solid residual of the distillation of pine resin, it is highly probable that abietic acid does not survive the thermal process but it is converted into oxidised derivatives as dehydroabietic acid and 7-oxo-dehydroabietic acid. during ageing, this conversion becomes complete. In some cases, however, when the distillation process takes place at lower temperature, small amounts of abietic acid survive. nevertheless, at the end of the ageing process, this component disappears.
In order to conclude this overview on diterpenic resins employed in fine arts, the results obtained by GC-Ms of Canada balsam are reported.
Canada balsam is a resinous material also known as Canada turpentine. It is produced by Abies balsamea. Figure 3  TIC chromatogram. several compounds are identified (Table 1), but none of these can be considered unmistakable markers of this resin. Luckily, this kind of resin was scarcely employed in painting.
An important issue of this paper has regarded the identification and mass spectrometric fragmentation of diterpenoids. The most abundant among abietic acids of diterpenic resins, in the different stages of oxidation, are abietic acid, dehydroabietic acid (dHA), 7-oxodehydroabietic acid (7-oxo-dHA), 15-hydroxydehydroabietic acid and 15-hydroxy-7-oxo-dehydroabietic acid. some spectra, or tables with relative peak intensities of the prominent peaks, are already present in literature, mostly methyl esters (Chang et al. 1971;Van den Berg et al. 2000). other spectra have been elucidated on the basis of fragmentation patterns and relative retention times.
The mass spectra of methyl esters of diterpenic abietic acids are characterised by a typical fragmentation pattern, which involves: (a) the loss of the ester group; (b) the expulsion of a methyl group (often in the C20 position); (c) expulsion of water or methanol from the hydroxyl moieties. In Table 2, characteristic fragment ions are listed for dHA, methyl abietate and 7-oxo-dHA (5, 6, 9, respectively).
In Figure 3, the fragmentation patterns for these three compounds are illustrated in more detail, whereas Table 2 shows characteristic fragment ions for methyl 15-hydroxy-7-oxodehydroabietate and methyl 15-hydroxydehydroabietate (8 and 10 respectively).

Triterpenic natural resins
The triterpenoid resins originate from numerous genera of broad-leaved trees, mainly but not exclusively tropical. several plant families are involved, the botanical origins do not allow any clear forecast of the chemistry of the resin, as it happens for diterpenoid resins. some triterpenoid resins (cyclic isoprenoid compounds with 30 carbon atoms) as dammar and mastic contain a polymeric fraction.
Among triterpenoid resins, dammar, elemi and mastic, in particular, have proved most valuable as spirit-soluble varnish resins for paintings. This is probably because they are less yellowing than most of the conifer resins and more easily soluble than the leguminous copals. Initially, they look pale in colour and even if they yellow with time, the discolouration never results serious. on the other hand, they do present disadvantages for, in addition to yellowing, they can also craze and become matt and very brittle. dammar and Elemi resins were just discussed in a previous paper (Russo & Avino 2012): here, we would like to discuss the results for mastic resin, which is another triterpenoic resin, widely employed in painting varnishes. The results obtained by the analysis of samples of fresh mastic are shown in Figure 4(a) Table 2. characteristic fragment ions and corresponding m/z values of dehydroabietic acid methyl ester (5), abietic acid methyl ester (6), 15-hydroxydehydroabietic acid methyl ester (8), 7-oxodehydroabietic acid methyl ester (9) and 15-hydroxy-7-oxo-dehydroabietic acid methyl ester (10). of the supplementary Material, whereas in Figure 4(b) of the supplementary Material the partial TIC of an artificially aged sample is presented. The ageing (de la Rie 1988;Mills & White 1994;Halpine 1995;Koller et al. 1997; Van der doelen, 1999) was obtained following the same procedure used for colophony, Venice turpentine, Manila copal and dammar. The complete list of the identified compounds is reported in Table 1. some triterpenic compounds of mastic resin can polymerise, as it happens for dammar, and thus gas chromatography analysis of aged samples is more difficult. However, it is always possible to identify characteristic markers of this resin: methyl oleanonate (40), methyl moronate (42), methyl isomasticadienonate (43). We identified two couples of isomers, with very similar mass spectra, whose precise attribution is difficult: compounds 43 and 44; compounds 45 and 46 (see Figure 4).
About the identification and mass spectrometric fragmentation of triterpenoids, in literature mass spectra and tables with relative peak intensities of the prominent peaks of many triterpenic compounds have already been reported (shiojima et al. 1992;Papageorgiou et al. 1997;Van der doelen, van den Berg, Boon, shibayama, et al. 1998).
The mass spectrometric fragmentation patterns of the triterpenic compounds, identified as markers of the resin examined, such as methyl ursonate (a) and ursonic (b) aldehyde, methyl moronate (c) and methyl oleanonate (d), are given in Figure 5.

Analysis and identification of samples drawn from old masters paintings
The protocol above reported was applied to analyse some samples collected from old masters paintings (kindly furnished by the Istituto Centrale per il Restauro of Rome) with the aim to determine the resins originally employed by both artists and restorers. μFT-IR analyses were also performed as a preliminary screening test to detect the presence of oleo-resinous material, before GC-Ms analysis. It should be considered that the analyses are also complicated by the small quantity of material that can be taken: this does not allow the repetition of the measurements in a statistically meaningful way.
In particular, two paintings were considered, i.e. Madonna con Bambino by niccolò Rondinelli, XV century, and Madonna con Bambino e Santi by Carlo Maratta, XVII century. Before performing the resin characterisation, the ratio between palmitic and stearic acids (P/s) (Cartoni et al. 2001), important marker of the ageing process, was determined. For Rondinelli's painting the ratio is 1.5 suggesting the presence of a mixture of drying oils whereas for Maratta's painting it is 2.5 evidencing the possibility that walnut oil was employed (Cartoni et al. 2003).
The analysis of a sample extracted by solvent action on a sheet of Japanese paper, according to the usual restorers' practice, from Maratta's painting is an interesting example of the micro FTIR potentialities. The spectrum (Figure 5(b) of the supplementary Material) is characterised by the following absorption bands: 3400 cm −1 (broad) due to the oH stretching; C=o stretching occurs at 1780, and 1730 (as shoulders) due to the presence of drying oils and 1707 cm −1 (strong) can be ascribed to resins. In the fingerprint region, the following bands occur: 1460 and 1384 cm −1 (medium), 1318, 1186, 1135 cm −1 (weak). However, in this case, μFT-IR analyses did not permit to identify directly the kind of resin or drying oil employed in a painting, but allows to address the subsequent analyses in order to optimise the employment of the micro-samples, avoiding losses of scarce and precious materials. so, in Figure  5(c) of the supplementary Material, TIC of the sample collected from Maratta's painting is shown evidencing the presence of two different natural resins: a diterpenic resin of the pinacea family and a triterpenic one. Indeed, it is possible to identify the two characteristic markers of the pinaceae resin: methyl dehydroabietate and methyl 7-oxodehydroabietate and two markers of mastic resin: methyl moronate and methyl oleanonate. Moreover, the peaks of azelaic acid, palmitic acid and stearic acid allowed us to verify the usage of a drying oil (Cartoni et al. 2004): in particular, the P/s ratio (2.5) suggested the presence of walnut oil.
The presence of a diterpenic resin (probably, Venice turpentine) and a triterpenic one in the same painting is very interesting. It is probable that a mixture of drying oil and diterpenic resin was used as binding medium, and mastic resin, added during a past restoration treatment, as varnish.

Conclusion
natural resins have been employed in easel paintings both as finishing varnishes (owing to their glossy effect) and in drying oil mixtures to modify medium viscosity. As natural resins could be found in binding media as trace components, it is important to test procedures to analyse resins efficiently. The choice of the most suitable solvent to use is a serious responsibility of conservators and is obviously related to the nature and the state of degradation of the varnish layer.
Preliminarily, μFT-IR analysis has proved to be a powerful tool to obtain additional, significant and non-destructive analyses on different materials, both organic (such as binders, fibres, polymers) and inorganic (pigments and fillers for instance), in particular, meaningful information on the oleo-resinous varnish extracted from the painting and on the cellulose of the fibre at the same time can be deduced. Then, GC-Ms analyses have allowed to characterise the main natural resins used in paintings All these considerations provide clear evidence of the necessity of deep knowledge of both the art history and the chemical composition of materials originally employed. In this way, it should be easier formulating a correct diagnosis of the state of conservation of a work of art and/or restoring a painting as well.