Formation of Metal Soaps In Oil Paintings Under Variable Environmental Conditions

Formation of metal soaps in model paintings under cyclic temperature and relative humidity conditions has been studied by ATR-FTIR and compared with the standard controlled conditions generally maintained in many museums and storerooms. Three different white pigments (lead, zinc and titanium white) and two types of linseed oils (cold-pressed and alkali-rened) have been tested. The behaviour in short times (up to 6 weeks) depends on the pigment and environmental conditions. The relative humidity conditions have been shown to inuence the metal soap formation in more degree than the temperature. Ageing under low cyclic relative humidity (30-50%) promotes metal soap formation in both lead and zinc white pigments, compared to high relative humidity (50-90%) conditions. Besides, the formation of zinc soaps continues for 6 weeks when low cyclic relative humidity is used, but it stops after 3 weeks under the other ageing conditions studied.


Introduction
In the second half of the 90s, some of the damage phenomena produced in oil paintings were associated, for the rst time, with the formation of metal soaps. Since then, the harmful effect of these compounds on the structural integrity and aesthetical appearance of works of art is well known [1,2].
Metal soaps are metal carboxylates containing both fatty acids and metal cations. In traditional oil paintings, metal soaps are formed by the chemical reaction between the fatty acids of the drying oil with the metal ions (Zn, Pb, Cu, etc) of some pigments or additives. In modern paintings, these compounds are sometimes added on purpose as dispersion agents, stabilizers, and extenders [3]. Metal soaps show up in the painting in a heterogeneous way, causing different kinds of damages: protrusions, e orescence, increased transparency, etc. [4][5][6][7].
The damage phenomena due to metal soaps have been pointed out in paintings of different age, since Rembrandt (S. XVII) to Salvador Dalí (S. XX), evidencing the extension of the problem and alarming the conservation departments of many museums all over the world. It is, indeed, a problem intrinsic to the painting components, unavoidable though, in some kinds of artworks. Although some metal soap aggregations have been successfully reduced using cleaning treatments [8], great caution is needed in the choice and use of cleaning agents. Furthermore, when the damage compromises the structural integrity of the paint layers, no restoration treatment exists that can return the paintings to their original condition. Preventive conservation is, therefore, the only strategy to adopt.
Although the formation mechanism of metal soaps has been extensively studied, the correlation with the environmental conditions of conservation still needs a deeper investigation. Some studies have demonstrated that certain conditions of temperature and relative humidity favour the formation of metal soaps and their subsequent crystallization [9][10][11][12][13][14][15]. However, in most of the published works, extreme conditions of ageing were used, which are far away from the real conservation environments. Besides, the normal uctuations of the environmental parameters with time, were not taken into account.
Hence, this research was born to study the formation of metal soaps in environmental conditions that are more similar to the real ones, trying to simulate both control and uncontrol storage rooms, with the aim of establishing the best values of temperature and relative humidity to avoid or delay the formation of metal soaps in oil paintings.
Lead white, Pb(CO 3 ) 2 ·Pb(OH) 2 , zinc white, ZnO 2 , and titanium white, TiO 2 , are the most used white pigments in oil paintings. Both lead and zinc cations are known to be prone to react with the free fatty acids of the oil. Lead and zinc are indeed the most common metals involved in the formation of metal soaps, although metal soaps of potassium, calcium, copper, and aluminium have been also identi ed [5,13,[16][17][18][19][20][21]. On the other side, titanium white does not present such reactivity since it is a photocatalytic pigment that does not behave as Lewis acid or alkali, so not able to react with the drying oil [22,23].
Besides the pigments, the nature of the binder agent has been proved to in uence the metal soaps formation. Although oil paintings are the works of art where the metal soap formation has been mostly studied, a few publications also report this phenomenon with beeswax, egg yolk and resins [24,25]. In this work, two different linseed oils were used: cold-pressed linseed oil and alkali-re ned linseed oil, whose concentration of free fatty acids could be different as a consequence of the extraction/re ning processes. Alkali re ning, indeed, reduces the amounts of free fatty acids, unlike cold pressing that does not alter the oil [26].
Different oil paintings based on lead, zinc and titanium white pigments, formulated following the traditional preparation methods, have been subjected to cyclic conditions of relative humidity (RH) and temperature. The formation of metal soaps was followed in time by ATR-FTIR, focusing on the spectral

Sample preparation
Titanium and zinc white pigments were purchased from Sennelier pigments manufacturer and lead (II) carbonate basic from Panreac Quimica S.A. Two types of linseed oils from Winsor&Newton were employed: cold-pressed and alkali-re ned drying oils. All the starting materials were rst characterized using X-Ray Powder Diffraction (XRPD) for the pigments and Attenuated Total Re ectance FTIR spectroscopy (ATR-FTIR) for the drying oils.
Painting models were made reproducing traditional paints, by mixing the pigments with the proper amount of drying oil in order to achieve good rheology of the mixture. Table 1 shows the relative amount of each component used and the acronym given to each paint. Such paintings were applied as uniform layers of 120 µm thickness in inert glass supports and let dry in laboratory uncontrolled conditions for 8-12 weeks. Figure 1 shows the prepared painting models after drying (time zero, t 0 ).

Ageing
Environmental conditions with cyclic temperature (T) and relative humidity (RH) were tested in 8-hours cycles, for every sample. In addition, xed parameters of 21 ± 2 ºC and 50 ± 5 % RH, were used as reference. To study the effect of variable temperature, samples were subjected, for 4 weeks, to cycles between 21ºC and 50ºC keeping RH xed to 50%. The variable RH tests were performed in two sets: low RH (between 30 and 50%) and high RH (between 50 and 90%) keeping temperature xed to 21ºC in both cases, for 6 weeks. The formation of metal soaps was weekly tracked by means of ATR-FTIR spectroscopy. As references, samples of pigments and both kinds of linseed oil ALONE, were subjected to the same ageing conditions.

Instrumentation
Controlled ageing was carried out in two different climatic chambers: a Vötsch VCL for temperature and low humidity tests and a Vötsch Heraeus HC 2020 for the high humidity tests.
XRPD patterns were collected at room temperature using a Bragg-Brentano geometry in a Panalytical X Pert MPD instrument using Cu radiation 45kV, 40mA, in an angular range of 5-90º (2θ), step size of 0.01º and 1s counting time. Phase identi cation was conducted using the PDF-2 2002 ICDD database using the X-Pert HighScore Plus software. ATR-FTIR measurements were carried out on the external surface of the model paintings, using a Nicolet 380 spectrometer. Spectra were recorded between 4000 and 400 cm − 1 after 64 scans. In order to study the progress of the metal soap formation, all spectra were normalized to the prominent CH stretching vibration band at 2924 cm − 1 to correct for variations in the amount of sample in direct contact with the ATR crystal.
Results And Discussion 3.1. Characterization of the raw materials Figure 2 shows the XRPD patterns of the commercial pigments.   Figure 4 shows the FTIR spectra obtained from the lead white paintings (PBCP and PBAR) after four weeks of ageing under cyclic temperature conditions. In the spectra, features of lead white are indicated with asterisk and correspond to the presence of the CO 3 2− ion as described in [28]. Bands at 2852-2954 and 1736cm − 1 are due to the siccative oil as explained before. The broad band at 3446 cm − 1 due to the OH stretching mode, con rms a certain degree of hydrolysis of the triglycerides. The band due to the asymmetric stretching of the carboxylate groups (ν a COO − ) in lead carboxylates has been reported to appear as a broad band between 1510 y 1550 cm − 1 [9,13] and, as can be observed, no evidence of metal soaps formation exists after four weeks, with any of the linseed oils used. Similar results to those above mentioned were obtained in the FTIR spectra of the lead white paints stored at standard conditions, in which no metal soaps bands are visible even after six weeks.

Lead white paints
On the other hand, a small band occurs at about 1544-1548 cm − 1 attributable to lead carboxylates, after six weeks of ageing at a low cyclic relative humidity (30-50%) (see Fig. 5). Although the signal intensity of the band is quite weak, and no other characteristic bands can help to identify the type of lead soaps, its wavelength matches a band assigned to lead stearate and palmitate in previous studies [19].
The use of cold-pressed or alkali-re ned linseed oil does not suppose any difference for the reaction time of the lead white to form metal soaps, and the respective FTIR bands of the lead soaps do not show differences in intensity. However, the bands are poorly de ned and longer ageing times could help to understand possible diversity in the behaviour of the oils. The formation of lead soaps takes place only at low relative humidity values. When the lead paintings are subjected to high relative humidity cycles or constant conditions at 50% RH, the band at 1544-1548 cm − 1 is not formed even after six weeks (see Fig.   6).
Unexpectedly, at these ageing times, the low relative humidity range favours the formation of lead carboxylates over the high relative humidity conditions. However, several studies have shown how increasing values of relative humidity speed up the formation of carboxylates as a consequence of hydrolysis phenomena of drying oil [9,13,29]. As hydrolysis proceeds, an FTIR band at about 1710 cm − 1 , attributed to the carbonyl stretching ν(C = O) of the free fatty acids, increases, while the carbonyl stretching band of the ester groups decreases [11,19,22]. In the studied lead white paints, the band of the free fatty acids cannot be distinguished because it is hidden by the stretching vibration band of the ester groups at 1735 cm − 1 . Nonetheless, no decreasing trend of this latter band is appreciable over time.
Moreover, the hydroxyl stretching band ν(OH) at 3452 cm − 1 , which usually increases as the hydrolysis proceeds, does not show substantial growths over time under all ageing conditions. Therefore, no hydrolysis phenomena seem to have been favoured by any of the ageing conditions tested, and the formation of metal soaps can be attributed to the reaction of lead ions with free fatty acids already present in the oil composition and, to a lesser degree, with those released during secondary reactions of oil oxidation. For this reason, when no hydrolysis phenomena are caused, the rate of saponi cation strongly depends on the mobility of free fatty acids and metal ions. In the case of the aged lead white paints, the cyclic variations between low values of relative humidity could have favoured such mobility more than the other ageing conditions. However, the behaviour in longer times cannot be discussed. Figure 7 shows the FTIR spectra obtained at different times during the cyclic temperature ageing of the zinc white paints, produced with cold-pressed linseed oil (ZNPC) or alkali-re ned linseed oil (ZNAR), together with the reference samples kept under standard conditions (21 ± 2ºC, 50 ± 5% RH). In both cases, a broad band at 1588 cm − 1 is seen at t 0 which strongly increases in intensity after two weeks. From week 2 to week 4, no signi cant change occurs. This behaviour is similar in standard conditions and under cyclic temperature, no matter the linseed oil used. FTIR spectra of zinc carboxylates previously detected in paintings, show a broad intense band between 1570 y 1590 cm − 1 associated with the asymmetric stretching of the carboxylate groups (ν a COO − ), whose intensity increases with the saponi cation reaction [9,14,15,17]. These observations allow us to assign the band at 1588 cm − 1 to the formation of zinc soaps. FTIR spectra obtained from linseed oil or pigment alone, aged under the same conditions, do not

Zinc white paints
show any new peak between 1570 and 1590 cm − 1 , con rming that the above-mentioned band is attributable to a secondary product of the reaction between the pigment and the drying oil, as metal soaps. Such band is already present at t 0 , meaning that the formation of zinc soaps started during the drying process prior to the accelerated ageing, under indoor laboratory conditions. The ageing at a cyclic temperature between 21 and 50ºC favours the formation of zinc soaps, as the intensity of the band at (see Fig. 7). At longer times (four weeks), the intensity barely changes. However, a further increase of the band at longer times is not discarded. Figure 8 shows the FTIR spectra obtained in the zinc white paintings (ZNCP and ZNAR) under different ageing conditions of cyclic RH and standard conditions, and the development of the band at 1588 cm -1 .
In all the humidity conditions tested, the FTIR spectra of the zinc white paintings show an increase of the band at 1587 cm -1 from t 0 to the rst three weeks. Unexpectedly, this band continue increasing at longer times only at low relative humidity values, while for higher relative humidity values (50-90%) or constant standard environmental conditions (21 ± 2ºC, 50 ± 5% RH), the intensity of this band is stabilized. This trend is observed for both the cold-pressed (ZNCP) and alkali-re ned (ZNAR) linseed oils. However, when uctuant conditions are used, both under low and under high RH values, the band at 1588 cm -1 shows a slightly higher intensity in the cold-pressed linseed oil-based paints (ZNCP) than in the re ned linseed oilbased paints (ZNAR) (see Fig. 8d-8f). This observation may be related with the higher percentage of free fatty acids present in cold-pressed linseed oil than in alkali-re ned linseed oil, due to their different production processes, and consequently the greater contribution of free fatty acids in soaps formation by the former.
Crystallization of metal soaps has been reported to produce a new band at (1510-1538 cm − 1 ) in the FTIR spectra, which is not observable in the samples studied, probably due to the short ageing period used [30].
As in the case of the lead white paints, an evaluation of hydrolysis of the zinc white paints is not possible at the ageing conditions. The absorption of the free fatty acids at about 1710 cm − 1 is hidden by the wide ester group band at 1738 cm − 1 , and the latter does not show a decreasing trend as the band of the carboxylates increases in intensity. However, an appreciable grown of the hydroxyl band at about 3400 cm − 1 is observed in the FTIR spectra of the paints aged with cyclic variations of relative humidity, both at high values and at low values. The hydroxyl bands are more pronounced especially when the ageing was carried out with cyclic values of low relative humidity, con rming once again the in uence of these conditions on the ageing processes of the oil paints. For this reason, no phenomena of hydrolysis are excluded at all under cyclic humidity conditions.
In the other environmental conditions, spectral changes attributable to hydrolysis phenomena are not observed and, consequently, the formation of metal soaps can be attributed to the reaction of zinc ions with free fatty acids already present and, to a lesser extent, with those released by secondary oxidation processes of the oil. The stabilization of the bands of the carboxylates, observed after three weeks of ageing, could therefore depend on the availability of free fatty acids. When all the available free fatty acids have reacted and no more fatty acids are reachable by metal ions, or leachable by hydrolysis phenomena, metal soaps stop forming. At low cyclic values of relative humidity, the band keeps increasing after the third week, probably due to its capacity to favour the mobility of the reagents, as observed before in the lead white paints. Moreover, as mentioned above, hydrolysis phenomena are not discarded and could have helped to form metal soaps in this speci c case.

Titanium white paints
Titanium white paintings do not show any evidence of metal soaps formation at any of the environmental conditions tested. Their FTIR spectra do not show indeed any difference in the region of the stretching vibrations of the carboxylates groups (1650 − 1380 cm − 1 ) [27]. Figure 9 shows, as an example, the FTIR spectra of titanium paintings aged at low cycling RH. Similar behaviour is observed in the high RH, cyclic temperature and standard environmental conditions. These observations con rm the scant propensity of this pigment to form metal soaps, already observed in previous studies [22,23].

Conclusions
The formation of metal soaps in oil paintings has been studied by ageing model samples in controlled cyclic conditions of temperature and relative humidity. Three different white pigments (lead, zinc and titanium white) and two types of linseed oils with different production process (cold-pressed and alkalire ned) have been tested. The zinc white-based paintings formed zinc soaps under all tested conditions, both under uctuations in temperature or humidity, and under standard conditions. However, in the case of the lead white-based paints, the formation of lead carboxylates only occurs, in the studied period, at low cyclic values of relative humidity (30-50%). Zinc white-based paintings have, therefore, a greater propensity to form metal soaps than lead white-based paintings. However, the formation of lead soaps at longer ageing times is not discarded. Titanium white-based paintings do not form metal soaps in any of the environmental conditions, as expected, con rming that this pigment is unaffected by saponi cation phenomena.
During the drying period (8-12 weeks), metal soaps are already formed in the zinc paintings and continue forming during the rst 3 weeks of ageing, but only under low RH (30-50%) further increase of the carboxylate band of FTIR is observed after the sixth week of treatment.
Surprisingly, both in lead white and zinc white paintings, the formation of metal soaps was favoured especially by low cyclic values of relative humidity (30-50%). Moreover, uctuant values of temperature and humidity promote faster saponi cation than in standard conditions, con rming the importance of preserving oil paints at optimal and constant environmental conditions. Fluctuant conditions could favour the mobility of free fatty acids and metal ions, thus promoting the formation of metal soaps.
Hydrolysis phenomena were not observed, even though they are not excluded under any environmental condition. The formation of metal soaps was therefore the product of the reaction between metal ions and free fatty acids already present in oil composition or released during secondary oxidation processes. Comparing the FTIR spectra, slightly more intense bands of the carboxylates are appreciated in the coldpressed oil-based paints than in the re ned linseed oil-based paints. The different processes of pressing or re ning linseed oil can therefore modify the quantity of free fatty acids present in its composition, and consequently affect the formation of metal soaps.          FTIR spectra of titanium-based paintings (TICP and TIAR) aged under cyclic conditions of relative humidity between 30 and 50% and 21ºC.