The cleaning of paintings: effects of organic solvents on oil paint films
Alan Phenix and Ken Sutherland
Abstract
The paper is a review of technical studies of the effects of solvents on oil paints in the context of the
removal of varnish from paintings. After an introduction describing the historical background to
technical studies of cleaning, the various effects of solvents on oil paints are discussed: swelling
and softening of the paint binder (which can contribute to the vulnerability of paints to pigment loss
during cleaning); solvent diffusion and retention; and leaching (i.e. the extraction of soluble
organic compounds from the paint). The paper closes with a discussion of methodological issues in
cleaning studies, particularly the relationship between studies on 'model' reference paint films and
realistic, 'clinical' studies of actual cleaning operations, considering also the related issue of the
ageing of oil paints.
Foreword
The first issue of Reviews in Conservation included a survey by Khandekar [1] of the use of
aqueous gel formulations for cleaning painted and varnished surfaces. These kinds of formulations,
which include as their active ingredients detergents, resin- or bile-acid soaps and enzymes, as well
as solvents in gelled form, are relatively recent additions to the range of chemical cleaning agents
available to the conservator. Their introduction and development owes much to the work of
Wolbers [2, 3]. Khandekar's review inevitably refers to some of the concerns in cleaning with the
more traditional organic solvents. The purpose of the present review is to examine these, and related
issues in more detail. In some respects the corpus of technical literature relating to the cleaning of
paintings with organic solvents is quite small and there is a real need for continued research in this
area in order to enlarge the body of fundamental data.
There have been reviews of the technical aspects of the cleaning of paintings before, and the present
survey acknowledges especially the detailed and comprehensive work of Michalski [4], which is
both essential reading and a valuable reference source. There has been further work since then and
we hope to incorporate this more recent information into the general framework of understanding. It
is perhaps important to be clear that the focus here is only on the technical and scientific aspects of
the cleaning of paintings. It goes without saying that this complex subject has many other
dimensions, of which the aesthetic and historical legitimately have supremacy. However, the
technical aspects of cleaning with solvents, in so far as they inform an understanding of risk and
enable the conservator to control the process, should be relevant independently from aesthetic
considerations. The technical concerns are the same whatever one's aesthetic preference in cleaning.
The role of scientific investigations of cleaning is to inform, guide and improve the art and craft of
cleaning practice: to help minimize the risk to objects and maximize the benefits of pictures being
displayed to their best advantage.
This review summarizes research carried out to date into solvent effects on oil paint films. Cleaning
research has tended to focus on the use of volatile solvents and their effects on oil paint, since these
are, respectively, the most widely used cleaning agents and the most commonly used medium in
traditional painting techniques. In addition, oil is a very versatile medium, which is associated with
a wide range of methods of application, including fine surface applications - glazes, scumbles, etc. that have contributed to the popular perception of oil paintings being particularly sensitive to
solvent cleaning, as demonstrated by the recurrent controversies surrounding cleaning treatments.
This has also made it an obvious focus for research into solvent effects. Studies on related subjects,
such as solvent effects on other paint media and the use of alternative cleaning agents, will not be
addressed here, and the effects of solvents on varnish and practical techniques of varnish removal
will not be reviewed except where the results pertain to effects on paint films.
Introduction
During the Select Committee Enquiry that followed the first cleaning controversy at London's
National Gallery, Charles Eastlake, who was Keeper of the Gallery at the onset of the controversy
in 1846, was questioned on the value of scientific research into conservation issues. The questions
related to statements by Michael Faraday, who had, at the request of the Committee, carried out
some simple experiments of his own on the removability of varnish from several paintings, and had
suggested that further research on issues such as cleaning would be useful [5, pp. 373-83]. While
Eastlake agreed that 'it would be desirable to have the assistance or advice of chemists, with
reference to the restoration of pictures', he believed that scientists should be consulted rather than
employed at the Gallery, stating that it would 'complicate the machinery very much' if the
Government decided to 'appoint chemists to consider how solvents and chemical applications act
upon pictures' [5, p. 469]. His remarks are of historical interest since, had his attitude been different,
research into cleaning and solvent effects might have been taking place at the Gallery almost a
century earlier than Graham's investigations in the 1950s [6, p. 184]. They also highlight the
longstanding difficulty of integrating scientific study with the practice of cleaning pictures.
Although there has long been a recognition of the risks inherent in exposing painted surfaces to
cleaning agents, systematic scientific studies of the effects of cleaning solvents on varnish and oil
films were not attempted until Stout's investigations in the 1930s, and it was not until the 1950s and
60s that thorough studies were made of solvent action on paint films, with a view to gaining a more
precise understanding of the mechanisms and effects of cleaning1. Of particular importance were
the phenomena of swelling of dried oil films, and the leaching out of soluble organic components,
described in detail by Stolow. During the same period, the cleaning of paintings had once more
become the subject of debate and controversy, and in presenting his research Graham reasonably
observed
No claim is made that the work we are doing will settle all the controversies, or provide any new
techniques for the restorer, but at least some of the processes he uses, being better understood, will
become more controllable, and some criticisms answerable. [7]
Since these early studies, our knowledge of solvent effects has been refined by further research, and
various efforts have been made to apply the theoretical data to practical situations. More recently,
an improved understanding of drying and ageing mechanisms of oil paints has raised some
questions about the relevance of model studies carried out on relatively young, laboratory-prepared
paint films. This has been the impetus for more direct, 'clinical' studies of cleaning effects, carried
out on paintings of significant age, and has highlighted a need for a more critical reassessment of
some of the earlier data.
Early studies: before 1957
In the early twentieth century there were numerous studies in industry of the properties and
behaviour of drying oils such as linseed oil, because of the importance of these materials in
commercial coatings applications. Although more attention was paid to the effects of outdoor
exposure conditions, such as water and prolonged ultraviolet (UV) irradiation, than to the effects of
organic solvents, some of the earlier investigations anticipated research in the conservation field.
Solvent extraction of oil films was used as an aid to elucidating drying mechanisms.
In 1915 Morrell described acetone-soluble fractions of linseed and poppyseed oil polymerized at
high temperature [8]. Similar measurements were later made by Elm, using trilinolenin as the filmforming material, and extracting dried films in acetone, benzene and tetrachloromethane (carbon
tetrachloride) [9]. The extractable material in these studies corresponded to the fraction of the oil
not yet polymerized into the solidified network, and was shown to diminish during the initial drying
process. Nelson carried out stress-strain measurements on paint films, including linseed oil, and
interpreted these in terms of solid and liquid phases in the films, with loss of elasticity resulting
from a reduction of the liquid phase in the early stages of drying and oxidation [10]. Work by Long
and co-workers supported this idea, with a reduction in the proportion of acetone-soluble phase in
trilinolenin films observed over a year of drying [11]. In addition, the dried films in this study were
extracted in a range of solvents, and greater proportions of material were found to be extracted by
oxygenated solvents, relative to hydrocarbons. Payne demonstrated the influence of factors such as
thickening pre-treatments and the use of driers on the amount of solvent-extractable material in
linseed oil films dried for three or six weeks, using acetone, methanol and benzene [12].
The first systematic studies of solvent effects in relation to cleaning paintings were made by Stout
in the 1930s. His experiments sought to investigate the mechanism of varnish removal, using a
combination of solvent exposure and mild friction, but since his series of 'varnishes' included
linseed oil, some general observations could be made regarding potential effects on oil paint films
[13]. A small percentage (3-4% by weight) of film material was found to be removed from sevenmonth-old films of linseed oil by brief exposure to acetone, ethanol or 1,2-dichloroethane,
presumably by a combination of leaching and abrasion of the softened film. Naphtha and toluene
had no measurable effect. Around the same time, Laurie described simple, empirical tests
demonstrating the softening effects of vapours of a series of solvents on the paint layers of a 40year-old painting, although details of the paint composition and solvent exposure were not provided
[14].
Graham's experiments in the 1950s, like Stout's, were designed to investigate varnish removal,
although he too examined linseed oil and interpreted the results in terms of potential effects on paint
layers [7]. Solvent was observed to diffuse into and swell oil films, with a sharp boundary between
the swollen and unswollen material. Different solvents were found to diffuse at different rates, in
some cases relating to the viscosity of the solvent. The proportion of soluble matter was measured
for a range of films comprising linseed stand oil and various linseed oil/copal and linseed oil/mastic
mixtures. The proportions of resin are not reported. For six-month-old linseed stand oil films, 4044% by weight was found to be extractable in ethanol, acetone and benzene. The potential for
embrittlement of paint films as a result of loss of this plasticizing, soluble material was cautiously
noted. Swelling of oil films - calculated from the volume of solvent imbibed per unit volume of
(stand) oil -was measured for a variety of solvents. Chloroform and cyclohexanone were found to
be among the most active pure solvents. The most significant observations were made for solvent
mixtures. The swelling effect of certain binary mixtures was found to be greater than that expected,
assuming a linear relationship between the fractional composition of the solvents and the degree of
swelling, as illustrated in Figure la. This finding cast doubt on the then common practice of using
restrainers in cleaning - in which an 'inactive' solvent, such as turpentine, is used to dampen the
cleaning action of an 'active' solvent, such as ethanol - since the swelling action of the resultant
solvent mixture may actually be greater than either of the solvents alone. The very high swelling
powers found by Graham of mixtures of an alcohol (ethanol) and a chlorinated solvent (chloroform)
are consistent with what is known of lipid solubility in the biological field [15].
Further observations on swelling and leaching of pigmented and unpigmented linseed oil films, in
organic solvents and water, were made by Browne in 1956 [16]. These were oriented to outdoor
coating applications: some of the films were 'weathered' by exposure to filtered UV radiation and
spraying with water. Paint films were generally found to have increased density after solvent
exposure and drying, and some of the films also became porous after solvent treatment. Volumetric
and linear swelling were measured, and the results were interpreted in terms of polarity of paint
films and solvents. Certain pigments (titanium and zinc oxides) caused increased swelling of paint
films in water relative to the organic solvents, and Browne suggested this was due to the more
hydrophilic nature of paints made with these pigments. A limited range of
1
In this context, cleaning refers to the removal of discoloured and deteriorated varnish, but in a
broader sense the term also encompasses removal of overpaint, surface dirt or any other unwanted
coating or deposit.
Fig. 1
Swelling of oil films in binary solvent mixtures:
a) swelling of an unidentified oil varnish in mixtures of turpentine and an unidentified polar
solvent (acetone or ethanol?) plotted from data in Graham [7];
b) swelling of a linseed stand oil film in mixtures of white spirit and acetone, from data in Stolow
[21];
c) swelling of linseed stand oil film in mixtures of benzene and methanol, from data in Stolow
[21].
solvents was tested: water, ethanol, n-butanol, dioxane, benzene and mineral spirits. Of these
solvents, dioxane had the strongest swelling effect in all cases. The paint films swelled to a much
lesser degree than unpigmented films of linseed oil, and accelerated ageing was found to reduce the
overall magnitude of swelling, but not significantly alter the relative swelling powers of the solvents
tested.
Swelling and leaching: principal risks in the cleaning of paintings with organic solvents
The investigations by Stolow in the 1950s-70s built substantially on the concepts identified by Stout
and Graham, providing detailed descriptions of swelling and leaching phenomena, and explaining
some of the important factors determining the response of paint films to solvent [17-20]. The
compilation of Stolow's findings remains perhaps the most important source that describes the
action of solvents on oil paints [21]. It is convenient to examine what is known of these two aspects
of risk separately, although in reality they are to some degree interrelated.
Swelling of oil paints
The findings of Stolow
From the point of view of the practical conservator, the most directly useful parts of Stolow's
research are his investigations of the swelling power of various organic solvents on reference paint
films. Swelling is a more immediate and tangible concern in the practice of cleaning paintings. If
swollen to a significant degree due to sorption of solvent, the binding power of the paint medium is
reduced and the pigment is vulnerable to removal, for example, by the mechanical action of the
cleaning swab delivering the solvent and removing the varnish. The preliminary cleaning tests
routinely conducted by conservators, are essentially directed towards assessing and controlling the
risk presented by swelling of the paint, set against the solubility of the varnish [22]. At the most
basic level, safe cleaning practice generally revolves around trying to select a solvent that will
effectively remove the varnish, but which causes minimal swelling to the original paint, through
differential solvent action or different rates of response, or a combination of both of these factors.
The significance of Stolow's swelling work is reflected in the fact that it has formed the foundation
for model descriptions of the cleaning process for more than 40 years.
Stolow made precise measurements of vertical swelling of supported paint films using a specially
designed non-contact instrument [23]. The principle behind this method is that the swelling liquid
impinges on the sample surface from a jet. As the sample swells or contracts, the jet is retracted or
advanced to maintain a constant fluid pressure. Once calibrated, the movement of the jet tip
provides a measure of the change in sample thickness over time without actual contact with the
surface. Using this technique, Stolow followed the processes of swelling and leaching for a variety
of unpigmented and lead white-pigmented films, up to 25 years old: swelling was found to reach a
maximum within a few minutes (depending on factors such as paint film thickness and the solvent
used), after which a degree of shrinkage occurred, corresponding to the continued leaching of
soluble material from the film. After drying, the
Fig. 2 Swelling of seven-year-old stand oil film in acetone, showing swelling and leaching of the
'virgin' film (a), and re-swelling of the same, leached film (b) (reproduced from Stolow [19]).
solvent treated films were found to be denser and more compact, in common with Browne's
findings, as well as more brittle. The same paint films exposed to solvent for a second time
exhibited predominantly swelling behaviour, gradually increasing to an equilibrium value, and
without any observable shrinkage from leaching [21, pp. 55-62]. The typical pattern of swelling and
leaching is illustrated in Figure 2.
The degree of equilibrium swelling was confirmed by Stolow to be an indicator of the magnitude of
interaction between the solvent and the organic phase of the paint. A key finding was the variation
in the degree of swelling of reference paint films of lead white/stand oil up to 14 years old caused
by organic solvents in relation to their Hildebrand solubility parameter 9 [21, p. 92].2 Maximum
swelling was found to correspond roughly to the solubility parameter region 8.8 - 10 (cal/cm3)1/2 (18
- 20.5 MPa1/2). This region includes butanone (methyl ethyl ketone), cyclohexanone, 2ethoxyethanol (cellosolve) and a number of chlorinated solvents (an observation which accounts for
the use of dichloromethane [methylene chloride] as a common ingredient in commercial paint
strippers [24]). Aromatic solvents (benzene, toluene, xylene) also showed significant swelling
behaviour. Solvents far removed from this region - with 9 values under 8 (cal/cm3)1/2 (various
mineral spirits and aliphatic hydrocarbons) or over 13 (cal/cm3)1/2 (methanol, water) - caused very
little swelling. Although absolute swelling values varied according to the age of the paint film, the
overall pattern, and the 'peak' swelling region, were generally the same. However, a slight shift to
higher polarity could be discerned for Stolow's oldest film. Results for two paint films, aged 27
weeks and seven years, are shown in Figure 3.
This treatment of the data helped to explain Graham's results from binary solvent mixtures: a
mixture of two solvents from either side of the peak swelling region, such as turpentine and ethanol,
will have a greater swelling effect than either of the pure solvents, since the solubility parameter of
the mixture
Fig. 3 Solubility parameter 8 (in MPa)'/!
Equilibrium swelling of lead white stand oil films in solvents, plotted as a function of Hildebrand
solubility parameter, 9, for films aged: a) 27 weeks; b) 7 years. The figure is re-drawn from data in
Stolow [19, 21] but using currently accepted values of 3 taken mostly from Marcus [59] and Barton
[60], will be intermediate between those of the individual solvents (Fig. 1). If both solvents are on
the same side of the swelling region however, as Graham observed for benzene/chloroform and
acetone/water mixtures, this effect will not be observed. Stolow examined a number of solvent
mixtures in this way [18, pp. 400-1]. He also suggested that, using his data, a solvent's swelling
properties could to some degree be predicted from its solubility parameter - this concept is
significant for cleaning practice and will be returned to below.
The swelling data were further interpreted by Stolow in terms of diffusion rates of the solvents, by
comparing the times taken for swelling to reach completion [21, pp. 95-105]. Solvents exhibited
trends according to their molecular volume and viscosity: those of low viscosity, such as acetone
and benzene, produced much more rapid swelling than more viscous solvents such as iso-butyl
alcohol. Solvents with low molecular volume, such as methanol, also showed rapid diffusion and
swelling. Unsupported paint samples exhibited more rapid swelling than supported films, as a result
of the greater surface area available for solvent penetration [21, p. 56]. Issues of diffusion and
retention of solvents will be discussed in more detail in a separate section.
Graphic presentation of swelling and solubility data
The application of Stolow's oil paint swelling data to cleaning practice owes much to later
interpretations using more sophisticated treatments of the solubility parameter element, in particular
the Teas chart, for better graphic presentation of solubility data [25]. These graphic interpretations
have been valuable and influential contributions to the technical and theoretical foundations of
solvent cleaning [26-8]. Of particular importance is the paper by Hedley (1980), who took selected
2
Current practice for the units of Hildebrand solubility parameter 3 uses the S.I. convention,
MPa1/2, rather than (cal/cm3)1/2 as used originally by Stolow. To convert from (cal/cm3)1/2 to MPa1/2,
multiply the former value by 2.0455.
swelling data of Stolow's and combined this with the empirical approach of solvent cleaning tests to
develop a broad model for solvent selection in varnish removal [29]. Hedley used this approach to
rationalize certain applications of Ruhemann's 'Safety Margin Test', which, in the form Ruhemann
presented, had the drawback of assuming a more linear relationship between the concentration of
'active' (i.e. polar) solvent in a mixture and its swelling/solubility properties than is probably the
case [22, pp. 192, 308-13]. As discussed above, and illustrated in Figure 1, such linearity is not
always followed. Using the Teas fractional solubility parameter chart, Hedley identified for oil paint
a 'peak swelling region', which represented a nominal zone of solvent power associated with
increased risk in varnish removal/cleaning (Fig. 4). For the common cleaning situation of a natural
resin varnish (moderately oxidized dammar or mastic) on oil paint, the conservator could effectively
use the Teas chart as a guide to map the respective solubility and swelling regions of varnish and
paint, and to locate regions where these do not coincide. Often, by identifying the boundary of
varnish solubility on the non-polar side, a solvent or solvent mixture can be found that is an
effective solvent for the varnish, but a relatively low sweller for the oil paint (Case A, Fig. 4). The
risks might be expected to increase as the varnish becomes more oxidized, such that the boundary
of solubility on the non-polar side corresponds with the zone of maximal swelling (Case B).
This approach was developed further by Michalski [4] who combined the swelling data of Stolow,
Graham, Browne, and Eissler and Princen [30] and re-introduced into the model an indicator of
solvent power neglected by the Teas system, namely the overall solubility parameter, 3.
Interestingly, the swelling results obtained by these various researchers using different experimental
methods present a broadly consistent picture, although the range of solvents covered is rather
limited in some cases.
However, the Teas chart approach is not without limitations. The Teas fractional solubility
parameter system itself has certain shortcomings that are now well established [4, 31, 32]. Phenix
described more recent systems that might offer better discrimination of factors, such as hydrogen
bonding, polarity
Fig. 4 Use of the Teas Chart for visualizing solvent action in cleaning (from Hedley [29]). See text
for explanation.
and acidity/basicity [32].3 The more critical constraints on the Hedley/Stolow model, however, are
those that derive from the reliance on young lead white/stand oil swelling data: these issues have
been discussed recently by Phenix [33].
Re-evaluating the Hedley/Stolow 'peak swelling region'
The 'peak swelling region' described by Hedley was based only on Stolow's swelling data for paint
films composed of lead white in linseed stand oil, all younger than 15 years old. Both their
comparative youth and the choice of stand oil as binder suggest that Stolow's data might
significantly under-estimate the response of older paint films to the more polar solvents, particularly
those that are strongly dipolar and/or engage in hydrogen bonding.
According to Stolow's data on the lead white/stand oil films, ethanol and methanol, for example, are
low-swelling solvents comparable in effect with the aliphatic hydrocarbons. In Stolow's doctoral
thesis and his publications outside conservation, data is presented on the swelling of other linseed
oil film types as a function of solvent solubility parameter, [17, pp. 199-200a; 18, p. 399]. These
data - which were not taken into consideration in Hedley's treatment - reveal the anomalous
behaviour of the stand oil film, with all the other films showing appreciable degrees of swelling in
the polar solvents, including lower alcohols. Stolow acknowledged that 'stand oil films behaved
somewhat differently from non-stand oil films' [17, p. 201]. The different swelling behaviour of
stand oil compared to other linseed oil films is shown schematically in Figure 5. This difference
almost certainly derives from the internal chemistry of the paint films - dried films of stand oil are
known to be considerably less oxygenated than those of blown or boiled oils, or linseed oils dried
without pre-thickening treatments. Solvents capable of causing significant degrees of
solubility parameter, 5 (cal/cm3)1/2
Fig. 5 Schematic representation of different swelling behaviour of stand oil films compared to
other linseed oil films, from data in Stolow, 1955 and 1957 [17,18].
3 Natural
resins are known to become increasingly acidic on ageing: acid/base interactions of
solvents are not described in the Teas chart.
swelling to oil paint films are spread over a broader range in the Teas Chart than is indicated by the
Hedley/Stolow model [32]. For example, strongly dipolar solvents such as N,N-dimethylformamide
(DMF) or N-methyl-2-pyrrolidone (NMP) are known by practical conservators to have high
swelling effects. Teas solubility parameters of such solvents place them removed from Hedley's
peak swelling region (Fig. 4).
Phenix [33] re-examined the swelling behaviour of non-stand oil paint films using a microscopical
method [34] derived from that of Eissler and Princen. For two 'virgin' reference paint films the
respective swelling powers are reported of more than 50 individual solvents plus a series of binary
solvent mixtures containing ethanol. The test films were of proprietary artists' oil paints (a mixture
of yellow ochre and lead white pigments in linseed oil): one film was aged by exposure to high light
dosage and the other was unexposed to light. The respective swelling powers of solvents were
presented as a function of solubility parameter d and Reichardt polarity ErN [35], and solvents were
classified into five categories by swelling power (Fig. 6) and by rate of swelling action. A more
complex relationship between swelling power and individual solvent characteristics, especially
polarity, is indicated than was reflected by previous studies. Several types of polar solvent were
found to generate a high degree of swelling in the paints tested. Although the lower alcohols
produced relatively low degrees of swelling in comparison with other strongly polar solvents,
methanol and ethanol were shown to be appreciably more active on these test paints than on the
lead white/stand oil films of Stolow. The data showed better correspondence with Stolow's results
for non-stand oil films. In most cases, the light-aged paint film showed distinctly lower degrees and
slower rates of swelling than its unexposed equivalent. A decrease in equilibrium swelling values of
paint films aged up to 25 years in certain solvents had also been observed by Stolow [21, p. 93], and
it may be that this reduction in swelling potential is an ongoing trend as paint ages.
Phenix also plotted swelling results for a series of ethanol/ white spirit and ethanol/water solvent
mixtures (Fig. 7). The broad shoulder towards higher polarity has parallels in Figure 5. That the
peak swelling is caused by relatively non-polar mixtures is an indication of the still relative youth of
these paint films. For binary mixtures of ethanol and white spirit containing 10% or more ethanol,
the initial rate of swelling was effectively the same and comparable with that for pure ethanol. This
result suggests that the rate of swelling and probably also the rate of penetration of a binary solvent
mixture are governed by diffusion of the more mobile component.
Leaching: the extraction of soluble organic compounds during cleaning
The risks associated with leaching have proved to be more contentious in the technical
considerations of cleaning with solvents, in a large part because the process of leaching - in contrast
to swelling - is less immediately tangible to the conservator in the course of cleaning. In addition,
the potential risks are longer-term than the more direct risks of physical damage associated with
swelling and softening of paint. The primary risks identified under experimental conditions are
embrittlement of the paint film, a consequence of the extraction of organic components that have
plasticizing properties in the
Fig. 6 Compilation of swelling data from Phenix [33]. Paint type - artists' oil paints containing
yellow ochre and lead white pigment bound in linseed oil, not-exposed to light ageing, 140(im
thick. Maximum swelling as a function of solvent solubility parameter 3. d values mostly from
Marcus [59 ].
film, and alterations to the optical properties of the paint surface resulting from disruption of the
binding medium.
The study of leaching is therefore problematic in that it is very difficult to correlate experimental
results with empirical experience, and to assess the magnitude of risk in practical situations. This is
reflected in the fact that discussions of this issue in particular have often been polarized, with
experimental
Fig. 7 Maximum swelling of two paint types in binary solvent mixtures containing ethanol, as a
function of Teas fractional dispersion force solubility parameter, fd. (from Phenix 2002 [33]). Paint
types - artists' oil paints containing yellow ochre and lead white pigment bound in linseed oil. Film
type A: 140µm thick, unexposed to light. Film type B: 130µm thick, artificially light-aged.
results either dismissed as irrelevant or taken at face value to represent cleaning effects on
paintings4. The value of model studies of leaching lies in the identification under controlled
experimental conditions of the various factors determining the susceptibility of paint films to
leaching, but it is only recently that more realistic, 'clinical' experiments carried out on paintings which will be discussed in a later section - have begun to allow these theoretical studies to be
interpreted in a more practical context.
The solvent extractable components
Knowledge of the chemical composition of material extractable from paint films by solvent is not of
direct practical significance, but it has been important in determining relationships between
chemical reactions occurring in a drying paint film - cross linking, oxidative scission, hydrolysis,
etc. - and the formation of these soluble components.
Stolow used infra-red spectroscopy (IR) and gas chromatography (GC) to study extracts from a
number of the pigmented and unpigmented films in his studies [21, 36}. The results were
interpreted in terms of a soluble fraction composed of unpolymerized linseed oil triglycerides those containing palmitic and stearic acid, as well as the less reactive, monounsaturated oleic acid in combination with lower molecular weight oxidation products, such as dicarboxylic acids,
aldehydes and hydroperoxides. A 'non-esterifiable portion' was also described, referring to material
that was not converted to methyl esters by the GC derivatization method used, but this fraction which formed a significant percentage of many of the extracts - was not further characterized in the
study. It was also speculated that the compounds extracted could include 'low molecular weight
polymers' [21, p. 81], although results from measurements of the quantities of material extracted
using solvents of different swelling power suggested that the molecular weight range of the
extractable material was limited to the smaller molecules, such as triglycerides and oxidation
products [21, p. 69]. The limitations of the analytical techniques available at the time prevented a
more detailed characterization of the extractable components.
A study by Jones [37], who used paper- and thin layer chromatography to analyse propan-2-ol (isopropyl alcohol) extracts from unpigmented raw and stand linseed oil films, did not find evidence for
intact triglycerides in the extracts. Instead, a heterogeneous mixture of material was found, which
was interpreted generally to represent 'fragments of polymer chains' and 'by-products of the
oxidation reaction'. However, more precise identification was not possible using these techniques.
A more recent study by Erhardt and Tsang [38] included GC and gas chromatography mass
spectrometry (GCMS) analyses of solvent extracts from a range of oil paint films, of different age
and pigmentation. The major low molecular weight components in the extracts were the saturated
fatty acids, palmitic and stearic, along with small amounts of monounsaturated and dicarboxylic
acids. Unlike previous studies, this observation pointed to the role of hydrolysis in the natural
ageing of oil paint, responsible for the conversion of triglycerides in the paint film to the free fatty
acids detected in the extracts. Other components of the extracts were attributed to paint film
components other than oil, including wax, plasticizers from synthetic polymers, natural resin
components, and, in some cases, soluble organic pigments.
Further evidence of hydrolytic decomposition in paint films, made from a variety of drying oils and
pigments, and aged under different conditions, was found by Schilling and co-workers [39]. Monoand diglycerides were detected in acetone extracts from these films using GCMS, in addition to
free, saturated and dicarboxylic acids. Another study, by Tumosa and co-workers, used GCMS to
identify mono-, di- and triglycerides of palmitic and stearic acids in acetone extracts from both a
six-year-old paint film and an eighteenth-century paint sample [40].
Sutherland and Shibayama [41, 42, chapter 3] also identified a range of compounds in solvent
extracts from oil paint samples, made with various pigments and aged between three and 63 years,
as well as older samples from paintings. In addition to GC and Fourier-transform infra-red
spectroscopy (FTIR) analyses, size exclusion chromatography (SEC), liquid chromatography/mass
spectrometry (LCMS) and direct temperature resolved mass spectrometry (DTMS) were used to
characterize larger chemical components in some of the extracts. In addition to fatty acids and
mono-, di- and triglycerides, a higher molecular weight fraction was detected in extracts from
several of the model films, corresponding to partially cross-linked, oligomeric components. Extracts
from samples from paintings, dating from the seventeenth to nineteenth centuries, were found to
contain a high proportion of free fatty acids, relative to glyceride species, suggesting that hydrolysis
of the oil glyceride structure of these samples had progressed to a greater extent than in the younger,
model films.
4
For example, the blanched appearance of a painting's surface after cleaning is often attributed ro
the leaching effect of the cleaning solvent, but it cannot be determined if this is a result of the
current treatment, a cumulative affect of a number of treatments, or related to other factors such as
the natural ageing and deterioration of the binding medium.
There is now an increasing body of evidence supporting the view that hydrolysis (or deesterification) of the glyceride ester polymer network may be a significant aspect of oil paint ageing
[39,43,44].
The physical effects of leaching
The increased brittleness of paint films as a consequence of leaching was recognized in the earliest
studies of solvent action. Later investigations made more detailed, comparative measurements of
this and other physical phenomena. Changes in physical properties of paint films due to solvent
leaching were studied by McGlinchey, using differential scanning calorimetry (DSC) [45], He
analysed films of poppy, walnut and linseed oil pigmented with titanium, lead and zinc white
pigments, which had undergone accelerated ageing treatments, and found the polar solvents acetone
and ethanol produced an increase in brittleness and density, whereas cyclohexane caused no
observable changes.
The effects of solvents on mechanical properties of paint films were also investigated by Hedley
and co-workers, who used thermal mechanical and dielectric analysis techniques to analyse 12year-old oil films pigmented with lead white and with burnt sienna [46]. It is likely that these paint
films were from the same set as those studied by Erhardt and Tsang [38]. Increased stiffness was
observed as a result of exposure to acetone and propan-2-ol, which again was attributed to the
leaching out of soluble components, although water was found to cause a reduced stiffness.
Changes were found to result from swab rolling with solvents, as well as from immersion in
solvent, although the effects of swabbing were less pronounced. Surface changes were also
investigated in this study using scanning electron microscopy (SEM), with erosion observed in
some samples as a consequence of loss of organic binder. Solvent effects were less in the lead white
samples, as compared to the burnt sienna, with respect to both stiffness and surface changes. This
can be attributed to the high medium content of the burnt sienna paint, as well as the different
catalytic properties of the pigments.
More recently, similar results were obtained by Tumosa and co-workers [40], who made stressstrain measurements on oil paint films made with various pigments, aged between six and 20 years.
Increased brittleness and stiffness were observed after exposure to acetone or toluene for as little as
30 seconds. Samples tested before the solvent had completely evaporated showed a reduced
stiffness in some cases, demonstrating the softening effect of the retained solvent. Exposure to the
nonvolatile solvent triethanolamine (TEA) was found to cause long-term softening of a paint film.
The results of this study were interpreted in terms of a painting's ability to tolerate stresses from
environmental fluctuations and handling following solvent treatment, on the basis that the solventexposed films in the tests retained a degree of elasticity and plasticity; but these conclusions are
somewhat generalized given the limited amount of data presented.
Changes in thickness, colour and gloss of a variety of oil paint samples due to solvent exposure
were measured by Erhardt and Tsang [38]. In general, these paint films showed the reduction in
gloss after solvent exposure that might be expected,
as a result of surface erosion from leaching, as well as from pigment loss in some cases.
Quantitative measurements of solvent-extractable material
A number of studies have used quantitative measurements of solvent leaching from reference paint
films as a general indicator of risk in cleaning treatments. Magnitude of leaching has been
determined either from weight changes in extracted paint samples or using quantitative analysis of
extracted fatty acids.
Stolow made leaching measurements on a wider variety of linseed oil paint samples than in his
swelling studies [21, pp. 62-68]. The majority were laboratory-prepared films, aged up to 25 years,
with some exposed to additional light or thermal accelerated ageing conditions. For lead white
pigmented films and unpigmented films of stand oil and alkali refined linseed oil, the maximum
amount of leaching was generally found to be between 15 and 30% of medium content by weight
for the range of films and solvents tested. With the exception of n-hexane, the leaching effects of
the solvents tested - which included alcohols, ketones and chlorinated solvents - were found to be of
the same order. The relatively low leaching effect of n-hexane was attributed to the low swelling
power of this solvent.5 Experiments using different times of solvent exposure indicated that
leaching initially occurred rapidly, with 50% of the total soluble material extracted from lead white
pigmented and unpigmented films in acetone within 100 seconds, although the process was slow to
reach completion, taking over three days in some cases.
Higher leaching values were found for oil films containing other pigments: between 46 and 56% of
medium was extracted from films pigmented with barium sulphate using various solvents, and as
much as 84-85% was extracted from titanium dioxide and iron oxide pigmented films in
dichloromethane and methanol. These more pronounced values reflect a physical disintegration of
the paint samples in solvent, a consequence of the pigments having a photocatalytic effect on the
degradation of the oil polymer, accelerated in these cases by UV ageing.
Quantities of extractable material showed some relationship with age of the paint film, with paint
films studied over a period of 25 years showing an initial decrease in the proportion of material
extractable, followed by a gradual increase. This was explained in terms of the different reactions
occurring in the film: it was suggested that initial rapid polymerization results in a reduced quantity
of extractable material, as demonstrated in earlier studies, but that after longer periods of ageing,
progressive oxidative degradation produces increasing quantities of extractables. Unpigmented
films in particular showed a pronounced increase in quantities of extractable material after longer
ageing periods, up to 15 years. This trend was not borne out, however, by preliminary analyses of
older paint samples (examples were from a 48-year-old palette, and a seventeenth-century painting).
These gave leaching values of the same order as, or lower than, those for the younger paint films
[21, pp. 65, 85].
5
The intrinsic solubilities of the various classes of compound likely to be present in the material
extracted received little attention from Stolow. The solubilities of fatty acids, hydroxy fatty acids,
mono-, di- and tri-glycerides are known to vary significantly both in the low polarity range [δ.
ca.7.2 - 9 (cal/cm3)1/2] and at higher polarity [δ. > 11 (cal/cm3)1/2] [15]. Solubility of dicarboxylic
acids would also be expected to vary with solvent polarity.
Studies subsequent to Stolow's provided additional information on leaching phenomena, which, in a
large part, corroborated the previous research. Jones studied in detail the effects of one solvent,
propan-2-ol, on one-year-old, unpigmented stand and raw linseed oil films [37]. Increased
quantities of extractable material were found for UV irradiated films, and from this Jones postulated
that oil paint films would generally show increasing quantities of soluble material with age,
corresponding to a progressive deterioration of the binding medium.
Erhardt and Tsang [38] compared the effects of four solvents (hexane, toluene, acetone and ethanol)
on a range of pigmented oil paint films, aged between five and 50 years. Solvent extraction was
measured by weight changes and by quantities of palmitic acid in the extracts, determined by GC.
Results were generally in agreement with Stolow's data, with the aromatic and oxygenated solvents
(toluene, acetone and ethanol) producing greater leaching, and hexane having a lesser effect.
Pigmentation was shown to have pronounced effect on the quantities of material extractable, with
the greatest proportions extracted from raw sienna films, and the least from lead white films. The
difference could be attributed in this case both to differences in medium content and to the greater
extent of cross-linking in the latter, catalysed by the lead white pigment. No clear correlations were
observed between the age of the paint films and solvent effects. Different exposure times to solvent
were used, including a few measurements of quantities of fatty acids extracted by swabs, but these
data were not compared with the immersion results. The pronounced solvent sensitivity of some of
the samples used in this study should be stressed. The two raw sienna films studied here, five and
10 years old, were reported to contain respectively 37% and 29% of oil in the paint. Weight losses
for these films in acetone and ethanol approach (and in one case exceeds) these values, suggesting
almost complete solubility of the binder phase and/or else significant loss of pigment.
The same authors later published additional data [47] using a broader range of solvents - now
including also cyclohexane, dichloromethane, cellosolve, methylcellosolve, methanol and water.
Leaching values were interpreted in terms of Hansen total solubility parameter δt. Lead white and
raw sienna films were found to differ only in the overall quantities of material extractable, not in
their relative responses to different solvents. 2-ethoxyethanol (cellosolve) had the strongest leaching
effect, followed by methanol.
The same paper described experiments demonstrating that the application of a solvent-based varnish
to an oil paint film can result in a similar extraction of soluble components as is found with
exposure of the paint films to free solvent. This was shown by the detection, using GC, of fatty
acids in dried varnish films removed mechanically from paint layers. The varnish was found to act
as a poultice, with the solvent diffusing into the paint film and drawing soluble oil components up
into the varnish layer.
This idea was explored further by Sutherland [48], who used GC to measure the quantities of fatty
acids extracted from a three-year-old, umber-pigmented oil film by a series of varnish formulations.
The quantities extracted were found to depend not only on the solvent used to apply the varnish, but
also on the resin component, with varnishes formulated with more polar, oxygenated resins
(represented in the study by the chemically reduced ketone resin, MS2A (Linden Chemicals), and
dammar) extracting more than those made from a non- polar, hydrocarbon resin, Regalrez 1094
(Hercules Inc.). The significance of the phenomenon of absorption by varnish to the depletion of the
organic binder phase in older paint films is worthy of further investigation. It is becoming
increasingly apparent that leaching is only one of several possible processes that may contribute to
loss of organic binder components. Constituents of the mobile, solvent-extractable phase of oil
paint, such as free fatty acids, may be lost by efflorescence and/or volatilization [4, 49], or possibly
even by sorption into the ground layer [50]. Volatilization of low molecular weight scission
products will also contribute to loss of organic matter from the paint [39].
Quantities of fatty acids extractable from paint films made with a variety of oil types and aged
under different conditions, using four solvents, have been measured in a recent study by Schilling
and co-workers [39]. This was done primarily to investigate potential changes in fatty acid ratios
that might influence medium analyses. Significant proportions of fatty acids were extracted from
the films tested, but this did not hinder identification of the oil type.
Quantitative measurements of material extractable from a range of reference paint samples, as well
as older samples from paintings, were made by Sutherland and Shibayama, using both weight
changes and GC analysis of fatty acids in the paint samples and solvent extracts to indicate the
magnitude of leaching [41, 42, chapter 4]. The quantities of material extractable, and the rate of
extraction in solvent, were found to be related to a number of factors, including the pigmentation
and medium content of the films. Greater amounts of soluble material were present, and extraction
occurred more rapidly, in paint films with slower drying pigments and high medium content.
Considerable quantities of extractable fatty acids were detected by Sutherland in paint samples as
much as 700 years old [42, p. 96], and whereas the more polar solvents used in
Fig. 8 Proportions of palmitic acid (solid figures) and azelaic acid (hollow figures) extracted from
lead white paint samples after 24-hour immersion in acetone, expressed as percentages of total fatty
acids in samples. Films made with • = raw linseed oil, ■ = alkali refined linseed oil, ▲ = sun
thickened linseed oil, ♦ = stand linseed oil. Trendlines were added using an exponential function
(from Sutherland 2000 [42]).
the study (acetone and ethanol) were generally found to extract the greatest quantities of material
from the younger paint films, dichloromethane - a solvent of intermediate polarity - was found to
have a pronounced leaching effect on the older (thirteenth-to nineteenth-century) paint samples.
Although this is known to be a high swelling solvent, it was not thought that these samples should
be especially prone to swelling, and it was speculated that the effect could be the result of specific
solvent interactions of the dichloromethane with free fatty acids, which form an increasing
proportion of the extractable material in older paint films.
Analysis of lead white pigmented films of different ages, up to 65 years, showed a trend of
decreasing proportions of acetone-extractable material with age (Fig. 8) suggesting that ongoing
chemical changes over extended ageing periods had produced a gradual stabilization (either from
increased cross-linking, or non-covalent interactions such as ionic bonds) in these paint films. The
samples had been formulated with different oil types, as indicated in Figure 8, but this did not
appear to be a significant factor in the overall pattern. This trend was borne out by a small number
of acetone extraction measurements carried out on lead white based samples from paintings, which
showed lower levels still of acetone-extractable fatty acids. However, despite this apparent overall
stabilization of paint films with age, analysis of previously leached films indicated that additional
extractable material will continue to form in a paint film after an initial, 'exhaustive' leaching, as a
consequence of ongoing degradation processes [42, p. 93].
Solvent sorption by paint films: diffusion and retention
Diffusion
In his review of 1990, Michalski [4] made the important distinction between two forms of solvent
penetration into paint - capillary penetration and diffusion. Capillary penetration into cracks, pores
etc. should not, in itself, have much impact on a paint film: its significant lies in the potential for
providing greater surface access for diffusion to occur. Capillary action is fast. It is influenced
primarily by porosity of the paint structure and by the viscosity and surface tension of the solvent.
Paint film porosity will increase on ageing and as a consequence of leaching.
Stolow demonstrated that, in young paint films, solvent penetration occurred by simple diffusion of
the solvent through the binding medium generally following Fick's Law, at least in the early stages
of swelling. In young, medium-rich paint and unpigmented films, solvent diffusion occurs with a
relatively sharp swelling front [4, 7]. The rate of paint film swelling is quite strongly dependent on
film thickness: doubling thickness quadruples the half-time to maximal swelling. Pigmentation does
not significantly slow down the diffusion process. Capillary penetration will cause many swelling
fronts, so speeding the overall swelling process. Stolow and Michalski essentially distinguished
solvents by virtue of their diffusion coefficients through (young) oil paint. Michalski defined three
broad groups: slow, medium and fast diffusers.
Retention
The concentration of solvent in a paint film will also be governed by the speed of diffusion of the
solvent outwards, as in drying or 'de-swelling'. Rapid diffusion (combined with rapid varnish
dissolution) is often exploited by restorers to prevent build-up of harmful levels of solvent in the
paint, and has been the justification for the use, in certain instances, of acetone [22]. The drying/deswelling process is slower than the swelling process due to the affinity of the paint film binder for
solvent. Stolow's de-swelling diffusion coefficients are lower than the equivalent swelling
coefficients, generally by a factor of two or more. Two phases of drying occur: an initial phase
involving evaporation of liquid solvent at the surface (controlled by solvent evaporation), and a
later phase of loss of absorbed solvent from within the paint
Two studies at Institut Royal du Patrimoine Artistique/ Koninklijk Instituut voor het
Kunstpatrimonium (IRPA/KIK) in Brussels have examined solvent retention by a test painting on
canvas c. 80 years old. Dauchot-Dehon [51] measured the retention of C'4 isotope-labelled solvents
applied by syringe drop and by swab to the surface of the painting. The solvents tested were
methanol, ethanol, propan-2-ol and acetone. Two phases of drying were observed, as described
above. Loss of solvent appeared mostly to follow an exponential pattern. Traces of solvents were
found to be left (somewhere) in the painting structure after many days; up to 120 days for methanol.
Ethanol was suggested as preferable for varnish removal over propan-2-ol, on the grounds of its
lower retention.
A related study by Masschelein-Kleiner and Deneyer [52] used gravimetric measurements to follow
solvent loss from similar paint samples. A wide range of solvents was classified according to the
time taken to evaporate from the samples, and also to their rates of penetration, which were based
theoretically on measurements of viscosity and surface tension. The stated intention was to support
an approach to solvent selection in cleaning paintings, polychromy, etc., partly on the basis of the
penetration of solvents into porous bodies. Solvent mixtures for cleaning were recommended by the
authors on the principle that those with strong penetration and retention were the most dangerous to
the paint layers. This approach was developed further in the didactic work by Masschelein-Kleiner
[53]. In addition to penetration and retention, solvents were further classified according to their
types of chemical interaction and their solutes, leading ultimately to a sequential scheme of more
than 20 solvents/mixtures for removing different kinds of coating. This scheme is widely followed
by conservators around the world.
Such specific recommendations are potentially misleading, since they fail to acknowledge other
important factors - most obviously, solvent power and associated swelling and leaching effects.
Solvents not retained strongly in the paint samples in these experiments included a number of
chlorinated solvents known as strong swellers of oil paint, and solubility parameters for some of the
(milder) recommended mixtures actually place them in the peak swelling region identified by
Stolow [21]. The scheme of solvents progresses rapidly to some quite potent formulations. Another
problem derives from the assumptions inherent in the methodology: paint is not simply a porous
body. While the system may be suited to cleaning, for example, fifteenth- and sixteenth-century
Netherlandish paintings bearing appreciably aged varnish and overpaints, the IRPA/ KIK scheme is
likely to lack the necessary subtlety, especially regarding polarity, for cleaning paintings that are of
more recent date and more intrinsically solvent-sensitive. More generally, the very idea of a simple
numerical list of solvents of increasing potency may, in itself, be considered problematic.
Furthermore, the criteria of penetration and retention do not discriminate neutral solvents from
acidic and basic compounds - ammonia and formic acid are included in some of the proposed
cleaning mixtures, and the distinct risks associated with these materials are not referred to in the
study.
The conflicting recommendations from different studies illustrate the complexity of the problem of
determining risks in solvent cleaning, and the numerous parameters involved. Diffusion and
evaporation are indeed important factors - one of Ruhemann's criticisms of Stolow's research, with
reference to the use of acetone for cleaning, was that his immersion and swelling data downplayed
the contribution of these properties [22, p. 201]. As Ruhemann correctly observed, much of the
safety in cleaning relies on careful manipulation of solvents by the conservator: scientific studies
necessarily use simplified models of cleaning, which in most cases can only provide general
indicators of risk. Any specific recommendations of cleaning solvents based on scientific studies,
such as those of Masschelein-Kleiner and Deneyer, must be carefully qualified, otherwise there is a
risk of creating a false sense of security in their use.
Model studies versus experiments on paintings
Some methodological problems associated with technical studies of cleaning have been alluded to
in the foregoing discussion. Ruhemann, for example, dismissed much of Stolow's work on the
grounds that the methods used did not properly replicate the conditions occurring in actual cleaning
practice. Concern for reliable replication of the process of cleaning is a consistent theme in
criticisms of technical studies of cleaning. The difficulty lies in the problem of making valid
measurements of an operation that involves so many internal and external variables. The more one
reduces the number of variables in pursuit of accurate, reliable data - for example, by immersing
samples in solvent, rather than rolling solvent over the sample surface with a swab - the further this
takes one away from the practical reality of the cleaning process. Conversely, studies that seek to
replicate and monitor the cleaning process as it actually occurs may involve so many variables that
the results or observations are largely uninterpretable. Somewhere between these two extremes,
hopefully, is a happy medium - there is a place for both practical, 'clinical' evaluations of the actual
cleaning process, and for more reductionist approaches, provided of course that correlation can be
made between them. The methodological constraints inherent within each approach demand explicit
consideration.
The majority of the studies described above have been carried out using laboratory-prepared paint
films rather than samples from paintings, the main reason being, as Stolow observed, 'it is
preferable to obtain data on films of precisely known composition and origin, than to experiment on
films whose past history is uncertain' [17, p. 10). Aside from practical limitations, there are also
obvious ethical constraints on the extent to which experiments can be carried out on old paintings,
unless fragments of paintings are available for test purposes. Laboratory-prepared films have the
additional advantage that they can be formulated with a variety of pigments, media, film
thicknesses, etc., to investigate the influence of these factors in a controlled manner. However,
model paint films are necessarily much younger than the paint layers they are intended to represent,
and from the time the first publications on the subject appeared there has been debate about the
extent to which data from model studies can be related to solvent effects on paintings [14, p. 34; 22,
pp. 197-202]. Exchanges on the subject have sometimes become quite fervent [54], but this debate
serves at least to demonstrate the importance of the issue.
The concern is a valid one. Although recently we have seen increased research activity in the area,
the chemistry of long-term oil paint ageing is still not thoroughly understood; but it is recognized
that paint films continue to undergo chemical and physical changes over long periods of time, as
has been highlighted in some of the recent studies [39, 43]. Paint films of considerable age, as in
paintings hundreds of years old, are therefore likely to show differences in their response to solvents
from the younger, model paint films. Some studies have used accelerated ageing treatments in an
attempt to produce paint films closer in character to significantly aged films; but for a chemically
complex system such as drying oil, particularly if pigment is present, doubt has been expressed as to
whether conditions of intense light and/or elevated temperature can be used to reproduce accurately
the same combinations of reactions that occur during long-term, ambient ageing [55, p. 598; 56,
57]. Despite their limitations, and given that one may never be able to replicate exactly the
combination of deterioration processes (both chemical and physical) that occur naturally,
accelerated ageing treatments in cleaning studies can provide valuable indications of the behaviour
of more mature paint films, provided the ageing treatments are designed carefully with reference to
what is currently known of deterioration mechanisms.
To avoid these difficulties, the 'clinical' approach to cleaning studies was adopted in a recent study
by White and Roy [57]. This presented comparative analyses of samples taken from a series of
more than 10 paintings, dating from the fifteenth to nineteenth centuries, in the course of cleaning
treatments. Paired samples were analysed from paint areas cleaned mechanically (representing the
paint film 'before cleaning') and using solvents. No general differences were observed between the
samples, either in their organic composition (studied by GCMS) or physical structure (porosity,
studied by SEM). One limitation of the organic analysis was that, at least in the form published, the
interpretations were only semi-quantitative, relying on visual comparisons of chromatograms, from
which it was difficult to judge what were stated to be 'negligible' or 'insignificant' differences.
Overall, however, the results offer a compelling demonstration that, at least for these paintings, the
magnitudes of leaching during varnish removal were so small as to be negligible. Of particular
interest were quantitative GCMS data presented on extracts obtained from swabs after gently
rubbing over exposed paint surfaces, which indicated only trace quantities of fatty acids, even after
prolonged exposure to acetone or propanol. Comparisons with data for extracted components
reported, for example, by Stolow and by Erhardt and Tsang are especially enlightening. In
attempting to draw wider conclusions from this study, though, we should be conscious of Stolow's
aforementioned point about uncertainty regarding the condition and past history of the paints.
Information provided here on the restoration histories of these paintings is limited and, apart from
one, all have probably been cleaned in the past, possibly several times. The influence of these past
cleaning treatments on the experimental results is hard to assess. For example, neither the
proportion of any soluble phase to the total organic phase in the paint is reported for these
measurements, nor the mass proportion of organic material in the paint. In one instance a detailed
quantitative treatment is given, in which the upper limit of fatty acids extractable with prolonged
solvent exposure is estimated to be 'one or two hundredths of a per cent' [57, p. 163]. This result is
for a swab extract from a lead white paint area in a sixteenth-century Venetian picture. Comparable
figures for younger, previously untreated, medium-rich paint would be highly informative.
Measurements of quantities of fatty acids in cleaning swabs had also been reported in an earlier
study by Ford and Byrne [58], who found variable quantities of palmitic and stearic acid in cotton
swabs that had been impregnated with various cleaning agents (including solvents and resin soap
gels) and rolled over the surface of an approximately 90-year-old, unvarnished oil painting. Details
of the paint tested, such as pigmentation, were not provided. Quantities extracted were in the order
of micrograms of fatty acid per cm2 paint area, with higher values for strongly alkaline reagents,
although the authors acknowledged these values were highly dependent on variable factors (time of
exposure, pressure, etc.) inherent in the swabbing procedure. These results are in contrast to the
upper limit of nanograms of fatty acids extractable with prolonged exposure to an acetone swab, for
a corresponding area of white paint in the sixteenth-century painting, which was found by White
and Roy [57, p. 163].
Experiments carried out by Sutherland [42, chapter. 6] were based on a similar approach to that of
White and Roy, analysing samples from paintings before and after solvent cleaning, and using GC
to determine the fatty acid composition of the samples. Instead of comparing analyses of 'total' fatty
acids in paint samples, however, measurements were made of the proportions of fatty acids in each
sample extractable with solvent, and any differences in these quantities of extractable material in
samples taken before and after local cleaning tests were used to indicate the extent to which
leaching had occurred. In cleaning tests carried out on a number of paintings from the seventeenth
to nineteenth centuries, a small, but measurable proportion of fatty acids was found to be removed
from the paint layers in some cases, whereas in others the effect was too small to be reliably
determined. Again, this was despite the use of polar solvents, and prolonged exposure to swabs in
most cases.
It is worth noting that these experiments are based on individual cleaning treatments, and do not
account for the possible cumulative effects of repeated treatments over time. Given that there can be
a considerable 'reservoir' of extractable material in paint films even of significant age [42, chapter
4], it seems likely that leaching is a phenomenon that will occur, although at a low level, on each
repeated exposure to solvent. Hence, although an individual cleaning treatment may present a
minimal risk to the paint layers from leaching, it is still important to avoid frequent re-treatments as
far as possible. To a large extent, this view supports conventional wisdom in cleaning.
Experiments carried out directly on paintings are important in that they begin to put the model
studies in a practical context, and the studies described above provide data to support the
generally held suspicion that cleaning effects on significantly aged paint layers - at least with
respect to leaching - are considerably less pronounced than the effects observed in studies of test
films under more extreme conditions. Such model studies have often been too readily dismissed,
however, and some of the interpretations of object-based studies over-generalized - such as a
reference to 'the long-held assumption of the safety and reliability of solvent cleaning methods for
old master paintings' [57, p. 160], neglecting to acknowledge the fact that many paintings are
certainly vulnerable to damage by solvents.
Conclusion
The recent trend towards direct studies of cleaning effects on paintings, as well as the study of
significantly aged paint samples in conjunction with younger, reference paint films, has been an
important development in cleaning research. Much work still remains to be done, however. For
example, investigations of alterations in mechanical and optical properties of paint layers as a result
of cleaning treatments would be valuable, to complement similar, 'clinical' studies of leaching. With
regard to leaching, data is presently available only for a fairly limited range of pure solvents.
Information on the leaching potential of a wider range of solvents that reflects the various practices
of restorers, particularly mixtures comprising a polar and an apolar solvent, would add significantly
to our understanding of this aspect of risk in the practical situation of cleaning. Correlation of the
models of paint/solvent interaction with further studies of the selective solubility of aged resin
varnishes in solvents of different type may also lead to useful practical developments. Along with
the improved understanding of drying oil chemistry, which is coming from current research,
continuing studies will help to build a more precise and constructive picture of the risks associated
with solvent cleaning.
Acknowledgements
Both authors would like to acknowledge the support of the MOLART project (Molecular Aspects of
Ageing in Painted Works of Art) for enabling this review to come to fruition. MOLART was a
research programme funded by the Dutch Organization for Scientific Research (NWO) and
organized by the FOM Institute for Atomic and Molecular Physics (FOM-AMOLF), Amsterdam.
This review has its origins in an unpublished internal report written by Alan Phenix for MOLART,
and in the Ph.D. thesis [42] of Ken Sutherland, which was published and partly funded by
MOLART/FOM-AMOLF. Ken Sutherland would also like to acknowledge the Charles E. Culpeper
Foundation and the Andrew W. Mellon Foundation for supporting his doctoral research at the
National Gallery of Art, Washington DC.
References
1
Khandekar, N., 'A survey of the conservation literature relating to the development of aqueous
gel cleaning on painted and varnished surfaces', Reviews in Conservation 1, 2000, pp. 10-20.
2
Wolbers, R., Notes for Workshop on New Methods in the Cleaning of Paintings, Getty
Conservation Institute, Marina del Rey, 1989.
3
Wolbers, R., Cleaning painted surfaces: aqueous methods, Archetype, London, 2000.
4
Michalski, S., 'A physical model of the cleaning of oil paint', in Mills, J.S. and Smith, P., eds,
Cleaning, Retouching and Coatings. Technology and Practice for Easel Paintings and Polychrome
Sculpture. Preprints of the Contributions to the Brussels Congress, International Institute for
Conservation of Historic and Artistic Works, London, 1990, pp. 85-92.
5
Report from the Select Committee on the National Gallery; together with the Proceedings of
the Committee, Minutes of Evidence, Appendix and Index, House of Commons, London, 1853.
6
Brommelle, N., 'Material for a history of conservation', Studies in Conservation 2, 1956, pp.
176-86.
7
Graham, I., 'The effect of solvents on linoxyn films', journal of the Oil & Colour Chemists'
Association 36, 1953, pp. 500-6.
8
Morrell, R., 'Polymerised drying oils', Journal of the Society of Chemical Industry 34, 1915,
pp. 105-9.
9
Elm, A., 'The drying and yellowing of trilinolenic glyceride', Industrial and Engineering
Chemistry 23, 1931, pp. 881-7.
10 Nelson, H., 'Stress-strain measurements on films of drying oils, paints and varnishes', in
American Society for Testing Materials, 24th Annual Meeting, New Jersey, 1921, pp. 1111-39.
11 Long, J., Rheineck, A. and Ball, G., 'Studies in the drying oils XVII. Influence of several
factors on the mechanism of drying of oil films', Industrial and Engineering Chemistry 25, 1933,
pp. 1086-91.
12 Payne, H., 'Linseed oil films. Permeability, adsorption and solubility', Industrial and
Engineering Chemistry 32, 1940, pp.
737-43.
13 Stout, G., 'A preliminary test of varnish solubility', Technical Studies in the Field of Fine Arts
4, 1936, pp. 146-61.
14 Laurie, A., 'Restrainers and solvents used in cleaning old varnish from pictures', Technical
Studies in the Field of Fine Arts 4,1935, pp. 34-5.
15 Schmid, P., 'Extraction and purification of lipids II. Why is chloroform-methanol such a good
lipid solvent?', Physiological Chemistry and Physics 5, 1973, pp. 141-50.
16 Browne, F., 'Swelling of paint films in water VIII. Swelling of linseed oil paints in water and
organic liquids', Forest Products Journal 6, 1956, pp. 312-18.
17 Stolow, N., Some Investigations of the Action of Solvents on Drying Oil Films, Ph.D. thesis,
University of London, 1955.
18 Stolow, N., 'The action of solvents on drying oil films', Parts I and II, Journal of the Oil and
Colour Chemists Association 40, 1957, pp. 377-402; 488-99.
19 Stolow, N., 'Application of science to cleaning methods: solvent action studies on pigmented
and unpigmented linseed oil films', in Thomson, G., ed., Recent Advances in Conservation:
Contributions to the IIC Rome Conference, 1961, Butterworth, London 1963, pp. 84-8.
20 Stolow, N., 'Solvent action', in Bromelle, N. and Smith, P., eds, Conservation and Restoration
of Pictorial Art, IIC/Butterworths, London, 1976, pp. 153-7.
21 Stolow, N., 'Solvent action', in Feller, R.L., Stolow, N and Jones, E.H., On Picture Varnishes
and Their Solvents, revised and enlarged edition, National Gallery of Art, Washington DC, 1985,
pp. 47-116.
22 Ruhemann, H., The Cleaning of Paintings, Faber and Faber, London, 1968.
23 Stolow, N., 'The measurement of film thickness and of solvent action on supported films',
Studies in Conservation 3, 1957, pp. 40-44.
24 Wollbrink, T., 'The composition of proprietary paint strippers', Journal of the American
Institute for Conservation 32, 1993, pp. 43-57.
25 Teas, J., 'Graphic analysis of resin solubilities', Journal of Paint Technology 40, 1968, pp. 1925.
26 Banik, G. and Krist, G., 'Losungsmittel in der Restaurierung' International Centre for the
Study of the Preservation and Restoration of Cultural Property (ICCROM)/Verlag der Apfel,
Vienna 1984.
27 Cremonesi, P., L'uso dei solventi organici nella pulitura di opere policrome, II Prato, Padova,
2000.
28 Stoner, J.H., 'The impact of research on the lining and cleaning of easel paintings', Journal of
the American Institute for Conservation 33, 1994, pp. 131-40.
29 Hedley, G., 'Solubility parameters and varnish removal: a survey', The Conservator 4, 1980,
pp. 12-18, re-printed, with corrections, in Villers, C, ed., Measured Opinions, United Kingdom
Institute for Conservation, London, 1993, pp. 128-34.
30 Eissler, R. and Princen, L., 'Effect of some pigments on tensile and swelling properties of
linseed oil films' Journal of Paint Technology 40, no. 518, 1968, pp. 105-11.
31 Blank, S. and Stavroudis, C, 'Solvents and Sensibility', Western Association for Art
Conservation Newsletter 11, no. 2, 1989, pp. 2-10.
32 Phenix, A., 'Solubility parameters and the cleaning of paintings: an update and review',
Zeitschnft fur Kunsttechnologie und Konservierung 12, 1998, pp. 387-409, Spring 2002.
33 Phenix, A., 'The swelling of artists' paints in organic solvents Part 2 - comparative swelling
powers of selected organic solvents and solvent mixtures', Journal of the American Institute for
Conservation, forthcoming, Spring 2002.
34 Phenix, A., 'The swelling of artists' paints in organic solvents. Part 1 - a simple method for
measuring the in-plane swelling of unsupported paint films', Journal of the American Institute for
Conservation, forthcoming, Spring 2002.
35 Reichardt, C, 'Solvents and solvent effects in organic chemistry', 2nd edn, VCH, Weinheim,
1988.
36 Stolow, N., 'The application of gas chromatography in the investigation of works of art', in
Application of Science in Examination of Works of Art, Museum of Fine Arts, Boston, 1965, pp.
172-83.
37 Jones, PL., 'The leaching of linseed oil films in isopropyl alcohol', Studies in Conservation 10,
1965, pp. 119-28.
38 Erhardt, D. and Tsang, J.S., 'The extractable components of oil paint films', in Mills, J.S. and
Smith, P., eds, Cleaning, Retouching and Coatings. Technology and Practice for Easel Paintings
and Polychrome Sculpture. Preprints of the Contributions to the Brussels Congress, International
Institute for Conservation of Historic and Artistic Works, London, 1990, pp. 93-7.
39 Schilling, M., Khanjian, H. and Carson, D., 'Fatty acid and glycerol content of lipids; effects
of ageing and solvent extraction on the composition of oil paints', Techne 5, 1997, pp. 71-8.
40 Tumosa, C, Millard, J., Erhardt, D. and Mecklenburg, M., 'Effects of solvents on physical
properties of paint films', 1COM Committee for Conservation, 12th Triennial Meeting, Lyon, 1999,
pp. 347-52.
41 Sutherland, K. and Shibayama, N. 'The components of oil paint films extracted by organic
solvents', /COM Committee for Conservation, 12th Triennial Meeting, Lyon, 1999, pp. 341-6.
42 Sutherland, K. Solvent extractable components of oil paint films; Ph.D. thesis, University of
Amsterdam, 2001.
43 van den Berg, J.D.J., van den Berg, K.J. and Boon, J.J., 'Chemical changes in curing and
ageing oil paints' ICOM Committee for Conservation, 12th Triennial Meeting, Lyon, 1999, pp. 24853.
44 Erhardt, D., Tumosa, C.S. and Mecklenburg, M.F., 'Natural and accelerated thermal aging of
oil paint films', in Roy, A. and Smith, P., eds, Tradition and Innovation: Advances in Conservation,
Preprints of the Contributions to the Melbourne Congress, 10-14 October, IIC, London, 2000,
pp.65-9.
45 McGlinchey, C, 'Thermal analysis of fresh and mature oil paint films: the effect of pigments
and driers and the solvent leaching of mature paint films', Materials Research Society Symposium
Proceedings 185, 1991, pp. 93-103.
46 Hedley, G., Odlyha, M., Burnstock, A., Tillinghast, J. and Husband, C, 'A study of the
mechanical and surface properties of oil paint films treated with organic solvents and water', in
Mills, J.S. and Smith, P.,eds, Cleaning, Retouching and Coatings. Technology and Practice for
Easel Paintings and Polychrome Sculpture. Preprints of the Contributions to the Brussels
Congress, International Institute for Conservation of Historic and Artistic Works, London, 1990,
pp. 98-105.
47 Tsang, J.S., and Erhardt, D., 'Current research on the effects of solvents and gelled and
aqueous cleaning systems on oil paint films', Journal of the American Institute for Conservation 31,
1992, pp. 87-94.
48 Sutherland, K., 'The extraction of soluble components from an oil paint film by a varnish
solution' Studies in'Conservation 45, 2000, pp. 54-62.
49 Singer, B., Devenport, J. and Wise, D., 'Examination of a blooming problem in a collection of
unvarnished oil paintings', The Conservator 19, 1995, pp. 3-9.
50 Boon, J.J. and van Och, J., 'A mass spectrometric study of the effect of varnish removal from a
19th century solvent sensitive wax oil painting', 1COM Committee for Conservation, 11th Triennial
Meeting, Edinburgh, 1996, pp. 197-205.
51 Dauchot-Dehon, M., 'Les effets des solvants sur les couches picturales. 1. Alcools et acetone',
Bulletin IRPA 14, 1973/4, pp. 89-104.
52 Masschelein-Kleiner, L. and Deneyer, J., 'Contribution a l'etude des solvants utilises en
conservation', ICOM Committee for Conservation, 6th Triennial Meeting, Ottawa, 1981, pp. 1-10.
53 Masschelein-Kleiner, L., Les Solvants, PInstitut Royal du Patrimoine Artistique, Cours de
Conservation 2, Brussels, 1981.
54 Wolbers, R., Erhardt, D., Bischoff, J., Phenix, A. and Stavroudis, C, 'Correspondence', Studies
in Conservation 39,1994, pp. 284-6; 40, 1995, pp. 207-12.
55 Miller, C, 'Degradation of drying oil films', Journal of the American Oil Chemists' Society 36,
1959, pp. 596-600.
56 Erhardt, D., Tumosa, C. and Mecklenburg, M., 'Can artists' oil paints be accelerated aged?',
Polymer Preprints 41, 2000, pp. 1790-1.
57 White, R. and Roy, A., 'GCMS and SEM studies on the effects of solvent cleaning on old
master paintings from the National Gallery, London', Studies in Conservation 43, 1998, pp. 159-76.
58 Ford, B. and Byrne, A., 'The lipid stripping potential of resin soap gels used for cleaning oil
paintings', Australian Institute for the Conservation of Cultural Materials Bulletin 17, 1991, pp. 5160.
59 Marcus, Y. 'The properties of solvents', John Wiley, Chichester, UK, 1999.
60 Barton, A.F.M., CRC Handbook of Solubility Parameters and Other Cohesion Parameters,
CRC Press, Boca Raton, 1983.
Authors
Alan Phenix was educated at Leeds University, receiving a B.Sc. Honours in Chemistry and Colour
Chemistry Combined in 1979. He studied Conservation of Easel Paintings for three years at the
Courtauld Institute of Art, University of London and was awarded the Institute's post-graduate
diploma in 1984. After an internship at the Tate Gallery London, he worked as paintings
conservator in Australia, returning to Britain in 1989 to take up a teaching and research position at
the Hamilton Kerr Institute, University of Cambridge. He was Lecturer in the Department of
Conservation and Technology at the Courtauld Institute of Art from 1991 to 2000, during which
time he spent 15 months on secondment to the MOLART project managed by the FOM Institute for
Atomic and Molecular Physics, Amsterdam. He is now Research Fellow at the Royal College of
Art/V&A Conservation Programme, London.
Ken Sutherland received a B.Sc. in biochemistry from University College London in 1992, and a
diploma in the conservation of easel paintings from the Courtauld Institute of Art in 1995. He spent
three years as Culpeper Fellow in the scientific research department of the National Gallery of Art,
Washington DC, and from 1998 to 2001 continued research at the National Gallery of Art into the
effects of organic solvents on oil paint films as part of the MOLART research programme, funded
by the Dutch Organization for Scientific Research (NWO), obtaining a doctoral degree at the
University of Amsterdam in 2001. He is currently a scientist in the conservation department of the
Philadelphia Museum of Art.