An innovative methodology for testing and selecting greener solvents for varnishing paintings

A primary purpose within cultural heritage science is preserving this heritage for current and future generations. Given this inherently sustainable perspective (within the United Nations 17 Sustainable Development Goals, Target 11.4 specifically refers to protecting and safeguarding the world’s cultural and natural heritage), the profession aligns with UN Sustainable Development Goals^1,2. Since our origins as a species, we have expressed and developed our characteristic artistic drive through painting^3,4. The cultural significance of paintings is thus incalculable, yet treatments to safeguard them can be harmful to the heritage professional as well as to the environment. Following recently suggested principles and definitions for green heritage conservation^5,6, the shift towards “green” practices and the use of “greener” solvents that are less harmful for the environment, human practitioners and artwork represents a significant step. These alternatives aim to match or surpass the performance of more traditional methods, and are part of a broader movement to promote sustainability in conservation^7,8,9.

Across professional disciplines, organic solvents may be used for cleaning (e.g., removal of non-original materials and coatings), consolidation, restoration and varnishing. Organic solvents still commonly in use include aromatic solvents such as xylene and petroleum-distillates¹⁰, which pose significant risks to both the art conservator’s health and the environment^11,12. The shift towards greener solvents in conservation is thus primarily driven by the desire to minimize the environmental impact and health hazards associated with conventional solvent-based methods, while also integrating them into more sustainable processes, and adhering to the principles of green chemistry that prioritizes safety, efficiency, and minimized environmental footprint^13,14. Whilst ‘green’ can refer solely to manufacture from biobased, renewable sources, ‘greener’ therefore acknowledges that all solvents carry hazards and impacts which should be comparatively assessed^6,13,14. Several general-purpose solvent selection guides have been published with the aim to reduce use of the most hazardous solvents¹⁵. A quantitative assessment framework for defining a greener solvent is provided in metrics toolkits (e.g., CHEM21), whose purpose is to measure the sustainability of chemical and biochemical processes based on scores for safety, health, and environment criteria¹⁶, as aligned with the Global Harmonized System (GHS) and European regulations¹⁷. However, the greenness of a solvent depends on the specific circumstances of its use¹⁸, a fact which must be increasingly reflected within the growing specialist research focused on particular conservation treatments^19,20. A nuanced perspective is critically needed since the conservator-restorer faces complications when looking to substitute their most problematic solvents, and the required solvent properties vary with both application and substrate/object. In the first place, correct solvent selection in conservation requires practitioner’s expertise and familiarity, since safe effectiveness often depends on combining observations with understanding and utilizing specific solvent properties. In paintings conservation, research into solvent action on artist oil paint films—spanning some 70 years—has clarified potentially undesirable effects^{21,22,23,24,25} and illustrated the complexity of modeling the interactions. Examining the potential longer-term effects of specific greener solvents expands this body of work^26,27. Some solvent characteristics (e.g., sensitivity to oxidation) and possible interactions (e.g., between free fatty acids and cyclic, aprotic solvents) which pose a generic risk to paint films can hereby be anticipated. Whilst various effects are directly observable during conservator’s tests (e.g., swelling of the paint film potentially leading to eventual pigment loss), there also exists the risk of less detectable, longer-term effects. Difficulties in predicting both are exacerbated by the individual response of paint films in any artwork.

It is acknowledged therefore that solvent selection in heritage conservation must satisfy highly variable criteria, and for painting cleaning treatments for instance these can change within a single artwork. The required solvent parameters/properties for dissolving resins (e.g., for varnish applications) meanwhile can be more clearly defined (note, switching to different water-soluble resin formulations could also be considered. However, paint films can be directly sensitive to water, and water can also strongly affect the chemistry of paint films. Further the stability criteria for new conservation resins and formulations are stringent and require considerable study). For varnish applications on paintings, generally ideal solvent properties include low toxicity, low to medium evaporation rate, and no interactions with the original materials. Single cases with successful replacement of solvents in protective coatings for conservation have accordingly been suggested²⁸. Meanwhile this current work presents a collaborative methodology to apply appropriate greener solvents for applications in painting conservation by combining preliminary selection, standard coating tests and application testing. The use of a software tool is instrumental in this endeavor, utilizing artificial neural networks to cluster solvents based on their physical properties and identify less harmful alternatives. To maximize significance, four common resins—Dammar, Paraloid B72, Laropal A81, Regalrez 1094—frequently used for varnishes in paintings conservation were tested²⁹, with xylene selected as the target solvent. In a recent worldwide survey¹⁰, xylene featured in the top 10 most used solvents in conservation, and although more hazardous solvents remain in use, exposure to xylene can result in serious health impacts¹². Its wide use, favorable solubility behavior and working properties for resin-based varnish applications make it an appropriate target^13,14,30. To the author’s knowledge, this is the first study to combine advanced solvent substitution approaches with field testing for this specific treatment.

The overall approach followed in this work includes three main steps. In step 1, substitute solvents are selected using two software tools in parallel, followed by a safety assessment. Step 2 includes the testing of the suggested solvents and resin solutions both individually and applied on test boards with analytical characterization, including viscosity and resin solubility measurements, static water contact angles, gloss and color. This results in a further reduction in the number of possible substitute solvents. Finally, the potentially successful resin solutions are tested as varnishes on actual paintings in step 3. This methodology is further discussed in detail in section “Methods and methodology”.

Terminology

Throughout: “Pure solvent” refers to a single solvent (as opposed to a mixture); “solvent(s) only” refers to solvent(s) without resin; “solvent effects” refers to effects observed on tested paint films when tested with the solvent(s) only; “resin solution” refers to the solutions of resin and solvent(s); these are further referred to as a varnish only after testing for this application on paintings in step 3.

Methodology and workflow

The methodology and workflow followed in this work for selection and application tests of alternative greener solvents for varnishes in conservation are presented in Fig. 1 and described in detail in the text below.

Step 1 (a), substitute solvent selection and safety assessment, Step 2 (b), testing of the suggested solvents and resin solutions on test boards with analytical characterization, Step 3 (c), testing of the potentially successful resin solutions as varnishes on actual paintings. Green boxes: substitute solvent selection frameworks, blue boxes: solvent and resin solutions for testing, gray boxes: substrates for testing, pink boxes: required solvent and film properties, yellow boxes: comments on assessments and requirements.

Full size image

Step 1: substitute solvent selection and safety assessment

Two approaches were used for selecting alternative solvents and their binary solvent blends to replace xylene. For one-on-one solvent substitutes, the SUSSOL (solvent selection and substitution) software was used³¹. Next, the solvents and binary solvent blends to dissolve each of the resins were calculated using the Hansen Solubility Parameters in Practice (HSPiP) software (version 5.2.02, 2015)³².

The SUSSOL software uses a solvent dataset containing 500 solvents with every solvent described by 22 physical properties. Employing a clustering algorithm, solvents are clustered into groups based on similarity in their physical properties while no information on the chemical structure is provided to the algorithm. Kohonen’s Self-organizing Map (SOM)³³ was used where clusters are ranked on a two-dimensional grid with the distance between two clusters as a measure of similarity. After cluster analysis, a list with appropriate candidate solvents was generated.

To limit computation time a subset was created for each resin, using Hansen Solubility Parameters (HSP) for the respective resins and calculating Hansen distance to each of the solvents in the 500 solvents database. The subsets contained between 236 and 451 solvents. In practice, 5000 clusterings were performed on a 9 × 9 and 11 × 11 SOM for each of the xylene isomers, i.e., o-xylene (CAS 95-47-6), p-xylene (CAS 106-42-3) and m-xylene (CAS 108-38-3). After clustering the resulting solvent candidate lists were combined and all solvents with health scores >3 were eliminated from the list.

Hansen Solubility Parameters (HSP) were originally proposed by Charles Hansen, as fractions of the cohesive energy density or Hildebrand solubility parameter³⁴. The total solubility parameter δ can be broken down in fractions related to the dispersion interactions δD, polar (permanent dipole) interactions δP, and hydrogen bonding (H-bond acceptor and donor) δH. The HSP of a resin can be experimentally determined by testing its solubility in a range of solvents with known HSP values. A solvent that dissolves the resin will have HSP values closer to that of the resin. By evaluating solubility across different solvents in the ‘solvent space’, a ‘solubility sphere’ can be created using the HSPiP software tool³². The 3D coordinates of the sphere’s center represent the resin’s HSP, and the radius defines the range of solvents in which the resin will dissolve. Once the HSP values for a resin are known, they can be used to select solvents or solvent blends likely to dissolve the polymer.

The final selection of suitable solvents was made using the CHEM21 assessment approach¹⁶, which is robust since it requires little data. This recognized methodology^13,14,30 assigns a score between 1 and 10 in three categories: safety (S), health (H), and environment (E), with 10 being the highest hazard. A color code is associated with the scores: green (1–3), yellow (4–6), and red (7–10). The combination of the three scores determines the solvent’s final ranking as either recommended, problematic, or hazardous. After compiling a list of solvents potentially suitable for the selected resins, all solvents with red CHEM21 scores were eliminated. The limitations of this assessment will be discussed further below.

Preliminary solubility testing was conducted for each solvent candidate with the target resin. To resemble the physical properties of the target solvent, xylene, further selection was based on solubility and relative evaporation rate. The selection process also placed particular emphasis on the potential for bio-based sourcing of solvents, including esters, certain ketones, alcohols, and terpenes.

Step 2: solvent and resin solution testing on reference substrates and test boards

In conservation practice, initial small-scale tests are commonly conducted to assess any visually detectable solvent effects on paint films, aiding in the safer selection of solvents for varnish removal as well as application. Whilst various systemized frameworks have been suggested for comparing cleaning approaches for paintings³⁵, a standardized methodology—the Solvent Star³⁶—was applied in this study. This approach, which uses quantified contact times to evaluate solvents effects on paint films, is designed for any solvent-based application. Although not all potential solvent effects can be hereby clarified (as noted above), initial results provide insights into the safety of solvents for paint films and the likelihood of unwanted interactions. Solvent Star tests were carried out on all self-prepared painted test boards and the individual paintings prior to testing resin solutions and varnishes.

The primary selection criteria for the different resin solutions were their working properties, drying characteristics and film formation for brushed varnish applications. On the painted test boards, variations in physical appearance of the coated surfaces when dry (esthetic quality), combined with the Solvent Star results were used to rank the resin solutions as varnishes, creating a pre-selection of appropriate candidates. The most successful varnishes were then further evaluated through standardized coating tests (see details in “Experimental” section) on reference substrates (i.e., Leneta cards and unpainted primed canvas boards) to assess their intrinsic properties and ensure they met the fundamental requirements for forming a protective film.

Step 3: varnish testing on individual paintings

After initially examining solvent sensitivity of the original paint films using the Solvent Star method, the top-ranking resin solutions from step 2 were tested and professionally assessed as varnishes on individual paintings by the conservator-restorers.

Materials

The following common resins applied in conservation practice were all purchased through Kremer Pigmente (Aichstetten, Germany):

• Laropal® A 81 (L), which is a condensation product of urea and aliphatic aldehydes (CAS 8931-47-7).

• Regalrez® 1094 (R), which is a low molecular-weight hydrocarbon resin (fully hydrogenated oligomers of styrene and alpha-methyl styrene, CAS 68441-37-2) produced by polymerization and hydrogenation of pure monomer hydrocarbon feedstocks. Special gum G 1650 is a rubber of styrene-ethylene/butylene-styrene block copolymer (CAS 66070-58-4) here added as a plasticizer in a 5% w/w ratio relative to the resin (e.g., 0.38 g plasticizer to 7.5 g resin).

• Paraloid™ B 72 (P), which is a general-purpose acrylic copolymer (ethyl methacrylate–methyl acrylate, CAS N/A).

• Dammar gum (D), which is a natural triterpenoid resin originating from Dipterocarpaceae trees, containing triterpenes (mainly low molecular weight compounds) and their oxidation products (CAS 9000-16-2).

The following solvents were purchased: xylene, (xyl), Xylenes, mixed, ≥97%, Thermo Scientific Chemicals (Geel, Belgium); eucalyptol (eucal), 1,8-Cineole, 99%, Thermo Scientific Chemicals (Geel, Belgium); isoamyl acetate (iaac), Isoamyl-acetate—Acetic acid isoamyl ester, ≥98%, extra pure, Carl Roth (Karlsruhe, Germany); anisole (anis), Anisole SOLVAGREEN® ≥99%, for synthesis, Carl Roth (Karlsruhe, Germany); cyrene (cyr), Cyrene^TM, BioRenewable, DMF and NMP Substitute, Sigma-Aldrich (Merck-group Darmstadt, Germany); dimethyl isosorbide (dimiso), Dimethyl isosorbide, BioRenewable, ReagentPlus®, ≥99%, Sigma-Aldrich (Merck-group Darmstadt, Germany); sangajol (sang), Mineral spirit, boiling point: 142–200 °C, Kremer Pigmente (Aichstetten, Germany), 2–25% aromatics; isopropanol (ipa), 98–100%, pure, Kremer Pigmente (Aichstetten, Germany); (Shellsol product line may be unavailable in some locations) Shellsol D40 (ShellD40), aroma free, Kremer Pigmente (Aichstetten, Germany), hydrocarbons, C9-C11, n-alkanes, isoalkanes, cyclics, <2% aromatics; ³Shellsol A (ShellsolA): aromatic content min. 98%, Kremer Pigmente (Aichstetten, Germany); n-pentyl propanoate (pp), Pentyl propionate, ≥99%, Sigma-Aldrich (Merck-group Darmstadt, Germany); isobutyl isobutanoate (ibib), Isobutyl isobutyrate, ≥98%, Sigma-Aldrich (Merck-group Darmstadt, Germany); Cyclohexanone, ACS reagent, ≥99.0%, Sigma-Aldrich (Merck-group Darmstadt, Germany); n-Butanol, Kremer Pigmente (Aichstetten, Germany) and (+)-Limonene, 96%, unstabilized, Thermo Scientific Chemicals (Geel, Belgium). The latter were used in a volumetric (vol %:vol %) ratio cylcohexanone:1-butanol 66:34 (cyclobut), and cyclohexanone:d-limonene 35%:65% (cyclolim).

As reference substrates, standardized checkerboard Leneta cards were purchased from Labomat Essor (Wevelgem, Belgium). The unpainted test boards were white acrylic primed canvas boards (8 × 10 cm²) sourced from Gerstaecker (code # 27271, Eitorf, Germany). The painted test boards were prepared following the same procedure historically used at SRAL for varnish application tests³⁷: Winsor & Newton™ Griffin Alkyd™ “fast-drying oil paint”, including ivory black, raw umber, ultramarine blue (Winsor & Newton, London, UK), were evenly brush-applied in 39 × 6 cm² strips (two strips for each paint type), onto an acrylic primed canvas board 2 months before testing, 40 × 50 cm², from Gerstaeker (code # 27282, Eitorf, Germany).

Finally, two historical artworks, part of a collection of deaccessioned paintings from churches and other buildings affiliated with the Bishopric of Roermond (The Netherlands), were donated for research and subjected to testing. Painting A, titled ‘Beheading of John the Baptist’ (unidentified painter, 18th century), measures 115.5 × 163.5 cm². Painting B is ‘The Crucifixion’ (unidentified painter, likely around 1800), measures 68 × 52.5 cm². Upon deaccession, both paintings had degraded, yellowed varnish coatings (UV examinations indicated natural resin-based), which were removed with ethanol 6 weeks prior to the testing described here. The varnish tests in this study reflect a typical treatment practice, applying varnish directly on top of a previously cleaned original paint layer.

Determination of HSP values and solubility testing

HSP values for selected resins were determined by simple solubility tests in 32 reference solvents. A small amount of resin (approx. 0.5 g) was mixed with 5 mL of solvent in a test tube and visually inspected after 17 h. HSPiP software was used to fit the solubility spheres and determine HSP values and radius for each resin.

Preparation of resin solutions

The resin solutions were prepared by dissolving the four resin types in selected solvents: 7.5 g of resin was weighed into a 100 mL Erlenmeyer flask and 50 mL of solvent was added, resulting in 15% w/v solutions. The samples were left overnight under mild shaking until the complete dissolution of the resins. Dammar samples were additionally filtered through a Whatman 54 filter paper to remove any fibrous residues from the natural product. The Dammar in anisole appeared cloudy, but the cloudiness disappeared after filtration. All other solutions were transparent.

Application of resin solutions

Standard application of the resin solutions was done on reference substrates, including Leneta contrast cards (5 × 10 cm²) and white acrylic primed canvas boards (8 × 10 cm²). The Leneta cards were coated with a manual spiral wire bar coater resulting in a wet film thickness of 70 µm. After evaporation of the solvent, the dry coating thickness was measured with a gauge, ranging from 50 to 55 µm. The unpainted canvas boards were manually coated with a brush to ensure full coverage of the surface. The brush applications were systematically carried out by the conservator-restorer, using the same brush for each resin mixture, applying five strokes per board. The samples were then dried under controlled environmental conditions (25 °C, 60% RH) for 2 months to allow complete evaporation of residual solvents before testing.

For applying the resin solutions on pre-painted test boards, an identical natural white hog bristle paint brush was used. While the aim was to apply a similar quantity of each mixture so that each of the four colors could be covered in one test, the brushing time varied depending on the viscosity and evaporation rate. This allowed for a realistic assessment of the key working properties prior to testing on the paintings.

The application on historical paintings followed a similar approach, but a larger brush was consistently used. Both similarly and differently colored areas within each painting were targeted for testing.

Characterization methods

For selected solvents, binary solvent mixtures and resin solutions, viscosity measurements were performed with a rotational viscometer (ViscoQC 300-L, Anton Paar GmbH, Graz, Austria) equipped with a peltier device (PTD 80, Anton Paar GmbH, Graz, Austria) for accurate sample temperature control at 25 °C. Samples were measured using a concentric cylinder spindle geometry with bob and cup (CC26 for pure solvents, CC12 for resin solutions).

Characterization of the dried films on Leneta cards and unpainted canvas boards included static contact angle measurements with deionized water (D.I.), performed on a goniometer (OCA 50, Dataphysics Instruments GmbH, Filderstadt, Germany) using droplets of 3 µL that were fitted with a tangent procedure in agreement with ISO 19403-2. The gloss measurements were taken by a micro-triglossmeter (BYK-Gardner Instruments, Geretsried, Germany) under an incident angle of 60° following ISO 2813. The color analysis was done with a CC-6807 color meter (BYK-Gardner Instruments, Geretsried, Germany) having spherical geometry d/8 according to ISO 7724-1, whereby the values for L*, a* and b* are recorded. L* represent the lightness of the color (L* = 0 yields black and L* = 100 indicates diffuse white), a* represents the color’s position between red and green (negative values indicate green and positive values indicate red), b* represents the position between yellow and blue (negative values indicate blue and positive values indicate yellow). The results of physical measurements are reported as average values with standard deviation taken from three independent measurements over the surface.

Professional observations on painted test boards and historical paintings were carried out by the conservator-restorers.

Step 1: substitute solvent selection and safety assessment

HSP values for selected resins were determined by simple solubility tests in 32 reference solvents. Following values were obtained: L resin: δD 16.8, δP 7.7, δH 8.5, R = 11.2; R resin: δD = 18.3, δP = 1.6, δH = 0.3, R = 9.2; P resin: δD 16.7, δP 10.0, δH 5.5, R = 9.8; D resin: δD = 18.8, δP = 5.5, δH = 4.0, R = 6.1. All raw data are available in Supplementary Information S1.

In searching for xylene substitutes based on clustering analysis, the SUSSOL software yielded a list of 46 potential solvent candidates with comparable physicochemical properties to xylene (see Supplementary Information S2), including 17 (carbonate) esters, 16 ketones, 5 alkyl- and alkoxy-substituted aromatic solvents, 3 (un)saturated hydrocarbons and 3 ethers (including 1,8-cineole).

The solubility properties of the selected resins (L, R, P, D) in 46 candidate solvents were monitored by HSPiP analyses, confirming good compatibility for most of the 46 solvent candidates found through clustering analysis. At this stage, employing the previously determined HSP values for the selected resins, ten additional solvents were selected using HSPiP.

Next, a further selection of the resulting 56 solvents was carried out based on E, H, S scores (CHEM21) with preferentially E ≤ 5, H < 3, S < 5, relative evaporation rate, dynamic viscosity and possible bio-based sourcing. This gave a list of 27 solvents for subsequent experimental solubility tests.

These 27 solvents are listed, indicating which software was used for the selection of the respective solvents, together with CHEM21 scores (see Supplementary Information S3). Results of experimental solubility tests for L, R, P, D resins are summarized (see Supplementary Information S4). From preliminary solubility testing, correlation between experimental data and software calculation was obtained. Based on the solubility profiles, the five most suitable solvents were selected for preparing resin solutions: anisole (anis), cyrene (cyr), dimethyl isosorbide (dimiso), eucalyptol (eucal) and isoamyl acetate (iaac). The relative properties of the selected solvents compared to m-xylene (m-xyl), p-xylene (p-xyl), and o-xylene (o-xyl) are presented in Fig. 2. The physical properties of the finally selected pure solvents were generally deemed sufficiently similar to xylene. Although, as reflected in their relatively low LogP (octanol/water) values, cyrene (cyr) and dimethyl isosorbide (dimiso) show significantly higher solubility in water at 20 °C compared to the other selected solvents and the target solvent xylene (m-xyl, p-xyl, o-xyl).

Y-axis is normalized, where 0 represents the lowest value in the 500 solvent SUSSOL dataset and 1 represents the highest value. Each solvent is represented as a polyline. More overlap between the polylines signifies similarity between solvents.

Full size image

In addition, binary solvent mixtures with an appropriate profile for dissolving the resins were proposed by HSPiP. For each L, R, P, D resin, HSPiP analysis generated thousands of potential solvent pairs. Using a combination of CHEM21 scores and Hansen distance as selection criteria, a list of 35 solvent blends was compiled, with at least four blends per resin (see Supplementary Information S5). The solubility of the L, R, P, D resins was then experimentally tested in each solvent blend (see Supplementary Information S6). Based on solubility, dynamic viscosity, relative evaporation rate differences between the solvents in the blend, and the potential for bio-based sourcing of the solvents, a final selection of two solvents blends was made for further testing.

Finally, using the software data, experimental solubility testing and CHEM21 scores for both pure solvents and blends, 14 novel resin solutions (15% w/v) were prepared for experimental application testing (solutions (a) through (n) as noted in Table 1), alongside ‘reference resin solutions’; namely solutions already used as varnishes in conservation, including those developed from practice^35,36, (noted as (o) through (t) in Table 1), and those prepared in the target solvent xylene (solutions (u) through (x) in Table 1)).

Step 2: assessment and ranking on painted test boards

The novel resin solutions of L, R, D, P resins (solutions (a) through (n) as noted in Table 1) alongside reference solutions of each resin prepared in xylene (solutions (u) through (x) in Table 1), were first tested by the conservator-restorer as a varnish on pre-painted test boards.

On the pre-painted canvas boards, the sensitivity of the paint films to the solvents alone (eucal, iaac, anis, dimiso, cyr, cyclobut, cyclolim, xylene) was first assessed using the Solvent Star method. The results for standard blue paint (Fig. 3a), brown paint (Fig. 3b) and black paint (Fig. 3c) are detailed in Fig. 3. It should be noted that the test paint films are very young, and the indicated solvent sensitivities below are only relevant for the purposes of this study. These results are not significant in relation to actual solvent effects on paint films being treated in conservation practice.

The results are shown for blue paint (a), brown paint (b), black paint (c). Used to quantify solvent contact in conservators’ tests, four of the axes variously quantify time (in seconds) and mechanical action (swab rolls) for observed solvent effects on the non-original (left side axes) and original materials (right side axes) in an artwork. The vertical axes relate to the CHEM21 rating of the solvent being tested and controllability (qualitatively assessed by conservator-restorer). Only axis 4 (contact time in seconds to point of visible effects on the paint film) is of relevance and completed here. The other axes remain unused (in blue).

Full size image

Being young paint films, almost all exhibited solvent sensitivity to each of the novel and reference solvents. Only eucal on the brown and blue paint films did not display direct solvent sensitivity. Paint film solvent sensitivity varies, and none of the resin solutions were rejected solely based on the Solvent Star results from the painted test board. These tests rather clarify the potential impacts of solvent-incurred disturbances, allowing for a qualitative evaluation of the novel resin solutions on these standardized paint films. The evaluation was based on their drying/working properties (primary criteria), as well as esthetic quality (secondary criteria), (see Supplementary Information S7). The solutions were then ranked as unacceptable, poor, ok, good, and very good.

The eight most promising novel resin solutions (solutions (a) through (g) plus (k) in Table 2), with at least one resin from each type, were selected for further testing, alongside the reference solutions.

Relative evaporation rate and viscosity are commonly referred to by practitioners as important properties when applying varnishes. The solvent evaporation rates, and solvent and solution viscosity of the novel resin solutions, were compared with their described working properties during brush-application. Three of the novel resin solutions, i.e., L cyr, L dimiso, L cyclobut, were excluded from further testing as varnishes due to unacceptable working and drying properties (i.e., they dried too quickly, making brushing out difficult, or took too long to dry). Their use in other possible conservation applications, such as consolidation etc., is not excluded. The eucal solutions had the highest solution viscosity among the different D, R, L resins. However, a higher viscosity was only noted by the conservator-restorers with the R resin. As anticipated, the P resin solutions consistently had the highest measured viscosity across all the tested solutions, yet they were not described as particularly viscous in practice. The solution viscosity and levelling ability of a varnish are directly and inversely related to the molecular weight of the resin³⁸. A higher solution viscosity will also be anticipated for solvents with a great molecular volume and when solvency is highest.

These physical properties and their relation to those experienced in practice are known to be nuanced³⁹. Also in these tests on the standardized paint films on the canvas board no clear link was observed between the ease of brushing evaluated by the conservator-restorers and the solvent/solution viscosities. Similarly, no direct correlation was found between measured solvent evaporation rates and practical brushing experience. For instance, isoamyl acetate solutions were consistently experienced as having a short brushing time, despite the relative evaporation rate being much lower than that of the reference xyl solvent. However, within the novel resin solutions, general trends could be observed between the solvent evaporation rates and perceived brushing time. Based on these outcomes, eight of the novel resin solutions (a) through (g), plus (k), were selected for further testing as varnishes, alongside the reference resin solutions.

Step 2: protective film properties from reference testing

To evaluate the protective properties of the resin solutions and the effects of solvent selection on surface properties, standard coating tests were conducted on reference Leneta cards and white canvas boards. These tests include static water contact angle measurements, as well as gloss and color evaluations.

Measurements on Leneta cards are a standard approach in the coatings industry for examining film quality. Measured water contact angles depend both on the chemistry and topography of the surface. With Leneta cards as a reference, a flat substrate (no roughness effects) is included and variations in contact angles can be only attributed to the chemistry of the coating. Static water contact angle measurements were used here to check for any significant variations that could be due to any residual solvent from the resin solution. Meanwhile, results from the primed only canvas test boards provide insights into the effect of brushed application on surfaces that more closely resemble the texture of an actual painting. These measurements across a resin film can also reveal surface uniformity and any variations in treatment or material properties. Moreover, the porosity of the primed canvas test boards may also cause water absorption and full coverage of the substrate by a closed polymer coating is required. For clarity of interpretation, the results of static water contact angles are either grouped per resin on Leneta cards (Fig. 4a), per solvent on Leneta cards (Fig. 4b), per resin on test board (Fig. 4c), or per solvent on test boards (Fig. 4d). On homogeneous Lenata cards, the contact angles seem to be primarily determined by the resin type rather than the solvent type, with higher values for R resin and some varnishes with D resin in the hydrophobic range (CA > 90°). The resin effects can be attributed to the intrinsic chemical composition of the resins containing hydrophobic moieties, either consisting of hydrocarbons with C-H bonds and lacking any other functional groups (R resin), or triterpenoids, containing some hydrophilic functional groups such as O-H, C=O, COOH (D resin). The latter is reflected in the respective HSP values (R resin: δD = 18.3, δP = 1.6, δH = 0.3, R = 9.2; D resin: δD = 18.8, δP = 5.5, δH = 4.0, R = 6.1). Consequently, interactions with the R resin are limited to London dispersion forces, whilst interactions with the D resin include minor permanent dipoles and hydrogen-bond donor and acceptor effects. However, the effect of these functional groups is limited given the large triterpenoid structure. The other L and P resins are less hydrophobic and provide contact angles in the hydrophilic region (CA < 90°). Their structure includes amide, ether and ester functions (L resin), and esters functions (P resin). This results in significantly more permanent dipoles and hydrogen-bond donor and acceptor effects as reflected in the HSP values (L resin: δD 16.8, δP 7.7, δH 8.5, R = 11.2; P resin: δD 16.7, δP 10.0, δH 5.5, R = 9.8).

Results are shown for Leneta cards grouped per resin (a), Leneta cards grouped per solvent (b), test boards grouped per resin (c), test boards grouped per solvent (d).

Full size image

Although the use of different solvents might influence the measured contact angles, solvent composition did not affect the contact angles in a uniform and identifiable way. However, a remaining influence of solvent on the equilibrium contact angles could possibly be related to the drying conditions and eventual film formation introducing local organization of the polymer molecules under evaporation of the solvent⁴⁰. Relative evaporation rates are typically lower for the selected alternative solvents (see Table 2 for relative evaporation rates), compared to the reference solvents. Alternatively, it has previously been demonstrated that solvents remained present for a long time in acrylic polymer films (e.g., P resin) when drying took place at room temperature and residual solvent had a considerable impact on the glass transition temperature value of the film⁴¹. Moreover, the physical and optical properties of several polymers used in conservation have been shown to vary when deposited from solutions of different quality⁴².

The trends in contact angles observed on standard Leneta cards are slightly reflected on the canvas boards, while inhomogeneity effects or variations in layer thickness on the boards may provide some lower contact angles as compared to the continuous films on Leneta cards. On the canvas boards, however, all coatings provided higher contact angle affirming the good coverage of the canvas structure and augmented protective properties compared to uncoated canvas boards. It can be concluded that there is no consistent influence of the selected solvent on the physical surface properties of the varnish. Anyhow, the protective properties against water are mainly determined by the selected resin type and hydrophobicity of coated canvas boards is higher compared to the non-coated board.

The results of gloss measurements on reference Leneta cards and primed (only) test boards are presented and again grouped per resin on Leneta cards (Fig. 5a), per solvent on Leneta cards (Fig. 5b), per resin on canvas board (Fig. 5b), or per solvent on test boards (Fig. 5d). Overall, there are very little variations noticed in gloss values, performing in the range of 85 ± 5 for most resins and some resins providing higher gloss values above 90. The high gloss values indicate the formation of smooth and homogeneous films with good compatibility between the resin and solvents not resulting in surface defects or local flow inhomogeneities at the surface of the coatings (except for sample (q) L sang:ipa). The gloss values for P resin show the least sensitivity to the used solvent in parallel with the previous observations of contact angles, while the low molecular weight R resin has highest gloss. The low values in gloss on white canvas boards might be due to the low reflection of light at the rough textile-like surface, while there are no significant variations in gloss noticed after application of the conservation coating. In parallel, there were no changes in color or whiteness of the canvas boards after application of the conservation coatings, either for the different resins or different solvents (see Supplementary Information S8). From these observations, it can be concluded that there are no identifiable trends in the visual characteristics on the test boards that can be directly related to either solvent or resin.

Results are shown for Leneta cards grouped per resin (a), Leneta cards grouped per solvent (b), canvas boards grouped per resin (c), test boards grouped per solvent (d).

Full size image

Step 3: characterization of dried film coatings in practice

The nine most promising novel and reference resin solutions were selected for further varnish testing. Solvent Star tests with the associated solvents were first carried out on the historical painting surfaces to identify any visible effects on the original paint films (Fig. 6). No solvents caused directly visible effects on the original paint layers. Consequently, brush-application tests of all selected novel solutions were conducted on the historical paintings, and their esthetic qualities were assessed. Rankings of poor, ok, good, very good, or excellent were assigned based on these assessments.

The results are shown for Painting A—The Beheading of St. John (a) and for Painting B—The Crucifixion (b). Used to quantify solvent contact and document effects in conservators’ tests, four of the axes variously quantify time (in seconds) and mechanical action (swab rolls) for observed solvent effects on the non-original (left side axes) and original materials (right side axes) in an artwork. The vertical axes relate to the CHEM21 rating of the solvent being tested and controllability (qualitatively assessed by conservator-restorer). Only axes 4 (contact time in seconds) and 5 (mechanical action with number of swab rolls) are of relevance here, and the others remain unused (in blue). These tests were completed for all the noted solvents with no effects observed.

Full size image

The results of application and visual inspection of the novel and reference solutions on the historical paintings are summarized for Painting A (Table 3) and Painting B (Table 4). The two novel varnishes with L resin (L eucal, L iaac), gave excellent results on Painting A, but a poor result on Painting B. This corresponded similarly with the results of the reference varnishes with L resin (L Shell-pp-ibib, L sang:ipa). The varnishes with P resin (P anis and P pp:ibib) gave ok to good results on both paintings A and B. On painting A, the P anis varnish was considered good and preferred to the reference P pp:ibib. Meanwhile the varnishes D anis and D sang (reference) gave an intermediate result on painting A. Both could be more easily applied on test painting B and gave better visual results. The novel varnish R iaac gave a good visual result on both paintings A and B, which was an especially positive result given the low expectations for this resin as a first varnish layer.

As anticipated, interactions between the specific properties of a particular paint and the resin solution strongly influenced both the experienced working properties of the solutions during their brushed application, and the visual result of the dried varnish film. Whilst the solvent carriers affected working properties (remaining within acceptable levels for varnish applications as indicated by tests on painted boards), no consistent trends were observed between solvent carrier and visual results on the actual paintings. Instead, the dried appearance of the varnish on a specific paint film was primarily determined by the resin type, rather than the solvent carrier(s). In other words, the visual result on a specific painting primarily depended on the resin in the varnish, rather than solvent carrier(s). For saturation, refractive index and levelling are of key importance³⁸.

A major problem with the use of naturally derived solvents including limonene or pinenes found in turpentine, is that residues can oxidize and accelerate the discoloration of varnishes⁴³. Unlike limonene or pinene, it is not expected that eucalyptol, the one natural solvent with promising results in this study, is prone to atmospheric oxidation. Oxidation requires a reactive site: a double bond, allyl- or alpha-hydrogen. When oxidation occurs at the double bond, epoxides are formed and are further converted to alcohols and ketones. Oxidation at the allylic- or alpha-position leads to the formation of (explosive) hydroperoxides⁴⁴. However, eucalyptol is a saturated, bicyclic ether without allyl- or alpha-hydrogens, hence no atmospheric oxidation for eucalyptol is expected and no work on eucalyptol oxidation was found in literature. Nevertheless, it remains an important topic for future research.

Refining greener solvent considerations of hazards and impacts for conservation practice

As mentioned above, CHEM 21 was used in this study in Step 1 of the substitute solvent selection process. Whilst providing a good initial framework for assessing solvent “greenness”, the often-close working conditions of solvent use in conservation practice requires closer examination of the hazards and impacts. For more information see Supplementary Information Table S9, an indicative detail of which is provided in Fig. 7, and the Greener Solvents Database⁴⁵.

A detail from Table S9 showing more specific information on the hazards and impacts of the solvents included in this study. See Table S9 for full information.

Full size image

The Globally Harmonized System of Classification and Labelling of Chemicals (GHS) hazard class and category codes, which describe the type and severity of hazard are listed for each solvent. The harmonized classifications, as well as all the classification and labeling (C&L) notifications submitted to the European Chemicals Agency (ECHA) are further included. Ideally, occupational exposure and worker threshold limits should be compared. Where available relevant data from sources such as the American Conference of Governmental Industrial Hygienists (ACGIH), the National Institute for Occupational Safety and Health (NIOSH), (ECHA), and safety data sheets (SDS) are incorporated. Table S9 also includes the notifications (see ECHA) where environmental hazards have been identified for air or water release. However, this data is only available for the more common solvents, which makes comparable assessment difficult. What remains evident is that every solvent carries hazards. For safe working practice consulting the associated SDS is a necessity and familiarity with the GHS codes is highly recommended. Appropriate ventilation (the first consideration) and personal protection equipment (PPE) must be used as necessary. For the conservation professional, understanding and assessing any associated hazards cannot depend solely on categories paired with pictograms. For instance, a highly flammable solvent (H225) may pose more acceptable risks for its use in conservation than a solvent that is harmful when inhaled (H332). For future work we would consider adapting our methodology to include a more detailed toxicology investigation after the initial CHEM21 assessment prior to practical experiments. Or in fact, a broader study to adapt the CHEM21 framework considering conservation purposes would be helpful¹⁴.

Also of pertinence, some of the solvents were noted to have a potent smell and these were categorized according to the Leffingwell database. Encouraging appropriate ventilation and the use of personal protection equipment (PPE) for acute exposures, there is a potential safety advantage in using solvents with some detectable odor. For instance, ‘..due to its strong smell, which can be perceived at low concentrations, the odor of isoamyl acetate..(ed. for instance)..is assumed to be detectable below the permissible exposure limit (100 ppm)’⁴⁶. However, regardless of relatively low associated hazards, typical working conditions for some conservation applications may negate the use of a potently odored solvent like isoamyl acetate.

Many of the substitute solvents suggested for researching during this study are currently significantly more costly than those traditionally used. With some options prohibitively expensive, it can only be hoped that wherever possible, in accordance with sustainability goals, the conservator-restorer might find the necessary financial or institutional support for greener solvent selection.

Conclusion

This paper successfully demonstrated an approach for suggesting, assessing and selecting solvents for varnish applications on paintings that can be safer for humans and the environment, whilst considering the individual nature of the artwork and its treatment requirements. Although certain specific physicochemical solvent properties can indicate a potentially successful solvent for an application, there are a great number of variables and specific requirements for selecting solvents and varnish solutions in painting conservation treatments. Combining the application of the SUSSOL and/or HSPiP software, standardized coating tests, and the practicing conservator-restorer’s expertise, proved an effective approach. The solution properties of specific resin/solvent combinations were examined and a quality control for the resulting films with the novel solvents provided. The evaporation rate of the substitute solvent was confirmed as an important physical property to consider. Testing the selected solvents on the paint films in a quantifiable and communicable way, and evaluating the novel varnishes on individual paintings, showed that certain substitute solvents in resin mixtures show promising potential for varnishing paintings. The novel varnish Regalrez 1094 in isoamyl acetate (CHEM 21 recommended rating) showed very good working properties and gave a good visual result on both paintings. The resin solution Paraloid B72 in anisole (CHEM21 recommended rating) showed good visual results on both paintings, while demonstrating acceptable working properties. Since its introduction through the present study the Paraloid B72 in anisole has been successfully used in practice at SRAL, as both an intermediate and final varnish on several canvas paintings. Use of this alternative solvent rather than the commonly used Shellsol A (CHEM21 hazardous rating), has promoted healthier conditions for the conservator-restorers whilst achieving excellent esthetic results. Adopting safer solvents is critical and should be strongly supported. This is whilst recognizing our lack of information on the longer-term/initially invisible effects on both humans and paint films, which are beyond the scope of this paper and acknowledged as ongoing research needs. Since all organic solvents carry hazards, safety regulations with the correct protective measures must always be applied in practice.