Exploring Benzyl Alcohol Derivatives and Related Compounds in the Cleaning of Oil Paintings

ABSTRACT This study examines five benzyl alcohol derivatives and three chemically similar compounds and compares them against benzyl alcohol in gelled emulsions for the removal of overpaint during cleaning of oil paintings. Comparative cleaning tests using xanthan gels, Pemulen® TR-2 gels, and neat solvent were made on overpaint on fragments of a seventeenth-century test painting. This approach demonstrated that molecular changes to a benzyl alcohol core resulted in enhanced control during the cleaning process. In some cases a benzyl alcohol derivative enabled selective removal of non-original material, when benzyl alcohol appeared to affect the original paint. Select derivatives were also tested in an area of overpaint on a sixteenth-century oil on panel painting by Jan van Scorel in the Rijksmuseum Collection through modifying the chemical activity of benzyl alcohol. Finally, two GC-MS-based methods were used to monitor benzyl alcohol retention and possible oxidation in paint layers post-treatment.


Challenges of selective cleaning of superimposed oil-based layers
The selective removal or reduction of layers of natural resin and synthetic varnishes on easel paintings is routinely successfully achieved with minimal risk to original paint layers.Such cleanings can be done using the traditional method of solvent wetted swabs, or a plethora of modern methods including controlled release through technical-grade tissues (i.e.Evolon) or immobilization in a wide variety of chemicals or hydrogels (Baij et al. 2021).However, it is also common when treating Old Master paintings to encounter non-original layers composed entirely of or in part by an oil-based binding medium.The polymeric process of drying in oil films renders them extremely difficult to solubilize once fully crosslinked (Wexler 1964).
This reality makes selective removal of an oil-based non-original layer from oil-based original layers challenging as the binding media of both share a common chemical basis.Therefore, what can solubilize the retouch also poses a risk of solubilizing the original.

Benzyl alcohol in cleaningattributes and limitations
Benzyl alcohol (BnOH), an organic solvent composed of an aromatic ring with a pendant alcohol at the benzylic (Bn) position, is an established reagent in paintings conservation for the reduction or removal of aged varnish or retouch layers containing aged oil (Dorge 2004).It has a particular aptitude for swelling oil-based materials (Phenix 2002), which is good for removing overpaint/ retouches, but potentially dangerous for the original.Its frequent success, where other solvents sometime fail, might suggests that the unsaturation present in the ring and the hydroxyl moiety together provide the necessary balance in Van der Waals and hydrogen bonding forces needed to obtain a good chemical match with polymerized oils.
Former studies comparing organic solvents used in paintings conservation have classified BnOH as a 'slow but very high sweller' of oil films (Phenix 2013).Due to this high swelling ability, in current treatment practice, BnOH is rarely used neat (as a pure solvent) in order to prevent damage to the original layers.Rather, it is more commonly immobilized in a gelled matrix where its absolute quantity can be reduced (often used in 0.5-20 wt% loads) and where interactions with the gelling matrix (xanthan, Pemulen®, borax, Car-bopol®, etc.) and other factors (pH and additives) are used to enhance or dampen the overall strength of the cleaning solution (Wolbers 2000).However, sometimes adjustments to the gel medium, BnOH concentration, dwell time, pH, and additives do not lead to a solution that permits selective removal of a non-original layer while keeping the original intact, forcing conservators to find a another treatment option such as mechanical removal.

Approaching cleaning through chemical modifications to the BnOH core
An additional approach to optimizing the cleaning solution is to use cleaning reagents with modified core structures of a specific cleaning reagent (in the context of this study -BnOH).Structural modifications to BnOH would retain the core molecular architecture responsible for the desired cleaning effect (Figure 1 (a)), while utilizing additional chemical modifications (Figure 1(b)) to confer more nuanced chemical characteristics to the original molecule.Using the same general principle of 'like dissolves like' (Zhuang et al. 2021), greater specificity in chemical design is hoped to lead to an increased ability to distinguish between chemically similar layers.
The addition of functional groups such as hydroxyl (-OH), methoxyl (-OMe), fluoryl (-F), alkyls, aryls, etc. can enable a molecule to have more or less of specific noncovalent (i.e.intermolecular) interactions (hydrogen bonding, dipole-dipole, and London dispersion forces) as well as electronically modify the strength of the π-π interactions between the aromatic ring and oil film (Martinez and Iverson 2012; Salonen, Ellermann, and Diederich 2011).Modifications can also cause adjustments to the steric bulk, or three-dimensional shape, of a molecule (i.e.use the same or similar functional groups, but attached at different locations on the BnOH core), which can change a compound's diffusion rate and the accessibility of specific functional groups present.Both factors consequently impact the nature and rates of solvent action with the oil film (Zumbühl 2019).
In essence, modifying the structure of BnOH and gelling this modified compound in place of BnOH enables adjustment to the power of the cleaning mixture using a single molecule, rather than through adjusting the entire gel formulation.
This strategy can be advantageous for multiple reasons -1.Increased efficiency in testing: once conservators have already identified a gelling system that shows promise and have taken the time to familiarize themselves with the in situ feel of how that gelling system interacts with their painting, they can continue testing with more options within that single gelling system, 2. Increased efficacy for solvation: the chemical modification in the key cleaning reagent confers the possibility to find a reagent that has a true improved chemical match to the target paint film (contrasted with currently used manipulations like adjusting the wt% load of BnOH, which only changes the rate of interaction, not its chemical nature), and 3. Increased safety in testing: if a desired effect can be achieved through modification of the key cleaning reagent, it may preclude attempts to further admix additional additives into a gel mixture, which can be dangerous due to adjuvant effect phenomena (Somervaille et al. 2012).

This work
The aim of this research is not to replace BnOH, but rather to broaden the range of options of cleaning solvents conservators could reach for to increase their ability to tailor cleaning mixtures.It specifically explores a selection of five BnOH derivatives (defined as compounds that retain the BnOH core, but that incorporate additional functional groups on either the aromatic ring or at the benzylic position, Figure 2) and three related compounds (defined as those that feature the same or similar functional groups, but that no longer retain the benzylic relationship between the aromatic ring and alcohol, Figure 2).All assessed compounds are commercially available, affordable, non-bioaccumulative from an environmental point of view, and lower in toxicity compared to normal BnOH (except 2phenyl-2-butanol, which has the same level of toxicity as BnOH; see https://comptox.epa.gov/ for more detailed information about toxicity).
The eight chosen compounds collectively have diverse chemical propertiesfrom incorporating functional groups with greater polarity and capacity for hydrogen bonding (H-bonding), to groups with increased lipophilicity, to chemical alterations that change the molecule's steric bulk.They are examined alongside BnOH for comparative effect, empirically testing and documenting under a stereomicroscope.
Gas chromatography-mass spectrometry (GC-MS) is also used to assess the amount of time BnOH remains detectable in the paint layers after cleaning and clearing.Finally, the retention of the solvent in the treated paint layer and its possible oxidative chemical changes that might occur over time will also be monitored.

Selection of BnOH derivatives and related compounds
Many BnOH derivatives and related compounds are commercially available.The eight chosen to be evaluated in this study were selected for these specific reasons: 3-fluoro BnOH: additional H-bond acceptor (F); strong dipole from F expected to make the compound more polar and may have significant electronic effects on the p-system.
4-isopropyl BnOH: increased lipophilicity and steric bulk due to the isopropyl group.
2-(3-hydroxyphenyl)ethanol: additional H-bond donor and acceptor (-OH); additional -OH directly connected to p-system (thus probes effect of altering electronics of p-system).
All evaluated compounds were obtained from Sigma-Aldrich with reported purities of 97-99%, as colorless, freely flowing liquids at room temperature with the exception of 2-(3-hydroxyphenyl)ethanol, which arrived as a highly viscous, slightly yellowed, translucent compound.

Cleaning solution delivery methods
The BnOHs were used neat and as 5 wt% gelled emulsions in xanthan gum and Pemulen® TR-2, both at pH 7 (verified with a Testo 206 pHmeter).Neat solvent tests were done to better understand cleaning differences based on structural modifications alone, while the gelled emulsions were chosen as more representative ways of application in conservation.For details on the preparation of the pH 7 xanthan and Pemulen® TR-2 gels, see Supplemental Material.

Test painting
Cleaning tests were performed on fragments from a seventeenth-century Spanish oil on canvasa heavily damaged painting belonging to the Rijksmuseum's Paintings Conservation Department that is used for research purposes.The painting is unlined, with a coarse, simple tabby weave canvas support and red-pigmented ground.The painted layers are executed in oil, with certain passages having fairly thick impasto.A layer of heavily yellowed varnish covers the painted layers.This varnish is readily removed with acetone, though upon removal residual (presumably older) varnishes remain.Retouches and fills are present in multiple passages.Layer stratigraphy and chemical composition were documented through cross-section analysis and thermally assisted hydrolysis-methylation-pyrolysis-gas chromatography-mass spectrometry (THM-Py-GC-MS) (see Supplemental Material).THM-Py-GC-MS analysis of varnish and paint scrapings from these various testing areas confirmed that both retouches and original painted layers are oil-based, and indicated that the bulk varnish removable by acetone was dammar-based while residual varnish not removable by acetone was a Pinaceae resin-based material.These results are consistent with the timeline of the painting; diterpenoids (i.e.Pinaceae) dissolved in spirits of wine and turpentine were used since the sixteenth century (Osete-Cortina and Doménech-Carbó 2006), while dammar resin varnishes were introduced in the nineteenth century (Colombini et al. 2000).
Fragments of a historical painting was chosen over mockups due to the availability of a suitable test painting and recognition of the difficulty in making mockups that are truly representative of aged oil paintings.

Cleaning test locations and pre-treatment
Cleaning tests were performed on multiple locations of the painting: (1) White/pink retouching over white original, appearing to be from a single retouching campaign.The retouch is predominantly light pink, with passages of bright white mixed wet-in-wet (Figure 3 All locations were pre-treated by swab cleaning with acetone until no more varnish or retouch could be removed.

Stereomicroscope and imaging software
All cleaning tests were performed under a Zeiss Stemi 2000-CS table microscope equipped with an Axiocam 105 color camera.All images were taken and processed with the ZEN (blue edition) software.

Cleaning test procedure
Two cleaning/documentation methods were used in an attempt to standardize the cleaning process, as shown in Table 1.Both methods are more qualitative than quantitative, however they are consistent with how paintings conservators routinely judge the efficacy of their cleaning tests.Note that Yamazen ultra-fine tipped pre-rolled cotton swabs were used for the cleanings so that a photograph of lifted material on the swab after each clean could be documented; in typical practice, alternative cleaning instruments may be used such as soft, white nylon bristled brushes.
Method A: was used to target retouches covering original paint.For each cleaning test, the site was photographically documented before cleaning and after each cleaning interval until all of the targeted retouching was removed (Supplemental Material).Time intervals were 20 s for gelled emulsions and 5 s for neat solvent evaluations.In all tests, Yamazen cotton swabs were dipped ∼1-2 mm in gel or neat solvent and agitated on the surface for the entirety of the timed intervals.After each interval, the gel or solvent was cleared using spit-wetted hand-rolled swabs.Interval cleaning continued until all of the retouching was removed, or until a clear lack of progress warranted stopping.
As the amount of the targeted retouching present will affect the time needed for its removal, a digital annotation tool from the ZEN stereomicroscopy software was used to measure surface areas to aid in homogenizing the quantity of retouching targeted for removal.Judgement of cleaning efficacy was based on the total number of intervals needed and the feel during the cleaning process.Method B: was used to evaluate the sensitivity of original paintyellow and black-pigmented.For each test, a ∼2-3 mm diameter droplet of gel was placed on the paint surface for 1 min, and subsequently cleared with a spit-dampened swab.Note that this is not a standardized protocol but it was done to avoid leaving residues.This clearing swab was then photographed in order to compare the quantity of pigment lifted off.The process was repeated two more times over the same test location, to give a total of three documented clearing swabs per test.

Neat solvent evaluations
Neat solvent cleaning tests were conducted using Method A on regions of white/pink retouching covering white original paint.A summary of findings is presented in Figure 4 and photographs tracking the cleaning progress using each compound are provided in the Supplemental Material.The 3-fluoro BnOH and control BnOH were the fastest cleaning agents, followed by five compounds with moderate efficacy.Two compounds -2-phenyl-2-butanol and 2-(3-hydroxyphenyl)ethanolwere ineffective, having had no observable progress in retouching removal after eight cleaning iterations.

Xanthan and Pemulen® TR-2 gel evaluations
Six different evaluations were conducted to provide a general idea of relative cleaning efficacy between compounds using gelled emulsions -Method A with xanthan, Method A with Pemulen® TR-2, Method B with xanthan (two evaluations), and Method B with Pemulen® TR-2 (two evaluations).
Method A was used on white/pink retouches over white original (same type as those used in the neat solvent evaluations), and Method B was used on original yellow for the first evaluation and original black paint for the second.Full data and photographic documentation for all tests using Method A are provided in the Supplemental Material.Also included are testing results and photodocumentation of locations where initial cleaning with an ineffective compound was subsequently cleaned with one of the more effective derivatives until the retouching was fully removed.Such tests had the advantage of being able to see performance differences on the exact testing locations.Full results for Method B testing are provided in Figure 5.
Overall, the xanthan gelled emulsions cleaned slower than the equivalent Pemulen® TR-2 gel.This was experienced while cleaning using Method A, and was manifested by a greater degree of pigment pick up on clearing swabs (Figure 5). 1 However, the relative efficacies of the evaluated compounds appear to stay fairly consistent.
While the cleaning efficacy differed slightly between the six evaluations, there was a perceptible consistency across all tests (Method A and B combined).As with the neat solvent evaluations, the 3fluoro and control BnOHs were clearly the most powerful at solubilizing the retouching.These two fast-acting compounds are followed by a trio consisting of the 3methoxy BnOH, 1-phenylethanol, and 3-phenyl-1-propanol as the next fastest at solubilizing paint.Following this group, there is a significant decrease in solubilization power with the 4-isopropyl BnOH, 2phenyl-2-butanol, and 2-(3-hydroxyphenyl)ethanol gels making either slow or no progress in retouching removal using Method A, and faint to no amounts of retouching lifted using Method B. For the latter, in situ evaluation with the stereomicroscope strongly suggests that in cases where faint pigmentation is observed on the clearing swabs, the material that is  being removed is more consistent with residual aged varnish rather than pigment.And finally, in all six evaluations, the 4-penten-1-ol gel appeared completely ineffective for this test panel.These results are summarized in a qualitative ranking provided in Figure 5, which represents the combination of results from all six evaluations.

Discussion
One of the key goals of this study was to determine if use of BnOH derivatives and related compounds would result in subtle changes to cleaning efficacy, while staying within a single gelling parameter.The results from both neat solvent, and xanthan and Pemulen® TR-2 gelled emulsions demonstrate that the cleaning efficacy can be controlled.

General trends and rationalizations
From all tests, it is clear that the 3-fluoro BnOH was faster at solubilizing the oil-based paint layers tested in this study compared to traditional BnOH.It was also clear that all seven other derivatives and related compounds tested work slower than traditional BnOH, but that between them, the range of cleaning power spans from slower than BnOH to ineffective.3-fluoro BnOH was expected to be a stronger cleaning agent than BnOH and was chosen for evaluation based on fluorine's ability to behave as a highly polarizing functional group and additional H-bond acceptor (van der Lubbe and Guerra 2019).Moreover, the relatively minimal change in steric bulk of 3-fluoro BnOH compared to traditional BnOH means its ability to diffuse through the paint film is not significantly impacted.
All other assessed compounds feature notable to extreme increases in steric bulk.Thus their slower solubilization compared to BnOH is likely a combination of increased sterics, lack of electronic changes that increase polar non-covalent interactions (most of these seven molecules simply feature added hydrocarbons), and physical properties that would impact diffusion rates (see section on importance of emulsification below).
Comparison of 4-isopropyl BnOH and 2-phenyl-2butanol is particularly interesting because these two molecules are constitutional isomers.Thus their chemistries are extremely similar (same molecular weight, same number and distribution of C, H, and O atoms), but the way the atoms are connected to each other differs.Despite their similar capacities for Van der Waals and H-bonding interactions, there was a significant difference between the two.4-isopropyl BnOH was capable of removing the retouching in Method A testing, albeit slowly, but 2-phenyl-2-butanol was ineffective.This difference could be explained by the general extreme steric bulk of the 2-phenyl-2-butanol resulting from the fact that it contains a quaternary carbon, a fully substituted sp 3 -hydbridized carbon.Quaternary carbons result in dramatic impacts on steric bulk, thus its presence may render the entire molecule less capable of diffusing through the paint film, rendering it ineffective in cleaning.Alternatively, since the alcohol is bonded directly to the quaternary carbon, the lack of efficacy could be a result of inaccessibility to the alcohol functional group (even if the entire molecule can still penetrate into the paint film).Both these processes could also be acting in concert.Nonetheless, structure plays as important a role as adding new or changing the chemistry of functional groups.

Functional groups and their relationships
The effectiveness of 3-phenyl-1-propanol both in neat solvent and gelled emulsion tests demonstrates that a benzylic relationship between the ring and alcohol functional groups is not essential for cleaning.
Moreover, 4-penten-1-ol demonstrates that having an aromatic ring is also not necessary for effective cleaning.4-penten-1-ol was the next fastest reagent after traditional BnOH when used as a neat solvent, suggesting that having a single, non-conjugated π −system in combination with an alcohol is enough to confer an appropriate chemical match for swelling oil-based films.

Importance of emulsification
4-penten-1-ol also illustrates another important finding: the ability to emulsify in gelled form is critical for successful cleaning.While effective as a neat solvent, this compound was ineffective in both xanthan and Pemulen® TR-2 gel tests.This result is attributed to a lack of emulsification in the physical gelall other compounds (except 2-(3-hydroxyphenyl) ethanol) resulted in a milky-white, opaque gel that when examined under the microscope has clear droplets (the oil in water emulsion) dispersed throughout the body of the gel.However, the 4-penten-1-ol and 2-(3-hydroxyphenyl)ethanol gels remain translucent and clear, with no droplets in its bulk when examined under the microscope.
With respect to 2-(3-hydroxyphenyl)ethanol, this compound was also ineffective when used as a neat solvent, which is inconsistent with what would be expected based on how 3-phenyl-1-propanol and 4penten-1-ol behaved.However, as mentioned earlier, this specific reagent was the only one to arrive as a highly viscous compound.Thus, a physical inability to penetrate into the paint film is attributed to its ineffective cleaning when used as a neat solvent (and could also play a role in explaining why it did not emulsify once gelled).

Xanthan vs. Pemulen® TR-2
As previously mentioned, the Pemulen® TR-2 gelled compounds appear to result in stronger cleaning gels than the equivalent xanthan series.This is unsurprising considering the known ability of Pemulen® TR-2 to have powerful internal surfactant qualities (Ravenel 2010).
However, one of the initial reasons behind assessing two gel systems was to determine if significant changes in relative cleaning efficacy of the nine compounds would differ due to the chemistry of the gel (i.e.adjuvant effects, etc.).For the two gel systems studied here, there does not appear to be significant differences in the behavior of the BnOH derivativesthe fluorinated species is consistently the strongest cleaning agent followed by BnOH.While some minor differences were observed in how 1-phenylethanol, 3-phenyl-1-propanol, 4-isopropyl BnOH, 2-phenyl-2butanol, and 2-(3-hydroxyphenyl)ethanol ranked against each other, it is not possible to distinguish between differences resulting from inherent inhomogeneity of the historical painting test surface and chemistry of the compounds themselves.

Induction period
In situ observation while cleaning with Method A of both xanthan and Pemulen® TR-2 gels indicated that there may be an induction period before the gels begin to work.Namely, the first one or two iterations resulted in less solubilization of the targeted paint and less material cleared onto the swab compared to later iterations.Often consistency in how the gel was working and the amount of material cleared off the painting only occurred after the third or fourth iteration of cleaning, suggesting an induction period is needed for the gel to properly swell the non-original paint layer.
This apparent induction period was also corroborated in several cases during Method B testing, and can be seen in the photographs of the three clearing swabs wherein the first swab has less pigmentation than the latter two (Figure 5).
An induction period is logicalthe paint film needs some time upon first exposure to solvent to swell.It is important to keep in mind that a few repeats of testing on the same location may be needed to obtain a truly representative sense of a gel's cleaning power and solvent used.However, depending on the nature of the location (i.e. if original layers would also be impacted), repeat testing may not be possible.

Application in conservation
As the nature of the paint films on each painting will differ, having multiple options utilizing faster and slower cleaning reagents gives more control and options in testing.Naturally, these derivatives and related compounds will not prove useful in all situations, however the following two examples provide realistic examples of advantages (and points of caution) when using these new compounds for treatment.

Selective cleaning of oil-bound layers
Selective cleaning of oil-based retouch (confirmed by GC analysis, see Supplemental Material) without impacting the original oil layer was demonstrated using the 5 wt% 4-isopropyl BnOH pH 7 xanthan gel on flesh-toned passages of the seventeenth-century test painting.Initial testing using the traditional 5 wt % BnOH pH 7 xanthan gel showed that the original paint film was sensitive to this gel.As confirmation, Method B testing using the BnOH xanthan gel, resulted in clear removal of the original paint after three iterations of cleaning and clearly pigmented clearing swabs (Figure 6(a)).However, when the equivalent 4isopropyl BnOH gel was used on a passage of original, even after three iterations of cleaning, the paint film remains undisturbed and the clearing swab free of pigment (Figure 6(b)).While the clearing swabs from this test have faint pigmentation, microscope evaluation indicates there are no pigment particles, and that the material picked up is more consistent with residual varnish.
Critically, when the same 4-isopropyl BnOH gel was used on a passage of retouch on top of the same previously tested original flesh-toned paint, this slower acting gel was demonstrated to work well enough to remove the targeted retouch (Figure 6(c)).Here a total of six cleaning iterations was needed (6 min total exposure time to gel).Due to the similarity in pigmentation, it is not possible to prove that none of the pigmented material cleared off the painting came from the original paint film, however based on the results shown in Figure 6(b), it is assumed that the original is not affected.Moreover, stereomicroscope analysis of the cleaned surface did not indicate the original paint was visually affected.
It is important to keep in mind that this 4-isopropyl BnOH xanthan gel was strong enough to result in faint pigment pick up in tests on passages of original yellow and black, meaning it is not 'safe' in all cases.However, this example demonstrates the potential to find solutions to sensitive cleaning challenges by modifying the chemical structure of BnOH.

Treatment of Jan van Scorel's Portrait of a Man from Harlem
Select BnOH derivatives xanthan gels were tested during the treatment of Jan van Scorel's Portrait of a Man from Haarlem (Figure 7), a sixteenth-century Dutch oil on panel in the Rijksmuseum's collection.
This painting featured a ca. 1 cm thick arc of red overpaint along the painting's arched upper edge.The red of the overpaint was darker and cooler compared to the original red layer beneath.Extensive cleaning tests employing a wide range of gel systems (Pemulen®, Carbopol®, xanthan, borax, and Klucel®) were carried out due to the difficulty in finding a system that would solubilize the overpaint.Unfortunately, none of these tested gels clearly solubilized the targeted overpaint.
GC analysis of the overpaint indicated it had high oil content.Operating on the assumption that the overpaint was considerably oxidized and the knowledge that existing tests using traditional BnOH emulsified in xanthan had succeeded in clearing aged varnish (but not the desired overpaint), both of the more polar derivatives/related compounds from this study -3fluoro BnOH and 2-(3-hydroxyphenyl)ethanolwere tested as 5 wt% pH 7 xanthan gelled emulsion forms.
Consistent with its performance on the seventeenth-century test painting, the 2-(3-hydroxyphenyl)ethanol gel proved completely ineffective at cleaning.On the contrary, the 3-fluoro BnOH gel did succeed in removing the dark red overpaint.However, despite this result, the decision was taken to not remove the red overpaint due to the sensitivity and difficulty in distinguishing between the underlying original red paint.
However, the 3-fluoro BnOH when gelled in a 40PVA borax gel proved both safe for the original and effective for the reduction of remains of aged oil-containing varnish present in certain passages of the painting.Initial tests using a 70/30 BnOH:methanol 10 wt% 40PVA 1 wt% (40% hyrolyzed polyvinylacetate) borax gel (Angelova et al. 2013) showed it was capable of reducing the residual oil-containing varnish without visible impact to the original paint film beneath.Tests using the same borax gel employing 3-fluoro BnOH in place of the BnOH was found to give the same result, but in ca.one-third the time.In the end, due to limited quantity of the 3-fluoro BnOH gel, the residual varnish reduction was completed using the initial 70/ 30 BnOH:methanol borax gel.

Post-treatment solvent retention monitoring
One of the well-recognized physical properties of BnOH is its high boiling point (205°C at 1 atm) and  slow evaporation rate.These physical characteristics paired with its known ability to undergo autoxidation in air and visible light to benzaldehyde, and subsequently to benzoic acid (Sudareva and Chubarova 2006), prompted us to use two complimentary GC-MS methods (pyrolysis and headspace) to monitor whether conversion to benzaldehyde and benzoic acid occurs over time scales in which BnOH remains unevaporated from the paint film.Previous studies have already examined residues of BnOHs remaining in paint films using FTIR (Dorge 2004).However, there are no studies focused on understanding the chemical conversion of the solvent while it remains in the painting.In this case we are concerned that the conversion from the alcohol to benzoic acid could potentially result in salt formation with metal ions present in paint films that could adversely affect paintings in the long-term.Moreover, with the technological advances made in recent years, headspace solid phase microextraction coupled to GC-MS employed in this study has the sensitivity to track traces of residual solvents, that are not detectable with traditional GC injection methods.

HS-SPME-GC-MS analysis
The monitoring of benzyl alcohol and benzaldehyde in the painting was performed using headspace solid phase microextraction coupled to gas chromatography-mass spectrometry (HS-SPME-GC/MS) (Supplemental Material for experimental conditions).This analytical method is particularly well suited to track volatile components such as benzaldehyde, and has been employed successfully in recent years for noninvasive solvent retention studies in paint films (Prati et al. 2017;Prati et al. 2019).
Two separate pieces from the same seventeenthcentury Spanish test painting used in the cleaning tests were treated with 5% w/w pH 7 xanthan gel and Pemulen® TR-2 gels, and monitored over time for solvent release using HS-SPME.
Results showed that both benzyl alcohol and benzaldehyde are detected in the test painting at the first time point (1 hr after cleaning) with both tested gels, and over the course of six days, the amounts of both compounds drops to <5% of their original amounts measured at 1 h post-cleaning (Figure 8).A separate experiment (data not shown) showed that we still could detect benzyl alcohol and the aldehyde over two weeks after cleaning with the 5% w/w xanthan gel.In this study we show for first time that while residual amounts decrease dramatically over the first week, it is likely that benzyl alcohol remains for several weeks inside a paint film post cleaning with a 5% w/w gel (thus retention could be longer if gels of higher concentration are used).The detection of benzaldehyde at all time points suggests that oxidation occurs rapidly, in this case even before the cleaning gel is applied to the painting.

THM-Py-GC-MS analysis
THM-Py-GC-MS analysis was concurrently used for the detection of benzoic acid, a compound that is not volatile enough to be sampled by SPME without heating or the addition of salts to enhance extraction efficiency (Tungkijanansin et al. 2020).Control sampling of an uncleaned area of the seventeenth-century test painting indicated that benzoic acid was already present in the paint layers, or it may have been generated during the pyrolysis (Tanczos, Rendl, and Schmidt 1999).To avoid misinterpretation of the results, xanthan gelled with the 3-fluoro BnOH derivative was used in place of BnOH to monitor the possible formation of 3fluoro benzoic acid, and being able to distinguish this fluorinated acid from the benzoic acid already present in the paint film.
Using THM, both the methylated species of 3-fluoro BnOH and 3-fluoro benzoic acid could be followed.Monitoring over two weeks showed a similar rate of decrease in detected benzyl alcohol as observed in the HS-SPME results (data in Supplementary Materials).Additionally, monitoring of the acid revealed that its abundance is directly related to the presence of alcohol (Supplementary Materials).The fact that the acid is already detected at the beginning of the monitoring can be linked to the early oxidation of the alcohol, already taking place in the gel even before it comes into contact with the test painting.This was confirmed through control GC runs analyzing freshly prepared xanthan and Pemulen® TR-2 gels.

GC-MS result discussion
SPME-GC-MS results indicate that the solvent is retained in the painting and slowly released over time and that both types of gels have the same capacity for solvent retention.Due to the conditions of the sampling experiment, in which the vials remained opened between sampling in order to avoid autoxidation of the alcohol in the headspace of the vial, it is not possible to show the results in an accumulative way.However, it is important to highlight that these results show that the solvent and its related species can be retained in the painting, even two weeks after the treatment is concluded.
Py-GC-MS results suggest that the presence of the oxidized species, 3-fluoro benzoic acid, is linked to the concentration of 3-fluoro BnOH retained in the paint and that this conversion is already initiated before the gel is applied in the painting.

Conclusions
After evaluating eight benzyl alcohol derivatives and related compounds, this work has shown that more controlled cleaning efficacy can be achieved attributable to the structural modifications of the benzyl alcohol core.Importantly, this improved cleaning efficacy can in certain situations provide solutions to cleaning challenges frequently encountered by conservators (as demonstrated with the use of the slower 4-isopropyl benzyl alcohol cleaning on the seventeenth-century test painting, and promising results in testing the BnOH derivatives for the removal of later overpaint during the van Scorel treatment).As also seen in this work, many of the compounds evaluated, though slower acting than benzyl alcohol, can still be dangerous to original paint (as with tests on passages of original yellow and black on the seventeenth-century test painting).
While the goal of this work is to expand the toolbox of cleaning solutions available for conservators for enhanced chemical matching, and promising preliminary results were demonstrated, this approach also has its challenges.The process of knowing how to select a structural modification is non-intuitive for a conservator without extensive experience and knowledge in organic chemistry.Additionally, while all the compounds tested here are considered non-persistent and non-bioaccumulative, the use of benzyl alcohols inherently implies continued use of solvent-based cleaning.Note that potentially lower toxicity is not assured.Additionally, this method of cleaning involves use of chemicals that are proven to stay long in paint films before evaporating.Finally, the increasing chemical specificity of benzyl alcohol derivatives or related compounds can render them difficult to obtain for conservators working in smaller institutions or private practice.
Despite the shortcomings, benzyl alcohol is often used for the removal of older, insoluble, oil-containing overpaint/retouches or varnishes.Thus, having the ability to fine-tune the cleaning ability of benzyl alcohol is a useful approach.There are many other commercially available derivatives and related compounds that were not explored in this study, providing much room for further exploration.However, even if difficult to implement widely, we hope the results herein provide at the very least insight into how paint films can respond to subtle structural changes in chemical structure.

Note
1. Pemulen ® TR-2's greater cleaning effect appears subtle/may not be evident from photographs of swabs used for clearance in Figure 5, however the difference is clearer when viewing directly under the microscope.

Figure 1 .
Figure 1.Defining a BnOH derivative: (a) benzyl alcohol parent compound; and (b) positions available for substitution.

Figure 2 .
Figure 2. BnOH-like compounds evaluated in this work: benzyl alcohol derivatives (structural differences highlighted in pink) and related compounds with their abbreviated names and CAS registry numbers.

Figure 3 .
Figure 3. Cleaning test surfaces used in the seventeenth-century test painting: (a) white/pink retouching on white original; (b) yellow original; (c) black original; and (d) flesh-toned retouching on flesh-toned original.(Scale bars = 1 mm).

Figure 4 .
Figure 4. Results of neat solvent evaluations: ranking of cleaning efficacy as judged by a combination of the objective number of cleaning iterations (numbers below compounds) needed to fully clear the targeted overpaint, and the subjective feel of the solvent efficacy during cleaning; depicted scaling (pink dots) intended to reflect relative cleaning efficacies.

Figure 5 .
Figure 5. Results from gelled emulsion evaluations: photographed clearing swabs (left to right for each image -first, second, and third cleaning test on the same location) from 5% w/w loaded pH 7 xanthan and Pemulen® TR-2 gelled emulsion cleaning evaluations, and a summary of overall performance.

Figure 6 .
Figure 6.Example of selective cleaning enabled by BnOH derivatives.Photographs documenting the progression of cleaning from before treatment (BT, far left) to the cleaned surface and its respective clearing swab after iterative cleaning employing Method B. Cleaning progress using: (a) the traditional BnOH gel on original; (b) the 4-isopropyl BnOH gel on original; and (c) the 4-isopropyl gel on overpaint.

Figure 7 .
Figure 7. Jan van Scorel's Portrait of a Man from Haarlem.Normal light overall image.