The Ketone-Based Varnish Removal from an Oil Painting by Wassily Kandinsky: Comparison and Assessment of Cleaning Methods Through Preliminary Test on Mock-Ups and Multi-Analytical Investigation

Abstract

This paper presents the removal of a non-original varnish from the surface of a painting by Wassily Kandinsky based on prior experimentation carried out on mock-ups, which were made according to the original artistic technique and painting stratigraphy. Due to a generally serious state of conservation, the artwork underwent several treatments over the years that have changed its original appearance. This work focused on the study and characterization of the ketone-based varnish present on the surface to outline how this non-original film layer interacted with paint layers and increased deterioration phenomena. Aiming to identify the most suitable approach for the artwork, some preliminary cleaning tests were made on artificially aged mock-ups. A multi-analytical investigation was carried out through non-invasive and invasive techniques to support all steps of the conservation treatment. By comparing cleaning results on mock-ups, a suitable protocol was developed and applied to the original painting. The best results were obtained by using the Nanorestore Gel^® Dry MWR (Medium Water Retention) with ethanol. The project has shown that the dissolving power of the solvent can effectively be controlled and limited at the surface by confining it to the nanostructure of the gel, thus minimizing the risk of interaction with the original paint layers.

1. Introduction

The project was developed thanks to the cooperation between the Galleria d’Arte Moderna e Contemporanea di Bergamo (GAMEC), the owner of the artwork, and the Centro Conservazione e Restauro “La Venaria Reale” (CCR), aiming to identify the most appropriate approach to the conservation treatment of the Kandinsky’s painting. Other institutions (DAFNE-LNF of INFN CHNet; Department of Chemistry, University of Torino) were also involved in the study. In the first phase, a thorough scientific investigation of the artist’s color palette and pictorial techniques was carried out by means of a multi-analytical approach, including both non-invasive and micro-invasive techniques. The main results achieved were presented in a paper by Ricci et al. [1]. Now, this article describes the second phase of the project in which the operational choices to face the conservation treatment of the painting have been defined based on the characterization of both original and non-original materials and their interaction with each other. By evaluating the effectson mock-ups, several cleaning systems were tested and compared.

Spitz-Rund is an oil painting on cardboard (73 cm × 33 cm × 3 cm) created in 1925, during Kandinsky’s Bauhaus period, when the artist’s language evolved into “pure” compositions. The painting shows a balanced composition of circles, triangles and squares on a pink background. All the elements are crossed by sharp, curved black lines, which give the impression of movement to the composition. Kandinsky used a 4 mm-thick cardboard as a support and then he applied a uniform preparatory layer composed of proteinaceous glue and zinc white. The IR reflectography shows the presence of an underdrawing with thin lines, likely realized with a carbon-based pigment, outlining the main shape distribution (Figure 1).

In accordance with other paintings by Kandinsky, Spitz-Rund was made by using several binding media to obtain different chromatic and surface-finish effects. Using different types of binders and varnishes, Kandinsky aimed to suggest a sense of depth in the composition, where some shapes seem to pop up toward the viewer and others seem to move backward in an elaborated balance [1,2]. The specific effect desired by Kandinsky was crucial to leading the project as the presence of a non-original varnish, described below, covered the whole painting, hiding the original balance gloss of the surface. Micro-samples of the painting, as shown by Ricci et al., confirmed the use of drying oil as binding media, where several organic and inorganic pigments were dispersed in [1]. Moreover, the analyses showed the presence of natural resin (both a diterpenic resin and a shellac resin), which may have been used in addition to oil medium or as a varnish layer. The observation under the microscope of cross-sections from the painting did not clearly prove the presence of a separated natural resin layer over the oil paint, thus possibly supporting the use of a mixture of different binders; however, the presence of a thin, uneven and poorly visible layer on the surface cannot be fully excluded. The non-original varnish, which was intended to be removed during the present conservation treatment, was widespread on the painting and was initially detected through UVF photography, which showed a blueish UV fluorescence on the whole artwork surface. As already mentioned in Ricci et al. [1] and discussed in detail in Section 3.1, the varnish was later confirmed as a ketone resin through Py-GC/MS analysis.

Spitz-Rund underwent several conservation treatments over the years due to the sensitivity of the constituent materials and to compensate for the effects of traveling around on the occasion of international exhibitions. Indeed, in the period between the 1930s and 1970s, the painting was glued on an auxiliary wooden support to make its transport easier. Several delaminations of the paint layers and a dense craquelure, probably caused by thermo-hygrometric variations, were previously treated with consolidation and filling processes. It is likely that the application of the ketone resin varnish occurred after the formation of these cracks. The craquelure is dense and extends through all layers down to the paper support. Additionally, there is a pronounced color deformation phenomenon, which resulted in significant alteration of the surface. The ketone-based varnish that was applied further hid the irregular surface texture and sipped into some of the paint cracks (Figure 2).

The first patent for this class of resins dates back to 1926, but the ketone-based products spread in the field of conservation treatments only in the fifties and sixties of the 20th century in order to replace natural resins [3]. Therefore, the application of this varnish on Spitz-Rund was most likely accomplished from the 1950s onwards. The film layer showed several alteration phenomena like yellowing, bleaching and hardening. In addition to the degradation of the ketone varnish, a substantial amount of dust accumulated over time.

The considerable adhesion between the synthetic varnish and the paint layers is another of the aspects investigated in this research. This led to the conclusion that the need to remove the non-original varnish depended both on aesthetic and conservation choices.

The removal of organic film layers from pictorial surfaces is one of the most complex phases of a conservation treatment, as the original paint layers are stressed by the chemical, physical and mechanical forces exerted by the cleaning systems, hardly foreseeable or assessable without the aid of suitable scientific instruments.

The water sensitivity phenomena of oil paints of the twentieth-century concern conservators as surface cleaning treatment can be complicated or prevented by it. There are many examples of not-varnished, water-sensitive oil paintings from the 20th century reported in the literature, including works by Karel Appel, Jasper Johns, Robyn Denny, Wassily Kandinsky, Kazimir Malevich, Piet Mondrian, Clyfford Still, Paula Rego, Patrick Heron, Francis Bacon and Per Kirkeby [4,5,6,7,8,9]. In the case of Spitz Rund, in addition to the issues normally related to the traditional cleaning systems, we had to take into account the fragility and great variability in the thickness of the brushstrokes, both of the original and non-original paint layers. Therefore, given that the goal was to prove the efficacy of several cleaning treatments for the removal of a ketone-based varnish, reducing as much as possible the effects on the original paint, it was essential to provide an initial test phase on mock-ups that reproduced the original stratigraphy. In order to set up cleaning tests, the solubility parameters of ketone resin were deduced from previous works in the literature that investigated its properties and response to treatments after accelerated aging [3,10,11,12]. Several tests have shown that ketone polymers are initially more stable than natural resins but become more difficult to dissolve over time exposure. Moreover, their films become weak and brittle over time.

According to the data acquired during the preliminary tests on mock-ups, different classes of materials reported to be used in recent studies have been compared, considering different modalities and times of application, depending on the properties of the pictorial films. The purpose of the various tests was primarily to identify a system capable of supporting the solvent mixture efficacy on non-original varnish while reducing the risk of interaction with the original paint layers. Then, different tools and materials, such as sponges, clothes or papers, were tested to remove any cleaning residues. The effectiveness of the different cleaning procedures tested on mock-ups was assessed by means of several analytical techniques, including both non-invasive techniques and sampling of fragments for cross-section observation. Conversely, the assessment of the cleaning of the painting was carried out exclusively with a non-invasive approach.

2. Materials and Methods

2.1. Analytical Methods

A complete and integrated diagnostic campaign was designed to support all steps of the cleaning processes on mock-ups and then the conservation treatment on Spitz-Rund.

The mock-ups underwent thorough documentation with visible diffuse and raking light photography (Vis-D and Vis-R), ultraviolet-induced visible fluorescence (UVF), and deep observation and documentation of surfaces with video microscopy and 3D microscopy.

Curing of films was assessed by mean of Fourier-transform infrared spectroscopy (FTIR) analyses, and multi-layered samples were mounted as cross sections to shed light on the cleaning process through a careful study of the stratigraphy with optical microscopy (OM) and scanning electron microscopy with energy-dispersive X-ray spectroscopy (SEM/EDS).

On Kandinsky’s paint microsample, Py-GC/MS was carried out to characterize the organic components. The cleaning processes on the paint were assessed by means of deep observation and documentation of surfaces with ultraviolet-induced visible fluorescence (UVF), video microscopy and 3D microscopy and by means of Reflection FTIR Spectroscopy.

Further details on the equipment and experimental parameters employed are reported in the Supplementary Materials.

2.2. Recreating Original Stratigraphy—Mock-Ups and Artificial Ageing

In order to define the conservation strategies on the painting, several mock-ups were made reproducing the original stratigraphy investigated through the combination of multi-analytical techniques, as described above.

A cardboard Paper Watercolour Arches 100% cotton was chosen as a painting support and was prepared with a white layer of zinc white and animal glue [1]. Then, the mock-ups were divided into three series according to the binder of the pictorial layer, as shown in the following list (Figure 3).

• Type 1: a pink paint layer with an oil-based binder.
• Type 2: a pink paint layer with a mixture of oil and natural resin (shellac) as binders.
• Type 3: a pink paint layer with an oil-based binder, then covered with a natural resin layer.

The pink color was recreated with the combination of two pigments as investigated on Spitz-Rund [1]: an organic red pigment (Karmin oil 308) and a white barium sulfate pigment (lithopone) mixed to obtain the tone of the painting background.

The colored powder was dispersed in linseed oil, and for Type 2, 2% of shellac, previously dissolved in alcohol, was added.

To facilitate the drying process of the oil binder, the samples were divided into two groups: one kept in an oven at 40 °C, and the other exposed directly to outdoor solar radiation, both for one month. At the end of the treatment, a micro-sample of each of them has been taken to verify the cross-linking degree of linseed oil using FTIR analyses. The decrease of 3010 cm⁻¹ (=CH (v)) signal was followed to assess the beginning of the drying process [13]. Then, on the dried films for Type 3, after 20 days of thermal aging at 40 °C, the mock-up was covered by a shellac-based layer.

Subsequently, all of these were treated again in an oven at 50 °C for 60 days, and to recreate the conservative varnish layer, a solid ketone resin Laropal K80 dissolved at 30% p/V in a solution made of Shellsol D70 and n-butyl acetate (70:30 v/v) was used. The film was applied using brushes and airbrushes in several steps to simulate the thickness of the Spitz-Rund varnish.

Finally, to recreate the phenomenology of degradation present on the surface of the painting, the mock-ups underwent more accelerated aging cycles, alternating heat treatment in the oven (30 days at 60 °C) to UV irradiation in a Solar Box (250 h Suntest CPS, Heraeus, Germany, equipped with a Xenon lamp and a UV filter that absorbs wavelengths lower than 300 nm, was used to simulate indoor solar exposure. Irradiation was set at 765 W/m², and the maximum temperature on the sample was kept below 50 °C by forced air circulation).

Their condition were checked every week with visible and UV light imaging.

2.3. Cleaning Test Methods

2.3.1. Solubility Parameters

In order to explore the range of solubility of the original paint layers and the varnish film, some preliminary solubility tests were carried out by using pure solvents or mixtures of them embedded in a cotton swab. The cotton swabs were rolled on the surface for a few seconds and then checked by observation under both visible and UV light to evaluate the presence of any paint or varnish residues. Solubility tests allowed us to evaluate the behavior of pictorial materials and to build effective alternative methods of solubilization.

Based on the Teas graph, solvent blends with increasing polarity were prepared considering their solubility parameters (Fd, dispersion force; Fp, polar force; Fh, hydrogen bonding force) [3]. The solvents used, either pure or in binary blends, were iso-octane (Fd 100, Fp 0, Fh 0), ethanol (Fd 36, Fp 18, Fh 46) and acetone (Fd 47, Fp 32, Fh 21).

More specifically, the Fd value, which describes the polarity of the solvent, is used to compare and assess different solvent blends. In accordance with the literature [3], the range of solubility of the materials of interest for Spitz-Rund are as follows:

• Aged drying oil films: Fd 45–70.
• Aged natural resin films (shellac): Fd 30–54.
• Aged ketone varnish: Fd 36–96.

Therefore, based on these values and aiming to remove the ketone varnish without any interaction with the underlying painting layers, a safe range of solubility with Fd 70–96 should be selected.

However, it is important to remark that the above-mentioned parameters should be referred only to free solvents. The practice suggests to us that whether using solvents loaded into gel systems, the safety range can be slightly reconsidered.

2.3.2. Gel Systems

Normally, gelling materials for the cleaning of painted surfaces should be chosen taking into account several features, including the following:

-

Transparency;

-

Capacity of loading polar solvents;

-

Adaptability to uneven and porous surfaces;

-

Ease of removal after application.

Finding a compromise among the above-mentioned criteria, the availability of the products and the specific conservation needs of the artwork, we selected the following gel systems:

• Rigid agar gels: By exploiting the ability of water to create emulsions with solvents in part or completely immiscible, it is possible to make two phases, physically separated, coexist in the presence of a gelling agent such as agar [14]. A small amount of benzyl alcohol (10% w/v) was added to the 3% gelled aqueous solution with agar and then applied when the gel became cold and rigid.
• PVA-borax gel: High viscosity polymer dispersion that can be charged up to 15–30% with some organic solvents. The gel was prepared by mixing PVA (4% w/w in water) and borax (8% w/w in water) in a ratio of PVA: B = 5:1. The ability of PVA-borax to adapt to the surface and to be easily removed makes it particularly suitable for paintings with uneven reliefs and thicknesses [15,16]. In this case, some gels were prepared with the addition of ethanol or benzyl alcohol (up to 20% v/v).
• Nanorestore Gels^®: They are water-based chemical gels, which have highly retentive networks that reduce their action on the surface. They were developed by the Center for Colloid and Surface Science of Florence under the UE-funded research project named NANOFORART [17]. Hydrogels, Dry or Peggy, can be loaded with polar solvents or water-based nanostructured fluids of the Nanorestore Cleaning^® series and they can be safely used on particularly water-sensitive substrates. In our research, we tested Dry gel Medium Water Retention MWR, Dry gel High Water Retention HWR, Peggy 5 gel and Peggy 6 gel. First, they were loaded with blend solutions of either water and ethanol or water and methyl ethyl ketone. Secondly, among the Nanorestore Cleaning^® series, Polar Coating B (a nanostructured mixture of methyl ethyl ketone and 2-butanol) and G (a nanostructured mixture of methyl ethyl ketone, 2-butanol, ethyl acetate and propylene carbonate) showed to be more effective and were tested loaded in different Nanorestore Gel^®. Based on the observation of limited results achieved with Polar Coating G, it was tested only with MWR gel and Peggy 6 gel. Instead, Polar Coating B was loaded into all the above-mentioned gels.

According to the technical sheet information, the two nanostructured fluids do not have a quantifiable polarity value, but they are designed for the removal of aged synthetic varnishes. The gels were carried out by immersing them in the chosen cleaning fluid for at least 12 h, and it can be used up to 4–5 times. All instructions for use and technical information were found in the technical sheets of the products [18].

2.3.3. Removing Varnish Residues

A range of materials usually used in dry cleaning treatments were selected to improve and integrate the action of solvent swelling on varnish film. Within a project carried out at the Cultural Heritage Agency of the Netherlands (RCE, formerly known as ICN) that investigated different dry-cleaning products and through further studies highlighted in the literature, the following products were tested and compared [19].

• Microfiber fabric (Evolon CR^®): according to some recent studies that prevent the use of solvent-resistant fabrics during paint removal, some tests have been carried out combining the action of polar solvents with a microfiber fabric [20,21].
• Bondina^® is a non-woven polyester material, chemically inert and with an extra smooth surface. The white type with a thickness of 30 g/m² was employed.
• Japanese paper is a class of products widely used in the field of conservation because it contains few impurities, is neutral and shows high performances for long-term preservation. Different features of Japanese paper depend on manufacturing processes, raw materials or thicknesses. In this study, we used two types of Hiromi Japanese paper with a grammage of 21.4 g/m² (Sekishu Tsuru) and 40 g/m² (Hosokawa-shi).
• High retention sponges (Blitz-Fix^®): They can be hydrated in water or in polar solvents and are able to leave few particles when put on the surface [22]. In this study, the sponges were first softened in water and then carefully dried on absorbent paper before application on the painted surface. Following this, they were cut into small pieces and applied to the surface of the samples using metal tweezers. In some tests, a Japanese paper layer (Hosokawa-Shi) or an Evolon CR^® layer has been interposed between the sponge and the surface to facilitate the absorption of the swollen varnish.

The Evolon CR^®, Bondina^® and Japanese paper were interposed between the gel and the painted surface to increase the peeling effect of the gel, while sponges Blitz-Fix^® were used to dab the surface after the gel application to remove the swollen varnish residues.

2.3.4. Cleaning Tests

Table 1. List of cleaning tests.

Number	Solvent	Gel System	Cleaning Tool	Time of Application 2
T1	Polar Coating G	MWR	Blitz-Fix® 1	3′
T2	Methyl ethyl ketone/water (1:1)	Peggy 5	Blitz-Fix®	3′
T3	Polar Coating G	Peggy 6	Blitz-Fix®	3′
T4	Polar Coating G	MWR	Sekishu Tsuru	3′
T5	Polar Coating B	HWR	Blitz-Fix®	3′
T6	Polar Coating B	Peggy 5	Blitz-Fix®	3′
T7	Polar Coating B	Peggy 6	Blitz-Fix®	3′
T8	Polar Coating B	HWR	Bondina®	3′
T9	Ethanol (20% w/v)	PVA-borax gel	Blitz-Fix®	4′
T10	Polar Coating BWater	HWRHWR	Hosokawa-shiBlitz-Fix®	4′2′
T11	Polar Coating BWater	MWRMWR	Evolon CR®Blitz-Fix®	4′2′
T12	Benzyl alcohol (10% w/v)	PVA-borax gel	Sekishu Tsuru	4′
T13	Benzyl alcohol (10% w/v)	Agar gel (3%) cold	Sekishu Tsuru	4′
T14	Ethanol/water (1:1)	MWR	Blitz-Fix®	3′

¹ In each application, the sponge Blitz-Fix^® was hydrated previously with water. ² The time of application changed according to the gel system used: PVA-borax gel and agar gel (T9, T12, T13) required longer time to reach the varnish removal effect, while in tests T10 and T11 a two-steps procedure was requested.

shows the main tests of varnish removal carried out on the mock-up series. The list reports how solvents, gels and tools have been mixed to highlight the most suitable cleaning methods. Therefore, times of application are indicated to complete the discussion.

In the first stage of the experimental work, we kept the same operative conditions for all tests to allow a systematic comparison of the different cleaning procedures and the interaction of the gel system and tools used with the painted surface. These cases are not entirely discussed in this paper since we described only the procedures for which the most promising results were obtained. In the second stage, the operative conditions of these latter procedures were slightly tuned to optimize the removal of the varnish, as reported in Table 1. This experimental phase on mock-ups later proved to be essential to define the most appropriate cleaning treatment to apply to the painting.

3. Results and Discussion

3.1. Identification of the Non-Original Varnish

Py-GC/MS analyses on micro-samples allowed the identification of a ketone resin, a low-molecular-weight polymer whose properties change depending on the degree of polymerization. Figure 4 and

Table 2. Peak assignment and source material in the pyrogram of sample 2.

Peak No.	Retention Time (min)	Assignment	Source
1	5.06	Cyclohexanone	Ketone resin
2	5.09	1,3-dimethoxy-2-propanol	Siccative oil
3	5.14	1,2,3-trimethoxypropane	Siccative oil
4	5.33	1-methoxycyclohexene	Ketone resin
5	5.98	2-methylcyclohexanone	Ketone resin
6	9.84	Octanoic acid methyl ester	Siccative oil
7	10.35	Nonanoic acid methyl ester	Siccative oil
8	10.68	Hexandioic acid dimethyl ester	Siccative oil
9	12.07	Heptandioic acid dimethyl ester	Siccative oil
10	13.41	Octandioic acid dimethyl ester	Siccative oil
11	14.35	Dodecanoic acid methyl ester	Siccative oil
12	14.64	Nonandioic acid dimethyl ester (azelaic acid methyl ester)	Siccative oil
13	15.84	Decandioic acid dimethyl ester	Siccative oil
14	16.68	Tetradecanoic acid methyl ester	Siccative oil
15	16.80–17.08	Methylcyclohexanone dimers	Ketone resin
16	18.79	Hexadecanoic acid methyl ester (palmitic acid methyl ester)	Siccative oil
17	20.45	9-Octadecenoic acid methyl ester (oleic acid methyl ester)	Siccative oil
18	20.63	Laccishelloic acid trimethyl derivative	Shellac
19	20.68	Octatadecanoic methyl ester (stearic acid methyl ester)	Siccative oil
20	21.42	Jalaric acid tetramethyl derivative	Shellac
21	21.86	Shelloic acid tetramethyl derivative	Shellac
22	22.74	Dehydroabietic acid methyl ester	Diterpenic resin
23	23.27	Aleuritic acid tetramethyl derivative	Shellac
24	23.63	7-methoxydehydroabietic acid methyl ester	Diterpenic resin
25	24.07	15-methoxydehydroabietic acid methyl ester	Diterpenic resin
26	24.77	7,15-dimethoxydehydroabietic acid methyl ester	Diterpenic resin

show the pyrogram of sample 02, along with the peak assignment and source material.

Table 3. List of micro-samples investigated on Spitz-Rund. The symbols + refer to semi-quantitative quantification of components in samples: (+) low amount; (++) medium amount; (+++) high amount.

Localization	Sample Number	Type of Sample	Drying Oil	Natural Resin	Ketone Resin
	01	Multi-layered sample, paint and varnish layers	++	++	+++
	02	Multi-layered sample, paint and varnish layers	++	++	+
	03	Multi-layered sample, paint and varnish layers	+++	+	+

shows where the three micro-samples were taken from the painting and the main results of the analytical investigation. In order to reduce the number of invasive analyses, micro-samples were focused on the pink background of the painting. Since the three samples include all layers of the whole painting stratigraphy, all the organic materials (drying oil, natural and ketone resins) were detected in each sample.

3.2. Results of Cleaning Tests on Mock-Ups

The results of the artificial aging experiments reveal a slight phenomenon of cracking on the surface of the mock-ups, which bears a resemblance to the cracking observed in Spitz-Rund. However, the extent of this alteration appears to be less pronounced on a macroscopic scale in the artificially aged mock-ups. Our findings indicate that while the alterations observed in the artificially aged mock-ups are similar to those of the painting, the severity of the cracks is reduced. This suggests that the aging process used in the experiments may not fully replicate the long-term environmental and material conditions of the original paint layers. Future work will need to address this limitation by refining the aging protocols to more closely match the conditions under which natural aging occurs, thereby enhancing the fidelity of the artificial aging process and improving the correlation between experimental and case studies.

Before proceeding to highlight cleaning test effects, it is important to report how few differences in solvent sensibility were observed between the mock-ups of the three series. In any of the tests, the solvent action had caused a complete removal of the varnish layer to the point of altering the underlying paint layers, so no differences in behavior between the three series could be verified. It is possible that similar effects to cleaning tests depend on the kind of raw materials employed or the way in which the layers were prepared. Moreover, the aging may not have been sufficient to induce an interference between one layer and the other. In addition, tests made by rolling the cotton swab with pure solvents have shown that paint layers of mock-ups without varnish film have a tendency to swell with low polar solutions.

At first, the removal tests showed that water, also by setting the pH or the ionic strength, had no effect on the ketone resin, so any water-based methods have not been taken into account.

Table 4. Main results of cleaning tests on mock-ups.

Number	After Cleaning Test		Effectiveness 1	Observation
	Vis	UV
T1			+	Slight softening of the varnish film; the gel does not absorb the dissolved residues and leaves varnish marks. The sponge causes a bleaching on the surface.
T2			+	Uneven removal of the varnish film. Marks along the test edges. Gel becomes more stiffer after loaded with the solvent mixture.
T3			+	Uneven removal of the varnish film. Marks along the test edges.
T4			+	Partial softening of the varnish film. The paper leaves several fiber residues.
T5			++	Good softening of the varnish film that was partially migrated into the gel, but the sponge caused a bleaching on the surface.
T6			++	Slight softening of the varnish film. The surface after cleaning is quite even.
T7			+	Slight softening of the varnish film, varnish marks along the test edge.
T8			+	The swollen varnish is partially absorbed into the gel system, but there are several residues on the surface.
T9			+	High softening of the varnish film but uncontrolled dispersion of the solvent on the surface. The surface, after cleaning, is bleached and uneven.
T10			+++	Few paper fiber residues on the surface. Adding a next mechanical action through using the sponge, the cleaning effect is good.
T11			+++	Good softening of the varnish film through using the tissue. The next step with the sponge supports the resin removal.
T12			++	Uncontrolled dispersion of the solvent that has a high interaction with the paint layers. The paper leaves some fiber residues and the surface after cleaning is uneven.
T13			++	Slight softening of the varnish film, partially absorbed by the gel. The test causes some bleaching on the surface.
T14			+++	Good softening of the varnish film. The surface after cleaning is even.

¹ Symbols express the effectiveness of the cleaning test in a range from (+) acceptable to (++) good to (+++) very good.

shows the main results of cleaning tests made on the Type 3 mock-ups, where the assessments were carried out by comparing Visible diffuse light photography and UV light photography. Based on these results, micro-samples of some of them were taken for cross-section analyses.

The tests have been successful in identifying both advantages and disadvantages for each cleaning method when applied to the paint stratigraphy, so how to suitably combine different kinds of solvents, gelling systems and cleaning tools.

Test 13 was carried out to investigate the high performance of rigid agar gels to incorporate into their mesh any particles dissolved, reducing any after treatment on the surface to remove the residues. The test has also verified that polysaccharide-based gel is able to control how the solvent is delivered and its solubility action during application. By using an emulsion water-in-oil, the advantage is that the two phases (water and benzyl alcohol) coexist but are distinct. This aspect is crucial in cleaning treatment to avoid an uncontrolled and non-selective action of a solvent mixture on paint materials. The final effect of Test 13 on the surface appears good in visible and UV light, but cross-section analysis shows that the thin shellac film under the ketone varnish has been partially altered (Figure 5). Due to its aromatic component and its polarity, benzyl alcohol can induce shellac and natural resin swellings, as observed in our test. Despite the effectiveness of the rigid gel, the high degree of retention and the medium volatility of the solvent may not be suitable for the underlying paint layer’s safety.

In the case of PVA-borax gel-based tests (T9; T12), it is difficult to control the gradual release of the solvent loaded, although the gel shows a high ability to adapt to uneven surfaces. The fast dispersion of polar solvent on the mock-up causes varnish marks and bleaching surface, especially along the edge of the gel application. By adding a paper layer between the painting surface and the gel, the effect is more even and satisfactory but not sufficiently controllable.

However, the nanostructured gel-based series shows good performances in dissolving the thin layer of varnish without affecting the underlying pictorial materials. To assess their effective capacity to incorporate swollen paint materials when applied on the surface, Dry and Peggy gels were further evaluated under the same conditions, including the same solvent loaded and application times (Polar Coating B, application time of 3′). Results are shown in Figure 6. Through UV light imaging, it is possible to clearly identify that the Dry gels were significantly more effective in removing the softened ketone varnish compared to Peggy gels. Moreover, the advantage of these gels’ transparency allows the monitoring of solvent action during application. These findings highlight the potential of these gels in applications requiring effective paint removal.

In general, all tests involving nanostructured gels on the samples yielded satisfactory results by visual control. However, a comparison of the cross sections from Test 8 and Test 14 reveals a notable difference in outcomes (Figure 7). In Test 8, it appears that the solvent was absorbed to such an extent that it led to the alteration of the underlying shellac layer. This indicates that the cleaning method might have affected not only the varnish film but also the integrity of the layers beneath it. In contrast, Test 14 exhibited highly satisfactory results. The paint was effectively thinned without causing damage to the underlying paint layers. This outcome suggests that the cleaning method used in Test 14 achieved a more controlled and selective action, preserving the integrity of the underlying layers while removing the varnish film. These findings highlight the effectiveness of nanostructured gels in certain conditions and emphasize the need for precise control over solvent interaction to avoid unintended alterations to underlying materials.

Another aspect observed during the tests is the tendency of Polar Coating B and Polar Coating G to cause bleaching or matting of the surface. This issue was particularly evident when these solvents were used coupled with the nanostructured gels. The bleaching effect suggests that the solvents may have interacted with the surface altering its appearance, possibly by damaging the surface finish or by affecting the optical properties of the paint. This observation highlights a potential limitation in the application of the two nanostructured fluids, which could impact the overall aesthetic quality of the treated surfaces.

Although all the tested gels loaded with solvents were effective in varnish removal, they all showed a tendency to leave on the surface some swollen varnish residues, which required the interposition of Japanese paper or fabric layer and an additional cleaning step with the use of sponges.

Indeed, the tests proved that the interposed layer of paper or fabric between the gel and the surface enhanced the migration of softened ketone varnish within the cleaning system. Nonetheless, it is worth noting that the use of highly volatile solvents can be challenging, as their rapid evaporation leads to the quick drying of the varnish, thus causing the paper or fabric to adhere to the mock-ups surface. Focusing on the Blitz-Fix^® sponges, they effectively permitted the capture of softened varnish without leaving behind residues. The use of small sponge pieces allows for a controlled and gradual removal process, enhancing precision and minimizing the risk of damaging the underlying surface.

3.3. Removing Varnish on Spitz-Rund

Based on the cleaning procedure assessed on the mock-ups, we intended to follow the same approach for the cleaning of Spitz-Rund. The starting point was the evaluation of the solubility of the non-original varnish by applying pure solvents on the surface through a cotton swab. By gradually increasing the polarity force of solvent blends, the range of swelling of the ketone varnish was confirmed to be included in the range previously considered [3]: indeed, due to the natural aging of the artwork, the ketone varnish of Spitz-Rund showed a tendency to be softened with solvent blend with an Fd lower than 87. Since the solubility action of ethanol on the painting proved to be more effective than other pure polar solvents or nanostructured fluids, it was tested and loaded into three kinds of gelling systems, as listed below:

• Nanorestore Gel^® Peggy 5 + ethanol/water (1:1).
• Nanorestore Gel^® MWR + ethanol.
• PVA-borax gel + 20% w/v ethanol.

Although the PVA-borax gel and Peggy 5 gel did not provide optimized results on mock-ups, they were included again among the products tested directly on the painting since they proved to be more respectful of the surface compared to other systems previously tested.

Test 2 (Nanorestore Gel^® MWR + ethanol), as already assessed on mock-ups, was the most effective in removing varnish film without bleaching (as happens for Test 1, Nanorestore Gel^® Peggy 5 + ethanol/water (1:1)) or leaving residues (Test 3, PVA-borax gel + 20% w/v ethanol). The loaded gel is kept on the surface only for 20–30 s and then removed by delicately peeling it with tweezers. The time of application was shortened in comparison to the tests on mock-ups to further ensure the safety of the artwork.

Another advantage of the use of Dry gel is that it can be cut down following the outline of the colored shapes of the painting, varying the time of application according to each specific paint layer. Thanks to the combination of the swelling power of the alcohol and the capillary absorption capacity of the nanostructured system, a good removal of the ketone varnish was successfully achieved. After the application of the gel, the surface was accurately observed with the support of non-invasive survey instruments such as a 3D microscope. The images acquired (Figure 8) show that the original paint surface was unaltered, but there still were several varnish residues.

Based on these considerations, a second cleaning step with the use of a Blitz-Fix^® sponge was applied. However, the result obtained was not fully satisfactory, and to reach a complete paint removal, another step was added. A cotton swab loaded with a low-polarity solvent mixture (iso-octane: ethyl alcohol = 80:20 v/v) was rolled on the surface. Furthermore, it has to be kept in mind that the Blitz-Fix^® sponge could not have been used for this purpose because it does not support apolar solvents.

The combination of these steps resulted in a satisfactory outcome upon visual inspection (Figure 9). Subsequently, the test area was analyzed by means of UV photography, 3D microscopy and Reflection FT-IR spectroscopy [23].

Reflection FT-IR spectroscopy measurements on the painting surface before cleaning showed the characteristic bands of ketone resin, areas K01 and K02 (1710, 1450, 1025 and 945 cm⁻¹, [23]), that are not present in the spectra acquired on the surface after removal treatment, areas K03 and K04, confirming the effectiveness of the treatment. In these latter areas, the analysis identified the presence of a lipid substance used as a paint binder (Figure 10).

In conclusion, every step of the cleaning treatment of Spitz-Rund was carefully carried out with the aid of a microscope. This approach enabled precise and detailed examination of the painting surface, ensuring respectful and effective removal of accumulated dust and non-original varnish while minimizing the risk of further damage to the delicate paint layers.

The overall removal of paint represented an opportunity to recover the original reflection/refraction index of the surface of the geometrical composition (Figure 11).

4. Conclusions

Due to the uniqueness and fragility of Kandinsky’s masterpiece, an initial test phase on mock-ups proved to be essential in defining the most appropriate procedure for the removal of a non-original varnish. A multi-analytical approach allowed us to identify the varnish as a ketone resin and to assess the effectiveness of several gelling cleaning systems. Based on the results of this preliminary phase, only minimal tuning of the cleaning procedure was requested to safely carry on the conservation treatment of the painting by minimizing the risk of interaction with the original paint layers.

Thus, it was possible to optimize a multi-step cleaning process on the painting: first, swelling of the varnish using a Nanorestore Gel^® MWR + ethanol; secondly, removal of the residues by coupling a dry cleaning with a low-polarity solution applied by a cotton swab.

Future research might focus on understanding the mechanisms behind some surface alterations that can occur during the cleaning process and be observed during the test phase on mock-ups, such as the bleaching or the matting of the varnish.