Introduction

The study is focused on Giacomo Balla’s “Ritratto d’uomo”/Eugenio Riva (1900), a work created during his Divisionist period when Balla refined his pointillist techniques to depict fine textures and subtle light effects. Balla (1871–1958), a pivotal figure in Italian art, transitioned from realism and neo-impressionism to Futurism, influencing the development of 20th-century modernism. Archival records from the National Gallery of Modern and Contemporary Art in Rome indicate that Ritratto d’uomo received a cleaning in the 1960s to eliminate discolored surface residue. Although food-based cleaning approaches fall outside modern conservation practice, they were occasionally employed in the past due to their mild ability to loosen surface grime without damaging fragile paint layers. Materials such as breadcrumbs, potatoes, and onions were historically used to polish and brighten painted surfaces. Onions, in particular, were valued for their acidic and polysaccharide-rich composition, which could enhance surface gloss, but also carried risks of abrasion and residue retention1,2,3,4,5. Arabinoxylans, which can be identified in onions and other products such as bread, may have helped to the cleaning effect due to their polysaccharide structure, which is recognized for gently removing dirt and residues6.

Cellulose is a linear polymer composed of β-(1,4)-D-glucopyranose units, forming tightly packed, crystalline structures through extensive hydrogen bonding. This regularity and crystalline organization render cellulose highly insoluble in most solvents7. In contrast, onion polysaccharides have branched structures, often incorporating arabinose side chains. These branches disrupt polymer regularity, preventing the formation of tightly packed crystalline domains. As a result, intermolecular hydrogen bonding is weakened, allowing organic solvents to penetrate and interact more effectively with the polysaccharide chains8. The branched structure of onion polysaccharides increases the number of exposed functional groups, creating more interaction points for organic solvents. Their reduced “excluded volume” compared to linear polysaccharides of similar molecular weight further enhances solubility9,10. Since the 1970s, hydrolytic enzymes, or hydrolases, have gained popularity in conservation, initially in paper conservation and later in other fields. These enzymes catalyze the hydrolysis of proteinaceous, polysaccharide, and lipid-based materials that can alter the appearance and structure of artworks11. Amylases, for example, target α-(1–4) glucosidic bonds in starch, enabling the effective removal of starch-based adhesives without affecting cellulose. α-Amylases cleave α-(1–4) bonds to progressively degrade starch, while β-amylases break down the polymer into shorter starch fragments12. Amylase hydrolyzes glycosidic bonds in polysaccharides, acting on both branched and linear structures with varying efficiency. Alpha-amylase, an endoenzymatic enzyme, randomly breaks α-1,4 bonds within chains, making it effective on branched substrates like amylopectin and linear ones like amylose, which it degrades into maltose and linear dextrins13. Beta-amylase, an exoenzymatic enzyme, cleaves maltose from the non-reducing ends but is ineffective at α-1,6 branching points14,15. Gamma-amylase hydrolyzes both α-1,4 and α-1,6 bonds, acting on both types of structures, though with a weaker preference for linear substrates compared to the other isoforms16.

However, enzymatic cleaning of thin polychrome paint layers presents challenges, as amylases can cause over-wetting or excessive penetration into delicate paint layers. Uncontrolled enzymatic activity may weaken or alter the paint structure. Additionally, enzymes such as amyloglucosidases and pullulanase, which act on both α-(1–4) and α-(1–6) linkages, may leave residues within paint layers, complicating future conservation efforts17,18,19,20. To mitigate these risks, enzyme-based applications on artworks are often conducted with gel supports that control the enzyme’s cleaning action. However, the necessity of a water-based medium for enzyme use can further exacerbate risks for sensitive materials21. Given these concerns, this study investigates alternative methods for safely removing the onion-based layer from the painting. While enzymes align with green chemistry principles, their application in restoration may not be ideal due to the sensitivity of certain painted surfaces. To confirm the nature of the onion-based residue, a multi-analytical approach was employed. Fourier-transform infrared (FTIR) spectroscopy and scanning electron microscopy (SEM/EDS) were used to determine the composition and thickness of the paint layers. To strengthen the identification of the organic material present on the surface of the painting, in addition to spectroscopic and morphological analyses, a DNA analysis was conducted. Additional analyses, including multispectral imaging (VIS, UV 365 nm, IR 960 nm), digital microscopy, and X-ray fluorescence (XRF), were conducted to assess the painting’s state of conservation and Balla’s palette, minimizing interactions between selected solvents and the paint layer. These analyses revealed significant bioactive onion compounds, including polysaccharides, sulfur-containing compounds (such as onion in A and cysteine sulfoxides), and phenolic compounds like rutin and quercetin22,23,24. Due to the polarity and bonding of these compounds with the paint layers, their removal typically requires solvents such as ethanol or ethyl acetate, which are also widely used in conservation contexts and considered relatively sustainable. However, this study explores alternative cleaning systems that further reduce solvent exposure and improve selectivity, particularly in the context of sensitive painted surfaces25. This research offers new insights into the unconventional use of onions in historical cleaning techniques—documenting, this method applied to a painting—and presents a approach to exploring alternative green cleaning solutions to enzymatic methods.

Methods

Sampling and analytical techniques

A multi-analytical approach was employed to confirm the composition of the residue, combining optical microscopy (OM), FTIR spectroscopy in Attenuated Total Reflectance mode (FTIR-ATR), and scanning electron microscopy with energy-dispersive spectroscopy (SEM-EDS).

  • Digital Microscopy: A DinoLite AM411-FVW digital microscope was used at 40× and 220× magnifications under visible and UV light to identify areas of interest and determine sampling locations for invasive analysis.

  • FTIR-ATR Spectroscopy: Conducted with a Nicolet Summit FTIR spectrometer equipped with an Everest™ Diamond ATR accessory, with a resolution of 8 cm−1 and 32 scans per sample, identifying organic compounds potentially indicative of onion residues.

  • SEM-EDS Analysis: Performed using a Tescan microscope with INCA 2000 EDS, operated in low vacuum mode (15 Pa) at 30 kV, eliminating the need for sample pre-treatment. This provided elemental composition and morphological analysis of different paint layers and residues, including high-resolution imaging of brownish particles in untreated areas.

Four samples were taken from the painting to ensure a representative analysis. Sample S1, collected from the surface, was analyzed specifically to confirm the presence of onion-derived compounds. Although anecdotal accounts mention the use of onion in past cleaning treatments, no onion-specific compounds have been identified. Any potential residues would likely consist of general plant-derived substances such as polysaccharides (e.g., arabinoxylans), simple sugars, or sulfur-containing compounds, which are not unique to onions and are unlikely to survive or remain detectable after extensive aging and conservation interventions.

The remaining three samples—S2, S3, and S4—were sampling from the edges of the artwork to examine the stratigraphic composition of the paint layers and potential interactions with the residue.

Molecular Analysis

To confirm the presence of onion residues on the painting surface, DNA was extracted from residual powder scraped from the area of the canvas adhered to the frame structure. Due to the high degree of degradation of the sample, a modified CTAB (Cetyltrimethylammonium Bromide) method, optimized for the purification of highly degraded DNA, was applied. The entire powder sample (7.8 mg), from the edge of the painting was suspended in 500 µL of preheated CTAB buffer at 65 °C with the addition of 2-mercaptoethanol to minimize biomolecule oxidation.

After an incubation of 30 min, the DNA was purified using phenol-chloroform extraction, followed by ethanol precipitation to concentrate nucleic acid sequences. To remove secondary contaminants, the DNA underwent further purification using silica-based columns (QIAamp DNA Mini Kit, Qiagen), repeating the process twice to optimize yield. The quality and quantity of the extracted DNA were assessed using UV spectrophotometry (Nanodrop 2000, Thermo Fisher Scientific) and fluorometry (Qubit 4 Fluorometer, Invitrogen).

Due to the high level of degradation, DNA concentrations were below the optimal threshold for direct sequencing analysis, necessitating PCR amplification. DNA amplification was performed using nested PCR, an approach aimed at optimizing the detection of degraded DNA. Specific primers targeting Allium cepa were selected, amplifying highly conserved regions of the rbcL, matK, and ITS (Internal Transcribed Spacer) genes, which serve as distinctive phylogenetic markers. Amplification reactions were conducted using a Veriti 96-Well Thermal Cycler (Applied Biosystems). The reaction mixture for the primary PCR contained a final volume of 25 µL, including 2 µL of DNA, 1X Taq buffer (Thermo Fisher Scientific), 2.5 mM MgCl2, 0.2 mM dNTPs (Invitrogen), 0.5 µM of each primer, and 1 U of Taq polymerase (Platinum™ Taq DNA Polymerase, Invitrogen). For the secondary PCR, 1 µL of the primary PCR product was used as a template for a second amplification with internal primers. The thermal cycling conditions included an initial denaturation at 95 °C for 5 min, followed by 40 cycles of 94 °C for 30 s, 55 °C for 30 s, 72 °C for 45 s, with a final extension at 72 °C for 5 min. Amplified products were analyzed on 2% agarose gel stained with SYBR Safe (Thermo Fisher Scientific) and compared to a 100 bp ladder (GeneRuler 100 bp DNA Ladder, Thermo Fisher Scientific) to confirm expected band size. Bands corresponding to the target fragments were purified using the QIAquick Gel Extraction Kit (Qiagen) and sent for Sanger sequencing (Eurofins Genomics).

The identity of the obtained sequences was verified through comparison with the GenBank (NCBI) database using BLAST (Basic Local Alignment Search Tool). Finally, to assess the similarity of the sample to modern onion cultivars, the obtained sequences were aligned using MUSCLE and analyzed through phylogenetic methods (Neighbor-Joining, UPGMA) implemented in MEGA X. The reliability of phylogenetic relationships was tested using bootstrap analysis with 1000 replicates.

Imaging and Non-Invasive Elemental Analysis

Beyond the invasive techniques, non-destructive analyses were carried out to further evaluate the painting’s condition and material composition.

Multispectral imaging, encompassing visible (VIS), ultraviolet (UV), and infrared (IR) reflectography, was employed to assess the presence of varnish layers, retouches, and potential preparatory sketches. Images were captured using a Madatec Samsung NX500 28.2 MP BSI CMOS camera with a HOYA UV-IR cut filter and a Yellow 495 filter for fluorescence capture. The near-infrared imaging at 850 nm proved particularly useful in detecting underlying alterations or concealed layers.

XRF analysis was conducted at 18 different locations on the painting to characterize the elemental composition of the pigments and potential contamination from previous restoration efforts. The analysis was performed using a portable XRF instrument equipped with a tungsten X-ray source and a Peltier-cooled silicon drift detector. Operating at 38 kV and 350 μA, the system allowed for the identification of both light and heavy elements, with data processed using PyMCA software for semi-quantitative analysis.

Cleaning Tests

To evaluate the most suitable method for removing the onion-based residue while preserving the integrity of the original paint layers, different cleaning solutions were tested. Among these, Nasier C03, amylase, and amylase in Klucel® G demonstrated selective removal through targeted enzyme-based actions capable of breaking down polysaccharides26,27. Nanorestore Cleaning® Polar Coating S and Polar Rescue Gel, on the other hand, are solvent-based systems featuring low-toxicity, eco-friendly organic solvents. Nanorestore contains anionic surfactants in a blend of 1-pentanol, ethyl acetate, and propylene carbonate, developed by CSGI for removing natural and synthetic polymers with reduced solvent content. Polar Rescue, developed by Lab4green, incorporates anionic surfactants and acetal, gelled with Klucel® G to limit solvent spread on the paint surface28,29,30.

The following formulations were tested (Fig. 1, area C):

  • Deionized water

  • Polar Rescue in Klucel® G (5 wt%)

  • Nanorestore Cleaning® Polar Coating S (containing 1-pentanol, ethyl acetate, and propylene carbonate)

  • Amylase (1 wt% in deionized water)

  • Amylase (1 wt% in deionized water, thickened with Klucel® G 5 wt%)

  • Nasier C03 (a stabilized enzyme-based hydrogel)

Fig. 1: Intervention points on “Ritratto d’uomo”/Eugenio Riva.
Fig. 1: Intervention points on “Ritratto d’uomo”/Eugenio Riva.
Full size image

The Figure shows analysis and testing areas: XRF points (green), sampling sites (yellow, S1–S4), and cleaning tests (red, CT).

Each cleaning system was tested for 1 min on a predefined area of the painting. The enzyme-based treatments, particularly the amylase solutions, were expected to selectively break down the onion polysaccharides without affecting the underlying paint layers. Meanwhile, solvent-based systems, such as Nanorestore Cleaning® Polar Coating S, were assessed for their ability to dissolve the residues efficiently while minimizing solvent penetration and mechanical stress.

Evaluation of cleaning efficacy

The effectiveness of each cleaning method was assessed using both qualitative and quantitative criteria. High-magnification images were captured before and after cleaning using the DinoLite AM411-FVW digital microscope, allowing for a detailed comparison of surface changes. Additionally, spectrocolorimetric analysis was performed using a Y3060 3nh spectrophotometer, measuring color shifts in the cleaned areas under a D65 illuminant. The measurements were taken three times per area to minimize uncertainty, with color variations calculated using the standard ΔE* formula:

$$\Delta E=\sqrt{{\left({L}_{2}-{L}_{1}\right)}^{2}+{\left({a}_{2}-{a}_{1}\right)}^{2}+({b}_{2}-{b}_{1})}$$

This provided an objective evaluation of any perceptible alterations caused by the cleaning agents. Further analysis of the cleaning swabs was conducted using FTIR-ATR to detect any residues left behind by the various solutions. Except for the amylase solution in Klucel® G, the swabs were dried at 60 °C before analysis, ensuring that any remaining compounds could be properly identified

Figure 1 illustrates the different areas subjected to XRF analysis, sampling, and cleaning tests. Based on the comparative evaluation, the most effective method for residue removal was determined to be a 1% amylase solution gelled with Klucel® G (5 wt%). This formulation provided optimal control over moisture exposure and enzymatic action, effectively breaking down the polysaccharide residues without compromising the paint layers. The final cleaning process was applied systematically across the painting in 4 × 4 cm sections, with each treatment lasting 1 min. To prevent excessive moisture absorption, two layers of Japanese paper were placed between the gel and the painted surface, ensuring controlled application and residue removal. This methodological approach facilitated the safe restoration of Ritratto d’uomo/Eugenio Riva and provided valuable insights into the specific challenges encountered when cleaning painted surfaces affected by historical restoration residues, particularly in works characterized by thin, delicate paint layers. By integrating a multi-analytical assessment with carefully controlled cleaning procedures, this study contributes to a broader understanding of conservation techniques suitable for modern and contemporary artworks.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

Identification of the onion-based residue

The analysis of Ritratto d’uomo/Eugenio Riva under OM provided valuable insights into both the artist’s technique and the condition of the pictorial surface. The brushwork, characterized by briskly applied, juxtaposed strokes, was observed to create vivid visual effects, consistent with Balla’s Divisionist style (Fig. 2a, b). In untreated areas, a homogeneous brownish layer was detected, covering the surface uniformly (Fig. 3a, b). This residue, presumed to be a remnant of past restoration treatments, was absent from the areas that underwent cleaning interventions in the 1960s. Upon closer examination, the thickness of the pictorial layer was found to be remarkably thin, particularly in the upper-left region, where the underlying canvas was visible in several areas. This feature was even more pronounced on the right side of the painting, where the prior restoration had been carried out. Microscopic analysis reveals that the cleaning treatment carried out in the 1960s led to partial abrasion of the paint layer (Fig. 4a, b). While solvents do not directly abrade the surface, they can cause the paint film to swell and soften, making it more prone to damage from mechanical actions such as swabbing.

Fig. 2: Optical microscopy analysis of “Ritratto d’uomo”/Eugenio Riva.
Fig. 2: Optical microscopy analysis of “Ritratto d’uomo”/Eugenio Riva.
Full size image

The Figures show the uncleaned facial area under visible light (a) and UV (b). Brownish surface residue is clearly visible.

Fig. 3: Optical microscopy analysis of “Ritratto d’uomo”/Eugenio Riva.
Fig. 3: Optical microscopy analysis of “Ritratto d’uomo”/Eugenio Riva.
Full size image

The Figures show brownish layer in untreated area under VIS (a) and UV (b) with a homogeneous residue with UV fluorescence.

Fig. 4: Optical microscopy analysis of “Ritratto d’uomo”/Eugenio Riva.
Fig. 4: Optical microscopy analysis of “Ritratto d’uomo”/Eugenio Riva.
Full size image

The Figures show the area cleaned during the 1960s restoration 1960’ in VIS (a) and (UV) (b) lights at 50x magnification. UV image reveals abrasions and irregular surface texture.

The comparison between untreated and cleaned regions (Fig. 5a, b) reveals that while the cleaning process removed some of the darkened material, it also disrupted the original paint surface, leaving residual brownish particles embedded in the paint layer.

Fig. 5: Optical microscopy analysis of “Ritratto d’uomo”/Eugenio Riva.
Fig. 5: Optical microscopy analysis of “Ritratto d’uomo”/Eugenio Riva.
Full size image

The Figures show detail of previously cleaned area of the face in VIS (a) and (UV) (b) lights. Chromatic alteration and surface disruption are visible.

The surface morphology of Sample S1, observed through SEM analysis in COMBO mode (backscattered and secondary electrons), is shown in Fig. 6. The image reveals a vegetal tissue characterized by a rough and stratified surface, composed of fibrous structures15. SEM-EDS analysis was conducted to examine the surface morphology and elemental composition of the residue. Imaging revealed a thin, surface-localized deposit, and elemental mapping detected signals for potassium (K), sulfur (S), and calcium (Ca) (Fig. 6, Table 1). This elemental profile is consistent with residues from biological materials, including the epidermal tissue of Allium cepa, where K and S play metabolic roles and Ca contributes to cell wall stability—features characteristic of onion tissue31,32. While this composition supports the hypothesis of an onion-based residue, it is important to note that these elements are not unique to onions and are also present in a variety of other organic and inorganic materials, such as environmental dust and traditional painting substrates like gesso. Therefore, although SEM-EDS confirms the presence of a surface deposit with a potentially biological origin, it does not offer sufficient chemical specificity to conclusively identify onion juice as the source. Additionally, SEM-EDS analysis of the paint fragment adhered to the organic layer revealed the presence of iron-based aluminosilicate pigments, commonly known as red earth pigments, identified through the concurrent detection of iron (Fe), aluminum (Al), and silicon (Si). The analysis also confirmed the presence of cadmium (Cd), suggesting the use of Cadmium Yellow/Orange pigments. A high lead (Pb) content was detected, indicative of Lead White, a pigment possibly also employed in the ground layer of the painting, along with zinc (Zn)33,34,35,36.

Fig. 6: Scanning electron microscopy of sample S1.
Fig. 6: Scanning electron microscopy of sample S1.
Full size image

SEM image (a) of sample S1shows fibrous organic residue, b analyzed areas selected for elemental analysis.

Table 1 EDS results acquired on sample n. 1
Full size table

Figure 7 reports the scanning electron microscopy (SEM) image, acquired at 358× magnification with a 30.0 kV electron beam on S1:B sample.

Fig. 7: Scanning electron microscopy of sample S1 B.
Fig. 7: Scanning electron microscopy of sample S1 B.
Full size image

The Figure shows fibrous organic residue.

The SEM imaging shows the microstructural features of the degraded onion-based residue on the painting surface. The surface presents a highly stratified, fibrous structure, indicative of the remnants of vegetal tissue. The network of collapsed partially fragmented cell walls is consistent with previous studies on dehydrated onion37,38,39. The fibrous network appears irregular, with sections exhibiting collapsed or fragmented cell walls, likely due to degradation processes and interactions with the painted surface over time.

The SEM image of Sample S2 (Fig. 8) provides a comprehensive view of the painting’s stratigraphy. Directly beneath the “onion” layer, a composite layer is visible, consisting of both the pictorial layer and the ground layer, with an estimated thickness ranging between 5 and 20 µm ± 3 µm. Below this stratigraphy, the canvas is clearly discernible. EDS analysis (Table 2) of the “onion” layer confirms the presence of the same elements previously identified in Sample S1, reinforcing the consistency of the findings. The analysis of the pictorial layer reveals a significant concentration of iron (Fe), likely acting as a chromophore responsible for the brown tonality observed on the surface. Additionally, silicon (Si) and aluminum (Al) were detected, suggesting the presence of aluminosilicate pigments or other mineral-based components. The ground layer exhibits a high concentration of zinc (Zn) and, in contrast to Sample S1, a notable absence of lead (Pb). This finding is consistent with Balla’s documented use of Zinc White as a priming layer on the canvas, while Lead White appears to have been used as well, either in localized priming or in pigment mixtures. Conversely, the presence of lead (Pb) in other areas is attributed to Lead White, which was historically employed to desaturate and modify the chromatic balance of the painted composition.

Fig. 8: Scanning electron microscopy of sample S2.
Fig. 8: Scanning electron microscopy of sample S2.
Full size image

The Figure shows the thin paint layers and stratigraphy.

Table 2 EDS results acquired on sample S2
Full size table

The electropherogram analysis (Fig. 9) shows a fluorescence signal corresponding to Allium cepa genetic markers, though with lower intensity compared to the modern reference, indicative of DNA degradation40,41. DNA degrades over time due to environmental factors such as heat, humidity, and oxygen exposure, as well as enzymatic processes. This leads to progressive fragmentation of the molecules, making it more difficult to amplify longer fragments during electrophoretic analysis. The reduced number of amplifiable fragments results in a weaker fluorescence signal. Moreover, in degraded samples, PCR tends to preferentially amplify shorter fragments, as they are more efficiently processed than longer ones. This can create disparities in electrophoretic signals, where some fragments appear well-defined while others are weak or absent42,43.

Fig. 9: DNA Electropherogram.
Fig. 9: DNA Electropherogram.
Full size image

Electropherogram comparing the Balla residue sample (blue) to modern Allium cepa (red).

Sequence alignment (Fig. 10) confirms the presence of multiple nucleotide substitutions and deletions, supporting the hypothesis of molecular degradation over time. Specifically, the Balla residue exhibits substitutions at position 4 (C → T), position 10 (A → G), position 16 (C → G), and position 20 (C → A). Additionally, a deletion is observed at position 5. These mutations suggest a degradation process that has altered the original nucleotide sequence, consistent with long-term molecular fragmentation44.

Fig. 10: DNA sequence alignment.
Fig. 10: DNA sequence alignment.
Full size image

Sequence alignment showing degradation of onion vs Balla residue sample.

The phylogenetic tree (Fig. 11) places the Balla residue sample within the Allium clade, confirming its genetic relationship with this genus. The sample exhibits greater similarity to Allium cepa than to wild Allium species, supporting the hypothesis that the organic residue originates from the domestic onion. However, the observed genetic divergence from modern cultivars suggests either degradation-related mutations or the use of a particular variety of onion distinct from those commonly cultivated today.

Fig. 11: DNA phylogenetic tree.
Fig. 11: DNA phylogenetic tree.
Full size image

Phylogenetic tree confirming Allium cepa origin.

FT-IR analysis in attenuated total reflection (ATR) mode was conducted on Sample S1 to investigate the brownish residues present in the uncleaned areas of the painting (Fig. 12). The resulting spectra were compared with those obtained from an onion layer applied to a commercial canvas mockup, allowing for a direct assessment of their similarities. The analysis revealed a strong correlation between the two, supporting the hypothesis that the uncleaned areas of the painting still contain an onion-based extract, likely a remnant of a past restoration treatment. The spectral features observed between 1800 and 750 cm−1 correspond to key biochemical components typically found in onions. Both spectra displayed a broad O-H stretching band at 3289 cm−1, which is characteristic of hydroxyl functional groups, alongside C-H stretching vibrations at 2849 and 2916 cm−1, associated with methylene (-CH₂) groups. The presence of an asymmetric CH₃ deformation band at 1405 cm−1 and an O-H bending signal at 1365 cm−1 further supported the identification of polysaccharide structures, which are a fundamental component of onion cell walls45,46,47,48. Additional spectral features reinforced this conclusion. A peak at 1734 cm−1 corresponded to ester carbonyl stretching (C=O), indicating lipid or carbohydrate derivatives. The band at 1608 cm−1 was associated with aromatic ring stretching (C=C), a common feature of phenyl structures found in flavonoids and other phenolic compounds. The signal at 1008 cm−1 was attributed to the vibrational frequency of terminal –CH₂OH groups, confirming the presence of carbohydrates. A minor band at 1255 cm−1, linked to amide III (random coil) vibrations, suggested the presence of protein residues further corroborating the organic nature of the layer.

Fig. 12: FTIR-ATR spectra of the painting residue and onion reference.
Fig. 12: FTIR-ATR spectra of the painting residue and onion reference.
Full size image

The spectral match supports a polysaccharide-rich organic origin.

Characterization of the painting materials

For SEM/EDS analysis, Sample S3 was divided into two fragments for analysis. The first fragment (Fig. 13) was examined at the surface level to identify the chromophores responsible for the red coloration. EDS analysis (Table 3) detected iron (Fe) and mercury (Hg), both commonly associated with red pigments. The high lead (Pb) content corresponds to the presence of Lead White, while cadmium (Cd) indicates the inclusion of a yellow pigment, possibly Cadmium Yellow. A detailed examination of Spectra 1 and 4 revealed elevated levels of iron (Fe), consistently accompanied by aluminum (Al) and silicon (Si), suggesting the use of iron-based aluminosilicate pigments. However, key differences emerged between the two spectra: Spectrum 1 exhibited higher concentrations of Fe, Si, and Al, along with a notable presence of mercury (Hg), strongly indicating the use of Cinnabar (HgS) as part of the red pigment mixture. Spectrum 4, in contrast, displayed higher levels of zinc (Zn), lead (Pb), and calcium (Ca), suggesting the presence of Zinc White and Lead White, likely used to modify and desaturate the tonality of red earth pigments.

Fig. 13: Scanning electron microscopy of first fragment of sample S3.
Fig. 13: Scanning electron microscopy of first fragment of sample S3.
Full size image

The Figure shows a thin paint layers and stratigraphy.

Table 3 EDS results acquired on the first fragment of sample S3
Full size table

The second fragment of Sample S3 offers valuable insights into the complete stratigraphy of the painting. The SEM image (Fig. 14) reveals surface areas with varying elemental contrasts, indicating differences in material composition across the layers. Regions exhibiting high-contrast white areas in the BSE (backscattered electron) image are characterized by a significant presence of lead (Pb), as confirmed by EDS analysis (Table 4). The detection of lead in the ground layer reinforces the thinness of the paint film, which was measured at ~10 µm ± 2 µm in the SEM image, suggesting a delicate application of pigment layers. Additionally, calcium (Ca) was detected in association with sulfur (S), suggesting the presence of calcium sulfate (CaSO₄). This compound is well documented as a common component of gypsum, which was historically used as a filler in pigments or as a substrate for dyes. Such use served to enhance pigment stability and improve color adherence in traditional painting techniques35. The elemental ratios observed in the SEM-EDS analysis further support this attribution.

Fig. 14: Scanning electron microscopy of the second fragment of sample S3.
Fig. 14: Scanning electron microscopy of the second fragment of sample S3.
Full size image

The Figure shows a thin paint layers and stratigraphy.

Table 4 EDS results acquired on the second fragment of sample S3
Full size table

The SEM image (Fig. 15) of Sample S4 reveals distinct areas with varying elemental contrast, indicating differences in material composition across the stratigraphy. EDS analysis (Table 5) differentiates these regions based on the presence of calcium (Ca) associated with magnesium (Mg), aluminum (Al), silicon (Si), and sulfur (S), along with iron (Fe). The presence of Fe, Si, and S suggests the potential formation of iron silicates or sulfates, both of which have been associated with blue to green-blue coloration, particularly when other transition metals are present49. Moreover, the detection of Ca and Mg may indicate the existence of a mixed-phase material, potentially resulting from natural weathering or alteration processes, which are known to produce blue-colored iron-bearing minerals such as vivianite or complex silicate-sulfate assemblages50,51.

Fig. 15: Scanning electron microscopy of sample S4.
Fig. 15: Scanning electron microscopy of sample S4.
Full size image

The Figure shows a thin paint layers and stratigraphy.

Table 5 EDS results acquired on sample S4
Full size table

Additionally, Spectra 2 and 4 clearly confirm the presence of Zinc White in the ground layer (Fig. 15, Table 5), reinforcing its role as a preparatory layer in Balla’s painting technique.

XRF analysis provided valuable insights into the elemental composition of the pigments used in Portrait of a Man/Eugenio Riva, helping to confirm their identification as known materials while identifying any potential anomalies related to past restoration The elemental analysis reveals the elemental composition of samples, it cannot conclusively identify specific pigments. Confirming pigments like cobalt blue, chrome oxide green, or Prussian blue (Fe₄[Fe(CN)₆]₃) requires complementary techniques such as Raman spectroscopy or XRD. Elemental signals (e.g., Co, Cr, Fe) may suggest certain pigments but are not definitive, as these elements can occur in various compounds. Therefore, pigment identifications presented here are provisional and based solely on elemental indicators (Table 6)52. The ground layer, observed in the upper-left white-grayish region, was applied directly onto the canvas, unaffected by later paint applications. The spectrum in this area revealed a significant presence of zinc (Zn), suggesting the use of zinc white (ZnO), a characteristic pigment in Balla’s works. Traces of calcium (Ca) were also detected, likely introduced as a filler in the form of calcium carbonate (CaCO₃) or calcium sulfate (CaSO₄). The presence of iron (Fe) and copper (Cu) suggests that a black pigment, possibly tenorite (CuO), was used to subtly alter the tonality of the preparation layer. In the reddish and blue areas, XRF detected significant amounts of iron (Fe), mercury (Hg), and lead (Pb), along with traces of cobalt (Co) and chromium (Cr). This composition indicates the use of:

  • Red ocher (Fe₂O₃) and cinnabar (HgS) for warm tones.

  • Lead white (2PbCO₃·Pb(OH)₂), present in varying concentrations, likely served as a lightening agent in mixtures.

  • Cobalt blue (CoAl₂O₄) and chrome oxide green (Cr₂O₃) were identified in cooler-toned areas, adjusting the chromatic balance.

Table 6 XRF Analysis of Pigments in Portrait of a Man/Eugenio Riva
Full size table

The dark blue areas showed strong signals for cobalt (Co), iron (Fe), and barium (Ba), supporting the presence of Prussian blue (Fe₄[Fe(CN)₆]₃), cobalt blue, and traces of barium sulfate (BaSO₄). The latter was likely used as a filler to modify opacity and texture. In light blue regions, the dominance of cobalt and zinc suggests a mixture of cobalt blue, zinc white34,35,36, and lead white36,37, with the latter pigments employed to adjust the tonality. Green regions were characterized by high levels of chromium (Cr), confirming the use of chrome oxide green (Cr₂O₃). This pigment was particularly intense in the bow tie, while in the jacket, it appeared in combination with Prussian blue and cobalt blue, producing a blue-green hue. In the brown regions, only zinc (Zn) was detected, pointing to the presence of zinc white as the primary component. The absence of significant metallic elements suggests that the brown color originated from an organic pigment, potentially an earth pigment or a lake dye. The orange areas exhibited cadmium (Cd), confirming the use of cadmium yellow (CdS) or cadmium orange (CdS·CdSe). This was frequently combined with cinnabar and red ochre, creating nuanced variations in hue. In red regions, the predominance of iron (Fe) alongside lower levels of mercury (Hg) indicates that red ochre was the main pigment, with cinnabar added selectively for enhanced vibrancy. The pink areas, such as the cheeks, did not show distinctive inorganic elements, suggesting the presence of an organic red lake. This hypothesis aligns with the pinkish fluorescence observed under UV light, a characteristic feature of lake pigments, and is consistent with Balla’s documented use of such materials in other works53. Similarly, the nostril displayed a combination of cinnabar, ochre, zinc white, and lead white, creating subtle variations in shading. The white areas of the painting revealed a deliberate distinction in pigment choice. The collar contained only lead white, while the background exhibited a green-yellowish fluorescence under UV light, pointing to zinc white as its primary component. This variation suggests that Balla strategically employed different whites to achieve specific visual effects, a technique also observed in other works by the artist. Taken together, the XRF results, supported by complementary analyses, reveal Balla’s sophisticated use of both traditional and modern pigments54,55. The findings confirm a deliberate and precise layering strategy, balancing the properties of different whites, blues, and reds to create depth and contrast. Additionally, the detection of zinc white as a priming layer aligns with its increasing adoption in early 20th-century painting, further situating the work within its historical and material context.

Additional observations from imaging

Additional multispectral imaging were performed to confirm the presence of previous restorations prior to cleaning testing, as well as to examine the consistency of the pigment palette with literature references.

Multispectral imaging provided an excellent visual record, revealing deviations or areas of overpaint that may indicate previous interventions.

Under visible light, the painting revealed thin paint layers characterized by the alternating use of colors to achieve a vivid and dynamic effect, a hallmark of the Divisionist style (Fig. 16).

Fig. 16: Visible Light Multispectral Imaging—Balla, “Ritratto d’uomo”.
Fig. 16: Visible Light Multispectral Imaging—Balla, “Ritratto d’uomo”.
Full size image

Multispectral imaging revealed areas of overpaint and irregularities that may point to past restorations.

UV fluorescence imaging highlighted chromatic variations with intense fluorescence in specific areas, such as the complexion of the cheeks and the gray background (Fig. 17). The cheeks exhibited a pink-orange fluorescence, while the gray background emitted a green-yellowish fluorescence. The latter was especially evident in regions of paint loss, such as along the collar’s edge, the painting’s perimeter, and some white brushstrokes on the figure’s face and hair. This emission was likely attributable to the presence of zinc white. In contrast, the absence of fluorescence in the white pigment of the collar indicated the use of a different pigment, most likely lead white56. Additionally, UV imaging detected a non-fluorescent substance on the left side of the painting, confirming the absence of fluorescent synthetic or natural varnishes, such as terpenoid, acrylic, or urea-aldehydic resins, which are typically used as protective coatings57,58,59.

Fig. 17: UV imaging of “Ritratto d’uomo” by Giacomo Balla.
Fig. 17: UV imaging of “Ritratto d’uomo” by Giacomo Balla.
Full size image

Left) Image in UV light (365 nm); right) details of the face in UV light (365 nm).

Infrared reflectography, commonly employed to examine subsurface layers, including compositional changes and preparatory sketches, relies on the ability of near-infrared radiation to penetrate thin paint layers and reflect from non-absorbing media. It is absorbed by carbon-based or other absorbing materials like underdrawings60,61. However, the IR analysis did not reveal significant compositional changes or preparatory sketches. This suggests that the artist did not employ a detailed underdrawing but instead directly applied swift, vibrant brushstrokes to define both the figure and the background (Fig. 18).

Fig. 18: IR imaging of “Ritratto d’uomo” by Giacomo Balla.
Fig. 18: IR imaging of “Ritratto d’uomo” by Giacomo Balla.
Full size image

Left) Image in IR light (850 nm); right) details of the face in IR light (850 nm).

Cleaning tests

Different water-based cleaning systems, including Polar Rescue in gel form (5 wt% in Klucel G) and Nanorestore Cleaning® Polar Coating S, were tested to assess their effectiveness in removing the degraded organic layer while ensuring the preservation of the underlying paint surface62. The goal was to identify a sustainable and selective cleaning method that minimizes risks to the original materials. Among the tested systems, Polar Varnish Rescue, Nanorestore Cleaning® Polar Coating S, and Nasier C03 demonstrated significant efficacy in removing the onion-based residue. OM observations confirmed a marked reduction in the aging layer following treatment, with treated areas appearing visually cleaner. In contrast, water alone had no observable effect on the organic layer, reinforcing the need for targeted cleaning formulations. Both amylase solution and amylase gel proved highly effective, successfully degrading the organic layer. Post-treatment microscopic analysis revealed increased brightness and enhanced saturation across the cleaned areas, with no apparent color loss or damage to the original paint layers (Table 7).

Table 7 Microscopical observations of the cleaning tests carried out with the proposed formulations
Full size table

The spectrocolorimetric analysis provided quantitative confirmation of the observations made through OM (Table 8). Since the human eye is the main tool for evaluating paintings, it’s worth noting that color differences (ΔE) below 1–2 are usually imperceptible, while differences above 2–3 may be noticeable, especially in side-by-side comparisons63,64. As expected, deionized water did not induce any chromatic alteration, reaffirming its lack of cleaning efficacy on the surface residue. Among the tested cleaning systems, systems 2 and 6 resulted in moderate variations in surface appearance, primarily shifting towards lighter and more reddish tones. This suggests that the removal of the organic layer was only partial, leaving behind residual material that subtly influenced the final color balance. Conversely, systems 4 and 5 demonstrated a more significant change in the L parameter, indicating a more effective and uniform removal of the degraded layer. These findings were further supported by notable variations in the a and b parameters**, confirming a visible modification in the overall chromatic properties of the treated surface.

Table 8 ∆L*, ∆a*, ∆b*, and ∆E* values obtained from the comparison between the treated areas before and after the cleaning tests
Full size table

The FTIR–ATR spectra obtained from the cleaning swab after the first pass confirm that water alone does not remove any detectable material from the surface. The spectrum is identical to that of the blank cotton swab, showing only the characteristic bands of cellulose. The broad absorption around 3400 cm−1 corresponds to O-H stretching, while bands in the 2900–2800 cm−1 region are attributed to C-H stretching. The region between 1500 and 1200 cm−1 presents signals related to C-O and C-C stretching, and the spectral window between 1100 and 1000 cm−1 exhibits the C-O-C vibrations typical of glycosidic bonds. The absence of additional peaks confirms that water did not solubilize or extract any compounds from the painting’s surface, in line with the observations from optical and spectrocolorimetric analyses (Fig. 19).

Fig. 19: IR spectra of cotton before and after cleaning with water.
Fig. 19: IR spectra of cotton before and after cleaning with water.
Full size image

Red line: IR spectrum of unprocessed cotton; blue line: spectrum of the swab after cleaning using water as the solvent. Distinct peaks confirm selective removal of organic residues by different systems.

The spectrum collected from the cleaning swab after using Polar Rescue (Fig. 20) as the solvent exhibits distinct features not present in the blank cotton swab, indicating the extraction of specific compounds from the surface. Notably, a band at 1497 cm−1 is observed, which can be attributed to the C = C stretching vibration in the central heterocyclic ring of flavonoid structures. Additionally, a band at 1250 cm−1 corresponds to C–O–C stretching, typical of ether linkages in flavonoid compounds. These spectral features suggest that Polar Rescue successfully solubilized flavonoid-based residues, supporting the hypothesis that the degraded layer contains onion-derived organic material. The presence of these bands, absent in the cotton blank, indicates that the solvent was effective in removing these organic residues.

Fig. 20: IR spectra of cotton before and after cleaning with Polar Rescue.
Fig. 20: IR spectra of cotton before and after cleaning with Polar Rescue.
Full size image

Red line: IR spectrum of the cleaning swab using Polar Rescue; Blue line: spectrum of the cotton swab. Distinct peaks confirm selective removal of organic residues by different systems.

The spectrum collected from the swab used with Nanorestore Cleaning® Polar Coating S exhibits clear differences compared to the blank cotton swab spectrum (Fig. 21), indicating the successful extraction of organic residues from the surface. Notably, well-defined bands appear in the 2900–3000 cm−1 region, with peaks at 2916 cm−1 and 2852 cm−1, which correspond to C–H stretching vibrations. These signals suggest the removal of phenolic compounds, likely originating from the onion residue. Additional bands at ~1250 cm−1 and 1070 cm−1 correspond to asymmetrical and symmetrical C–O–C stretching, consistent with ether linkages typically found in flavonoids and other polysaccharide derivatives. The presence of these features, absent in the cotton blank, confirms that Nanorestore Cleaning® Polar Coating S effectively extracted components from the degraded layer, reinforcing its role in selectively removing the organic residue while preserving the underlying paint.

Fig. 21: IR spectra of cotton before and after cleaning with Nanorestore Cleaning® Polar Coating.
Fig. 21: IR spectra of cotton before and after cleaning with Nanorestore Cleaning® Polar Coating.
Full size image

Red line: IR spectrum of the cleaning swab using Nanorestore Cleaning® Polar Coating S; Pink line: spectrum of the cotton swab. Distinct peaks confirm selective removal of organic residues by different systems.

The FTIR–ATR spectrum obtained after cleaning with 1% amylase in deionized water (Fig. 22) shows a notable increase in intensity between 1500 and 1700 cm−1, forming a broad peak that is not present in the cotton swab blank. This spectral region is associated with aromatic structures and phenolic compounds, both of which are commonly found in onion extracts. The band at 1679 cm−1 corresponds to C=O stretching vibrations in hydrogen-bonded carbonyl groups, suggesting interactions with organic residues. At 1630 cm−1, a distinct peak appears due to C=C ring stretching within a phenyl structure, supporting the presence of flavonoid components. Further down, a band at 1559 cm−1 is linked to fundamental vibrations of the aromatic ring system, while the band at 1510 cm−1 corresponds to in-plane CH bending vibrations within the phenyl rings. These findings indicate that enzymatic hydrolysis by amylase, which specifically targets α-1,4-glycosidic bonds, disrupts the polysaccharide matrix of the onion residue, making previously bound phenolic compounds more accessible. The spectral changes observed suggest that the enzymatic treatment effectively breaks down the degraded organic layer, improving the removal of residues that were otherwise less soluble. The increased visibility of these phenolic structures in the spectrum reinforces the role of amylase as a selective and efficient cleaning agent, successfully releasing the organic components from the surface while preserving the integrity of the underlying paint.

Fig. 22: IR spectra of cotton before and after cleaning with Amylase.
Fig. 22: IR spectra of cotton before and after cleaning with Amylase.
Full size image

Red line: IR spectrum of the cleaning swab using 1% Amylase in deionized water; pink line: spectrum of the cotton swab. Distinct peaks confirm selective removal of organic residues by different systems.

The spectrum obtained from the swab after using Nasier C03 (Fig. 23) shows a pattern similar to that observed with amylase, indicating its effectiveness in interacting with the organic residues present on the surface. However, the cleavage of glycosidic bonds and the subsequent release of phenolic compounds appear to be less pronounced compared to enzymatic treatment. While the spectral region between 1500 and 1700 cm−1 still exhibits bands associated with aromatic structures and phenolic groups, their intensity is lower than that observed after amylase application. This suggests that Nasier C03 partially facilitates the solubilization of organic residues, though its ability to break down the polysaccharide network is not as efficient. The presence of characteristic C=O and C=C stretching vibrations, albeit weaker, confirms that the cleaning action of Nasier C03 is based on a similar principle, though likely relying on a different mechanism or degree of interaction with the aged organic layer.

Fig. 23: IR spectra of cotton before and after cleaning with Nasier C03.
Fig. 23: IR spectra of cotton before and after cleaning with Nasier C03.
Full size image

Red line: IR spectrum of the cleaning swab using Nasier C03 in deionized water; Pink line: spectrum of the cotton swab. Distinct peaks confirm selective removal of organic residues by different systems.

Final cleaning outcome

The surface cleaning of the painting was carried out using a 1% amylase solution gelled with Klucel® G (5 wt%), ensuring controlled application and minimal moisture penetration (Fig. 24). The gel was applied for 1 min, with two layers of Japanese paper interposed between the gel and the painted surface to aid in both controlled diffusion and efficient removal of excess material.

Fig. 24: Controlled cleaning with Amylase Gel and Japanese paper.
Fig. 24: Controlled cleaning with Amylase Gel and Japanese paper.
Full size image

Application of the 1% amylase solution gelled with Klucel G (5 wt%) by interposing two layers of Japanese paper for controlled cleaning.

The procedure was repeated in successive applications until the removal of the degraded coating was deemed satisfactory (Fig. 25a, b). One of the main challenges during the cleaning process was ensuring a visually uniform transition between the areas that had already been treated and those still containing residues. Achieving a balanced result required careful modulation of the cleaning process, particularly in preventing over-cleaning in certain regions while gradually reducing the degraded layer. Fortunately, the darker tones on the left side of the face facilitated a smoother transition, allowing the cleaned sections to blend naturally with the untreated areas without causing disruption to the original paint layer.

Fig. 25: “Ritratto d’uomo” by Balla before and after cleaning.
Fig. 25: “Ritratto d’uomo” by Balla before and after cleaning.
Full size image

Before (a) and after (b) cleaning, with clearer details and stronger color contrast after removing the residues.

Discussion

The integrated analysis of FT-IR ATR spectroscopy, SEM/EDS, DNA sequencing, spectrocolorimetric evaluation, and XRF provides a comprehensive characterization of the organic residue identified on the surface of Giacomo Balla’s painting. The results confirm that the brownish material found in uncleaned areas corresponds to an onion-based layer, historically applied during a previous restoration, and provide new insights into the artist’s palette and the impact of conservation treatments.

FT-IR ATR spectra from Sample No. 1 exhibited strong similarities with the onion layer applied to the mockup, particularly in the spectral range associated with hydroxyl (-OH) stretching at 3289 cm−1, methylene (-CH₂) stretching at 2849 and 2916 cm−1, and key carbohydrate peaks such as the ester carbonyl (C=O) at 1734 cm−1 and aromatic (C=C) vibrations at 1608 cm−1. The presence of these signals, alongside minor protein-related bands, supports the attribution of the residue to an onion-based extract that has undergone oxidative and structural modifications over time.

SEM/EDS analysis confirmed the fibrous nature of the brownish residue, revealing cellular structures consistent with plant material. The detection of sulfur (S) and potassium (K) further aligns with the known elemental composition of onions, reinforcing the organic origin of the layer.

Beyond the residue analysis, SEM investigations of additional samples provided valuable information on the structural composition of the painting. Cross-sectional analysis revealed the presence of a preparatory ground layer applied directly onto the canvas, primarily composed of zinc white (ZnO) with minor calcium carbonate (CaCO₃) inclusions. This layer was found to be thin and homogeneously distributed, providing a smooth base for the subsequent paint layers.

The stratigraphy of the painted surface varied depending on the color areas examined. In the red and pink regions, SEM/EDS analysis identified the co-presence of red ochre (Fe₂O₃), cinnabar (HgS), and lead white (2PbCO₃·Pb(OH)₂), with the latter being used in higher concentrations to achieve lighter tones. In the blue areas, cobalt blue (CoAl₂O₄) and Prussian blue (Fe₄[Fe(CN)₆]₃) were detected, with traces of barium sulfate (BaSO₄) as a filler. Green regions contained chrome oxide green (Cr₂O₃), sometimes mixed with Prussian blue to obtain more nuanced shades. The analysis of the black areas suggested the presence of carbon-based pigments, potentially tenorite (CuO) or a combination of organic blacks. SEM analysis also provided insights into the physical degradation of the paint layers. Microfractures and detachment phenomena were observed in samples areas, particularly where the ground layer was thinnest. These structural vulnerabilities highlight the challenges associated with conservation treatments and emphasize the importance of using controlled cleaning methods to prevent further instability.

The DNA sequencing results further support this identification, placing the sample within the Allium clade and showing greater genetic affinity to Allium cepa than to wild Allium species. However, the observed divergence from modern cultivars, characterized by nucleotide substitutions and deletions, suggests degradation-related sequence modifications or the use of a historical onion variety. In particular, the presence of transitions from cytosine to thymine (C → T) and adenine to guanine (A → G) suggests spontaneous deamination processes, where cytosine converts to uracil and adenine to hypoxanthine, leading to replication errors. The identified deletions could result from double-strand breaks, which occur in response to oxidative or hydrolytic stress and reduce the quantity of amplifiable sequences.

Spectrocolorimetric analysis revealed a measurable brightening effect following the removal of the onion-based layer. The ΔL values indicated an increment of luminosity values, while Δa and Δb shifts confirmed a return to a more balanced chromatic spectrum. These results confirm that the organic layer played a significant role in the surface darkening observed in untreated areas, likely due to oxidative and structural modifications over time. The cleaning treatment using amylase-based gel proved highly effective, allowing selective removal of the organic deposit while preserving the underlying paint. OM confirmed the uniformity of the treated areas, with no visual alterations to the pictorial layers. The comparative evaluation of cleaning systems highlighted the differential efficacy of enzymatic and solvent-based treatments. The Hansen solubility parameter diagram (Fig. 26) illustrates the power solvent of the tested solvents, explained as the Polar Rescue Gel and Nanorestore Cleaning® Polar Coating S interacted differently with the phenolic and polysaccharide components of the onion layer. Polar Rescue Gel, positioned in a lower-polarity region, exhibited greater efficiency in extracting phenolic compounds, whereas Nanorestore Cleaning® Polar Coating S was more effective in solubilizing carbohydrate-based residues. The polarity of the cleaning solutions plays a crucial role in their ability to break intermolecular bonds within the onion residue, facilitating selective solubilization.

Fig. 26: Hansen solubility parameter plot.
Fig. 26: Hansen solubility parameter plot.
Full size image

Relative position of Polar Rescue (blue) and Nanorestore systems indicates differing solubility efficiency on onion residue components.

This highlights the importance of solvent selection in conservation treatments and reinforces the advantages of enzymatic systems, which provide precise removal without excessive mechanical action or solvent penetration that could affect the underlying paint layer.

XRF analysis provided further insight into Balla’s materials, confirming the presence of zinc white (ZnO) in the preparatory layer, lead white (2PbCO₃·Pb(OH)₂) in highlights, and a diverse pigment palette including cinnabar (HgS), red ochre (Fe₂O₃), cobalt blue (CoAl₂O₄), Prussian blue (Fe₄[Fe(CN)₆]₃), and chrome oxide green (Cr₂O₃). The identification of these pigments aligns with previous studies on Balla’s works and contextualizes the painting within the artist’s broader material choices. The integration of these findings with cleaning test results highlights the need for tailored conservation strategies that consider both the chemical stability of the original materials and the impact of past interventions. The XRF analysis reinforced and contextualized the elemental data obtained through SEM/EDS on micro-samples. For instance, the detection of mercury and iron by SEM/EDS in red areas (sample S3) correlates with the broader identification of cinnabar and red ochre by XRF in the same chromatic regions. Similarly, the presence of cobalt, chromium, and zinc—mapped extensively by XRF across blue and green areas—corresponds with SEM/EDS detection of cobalt blue and chrome oxide green in selected cross-sections. This integrated interpretation confirms Balla’s elemental palette and the possible combined use of lead white and zinc white across both ground and pictorial layers.

The artist’s paint choices are important for understanding cleaning, as pigments and binders differ in sensitivity to enzymatic treatments. Inorganic pigments like lead white are generally stable and compatible with amylase cleaning, which targets starch residues. However, organic pigments may be more vulnerable to enzyme or moisture damage. Highlighting these differences strengthens the connection between material composition and cleaning safety.

Analytical observations suggest that the onion-based residue remained confined to the surface and did not chemically interact with the underlying pictorial layers. Its darkening and structural alteration likely result from long-term exposure to light and air, rather than from any reaction with the original paint film. Moreover, the presence of transition and heavy metals—such as iron, copper, zinc, and lead—may have catalyzed oxidative degradation processes in the onion-based residue. These metals, detected in both the paint and residue layers, are known to facilitate the formation of free radicals and promote crosslinking reactions in carbohydrate- and phenol-rich substances, potentially accelerating darkening and insolubilization over time65,66.

Conclusion

The restoration of Giacomo Balla’s “Ritratto d’uomo”/Eugenio Riva underscores the complexities associated with past restoration treatments and the critical role of modern scientific methodologies in overcoming these challenges. The identification of an onion-based residue, likely a remnant of historical cleaning practices, demonstrated how empirical methods, despite their initial effectiveness, can lead to unintended long-term alterations in the visual and structural integrity of an artwork.

The integration of FT-IR ATR, SEM/EDS, DNA analysis, and spectrocolorimetric techniques allowed for the characterization of the organic onion layer, while XRF was used exclusively to identify associated inorganic elements.

The discovery of molecular degradation in the DNA extracted from the residue further reinforced the understanding of how organic materials interact with painted surfaces over time.

Enzyme-based cleaning formulations, particularly amylase solutions, proved to be the most effective and controlled method for selectively removing the residue while preserving the original paint layers. The comparative study of solvent-based alternatives, analyzed through Hansen solubility parameters, highlighted their role in targeting different components of the residue, offering valuable complementary options in conservation strategies. The ability to tailor cleaning protocols to the specific chemical properties of aged residues ensures greater precision and minimal risk to fragile artworks.

Beyond the immediate restoration outcomes, this study contributes to the broader field of conservation science by providing a replicable methodology for assessing and mitigating the impact of past treatments. The multidisciplinary approach—bridging analytical chemistry, materials science, and art conservation—reinforces the necessity of integrating empirical data with innovative restoration techniques. By refining the understanding of material interactions in historical paintings, this research enhances the ability to safeguard cultural heritage while preserving both esthetic and material authenticity for future generations.