Introduction

Copper-containing compounds, both natural and synthetic, have historically been used as pigments in Mediterranean wall paintings1. The most frequently reported are basic copper carbonates such as azurite (Cu3(CO3)2(OH)2) and malachite (Cu2(CO3)(OH)2), copper silicates such as Egyptian blue (CaCuSi4O10) and chrysocolla (Cu,Al2-x(H2-xSi2O5)(OH)4·nH2O), and copper acetates such as verdigris (Cu(CH3COO)2·H2O)1,2. Copper chlorides and copper sulphates have also occasionally been identified (e.g., atacamite and brochantite respectively), although it remains unclear whether their presence was intentional, accidental or the result of alteration2. Since the 18th century, the synthesis of new inorganic pigments expanded artist’s palettes with copper arsenates (e.g., emerald green and Scheele’s green)3, although their use in wall paintings was not as widespread.

Among them, azurite and malachite were historically the most widespread, often found together in nature as copper ore deposits4,5. However, their use declined in the early modern era (15th century) with the introduction of their artificial analogues, known as verditers (blue or green)4,5. These had identical chemical composition and crystalline structure to their natural equivalents, although they differed in the morphology of their particles, which was angular in the natural basic Cu-carbonates and spherical in the artificial ones2. Used since the 3rd millennium BCE, Egyptian blue was produced by firing copper, lime and silica sand in the presence of flux (soda, potash, etc.) at 800–1100 °C, yielding the mineral cuprorivaite6. Another Cu-silicate was chrysocolla, although there is debate about its use as a pigment, as it is easily confused with Egyptian green frit ((Ca,Cu)3Si3O9)7. Even so, this natural greenish-blue copper silicate hydroxide hydrate has been identified in some wall paintings8,9,10. Lastly, verdigris, a synthetic copper acetate, had been used since antiquity and was first reported in the 3rd century BCE11. Its manufacturing process was described by many authors and generally involved exposing copper or copper alloys to an acetic atmosphere (vinegar). This produced a bluish rust, which was scraped off and used as a pigment11,12. Still, other recipes included exposing copper to wine lees, milk, urine, or honey—which would yield different chemical compounds11.

In historical wall paintings, pigments could be applied a fresco (from the Italian word affresco, meaning ‘fresh’), a mezzo-fresco (mezzo meaning ‘half’, ergo ‘half-fresh’) and a secco (meaning ‘dry’)13,14,15. In fact, most of the Medieval wall paintings were executed following these techniques16. In the fresco technique, pigments are mixed with water and applied onto a freshly laid lime-based mortar (which therefore has a high degree of moisture). The pigments then bond to the surface via a carbonation process13. In the case of the mezzo-fresco technique—also referred to as lime paint—the pigments are mixed with lime water or lime paste and painted onto a semi-dry lime-based mortar14,15. They are then fixed to the surface of the mortar due to the reactivation of the carbonation process14,15. For its part, in the secco technique, the pigments are mixed with an organic binder, generally proteinaceous in nature (e.g. egg yolk or rabbit glue), and painted onto a completely dry mortar, made of lime and/or other binders (e.g. gypsum)13.

A substantial body of research has examined the degradation of copper-based pigments in wall paintings. Azurite, for example, may change colour towards green through the formation of copper chlorides (atacamite, paratacamite or clinoatacamite), copper sulphates (brochantite or posnjakite) or secondary carbonates (malachite)8,17,18,19,20. The most commonly reported alteration, however, is blackening, generally attributed to the formation of tenorite (CuO) under conditions of high humidity and alkalinity in lime-based substrates2,21,22,23,24,25,26,27. Tenorite formation may also result from thermal damage28 or laser irradiation29,30. In addition, interactions with pollutant gases and acids can produce dark copper compounds such as copper sulphides (covellite)31,32. Despite this extensive research, the mechanisms driving CuO formation under alkaline fresco conditions remain largely unclear. Importantly, most studies have focused on natural azurite and malachite powders, while other copper pigments—particularly silicates and acetates—have received far less attention. Furthermore, factors such as pigment origin, impurities, and the formation of copper species other than tenorite have rarely been explored in lime-based mural systems. Consequently, the causes and variability of blackening in fresco paintings are still not fully explained.

The present study addresses this gap by examining a representative set of copper pigments applied directly onto lime-based substrates in order to clarify the mechanisms responsible for blackening. Eight pigments of different origin (natural and synthetic) and with different chemical composition were selected: three blue basic copper carbonates (two natural and one synthetic azurite), two basic green copper carbonates (natural and synthetic malachite), two copper silicates (Egyptian blue and chrysocolla) and one copper acetate (verdigris). A preliminary investigation comparing fresco, mezzo-fresco and secco (with egg yolk and with rabbit glue) applications confirmed that blackening was widespread in fresco, limited in mezzo-fresco paints and negligible in secco33. Building on these results, the present work focuses specifically on fresco systems to assess the influence of pigment composition, origin and impurities, and to evaluate the possible role of copper phases other than tenorite. The drying process of the paint mock-ups was monitored by digital photography over a period of 28 days to evaluate the changes on the surface, followed by stereomicroscopy examination after one year under laboratory conditions (20 ± 2 °C and 60 ± 10% RH). The blackening process was further evaluated using physical, mineralogical, micro-textural and chemical analysis, providing new insight into the behaviour of copper pigments in highly alkaline wall painting environments.

Methods

Materials and mock-up preparation

The mortar used to prepare the paint mock-ups consisted of calcitic lime putty from the Department of Mineralogy and Petrology of the University of Granada (Spain) and silica aggregates. Coarse (1.6–4 mm) and fine silica (0.125–1.6 mm), obtained from a local store in Granada, and Carrara marble powder (<0.7 mm), supplied by CTS S.L. (Madrid, Spain), were used as aggregates. The eight copper-based pigments used in this study, all supplied by Kremer Pigments GmbH & Co. KG (Aichstetten, Germany), are listed in Table 1.

Table 1 Pigments used in the study, along with their identification code (ID)
Full size table

Mock-ups were prepared as circular discs (≈8 cm diameter, ≈2.5 cm height) reproducing the layered structure of historical wall paintings1. The inner layer (arriccio) was made out of lime and sand with a lime-to-sand (L:S) ratio by volume of 1:3, using 2 parts coarse silica and 1 part fine silica. The outer layer (intonaco) had an L:S ratio by volume of 1:2, composed of 1 part fine silica and 1 part marble powder. Further characterization of the raw materials is reported in ref. 34. For fresco application (hereinafter -F), pigments were mixed with deionised water to form a paste and applied during the so-called ‘golden-hour’ (4–6 h after applying the intonaco) during which the mortar still retains a high degree of humidity14. Paint was applied with a fine paint brush, producing a 3–5 mm thick paint layer. The drying process was monitored by digital photography over 28 days, after which the samples were left for one year under laboratory conditions (20 ± 2 °C and 60 ± 10% RH). After this period, the paint samples were further analysed.

Analytical methods

Pigments and paint mock-ups were characterised using complementary physical, mineralogical microtextural and chemical techniques:

  • The particle size of the pigment powder was characterised by laser diffraction (LD), using laser particle size analyser Mastersizer 2000LF (Malvern Instruments, Worcestershire, United Kingdom). Samples were previously dispersed in alcohol. One measurement was acquired per pigment. The values reported here are based on volume distribution.

  • The mineralogical composition of the pigments was carried out by X-ray diffraction (XRD) according to the dispersed powder method. The pigments were analysed with an XPert PRO (PANalytical B.V., Almelo, The Netherlands) system using Cu-Kα radiation, Ni filter, 45 kV voltage, and 40 mA intensity. The exploration range was 3° to 60° 2θ and the goniometer speed was 0.05° 2θ s−1. The mineralogical composition of the paint mock-ups was determined using µXRD with a BRUKER D8 DISCOVER diffractometer (Massachusetts, United States) with a PILATUS3R 100K-A detector. The most superficial part of each paint was scraped, ground in an agate mortar, and the powders were analysed using Cu-Kα radiation, a Ni filter, a voltage of 50 kV, and a current of 1 mA. The exploration range was 5° to 75° 2θ and the goniometer speed was 0.02° 2θ s−1. The mineral phases for both the pigments and the paint mock-ups were identified using X’Pert HighScore software (version 4.9.0.27512).

  • The elemental analysis of the pigments was carried out by inductively coupled plasma-optical emission spectroscopy (ICP-OES) with a PERKIN-ELMER OPTIMA 8300 spectrometer (Massachusetts, United States). The instrument was calibrated in line with the standards for the elements to be identified. These standards are acid solutions of PERKIN.ELMER S.10. In this case, the powder pigments were previously diluted in hydrofluoric acid.

  • Digital photographs of the mock-ups were taken with a Nikon D3400 digital camera (Shinagawa, Tokyo, Japan) equipped with a Nikon Nikkor Zoom Lens, DX 18–105 mm. Textural and chromatic features of the powder pigments and the paint mock-ups were assessed using a Nikon SMZ 1000 microscope (Melville, New York, United States). The paint thin cross sections were studied by polarised light microscopy (PLM) with reflected light (RL) in plane-polarized light (analyser removed) or crossed polarizers (analyser inserted) using a Carl Zeiss Jenapol U microscope (Oberkochen, Germany). For this, the specimens were embedded in epoxy resin (EpoThin 2 Epoxy Resin and EpoThin 2 Epoxy Hardener) and then cut and polished to a mirror finish.

  • The microtexture and elemental composition of the pigments were studied using a JEOL JSM 6010LA scanning electron microscopy with energy-dispersive X-ray spectroscopy (SEM-EDS) in both secondary (SE) and backscattered electron (BSE) detection modes. Observation conditions included a working distance of around 10 mm, an accelerating potential of 20 kV and a specimen current of ~60 mA.

  • The microtexture and elemental composition of the paint mock-ups (surface and thin cross sections) were studied by means of SEM-EDS with a Zeiss EVO60 scanning electron microscope (Oberkochen, Germany) equipped with a lanthanum hexaboride (LaB6) cathode and a silicon drift detector (SDD), coupled with a 40 mm2 Oxford Ultim Max EDS microprobe for semi-quantitative elemental analysis. Samples removed from the mock-ups surface were analysed with no pretreatment, in variable pressure mode (VP), using an accelerating voltage of 20 kV and a pressure of 10−3 Pa. The thin cross sections of paint were coated with a nanometric layer of carbon and analysed in high vacuum mode, under the same conditions. An additional high-resolution (HR) variable pressure (VP) field emission (FE) SEM (VP-FESEM) SUPRA40VP (Carl Zeiss, Jena, Germany) equipped with the Aztec 3 microanalysis system was used to study the paint thin cross sections. Images and composition analyses were done in non-carbon coated chip samples, at low pressure and 15–20 kV beam accelerating voltage to study the texture characteristics of the painting surface. Micro-Raman spectroscopy (MRS) analysis was performed in the painting’s cross sections using a Jasco NRS-5100 confocal Raman spectrometer to identify the nature of the pigments and alteration compounds. Spectra were recorded by placing the thin section on the microscope (OLYMPUS) stage and observing them with 20 x and 100 x objectives. Paints were excited with a 532 nm laser (solid-state Torsana Starbright). Five spectra were collected from each sample spot and averaged, with exposure times of 10 s (25% power). The Raman spectra were recorded in the 3000–150 cm–1 range with a spectral resolution of 1.6 cm–1. The software JASCO Spectra Manager™ II was used to process the Raman spectra.

  • A molecular characterization of the powder pigments and the paint mock-ups was carried out by attenuated total reflectance - Fourier-transform infra-red (ATR-FTIR) spectroscopy using a Thermo Nicolet 6700 (Thermo Fisher, Massachusetts, United States) at a resolution of 2 cm–1 over 100 scans in the mid (400–4000 cm–1) and far infrared (50–400 cm–1) spectral regions. Pigments were individually analysed by placing less than 1 g on a diamond crystal and pressing the plunger against it, exerting the maximum pressure permitted by this instrument so as to ensure the best possible contact with the diamond crystal. The same procedure was followed for the paint mock-ups, which samples were obtained by scraping off the most superficial portion of paint.

  • The quantitative elemental composition and the oxidation state of the copper (Cu) and carbon (C) in the pigments and paint mock-ups were determined by X-ray photoelectron spectroscopy (XPS) with an Axis Ultra-DLD (Kratos Analytical Ltd., United Kingdom instrument), using monochromatic Al Kα radiation. Equipment settings: survey spectra at 75 W X-ray source power and 160 eV pass energy, and high-resolution spectra of Cu, C and O at 225 W source power and 20 eV pass energy, 1.33 × 10−8 Pa vacuum. The analysed area measured ~300 × 700 µm in size. All binding energies (BE) were determined using the C1s transition at 284.8 eV as a reference. Peak fitting with multiple components, aimed at distinguishing different oxidation states and/or chemical environments (deconvolution process), was performed on the high-resolution Cu2p and C1s core-level regions. Baselines were applied to the Cu2p(₃/₂) region (linear) and the C1s region (U 2 Tougaard). As usual, the components were modelled with a GL(30) line shape (i.e., a 70% Gaussian and 30% Lorentzian mixture) and fitted using the Marquardt–Levenberg algorithm. The fitting procedure was guided by both physical/chemical plausibility and the minimisation of the residual standard deviation (residual STD) relative to the experimental data. One spectrum was obtained per sample. The deconvolution of the spectra was done by using CasaXPS software. This was essential for comparing the presence and intensity of the peaks for oxidised Cu (Cu2+) and/or reduced Cu (Cu0/Cu+) species. Oxidised species were positioned at a BE of around 933.4–935.0 eV and reduced species at around 932.0–933.2 eV, as reported by ref. 35. The increase in the ratio of reduced/oxidised Cu was calculated for all the samples.

Results

Pigment characterization

Comparison between our analytical results and the data provided by the supplier (Table 1) revealed that commercial specifications were not always accurate, particularly regarding purity. A clear trend emerged between natural and synthetic pigments. Natural pigments systematically contained mineral impurities associated with their geological origin (e.g., quartz, feldspars, iron oxides and secondary copper phases), whereas synthetic pigments showed residual phases linked to their manufacturing processes (e.g., calcite or quartz from raw materials)4,5,6. Even pigments marketed as highly pure (i.e., AZP) still contained detectable impurities, indicating that complete compositional purity is uncommon in commercial copper pigments. The semi-organic copper acetate (verdigris) differed from the inorganic pigments in being composed of a single crystalline phase, consistent with its synthesis route11. Regarding elemental composition, pigments rich in basic copper carbonates showed the highest Cu contents, whereas copper silicates displayed lower values. Overall, these results demonstrate that pigment origin (natural vs. synthetic), impurity content and chemical nature are key variables that must be considered when evaluating the stability and alteration behaviour of copper-based pigments.

Physical characterization of the fresco paint mock-ups

The detailed monitoring of the drying process (Fig. 1) showed that most colour changes occurred between 24 and 48 h after paint application, with little variation afterwards. Clear trends emerged depending on pigment chemistry: basic copper carbonates exhibited the most pronounced and widespread blackening (AZN-F, AZS-F, AZP-F, MAN-F and MAS-F), whereas copper silicates remained comparatively stable, showing only minor total shifts in chrysocolla (CHR-F). In contrast, the copper acetate mock-up (verdigris, VER-F) showed the largest chromatic variation together with poor cohesion, resulting in a powdery and highly heterogeneous surface. All inorganic pigments produced compact, homogeneous layers.

Fig. 1: Digital photographs documenting the monitoring of colour changes in the fresco painted mock-ups during the drying process (28 days) and after 1 year under laboratory conditions (20 ± 2 °C and 60 ± 10% RH).
Fig. 1: Digital photographs documenting the monitoring of colour changes in the fresco painted mock-ups during the drying process (28 days) and after 1 year under laboratory conditions (20 ± 2 °C and 60 ± 10% RH).
Full size image

See Table 1 for an explanation of pigment identification codes (ID).

Stereomicroscopy observations (Fig. 2) supported these trends. The powder pigments of natural origin frequently displayed particles of different colours due to associated mineral impurities, while synthetic pigments appeared more uniform. Regarding paints, the basic Cu carbonate-based F paints (Fig. 2a) underwent the most pronounced blackening, whereas in Cu-silicates (Fig. 2b), Egyptian blue (EGB-F) remained unaltered, as opposed to chrysocolla (CHR-F), which underwent slight blackening. As for the copper acetate-based paint verdigris (VER-F), discolouration of varying shades formed on the surface (Fig. 2c). Overall, the degree of alteration in fresco samples followed consistent patterns linked to pigment chemistry (carbonate vs. silicate vs. acetate), origin (natural vs. synthetic), and surface cohesion (homogeneous vs. heterogeneous), indicating that these factors strongly influenced the early-stage colour stability of copper-based fresco paints.

Fig. 2
Fig. 2
Full size image

Micrographs of powder pigments (top row) and fresco paint mock-ups (bottom row) for a) basic copper carbonates, b copper silicates and c copper acetate. All images were taken after 1 year left under laboratory conditions (20 ± 2 °C and 60 ± 10% RH). See Table 1 for an explanation of the pigment identification codes (ID).

Mineralogical characterization of the fresco paint mock-ups

Through mineralogical analysis (Table 2) we identified three main groups of phases: (i) compounds inherent to the fresco technique (portlandite, calcite, quartz), (ii) primary minerals already present in the raw pigments, and (iii) alteration products formed. In general, the main pigment phases detected in the powder were preserved in the fresco paints, confirming limited mineralogical transformation in most cases. The only exception was the verdigris fresco paint (VER-F) in which hoganite was no longer detected, and vaterite appeared instead. This compound is a metastable polymorph of calcium carbonate that becomes stabilised in the presence of organic molecules36. Since hoganite is a semi-organic pigment and is water-soluble [https://www.mindat.org/], it is likely that it dissolved, allowing the organic part of the pigment to influence vaterite precipitation. Nevertheless, metastable vaterite would eventually dissolve, with stable calcite forming in its place due to the rapid kinetics of the vaterite-to-calcite transformation36. Such behaviour of verdigris highlights the distinct response of copper acetates compared with inorganic pigments. The identified alteration products also displayed clear compositional trends. In general, copper oxides were only detected as alteration products in basic copper carbonate-based paints and in copper acetate, while silicate-based showed no detectable secondary phases. These observations indicate that pigment chemistry was the primary factor governing mineralogical change. However, the evident blackening observed a visu was not always matched by the consistent identification of copper oxides in the analytical results (Table 2). This discrepancy suggests that additional factors, beyond mineralogical transformation, may have contributed to the chromatic alteration.

Table 2 Mineralogical characterization by µXRD of the fresco paint mock-ups
Full size table

Micro-textural characterization of the fresco paint mock-ups

SEM-EDS analysis of the surface of the paints revealed clear micro-textural trends controlled mainly by pigments' chemistry and origin. Most Cu-carbonate fresco paints (Fig. 3a–e), as well as Cu-silicate CHR-F (Fig. 3g), showed heterogeneous surfaces characterised by voids exposing Cu-rich particles (observed with higher contrast in Fig. 3). Synthetic Cu-carbonates exhibited a higher density of voids than natural ones, indicating greater reactivity. In contrast, EGB-F displayed a compact, homogeneous surface covered by a continuous Ca-rich layer, with no exposed pigment particles (Fig. 3f). Verdigris (VER-F) behaved differently from all inorganic pigments, showing a highly heterogeneous surface composed of morphologies with distinct compositions (Fig. 3h–j). EDS analysis confirmed great variability in elemental composition depending on morphology (Supplementary Figure 1-2). For instance, acicular particles (Fig. 3i) were enriched in O and Cu with minimal Ca, whereas more massive morphologies (Fig. 3j) were richer in O and Ca, with lower Cu contents.

Fig. 3: SEM micrographs of the surface of the Cu-based F paints.
Fig. 3: SEM micrographs of the surface of the Cu-based F paints.
Full size image

The arrows mark the presence of voids. The white rectangles highlight the areas observed at higher magnifications. See Table 1 for an explanation of the pigment identification codes (ID).

Cross-section observations by PLM-RL (Supplementary Fig. 3) confirmed these trends. Blackening was heterogeneous in all paints except for Egyptian blue (EGB-F), which has homogeneous blue. Carbonate-based paints showed the highest degree of blackening, silicates an intermediate behaviour (stable EGB-F but slightly darkened CHR-F), and finally verdigris (VER-F) with a complex colour variation rather than uniform blackening. Furthermore, micro-Raman analysis of these cross-sections did not yield spectra that could enable the identification of a distinct mineral phase associated with the blackened particles. As illustrated in Supplementary Fig. 4 for natural malachite (MAN-F), both the green malachite grains and the visually blackened particles displayed the characteristic Raman peaks of malachite. This result suggests that the darkening may not be linked to the formation of a new, easily detectable crystalline phase.

Regarding pigment origin, differences were observed by SEM between natural and synthetic basic copper carbonate-based paints. Both natural and synthetic showed pitting patterns and reaction halos (marked with white arrows in Figs. 4a, b, respectively), indicating surface alteration at the pigment-mortar interface. These reaction ‘halos’ were detected around the black particles observed by PLM-RL, as seen in AZN-F (marked with yellow arrows in Fig. 4c). Chemical analysis of the particles by EDS line-scan identified higher Cu and lower O contents on the ‘reaction halos’. However, these alteration patterns were more pronounced and widespread in synthetic copper carbonates. Additional deposits of Cu- and O-rich particles were depicted around the azurite grains (marked with white arrows in Fig. 5a), as well as O-rich ‘reaction halos’ (marked with yellow arrows in Fig. 5b), evident by EDS mapping analysis (Fig. 5b). The main difference between natural and synthetic Cu-carbonate-based F paints was the presence of Cu-rich acicular morphologies (marked with red arrows in Fig. 5c, d), suggesting more advanced alteration processes. This acicular habit could be related to the formation of tenorite (CuO)37,38.

Fig. 4: PLM and SEM micrographs of natural Cu-carbonate-based F paint cross sections.
Fig. 4: PLM and SEM micrographs of natural Cu-carbonate-based F paint cross sections.
Full size image

a, b The white rectangles in the PLM micrographs correspond to the areas shown in the SEM micrographs on their right. These also include white rectangles that highlight the areas observed at higher magnifications. The white arrows indicate pitting of the pigment particles. The yellow arrows highlight the ‘reaction halos’. c Numbers 1, 2 and 3 in the SEM micrograph correspond to the EDS analysis marked with black dashed-line rectangles in the line-scanning spectra. For more information about the references to colour in this figure legend, readers should consult the web version of this article. See Table 1 for an explanation of the pigment identification codes (ID).

Fig. 5: PLM and SEM micrographs of synthetic Cu-carbonate-based F paint cross sections.
Fig. 5: PLM and SEM micrographs of synthetic Cu-carbonate-based F paint cross sections.
Full size image

The white rectangles in the PLM micrographs correspond to the area shown in the SEM micrographs at right. The white arrows highlight pitting of the pigment particles. The yellow arrows indicate the ‘reaction halo’. The red arrows highlight the presence of acicular morphologies. For more information about the references to colour in this figure legend, readers should consult the web version of this article. See Table 1 for an explanation of the pigment identification codes (ID).

Among the Cu-silicate-based F paints, two distinct behaviours were observed. Egyptian blue (EGB-F) remained chromatically stable, retaining its bluish hue and showing no reaction halos under SEM, confirming the high stability of cuprorivaite particles. Sporadic dark Cu-rich particles were occasionally detected (Fig. 6a, b), likely related to the raw materials used during pigment synthesis8. EDS line-scan analysis of this particle (Fig. 6b) showed a similar spectrum to that observed in basic Cu-carbonates. However, these blackish particles did not affect the overall appearance of the paint layer. In contrast, chrysocolla (CHR-F) exhibited slight blackening associated with impurities. SEM-EDS and PLM-RL analyses revealed Cu- and Co-rich particles likely linked to accessory minerals such as malachite or other ore-related phases (Fig. 6c, d). The Cu-rich phases showed reaction halos similar to those observed in Cu-carbonate paints (marked with white arrows in Fig. 6c, d). Other additional reddish-brownish particles, rich in Fe, were also present (marked with orange arrows in Fig. 6e, f).

Fig. 6: PLM and SEM micrographs of Cu-silicate-based F paint cross sections.
Fig. 6: PLM and SEM micrographs of Cu-silicate-based F paint cross sections.
Full size image

The white rectangles in the PLM micrographs correspond to the area shown in the SEM micrographs at right. The red circle highlights the presence of Cu-rich particles. The white arrows indicate pitting of the pigment particles. The orange arrows highlight the presence of Fe-rich particles. For more information about the references to colour in this figure legend, readers should consult the web version of this article. See Table 1 for an explanation of the pigment identification codes (ID).

Finally, the organic Cu-acetate verdigris paint (VER-F) showed a completely irregular surface, with shades ranging from blue (Fig. 7a) to brownish-black (Fig. 7b). Under SEM-EDS observation, the paint showed irregular micro-textural features, with fissures and cracks, and a layer that was completely detached from the lime-based mortar substrate (marked with white arrows in Fig. 7). This heterogeneous paint layer was composed of varying amounts of C, O, Ca and Cu, as observed by EDS mapping analysis (Fig. 7).

Fig. 7: PLM and SEM micrographs of a cross section of the Cu-acetate-based F paint.
Fig. 7: PLM and SEM micrographs of a cross section of the Cu-acetate-based F paint.
Full size image

The arrows indicate locations where the paint layer detached from the mortar substrate. See Table 1 for an explanation of the pigment identification codes (ID).

Chemical characterization of the fresco paint mock-ups

ATR-FTIR spectra (Supplementary Fig. 5–6) of the powdered pigments were compared to those obtained from the F mock-ups. Based on a comparison with the available literature, band wavenumbers (cm−1) were assigned to the vibrational modes characteristic of each mineralogical phase for both the powder pigment and its corresponding fresco paint (-F). These data are presented as Supplementary Table 1. As expected, all fresco paints showed the characteristic bands for CaCO3 related to the lime-based substrate, confirming the influence of the fresco technique, through the presence of a newly formed band at ≈1395 cm−1 ν(CO32−) and 870 cm−1 δasym(CO32−)39,40. For most pigments, the main spectral features remained similar before and after application, indicating that the primary mineral phases were preserved. However, differences emerged depending on the pigment chemical nature.

Cu-carbonates and Cu-silicates displayed only minor spectral changes. In the mid-IR region, characteristic bands of the pigments were retained with slightly lower intensity, suggesting no structural modification had occurred. In the far-IR region, however, subtle changes in Cu–OH and Cu–O bands were detected in some Cu-carbonate samples (Fig. 8a, b), indicating local variations in the chemical environment of Cu2+ ions. In contrast, the Cu-acetate pigment (VER-F) showed pronounced spectral modifications (Fig. 8c, d). In the mid-IR spectra, bands assigned to the water molecules and methyl (CH3) groups disappeared, while a broad ν(–OH) band appeared, indicating increased hydration (marked with a grey rectangle in Fig. 8c). Characteristic COO– and O–C–O bands shifted to lower wavenumbers (see Supplementary Table 1), and new bands consistent with calcium acetate species were detected (marked in red in Fig. 8c), suggesting interaction between verdigris and the lime substrate41. In addition, calcite bands were detected in VER-F, confirming the formation of calcium carbonate. Finally, Cu-related bands in the region below 400 cm−1 (Fig. 8d) decreased markedly or appeared only as shoulders (see Supplementary Table 1), indicating modification of the original verdigris chemical structure.

Fig. 8: ATR-FTIR analysis of the altered frescoes.
Fig. 8: ATR-FTIR analysis of the altered frescoes.
Full size image

Spectra comparing the powder pigment and the fresco paint (-F) of a natural azurite (AZN) in the far-IR (600–200 cm−1) region; b synthetic malachite (MAS) in the far-IR region; and verdigris (VER) in the c mid-IR (4000-400 cm−1) and d far-IR regions. In c, d, the grey rectangle indicates changes in –OH and CH3 bands; other acetate compounds different to hoganite are marked in red; and the red rectangles indicate new Cu-assigned bands. See Table 1 for an explanation of the pigment identification codes (ID).

The XPS high-resolution analysis was performed by comparing the Cu2p spectra (only the 3/2 region is shown for clarity) and C1s of each powder pigment with those from the corresponding F-paint. The increase in the ratio of reduced/oxidised Cu was calculated for all the samples (Table 3), while all Cu2p(3/2) spectra, including the peaks for oxidised Cu (Cu2+) and reduced Cu (Cu0/Cu+), are presented in Supplementary Figs. 7–8. The most representative Cu2p(3/2) XPS high-resolution spectra are shown in Fig. 9, where Cu2+ peaks are represented in green and Cu0/Cu+ peaks in red. An increase in the reduced Cu species was observed in all the F paints to a greater or lesser extent (Table 3). The most pronounced increases occurred in the basic Cu-carbonates (AZN-F, AZS-F, AZP-F, MAN-F and MAS-F), whereas EGB-F (Cu-silicate) exhibited minimal changes. CHR-F (Cu-silicate) and VER-F (Cu-acetate) exhibited similar trends to those of carbonates, indicating partial reduction of Cu2+ to Cu0/Cu+ after fresco carbonation. In such a complex system, it is difficult to determine the exact origin of the reducing effect observed in the Cu2p spectra. Nevertheless, deconvolution of the C1s XPS high-resolution spectra provided important information. The C1s spectra for all samples (pigment and fresco) are included as Supplementary Figs. 9–10, representing the peak position of reduced carbon (C–C), oxidised carbon not related to carbonates (C–O and C=O/O=C–O), and oxidised carbon related to carbonates (CO32–)35. Reduced carbon (Creduced) was the predominant form, suggesting that carbon could act as a reducing agent for Cu²⁺. Its atomic concentration (at.%) in the frescoes, reported in Table 4, increased after carbonation in most samples, except for VER-F, with EGB-F showing the smallest increase (0.1 at.%). As expected, F painting mock-ups showed higher carbonate-related carbon (Ccarbonate) than the corresponding pigments, except for the AZP-F.

Fig. 9: Deconvoluted high-resolution XPS spectra of the Cu2p(3/2) region.
Fig. 9: Deconvoluted high-resolution XPS spectra of the Cu2p(3/2) region.
Full size image

a AZS (right) and AZS-F (left); b MAN (right) and MAS-F (left); and c EGB (right) and EGB-F (left). See Table 1 for an explanation of the pigment identification codes (ID).

Table 3 Increase (%) in the ratio of reduced/oxidised copper for all fresco paint samples.
Full size table
Table 4 Atomic concentration (at.%) of total carbon (Ctotal), oxidised carbon assigned to carbonate groups (Ccarbonate), oxidised carbon different to that from carbonates (Coxidised), and reduced carbon (Creduced) for the pigments and the painting mock-ups See Table 1 for an explanation of the pigment identification codes (ID)
Full size table

Discussion

In the fresco technique, where copper pigments are mixed with water and applied onto a freshly laid lime-based mortar, the blackening process is strongly influenced by the highly alkaline environment, as widely reported in the literature2,25. However, the precise mechanism underlying copper pigment blackening remains a subject of ongoing debate42. One prevailing hypothesis suggests that during the carbonation of Ca(OH)2, hydroxyl groups (OH– ions) are released, which react with the Cu2+ ions in the pigment to form Cu-hydroxides, such as spertiniite (Cu(OH)2), which are metastable phases that are difficult to isolate in pure crystalline form23,43. Consequently, at room temperature and under high pH conditions, Cu(OH)2 forms a complex anion (Cu(OH)42−) that acts as a precursor in the formation of black Cu-oxides43. Although Cu-hydroxides were not identified by µXRD in our samples, their formation as an intermediate phase cannot be excluded, particularly given the strongly alkaline and humid environment of frescoes. Importantly, copper hydroxides are thermodynamically unstable under ambient conditions and tend to dehydrate or transform into copper oxides44. In the mock-ups here evaluated, tenorite (CuO) and cuprite (Cu2O) were identified only in some samples, suggesting that hydroxides and other intermediates may evolve towards CuO over time. However, our results also indicate that additional factors influence blackening: pigment-related features (pigment nature, origin and impurities) and reduced copper species, as discussed below.

It was demonstrated that pigment composition strongly controls the extent of blackening in fresco paints. Basic Cu-carbonates (azurite and malachite) were consistently more prone to blackening than Cu-silicates (Egyptian blue and chrysocolla), showing pitting patterns and ‘reaction halos’ around the Cu-carbonate particles. In turn, Cu-silicates retained their original hue, except for minor blackening associated to impurities. Among these, natural chrysocolla (CHR-F) exhibited slight blackening due to Cu-, Co- and Fe-rich impurities, whereas synthetic Egyptian blue (EGB-F) remained entirely stable. PLM-RL and SEM-EDS suggest that these impurities were likely to be malachite, heterogenite (Co3+O(OH)), and haematite (Fe2O3), generally found together in the mineral ore45,46. The high stability of the Cu-silicate particles—namely cuprorivaite and chrysocolla—is related to the tight arrangement of the copper ions within the silicate matrix, not easily removed by chemical and physical means47. These observations highlight the role of minor impurities in promoting localised alteration, even though the bulk composition remains stable. Origin also affected reactivity. Interestingly, synthetic Cu-carbonates (AZS-F and MAS-F) often showed more pronounced pitting and acicular features compared with their natural analogues (AZN-F, AZP-F and MAN-F), despite being chemically purer. This suggests that factors related to the synthesis process—most likely the smaller particle size in synthetic pigments—enhanced their chemical reactivity and susceptibility to blackening.

The organic Cu-acetate pigment, verdigris, displayed a highly heterogeneous, powdery and extensive alteration. ATR-FTIR analysis showed shifts in carboxylate vibrations, suggesting the formation of a mixture of compounds48. The presence of a broad –OH band, indicative of hydrated phases, together with bands characteristic of calcium acetate species, suggests the formation of calcium acetate monohydrate (Ca(CH3COO)2·H2O) and/or calcium acetate hemihydrate (Ca(CH3COO)2·(H2O)0.5) as degradation products of verdigris49. A likely reaction pathway involves calcium hydroxide (Ca(OH)₂) from the lime-based substrate, which releases Ca²⁺ and OH⁻ ions. In the presence of moisture, hoganite (Cu(CH₃COO)₂·H₂O) partially dissolves, releasing copper Cu²⁺ and CH₃COO⁻ ions. An ion-exchange reaction can then occur: calcium ions combine with acetate ions to form calcium acetate compounds, while copper ions precipitate as secondary copper species. Supporting this pathway, bands assigned to ν(Cu–OH) were also identified41, confirming the presence of secondary copper compounds. Such transformations emphasise the intrinsic instability of organic copper pigments in alkaline fresco environments.

Tenorite (CuO), the oxidised copper species (Cu2+) commonly assumed as the main contributor to blackening2,25,50, was only detected in the azurite samples (AZN-F and AZP-F). However, XPS results suggested an additional pathway of blackening, which consists in the formation of reduced copper species (Cu0/Cu+), recognised due to the increase in the atomic ratio of reduced copper in all the paints. In the Egyptian blue fresco, which was the only paint that did not undergo blackening, this increase was almost negligible compared to the others (see Table 3), supporting the conclusion that species containing copper in a low oxidation state (Cu0 and/or Cu+) are also contributors to blackening. This was further confirmed by ATR-FTIR results: when the oxidation state of copper changes (e.g., from Cu²⁺ to Cu0 or Cu⁺), it can affect nearby bond strengths and electron distributions—trend we observed in our spectra. In fact, Kumashiro et al.51 observed that the reduction of Cu²⁺ to Cu⁺ by the effect of carbon monoxide (CO) was accompanied by shifts in IR absorption bands. However, the characteristic ν(C–O) bands formed upon reduction, assigned to CO absorbed on Cu+ species—present around 2160, 2150 and 2135 cm–151—were not identified in our spectra.

Among the potential alteration products, cuprite (CuO2) represents a likely candidate, being a reduced copper species (Cu+) that, despite being reddish in bulk form, can also occur as a blackish nanometre-scale layer during early-stage corrosion of copper and copper alloys52,53,54. Cuprite was indeed identified in verdigris (VER-F) and natural malachite paints (MAN-F), where it was already present in the original pigment as an impurity, although its further formation as an alteration product cannot be discarded29. In the other paints, cuprite may have formed in amorphous or poorly crystalline form, explaining why it was not detected by µ-XRD and µ-Raman.

Copper hydroxides may also form transiently in humid, alkaline fresco environments43. Cu(OH)2), a Cu2+ species, can dehydrate to tenorite or undergo partial reduction to Cu+ species, acting as an intermediate in the formation of both oxidised and reduced copper phases44. Although crystalline hydroxides were not detected in this study, their presence cannot be excluded, particularly as a poorly crystalline or amorphous compound. Over time, these intermediates are expected to evolve towards tenorite, the most stable copper oxide under environmental conditions and the phase most frequently reported in the literature.

Another possible alteration compound is paramelaconite (Cu4O3), a mixed-valence copper oxide containing both Cu+ and Cu2+ species55, previously reported by Špec et al.27 as a possible alteration product of malachite. Structurally, it represents an intermediate phase between cuprite and tenorite, arising from partial oxidation of Cu+ or incomplete oxidation of Cu2+55. Due to its mixed-valence structure, paramelaconite is difficult to synthesise and stabilise as a well-crystallised phase56. Nevertheless, its occurrence in amorphous or poorly crystalline form, potentially alongside cuprite or metallic copper, cannot be excluded.

Since neither cuprite nor paramelaconite was consistently identified by µXRD, other explanations may involve the formation of amorphous or poorly crystalline non-stoichiometric copper oxides (Cu₂₋ₓO) and partial re-oxidation processes. In alkaline and humid fresco environments, Cu⁺ species can undergo incomplete oxidation to Cu²⁺, producing amorphous surface layers that are not detected by µXRD but contribute to dark coloration. Nanoscale effects, particle heterogeneity, and the coexistence of multiple Cu species can further enhance light absorption, yielding a blackish appearance even in the absence of crystalline tenorite. These mechanisms are consistent with observations in copper corrosion studies52,53,54, where early-stage cuprite layers often appear dark or black when poorly crystalline or mixed with other alteration products. Overall, these processes highlight the multifactorial nature of copper pigment degradation and support a cautious interpretation of XPS evidence when explaining blackening phenomena.

Regardless of the reduced copper species formed, the mechanism of copper reduction is not straightforward due to the complexity of the system, in which not only CO₂ (which by itself does not exhibit reducing behaviour) but also water (H₂O), pH, and carbon monoxide (CO)—always present in the air (outdoor and indoor)57—play a role. Indeed, since carbon in CO is not in its highest oxidation state, it could act as a reducing agent, as reported by refs. 51,58. In addition, this reducing C could originate from carbon-based species from the air. Atmosphere typically contains, on the one hand, volatile organic compounds (VOCs)—such as hydrocarbons (e.g., methane) and other low-molecular-weight organic species like formaldehyde, ethanol, or acetone, originated from pesticides, solvents, and cleaning agents59—and, on the other hand, soot particles from carbonaceous fuel combustion60. These carbon-containing species are ubiquitous and adsorbed on all surfaces, as adventitious carbon61. Additionally, the carbonation process could favour their incorporation into the fresco-applied pigments, becoming trapped within the forming carbonate matrix. As a result, the total carbon content increases after the pigment is applied a fresco, comprising carbonate species, other oxidized carbon atoms, and C–C-type carbon. Moreover, there was no correlation between the generation of carbonate ions (CO32–) and the higher content of reduced Cu. These compounds would imply an increase in reduced carbon (C–C), as detected by XPS analysis, which could have acted as a reducing agent. Yet, whatever the origin, these reduced carbon species (C–C-type) were incorporated into the medium during the fresco carbonation process, acting as a reducing agent of the copper ions, which subsequently contributed to the blackening. Nevertheless, the exact origin of the reducing mechanism of copper remains unclear.

Overall, the latest research has assigned the blackening of copper pigments in alkaline environments to the formation of tenorite (oxidation state Cu2+). Despite the pronounced, widespread blackening observed in the frescoes, tenorite was detected in only a small number of samples. Our results indicate that reduced copper species (oxidation state Cu+ and/or Cu0), present in amorphous or poorly crystalline forms, also played a significant role in the blackening process. Nevertheless, it remains uncertain whether the reduction of copper ions was driven by atmospheric CO or by other compounds containing reduce carbon atoms. To obtain a clearer picture of the underlying physical, mineralogical, and chemical alteration mechanisms, the use of advanced techniques such as transmission electron microscopy (TEM), synchrotron radiation micro-X-ray powder diffraction (SR-μPXRD), or synchrotron radiation X-ray absorption near-edge structure spectroscopy (SR-μXANES) is recommended for future studies.

These findings have important implications for the conservation of wall paintings. Accurate identification of pigment type, origin, and chemical interactions with the fresco substrate is essential for predicting the long-term colour stability of copper-based paints. The higher susceptibility of basic Cu-carbonates (azurite and malachite) and Cu-acetate verdigris to alkaline-induced blackening highlights the importance of careful pigment characterization, which can guide conservation-restoration decisions, such as avoiding alkaline-based treatments to preserve paint integrity. Furthermore, the identification of reduced copper species (Cu⁺/Cu⁰) as contributors to blackening suggests that environmental control (e.g., such as limiting exposure to pollutants or carbon-based reducing agents) could help prevent deterioration.