A Multi-Spectroscopic Investigation into the Degradation Mechanisms and Lightfastness of CdS and Cd₁₋ₓZnₓS Yellow Acrylic Artists’ Paints Under Artificial Daylight: Competition Between Photocorrosion and Binder Degradation

1. Introduction

The degradation of pigment properties, particularly color alteration in visual art, presents a critical challenge for the museum community, including curators, researchers, and conservators.[1,2,3,4,5,6,7,8,9] While irreversible changes in various artistic media have long been recognized, the vulnerability of specific historical pigments remains a focal point of modern heritage science. [2,4,10,11,12,13,14] Among these, cadmium yellow pigments (composed of cadmium chalcogenides) have been highly prized by artists since the mid-19th century for their exceptional opacity, high tinting strength, and brilliant hue. Over the past few decades, the deterioration of these pigments has received significant attention due to its manifestation in masterpieces by iconic modernist artists such as Pablo Picasso [5,15], Vincent van Gogh[16], Henri Matisse[16,17,18], Edvard Munch[19], and James Ensor[7]. These alterations exhibit a diverse range of macroscopic and microscopic features, including paint chalking, fading, flaking, spalling, darkening, and the formation of surface protrusions or white crystalline crusts. In advanced stages, grey-orange alteration crusts develop, completely obscuring the original yellow paint layer[5,6,7,15,20,21,22,23,24,25,26].

Consequently, the international scientific community actively investigates the underlying degradation mechanisms of cadmium yellow, as evidenced by a growing body of literature. [1,5,7,8,9,10,15,16,17,18,26,27,28,29,30,31,32,33,34,35,36,37,38] For instance, studies on Joan Miró's paintings and studio materials have highlighted how environmental factors, chemical composition, and historical manufacturing technologies dictate subsequent aging pathways[23]. Furthermore, analytical research on works by Pablo Picasso employing synchrotron radiation and photoluminescence spectroscopy has elucidated complex physicochemical transformations, such as cadmium oxidation and the subsequent formation of degradation products (e.g., cadmium sulfate) that cause the paint layer to lighten and fade[15]. Collectively, these studies demonstrate that cadmium degradation is a widespread phenomenon rather than an isolated occurrence in modernist painting.

To systematically address these challenges, this experimentally based project was developed and executed in close collaboration with the State Tretyakov Gallery Moscow Russian Federation. As as a leading institution housing a vast collection of Russian painting from the 12th to the 21st century and maintaining a rigorous commitment to technical art history, the Gallery is deeply invested in the comprehensive understanding of material properties and aging processes. A significant part of the collection consists of contemporary art. Researching these materials requires a dual foundation in both the photophysical/photocatalytic processes of semiconductor particles and the physicochemistry of polymer aging. This is particularly fair for twentieth-century and contemporary artworks; while traditional oil paints embrittle predominantly via drying oil oxidation [7,8,12,16,19,39], acrylic binders undergo multifaceted degradation driven by additive migration, macromolecular alterations, and photochemical sensitivity[3,40,41,42,43,44,45,46,47,48]. Acrylic paints formulated with cadmium sulfide (CdS) [1,27,49] and cadmium zinc sulfide (Cd_xZn_1-xS) [23] are valued in contemporary art for their mechanical flexibility, yet they remain complex chemical systems vulnerable to environmental degradation. Mapping these coupled organic–inorganic transformations necessitate a robust multianalytical approach that evaluates binder degradation and pigment alteration in parallel, as the binder's chemical composition, structural evolution, and aging trajectory serve as primary determinants for the degradation rates and mechanisms in cadmium yellow systems.

Ultimately, a precise understanding of the physical and chemical mechanisms occurring within these paint layers will refine existing conservation protocols, optimize museum display conditions, and aid in developing preventive strategies to safeguard unique artistic heritage against environmental and time-dependent degradation. To this end, the aim of this work is to compare the lightfastness and degradation kinetics of two commercial, artist-grade, cadmium-based acrylic paints under artificial daylight: Cadmium yellow medium and Cadmium lemon. Both formulae utilize a styrene-acrylic copolymer binder and carry a Pigment Yellow 35 (P.Y.35) classification with a commercial lightfastness rating of three stars (excellent/fully lightfast). To evaluate these materials, a multi-analytical framework was deployed, combining colorimetry, scanning electron microscopy with energy-dispersive X-ray spectroscopy (SEM-EDS), attenuated total reflection Fourier-transform infrared spectroscopy (ATR-FTIR), and Raman microscopy. The resulting colorimetric, spectral, and structural datasets were analyzed using chemometric methods, including Principal Component Analysis (PCA) and Partial Least Squares (PLS) regression. For conciseness, Cadmium lemon (Code 203) and Cadmium Yellow Medium (Code 201) are abbreviated throughout this study as LY203 and Y201.

2. Materials and Methods

2.1. Sample Preparation and Artificial Daylight Aging

Y201 and LY203 paints were evenly applied to aluminum foil with a medium-sized brush. All samples were then dried for two weeks at room temperature (22°C and 40% relative humidity).

Accelerated artificial aging of paints was conducted in a Q-Sun xenon arc chamber under controlled conditions (38°C and 65% relative humidity). Samples were irradiated with a xenon lamp (λ > 290 nm), simulating sunlight, at an intensity of 60 W/m². The degradation of the samples was studied after 287 and 455 hours of exposition, which corresponds to 1.2 × 10⁷ and 1.9 × 10⁷ lux-hours. Colorimetric characteristics and reflectance spectra (400–700 nm) were determined using a ColorFlex spectrophotometer (HunterLab, USA) configured with a 45/0° geometry, a 10° observer, and a D65 illuminant after 120, 240 and 489 hours, which corresponds to 0.5× 10⁷, 0.99× 10⁷, 2.02× 10⁷ lux-hours. The color difference (ΔE), calculated according to the CIELab-76 system [50], was used as the main evaluation criterion. The measured L*a*b* values were plotted using the Chromaticity Diagram App (OriginLab, version 1.82). Color swatches were generated by converting the measured L*a*b* values to the sRGB color space using an online color calculator (e-paint.co.uk).

2.2. Fluorescence Microscopy-Spectroscopy

Wide-field luminescence images of the samples were captured using a Ximea xiD color camera in a fluorescence optical microscope (Biomed PR LUM) with excitation from a mercury lamp and a set of excitation/emission dielectric filters: UV (330–385 nm, emission λ>430 nm), and Green (480–550 nm, emission λ>590 nm).

To measure luminescence spectra and luminescence decay curves in different regions of interest (ROI), samples located in the focal plane of an optical microscope (Olympus IX71, 20x, N.A.=0.5) were excited by a femtosecond laser pulse with a wavelength of 344 nm, an energy of 5 pJ, and a duration of 350 fs. UV femtosecond pulses were generated by a BBO nonlinear crystal as the third harmonics of a 1033 nm Yb-doped diode-pumped solid-state amplifier system (TETA, Avesta Project). The luminescence emission signal was collected by the same objective lens and coupled via a microscope side channel to an input of a monochromator/spectrograph (Acton SP300i), while blocking the excitation laser with a dielectric filter (Thorlabs FESH650). Fluorescence emission spectra in the spectral range from 430 to 680 nm were registered with an electron-intensified CCD camera (PI MAX 2, Roper Scientific) installed at the monochromator output port. A GaAs hybrid photodetector of a time-correlated single-photon counting system (Becker&Hickl SPC-150N) was installed at the second monochromator’s output and was used to record fluorescence decay kinetics in a range of 0-200 ns at an emission wavelength selected with a monochromator.

2.3. Scanning Electron Microscopy (SEM)

Electron microscopy of unaged and aged paints samples was performed using a Thermo Fisher Scientific Prisma E microscope (Brno, Czech Republic). Samples for secondary electron imaging were sputter-coated with a 40 nm gold layer using a Quorum Technologies Q150R ES plus coater (Laughton, UK) and mounted with copper tape to enhance charge drainage. Chemical mapping and elemental analysis were conducted via energy-dispersive X-ray spectroscopy (EDS) operated in low-vacuum mode.

2.4. X-Ray Fluorescence Analysis

The paint pigment composition was determined using an EKROS XRF-9720 STINGRAY analytical X-ray scanning microprobe (EKROSCHIM, Russia) equipped for X-ray fluorescence analysis.

2.5. X-Ray Photoelectron Spectroscopy (XPS)

X-ray photoelectron spectroscopy (XPS) was performed using an Axis Ultra DLD spectrometer (Kratos Analytical, UK) equipped with a monochromatic Al Kα source (hν = 1486.7 eV, 150 W). Pass energies of 160 eV and 40 eV were applied for survey and high-resolution scans, respectively. The energy scale was calibrated using standard core levels of high-purity metals: Au 4f_5/2 (83.96 eV), Ag 3d_5/2 (368.21 eV), and Cu 2p_3/2 (932.62 eV). Samples were mounted on the holder using double-sided non-conductive tape. To compensate for sample charging, an electron neutralizer was employed, and the binding energy scale was referenced to the adventitious C 1s peak at 285.0 eV.

2.6. Attenuated Total Reflectance–Fourier Transform Infrared (ATR-FTIR) Spectroscopy

The infrared spectra of both paints’ samples were acquired using a Bruker Vertex-70 FTIR spectrometer equipped with a Platinum Diamond ATR accessory (Bruker, Germany). Crystals generally have a diameter of approximately ~2 mm. The spectra were recorded in attenuated total reflectance (ATR) mode over a range of 600–4000 cm⁻¹ with a resolution of 4 cm⁻¹, averaging 64 scans per measurement. For each sample, eight ATR-FTIR measurements were taken from different, randomly selected surface areas.

2.7. Micro-Raman Spectroscopy

Micro-Raman spectroscopy was employed as a complementary technique to ATR-FTIR to provide comprehensive chemical analysis and complete molecular fingerprints. Spectra were acquired using a Senterra Raman microscope spectrometer (Bruker, Germany), equipped with a 785 nm excitation laser and a 50× objective (numerical aperture, NA = 0.75). The spatial resolution calculated taking into account the diffraction limit was 1.22λ/NA = 1.3 μm. Raman spectra were obtained from ten randomly selected points on each sample.

2.8. Chemometric Analysis

The complexity and spectral variations observed in the obtained Raman and ATR-FTIR spectra necessitated the use of chemometric analysis, specifically principal component analysis (PCA), partial least squares, or projection to latent structures (PLS). These methods enable an effective comparison and interpretation of the photodegradation pathways of the two studied paint systems. Raw data pre-processing included smoothing, baseline correction, selection of chemically significant spectral regions, and normalization. Smoothing was performed using the Savitzky–Golay (SG) algorithm to remove random noise. Baseline correction was achieved through a combined sliding window method and polynomial fitting. Differentiating spectra allows for the isolation of peak positions while reducing or eliminating the influence of background noise and baseline. Peak positions were determined by recording the minima of the second derivative spectra; for poorly resolved bands, the maxima of the fourth derivative spectra were additionally evaluated. PLS with variable importance in projection (VIP) estimation was employed to assess the contribution and significance of each wavenumber (i.e., specific ATR-FTIR bands) in predicting the analyzed property. All pre-processing and subsequent mathematical analyses were performed using custom scripts and procedures developed in MATLAB.

3. Results and Discussion

3.1. Composition of Unaged LY203 (Cadmium Lemon) and Y201 (Cadmium Yellow Medium) Paints

Since the manufacturer does not disclose the exact composition of the paints, their composition was analyzed in this study. Scanning electron microscopy coupled with energy-dispersive X-ray spectroscopy (SEM-EDS) of the LY203 and Y201 samples determined their elemental composition (SI, Figure SI 1.1, SI 1.2/Table SI1.1, SI 1.2). EDS analysis of the LY203 paint identified the pigment as Cd_xZn_(1-x)S, where x=0.66±0.04. Subsequent X-ray fluorescence (XRF) analysis of this LY203 Cd_xZn_(1-x)S pigment yielded an x value of 0.59, which is consistent with the SEM-EDS results. For the Y201 paint, both SEM-EDS and XRF analyses confirmed that the pigment is CdS.

SEM-EDS detected carbon (C) and oxygen (O) originating from the polymer binder in both paints. Additionally, trace amounts of phosphorus (P) and sodium (Na) were detected via SEM-EDS in both formulations, likely deriving from sodium oligophosphate salts commonly used as dispersing agents. A minor concentration of aluminum (Al) indicated the presence of kaolin Al₂O₃ ·2SiO₂ 2H₂O. Furthermore, XRF analysis identified trace levels of transition metals, specifically Fe, Mn, Co, Cr, and Ni.

Attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectra of LY203 and Y201 demonstrated that the binder consists of a styrene-acrylic copolymer (Figures SI2.1 and SI2.2; peak assignments are detailed in Section/Table SI2).

Finally, Raman spectra recorded using a 785 nm laser (near off-resonance conditions) provided comprehensive molecular data on both the pigments and the binders. Figures SI3.1 and SI3.2 display the averaged Raman spectra of the unaged Y201 and LY203 paints, with peak assignments summarized in Section/Table SI3. Detailed analysis of ATR-FTIR and Raman presents below.

3.2. Comparison of SEM Images of Paint Surface Before and After Aging

Figure 1 displays scanning electron microscope (SEM) images of the Y201 and LY203 paints before and after 455 hours of aging. The micrographs reveal surface microheterogeneity within both the pristine and aged paint layers. SEM analysis further indicates morphological alterations, including minor chalking and enhanced surface roughness across both formulations (Figure 1). Notably, at high resolution (see insets in Figure 1; scale bar: 5 nm), this microheterogeneity is more pronounced in the aged LY203 paint than in its Y201 counterpart.

3.3. Fluorescence Microscopy

3.3.1. Wide-Field Fluoresence Micrscopy

Wide-field luminescence imaging of the Y201 and LY203 samples reveals that the color of the paint layer depends on the excitation wavelength (SI 4, Figure SI4.1). This dependency suggests that distinct optical transitions drive image formation across different ranges of the excitation spectrum. Fluorescence microscopy (SI 4, Figure SI4.1) further indicates that the heterogeneity observed via SEM (Figure 1) and widefield microscopy (Figures 2E, 2F, 2H, and 2I) stems from pigment particle agglomeration or coarse grinding. As shown in Figure 2, both the luminescence spectra and decay curves vary between the region of interest (ROI) associated with the dark spot (D2) and the ROI corresponding to the uniform layer (L1).

3.3.2. Luminescence Spectra in Different Region of Interest (ROI)

Figure 2A,2B,2C,2D displays the luminescence spectra of the pristine and aged Y201 (CdS) and LY203 (Cd₀.₆₆Zn₀.₃₄S) paints, each featuring a pronounced peak alongside a broad, long-wavelength band. The sharp peaks are attributed to near-band-edge exciton transitions, whereas the broad long-wavelength band is assigned to trap-state luminescence. Due to zinc doping, the peaks for the LY203 (Cd₀.₆₆Zn₀.₃₄S) dye are blue-shifted by approximately 20–25 nm relative to those of the Y201 (CdS) dye. Spatial variations in the luminescence spectra across different ROIs can be explained by localized differences in nanoparticle agglomeration or grinding defects. Across various ROIs, the peak positions range from 513 nm (2.42 eV) to 547 nm (2.27 eV) for the Y201 (CdS) paint, and from 480 nm (2.58 eV) to 499 nm (2.48 eV) for the LY203 (CdZnS) paint. Because these peak positions closely match the bulk band gaps of CdS (2.42 eV)[51,52] and Cd_0.66Zn_0.34S (~2.59–2.88 eV)[53,54], quantum confinement effects on the exciton peak position are negligible for the pigment particles in both paints. Typically, the ROI associated with the D2 dark spot exhibits a red-shifted luminescence peak compared to the L1 ROI, which may be linked to surface traps within the aggregates. Spectral evolution during paint aging varies by ROI, showing a general trend toward decreased exciton peak intensity and a relative increase in trap-band intensity. This aging trend is more pronounced in the LY203 paint than in the Y201 paint (Figure 2A, 2B). Splitting of the exciton peak was observed in specific D2 ROIs of the LY203 samples (Figure 2B). The short-wavelength component of this split peak is likely attributable to an increase in zinc concentration during pigment aging, alongside a corresponding blue shift of the edge exciton in CdxZn1-xS[54,55,56,57]. This localized increase in the zinc fraction within the solid solution is driven by photocorrosion processes, which selectively reduce cadmium ions, thereby enriching the remaining crystal lattice with zinc[58], as discussed below.

3.3.3. Luminescence Decay

Luminescence decay curves exhibit complex profiles that depend on the region of interest (ROI) within the paint layer (Figure 2G, 2J). The average exciton luminescence lifetime <τ> (Equation SI4.1) ranges from 0.8 to 5 ns for the Y201 paint and from 3 to 11 ns for the LY203 paint, depending on the ROI. However, it should be noted that in the D2 type ROI, the luminescence decay occurs somewhat slower than in the L1 type ROI. Meanwhile, the trap luminescence decay lifetime spans 40 to 65 ns. At a probing wavelength of λ>550 nm, the exciton decay is faster than the trap luminescence band, a behavior typical of semiconductor luminescence (SI, Figure SI4.2). Because average lifetimes for both transitions vary significantly across different ROIs, no clear pattern could be identified in the pigment's lifetime changes during paint aging.

3.4. Color Change During Aging

Upon exposure to light and moisture, the crystal lattice of cadmium yellow pigment can undergo chemical transformations, including lattice corrosion, alongside binder degradation. These processes generate new chemical species that can significantly alter the optical properties of the paint. Colorimetric measurements of photochemically aged samples quantified these changes using the total color difference (∆E*) (Table 1, Figure 3). The main criterion, the ∆E* value in the CIELab-76 system, was calculated using Equation (1)[59]:

ΔE = [(ΔL*)² + (Δa*)² + (Δb*)²]^1/2

(1)

where ∆L* = L*_i − L*₀, ∆a* = a*_i - a*₀, and ∆b* = b*_i - b*₀.

In this equation, index 0 denotes the sample prior to aging, and index i denotes the sample after a given aging period.

The results revealed a pronounced color change in the LY203 (Cadmium Lemon) paint, whereas the Y201 (Cadmium Yellow Medium) paint exhibited minimal alteration. Specifically, the ∆E* value for Cadmium Lemon increased sharply to 15.52 after 455 hours of aging, indicating a visually perceptible color shift toward a darker, less yellow hue. In contrast, the ∆E* for Y201 (Cadmium Yellow Medium) remained relatively stable at approximately 1.68 after 287 hours before slightly declining to 1.09 at 455 hours. This distinct variation provides critical insights for predicting the degradation pathways of these two paint formulations.

For Cadmium Lemon, while the a* value shifted slightly toward the red component and the b* value shifted toward the blue region, a significant decrease in the L* (lightness) value indicated pronounced darkening of the paint. This severe darkening confirms the formation of dark degradation products. Conversely, in the Cadmium Yellow Medium paint, the initial color change ((∆E* = 1.68 after 287 h) was driven by a slight decrease in L*and a* values, coupled with an increase in the b* value toward the yellow axis, indicating enhanced yellowness. This yellowing likely results from the photo-oxidation of the binder, which induces polymer degradation and generates new chromophores, thereby exposing the underlying yellow pigment particles. After 455 h, both the b* and a* values decreased, while the L* value increased, signifying surface chalking. This trend suggests pigment bleaching and the formation of common white degradation products, such as CdCO₃ and CdSO₄, which counteracted the initial color change induced by binder loss. Furthermore, the degradation of the paint film surface led to a rough, irregular topography that scatters light diffusely, further contributing to the pronounced lightening effect

3.5. ATR-FTIR Spectra

3.5.1. Averaged ATR-FTIR Spectra: Binder Changes During Artificial Aging

Figure 4A and 4C show the averaged ATR-FTIR spectra of Y201 and LY203 paints at different aging times. Minor spectral changes due to aging are resolved using the difference spectra ΔATR-FTIR, defined as: ΔATR-FTIR(t) = ATR-FTIR(t) - ATR-FTIR(t=0) (Figure 4B, 4D). The main spectral features of the differential ΔATR-FTIR(t) spectra of LY203 and Y201 paints in Figure 4 demonstrate the rise and decline of several characteristic bands. Specifically, a decrease (ΔATR-FTIR(t) < 0) is observed for the acrylic ester group peaks at 1724 cm⁻¹ (C=O carbonyl stretching)[2] and near 1159 cm⁻¹ (C–O ester stretching).[2][60][61] Bleaching (ΔATR-FTIR(t) < 0) also occurs in the 3100–3000 cm⁻¹ region, which is associated with =C–H stretching in the styrene phenyl ring and side-chain vinyl hydrogens[62][63][64], alongside a decline (ΔATR-FTIR(t) < 0) in the aromatic ring-breathing bands at 752 and 690 cm⁻¹. Conversely, the rise of bands at ~1753 cm⁻¹ and 1783 cm⁻¹ on the blue side of the 1724 cm⁻¹ acrylic ester peak (ΔATR-FTIR(t) > 0) can be tentatively assigned to the formation of anhydrides and γ-lactones. [2,30,43,44,49,65,66,67,68,69] Furthermore, an increase in intensity within the ~1700–1600 cm⁻¹ range (on the red side of the 1724 cm⁻¹ band) indicates the formation of unsaturated molecules with C=C bonds and new carbonyl compounds, such as aldehydes, ketones, and carboxylic acids. Finally, a significant increase (ΔATR-FTIR(t) > 0) is observed in the 3600–3100 cm⁻¹ region, corresponding to hydroxyl group vibrations. Taken together, these spectral variations demonstrate that aging induces significant binder oxidation in both LY203 and Y201 paints.

3.5.2. The Averaged ATR-FTIR Spectra. Spectral Features of Inorganic Additives and Pigments

Differential ATR-FTIR spectra ΔATR-FTIR(t) reveal chemical alterations in the pigments during aging. Specifically, positive bands (ΔATR-FTIR(t)> 0) emerge in the 1140–1000 cm⁻¹ range for both paints. These bands likely indicate the formation of cadmium sulfate (CdSO₄) as a pigment degradation product. This is supported by the characteristic ν₃ stretching vibration of the sulfate group, which typically appears in the 1140–1000 cm⁻¹ region. Furthermore, the bands at 563 and 636 cm⁻¹ can be tentatively assigned to CdSO₄.[5,18,70] While X-ray Photoelectron Spectroscopy (XPS) analysis confirmed the presence of S⁶⁺ sulfur in the aged samples (SI, Figure SI 5.1) , the baseline presence of a BaSO₄ filler makes this XPS analysis only an indirect confirmation of CdSO₄ formation. It is noteworthy that the increase in the intensity of the sulfate group bands is more pronounced in LY203 paint than in Y201 paint.

3.5.3. Chemometric Analysis of ATR-FTIR Spectra

Principal component analysis (PCA) and partial least squares (PLS) regression were utilized to identify the most significant spectral features in the ATR-FTIR spectra of aged paints. The PCA results presented in Figure 5 provided a preliminary overview of the primary differences between the unaged and aged paint samples (LY203 and Y201). The score plots illustrate these differences for both paint types as a function of aging time, revealing well-defined sample clustering. Specifically, Figure 5C and 5D display the PC1 versus PC2 score plots. For Y201, PC1 and PC2 explained 97.5% and 1.78% of the variance, respectively, accounting for a cumulative variance of 99.3%. For LY203, PC1 and PC2 captured 94.7% and 4.5% of the variance, respectively, totaling 99.2% of the data variance. The samples were primarily discriminated along the PC1 axis according to their aging stage. Evaluation of the corresponding PC1 loading vector (Figure 5A) indicated that this behavior was associated with the cleavage of ester bonds (negative peaks at ~1732 and 1159 cm⁻¹), the formation of anhydrides and γ-lactones (positive band on the higher-wavenumber side of ~1732 cm⁻¹), and the generation of unsaturated molecules containing C=C bonds alongside new carbonyl species, such as aldehydes, ketones, and carboxylic acids (positive band on the lower-wavenumber side of ~1732 cm⁻¹). Concurrently, the loading vector revealed an aging-induced increase in peak intensities within the 1140–1000 cm⁻¹ range, which is characteristic of CdSO₄.

To identify the ATR-FTIR spectral features that change significantly during paint aging, partial least squares (PLS) regression analysis was performed on the Y201 and LY203 paints. Figure 6 illustrates the variable importance in projection (VIP) scores derived from the PLS regression models. The VIP metric identifies the predictor variables that exert the greatest influence on the regression; specifically, VIP scores greater than 1 correspond to spectral bands that become statistically significant during aging, driving the spectral differentiation over time. As shown in Figure 6, the most pronounced aging-induced spectral changes for both paints occur in the 1800–1600 cm⁻¹ region. These bands indicate the formation of unsaturated molecules containing C=C bonds and new carbonyl compounds—such as aldehydes, ketones, carboxylic acids, anhydrides, and γ-lactones—concurrent with the degradation of ester bonds in the acrylic matrix. Conversely, the VIP scores in the 1200–1000 cm⁻¹ range reveal distinct behaviors between the two paints. For Y201, the VIP analysis highlights a prominent peak at 1159 cm⁻¹ (assigned to C–O stretching of ether bonds) alongside a weak band around 1120 cm⁻¹ associated with CdSO₄. In contrast, LY203 exhibits significant bands spanning the 1130–1020 cm⁻¹ range, which are strongly attributed to CdSO₄. This discrepancy suggests that CdSO₄ formation is more kinetically active in LY203 than in Y201. Furthermore, Figures SI6.1 and SI6.2 display the PLS regression scatterplots for the ATR-FTIR datasets, demonstrating clear clustering and differentiation based on aging duration (SI, Figure SI 6.1, SI 6.2, SI 6.3, SI 6.4)

3.6. Characterization by Micro-Raman Spectroscopy

Micro-Raman spectroscopy complements the results of ATR-FTIR spectroscopy by extending the vibrational spectra into the low-frequency spectral range characteristic of the phonon bands of CdS and Cd_xZn_1-xS pigments and providing higher spatial resolution in sample analysis. Thus, new features of the CdS and CdZnS pigments degradation process are revealed in the Raman spectra.

3.6.1. Averaged Micro-Raman Spectra in the Range of 180÷700 cm⁻¹ [9,10,36,71,72,73,74,75,76,77,78,79,80,81,82]

With 785 nm non-resonant excitation, the bulk Resonant Raman Scattering enhancement is absent. This allows weak, local symmetry-breaking configurations to stand out. Figure 7 shows the averaged micro-Raman spectra of both paints in the region of phonon Raman bands. Since CdS, Cd_xZn_1-xS are polar semiconductor, long-range Coulomb interactions lead to the formation of longitudinal (LO) and transverse (TO) optical branches. The Raman bands at 236, 253, and ∼301–305 cm⁻¹ represent the TO and LO contributions from h-CdS or c-CdS (Figure 7A). The band at ~210 cm⁻¹ can be tentatively assigned to surface or multi-phonon modes. The strong contribution around 590-618 cm⁻¹ clearly corresponds to the 2LO second harmonic. Raman assignments for Cd_xZn_1-xS are: peak appear close to ~210 cm-1, representing surface optical (SO) modes or surface disorders; ~213–240 cm-1 weak peaks associated with TO modes or multi-phonon scattering; LO mode of the ternary alloy with a frequency of ~305 cm⁻¹ and the corresponding overtone (2LO) modes in the range of ~590-618 cm⁻¹. The I_2LO/I_LO intensity ratio serves as an indirect indicator of exciton–phonon coupling and is frequently correlated with crystallinity in cadmium chalcogenides and their ternary alloys.[71,78,81,82] Generally, a higher ratio signifies stronger coupling and superior crystal quality, whereas a lower ratio implies an increased degree of structural disorder. For the CdS pigment in the unaged Y201 paint, the I_2LO/I_LO ratio is close to 0.9; in contrast, that of the Cd_0.66Zn_0.34S pigment from unaged LY203 is lower by a factor of 1.27, at approximately 0.71. This indicates that the incorporation of zinc induces substitutional disorder and lattice deformation, thereby weakening the exciton–phonon coupling. Over time, a progressive decline in the I_2LO/I_LO intensity ratio is observed in both systems (Figure 7), signifying structural degradation of the crystal lattice and defect formation during paint aging. Notably, the rate of decline is more pronounced for Cd_0.66Zn_0.34S than for CdS: the ratio drops from 0.9 in pristine CdS to 0.65 after 455 hours of aging, whereas it falls from 0.71 to 0.48 for Cd_0.66Zn_0.34S. Further evidence of Cd_0.66Zn_0.34S degradation in the LY203 paint is the appearance of a broad band at ~385 cm⁻¹ (Figure 7B), which indicates the formation of cadmium oxide (CdO).

3.6.2. Averaged Micro-Raman Spectra in the Range of 800-1900 cm⁻¹ [49,66,67,68,69,83,84,85,86,87]

Averaged micro-Raman spectra within the spectral region characteristic of the binder (SI, Figure SI 7.1) corroborated the ATR-FTIR spectroscopy findings. Specifically, the attenuation of the peaks associated with the styrene and acrylate components at ~1000 and ~1730 cm⁻¹, alongside the intensification of the oxidation product peaks at ~1770 cm⁻¹, confirms the photo-oxidation of the binder and the subsequent formation of γ-lactones and anhydride structures (SI, Figure SI 7.1).

3.6.3. Spatially Resolved Micro-Raman Spectra

Micro-Raman spectroscopy provides highly detailed spatial information regarding the paint layer surface, offering significantly greater spatial resolution than Attenuated Total Reflection Fourier-Transform Infrared (ATR-FTIR) spectroscopy. While ATR-FTIR records spectra from a relatively large spot size d≈ 2 mm, micro-Raman spectroscopy achieves a spot size of approximately d≈ 2 μm (SI, Figure SI 7.2, SI 7.3). Consequently, micro-Raman spectra reveal localized micro-inhomogeneities within the sample to which ATR-FTIR remains insensitive. Figure 8 displays the micro-Raman spectra and corresponding optical micrographs of the pristine and aged paint samples in the spectral range of 180 ÷ 700 cm⁻¹ . Following aging, significant spectral variations emerge for the LY203 paint within the regions of ~300 ÷ 400 cm⁻¹ and ~ 500 ÷ 600 cm⁻¹ (Figure 8G); conversely, the pigment spectral region for the Y201 paint remains virtually unchanged (Figure 8C). Under non-resonant excitation at 785 nm, bulk resonant Raman scattering enhancement is absent, allowing weak, localized configurations that break crystal symmetry to become prominent. Consequently, features within the ~300 ÷ 400 cm⁻¹ range serve as highly sensitive indicators of CdZnS degradation. From a chemical point of view, the Cd-S bond is less stable and undergoes photocorrosion much faster than the Zn-S bond. Within the Cd_0.66Zn_0.34S alloy, photocorrosion induces the selective leaching of cadmium ions (Cd²⁺) from the surface lattice. Because the fundamental 1LO phonon mode of pure ZnS is located precisely at 350 cm⁻¹, advancing photocorrosion enhances this peak at specific localized points during micro-Raman measurements, as demonstrated in Figure 8G. This phenomenon directly monitors the transformation of the balanced alloy surface into a highly defective, Zn--terminated shell, a finding consistent with the blue-shifted luminescence peaks identified in Figure 2B and 2B1. At alternative measurement points, the intensity of the micro-Raman peaks at 350 cm⁻¹ decreases, accompanied by the appearance of localized structural modes associated with CdSO₄ ⋅ H₂O or ZnSO₄ ⋅ 7H₂O. The broad shoulder spanning 500–540 cm⁻¹ is tentatively attributed to the ν₄(SO₄²⁻) bending modes of weakly crystalline or amorphous metal sulfates forming at the interfaces of the paint film. Furthermore, the spectral region between 540–570 cm⁻¹ likely corresponds to amorphous ZnO or CdO phases characterized by disordered surfaces associated with surface optical phonons or localized defect complexes[88]. Due to the violation of crystal symmetry rules in the degraded pigment, these phases manifest as broad, low-intensity bands. Notably, optical microscopy reveals that black spots develop mainly on the aged LY203 paint, while the aged Y201 paint remains almost free of such defects. All these observations indicate that LY203 paint undergoes highly heterogeneous photocorrosion.

3.6.4. Principal Component Analysis of Micro-Raman Spectra in the 900–1900 cm⁻¹ Window

The 900–1900 cm⁻¹ spectral window captures specific variations in the acrylic binder. The microheterogeneity of the samples is reflected by a wider spread of observations in the micro-Raman PCA scatterplot (Figure 9, SI, Figures SI 7.4, SI 7.5, SI 7.6, SI 7.7) than in the ATR-FTIR PCA plot (Figure 5). PC1 (71.6% variance) clearly separates the groups: Y201 paint spectra display negative scores, while LY203 spectra display positive scores. Additionally, Y201 forms a more compact, well-defined cluster than LY203.

The PC1 loading profile (L1-PC1) features pronounced, broad bands with positive coefficients. Given that the PC1 vector progresses from the negative Y201 cluster to the positive LY203 cluster, these positive loadings represent spectral features with greater intensity in the LY203 sample. Specifically, the broad band at 1000–1200 cm⁻¹ (maximum at ~1066 cm⁻¹) stems from sulfate groups (SO₄²⁻). Moreover, the bands at 1220–1400 cm⁻¹ (~1288 cm⁻¹) and 1700–1820 cm⁻¹ (~1766 cm⁻¹) reflect photo-oxidation and chain scission of the aging acrylic emulsion matrix. This degradation pathway produces oxygen-containing fragments (e.g., peroxides, hydroperoxides, and secondary oxidation intermediates) that display weak C–O skeletal vibrations within 1000–1150 cm⁻¹. Concurrently, main- and side-chain cleavage yields secondary esters and monomers. These terminal segments and isolated fragments modify the original, broad C–O single-bond stretching bands of the acrylic skeleton (1100–1200 cm⁻¹). Together, these PCA scores and loading plots demonstrate that photo-oxidative degradation is markedly more advanced in the LY203 (CdZnS) matrix than in Y201 (CdS). This accelerated degradation drives concurrent deterioration of both the pigment and binder, corroborating the complementary ATR-FTIR data.

The PC2 loading profile (L2-PC2, 5.4% variance) reflects binder degradation, displaying negative peaks at 1776 cm⁻¹ and 1350 cm⁻¹. Based on the orientation of the PC2 vector relative to aging time, negative values signify features that intensify with aging. The 1776 cm⁻¹ band is attributed to γ-lactone or anhydride formation[49], where anhydrides result from the dehydration of neighboring carboxylic acid groups generated during acrylic aging. This accumulation confirms advanced oxidation of the polymeric binder. Because the carbon G-band near 1580 cm⁻¹ is absent, the 1350 cm⁻¹ peak does not indicate disordered carbon[89]. It instead stems from C–H vibrations, aliphatic chains, or the skeletal vibrations of cross-linked polymers. Artificial aging breaks long polymer chains into shorter, disordered fragments. This chain scission raises the concentration of terminal, localized aliphatic groups, amplifying the 1350 cm⁻¹ signal. PCA of individual spectra reveals that these variations occur predominantly in LY203 rather than Y201 (SI Figures SI 7.4, SI 7.5).

3.7. Photocatalytic Degradation Mechanisms of Cd₀₆₆Zn_0.34S (Cadmium Lemon, LY203) and CdS (Cadmium Yellow Medium, Y201) Paints

3.7.1. Energy Levels in Semiconductor Pigments CdS and CdxZn1-xS

A comparison of the aging processes of the two acrylic paints—cadmium sulfide (CdS) in Y201 and alloyed Cd_xZn_1-xS (x = 0.66 ±0.04) in LY203—reveals that LY203 undergoes a more pronounced color change. This chromatic alteration is accompanied by a more severe degradation of the inorganic pigment and active oxidation of the organic binder via polymer chain scission. Because the two paints differ primarily in their inorganic pigment composition, their distinct aging behaviors can be attributed to differences in the photoactivity of the respective semiconductor photocatalysts. Luminescence measurements yielded a bandgap of 2.45 ± 0.07 eV for Y201 (CdS) and 2.63 ± 0.05 eV for LY203 (Cd_0.66Zn_0.34S), aligning well with reported literature values (2.47 eV and 2.61 eV, respectively)[51,52,53,54,90,91]. The documented conduction band (CB) and valence band (VB) potentials are −0.55 V and +1.92 V vs. NHE for CdS, and approximately ~0.6 V and ~+2.0 V vs. NHE for Cd_0.66Zn_0.34S.[16,52,75,92,93,94] Notably, the bandgap for the LY203 paint is roughly 200 meV wider than that of the Y201 paint, and the photogenerated electrons and holes in LY203 possess slightly stronger redox capabilities. Incorporating Zn elevates the CB and VB energy positions, thereby enhancing the thermodynamic driving force for electron- and hole-mediated redox reactions.

Excitation with a photon energy greater than the semiconductor bandgap promotes an electron to the conduction band (CB), generating a corresponding hole in the valence band (VB) (Scheme 1). These photogenerated charge carriers can undergo radiative recombination, emitting a photon that manifests as an exciton luminescence peak (Figure 2). Alternatively, the electrons and holes can be captured by bulk or surface traps. This trapping lowers their energy states, leading to the emission of photons with energies below the bandgap. Given the broad energy distribution of these trap states, their radiative recombination generates broadband emission (Figure 2). Concurrently, nonradiative recombination can occur via thermal dissipation. This process takes place predominantly through trap states because their energy steps are substantially smaller than the pigment bandgaps, allowing the released energy to be readily absorbed by low-energy lattice phonons (25–35 meV; 200–300 cm^-1). In an ideal semiconductor photocatalyst, the relationship between photocatalytic efficiency and excitation wavelength (λ) follows a distinct threshold pattern (λ_edge)., where efficiency rises sharply (λ < λ_edge) and typically plateaus or mirrors the absorption profile.[92] Radiative and nonradiative recombination pathways directly compete with photocatalysis and photocorrosion, both of which are mediated by these photogenerated electrons and holes. Crucially, neither radiative nor nonradiative recombination pathways induce damage to the pigment particles themselves.

3.7.2. Photocorrosion in CdS and Cd_xZn_1-xS [16,52,91,93,94,95,96,97,98,99,100,101,102]

In chalcogenide materials, the anodic potential for hole-mediated photocorrosion is often less positive than the water oxidation potential, making pigment self-destruction thermodynamically favorable. Photocorrosion in CdS and Cd_xZn_1-xS is driven by photogenerated holes h⁺ that oxidize surface sulfide ions (S^2-). This hole-mediated photocorrosion (Reactions 1 and 2) is a non-elementary process involving an intermediate sulfur radical S^·- culminating in metal ion dissolution and elemental sulfur precipitation. In Cd_xZn_1-xS , photocorrosion is spatially non-uniform. Because Cd–S bonds have lower energy than Zn–S bonds, holes preferentially attack Cd–S surface states. The resulting leaching of Cd²⁺ leaves an elemental sulfur shell that scatters light and deactivates photocatalytic sites. Recent DFT and experimental studies confirm this selective bond breakage, often termed "controllable photocorrosion". This mechanism explains the post-aging short-wavelength luminescence peak (Figure 2B); photocorrosion induces surface Zn-enrichment in Cd_xZn_1-xS, blue-shifting the emission spectrum[58,97,101].

Cd_xZn_1-xS + 2h⁺ → x₁Cd²⁺ + x₂Zn²⁺ + S⁰ in the absence of O₂

(1)

In the presence of moisture and oxygen, degradation accelerates due to the multielectron oxidation of sulfur into sulfates. Rather than stopping at elemental sulfur (S⁰), the lattice is deeply oxidized to soluble sulfate ions, destroying the surface structure. This pathway is corroborated by the presence of CdSO₄ and CdCO₃ in both paints, as confirmed by ATR-FTIR and micro-Raman data. Ultimately, the proliferation of trap states during photocorrosion drives the steady increase in broadband luminescence observed over time (Figure 2).

Cd_xZn_1-xS + 4h⁺ + O₂+ 2H₂O → x₁Cd²⁺ + x₂Zn²⁺ + SO₄^2-+4H⁺

(2)

While anodic photocorrosion proceeds via hole-mediated pathways, photogenerated electrons can concurrently reduce dissolved or surface-bound metal ions (Reaction 3) through a thermodynamically favorable pathway (Scheme 1). This process is recognized as reductive or electron-mediated photocorrosion.[96]

Cd²⁺ + 2e^- → Cd⁰

(3)

The formation of metallic cadmium (Cd⁰) likely induces the darkening of the paint film. This specific mechanism provides a robust explanation for the colorimetric data, which revealed a significantly more pronounced darkening in LY203 relative to Y201. Concurrently, photocatalytic reactions generate highly reactive radical species. Due to their elevated oxidation potentials, these radicals can actively oxidize surface sulfide anions (S^2-), —a degradation pathway formally recognized as indirect (or radical-mediated) photocorrosion.[103]

Cd-S + RO₂^. (HO₂^.) + H⁺ → Cd₂⁺ + S^-. + ROOH (HOOH)

(4)

The photocatalytic mechanism of polymer degradation is governed by the photogeneration of electron–hole pairs and the subsequent formation of reactive radical intermediates. The binder matrix consists of a styrene-acrylic copolymer that is highly susceptible to radical attack. During the photocatalytic process, photogenerated electrons reduce ambient oxygen to yield superoxide radical anions O₂^-.; Reaction 5), which protonate to form hydroperoxyl radicals HO₂• under weakly acidic conditions (Reaction 6). The HO₂• species dominates this environment, acting as a significantly more aggressive electrophilic oxidant (Reaction 7). An alkyl radical R•, which is subsequently transformed into peroxyl RO₂• and alkoxyl RO• radicals (Reactions 8–10). Polyethylene glycol (PEG) is a polyether compound with terminal hydroxyl groups that serves as a surfactant and wetting agent. Usually adsorbed on the CdS surface or bonded with Cd²⁺ as an alkoxide, it can be directly oxidized by holes (11). The hole abstracts one electron from the oxygen atom (or the $\alpha$-carbon), forming a highly reactive $\alpha$-hydroxyalkyl radical. Similar reactions can occur with other alkoxides in the binder. Thus, different pathways for the formation of radical intermediates can be realized, both through the formation of active oxygen species via the primary O₂^.- radical and through a direct reaction with the hole. Ultimately, the β-scission of these polymer-bound radicals (Reaction 12) serves as a fundamental reaction pathway driving backbone cleavage and overall binder degradation.

O_{2 (ads)} + e⁻_cb → O₂^•− _(ads)

(5)

O₂^•− _ads + H⁺ ↔ HO₂^• pKa ~ 4.80÷4.88

(6)

HO₂^• + R-H → HOOH + R^•

(7)

R^• + O₂ → RO₂^•

(8)

RO₂^• + R’H → ROOH + R’^•

(9)

ROOH → RO^•+ HO^•

(10)

H-(OCH₂CH₂)-OH + h⁺ → H-(OCH₂C^•H)-OH + H⁺

(11)

R^•→ β-scission → R′^• + vinyl groups

(12)

RO₂^•, RO^•→ intermediate products

(13)

The likelihood of direct oxidation of the acrylic component within the binder matrix is low. The oxidation potential of acrylate by photogenerated holes h⁺ is exceptionally high; direct electrochemical oxidation or decomposition typically requires elevated anodic potentials exceeding +2.0 V to +2.5 V vs. NHE.[104] Styrene (monomer) electro-oxidation generally initiates at a much lower potential of approximately +1.33 V vs. RHE (Reversible Hydrogen Electrode). Polystyrene (PS) features a saturated, insulating hydrocarbon backbone where the electroactive vinyl group has been polymerized. Polystyrene does not exhibit a distinct direct thermodynamic oxidation potential under standard voltammetric conditions. Concurrently, surfactants and dispersants—specifically polyethylene oxide (PEO)-based surfactants required for emulsion stability—exhibit significantly lower oxidation potentials than the acrylate core.[105] For instance, the terminal hydroxyl groups (–OH) on unfunctionalized PEO undergo oxidation at just +0.16 V vs. NHE. Therefore, their oxidation by h⁺ is highly feasible serving as a key alternative trajectory for radical generation.

Concurrently, degradation of the styrene-acrylic copolymer paint binder may proceed photochemically under simulated environmental exposure with daylight filters.[106,107] Styrene-acrylic binders typically exhibit greater susceptibility to photochemical degradation than pure acrylic binders. [108,109,110] This vulnerability is attributed to the aromatic styrene rings, where conjugation between the benzene ring and the vinyl group reduces ultraviolet (UV) stability. The susceptibility of styrene increases under harsher UV conditions due to π-π* transitions in the UV-C range (λ<280 nm). Styrene also exhibits weaker secondary absorption bands in the UV-B range (280–290 nm), while remaining photochemically inactive in the UV-A region (315–400 nm). Because the daylight filters utilized in these aging experiments feature a sharp cutoff at approximately 290–295 nm, they effectively block UV-C radiation and transmit a minimal natural UV-B fraction of only 0.5%–1.1%. Consequently, the observed binder degradation occurs predominantly through pigment-mediated photocatalytic pathways rather than direct, intrinsic photochemical reactions.

4. Conclusions

In summary, this investigation evaluated the environmental aging dynamics of Y201 (CdS) and LY203 (Cd_0.66Zn_0.34S) acrylic artist's paints, yielding several key insights into heritage material preservation:

• Colorimetric Stability: Under daylight exposure, the Y201 paint demonstrated exceptional color stability ΔE*= 1.1, whereas the alloyed LY203 paint underwent severe alteration ΔE*= 15.52.
• Microscopic Inhomogeneity: Light and fluorescence microscopy revealed large pigment agglomerates throughout the paint layer. Photochemical activity and spectral properties varied significantly between regions of uniform pigment distribution and those surrounding these coarse clusters.
• Because natural daylight contains a small fraction of UV radiation, photooxidative degradation of the binder can occur. Furthermore, binder degradation is driven by photocatalysis rather than direct photolysis. Visible light with energies above the semiconductor bandgap λ<λedge ≈ 500 nm photoexcites the semiconductor pigment, thereby accelerating polymer oxidation.
• Inorganic vs. Organic Drivers: Vibrational spectroscopy (ATR-FTIR and micro-Raman) paired with chemometrics showed that while binder oxidation profiles are nearly identical for both paints, the macroscopic color change is dictated entirely by the photocorrosion of the inorganic pigment.
• Lattice Disordering & Selective Cleavage: Photocorrosion induces structural disorder in the crystal lattice, as evidenced by a decrease in the Raman I_2LO/I_LO intensity ratio. In LY203, this selective attack results in surface Zn-enrichment, altering its luminescence profile.
• Reductive Darkening: Electron-mediated photocorrosion can promote the reduction of Cd²⁺ to metallic cadmium Cd0, providing an additional mechanism for paint darkening.
• Trap-State Coloration: Disruption of the semiconductor lattice generates a dense, continuous distribution of sub-bandgap electronic trap states. These traps express themselves as broad electronic absorption bands that directly cause the observed color shifts, while concurrently giving rise to the broad emission bands seen after aging.
• From a practical standpoint, mitigating photocorrosion is essential to improve the durability of Cd_xZn_1-xS (x=0÷1) artistic pigments. While existing strategies primarily target hole-mediated photocorrosion, this paper highlights the critical need to address electron- and radical-mediated degradation pathways including reaction oxygen species as well. To combat these diverse mechanisms, several advanced strategies can be deployed, including elemental doping, defect engineering, protective layer deposition, polarization management, heterojunction design, morphology optimization, and reaction environment modulation. [95]