In Situ Ultraviolet Fluorescence Spectroscopy for the Characterization of Translucent Calcite Coatings over Prehistoric Cave Art

Abstract

The conservation of prehistoric cave art involves complex taphonomic processes such as the formation of calcium carbonate coatings. Often invisible to the naked eye, these deposits are essential for understanding the evolution of the artworks and monitoring their preservation. While few noninvasive techniques probe the concretion layers, this work explores the potential of ultraviolet fluorescence to characterize calcite veils in the Chauvet cave, an Upper Paleolithic site displaying some of the oldest drawings known to date. Here, we show this technique is capable of distinguishing limestone from speleothem calcite, despite their similar crystallographic nature, thanks to organic impurities embedded in the carbonate matrices. A custom-built installation set up in the harsh conditions of the cave enabled ultraviolet fluorescence spectra to be measured at working distances of up to 2.5 m. Data collected in situ were processed according to a suitable fluorescence model, providing stratigraphic information on the translucent calcite layer covering the drawings and highlighting significant spatial heterogeneity in the concretions. Additionally, we observed localized orange emissions suggesting local thermal alteration of calcium carbonate by hot charcoals.

Introduction

Among the oldest forms of human artistic expression, Paleolithic rock art has been exceptionally preserved in caves for tens of thousands of years, across a few hundred archeological sites particularly in Spain and France. However, the artworks found in subterranean environments remain very fragile. Their condition depends on a delicate equilibrium of environmental parameters, raising major concerns for the management of prestigious sites and their accessibility to the general public.

Cave art, as observed today, results from a variety of alteration processes, prompting a multidisciplinary approach to the decorated walls where taphonomy is considered an essential part of their analysis. , Chemical, biological and geomorphological phenomena all modify the perception of the artworks, for instance by distorting observed colors or even obliterating panels, thereby biasing the apparent spatial distribution of art in the site. Among these processes, the crystallization of mineral coatings over the decorated panels stands out for its ambiguous role relative to the preservation of rock art. While overlying thick or opaque crusts can conceal the artworks, a translucent mineral coating may on the contrary behave like a natural varnish. In this respect, the case of the Grande Grotte of Arcy-sur-Cure (Yonne, France) is exemplary: the presence of an opaque calcite layer completely masked the Paleolithic paintings until the crust was abraded during restoration work. Calcium carbonate layers growing on paintings in karst-related contexts are particularly interesting for their potential as chronological indicators. Uranium thorium dating of overlying layers can provide terminus ante quem for paintings, although the application of this technique is controversial.

Few analytical methods have been documented for the characterization of mineral coatings on decorated panels. While the analysis of cross sections from microsamples gives valuable information on the surface condition, − it is unable to provide a global overview of the mineral veils. Noninvasive in situ methods are better suited in this regard and align with conservation policies, but they complicate the interpretation of the stratigraphy. Conventional techniques focusing on pigment identification such as X-ray Fluorescence (XRF) and Raman spectroscopy can detect components associated with mineral crusts. Studies mostly identify calcite, , as well as a wider range of mineral coatings such as hydroxyapatite, calcium oxalates , and gypsum. , However, these techniques fail to provide stratigraphic information and cannot differentiate certain key components by elemental (oxalate and carbonate) or crystallographic analysis (calcium carbonates of distinct diagenetic origins). Confocal XRF has demonstrated its potential for the characterization of a calcite crust overlying the paint layer although the measurements were not carried out in situ.

A less conventional technique, ultraviolet fluorescence (UVF) spectroscopy and imaging have recently revealed the presence of a silica skin on Paleolithic paintings through the emission of uranyl complexes. Likewise, calcium carbonate can be detected indirectly by UVF due to various organic and inorganic fluorescent impurities incorporated into the carbonate matrix. The different fluorophores associated with speleothems and limestone can enable their differentiation despite their similar elemental and crystallographic nature. In particular, the incorporation of organic matter from surface ecosystems makes the fluorescence of speleothems dependent on historical environmental parameters.

The characterization of the calcite veils is of particular interest for the Chauvet cave (Ardèche, France). Hundreds of paintings and engravings of rhinoceroses, mammoths, lions, horses, bisons, and other fauna remain in an exceptional conservation state, owing to multiples collapses that sealed the cave during the Upper Paleolithic. With its decoration dated at 37 ky cal. BP, this UNESCO World Heritage site gives a unique insight into the very beginning of art. The ongoing multidisciplinary scientific study of the cave highlights the complexity of the site, characterized by diverse conservation states of the panels due to several taphonomic processes. Among them, the calcite veils are investigated with the aim of understanding the original aspect of the artworks and their vulnerability, especially in the context of the cave’s exposure to climate change. However, this study faces a major analytical challenge: the preservation of prehistoric soils, with well-preserved animal traces and human footprints, restricts access to permanent walkways. As a result, the working distances to the panels range from 1 to 6 m, rendering conventional analytical techniques inadequate. The objective of this study is therefore to develop an analytical protocol tailored to the site’s constraints and to conduct in situ characterizations of the calcite coatings on the artworks.

In this study, we explore the potential of UVF techniques for the characterization of these coatings, from a global survey with RGB fluorescence imaging to chemical identification using emission spectroscopy. We demonstrate the ability of this approach to distinguish the limestone from calcium carbonate concretions owing to distinct fluorophores incorporated in the carbonate matrices, overcoming conventional analytical limitations faced by XRF or Raman spectroscopy. A simplified modeling of the decorated wall is proposed, enabling appropriate interpretation of the spectroscopic data. Using a custom spectroscopy setup tailored to the conditions of the cave, we perform noninvasive in situ measurements at a working distance of 2 m. The data provide evidence that a translucent calcite veil is covering the paintings, with significant variability in certain areas. Further discussion will be proposed on an unusual fluorescence signal and on the semiquantitative nature of the method.

Materials and Methods

Panels

This study focuses on an area of the Chauvet cave named Secteur des Chevaux in the Hillaire chamber (Figures and S1). It comprises the Horses Panel, the Alcove of the Lions, and the Reindeer Panel, of which the Horses Panel and the left-hand wall of the Alcove were particularly studied here. These panels display a complex range of techniques despite their monochromatic appearance and the extensive use of charcoal as pigment. Prehistoric artists prepared the bedrock support by scraping away the overlying clay; they obtained subtle shades by blending the pigment and occasionally incorporating brown clay; engravings were added to enhance the contours. For our measurements, the working distance to these panels ranged from 1.5 to 3 m. The majority of the panels exhibit no visible encrustations, except for the Alcove of the Lions.

1.

Custom UVF spectroscopy setup. (A) Diagram of the setup. (B) Setup operating on the Horses panel.

Reference Samples

Natural materials were sampled in and around the cave and artificial references were synthesized for comparison:

- (S1) Limestone rocks representative of the bedrock were sampled in the vicinity of the site, from the same geological context (Barremian-Bedoulian limestone layers of the Urgonian facies)

- (S2) Three speleothem samples were obtained from previous studies (Figure S2). Different strata are visible on cross sections due to variations in crystallinity and trace organic matter content.

- (S3) Charcoal samples were synthesized from Pinus sylvestris wood at various temperature. Cubes of 1 cm³ were placed in a tubular furnace under continuous N₂ flow (2 L/min) for 20 h at temperatures of 400, 500, or 700 °C. The temperatures represent different degrees of wood carbonization.

After being ground in an agate mortar, samples S1 and S2 were analyzed by X-ray Diffraction using a D8 Discover (Bruker) with nonmonochromated Cu Kα radiation (40 kV, 30 mA). The step size was set at 0.01° with an integration time of 0.1 s. The diffractograms show that both limestone and speleothem references consist of calcite (Figure S3).

We analyzed these samples before and after chemical or thermal treatments to investigate transformations that may have occurred in the cave, at the time of the occupation or during aging, and to assess the presence of organic molecules in the rocks:

- Samples S1 and S2 were subjected to acid digestion by progressive addition of a 6 M HCl solution to avoid altering the solution fluorescence. The remaining solid was removed by centrifugation.

- Samples S1 and S2 were heated in air at 500 °C in a muffle furnace for 1 and 24 h. This procedure aimed to mimic the effects of thermal treatment on the optical properties of these materials.

- To investigate potential long-term alteration of charcoals, samples S3 were oxidized following an adapted protocol from Deldicque et al.: powdered chars were added to 30% H₂O₂ solutions (unstabilized, VWR AnalaR NORMAPUR). The solutions were stirred at 90 °C for 48 h, resulting in a complete disappearance of the chars. The validity of this protocol was assessed by comparing the resulting infrared spectra with archeological data.

UVF Imaging

Luminescence imaging was performed under excitation by a monochromatic LED centered at 365 nm using a setup developed in cooperation with LUMETIS (Labège, France). Images were acquired with a 12-bit 20 Mp FLIR Blackfly S camera. It was mounted with a long pass filter and a 16 mm or 50 mm Kowa HC-VIS-SW camera lens.

UVF Spectroscopy

The instrument design was determined by the size desired for the measurement area. A 1 mm² spot size was deemed necessary for this study due to the heterogeneity of the cave walls. With a fixed light source, the conservation of etendue imposes a compromise between distance and spot size. Given the constrained working distances and the weak intensity of fluorescent emission, we chose to bring a lightweight optic close to the object. To achieve this, a 3-m-long coaxial optical fiber bundle (Vis-NIR Ocean Optics) was used to bring the collimated light from the wall to the walkway, resulting in an operator-to-panel distance of up to 2.5 m. Among the seven 600 μm-thick fibers, six were used for collection and one for excitation. The fiber bundle was mounted on a carbon fiber pole controlled by precision translation stages (Figure ).

The excitation source was a 365 nm UV LED (OSRAM LZ1-10UV0R, 1.3 W optical power, 64° beam angle) mounted with an OD6 bandpass filter blocking above 380 nm. This excitation wavelength allows for an efficient excitation of organic molecules incorporated into calcium carbonate, as shown by excitation–emission matrices acquired on speleothems, , while allowing a safe handling by the operator. Exposure of dyed papers to the beam showed no fading, confirming safe operation on cave walls. On the detection side, the light propagating in the second arm of the coaxial fiber was filtered by a long-pass OD4 filter (400 nm cutoff) which blocks the excitation peak while transmitting the visible part of the spectrum. To maximize optical throughput and maintain paraxial conditions, the light beam was collimated by a fiber lens prior to the filter and then collected by a second collimator connected to a second optical fiber (SR Ocean Optics). This fiber delivered the light to an Ocean Optics USB4000 spectrometer with a 200-μm slit and a resolution of 10 nm. At the distal end of the coaxial fiber, a plano-convex lens collimated the light beam, resulting in a probe-to-surface distance of 5 cm determined by positioning the probe to maximize signal intensity. This allows for safe operation of the setup even on the relatively unstable walkways. Due to the coaxial geometry, the excitation and detection fibers terminate in close proximity, leading to a strong specular reflection on the plano-convex lens. For fluorescence measurements, specular reflection of the excitation beam is not expected to change the emission spectrum because excitation and emission occur at distinct wavelengths. However, a weak signal was detected by the spectrometer without a target present. The dark signal was therefore recorded with the excitation light active in the absence of a target. The measured signal included both the spectrometer electronic noise and the residual signal arising from specular reflection on the lens.

The setup was powered by a 24 V battery. The data were acquired with the spectrometer through a custom-built Python interface using the python-seabreeze library. Integration times ranged from 1024 to 16384 ms with 5 averaged scans. Data were first normalized to their integration time. The dark signals acquired with the corresponding integration time were then subtracted. Finally, the data were multiplied by the spectral calibration function which was obtained using a factory-calibrated Ocean Optics SR4 spectrometer as a reference and an Ocean Optics HL-2000-FHSA lamp as a broadband light source. However, absolute flux calibration was not performed, resulting in the correction of the instrument response function but not the expression of measured intensity in absolute irradiance. Specular effects were considered negligible because the characteristic size of the surface roughness is much smaller than the beam size. Non-Negative Least Squares (NNLS) optimization was performed using the scipy.optimize.nnls function from scipy library on the spectral data restricted to the 430 to 610 nm region. The root-mean-square error (RMSE) between observed and NNLS-fitted spectra was normalized by the range (maximum minus minimum) of the observed signal, yielding a dimensionless normalized RMSE that enables comparison of fit quality across spectra of differing amplitude.

Sixty emission spectra were collected on the Chauvet Cave panels with the custom-built instrument, under a relative humidity rate of approximately 100% and a temperature of 13 °C. Measurement points were chosen to sample both limestone surfaces and areas covered with black pigment. This setup was also used to characterize natural and artificial samples S1 and S2 before and after chemical and thermal treatments.

Analytical Development

Fluorescence of the Components of the Decorated Panels

Reference samples were measured to evaluate the capability of UV fluorescence spectroscopy to differentiate limestone from speleothems.

Fluorescence emission from the speleothem reference shows a single peak centered at approximately 450 nm. Since the literature indicates this blue fluorescence arises from incorporated organic matter from the surface ecosystems, we extracted the organic molecules from the speleothem by acid digestion of the mineral matrix. Fluorescence emission from the resulting solution was very similar to that of the intact speleothem (Figure A), confirming the organic origin of the signal.

2.

Analytical principle. (A) UVF emissions recorded from the reference samples (in black) and from the solutions obtained after acid digestion (in gray). (B) Model of the decorated panels.

The limestone reference exhibits a fluorescence spectrum with two peaks centered at approximately 450 and 580 nm. After acid digestion of the mineral matrix, only the blue peak remained (Figure A), indicating an organic origin attributable to diagenetic organic residues. , The second peak was isolated after heating limestone and, unexpectedly, speleothem samples as well. As reported in the literature, several sources can be proposed, such as aromatic fluorophores, structural defects or substitutions of metal ions in the carbonate matrix. The interpretation of this second component remains challenging and is addressed in the Discussion section.

The diversity of organic emissions from speleothems results in a variation of the fluorescence peak in a range of up to 20 nm. ,− Fifty fluorescence spectra were measured at 500 μm intervals along the growth axis of the speleothem sample, demonstrating very limited spectral variability among the Chauvet Cave concretions (Figure S2A–C). In contrast, the limestone exhibits an additional peak with comparable intensity at 580 nm. The variability in concretion emission can therefore be considered negligible compared to the difference between the emissions of limestone and speleothems. The study of reference materials thus demonstrated that UV fluorescence spectroscopy can distinguish between these two materials in the context of the Chauvet Cave. The sensitivity of fluorescence to impurities embedded in the carbonate matrix, particularly organic compounds, enables distinction between limestone and speleothems although both are calcite, overcoming the limits of elemental (XRF) and crystallographic (Raman) approaches.

Theoretical Considerations on the Stratigraphic Model of the Decorated Panels and UV Fluorescence

As discussed above, noninvasive approaches generally fail to probe the stratigraphy of cave walls. Building on the ability of UV fluorescence spectroscopy to differentiate limestone and concretions, we propose a simplified model of the wall to assess the technique’s potential for probing its layered structure.

Our model for the panels comprises the limestone bedrock support, an occasional layer of pictorial matter and an overlying layer of calcium carbonate corresponding to the natural calcite veil deposited over cave walls (Figure B). The charcoal pigment absorbs but does not fluoresce, while the calcium carbonate layers fluoresce. Therefore, when UV light excites an area covered with pigments, the excitation is absorbed by the pigment layer, leaving only the overlying calcite to emit. Where pigment is absent, both UV excitation and visible emission can penetrate the translucent calcite layer, enabling detection of fluorescence from both speleothem and limestone. This model thus explains fluorescence variations induced by the presence of nonfluorescent pigments.

A priori, fluorescence from a multilayered system containing both colored and fluorescent components does not permit easy separation of individual layer contributions, hence the need for simplifying assumptions. We model the radiation as two counterpropagating diffuse fluxes oriented normal to the layers. As pigment is never perfectly covering even at the millimeter scale, the area probed by the beam contains pigment-covered and pigment-free regions, all of which are overlain by the concretion layer. In this context, the measured emission comprises contributions from both speleothem and limestone, provided the limestone is not obscured by pigment. The intensity emitted at a surface S under an excitation of intensity I₀ can be expressed as

$$\frac{\mathrm{I}(\lambda )}{I_0}={\int }_S(\left({\int }_0^{\mathrm{t}}{\alpha }_{\mathrm{speleothem}}(\lambda ,\mathrm{x},\mathrm{y},\mathrm{z}){\mathrm{T}}_{\mathrm{speleothem}}(\lambda ,\mathrm{x},\mathrm{y},\mathrm{z}){\mathrm{T}}_{\mathrm{speleothem}}({\lambda }_{\mathrm{exc}},\mathrm{x},\mathrm{y},\mathrm{z})\mathrm{dz}\right)+{\alpha }_{\mathrm{limestone}}(\lambda ,\mathrm{x},\mathrm{y}){\mathrm{T}}_{\mathrm{speleothem}}(\lambda ,\mathrm{t},\mathrm{x},\mathrm{y}){\mathrm{T}}_{\mathrm{speleothem}}({\lambda }_{\mathrm{exc}},\mathrm{t},\mathrm{x},\mathrm{y}){\mathrm{T}}_{\mathrm{pigment}}(\lambda ,x,y))\mathrm{dxdy}$$ 1

Where t­(x, y) denotes the local thickness of the calcite veil, α_speleothem and α_limestone the emission yields of an elementary volume of calcite and limestone under a λ_exc excitation, T_speleothem the transmission function of a thickness z of calcite and T_pigment the transmission function of the pigment layer. We consider the optical properties of the limestone substrate to be homogeneous across the panels in a given geomorphological context, thereby removing the spatial dependence of α_limestone. Given the consistent spectral shapes measured for the fluorescence of the reference speleothem (Figure S2A–C), we can decompose the concretion emission yield as the product of a spatially independent envelope α_speleothem with a wavelength-independent factor accounting for potential spatial variations in fluorophore content. The presence of a coating, if any, does not affect visibility to the naked eye, allowing the concretion layer to be treated as having uniform absorption and scattering across the visible spectrum. Likewise, since we only consider charcoal pigments with a characteristic broadband absorption in the visible region, we assume they also exhibit spectrally uniform absorption and scattering. This further assumption eliminates the dependence of T in λ. As a consequence, the spatial integrals multiplying α_speleothem and α_limestone depend only on the distribution of concretion thicknesses and fluorophore contents, i.e., on the considered surface S:

$$\begin{array}{c}\frac{I(\lambda )}{I_0}={\alpha }_{\mathrm{speleothem}}(\lambda )\mathrm{f}(\mathrm{S})+{\alpha }_{\mathrm{limestone}}(\lambda )\mathrm{g}(\mathrm{S})\left({\int }_S{\mathrm{T}}_{\mathrm{pigment}}(\mathrm{x},\mathrm{y})\mathrm{dS}\right)\\ ={\alpha }_{\mathrm{speleothem}}(\lambda ){\mathrm{C}}_{\mathrm{speleothem}}+{\alpha }_{\mathrm{limestone}}(\lambda ){\mathrm{C}}_{\mathrm{limestone}}\end{array}$$ 2

Under these assumptions, the fluorescence emission collected from a decorated panel can be decomposed as a linear combination of emissions from concretions and limestone with coefficients C_speleothem and C_limestone. The thickness of the calcium carbonate veil modifies the intensity of the coating’s emission (f) and the attenuation of the support’s emission (g). The pigment distribution over the spot area determines the T_pigment spatial integral, reflecting the degree to which pigment masks the underlying limestone fluorescence. It is therefore possible to separate the contributions of limestone and speleothem and focus on the specific emission of the calcite coating. In these conditions, UV fluorescence spectroscopy is capable of probing the stratigraphy of cave walls. However, this linear decomposition may not hold in all cases. Specifically, the assumption of wavelength-independent absorption no longer holds in the presence of colored concretions. The contribution of the substrate to the total fluorescence signal would then correspond to the limestone emission modulated by the unknown absorption of the calcite veil. Additionally, caves with different geomorphological histories may have more diverse organic fluorophores inputs throughout concretion growth. This would invalidate the use of a single fluorescence spectrum reference for the concretions, similarly preventing the decorrelation of the limestone and calcite veil emissions.

Results

The Black Drawings under the UV Light

The study of the emissions from cave wall components provides an interpretative framework for UVF imaging (Figure ): calcite concretions exhibit distinctive blue emission, whereas limestone appears gray because of the two peaks described above. Accordingly, the Horses panel reveals an inhomogeneous distribution of speleothem deposited on the decorated limestone. In particular, UVF imaging highlights major calcite encrustations, such as those in the lower right corner of the panel. Additionally, the images reveal two other sources of fluorescence. First, fossils exhibit strong fluorescence, particularly those with orange emission. Second, an orange fluorescence is observed near certain black charcoal tracings, a phenomenon that is further investigated in the Discussion section. However, most of the surface exhibits only the limestone fluorescence, occasionally obscured by the black drawings where no signal is detected. Despite the 12-bit dynamic range of the image acquisition, emission from the black drawings remains too weak, thus requiring a spectroscopic approach.

3.

Ultraviolet fluorescence from the Horses panel. (A) Visible and (B) UVF imaging. (C) Spectroscopic data visualization: each diagram represents a measurement point of UVF spectroscopy. The radii of the half-disks are proportional to the normalized magnitudes of the concretion and limestone components. Corresponding normalized spectra are plotted for three measurements, highlighting the relative attenuation of the second peak (red arrow).

The 60 spectra measured in the cave show varying intensities of the two peaks described above, reflecting varying contributions of limestone and speleothem to the total emission (Figure S4). In contrast to imaging results, spectroscopy revealed a fluorescence signal on the black drawings, though with intensities up to 20 times lower than from bare limestone. Given the linear model described above, we treated the spectroscopic data with the NNLS optimization algorithm, using reference spectra of limestone and speleothem emissions from the panel. This decomposition yields the coefficients for limestone (C_limestone) and speleothem (C_speleothem) contributions to the measured spectra (see eq ). Negligible magnitudes of fit residuals compared to total spectra were considered indications of satisfactory performances of the algorithm (Figure S5A–D). The histogram of normalized RMSE of the NNLS fits shows that most of the spectra are correctly decomposed (Figure S5E). Some measurements corresponding to orange emissions yield poorer fits with higher normalized RMSE and will be investigated in the Discussion section (Figure S5C).

Given the significant differences in fluorescence intensities, the data were normalized to enable visualization of the analysis results according to the formulas:

$${\mathrm{C}}_{\mathrm{limestone}}/({\mathrm{C}}_{\mathrm{limestone}}+{\mathrm{C}}_{\mathrm{speleothem}})$$

$${\mathrm{C}}_{\mathrm{speleothem}}/({\mathrm{C}}_{\mathrm{limestone}}+{\mathrm{C}}_{\mathrm{speleothem}})$$

These values could then be mapped to their acquisition locations (Figure S6) with each measurement point being represented by two semicircles whose radii are proportional to the normalized values of C_speleothem (blue) and C_limestone (green).

The horse in the center of the image provides a detailed case study (Figure C). With C_limestone close to zero, the emission measured from the horse’s lip is dominated by the blue peak characterizing the speleothems (spectrum 1) reflecting the masking of limestone emission by the pigment (low T_pigment spatial integral). In contrast, spectrum 3 from bare limestone exhibits the additional peak at 580 nm: limestone emission is not attenuated by any pigment layer (integrated T_pigment equals 1 and C_limestone is high compared to C_speleothem). At the horse’s jaw, spectrum 2 shows intermediate characteristics with partial pigment coverage resulting in an intermediate value of the T_pigment integral and yielding balanced fluorescence contributions from limestone and concretions. These measurements illustrate the relative increase of the speleothem contribution in charcoal-covered areas (highlighted by the red arrow in Figure C). Building on the linear model and the study of reference materials, this pattern indicates the presence of a concretion layer overlying the artworks. Despite the challenging working distance, UV fluorescence spectroscopy successfully detects the calcite coating overlying the decorated limestone, enabling noninvasive remote taphonomic studies of prehistoric artworks.

Assessing the Heterogeneity of Taphonomic Processes

Visualization of the spectral data in Figure reveals heterogeneous concretion emission intensities, demonstrating the importance of local taphonomic characterization. This spatial heterogeneity is particularly evident in the Alcove of the Lions, which experienced calcite flowstone formation resulting from water infiltration after human occupation ceased. Under UV excitation, calcite flowstones exhibit strong blue fluorescence revealing the pathways of past water flow. Eleven UVF spectra were acquired in this area to investigate calcite deposition over this panel. Results are shown in Figure , using the same normalization and visualization as for the Horses panel.

4.

Detail of the Alcove of the Lions. (A) Spectroscopic data visualization comparable to Figure : the radii of the half-disks in blue and green scale with the magnitudes of the concretion and limestone components, respectively. (B) UVF image with a yellow rectangle indicating the area of the charcoal line covered by a calcite concretion.

Detailed examination of the UVF image shows different taphonomic facies (Figure ). On the left, a large flowstone emerges from a horizontal crack crossing a horse’s head. Blue fluorescence from the charcoal-covered surface indicates that a calcium carbonate deposit has formed over the drawings. Unlike the Horses panel, the calcite crust fluorescence is strong enough for detection by UVF imaging, indicating a greater thickness of the layer or a higher concentration of fluorophores. In UVF spectroscopy, these variations in concretion thicknesses are reflected in contrasted magnitudes of the C_speleothem parameter derived from the NNLS treatment.

On the horse’s breast line, imaging and spectroscopy reveal a small region covered by a calcite concretion (highlighted in yellow on Figure B) whereas only the limestone signal was detected along the rest of the line. This heterogeneous concretion distribution across the line results from the large calcite flowstone adjacent to the line on the right. Further down the breast, we observe a clear delineation between areas with calcite deposits and those with pigment. The charcoal lines in this area remained apparently free of concretion while calcium carbonate was deposited on pigment-free areas. Notably, the lower area of the image shows that the lines delimit the calcite flow. These findings were confirmed by UVF spectroscopy: no speleothem contribution is detected from these black lines (C_speleothem equals zero).

These results demonstrate the value of combined UVF imaging and spectroscopy to characterize the concretions overlying the artworks. In the Alcove of the Lions, they are heterogeneous, not only across the panel but even within a single painted line. Local disparities in the calcium carbonate deposits suggest interactions between the charcoal pigment and the crystallization process. This contrasts with the behavior observed by Chalmin et al. for mineral coloring materials (manganese oxide, iron oxide), where no correlation between the pigment layer and the calcite crust was observed.

Discussion

Semiquantitative Relevance in Question

Since our method isolates the contribution of concretions to the fluorescence spectra, we can discuss to what extent this approach may provide a quantitative assessment of the speleothem coating thickness. Provided that this layer exhibits negligible UV absorption, C_speleothem should scale with the number of excited fluorophores and thus with the layer thickness. However, this parameter cannot be translated into an absolute thickness value, as no calibration using synthetic calcite layers was performed. To visualize the distribution of the C_speleothem values, we plot in Figure the histogram of speleothem emission intensities for the acquisitions from the Horses Panel and the Alcove of the Lions. The histogram shows two populations of C_speleothem values derived from the NNLS algorithm. The measurements from calcite flowstones that are visible to the naked eye in the Alcove of the Lions fall in the highest range of C_speleothem values, confirming the stronger contribution of concretion emission.

5.

Distribution of the speleothem contribution to the fluorescence spectra as computed by the NNLS treatment among the spectra measured on the Horses panel and the Alcove of the Lions.

The correlation between the C_speleothem values and the taphonomic processes manifested in the calcite coatings raises questions about the origin of organic fluorophores. Calcite flowstones results from water infiltration, transporting organic matter from the surface ecosystem. The origin of the fluorescent organic matter in the calcite veils, however, is less clear. In the absence of water infiltration, partial remobilization of limestone organics and microorganisms are potential sources, requiring further study. Regardless, organic fluorophores incorporated in calcite coatings cannot be assumed a priori to be identical, despite their similar emission signature. Thus, C_speleothem variations may reflect not only thickness differences but also disparities in the nature and concentration of fluorophores. Our method therefore remains semiquantitative only when calcite formed under similar conditions. This applies to most of the Horses Panel, for instance, which is not affected by calcite flowstones. However, a quantitative extension of the method implying calibration on synthetic calcite veils seems out of reach given the heterogeneity of fluorophores across the panels and the difficulty of reliably reproducing the optical properties of naturally grown calcite veils.

Heating and Orange Fluorescence: Hypothetic Archeological Interpretation

While the blue peak of the UVF spectra could be attributed to organic fluorophores, interpreting the second component of the limestone emission is more challenging. Previous work has suggested an organic origin for this large peak around 580 nm, with bitumen considered a strong candidate. ,, Based on this hypothesis, we expected this contribution to disappear upon heating to 500 °C due to thermal degradation of bituminous organic molecules. Instead, heated samples of bedrock limestone exhibit significant reduction in acid-extractable organic fluorescence and the appearance of an orange peak. Similarly, heated speleothems show not only the disappearance of the blue peak, but also the appearance of a similar orange emission.

Fluorescence spectra of heated limestone and speleothems present similar shapes but differences in intensity (Figure A–B). Understanding this requires considering light absorption and scattering by the material. Heating induces charring of organic matter, thereby darkening the material and increasing absorption of both excitation and emission light. Fluorescence from heated samples therefore reflects the combined effects of the disappearance of the blue peak, appearance of the orange peak and increase in absorption.

6.

Spectroscopy of orange emissions. (A, B) UVF spectra of the reference samples after annealing. (C) UVF spectra of tar and degradation products of charcoal. (D) UVF imaging of a detail showing the position of the acquisitions. The location on the panel and the corresponding visible image are shown on the left. (E) UVF spectra recorded from the detail showed in (D).

The sensitivity of luminescence to annealing is well established and generally explained by a model initially developed for quartz. The mechanism of luminescence enhancement by holes transferring from a reservoir to luminescent centers is supported in both synthetic and geological calcite. However, the chemical origin of these annealing-sensitive emissions is uncertain: some studies invoke Mn²⁺ or rare earth ions, others structural defects. Aromatization of organic fluorophores could also explain the observed red shift. Although the orange emission is reported in other studies about natural calcite, , it is important to bear in mind the difficulty of generalizing this behavior, which appears to depend on ionic and molecular impurities (fluorophores, activators, inhibitors), as well as crystal structure (defects, crystallites).

In situ UVF imaging revealed localized orange emissions in several areas (Figure ). Remarkably, these emissions are located immediately beneath some charcoal marks. Fluorescence spectra from these areas consistently exhibit an emission peak centered around 580 nm (Figure E). To understand this phenomenon, several potential fluorescent emissions associated with charcoal were examined.

7.

Imaging of orange emissions. Each subfigure contains a visible image and an UVF image of the same area. (A) Torch smears. (B) Example of an orange emission not situated only beneath a charcoal mark. (C) A charcoal mark exhibiting orange emission near another unaffected mark. (D) Depictions showing the most intense orange emissions.

We evaluated fluorophores derived from wood combustion by measuring the fluorescence of Pinus sylvestris tar, as this wood species was used to produce the charcoals found in the cave. Fluorophores generated by charcoal degradation were also investigated since this pigment can undergo significant oxidation over more than 30,000 years. The protocol from Deldicque et al. was adapted to oxidize reference charcoals carbonized at various temperatures. The reaction was carried out until complete degradation of the solid, with degradation products dissolved in the solution.

However, tars and charcoal oxidation products both exhibit a blue emission peak, even after incorporation into a carbonate matrix, and thus cannot explain the orange emission (Figure C). Furthermore, even at the local scale of a single drawn figure, orange fluorescence is not observed on all charcoal lines (Figure C). Given the similar preservation conditions of nearby features, this observation argues against fluorophores derived from charcoal degradation. Likewise, these fluorescent zones do not form a halo around the lines, instead appearing beneath them or, occasionally, to one side (Figure B). This implies that the orange-emitting areas do not result from gravity-driven flow or chemical diffusion of fluorophores.

Having excluded these hypotheses, and given the results from standard samples, we considered local heating of calcium carbonate as a potential explanation for the orange fluorescence observed in the cave. First, heated speleothem and bedrock limestone are the only materials exhibiting an emission peak at 580 nm (Figure ). Second, this orange emission also appears around torch smears, the traces left on the wall when reviving a torch (Figure A), although some marks that could be interpreted as such do not display the emission. UVF intensity differences between heated speleothems and heated limestone in reference samples suggest that variable calcite coating thickness on the wall may account for the observed fluorescence heterogeneity among torch smears.

In archeological terms, this hypothesis implies that certain lines in the drawings were executed using charcoals with hot, possibly burning tips. While not providing direct evidence, UVF spectroscopy offers several arguments supporting this hypothesis.

Conclusion

This work aimed to develop an analytical method to characterize the calcite coatings overlying prehistoric cave art. We showed that the sensitivity of fluorescence to impurities in the calcium carbonate matrix makes UVF spectroscopy able to differentiate speleothems from limestone, thus overcoming limitations of conventional XRF and Raman spectroscopies. We proposed a simplified model of the decorated wall with several assumptions based on the characteristics of the Chauvet cave, enabling linear decomposition of the total fluorescence signal. It was therefore possible to highlight the distinctive emission of the calcite coating and extract stratigraphic information.

A custom setup was designed to implement this technique in the challenging conditions of the Chauvet cave, achieving in situ noninvasive UVF spectroscopy at working distances up to 2.5 m. Combined fluorescence imaging and spectroscopy of charcoal drawings provided new insights into the surface condition of the artworks. On the Horses panel, this approach revealed the presence of a translucent calcite veil covering the drawings. Identifying this natural varnish layer is crucial to assess their conservation state and current stability. We also observed significant heterogeneity in the calcite concretions, especially in the Alcove of the Lions, where charcoal drawings influenced calcium carbonate deposition. This panel demonstrated the relevance of the spatially resolved characterization that our method provides for calcite coatings.

The semiquantitative nature of the method was discussed in light of the various factors likely to modify calcite fluorescence. Comparing and quantifying calcite layer thicknesses is challenging when fluorophore concentrations vary, representing the technique’s main limitation. Quantitative assessment of the fitting parameters uncertainties also represents a significant perspective for our protocol, which focuses on the examination of residuals.

Finally, we proposed a hypothetic interpretation for the orange emission observed near certain charcoal marks in the Chauvet Cave. Our results suggest that this fluorescence may result from the thermal alteration caused by hot or burning charcoals used for drawing. This hypothesis still requires confirmation by complementary approaches, such as a study of charcoal combustion states, and examination of how heating affected clays with variable iron oxide content.

Overall, these results highlight the relevance of a global approach to decorated cave walls, considering the support, the pigment, and the concretion as components of a complex system. Characterizing the speleothems is essential for a taphonomic analysis of the panels, to understand the transformation of their appearance and reconstruct the gestures of prehistoric artists. The identification of translucent calcite veils forming a natural protective layer over the artworks also contributes to assessing their vulnerability and forecasting their evolution. Our method therefore constitutes a promising tool for decision-making regarding the conservation and accessibility of prehistoric caves.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.6c05061.

• Figure S1: the Chauvet cave. (A) Location of the site. (B) Location of the studied region on a topographic map of the cave (Chauvet cave scientific team – 2019). (C) General view of the area (C. Fritz, Chauvet cave scientific team, MC, CNRS). From left to right: the Horses panel, the Alcove of the Lions, and the Reindeer panel. Figure S2: speleothem samples from the Chauvet cave. On the main sample (A), fluorescence measurements were performed at 500 μm intervals along the colored line. The color scale indicates the positions of the corresponding UV fluorescence spectra plotted in (B) and in (C) after normalization by their intensity at 450 nm. Samples used for (D) annealing and (E) acid digestion. Figure S3: X-ray diffraction of reference samples. The main peaks of calcite are indicated in red. Figure S4: complete set of fluorescence spectra measured on decorated panels. The peaks discussed in the text appear at approximately 450 and 580 nm. Figure S5: Non-Negative Least Squares results on experimental spectra. The speleothem and limestone spectral contributions are represented by light and dark gray areas, respectively. The experimental spectrum is plotted in black and the residual of NNLS in red. (A, B) Satisfactory fitting of fluorescence signal by reference spectra. (C) Spectrum of the orange emission studied in the Discussion section. (D) Illustration of the performance of the data treatment on a weak signal. (E) Histogram of normalized root mean square errors (RMSE) computed from the NNLS treatment of the experimental spectra. Figure S6: visualization of the 47 fluorescence spectra recorded on the Horses panel, including those already shown in Figure C with the same legend. The radii of the half-disks are proportional to the normalized magnitudes of the concretion (blue) and limestone (green) components (PDF)

C.F., M.J., and P.W. conceptualized and supervised the study. C.F. led the scientific team of the Chauvet Cave. Y.B., T.G., and P.W. designed the spectroscopy setup. Y.B., G.T., C.F., M.J., and P.W. discussed the methodology and investigated the application to the cave drawings. Y.B., M.J., and P.W. wrote the first draft of the manuscript, which was subsequently edited by T.G., C.F., and G.T. All authors contributed to the review of the manuscript and approved the final version.

The authors declare no competing financial interest.