Introduction

The Benedictine monastery of St. John in Müstair is a UNESCO World Heritage Site, hosting an extraordinary cycle of Carolingian wall paintings of the late eighth/early ninth century. Around 1200, the eastern wall of the church, as well as the three apses, were redecorated with a Romanesque wall painting cycle, covering the Carolingian one. In 1492 the construction of gothic vaults isolated the upper register of both painting cycles in the attic of the church. In the later epochs, the wall paintings in the church, with the exception of the attic, were covered with successive decorations and limewash, and their existence was subsequently forgotten [1]. The history of the church is marked by the occurrence of several fires, three of which are documented. Two of them happened before the construction of the gothic vault, in first half of the tenth century and the second half of eleventh century (between 1079 and 1087) respectively; a third fire happened in the attic in 1499 [1].

In 1894 the art historians Joseph Zemp and Robert Durrer rediscovered the upper register of the wall paintings in the attic of the church and, shortly after, documented them with watercolour [2] (Fig. 1a). In 1908–1909, the upper register of both the Carolingian and Romanesque cycles were detached by the company Christian Schmidt & Söhne from Zurich, generating 26 large fragments which since then are part of the collection of the Swiss National Museum (SNM). The detachment process for the Carolingian cycle was performed using the technique of strappo, removing the pictorial layers only. However, in this specific case, the underlying intonaco was sometimes also detached. The detached wall paintings were transferred to a textile support and mounted onto wooden strainers [3] (Fig. 1c, d, Fig. S1) (these will be referred to as “Carolingian objects”). The strappo process was not always fully successful, therefore traces of the original paint layer(s) (so-called imprints) remain in the church’s attic (Fig. 1b, Fig. S2). Six Romanesque fragments were detached from the eastern wall by removing the intonaco together with the paint layers (i.e. stacco technique). They were subsequently embedded in a rigid support made of a gypsum plaster [3] (Fig. 1e, f, Fig. S3) (these will be referred to as “Romanesque objects”). Past investigation [4, 5] allowed the characterisation of the plaster support of both Carolingian and Romanesque wall painting. It was found out that, the Carolingian support consists of lime-based plasters; a first layer, applied direct onto the stone masonry, is 1.5–3 cm thick and contains quarry sand ranging from 0–20 mm in size. On top of this, a thick limewash was applied, finally a 5–20 mm thick intonaco, containing crushed brick and sand was applied, in trowel portions, then flattened and painted [5]. The support of the Romanesque wall paintings consists of a single lime plaster layer, 2–9 mm thick, dense and well evened, which lies directly on the Carolingian surfaces in the east wall. The mixture is very rich in lime putty with a little sand and chopped hay as reinforcement [5]. Prior to the application of the Romanesque plaster, the Carolingian surface was roughened by keying in order to guarantee a good adhesion [5]. In the past, keying was a common process which, through the use of a specific tool (e.g. hammer), produced regularly spaced indentations in order to roughen a smooth surface and obtain an increased adherence between the support and new plaster layer.

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

© Eidgenössische Archiv für Denkmalpflege/ Federal Archives of Historic Monuments); b imprints in the attic, scene Mü 86; c, d corresponding Carolingian object LM-11990.1 front side and back side; e, f Romanesque object LM-11995.1, front and back side. (b–f, photos by Focus)

a Watercolour by Zemp and Durrer depicting part of the East wall before the detachment, fragment of the Romanesque cycles lie above the Carolingian wall painting (

The Carolingian objects are fragmented and unstable and have required repeated conservation and restoration treatments (including retouching, cleaning and fixing) over the last century. Their instability and very heterogeneous appearance severely limit their exhibition potential. The instability is to be linked with the condition at the moment of the detachment, as well as to the detachment it-self, which is per se a very invasive treatment for any wall painting. In addition, the very thin detached wall painting layer, transferred from a rigid wall to a flexible support developed deformation patterns typical of canvas paintings. These can occur due to the movement of the canvas caused by changes in relative humidity and various mechanical stresses. Finally, the aging of the materials, including the original materials and those used for the detachment as well as for conservation-restoration, could also engender further instability. Whichever the reason, a need for further intervention was identified by several investigations of the condition of the objects, the most recent being undertaken in 2016 and 2018. The situation is different for the Romanesque paintings which maintain a better conservation condition. Unlike the remaining Romanesque mural paintings in the church of St John’s, which have undergone repeated restorations since 1947 [6], only a few interventions are documented for the Romanesque objects in the SNM collection. These concerned the stabilisation of a few loose flakes present in four of the detached paintings (LM-11995.3, 4, 5, 6), while the original pictorial layers do not appear to have been altered.

A multidisciplinary research project involving curators, conservators, art historians, and conservation scientists was launched in 2019 with the intent to develop a conservation concept for these valuable objects [3]. The project involved comprehensive analytical investigation of the original painting materials used for the execution of the wall paintings, as well as the added materials used for detachment and subsequent restoration-conservation interventions. A comprehensive publication on this topic is planned.

Research aim

The main tasks of this project included comprehensive archival research, an extensive technical photographic documentation of all objects (orthophotos and technical images), an overall assessment of their condition, and analytical investigation. The latter included characterisations of the most representative situations through non-invasive analyses performed on the objects and on the imprints in the attic of the church of St. John. A second phase involved targeted sampling and subsequent analyses using various complimentary analytical techniques. The results have contributed to the understanding of the materials used during detachment and restoration interventions, as well as the materials of the original wall paintings. This paper focuses on the results concerning the original painting materials and techniques used for the execution of both Carolingian and Romanesque wall paintings. One of the objectives of the study is to reveal the pigment colour change, which has not been sufficiently emphasized in previous studies.

Material and methods

Description of the Swiss National Museum objects and of the attic’s imprints

The upper register of the Carolingian painting cycle on the 4 walls of the attic includes the iconographic scenes of the Ascension of Christ (eastern wall) and several scenes of the life of King David (southern, western, and northern walls) (see Table S1 and Fig. S4). The wall painting of the Ascension of Christ, because of its dimension, could not be detached as a whole. It was therefore divided in 4 parts generating originally 4 individual objects (LM-11990.1-4); in 1968 one of the objects was further separated into two objects (LM-11990.4 and LM-11990.4a). The paintings on the other walls are subdivided into individual scenes which were independently detached generating 13 objects (LM-11991.1-8 and LM-11992.1-5).

Due to the heterogeneous condition of the wall paintings and the difficulties associated with detachment, the pictorial layer removed and transferred to canvas varies from very thin to a few millimetres thick, sometimes including portions of intonaco. Very often, the objects present lacunae (Fig. 1c, Fig. S1a-c): areas where the original wall painting was already missing before detachment or zones where the detachment was unsuccessful (transfer losses) (Fig. 1b, Fig. S1a–c). The imprints usually still maintain painting traces (Fig. 1b, Fig. S2a–c). In addition, on some of the Carolingian paintings of the eastern wall, there are numerous keying holes (Fig. 1c, Fig. S1a, b). They had the function to improve the adherence of the intonaco of the Romanesque painting cycle on the Carolingian wall painting.

The Romanesque cycle was located only on the east wall. When it was rediscovered, the scenes depicted were very fragmented and only partially covered the foregoing representation of the Ascension of Christ (Fig. 1a). The scenes include themes from the Old and New Testament (Fig. S3). The detachment technique and the gypsum structure used to support the detached fragments, which was more similar to the original wall support, lead to a better conservation condition for the Romanesque objects. A limited number of unstable flakes and salt efflorescences were detected, which can potentially represent a risk for the object’s stability.

Considering the heterogeneous conditions of the Carolingian objects, three of them were chosen, namely LM-11990.1 (Fig. S1a), LM-11990.2 (Fig. S1b) and LM-11990.3 (Fig. S1c). These strappi seem less invasively restored, present a wider colour palette than other objects and are also characterised by a better conservation state. The corresponding attic’s imprints where also broadly investigated (Fig. S2a-c). Concerning the Romanesque objects, most of the investigation were done on object LM-11995.4 which presents a larger variety of original colour (Fig. S3b), few additional measurements were performed on three additional objects in order to investigate the entire palette (Fig. S3a-b).

In situ techniques

The diagnostic campaign started with non-invasive investigations on the SNM objects and on the imprints. The most representative areas were selected after careful visual inspection by the conservators, supported by a systematic study and comparison of technical images.

Technical imaging

Technical imaging was undertaken on both SNM objects and imprints (Phase One XF IQ4). For each object/scene a set of images was generated (stack): visible (reflectance VIS-R_vis), UV reflectance (UV-R_uv365), UV false colour (UVFC), UV Fluorescence (VIS-L_uv365), IR Reflectance(IR-R_ir), IR False Colour (IRfc, IR-R_ir), Visible Induced IR luminescence (VIL, IR-L_vis), and UV Induced IR luminescence (IR-L_uv365), following standards described in 2013 by [7] and further developed in [8]. For more specific technical information see Suppl. Mat 1. Areas with similar appearance in the VIS images but different spectral answers in the other images were selected for further investigation. Pigment identification by technical imaging was only carried out in the case of Egyptian blue (Fig. 5b, d).

Handheld-X-ray fluorescence spectrometry (h-XRF)

Non-invasive element detection was performed using two different handheld instruments (h-XRF): Tracer 5G (Bruker BSI Nano, Billerica, MA, USA) mounted on a tripod for the SNM objects, and Niton Xl3t (ThermoFisher Scientific, Waltham, MA, USA) for the imprints. No tripod was used in the attic. The Tracer 5 g is equipped with a graphene window, a rhodium thin window X-ray tube, and a 40 mm2 silicon drift chamber detector. It was operated under the following conditions: X-ray tube 40 kV and current at 20 µA, 3-mm spot size collimator, no filter and acquisition time 60 s. With these conditions it was possible to detect elements with atomic number Z ≥ 12 (from Mg on). The Niton Xl3t X-Ray has a Si Pin diode detector with an Ag anode. It was operated at X-Ray tube 50 kV, 40 µA max, and acquisition time 60 s, allowing the detection of chemical elements with atomic number Z ≥ 16 (from S onwards). Although two spectrometers were used, the relative spectral intensities of the elements detected in different points can be used to understand the chemical composition, at least for major compounds, of the pictorial layer(s).

Reflectance Fourier transform infrared spectroscopy (FTIR)

Non-invasive infrared measurements were performed using the portable FTIR spectrometer ALPHA II (BSI Nano, Billerica, MA, USA) equipped with a CenterGlow™ source, a RockSolid™ interferometer and a temperature controlled DTGS detector. The infrared radiation reflected from the examined surface is collected in a non-contact mode (at about 1 cm distance) by an external reflectance module with specular optics (22°/22°). The sampling area of ca. 3 mm diameter is visible through a built-in USB video camera. The reflection FTIR spectra were acquired between 8000 and 350 cm−1 using a resolution of 4 cm−1 and an average of 186 scans. The spectra are provided in pseudo-absorption units [A’ = Log(1/R); R = reflectance]. The background correction was performed collecting spectra from a flat gold mirror.

Laboratory methods

To answer some of the open questions about the pigment palette and the artistic technique, seven samples were taken on three Carolingian objects (LM-11990.1, LM-11990.2, and LM-11990.3) (Suppl. Mat. Table S2) and ten samples on three Romanesque objects (LM-11995.1, LM-11995.4, and LM-11995.6) (Suppl. Mat. Table S3). Twelve samples were embedded in epoxy resin and polished to obtain cross-sections, while 5 of them originating from the Romanesque objects were analysed with no preparation.

Optical microscope (OM)

The optical examinations of the cross-section samples (identification of sequence, thickness, and morphology of the layers) were performed with a Zeiss Axioscope 5 microscope (Carl Zeiss, Oberkochen, Germany) equipped with an Axiocam 305 color camera. Images of the cross-sections in both reflected visible light and UV radiation were collected with the software ZEN core v3.1 at different magnification (50×, 100× and 200×).

Scanning electron microscopy with energy dispersive spectroscopy (SEM–EDS)

Selected cross sections have been investigated using a Zeiss EVO MA 10 SEM (Carl Zeiss, Oberkochen, Germany) equipped with Thermo NORAN System 7 (Peltier cooled ULTRADRY detector 30mm2) for EDS (Energy Dispersive Spectroscopy) (ThermoFisher Scientific, Waltham, MA, USA). Spot analysis and maps were performed at 20 kV, and current at 100, 200, and 500 pA according to the sample, 8.5 mm working distance, in high vacuum (HV). Back scattering maps were acquired using a 5 segment HDBSD detector (High Definition Backscattered Electron Detector). Samples were carbon coated (Cressington Carbon Coater 108C with carbon rods at a pressure of 0.01 mbar). Data was processed using COMPASS & XPHASE software.

Micro-Raman spectroscopy (µ-Raman)

Cross sections were investigated using a Raman spectrometer Lab-Aramis (Horiba Jobin–Yvon, Palaiseau, France) equipped with three lasers (532, 633, and 785 nm) with laser power and duration which were adapted to individual cases and kept to low level (< 10 mW) to avoid damage to the samples and laser induced modification (especially in the case of sensitive pigments such as red ochres).

Micro-FT-IR spectroscopy (Micro-FTIR)

Three different spectrometers were used. The first one is an Excalibur Series FTS 3500GX spectrometer equipped with an UMA-500 microscope (Biorad, Hercules, California - USA) in reflectance mode, in the range 4000–650 cm−1, with a resolution of 4 cm−1, and 512 scans. The second equipment used was a Spotlight 400 FTIR microscope fitted with a single element MCT detector and coupled to a Frontier FTIR spectrometer (PerkinElmer, Switzerland). Micro-FTIR spectra were collected in Attenuated Total Reflection (ATR) mode with a germanium crystal having a 100 µm diameter, in the spectral range 4000–600 cm−1 with a spectral resolution of 4 cm−1 and by co-adding 64 scans. The third spectrometer is a Jasco FTIR 6600 with Microscope IRT 7100 and a micro-ATR 5000-D with a diameter of 100 µm. Measurements were performed as single point or mapping, or with diamond Anvil cell (Cassegrain X16), in the range 4000–600 cm−1, with a resolution of 8 or 4 cm−1, and 1024 or 128 scans respectively.

Micro-XRF (µ-XRF)

Some cross sections were analysed with a lab based micro-XRF, Eagle III XXL (Edax, Mahwah, NJ, USA) equipped with a Rh tube and a Si-(Li-) semiconductor. The analyses were performed at 20 kV and 100 µA, point measurements (50 µm measurement spot) of 200 s (600 s respectively in several cases to obtain a better signal-to-noise ratio). To evaluate the results the program Vision32 was used.

Results

Characterisation of the Carolingian pigment’s palette

The colour palette was widely characterised (Figs. S1 and S2). The points/areas of investigation were chosen according to the different colour tones and optical behaviour. However, it must be pointed out that the visual appearance is quite heterogeneous depending on object’s condition, presence of surface coating, extent of the retouching and amount of the original material. In addition, the colours of the Carolingian paintings/imprints were highly modified by fires which occurred in the tenth, eleventh and fifteenth century [1, 9].

The following colour shades were detected on the Carolingian objects and on the imprints (Table 1):

  • White: is very common and is mainly used for the garments of the various figures, as highlight for floral decorations, faces, and architectural structures surrounding the figures, as well as in alternation with yellow-orange and red to create the horizontal and vertical frames delimiting the single scenes. In addition, in the Carolingian painting layers, the white also appears in the form of later limewashes and decorative beads added in the Romanesque period directly onto the painted surface (Fig. S1b and c). Most of the white colours analysed showed the presence of Ca (Ca- and/or Mg-rich carbonates were detected by Raman, although Mg was not clearly detected by h-XRF) indicating the use of lime and/or bianco di San Giovanni (white 1). Bianco di San Giovanni is a pigment consisting of calcium carbonate and calcium hydroxide, first described by Cennino Cennini in the early fifteenth century [10]. In addition, Pb-rich white areas (white 2) were identified in some places on object LM-11990.1, on the decorative beads of object LM-11990.2, which are Romanesque additions (Fig. S1 a and b), and on the northern wall (scene Mü-14, nose of one of the character). Due to the lack of stratigraphy, it is not possible to discern if the Pb is present as lead white in the uppermost white layer or if it arises from the underneath layers. In previous investigations, the white highlights of the Carolingian cycle were always found to consist of a Ca-based pigment [11] with no lead white admixtures.

  • Flesh tones: are generally made with ochres/Fe-oxides mixed with Pb-pigments, while Ca can be related to either a lime binder or bianco di San Giovanni. Traces of As were also detected in some cases (object LM-11990.1, scene Mü-85 and -86) (Table 1). Local darkening of the flesh tones is sometimes observed. This could be correlated to deterioration of Pb-based pigments.

  • Yellow: was used mostly for the execution of the central band of the decorative frames delimiting the single scenes and was detected on one object from the East wall (Figs. 1c and S1) (yellow 1, Table 1). A sample collected on a vertical decorative band shows a very thin layer directly lying on the transfer adhesive (sample LM-11990.2_str_02, Figs. S1a, S5). This latter was used to attach the removed wall painting to the canvas. In this case it is most likely that the detachment process was incomplete and only a fraction of the paint layer was removed. µ-XRF analysis shows Fe as a major element together with Ca, Si and K, suggesting the use of an ochre mixed with lime. A second yellow (yellow 2, Table 1) was observed as small residues within the black decorative bands on the scenes of Mü-1 (southern wall), Mü-11 (western wall), and Mü-14 and -18 (northern wall). In all cases, h-XRF analysis shows Pb as main element, either alone or in association with low amounts of Fe and Ca. No further investigation were however carried out.

  • Brown–red: several shades ranging from dark orange to brown and dark red are visible on the objects and the imprints, especially for horizontal and vertical decorative bands, figure’s halos, hair, facial details and elements of dresses, masks, clouds, and architectural features. The analysis revealed two main pigment mixtures, a first one containing Ca, Fe and Pb in very different proportions (brown–red 1, Table 1, Fig. S7 black dots and spectra), and a second one containing additionally As in different amounts (brown–red 2, Table 1). The high variety of these two mixtures accounts for the large variability of the observed shades. The brown–red 2 was detected on two objects and on their corresponding imprints (scene Mü-85, 86, 87), as well as in scenes Mü-12 (western wall), Mü-14, Mü-18 (northern wall), and Mü-7 (southern wall). A sample of brown–red 2 was collected from the halo of the second apostle from the left depicted on object LM-11990.1 (sample LM-11990.1_str_01, Figs. S1a, S6a-b). It shows a thin red–orange paint layer (20 µm thick) applied directly on the intonaco (Fig.S6c). Haematite and calcite were identified by µ-Raman (Fig. S6d). No organic material was identified by µ-FTIR. Although traces of Pb and As were detected by non-invasive analyses, no related pigments were identified in the sample by Raman spectroscopy. SEM–EDS analysis confirmed the presence of traces of As, together with Fe, Mg, K, Ca and Ti in the pictorial layer (Fig. 2c–h), as well as Si and Al (not shown in the figure), but no Pb. Ti is localised in specific grains only (Fig. 2h), while As is either localised in specific grains associated with Mg (whose origin is unclear) or dispersed in the painting layer (Fig. 2f, g). Furthermore, the presence of Ca and Mg in this cross section (Fig. 2c and f) can be associated with the Mg-lime of the intonaco, as found in other previous studies [12, 13].

  • Bright red: these tones are present as isolated patches inside blackened colour fields on the eastern wall in correspondence of the horizontal and vertical decorative frame (objects LM-11990.1-2, scene Mü-86, eastern wall) as well as in the clouds (object LM-11990.2, scene Mü-86), and in the crown of the personification of the Sun (object LM-11990.2) (Fig. S1a-b); it was also detected on the scene Mü-10 (western wall). Two samples of bright red were collected. The first one was taken from one of the horizontal decorative frames (sample LM-11990.2_str_04, Fig. S1b, Fig. 3a, b). The cross section shows a single pictorial layer (approximately 50 µm thick), lying directly on the intonaco, which contains finely dispersed orange and a few yellow grains (Fig. 3c). Raman analyses revealed a mixture of lead-based minerals, namely minium, massicot, litharge, lead carbonate. They are mixed with a Ca- and Mg-rich carbonate, either dolomite or magnesite. The discrimination of these two minerals is not possible, since only the major band at ca. 1097 cm−1 common to the two minerals is observed in the spectra, and rare calcite (Fig. 3d). The second sample was taken from a painted cloud where three partially overlapping layers were applied, which today appear black (probably an altered pigment), bright red and red (sample LM-11990.2_str_06, Figs. 4a-b, S7b). The in-situ h-XRF clearly show that the bright red (red dot in Fig. S7b) is mainly composed of Pb-containing minerals (Fig. S7d). The stratigraphy shows that the depiction of a cloud was created with at least three superimposed layers (2 to 4) (Fig. 4c). Layer 2 is made of calcite (lime) containing a few fine grains of Pb-based pigment (since massicot and litharge present a main peak at about 145 cm−1, the discrimination between these two minerals is impossible) (Fig. 4e) and rare lead white (Fig. 4d). Layer 3 is a thick layer (approx. 40 µm) with minium, massicot and litharge mixed with lead carbonate, Ca- and Mg-rich carbonate and rare lead white (Fig. 4f). Layer 4 is a thin layer (5–10 µm) with a more intense orange-red colour made with massicot and litharge mixed with lead carbonate, a Ca- and Mg-rich carbonate and calcite (Fig. 4g). Because of the use of Pb-based pigments, it is assumed that the painting was applied a secco, therefore the calcite and the Mg-rich carbonate might not stem from a lime binder but from the pigment bianco di San Giovanni, added to lighten the tone of the paint.

  • Blue: two blues were found. Egyptian blue (light blue 1) was easily identified by the strong VIL signal in the technical images [14] (Fig. 5). Its presence was further confirmed by the presence of high Cu counts associated with Ca. A second type of light blue was identified as lapis lazuli by p-FTIR on only one object (LM-11990.2) (Fig. 6a, b). Specifically, the IR spectrum collected (Fig. 6c) shows the marker band at ca. 2345 cm−1 of the carbon dioxide encapsulated in natural lapis lazuli [15, 16]; probably of Afghan origin [17]. However, further investigations would be needed to confirm this origin [18]. The additional bands of CH and C = O groups, instead, revealed the presence of beeswax [19] used during the restoration of the object. A sample collected from the angel’s dress (sample LM-11990.2_str_05, Fig. S1b, Fig. 6d) shows a very thin and fragmentary paint layer with grains of lapis lazuli bound with lime (Fig. 6e). This is consistent with the general rarity of this pigment in the eighth and ninth centuries. Until now, lapis lazuli from this period has been only found on wall paintings in San Saba in Rome [20] and, most recently, at S. Benedict church in Mals/Malles [21]. The use of lapis lazuli is also evoked for the wall painting of the tower of Torba in Lombardy, although no analytical evidence have been published [22, 23].

  • Purple to dark blue: purple/blue areas were found on the objects LM-11990.2 and LM-11990.3 and on the relative imprints on the eastern wall (scenes Mü-85, -86 and -87). These colour tones were used for backgrounds, mandorlas, and halos. Two different compositions were identified. The first one is found in the mandorla of the Christ (objects LM-11990.2-3, scene Mü-85) (Fig. S1b-c), and contains Ca and Cu as main elements, associated with traces of Fe and Pb (purple to dark blue 1, Table 1). It presents a bright red appearance in IRfc and a strong VIL signal indicating the use of Egyptian blue (Fig. 5a, b). In the past, a sample from this layer was investigated allowing the identification of a lake (possibly madder) mixed with Egyptian blue [24]. The second shade (purple to dark blue 2, Table 1) is observed in the background of the Ascension of Christ in the middle of object LM-11990.3 (Fig. S1c) and in the corresponding imprint (scene Mü-87) (Fig. S2b), in some of the halos depicted in the scenes Mu-86 and -87, and in the hair of one of the angels in scene Mü-85. This colour appears bright red in IRfc (Fig. 7c) and presents no VIL signal (Fig. 7d). The non-invasive chemical analysis (carried out on the detached painting only) shows different proportions of Ca, Pb, Fe and Cu. A sample was taken for further investigation (sample LM-11990.3_str_07, Fig. S1c, Fig. 7a, b and S8). At high magnification five sub-layers (2a-e) (Fig. S8a) overlying the transfer adhesive (layer 1, Figs. 7e and S8a) are observed. There are no apparent discontinuities between the paint layers. Raman analysis was able to reveal the presence of calcite, litharge/massicot and possibly plattnerite (Fig. 7f). SEM–EDS investigation showed that sub-layer 2e is mainly composed of Ca, Mg and lesser Pb, the latter being dispersed both in the matrix and in the grains. The sub-layer 2d shows high amount of Pb, with some Ca. The sub-layer 2c is composed of Fe, associated with Mg, Si and Al (these latter are not shown in Fig. S8). On top two very thin and discontinuous sub-layers are present, one enriched in Cu (2b), and the other containing only Ca (layer 2a). Since the sample comes from an area where a painted cloud is present below the upper purple layer, the lower layers identified in this sample can be attributed to the cloud, while the upper layers (2a and b) arise from the purple shade, which was painted over part of the cloud. Indeed, a past investigation performed on a sample of purple to dark blue 2 collected from the imprint (Mü-85) and belonging to the same background area, but far from the cloud, showed the presence of only two layers on the intonaco,  a lower one containing a lake covered by a layer consisting of tenorite [25].

  • Grey: a first type (grey 1, Table 1) has been identified in objects LM-11990.1-2, used in the shadows over flesh tone, in garments, in some architectures, as first layer in the clouds (Fig. S7a-b, blue dots) and in the corresponding imprints (Mü-85, -86, and -87, eastern wall), as well as in scenes Mü-12 (western wall) and Mü-1 (southern wall). This grey contains Ca and Pb (Fig. S7, blue dots and spectra). In a sample of this grey, used to paint the roof of a temple (sample LM-11990.1_str_03, Fig. S1a, Fig. S9a, b, c), plattnerite was identified as a probable alteration product of Pb-based pigments [26, 27] (Fig.S9d). In addition, a mixture of litharge and massicot with calcite and a Ca- and Mg-rich carbonate was detected (Fig. S9e). The presence of oxalates, detected by Raman (Fig. S9e) and µ-FTIR, could be correlated to the oxidation of an organic paint binder. A second grey tone was identified only in the attic (Mü-85—eastern wall and Mü-8—southern wall) (grey 2, Table 1). It was used to represent shadows on the flesh tone of the mask at the intersection of the decorative bands, for the halo of an angel and for decorative patterns in the halo of Christ. This grey is probably made of a mixture of Egyptian blue (strong VIL signal) with ochre and/or lead-based pigments and lime (Ca, Cu, Fe, and Pb detected as main elements).

  • Black: two types of black were used. One identified in the attic (Mü-85) on a sinopia observed on the arriccio, in which only Ca (from lime) was detected by XRF, suggesting the presence of a carbon-based black (black 1, Table 1). The second type (black 2, Table 1), widespread on both the SNM objects and on the imprints on several walls, was used to create the horizontal and vertical bands, the leaves within the festoons of the decorative frame, the clouds (associated with the bright red areas) and the buildings’ roofs in object LM-11990.1. This colour is mainly composed of Pb (Fig. S7, magenta dots and spectra). Although no specific sample was taken, it most probably consists of degradation products of Pb-based pigments.

Table 1 Summary of the main colour shades identified on different Carolingian objects
Full size table
Fig. 2
Fig. 2The alternative text for this image may have been generated using AI.
Full size image

a Sample LM-11990.1_str_01, brown–red 2, micro photo in Vis-light where the pictorial layer (in red) lies on the intonaco (white layer); b BSE image (×1500 magnification) taken at 20 keV and 500 pA,; c–h EDS maps with distribution of Kα intensity for Ca, Fe, K, Mg, As, and Ti. For the location of the sample see Figs. S1a and S6a–c

Fig. 3
Fig. 3The alternative text for this image may have been generated using AI.
Full size image

a-b Sample LM-11990.2_str_04, bright red, location on the object; c micro photo in Vis-light where three layers are visible: 3- pictorial layer, 2-intonaco, 1-transfer adhesive; d Raman spectra of layer 3. For the general location of see Figs. S1b

Fig. 4
Fig. 4The alternative text for this image may have been generated using AI.
Full size image

a, b Sample LM-11990.2_str_06, bright red, location on the object; c micro photos in Vis-light where 3 layers are visible, layer 1, not visible in this picture, is the transfer adhesive; d FTIR spectrum of layer 2. e–g Raman spectra of layer 2 (e), layer 3 (d), and layer 4 (g). For the general location of see Fig. S1b. Evidence of an organic material (either a drying oil or beeswax) were found in L2, This can be attributed to a coating applied to the surface of the strappo by the restorers of the SNM. Layer 2 corresponds to the grey areas indicated in Fig. S7a-b with purple dots, while layers 3 and 4 correspond to the areas with red and magenta dots

Fig. 5
Fig. 5The alternative text for this image may have been generated using AI.
Full size image

a, b Details of object LM-11990.2 in visible light and Visible Induced Luminescence—VIL; c, d imprints of scene Mü-17 in visible and VIL. Photos ART Imaging

Fig. 6
Fig. 6The alternative text for this image may have been generated using AI.
Full size image

a, b Detail of object LM-11990.2, area presenting light blue 2, with location of Str_05, c p-FTIR analysis; d micro photo in Vis-light of cross section LM11990.2_str_05 where 2 layers are visible: 2- lapis lazuli and lime, 1-transfer adhesive; e Raman spectrum of layer 2. For the general location of the sample see Fig. S1b

Fig. 7
Fig. 7The alternative text for this image may have been generated using AI.
Full size image

a, b Sample LM-11990.3_str_07, dark blue, location on the object; c IR false colour image; d VIL image; e micro photo in Vis-light of the stratigraphy; f Raman spectra collected in layer 2, where massicot or litharge, plattnerite and calcite are detected. The white rectangle in “e” corresponds to the location of Fig S8a. For the general location of see Fig. S1c

Characterisation of the Romanesque pigment’s palette

As the Romanesque wall paintings suffered at least one fire that caused their discolouration, the original colours (green, purple, red–orange, yellow and white) can only be observed in limited areas on two objects (LM-11995.3-4) (Fig. S3b-c). These areas were protected from the fire by debris lying on the gothic vaults. The other areas/objects show only shades of brown-beige and dark blue.

The pictorial layers were investigated on cross sections. Usually the paint layers are 20–30 μm thin, while the thicker ones seem to be obtained by superimposing several layers. The results can be summarised as follows:

  • Purple: (sample LM-11995.4_str_14, Fig. S3c, Fig. 8a, b) found on a garment. It consists of three overlapping layers in which haematite and carbon black are mixed in varying proportions (Fig. 8c). The innermost (layer 2, Fig. 8c), in contact with the intonaco (or in some cases, with a limewash applied to the intonaco), is a pink/light red layer containing round red haematite grains and black carbon grains in a calcite matrix (Fig. 8d). Above this, there is a homogeneous layer of intense orange-red colour, approximately 20–40 μm thick, with fine black grains of black carbon and a few red grains of haematite in a calcite matrix (layer 3 in Fig. 8c, and Fig. 8e). On the surface, there is a black layer (layer 4, Fig. 8c), about 20 μm thick, composed of fine black grains of black carbon and a few red grains (Fig. 8e). The lack of a well-defined interface between the layers suggests that they were probably applied using a fresco technique.

  • Red (sample LM-11995.4_str_11, Fig. S3c, Fig. S10b,d): it is a thin and homogeneous layer consisting of haematite (probably an anatase rich ochre) with calcite and possibly dolomite (Fig. S10d-e). It was probably applied by a fresco technique. Gypsum was also identified; it could arise from the gypsum plaster used to embed the stacco or it could be related to the salt efflorescences visible on this object.

  • Yellow (sample LM-11995.4_str_12, Fig. S3c, Fig. S10c,f): it is obtained by a single thin layer of paint composed mainly of fine-grained yellow mineral and a few large orange, black and yellow grains. Spectroscopically only calcite was detected, but the high Fe content of this layer suggests the presence of ochre (goethite) (Fig. S10g). This colour appears to have been applied by a fresco technique.

  • Green (sample LM-11995.4_13_Str, Fig. S3c, Fig. S11a, b): at high magnification three unresolved sub-layers are visible (Figs. 9a and S11c): an underlying greyish layer (2a) with abundant black angular grains in a whitish matrix, an intermediate greenish-blueish layer (2b) with some black and rare blue grains in a greenish matrix, and the top layer (2c) with a greyish-greenish appearance and no discernible grains. The colour of this layer tends to veer towards the blue in some areas. Sometimes the three layers mix together, sometimes one predominates. Green earth, carbon black and calcite were identified by Raman (Fig. S11e). A proteinaceous material was detected by ATR-FTIR (Fig. S11d), which could be a residue of the glue used for the detachment, or a paint binder. However, the latter hypothesis seems unlikely due to the apparent a fresco application of the layers. SEM-EDS analysis shows that layer 2b is dominated by Fe, Si, K, Al, and Ca (Fig. 9c-f, Al and K not shown) confirming the use of green earth mixed with calcite, while in the sub-layer 2a Ca predominates (Fig. 9c) indicating the use of lime mixed with black carbon. Finally, traces of Cu were found in localised areas on the top of the sample (Fig. 9b and g, pt1). However, the nature of the Cu-rich layer remains undetermined. The partially blueish colour of the top layer could indicate that azurite was used.

  • Dark blue: this tone was identified only in limited areas on two objects (LM-11995.1, Fig. S3a and LM-11995.2, not shown). It appears pinkish-red in the IR false colour (Fig. S12c). A sample (LM-11995.1_str_10, Fig. S3a, Fig. S12a-b) shows a thin layer of paint applied on a limewash layer, most likely when it was still wet (Fig. S12d). At high magnification two sub-layers are visible (Fig. S12d). The upper one (layer 3b) is rather discontinuous and is enriched in Cu, often localised in patches, associated with Ca (Fig. S12f,i); the lower layer (layer 3a) contains mainly Ca (Fig. S12f). The fact that the layer 3a has a purple coloration but contains only Ca (lime) could indicate the use of an organic colorant. However, no clear spectrum has been acquired. In addition to calcite, rare Ca-oxalates were detected, suggesting the alteration of an organic compound (Fig. S12j).

  • Blue (sample 17_PL, Fig. S13a-b): a grain of the blue used for the lamb’s fur in object LM-11995.3 (Fig. S3b) was analysed. It contains lapis lazuli and calcite with a very small amount of an organic substance whose characterisation is unclear (Fig. S13c-e). Oxalate (weddellite) was also detected (Fig. S13c-d).

Fig. 8
Fig. 8The alternative text for this image may have been generated using AI.
Full size image

a, b Sample LM-11995.4_str_14, purple, location in the object, c micro photo in Vis-light of cross section where 4 layers are visible: 2–3-4-pictorial layers, 1-intonaco; d, e Raman spectra of layers 2, 3 and 4. For the general location of see Fig. S3c

Fig. 9
Fig. 9The alternative text for this image may have been generated using AI.
Full size image

a Sample LM-11995.4_str_13, micro photo in Vis-light; b BSE image (×600 magnification) taken at 20 keV and 500 pA; c–f EDS maps with distribution of Kα intensity for Ca, Fe, Mg, and Si, g EDS-spectrum of pt.1. The dashed lines in “a” indicate the interfaces of the sub-layers. For the location of the sample, see Figs. S3c and S11a-b

Four additional samples were collected on object LM-11995.4, in the zone characterised by a chromatic alteration due to fire (“PL samples”, Table S3). Thanks to the survival of parts of the objects that were not altered by the fire, it is know that the present reddish-brown shades correspond to former yellow, red, green, and purple shades respectively (Fig. S3c). These samples were analysed in order to determine the mineralogical changes caused by the fire. They show a clear modification only in the case of the green colour, where haematite was identified as the fire deterioration product of the green earth.

Discussion (new findings on Carolingian and Romanesque technique)

Regarding the Carolingian wall paintings, some of the results are in agreement with previous investigations carried out on samples from the imprints in the attic [4, 24] or in situ [21]. For example, altered Pb-pigments, red and yellow Fe-rich pigments, carbon black, and Egyptian blue were already found by [4] while lapis lazuli was only mentioned in an unpublished report by [24]. The presence of these pigments was confirmed by [21]. In agreement with the previous studies, no green pigments were found. The presence of madder, which has been found in samples by [4, 24] and hypothesised based on in-situ analyses by [21], was not confirmed spectroscopically, but, its use is likely, on the basis of the present investigation.

Compared to the previous studies, the present results provide additional spectroscopic evidence that several Pb-based minerals such as minium, massicot and litharge are present in different mixtures in the bright red, grey and black areas. They also add new information about their use. In the case of the Pb-rich grey and black areas, their present colours can be attributed to a degradation of lead pigments. The presence of plattnerite as a degradation product is suggested by the spectroscopic data. Grains of minium, litharge and massicot have been detected within the grey/black layers. It is likely that minium and massicot were intentionally used as pigments by the artists. However, the use of litharge as a pigment is doubtful because only small amounts have been found, always mixed with minimum, massicot, or both; in addition, the use of litharge as a pigment is rarely attested ([28] and references therein). Pb-carbonate was spectroscopically identified in both samples of bright red, while the presence of lead white (characterised by the presence of hydrocerussite) is rather rare. Three hypotheses could be formulated to explain the sporadic presence of Pb-carbonate and litharge. Firstly, both Pb-carbonate and litharge could be interpreted as a by-product of the production of minium by roasting of hydrocerrussite ([28] and references therein). This hypothesis would be valid for the bright red shade. A second hypothesis is that litharge is present as an impurity from the production of massicot which was used as pigment to obtain a yellowish shade; this would be the case of grey 1. Thirdly, Pb-carbonate, litharge and massicot could be degradation products of lead white formed as a result of the three fires ([28 and references therein) that occurred in the church over the course of its history. This third hypothesis seems improbable. If lead white had been used in the painting of the Carolingian cycle, given the size of the church and the spatial distribution of past invasive and non-invasive analytical campaigns, the detection of this pigment in some of the areas less affected by the heat of the fires would be expected. To the contrary, lead white was seldom detected.

Another significant finding concerns the study of the As-containing brown–red. For this colour, [21] proposed that As originated from a surface layer of orpiment that has now disappeared. The authors based their hypothesis mainly on the fact that As was detected exclusively in correspondence with the halos of several figures which now appear red, but which could be expected to have originally had a yellow/gold colour. While this cannot be completely ruled out, a new hypothesis has emerged from the present work. Firstly, evidence has been found that this As-containing shade is present in several figurative details and not only in the halos. Furthermore, if an As-rich mineral was present, its detection by Raman would have been highly probable since both realgar and orpiment have a relatively high Raman scattering cross section. It must, therefore, be inferred that the As in the painting layer is dispersed within the ochre rich matrix. Arsenic is often detected in amounts ranging from traces to minor or major elements (up to about 10%) in ochres from several sites, some of which were used in the past for pigment production [29, 30]. The detection of As by h-XRF on the Müstair wall paintings might therefore indicate the use of a specific As-rich raw material, and would merit further investigation.

The present project has confirmed and extended the awareness that today’s chromatic aspect of the Carolingian wall paintings is very different from the original one. Analytical evidence and iconographic interpretation suggest that the black areas were originally either red–orange or yellow. The degradation of the Pb-based pigments preceded the detachment, as evidenced by the documentation made before the intervention (Fig. 1a). This fact was already noted by [5]. The observed grey shades are the results of deterioration as well. While grey 1 might have been yellow (altered massicot), grey 2 could have been greenish to blueish (compatible with the presence of Egyptian blue, sometimes mixed with yellow ochre). Indeed this latter colour is found in areas where, from an iconographic point of view, no grey would be expected (halo of an Angel, halo of Christ, and shading of a face). Furthermore, the observed grey sometimes veers slightly into a turquoise colour, for example in the halo of the angel. As no stratigraphic samples were taken, the mineralogical composition cannot be assessed and the original appearance cannot be inferred at present. The appearance of the areas with the purple to dark blue 2 colour has also profoundly changed. The analysis suggests the use of an organic purple dye and a Cu-based pigment. Iconographically, this background, which frames the Ascension of Christ and imposes itself over the sky and the clouds, could have been either purple or blue. Therefore, as Egyptian blue can be excluded because of the absence of VIL, azurite is the most probable candidate for the Cu-pigment. In this case, the present very dark appearance could be due to the modification of azurite into tenorite.

New insights have also been gained regarding the execution technique. First, the study of the cross sections showed that most of the samples present a single pictorial layer, and probably only a portion of it. This is a consequence of the strappo technique. When removing the paint layers with this method, the detachment often takes place not between, but within layers. Only three samples showed a thick fragment of the intonaco. While the layer of red ochre (brown red 2) in sample LM-11990.1_str_01 (Fig. 2, Fig. S6) seems to have been applied on a still wet intonaco, the layer of minium in sample LM-11990.2_str_04 (bright red) appears to have been applied on a dry surface. This is suggested by the microscopic observation of a very sharp interface between the intonaco and the pictorial layer (Fig. 3c), but could not be fully corroborated by the detection of a specific organic binder. The latter was prevented or made difficult by the possible presence of restoration-conservation materials which cannot be ruled out in the case of the SNM objects. Only in the case of the grey 1 in sample LM-11990.1_str_03 (Fig. S9), the detection of oxalate might indicate the former presence of a binder. In addition, Minium is often found on medieval wall paintings, usually associated with organic binders [31], since these would mitigate deterioration risks. Finally, calcite and a Ca- and Mg-rich carbonate have been detected in conjunction with Pb-based pigments. Since the use of a lime-binder would increase the risk of alteration of Pb-based pigments, the presence of calcite and a Ca- and Mg-rich carbonate within the paint could be due to the use of a calcium-based pigment like bianco di San Giovanni to enhance the layer's brightness. However, since bianco di San Giovanni contains calcium hydroxide [32], it could be argued that it could have a binding role. Unfortunately, distinguishing between the use of calcite, or Ca- and Mg-rich carbonate as pigment (bianco di San Giovanni) or as a lime binder is challenging, making it impossible to rule out the latter possibility.

Some additional macro-observations regarding the painting procedures and the construction of the different scenes have been made in the course of the present project. On the east wall different blue backgrounds are observed. The different hues were probably used to enhance specific scenes or representations. On the two sides of the scene depicting the Ascension of Christ, a light blue background made of Egyptian blue was used. In the centre, where the ascending Christ is depicted, the background consisted of a purple/blue tone probably made with a red organic dye, on top of which was applied a glaze made of a copper-based pigment, probably azurite. In correspondence with the mandorla of Christ another shade of blue was used, obtained using an organic dye (probably of a red colour) mixed with Egyptian blue.

Finally, the study of three of the clouds present on the east wall (Figs. 4, 7, S7 and S8) highlighted a complex execution technique. First, the cloud was outlined using Egyptian blue (Fig. 7d); then the outline was filled with a thin paint layer containing massicot, litharge, lead white and lime or bianco di San Giovanni (Fig. S7 blue dots, corresponding to layer 2 in sample LM-11990.2_06_str in Fig. 4c, and layer 2e sample LM-11990.3_07_str in Fig. S8a). Despite the fact that today it appears grey, originally it must have been a shade of yellow. On this base tone, two thin layers where then applied. The first one contains minium, massicot, litharge and lead white, mixed with calcite and a Ca- and Mg-rich carbonate, while the second only contains massicot, litharge and rare lead white mixed with a Ca-and Mg-rich carbonate (Fig. S7 red and magenta dots, corresponding to layers 3 and 4 in Fig. 4c, and layer 2d in Fig. S8a). Although minium-was not detected in sample LM-11990.3_str-07, its use is very likely and the lack of detection must be due to its degradation, as the layer which we would expect to be bright red to orange appears partially or fully darkened (Fig. S7a,b magenta dots). The last layer, appearing brown red, is composed of ochre (Fig. S7a,b black dots, corresponding to layer 2c in Fig. S8a). The colour of this last layer could be the original one or could have been modified by the fire (from yellow to red ochre). Based on these observations, a sequence of three shades within the clouds can be reconstructed. From top to bottom, these would have been pale yellow, yellow/orange, and red. This resembles the appearance of cumulus clouds in the sunset, which might be the effect intended by the artist. The complexity of execution of this comparatively simple element of the painting shows that the Carolingian cycle must once have had a much more sophisticated appearance than it does today.

The Romanesque painting technique differs substantially form the Carolingian one. In the Romanesque cycle the colour was obtained by applying either a single layer 20 to 30 μm thick (Figs. S10), or by overlapping several layers (Figs. 8, 9, S11, and S12). The first layer is applied directly on the intonaco or on a limewash. The identified pigments are mainly ochres (haematite, while goethite rich ochres were not identified but their use can be assumed) and green earths. Lapis lazuli and two different unspecified Cu-based pigments were also used. Most of the samples suggest an application a fresco. In the case of the non-identified Cu-based pigments present in the dark blue and green areas, they could have been applied with an organic paint binder. The results of the analyses on the green background (sample LM-11995.4_str_13) concord with past observations made in the church, where a ground layer made of carbon black and lime (the so called veneda) was observed underneath both green and blue colours. This layer seems to have been used in order to obtain a more intense colour effect [11, 33]. In addition, glazes executed with a copper pigment have previously been observed in the scene The feast of Herod (middle apse) [11], in the south apse (unspecified Cu-based pigments, [34]) and the north apse (a mixture of malachite with black carbon and rare lapis lazuli [35]). The pigments used for the dark-blue ground analysed in this research (sample LM-11995.1_str_10) could not be clearly identified, but the use of an organic dye associated with a glaze of a Cu-rich pigment is likely. Its chemical composition and texture differ from previously investigated samples from the Romanesque paintings in the three apses, which showed the use of veneda and lapis lazuli [11]. This kind of dark blue ground has previously been only found in a dislocated fragment from the attic [36] which could not be clearly attributed specific depiction in the Romanesque cycle. The presence of a possible dye therefore represents a hitherto unknown element of the Romanesque paintings.

Conclusion

The results of the present study confirm, on the one side, the data obtained from previous investigations and, on the other side, opened up interesting new perspectives on the painting materials and techniques of these unique artworks.

Concerning the Carolingian painting cycles, this study has shown an extensive use of a complex mixture of Pb-based minerals in varying amounts. Massicot is always found associated with smaller amounts of litharge. Minium on the other hand was associated with smaller amounts of lead white, litharge and massicot. The latter three minerals could be residues from the production process of minium. If massicot was added to alter the tone of minium, then the litharge and lead white could alternatively be attributed to the production of massicot, which, just as for minium, consists in the roasting of lead white. The small amounts of lead white and litharge detected make it unlikely that they were used as a primary colour or intentionally added to the mixture. Therefore, the areas painted with lead pigments (now mostly blackened) would have been executed with massicot, minium, and varying mixtures of the two. Bianco di San Giovanni appears to have been added to lighten the colours.

In the analysed samples where the intonaco was preserved, the paint layers were applied to the intonaco either in a fresco technique (in the case of ochre) or a secco (in the case of Pb-based pigments). The use of Pb-based pigments on wall paintings is counterintuitive, given their susceptibility to alterations such as blackening in alkaline environments. However, in the objects from Müstair, calcite and Ca- and Mg-rich carbonate were present in all layers containing Pb-based pigments, suggesting possible functions as a pigment (bianco di San Giovanni) or as binder. Additionally, detection of oxalates suggests the presence of an organic binder within Pb-rich layers, even if a strong contribution from old conservation and restoration products must considered. A protein-based substance such as egg white or casein would provide a strong binder in conjunction with the calcium hydroxide. However, the analytical methods used do not allow us to definitively discern between these possible functions of the carbonate component.

Another artificial pigment widely used on the Carolingian paintings is Egyptian blue. This pigment, like the Pb-based ones, was used in a variety of mixtures to give different hues and colours. In the present study, it was found mixed with an organic dye, with ochres and with lead-based pigments. The VIL images show a widespread application for outlines, backgrounds, shadings and details, both pure and as admixture.

Lapis lazuli has been detected only on the east wall on small areas. Given the fact that lapis lazuli is difficult to detect by non-invasive analyses, especially when mixed with other pigments such as Egyptian blue, a more widespread use in Müstair cannot be excluded.

The presence of arsenic in some red areas with ochres and not in others suggests the use of two different types of red-brown pigment. While today, due to the fire alterations, no difference can be seen between the two pigments, in the past this might have been otherwise.

Thanks to the stratigraphic samples, a complex execution technique for the background of the central scene on the east wall has been discerned. The Ascension of Christ stands out against a backdrop of a pale blue sky adorned with multi-coloured clouds. The blue hue of the sky was achieved using Egyptian blue, which was also used to outline the clouds. These clouds were painted using three distinct shades, employing massicot, minium, and ochre in various proportions. The gradation of shades probably created a transition from a light yellow at the top to orange and eventually dark red towards the bottom of the clouds. Notably, the purple-black frame surrounding the Ascension of Christ was partially painted over one of the clouds. An organic dye was probably employed for this background, overlaid with a glaze likely composed of azurite. The original transparency of this layer remains uncertain, leaving it unclear if the cloud partially covered by it would have still been visible. If so, the purple background would have appeared to form a transparent framing of the ascension against the sky. The mandorla of Christ, which now shares the same purple-black colour as the background, was created using an organic dye mixed with Egyptian blue, likely giving it a blueish tint compared to the surrounding purple backdrop.

These newfound details shed light on the intricate execution technique of the paintings, revealing how much of the original complexity is hidden to the naked eye today due to their deterioration to several factors. Alongside the mechanical damage incurred during the detachment process, the adverse effects of (at least) three fires that the paintings endured in the church attic must have had a role. These fires undoubtedly inflicted profound harm, especially on the earth pigments. Additionally, certain pigments, such as those containing lead and copper, have blackened over time because of their inherent instability and as result of oxidative processes. The original appearance of the paintings can therefore only be reconstructed through careful scientific analysis.

Similar to the Carolingian paintings, the Romanesque phase has faced challenges stemming from fires and pigment degradation. Presently, hues of red, brown and beige dominate the Romanesque artworks detached from the attic of the church. Unlike their Carolingian counterparts, these paintings were executed using primarily mineral pigments in a fresco technique, thereby mitigating some problems regarding chemical alterations of the pigments. Therefore, the discoloration observed in the Romanesque wall paintings primarily stems from the intense heat generated by the fires. Notably, two stacchi sections have retained the original colours (including green, purple, red–orange, yellow, and white) in their lower portions. This preservation is attributed to the shielding provided by debris accumulated atop the vaults in the attic, which effectively insulated these areas from the heat. This preservation afforded researchers a unique opportunity to study the mechanisms of colour alteration. In the case of the green pigment, a distinct mineralogical modification was evident, with haematite identified as the product of fire-induced deterioration of the green earth pigment.

Moreover, the examination of the stacchi yielded another significant discovery: the identification of a pigment previously unrecognised in the Romanesque cycle. A purple hue, likely achieved through the use of an organic substance, and seemingly overlaid with a copper-rich layer, possibly a glaze containing azurite. Should this hypothesis prove accurate, it would suggest a technique akin to that employed in the Carolingian cycle on the eastern wall. It would also be the first indication of the use of azurite on the Romanesque wall paintings.

Although the primary motivation behind the analyses conducted on the stacchi and strappi from Müstair, as presented in this paper, was rooted in concerns about the conservation of the paintings, they have yielded significant insights into the execution techniques employed. The findings highlighted in this paper hold significant importance, shedding light not only on the wall paintings in Müstair but also contributing to a deeper understanding of Carolingian and Romanesque art on a broader scale. This underscores the symbiotic relationship between conservation and research, showcasing how conservation science can provide invaluable insights to art historians and archaeologists, enriching our comprehension of medieval wall painting techniques.