Material characterisation of a painted beehive panel by hyperspectral imaging in 3 combination with advanced spectroscopic and chromatographic techniques

: 19 In this study, a painted beehive panel from the collection of Slovene Ethnographic Museum was 20 examined with respect to its material composition with the aim to reveal the painting technique. 21 Due to the state of degradation due to outdoor weathering (UV irradiation, rainfall, extreme 22 temperature and humidity fluctuations), as well as past conservation interventions, the object 23 represented a complex analytical challenge. We aimed for non-invasive techniques (FTIR in 24 reflection mode, Raman spectroscopy and hyperspectral imaging in the range of 400 to 2500 nm), 25 however in order to explore paint layers, cross-sections were also analysed using Raman 26 spectroscopy. FTIR spectroscopy in transmission mode and gas chromatography coupled to mass 27 spectrometry were also used on sample fragments. Various original materials were identified such 28 as pigments and binders. The surface coating applied during restoration interventions was also 29 characterised. Additionally, organic compounds (oxalate, carboxylate), representing 30 transformation products, were found. The potential use of Prussian blue as a background paint 31 layer is discussed. 32


conservation. 48
Most existing research is art-historical or ethnographical and our current understanding of the 49 painting process is based on 20 th century texts and observational studies [3, 4, 5], which present 50 following possible characteristics. A base layer of paint, most often calcium carbonate [4], might 51 have been applied directly across the entire front of a wooden panel forming the background of 52 the composition. Once this foundation layer dried, the composition was stencilled or drawn with a 53 pencil; this was then painted out such that forms were filled with mostly basic colour tones 54 (modelled or not), and then painted further with details and contours. Each layer of paint was left 55 to dry before the next one was applied. Oil paints, probably hand-made of locally available linseed 56 oil or oil of turpentine and durable mineral pigments [5], might have been used in most cases, 57 although tempera paints and binders such as poppy seed oil, mastic resin, egg white and egg-coated/varnished with bleached linseed oil mixed with mastic resin [4]. 60 Spectroscopic and separation techniques such as Raman, FTIR and GC-MS are well established in 61 the investigation of heritage materials [6,7] including in the analysis of panel paintings [8,9], while 62 Delaney et al. explored the used hyperspectral imaging and imaging spectroscopy [10]. In a study 63 of an Italian early renaissance panel painting, a combination of different molecular and elemental 64 spectroscopic imaging methods was shown to provide insight into artistic materials, pigment 65 distribution and underdrawings [11], while hyperspectral imaging and other spectroscopic 66 techniques were used for another similar object [12], where mapping of wax and other organic 67 materials is described and discussed. 68 For characterisation of drying oils gas chromatography coupled to mass spectrometry (GC-MS) is 69 often used to determine the palmitic-stearic and azelaic-palmitic acid ratios, on the basis of which 70 the most frequently used drying oils (linseed oil, poppyseed oil and walnut oil) can be reliably 71 identified [13,14] and other organic materials such as waxes and triterpenoid resins can also be 72 revealed [14]. 73 In the present study, the material composition of the painted beehive panels was studied in order 74 to explore the painting technique. Paint stratigraphy was researched to investigate the hypothesis 75 that contours are generally better preserved as they are constituted of several paint layers, and to 76 explore the background paint layer which is mostly deteriorated.  EOS 350D camera with EF-S18-55 mm f/3.5-5.6 II lens and UV-cut-off filter were used. 91 The locations of the point-based non-invasive analyses are marked in Figure 1a ("A" indicates FTIR 92 spectroscopy in reflection mode, "R" indicates Raman spectroscopy). Destructive sampling 93 locations (PK2-1, PK2-2, PK2-2b, PK2-3, PK2-4) are denoted in Figure 1a as well. A part of these 94 samples containing strata from the support to the uppermost layer was used for FTIR spectroscopy 95 (PK2-1, PK2-2, PK2-3), GC-MS analyses (PK2-1 and PK2-2), and for preparation of cross-sections 96 used in optical and Raman microscopy. The exception was sample PK2-4 used in its raw form for 97 were taken from the painted beehive panels, placed between the windows of a diamond anvil cell 115 and examined under microscope. 116 Non-invasive FTIR analysis of the panel's surface was carried out with a portable Alpha-R 117 spectrometer from Bruker Optics. The pseudo-absorption spectra (A'=log (1/R), R=reflectance) 118 spectra were collected in reflection mode between 7500 and 400 cm -1 , at 4 cm -1 spectral 119 resolution. 160 scans per sample were averaged and for the background measurement, a gold 120 mirror was used. An integrated video camera controlled and monitored the sampling area. 121 Processing of the FTIR data was implemented using Bruker OPUS software. 122

Raman spectroscopy 123
The spectra were recorded using a 785 nm and 514 nm laser excitation lines with a Horiba Jobin 124 Yvon LabRAMHR800 Raman spectrometer coupled to an Olympus BXFM optical microscope. The 125 spectra were recorded using ×50 LWD objective lens and/or ×100 objective lens and a 600 126 grooves/mm grating. A multi-channel, air-cooled CCD detector was used. Experimental parameters 127  The dimensions of the panel allowed investigation directly under the microscope in a non-invasive 131 manner. The panel was placed directly under the objective and the spectra were then collected 132 using the x50 LWD objective from the locations of the interest. Further analysis was done also on 133 the cross-sections of the samples. In such case, the cross-section was placed under the microscope 134 and investigated using x100 objective lens. Spectral interpretation and identification of the 135 materials were done in comparison with own spectral database and the literature [15,16]. 136

Hyperspectral imaging 137
The system comprises of a high-resolution ClydeHSI Hyperion Art Scanner that can accomplish 138 scan areas up to 2.2 m x 2.2 m with an optical spatial resolution better than 25 μm. This scanner 139 was used with either a push-broom VNIR (400 to 1,000 nm, Δλ = 3 nm (FWHM)) or SWIR (900 to 140 2,500nm, Δλ = 10 nm (FWHM)) hyperspectral cameras each capable to provide a spatial resolution 141 on the panel of better than 0.3 mm. The scanner was fitted with a dual distance sensor to ensure 142 that even curved surfaces will remain in focus. Illumination was made using a tungsten light source 143 that has a smooth spectral emission from approx. 350 nm to 3,500 nm. The illumination level in 144 the visible spectrum was ca. 2,000 lux. At the start and end of each scan a reflective white tile was 145 measured to record the instrument spectral-spatial response function, and this was used to 146 convert the raw data signal into reflectance and absorption data. Subsequently, data analysis was 147 made using Principle Components Analysis (PCA) and Spectral Angle Mapping (SAM) methods to 148 extract the location of materials and their distributions across the panel. The cross-sections of the samples were examined using an Olympus BX 60 microscope connected 153 to an Olympus SC-50 video camera using visible and ultraviolet (UV) illumination, the latter 154 emitted from a Hg bulb Ushio USH-1030L. 155

GC-MS Analysis 156
A sample (0.5-2 mg) was treated with 3 mL of 0.5 M methanolic solution of NaOH and 300 µL of 157 dichloromethane. After flushing with nitrogen, the closed vials were heated for 10 min at 90 °C. 158 The vials were cooled briefly before addition of 3 mL of a 12% methanolic solution of H2SO4. After 159 nitrogen flushing, the closed vials were heated for 10 min at 90 °C. The vials were then cooled to 160 room temperature. 3 mL deionized water and 1.5 mL hexane were added to the vials and then 161 fatty acids methyl esters (FAMEs) were extracted by vigorous shaking for about 1 min. Following 162 centrifugation, the top layer was transferred into a vial for GC-MS analysis. 163 GC-MS analyses were carried out using Thermo Scientific Focus GC with a mass spectrometric 164 detector Thermo Scientific ISQ. Chromatographic separation was achieved using a Supelco, 165 Omegawax 320 capillary column (bonded polyethylene glycol stationary phase; 30 m x 320 µm x 166 0.25 µm). The injector temperature was set at 200 °C and the interface temperature at 250 °C. The 167 oven was programmed from 185 °C, then increased at 1 °C/min to 215 °C, stayed constant for 9 168 min and then decreased at 10 °C/min to 185 °C. The injection volume was 2 µL and the inlet was 169 operated in split mode, with a 1:5 split ratio. The carrier gas was helium at a constant flow of 2 170 mL/min. 171 The painted beehive panel was examined to obtain a comprehensive picture of the applied 173 materials and their potential transformations ( Table 1) The number of detected pigments is significant for an object of folk art, where the colour palette 189 was usually limited. As these objects faded due to outdoor weathering, they tended to get 190 frequently "refreshed", i.e. painted over, while in use; this could have been done in situ by folk artists themselves or by owners. Once acquisitioned, curators or conservators may have done the 192 The best preserved areas of PK2 are the contours of the painted figures and objects, mostly 194 executed in a darker red-brown colour, along with the draperies painted in red, and some of the 195 partially remaining whitish, green, brown, blue and black painted sections. In the green coloured 196 regions such as tree foliage (R4, R5, R18, A5, Figure 1a) and the lighter green apron of the female 197 figure (R7, R19, A3), barium sulfate was detected using Raman spectroscopy, along with Prussian 198 blue (R5, R18). Both were present in the FTIR spectra as well. Raman analyses corroborated the 199 detection of cinnabar in the red draperies (R1, Figure 1a), although in one red area (A1, Figure 1a Figure 1a); however, using FTIR in reflective mode, Prussian blue was 206 also detected (A2, Figure 1a) which could either be a part of the same or of an underlying layer. 207 There are several whitish areas still preserved on the panel; however, it wasn't possible to identify 208 the pigment using Raman analysis, with the exception of a possible presence of calcium carbonate 209 in one location (R6, Figure 1a). Furthermore, FTIR again showed the presence of Prussian blue in 210 one of the white areas (A4, Figure 1a). Calcium carbonate was detected also on PK2-1 in the layer 211 containing ultramarine, although this could be considered as an impurity. Using Raman 212 the cross-sections of PK2-1, PK2-2b and PK2-3, in the paint layer closest to the wooden support. 214 Barium sulfate was found in the majority of areas of both darker and lighter green colour, and in 215 particular in all point analyses where Prussian blue was identified as well (R5, R18, A1, A3-A5, A7, 216   To investigate the matter further, the sample PK2-1 was taken from the dark blue contour of a 290 garment (Figure 1a). This consisted of all stratigraphic layers -from the wooden support to paint 291 layers as evident in the optical micrograph of the cross-section (Figure 5a). Two paint layers are 292 present: a lower one of lighter blue/greenish-blue colour and an upper one of darker blue colour. 293 The Raman spectrum (Figure 5b) at the spot R1 in Figure 5a  The hyperspectral images showed a uniform distribution of wax on the panel surface; therefore, 342 waxes are most likely part of the resin varnish. However, beeswax could have been secreted by 343 bees as well, depending on how thorough conservation cleaning was. Important information could 344 potentially be obtained from such "original" wax. Chemical characterisation of beeswax using GC-345 MS can enable nest-mate recognition based on fatty acid composition. Beeswax plays a role in the 346 chemical communication within a honey bee colony and provides a chemical signature so that 347 subtle differences in its composition may be important and have never been studied historically. 348 Isotope ratio mass spectrometry (IRMS) was found to be very useful in establishing the 349 geographical origin of beeswax [22]. All of this indicates that much can still be learned from  Optical micrograph of the cross-section of PK2-3 ( Figure ) shows that all the paint layers were 376 removed as the wood structure is clearly visible. This sampling location is marked with a white 377 circle A in Figure 9, and was used to derive reference spectral features for wood in order to 378 perform hyperspectral analysis of distribution of exposed wood. The red areas in the false colour 379 image in Figure 9  Relatively large collections of painted beehive panels, which are relatively uniform objects, 397 represent an interesting source for systematic studies under various aspects. Therefore, a 398 comprehensive material characterisation of a painted beehive panel was carried out using FTIR 399 and Raman spectroscopy, gas chromatography coupled to mass spectrometry and hyperspectral 400 imaging. The investigated object was degraded, likely as a consequence of outdoor weathering. As 401 such it represented a complex analytical problem requiring complementary analytical techniques. 402 Many different compounds were identified representing pigments and their components 403 (cinnabar/vermillion, iron oxide, lead oxide, iron hydroxide, carbon-based black, Prussian blue, 404

A B
Emerald green) as well as binders (linseed oil). A surface conservation varnish was identified as a 406 mixture of a triterpenoid resin and beeswax, with a possible addition of carnauba wax. 407 Compounds indicative of degradation were identified such as oxalates and carboxylates. 408 Prussian blue was identified in many locations using FTIR and Raman spectroscopy and the 409 conclusion was made that it was used as a background colour. 410 Beeswax, however, could have two origins and some of it could have been deposited by bees 411 themselves. Studying these depositions might allow us to understand bee activities at the entry 412 into the beehive or their affinity for different colours. Beeswax may also hold clues as to the origin 413 of panels, however, this may no longer be possible due to the conservation varnish, which 414 represents the second source of beeswax. 415 This study represents a comprehensive material study of a painted beehive panel and will serve as 416 the blueprint for further analysis of objects in the extensive collections of such panels in Slovenian 417 museums, with the aim to develop a procedure for condition evaluation. Based on this preventive 418 conservation strategies and storage recommendations could be developed in the future. 419 420 Acknowledgments 421 This study is part of the InnoRenew CoE project (start-up project 6.1. Advanced materials for 422 cultural heritage storage). The authors are grateful to the Slovene Ethnographic Museum for