**2. Materials and Methods**

The icon (dimensions: 148.2 × 58.7 cm) was initially pictured using a MuSIS-MS multispectral camera (FORTH-Photonics, Heraklion, Greece) in the 1000 nm and the false-color infrared (IRFC) modes. Micro-samples (~1 mm × 1 mm) were removed from damaged areas using surgical scalpels (Figure 1), and, after preliminary stereoscope investigation, they were embedded in polyester resin, cross-sectioned and subjected to grinding and polishing (Pedemin-2, DAP-7, Struers, Ballerup, Denmark). Cross-sections were examined under an optical microscope (OM, DMRXP, Leica Microsystems, Wetzlar, Germany) at magnifications up to 200×, and, upon carbon coating (for conductivity purposes, using a Balzers' CED030 carbon vaporizer, Leica Microsystems, Wetzlar, Germany), through a scanning electron microscope coupled with an energy dispersive analyzer (SEM-EDX, Quanta Inspect D 8334, FEI, Hillsboro, Oregon, USA). Elemental compositions were estimated by using the built-in 'Genesis-Spectrum' software (EDAX Company, Mahwah, NJ, USA), in a standard-less quantification method mode that incorporates ZAF matrix corrections [16], and in combination with high accelerating voltage (25 kV) and optimal spectra collection parameters (a high count rate, long collection times, adequate DT%, etc. leading to high elemental peak to background ratios). Through the analysis of multi-elemental standard targets, it was demonstrated that this approach results in quantitative analysis with errors of circa ±3% for high concentration elements, and ±20% for low concentration ones (<5%). For each distinct pigment/phase, at least three EDX analyses were undertaken, targeting on different grains/areas; results were automatically normalized to 100% and the mean values were calculated. Due to the presence of the conductive carbon layer, carbon was not quantitatively estimated in organic/lake-type pigments. Micro-morphological characteristics were recorded using the SEM's backscattered electron detector (BSE), which permits for the differentiation of the observed phases on the basis of their atomic number. Also, the size of the various pigment grains along with the thickness of the gold leaves and the pertinent adhesives were determined using a built-in image processing tool of the SEM device (Table 1. Samples cross-sections were further examined under a µ-Raman device (inVia, Renishaw, Wotton-under-Edge, UK) using a low power (~0.01–1 mW) 514 nm laser; spectra were collected through a 100× magnification lens with repeated acquisitions of varying durations, and recorded in frequencies of 100–1800 cm−<sup>1</sup> . A minor ground/gesso sample (<1mg) was pulverized and analyzed by using X-ray diffraction (XRD, 'D5000 , SIEMENS, Munich, Germany, equipped with a Cu-Kα anticathode, diffraction pattern recorded in the range of 2–90◦ (2θ) with a step size of 0.04◦ and a scan speed of 2 s per step). Note that during older conservation interventions, the painting and ground layers of the icon were detached from their original wooden substrate and placed onto a new one [7]; therefore the present authors did not employ techniques that pertain to wooden panel identification (e.g., x-ray radiography).



*Heritage* **2020**, *3*

### **3. Results** *Heritage* **2020**, *3* FOR PEER REVIEW 5

First, we present the results of the multispectral imaging, and the data that pertain to materials identification follow. The latter are presented in terms of the stratigraphy of a typical icon [17]: first the data on the ground/preparatory layer are presented, then the pigment palette is disclosed through the paint layers analysis results, and finally the data that pertain to gilded decorations are discussed. **3. Results**  First, we present the results of the multispectral imaging, and the data that pertain to materials identification follow. The latter are presented in terms of the stratigraphy of a typical icon [17]: first the data on the ground/preparatory layer are presented, then the pigment palette is disclosed

### *3.1. Multispectral Imaging* through the paint layers analysis results, and finally the data that pertain to gilded decorations are discussed.

The potential of infrared radiation to penetrate through the upper layers of paintings has been exploited in order to reveal layers that are invisible to the naked eye (such as underdrawings) [18,19] as well as for pigments identification [20]. In case of the St Theodoros icon, the inspection at 1000 nm revealed a wealth of information pertaining to the painting technique. The preliminary drawing is of a notably confident character, created by employing two techniques, namely brushstrokes and extremely thin (<30 µm, see next) incisions (Figure 2a,b). It is worth noting that in the case of the Saint's face and curly hair (where accuracy in sketch is of utmost importance), the drawing was rendered through thin brushstrokes (no incisions), while only few minor sketch-corrections were spotted in the corresponding areas (Figure 2c,d, arrow B). The preparatory paint layers that followed drawing (base colors/underpaintings [17]) were freely applied onto the ground (Figure 2d, arrow A), while the subsequent lighter tones and highlights were rendered with extremely accurate/skillful and fine brushstrokes (Figure 2). On the other hand, IRFC photography gave some hints on the employed pigments. For instance, the red mantle is rendered in an intense yellow-orange false color, thus implying the presence of cinnabar, while the differences in the false color of the "greenish" armor parts and the underwear garment around the Saint's waist indicate employment of different pigments (Figure 2e,f) [21]. *3.1. Multispectral Imaging*  The potential of infrared radiation to penetrate through the upper layers of paintings has been exploited in order to reveal layers that are invisible to the naked eye (such as underdrawings) [18,19] as well as for pigments identification [20]. In case of the St Theodoros icon, the inspection at 1000 nm revealed a wealth of information pertaining to the painting technique. The preliminary drawing is of a notably confident character, created by employing two techniques, namely brushstrokes and extremely thin (<30 μm, see next) incisions (Figure 2a,b). It is worth noting that in the case of the Saint's face and curly hair (where accuracy in sketch is of utmost importance), the drawing was rendered through thin brushstrokes (no incisions), while only few minor sketch-corrections were spotted in the corresponding areas (Figure 2c,d, arrow B). The preparatory paint layers that followed drawing (base colors/underpaintings [17]) were freely applied onto the ground (Figure 2d, arrow A), while the subsequent lighter tones and highlights were rendered with extremely accurate/skillful and fine brushstrokes (Figure 2). On the other hand, IRFC photography gave some hints on the employed pigments. For instance, the red mantle is rendered in an intense yellow-orange false color, thus implying the presence of cinnabar, while the differences in the false color of the "greenish" armor parts and the underwear garment around the Saint's waist indicate employment of different pigments (Figure 2e,f) [21].

**Figure 2.** (**a**) Visible macro-detail of the armor. (**b**) Same area as in (**a**), pictured at 1000 nm. Preliminary drawing executed by incision (arrows "A") and brushstrokes ("B"); insert picture (lower left corner) shows an incision cross-section (scanning electron microscope (SEM), backscattered electron detector (BSE), 2000×). (**c**) St Theodore face, detail on visible light. (**d**) Same area as in (**c**), infrared (1000 nm). Arrow "A" points on preliminary paint layer brushstrokes, "B" on a minor **Figure 2.** (**a**) Visible macro-detail of the armor. (**b**) Same area as in (**a**), pictured at 1000 nm. Preliminary drawing executed by incision (arrows "A") and brushstrokes ("B"); insert picture (lower left corner) shows an incision cross-section (scanning electron microscope (SEM), backscattered electron detector (BSE), 2000×). (**c**) St Theodore face, detail on visible light. (**d**) Same area as in (**c**), infrared (1000 nm). Arrow "A" points on preliminary paint layer brushstrokes, "B" on a minor sketch correction. (**e**) Detail, visible light. (**f**) The area figured in (**d**) as it was pictured in the false-color infrared (IRFC) mode.

## *3.2. Ground*/*Gesso 3.2. Ground/Gesso*

false-color infrared (IRFC) mode.

During the microscopic probing of the cross-sections, it was observed that the preparatory ground layer consists of up to eight distinct sub-layers of ~50–150 µm thickness, which correspond to the successive gesso coatings applied onto the wooden panel (Figure 3a). XRD and µ-Raman analyses revealed that the inorganic ground component is gypsum (CaSO4·2H2O), which was probably mixed with an organic gluing agent [17]. For instance, the relevant µ-Raman spectrum shows a characteristic shift at ~1008 cm−<sup>1</sup> that corresponds to the v<sup>1</sup> (SO4) symmetric stretching mode of gypsum (Figure 4a) [22], In addition, the ground layer contains minor admixtures of black, red and yellow pigments (see insert on Figure 3a). During the microscopic probing of the cross-sections, it was observed that the preparatory ground layer consists of up to eight distinct sub-layers of ~50–150 μm thickness, which correspond to the successive gesso coatings applied onto the wooden panel (Figure 3a). XRD and μ-Raman analyses revealed that the inorganic ground component is gypsum (CaSO4∙2H2O), which was probably mixed with an organic gluing agent [17]. For instance, the relevant μ-Raman spectrum shows a characteristic shift at ~1008 cm−1 that corresponds to the v1 (SO4) symmetric stretching mode of gypsum (Figure 4a) [22], In addition, the ground layer contains minor admixtures of black, red and yellow pigments (see insert on Figure 3a).

*Heritage* **2020**, *3* FOR PEER REVIEW 6

**Figure 3.** (**a**) Multiple ground layers, optical microscope (OM), 50×. The insert picture shows scattered grains of black and yellow pigments in the ground (OM, 50×). (**b**) Natural ultramarine grains and their maximum dimensions; arrows point on charcoal particles that lay into the lazurite substrate (SEM, BSE, 4000×). **Figure 3.** (**a**) Multiple ground layers, optical microscope (OM), 50×. The insert picture shows scattered grains of black and yellow pigments in the ground (OM, 50×). (**b**) Natural ultramarine grains and their maximum dimensions; arrows point on charcoal particles that lay into the lazurite substrate (SEM, BSE, 4000×).

### *3.3. Paint Layers 3.3. Paint Layers*

Pigments employed in the St Theodoros icon were identified through SEM-EDX and μ-Raman spectroscopy (Table 1 and Figure 4). Thus, a palette consisting of nine distinct pigments was revealed: natural ultramarine, green earth, two types of iron ochre, cinnabar, minium, red lake, charcoal and lead white (Table 1 and Figure 4). The extremely expensive and rather rare ultramarine pigment was identified through its characteristic Raman spectrum (v1 stretching vibration mode and v2 bending vibration mode of S3- , at 548 cm−1 and 258 cm−1, respectively [23]) and its elemental composition (Table 1), while the characteristic conchoidal fracture features of the relevant grains and the detection of minor calcite (natural impurity) verify the natural origin of the particular pigment (Figures 3b and 4b, Table 1) [23,24]. Pigments employed in the St Theodoros icon were identified through SEM-EDX and µ-Raman spectroscopy (Table 1 and Figure 4). Thus, a palette consisting of nine distinct pigments was revealed: natural ultramarine, green earth, two types of iron ochre, cinnabar, minium, red lake, charcoal and lead white (Table 1 and Figure 4). The extremely expensive and rather rare ultramarine pigment was identified through its characteristic Raman spectrum (v<sup>1</sup> stretching vibration mode and v<sup>2</sup> bending vibration mode of S<sup>3</sup> - , at 548 cm−<sup>1</sup> and 258 cm−<sup>1</sup> , respectively [23]) and its elemental composition (Table 1), while the characteristic conchoidal fracture features of the relevant grains and the detection of minor calcite (natural impurity) verify the natural origin of the particular pigment (Figures 3b and 4b, Table 1) [23,24].

In the case of the green pigment, authors were unable to collect Raman spectra. However, the EDX analysis revealed that the pertinent grains are mainly composed of silicon, iron, potassium and magnesium, and this elemental composition evidently shows employment of green earth (Table 1) [28,29]. Similarly, the use of two iron ochre varieties was attested to through SEM-EDX analyses, as the deep-red and the yellowish ochre differ drastically in terms of elemental composition (especially as regards the content of iron, calcium, silicon and chlorine, see Table 1). In addition, the grains of these two pigments are of a notably small size (0.5–5 µm), and this is so in the case of cinnabar and lead white as well (0.5–8 µm, see Figure 5a). Note that a cinnabar Raman spectrum is displayed on Figure 4c; the characteristic shifts at 253, 282 and 343 cm−<sup>1</sup> originate from a totally symmetric A<sup>1</sup> and degenerated E transverse modes (ETO) respectively [30]. The employment of these thin-grained pigment fractions reflects intense grinding and suggests meticulous pigment preparation.

*Heritage* **2020**, *3* FOR PEER REVIEW 7

*Heritage* **2020**, *3* FOR PEER REVIEW 7

**Figure 4.** Characteristic μ-Raman spectra of St Theodoros icon ground and pigments. (**a**) Gypsum, characteristic peak at 1008 cm<sup>−</sup>1. (**b**) Natural ultramarine, Raman shifts at 258, 548, 815 and 1091 cm<sup>−</sup>1. (**c**) Cinnabar, shifts at 253, 282 and 343 cm<sup>−</sup>1. (**d**) Carbon black, characteristic shifts at 1363 and 1603 wavenumbers. Insert figures show indicative individual pigment grains that were analyzed. For libraries of pigment Raman spectra, the reader is directed to [25–27]. **Figure 4.** Characteristic µ-Raman spectra of St Theodoros icon ground and pigments. (**a**) Gypsum, characteristic peak at 1008 cm−<sup>1</sup> . (**b**) Natural ultramarine, Raman shifts at 258, 548, 815 and 1091 cm−<sup>1</sup> . (**c**) Cinnabar, shifts at 253, 282 and 343 cm−<sup>1</sup> . (**d**) Carbon black, characteristic shifts at 1363 and 1603 wavenumbers. Insert figures show indicative individual pigment grains that were analyzed. For libraries of pigment Raman spectra, the reader is directed to [25–27]. these two pigments are of a notably small size (0.5–5 μm), and this is so in the case of cinnabar and lead white as well (0.5–8 μm, see Figure 5a). Note that a cinnabar Raman spectrum is displayed on Figure 4c; the characteristic shifts at 253, 282 and 343 cm−1 originate from a totally symmetric A1 and degenerated E transverse modes (ETO) respectively [30]. The employment of these thin-grained pigment fractions reflects intense grinding and suggests meticulous pigment preparation.

pigment fractions reflects intense grinding and suggests meticulous pigment preparation. **Figure 5.** (**a**) Notably small grains of ochre (bottom) and cinnabar (up); SEM, BSE, 10,000×. (**b**) Minium grains (big bright particles) among ochre (gray particles); SEM, BSE, 5000×. (**c**) Lake glaze (upper layer, uniform) on top of an ochre and lead white substrate (bottom layer, spotted) (SEM, BSE, 3000×). Insert picture: the same sample under OM, the arrow points on the glaze (100×). **Figure 5.** (**a**) Notably small grains of ochre (bottom) and cinnabar (up); SEM, BSE, 10,000×. (**b**) Minium grains (big bright particles) among ochre (gray particles); SEM, BSE, 5000×. (**c**) Lake glaze (upper layer, uniform) on top of an ochre and lead white substrate (bottom layer, spotted) (SEM, BSE, 3000×). Insert picture: the same sample under OM, the arrow points on the glaze (100×).

**Figure 5.** (**a**) Notably small grains of ochre (bottom) and cinnabar (up); SEM, BSE, 10,000×. (**b**) Minium grains (big bright particles) among ochre (gray particles); SEM, BSE, 5000×. (**c**) Lake glaze (upper layer, uniform) on top of an ochre and lead white substrate (bottom layer, spotted) (SEM, BSE, 3000×). Insert picture: the same sample under OM, the arrow points on the glaze (100×). On the other hand, a few minium grains were spotted among red ochre particles, therefore it seems probable that the minium was added in order to slightly adjust the hue of the ochre (Figure On the other hand, a few minium grains were spotted among red ochre particles, therefore it seems probable that the minium was added in order to slightly adjust the hue of the ochre (Figure 5b). Of special interest is the case of the deep-red lake, which was used as a glaze (translucent paint layer) that covers an ochre plus lead white paint layer (Figure 5c), which is in fact a technique quite commonly applied in Cretan icons [31]. Here the lake organic coloring compound could not be identified, yet the elevated phosphorous (4.1 wt%) is compatible with the employment of insect dye [32]. Finally, charcoal of plant origin was applied as a preliminary paint layer in the areas rendered in lazurite (Figure 3b) and as a minor addition in various paint layers. Charcoal was also used to render the preliminary drawing (Figure 6b), while a minute amount of the same pigment was included in the ground layer/gesso (Figure 3a). The corresponding Raman spectra show the typical G and D bands of carbon at ~1600 cm−<sup>1</sup> and ~1360 cm−<sup>1</sup> , respectively [33].

On the other hand, a few minium grains were spotted among red ochre particles, therefore it seems probable that the minium was added in order to slightly adjust the hue of the ochre (Figure

*3.4. Gilded Pictorial Elements* 

therefore some overestimation is possible [35,36].

*Heritage* **2020**, *3* FOR PEER REVIEW 8

5b). Of special interest is the case of the deep-red lake, which was used as a glaze (translucent paint layer) that covers an ochre plus lead white paint layer (Figure 5c), which is in fact a technique quite commonly applied in Cretan icons [31]. Here the lake organic coloring compound could not be identified, yet the elevated phosphorous (4.1 wt%) is compatible with the employment of insect dye [32]. Finally, charcoal of plant origin was applied as a preliminary paint layer in the areas rendered in lazurite (Figure 3b) and as a minor addition in various paint layers. Charcoal was also used to render the preliminary drawing (Figure 6b), while a minute amount of the same pigment was included in the ground layer/gesso (Figure 3a). The corresponding Raman spectra show the typical

The icon background ("campus") along with the highlights of the armor and certain vestment details (e.g., bracelets) are rendered in gold tones. Micro-samples investigation revealed that these particular pictorial elements are in fact gilded with high purity (Au > 99 wt %) and extremely thin (<1 micron) gold leaves (Figure 6, Table 1). The latter have been applied by employing two distinct gluing agents, a yellow iron-rich clayey bole in case of the background and a lead-containing mordant in the highlights (Table 1). These adhesives pertain to the two most common—in the framework of painting—gilding techniques, namely water and mordant/oil gilding, respectively [34,35]. It shall be mentioned that the gold leaf thickness determination was achieved through

G and D bands of carbon at ~1600 cm−1 and ~1360 cm−1, respectively [33].

**Figure 6.** (**a**) Background image: thin gold leaf (bright layer) on a bole substrate (dark gray substrate, 3.12 μm marker); BSE, 10,000×. Insert picture: same sample, OM, 200×; note the yellow bole layer. (**b**) Background image: Double gold leaf (uppermost bright layers) on a lead-mordant (bright substrate); BSE, 8000×). Insert picture: same sample under OM, arrow points on charcoal grains that lay on the white gesso and correspond to preliminary drawings (200×). **Figure 6.** (**a**) Background image: thin gold leaf (bright layer) on a bole substrate (dark gray substrate, 3.12 µm marker); BSE, 10,000×. Insert picture: same sample, OM, 200×; note the yellow bole layer. (**b**) Background image: Double gold leaf (uppermost bright layers) on a lead-mordant (bright substrate); BSE, 8000×). Insert picture: same sample under OM, arrow points on charcoal grains that lay on the white gesso and correspond to preliminary drawings (200×).

### **4. Discussion**  *3.4. Gilded Pictorial Elements*

Through the analytical investigation of the St Theodoros icon, authors were able to identify the employed painting materials (except of the organics) and techniques, and now, a crucial question arises: how can these data contribute towards the assessment of painter's identity? To this end, the analytical data were compared to the findings of previous studies of Angelos' known (signed) works [10–12,14,15], and evaluated in the light of analytical investigations of other high-quality Cretan icons [31,37–39]. It is thus shown that the icon in consideration can indeed be assigned to Angelos. According to the pertinent studies, Angelos' works show a series of specific technical characteristics, that when seen as a whole constitute a rather idiomorphic painting manner. In detail, The icon background ("campus") along with the highlights of the armor and certain vestment details (e.g., bracelets) are rendered in gold tones. Micro-samples investigation revealed that these particular pictorial elements are in fact gilded with high purity (Au > 99 wt %) and extremely thin (<1 micron) gold leaves (Figure 6, Table 1). The latter have been applied by employing two distinct gluing agents, a yellow iron-rich clayey bole in case of the background and a lead-containing mordant in the highlights (Table 1). These adhesives pertain to the two most common—in the framework of painting—gilding techniques, namely water and mordant/oil gilding, respectively [34,35]. It shall be mentioned that the gold leaf thickness determination was achieved through inspection of high magnification SEM images, using a built-in image processing software (Figure 6) therefore some overestimation is possible [35,36].
