RBS, PIXE, Ion-Microbeam and SR-FTIR Analyses of Pottery Fragments from Azerbaijan

: The present work is aimed at the investigation of the ceramic bulk and pigmented glazed surfaces of ancient potteries dating back to XIX century A.D. and coming from the charming archeological site located in the Medieval Agsu town (Azerbaijan), a geographic area of special interest due to the ancient commercial routes between China, Asia Minor, and Europe. For the purpose of the study, complementary investigation tools have been exploited: non-destructive or micro-destructive investigation at elemental level by ion beam analysis (IBA) techniques, by using Rutherford Backscattering Spectrometry (RBS), Proton-Induced X-ray Emission (PIXE) spectroscopy and ion-microbeam analysis, and chemical characterization at microscopic level, by means of synchrotron radiation (SR) Fourier transform infrared (FTIR) microspectroscopy. The acquired information reveals useful for the identiﬁcation of the provenance, the reconstruction of the ﬁring technology, and ﬁnally, the identiﬁcation of the pigment was used as a colorant of the glaze. AZR7, and AZR1 potsherds. Several steps can be seen in the RBS spectra, corresponding to the various detected elements.


Introduction
Pottery represents the most abundant finds in the archaeological excavations and its appearance (vessel shape, style) is highly influenced by the cultural-social changes. Thus, the scientific investigation of ancient potteries, i.e., the analysis of their mineralogical, chemical, and physical properties, represents a first crucial step in order to achieve information on many aspects of the past including provenance, dating, trade, and technology [1][2][3]. Their study can therefore be useful in identifying specific cultural groups and their mutual influences are of great relevance from both restoration and conservation perspectives.
Nevertheless, the characterization of archaeological pottery is a rather difficult procedure due to the presence of a variety of minerals and to the complex features of the firing process [4,5]. Furthermore, decorated potteries are even more complex to be analyzed, because of the difficulty to isolate the glazed and silicon can be detected because of their non-Rutherford backscattering cross sections for protons with energies above 2 MeV. Hence, the simultaneous application of these two methodologies allows for the measurement of almost all elements present in the pottery. In addition, when a micro-beam is employed, it is possible to map the material properties exploring an area of about 1 × 1 mm 2 .
Synchrotron radiation Fourier transform infrared (SR-FTIR) has been proved to be a powerful molecular spectroscopy technique for the characterization, other than that of the raw minerals constituting the bulk, of pigments, and binding media of painted materials [20]. It represents an advance over conventional FTIR spectroscopy since it guarantees a higher signal-to-noise ratio at diffraction-limited lateral resolution, thanks to the superior brightness of infrared synchrotron radiation (100 to 1000 times higher than conventional IR sources) [21,22]. Therefore, SR-FTIR results particularly suitable in the case of analysis of small and/or heterogeneous samples, such as ancient paintings as well as corrosion and alteration products, made by a mixture of organic and inorganic compounds distributed in a layered structure.
In the present study, IBA has been first of all applied to some of the previously investigated glazed pottery fragments coming from the Agsu site and dated back XIX century A.D. in order to deduce their almost total elemental composition, both in depth and at surface level. In fact, IBA techniques permit detection of the depth profile of analyzed elements through the control of the ion beam energy and of the known ion stopping powers in the irradiated matrix. By increasing the ion beam energy, it is possible to increase the ion range and the analysis depth; by decreasing the ion beam energy, the analysis is referred to the first surface layers.
In particular, the ceramic body and the decorated surface of the samples have been preliminary characterized at elemental scale by particle Rutherford Backscattering Spectrometry (RBS) and Proton Induced X-ray Emission (PIXE) spectroscopy. After that, the ion-microprobe has been used to obtain the micrometric elemental spatial distribution map of the representative samples.
Then, SR-FTIR analysis was performed on the same samples, in order to define, at the µm scale, raw materials, pigments, and binders used for the potteries production, so achieving crucial information in order to clarify technological features and preparation processes typical of the local population. Starting from them, the reproduction of similar materials in accordance with the ancient recipes, to be used in the restoration works, can be attempted.

Materials and Methods
We analyzed four pottery fragments (Figure 1), labelled as AZR3, AZR5, AZR7, and AZR1, taken in the medieval ruins of Agsu.
The shards probably come from domestic objects such as bowls, dishes or vessels. Based on the aforementioned preliminary archaeometric investigation [7], samples were selected as representative of Group 1 (AZR3, AZR5, and AZR7) and Group 2 (AZR1), as the variety and differences of colors found in the site are concerned.

Sample
Typology Description AZR3 Glazed pottery Dark beige ceramic body, black glaze AZR5 Glazed pottery Reddish ceramic body, dark yellow glaze AZR7 Glazed pottery Reddish ceramic body, yellowish glaze AZR1 Faience White ceramic body, light blue glaze Rutherford Backscattering Spectrometry (RBS) and Proton Induced X-ray Emission (PIXE) analyses were performed for all investigated samples on the external decorated convex side and on the back. Ion-microbeam analyses were performed in cross-section for AZR3, AZR5, and AZR7 fragments, and on the external glazed surface in the case of AZR1 shard.
RBS and PIXE investigations were carried out at the Nuclear Physics Institute in Rez (Czech Republic), using the 3 MeV Tandetron CANAM accelerator [23] to accelerate 3.1 MeV proton beams, collimated to 1.5 mm × 1.6 mm.
RBS was performed in high vacuum (10 −6 mbar) with an acquisition time of 500 s. During each measurement, the ion current was maintained at ∼6.0 nA. The backscattered H + ions were monitored by an Ultra-Ortec PIPS silicon detector placed at a scattering angle of 170°, having 100% detection efficiency for the proton energy detection analysis. RBS spectra were fitted using the SIMNRA simulation code [24] and IBANDL database [25] for experimental nuclear reaction data, cross section type, and resonance parameters. The measured RBS spectra were available for quantitative elemental analysis, here reported in terms of atomic percentages, by assuming that each element exhibits a RBS yield whose intensity is proportional to the content of that element in the investigated thickness of the sample, and that the total RBS profile will contain the weighted sum of the single elements contributions. The minimum detection limit (MDL) was of the order of 0.1% in atomic composition.
The same vacuum chamber was employed for PIXE, using two low energy X-ray solid-state detectors, i.e., a silicon drift SDD detector (Amptek XR-100FASTSDD) and a germanium Ultra-LEGe detector (CANBERRA model GUL-0110P). The X-ray detection energy window was in the (0.5 ÷ 20) keV range. The used detectors were located at scattering angular positions of 135° (SDD) and of 150° (LEGe), respectively. The detectors energy resolution was ∼122 eV for SDD and ∼145 eV for LEGe, respectively, measured at the 5.9 keV Kα-line of Mn. A polyethylene filter ∼115 µm thick (ρ = 0.98 g/cm 3 ) was interposed in front of the Ultra-LEGe detector to stop the backscattered particles and to  Rutherford Backscattering Spectrometry (RBS) and Proton Induced X-ray Emission (PIXE) analyses were performed for all investigated samples on the external decorated convex side and on the back. Ion-microbeam analyses were performed in cross-section for AZR3, AZR5, and AZR7 fragments, and on the external glazed surface in the case of AZR1 shard.
RBS and PIXE investigations were carried out at the Nuclear Physics Institute in Rez (Czech Republic), using the 3 MeV Tandetron CANAM accelerator [23] to accelerate 3.1 MeV proton beams, collimated to 1.5 mm × 1.6 mm.
RBS was performed in high vacuum (10 −6 mbar) with an acquisition time of 500 s. During each measurement, the ion current was maintained at~6.0 nA. The backscattered H + ions were monitored by an Ultra-Ortec PIPS silicon detector placed at a scattering angle of 170 • , having 100% detection efficiency for the proton energy detection analysis. RBS spectra were fitted using the SIMNRA simulation code [24] and IBANDL database [25] for experimental nuclear reaction data, cross section type, and resonance parameters. The measured RBS spectra were available for quantitative elemental analysis, here reported in terms of atomic percentages, by assuming that each element exhibits a RBS yield whose intensity is proportional to the content of that element in the investigated thickness of the sample, and that the total RBS profile will contain the weighted sum of the single elements contributions. The minimum detection limit (MDL) was of the order of 0.1% in atomic composition.
The same vacuum chamber was employed for PIXE, using two low energy X-ray solid-state detectors, i.e., a silicon drift SDD detector (Amptek XR-100FASTSDD) and a germanium Ultra-LEGe detector (CANBERRA model GUL-0110P). The X-ray detection energy window was in the (0.5 ÷ 20) keV range. The used detectors were located at scattering angular positions of 135 • (SDD) and of 150 • (LEGe), respectively. The detectors energy resolution was~122 eV for SDD and~145 eV for LEGe, respectively, measured at the 5.9 keV K α -line of Mn. A polyethylene filter~115 µm thick (ρ = 0.98 g/cm 3 ) was interposed in front of the Ultra-LEGe detector to stop the backscattered particles and to attenuate the low-energy X-rays (K-lines of Na, Al, Si, P, S, Cl, and K). This filter increases the detector sensitivity for the X-rays emitted from Ca, Ti, Mn and Fe. Both detectors were calibrated with a standard Al-Mg alloy from BAM with Ref. ERM-EB313. A DXP-XMap control and acquisition system for X-ray data mapping from company XIA LLC was used for measurements performed by using both detectors. The GeoPIXE software [24] was used for the analysis of the PIXE data, also taking into account various effects such as pile-up, escape peaks from Si and Ge, and secondary electron bremsstrahlung.
For each PIXE spectrum, the background was fitted to a SNIP algorithm. The peak tail and the full width at half maximum (FWHM) of the Mn K α -line (5.9 keV) were considered as free parameters. The GeoPIXE software furnished the elemental mass concentration and the corresponding peak areas resulting from the major lines of all the detected elements, together with the minimum detection limit (MDL) of the order of ppm.
Ion-microbeam analyses were performed in vacuum chamber, the energy of the micrometric proton beam (2 µm in diameter) was 2.6 MeV, the spot size 3.1 µm 2 , and the beam current~8 pA. Target was moved in the X-Y plane to achieve a maximum scanning area of 1 × 1 mm 2 in size. The presented results were obtained by the PIXE investigation, using an electronic filter to select the characteristic peak of interest and mapping in false colors the X-ray emission from single elements coming from the scanning areas of interest of the investigated sample. Ion range, stopping powers, straggling, and energy losses were calculated using the SRIM simulation code [26].
SR FTIR absorbance data were collected at the Life Sciences branch of the Synchrotron Infrared Source for Spectroscopy and Imaging (SISSI) beamline in Elettra Sincrotrone Trieste [27]. The instrument consists of a Bruker VERTEX 70 FTIR spectrometer coupled with Hyperion 3000 Vis/IR microscope. Firstly, single-point analyses were performed on small powdered portion of the samples (about 2 mg). Grains of interest for the analysis were selected under a stereomicroscope, and transferred by a sharp needle into a DAC (Diamond Anvil Cell) compression cell, flatten and then measured in transmission mode using the MCT detector. The approach allowed us to improve the selectivity of the FTIR analysis with respect to conventional spectroscopy since we selected the material of interest by visual inspection (reduced background effect). Due to the micrometric size of the grains, a microscopic approach that guaranteed a good S/N ratio was mandatory. For each powder sample, we collected 15-20 spectra from representative spots, in the (650-4000) cm −1 range with a spectral resolution of 4 cm −1 , an aperture size of 25 × 25 µm 2 , and averaging 256 scans per spectrum. Due to the complexity of the experimental FTIR profiles, the spectra were initially compared with those of standard minerals and/or pigments from databases [28,29] and literature [30] for a reliable assignment of the bands. Hence, spectral analysis included baseline removal and curve fitting through the PeakFit TM version 4.05 software (SPSS Inc., Chicago, IL, USA). The strategy adopted for the curve fitting procedure was to use well-defined Gaussian shape profiles, with all the parameters left free to vary upon iteration. The statistical parameters were used as a guide to "best-fit". While we are aware that the spectral decomposition procedures have no unique solution, we remark that the one we adopted here uses the minimum number of parameters and, at the same time, it furnishes extremely good fits to the data. The obtained results were in fact characterized by r 2~0 .9999. Furthermore, the choice of the components in the experimental spectra, with the assigned center-frequencies, was also suggested by the analysis of the second derivative profiles (data not shown) that showed minima approximately corresponding to the maxima of each band component.  Assuming the bulk matrix to be composed mainly by SiO2 in agreement with the literature [7], the RBS analysis refers to ∼a 40-50 μm depth, being 100 μm the penetration depth of the used proton beam in a SiO2 matrix. Therefore, in the case of measurements performed on the external surface, the RBS signal will reasonably reflect the contribution coming from both the glaze and the ceramic body. Instead, the RBS data collected from the back-side of the samples will be informative for the elemental composition of the bulk and can be used for comparison.

RBS-PIXE Analysis
The relative atomic concentrations of detected elements as obtained from the RBS analysis are summarized in Tables 2 and 3.  Assuming the bulk matrix to be composed mainly by SiO2 in agreement with the literature [7], the RBS analysis refers to ∼a 40-50 μm depth, being 100 μm the penetration depth of the used proton beam in a SiO2 matrix. Therefore, in the case of measurements performed on the external surface, the RBS signal will reasonably reflect the contribution coming from both the glaze and the ceramic body. Instead, the RBS data collected from the back-side of the samples will be informative for the elemental composition of the bulk and can be used for comparison.
The relative atomic concentrations of detected elements as obtained from the RBS analysis are summarized in Tables 2 and 3. Assuming the bulk matrix to be composed mainly by SiO 2 in agreement with the literature [7], the RBS analysis refers to~a 40-50 µm depth, being 100 µm the penetration depth of the used proton beam in a SiO 2 matrix. Therefore, in the case of measurements performed on the external surface, the RBS signal will reasonably reflect the contribution coming from both the glaze and the ceramic body. Instead, the RBS data collected from the back-side of the samples will be informative for the elemental composition of the bulk and can be used for comparison.
The relative atomic concentrations of detected elements as obtained from the RBS analysis are summarized in Tables 2 and 3. Worth of notice, the use of the SIMNRA code for the best-fit of the spectra allowed us to put in evidence three light elements, i.e., C, N, and O. The O-element is bound to the other detected elements, giving rise to oxides such as SiO 2 , Al 2 O 3 , CaO, Na 2 O, MgO, PbO 2 , and Fe 2 O 3 , BaO.
Regarding the PIXE analysis, the analyzed thicknesses are comparable to the RBS ones performed with the same ion beam probe. However, as already remarked, PIXE analysis has also allowed us to reveal elemental trace composition. Additionally, in this case, PIXE was applied to both the external glazed surface, in order to have information on the glaze/bulk composition, and to the unglazed back, in order to detect the bulk composition to be used for comparison.  Worth of notice, the use of the SIMNRA code for the best-fit of the spectra allowed us to put i evidence three light elements, i.e., C, N, and O. The O-element is bound to the other detecte elements, giving rise to oxides such as SiO2, Al2O3, CaO, Na2O, MgO, PbO2, and Fe2O3, BaO.
Regarding the PIXE analysis, the analyzed thicknesses are comparable to the RBS one performed with the same ion beam probe. However, as already remarked, PIXE analysis has als allowed us to reveal elemental trace composition. Additionally, in this case, PIXE was applied t both the external glazed surface, in order to have information on the glaze/bulk composition, and t the unglazed back, in order to detect the bulk composition to be used for comparison.      As a general result, a moderate homogeneity is observed in the elemental components of the potteries, which could suggest an elaborate elutriation, mixture, and processing in the pottery manufacture, corresponding to a high productivity level at that time.
In agreement with previous results [7], the composition of the ceramic body of all the investigated specimens indicates the use of raw materials based on alumina and silica. According to literature [31,32], alumina is considered the characteristic oxide testifying a Chinese provenance. In addition, the presence of calcium magnesium silicates (e.g., diopside, CaMgSi 2 O 6 ) as neo-formation minerals, indicating a high firing temperature, is conceivable for samples, like ours, that present such minerals in their raw materials, as occurs for ancient Chinese pottery [33].
Based on the aforementioned considerations, a Chinese provenance is hypothesized for all the investigated fragments.
Interestingly, a high content of CaO is observed for AZR3 sample with respect to the other fragments, for which the amount of these elements is almost comparable. This excess can be associated with the presence of calcite (CaCO 3 ) of secondary origin [7].
As far as external glazed surfaces are concerned, the results indicate, for samples AZR5 and AZR7, a glaze composition based on Pb-and Si-oxides [34,35], with iron-oxides responsible for the yellowish pigmented layers [35,36]. In the case of AZR3 shard, the dark coloration can be hypothesized to be due to iron oxides. The presence of such oxides, as reported in Reference [7], suggests that firing was carried out in an oxidizing atmosphere at temperature around 850 • C. The glazed surface could be of alkaline nature [37]. The glaze for AZR1 sample turns out to be Si-and Na-oxides rich, whereas Cu can be hypothesized as responsible for the blue pigment.
Finally, it is worth remarking that, contrary to PIXE, RBS has allowed, with fairly good accuracy, for the estimation of the content and the depth distribution of the metallic elements.

Ion-Microbeam Analysis
µ-PIXE elemental maps have allowed us to obtain the distribution of the major elements. Figure 6 displays the micro-beam images of some elements (Ca, Al, Fe and Si) present in the cross-section close to the external glazed surface of AZR3 sample. Mesovoids are observed, with a size ranging from some hundred of µm down to tens of µm. The sample is highly damaged on the surface, with the dark glaze present only in some points, and presumably, responsible for the red spot observed in the Fe distribution map. Si and Al are widespread in the ceramic body, whereas Ca is visible in a low amount and appeared mainly concentrated on the surface, supporting the presence of calcite of secondary origin, as previously hypothesized [7]. Ti is present only in trace. Figure 7 shows the Ca, Pb, Fe, Si, and Al µ-PIXE elemental maps of a 1 × 1 mm 2 portion from the external glazed surface of AZR5 sample. A general inspection of the maps reveals evident changes in the distribution of all elements at~200 µm below the surface, allowing us to identify two main zones that can be ascribed to the glaze (external) and the ceramic body (internal). Si appears to be the main element, widely spread in all the investigated area, with a major concentration in the ceramic body. Ca is detected in low amount, mainly concentrated on the external layer, whereas the opposite occurs for Al. Interestingly, as far as Fe is concerned, it appears concentrated on a thin layer marking the two different observed zones, furnishing an evidence of the yellowish pigmented layer underlying the glass. Below it, the distribution map of Pb evidences presence in some amount of this element down to -50 µm, and then the element tends to disappear. According to this, the existence of an engobe as an intermediate layer between glaze and ceramic body, hypothesized in previous measurements [7], is confirmed. In addition, Pb is not uniformly distributed on the glazed surface, testifying a not good conservation state. Heritage 2019, 2 FOR PEER REVIEW 2  Figure 7 shows the Ca, Pb, Fe, Si, and Al μ-PIXE elemental maps of a 1 × 1 mm 2 portion from the external glazed surface of AZR5 sample. A general inspection of the maps reveals evident changes in the distribution of all elements at ∼ 200 μm below the surface, allowing us to identify two main zones that can be ascribed to the glaze (external) and the ceramic body (internal). Si appears to be the main element, widely spread in all the investigated area, with a major concentration in the ceramic body. Ca is detected in low amount, mainly concentrated on the external layer, whereas the opposite occurs for Al. Interestingly, as far as Fe is concerned, it appears concentrated on a thin layer marking the two different observed zones, furnishing an evidence of the yellowish pigmented layer underlying the glass. Below it, the distribution map of Pb evidences presence in some amount of this element down to -50 μm, and then the element tends to disappear. According to this, the existence of an engobe as an intermediate layer between glaze and ceramic body, hypothesized in previous measurements [7], is confirmed. In addition, Pb is not uniformly distributed on the glazed surface, testifying a not good conservation state.   Figure 7 shows the Ca, Pb, Fe, Si, and Al μ-PIXE elemental maps of a 1 × 1 mm 2 portion from the external glazed surface of AZR5 sample. A general inspection of the maps reveals evident changes in the distribution of all elements at ∼ 200 μm below the surface, allowing us to identify two main zones that can be ascribed to the glaze (external) and the ceramic body (internal). Si appears to be the main element, widely spread in all the investigated area, with a major concentration in the ceramic body. Ca is detected in low amount, mainly concentrated on the external layer, whereas the opposite occurs for Al. Interestingly, as far as Fe is concerned, it appears concentrated on a thin layer marking the two different observed zones, furnishing an evidence of the yellowish pigmented layer underlying the glass. Below it, the distribution map of Pb evidences presence in some amount of this element down to -50 μm, and then the element tends to disappear. According to this, the existence of an engobe as an intermediate layer between glaze and ceramic body, hypothesized in previous measurements [7], is confirmed. In addition, Pb is not uniformly distributed on the glazed surface, testifying a not good conservation state. Worth of note appears, instead, the distribution gradients of some elements such as Fe and Pb observed in the cross-section close to the external glazed surface, reported in Figure 8. In particular, the surface layers show a significant Fe depletion, whereas Pb is detected in a very high amount.
glazed surface of ARZ5 sample. Typical scale of the colors associated to the elemental concentrations is also displayed. AZR 7 sample exhibits a uniform distribution of Si, Al, Ca and Mg-elements (data not shown).
Worth of note appears, instead, the distribution gradients of some elements such as Fe and Pb observed in the cross-section close to the external glazed surface, reported in Figure 8. In particular, the surface layers show a significant Fe depletion, whereas Pb is detected in a very high amount. As far as AZR1 sample is concerned, it is too small size avoided any cross section investigation, so the mapping was performed scanning a 1 × 1 mm 2 area on the external glazed surface, including both the zones with and without pigment.
The obtained distributions for some of the main elements (Ca, Cu, Fe, Si, Ti, and Zn) are reported in Figure 9. For all the detected elements, the distribution appears homogeneous, Si and Ca representing the main constituents of the glassy layer, Cu the main responsible of the blue coloration. TiO2 and As far as AZR1 sample is concerned, it is too small size avoided any cross section investigation, so the mapping was performed scanning a 1 × 1 mm 2 area on the external glazed surface, including both the zones with and without pigment.
The obtained distributions for some of the main elements (Ca, Cu, Fe, Si, Ti, and Zn) are reported in Figure 9.
AZR 7 sample exhibits a uniform distribution of Si, Al, Ca and Mg-elements (data not shown).
Worth of note appears, instead, the distribution gradients of some elements such as Fe and Pb observed in the cross-section close to the external glazed surface, reported in Figure 8. In particular, the surface layers show a significant Fe depletion, whereas Pb is detected in a very high amount. As far as AZR1 sample is concerned, it is too small size avoided any cross section investigation, so the mapping was performed scanning a 1 × 1 mm 2 area on the external glazed surface, including both the zones with and without pigment.
The obtained distributions for some of the main elements (Ca, Cu, Fe, Si, Ti, and Zn) are reported in Figure 9. For all the detected elements, the distribution appears homogeneous, Si and Ca representing the main constituents of the glassy layer, Cu the main responsible of the blue coloration. TiO2 and For all the detected elements, the distribution appears homogeneous, Si and Ca representing the main constituents of the glassy layer, Cu the main responsible of the blue coloration. TiO 2 and ZnO could both have been used as opacifiers to tone down the classic shades in the case, for example, of dark blue [38].

SR-Based FTIR Spectroscopy Analysis
As already reported in Materials and Methods, the examination of small powdered portions of the samples by optical microscopy allowed distinguishing among grains from the bulk and from different pigmented surfaces. In both cases, the chemical characterization was performed by SR FTIR microscopy.
In the case of grains coming from the bulk, the analysis was aimed at the identification of current and neo-formation mineralogical phases in order to achieve information on the production process, in terms of raw materials and firing conditions.
As far as samples belonging to Group 1 are concerned, Figure 10 reports the photomicrographs, in white reflected light, of AZR3, AZR5, and AZR7 samples relative to the bulk area, after being pressed within the DAC, together with their corresponding average SR-FTIR micro-spectra.

SR-Based FTIR Spectroscopy Analysis
As already reported in Materials and Methods, the examination of small powdered portions of the samples by optical microscopy allowed distinguishing among grains from the bulk and from different pigmented surfaces. In both cases, the chemical characterization was performed by SR FTIR microscopy.
In the case of grains coming from the bulk, the analysis was aimed at the identification of current and neo-formation mineralogical phases in order to achieve information on the production process, in terms of raw materials and firing conditions.
As far as samples belonging to Group 1 are concerned, Figure 10 reports the photomicrographs, in white reflected light, of AZR3, AZR5, and AZR7 samples relative to the bulk area, after being pressed within the DAC, together with their corresponding average SR-FTIR micro-spectra. Figure 10. Photomicrographs, in white reflected light, and SR-FTIR spectra (average ± standard deviation) relative to the bulk areas of sample AZR3 ((a,b), respectively), AZR5 ((c,d), respectively) and AZR7 ((e,f), respectively). Average spectra (black line) have been obtained averaging 30, 40 and 18 sampled points within the shown areas for AZR3, AZR5, and AZR7 samples, respectively. Standard deviation of the spectra is graphically represented as light grey shadowed areas.
The SR-FTIR spectra are rather similar, indicating a compositional homogeneity of these shards for what concerns the mineralogical phases. All spectra evidence a main peak at ∼1087 cm −1 , together with a shoulder at ∼1162 cm −1 , that can be ascribed to the presence of quartz (SiO2), together with another main peak at ∼1047 cm −1 , and a less intense one at ∼1633 cm −1 , and features at ∼3450 cm −1 , associated to montmorillonite ((Na,Ca)0,3(Al,Mg)2Si4O10(OH)2·n(H2O)). In addition, the characteristic band of calcite (CaCO3) at ∼1445 cm −1 is observed in the spectra of AZR3 and AZR5 samples.
The high quality of our spectra has allowed for a quantitative analysis of the observed profiles, by means of deconvolution into symmetrical Gaussian profiles and curve-fitting. This is particularly helpful in the (800-1400) cm −1 spectral range, since, as has been already demonstrated [39], the complex broad band at ~1000 cm −1 contains several components that are fingerprint of the main Figure 10. Photomicrographs, in white reflected light, and SR-FTIR spectra (average ± standard deviation) relative to the bulk areas of sample AZR3 ((a,b), respectively), AZR5 ((c,d), respectively) and AZR7 ((e,f), respectively). Average spectra (black line) have been obtained averaging 30, 40 and 18 sampled points within the shown areas for AZR3, AZR5, and AZR7 samples, respectively. Standard deviation of the spectra is graphically represented as light grey shadowed areas.
The SR-FTIR spectra are rather similar, indicating a compositional homogeneity of these shards for what concerns the mineralogical phases. All spectra evidence a main peak at~1087 cm −1 , together with a shoulder at~1162 cm −1 , that can be ascribed to the presence of quartz (SiO 2 ), together with another main peak at~1047 cm −1 , and a less intense one at~1633 cm −1 , and features at~3450 cm −1 , associated to montmorillonite ((Na,Ca) 0,3 (Al,Mg) 2 Si 4 O 10 (OH) 2 ·n (H 2 O)). In addition, the characteristic band of calcite (CaCO 3 ) at~1445 cm −1 is observed in the spectra of AZR3 and AZR5 samples.
The high quality of our spectra has allowed for a quantitative analysis of the observed profiles, by means of deconvolution into symmetrical Gaussian profiles and curve-fitting. This is particularly helpful in the (800-1400) cm −1 spectral range, since, as has been already demonstrated [39], the complex broad band at~1000 cm −1 contains several components that are fingerprint of the main mineralogical phases. The results of the best-fit procedure in the case of AZR3 and AZR7 shards are reported in Figure 11 as examples.
Concerning AZR1 fragment, belonging to Group 2, Figure 12a shows the photomicrograph of the sample, relative to the bulk area. The corresponding SR-FTIR micro-spectrum is reported in Figure 12b.
Concerning AZR1 fragment, belonging to Group 2, Figure 12a shows the photomicrograph of the sample, relative to the bulk area. The corresponding SR-FTIR micro-spectrum is reported in Figure 12b. With regard to the mineralogical phases, a first examination of the spectrum has allowed us to clearly recognize the presence of montmorillonite, as indicated by the peaks at~1034 cm −1 ,~1630 cm −1 , 3348 cm −1 , and~3414 cm −1 . In particular, the former two peaks are typically assigned to the hydration water molecules of this mineral. Looking further into the spectrum, we could postulate the presence of quartz by the peaks at~1087 cm −1 and~1164 cm −1 . Metal carboxylates (ν sym COO − at~1424 cm −1 ) and oxalates (ν sym COO − at~1312 cm −1 , respectively) peaks are also detected, that can derive from the degradation of some organic compound. About this, it is worth noting that the C-H stretching contributions that appear in the same spectrum at~2846 cm −1 and~2909 cm −1 support the presence of a proteinaceous material subjected to degradation processes [40]. Finally, the shoulder at~1448 cm −1 is associated with the presence of calcite.
Deconvolution into symmetrical Gaussian profiles and curve-fitting of the (800-1250) cm −1 spectral range was performed, and the results of the best-fit procedure are reported in Figure 13.
With regard to the mineralogical phases, a first examination of the spectrum has allowed us to clearly recognize the presence of montmorillonite, as indicated by the peaks at ~1034 cm −1 , ~1630 cm −1 , ~3348 cm −1 , and ~3414 cm −1 . In particular, the former two peaks are typically assigned to the hydration water molecules of this mineral. Looking further into the spectrum, we could postulate the presence of quartz by the peaks at ~1087 cm −1 and ~1164 cm −1 . Metal carboxylates (νsym COO − at ~1424 cm −1 ) and oxalates (νsym COO − at ∼1312 cm −1 , respectively) peaks are also detected, that can derive from the degradation of some organic compound. About this, it is worth noting that the C-H stretching contributions that appear in the same spectrum at ~2846 cm −1 and ~2909 cm −1 support the presence of a proteinaceous material subjected to degradation processes [40]. Finally, the shoulder at ∼1448 cm −1 is associated with the presence of calcite.
In Table 6, the qualitative mineralogical phase composition relative to the ceramic bulk of all the investigated shards is reported. Table 6. Mineralogical composition (relative abundances of phases were estimated on the basis of the area of the characteristic IR peaks) obtained by SR-FTIR for the ceramic body of all the investigated shards.

Mineral Quartz Oligoclase Orthoclase Diopside Montmorillonite Anorthite Calcite
Diopside and anorthite are reported to be formed at (850-950) °C through the reaction between silica and carbonate materials [41]. Hence, their presence testifies a maximum firing temperature around these values, whereas quartz and feldspars can persist above 1000 °C [42]. This hypothesis is also supported by the absence of gehlenite, that, as is well known, tends to vanish at -900 °C, as well as of high-temperature neo-formation minerals. Going on, montmorillonite could be derived from Other than the aforementioned montmorillonite, quartz and calcite, also orthoclase (KAISi 3 O 8 , bands at~1002 cm −1 ,~1118 cm −1 ), diopside (bands at~947 cm −1 ,~1056 cm −1 ,~1203 cm −1 ) and anorthite (band at~978 cm −1 ) have been identified.
In Table 6, the qualitative mineralogical phase composition relative to the ceramic bulk of all the investigated shards is reported.
Diopside and anorthite are reported to be formed at (850-950) • C through the reaction between silica and carbonate materials [41]. Hence, their presence testifies a maximum firing temperature around these values, whereas quartz and feldspars can persist above 1000 • C [42]. This hypothesis is also supported by the absence of gehlenite, that, as is well known, tends to vanish at -900 • C, as well as of high-temperature neo-formation minerals. Going on, montmorillonite could be derived from the hydrolysis process occurring during the burial period of the findings [43]. Finally, as far as the presence of calcite revealed for AZR3, AZR5, and AZR1 samples is concerned, it is well known that at temperatures above (750-800) • C calcite dissociates in calcium oxide (CaO) and carbon dioxide (CO 2 ). Its observation for these three samples, could then testify a firing process performed at low temperature, around (650-750) • C, or incomplete, probably due to an inhomogeneous heat distribution inside the kiln, that didn't permit the completion of decomposition and formation reactions. Nevertheless, decarbonation of calcite may extend to (1000-1100) • C for calcite rich systems and/or in case of coarse mineral grains and rapid heating rate [44]. Again, calcite can probably be of secondary origin, in agreement with what observed for AZR3 fragment by PIXE and ion-microbeam analyses, due to conservation circumstances such as burial conditions in humid soil, because of the gradual interactions of hygroscopic lime (CaO)-moisture and calcium hydroxide [Ca(OH) 2 ]-atmospheric CO 2 .
Hence, the whole observations suggest a firing temperature surely higher that 850 • C and in particular in the range (850-950) • C for all the investigated samples, despite the presence of calcite.
As far as the grains coming from the different pigmented surfaces are concerned, the analysis was aimed at the identification of pigmenting agents used for decoration. Figure 14 shows the photomicrograph, in white reflected light, of sample AZR3, relative to the black pigmented area, together with the corresponding average SR-FTIR spectrum.
Heritage 2019, 2 FOR PEER REVIEW 7 the hydrolysis process occurring during the burial period of the findings [43]. Finally, as far as the presence of calcite revealed for AZR3, AZR5, and AZR1 samples is concerned, it is well known that at temperatures above (750-800) °C calcite dissociates in calcium oxide (CaO) and carbon dioxide (CO2). Its observation for these three samples, could then testify a firing process performed at low temperature, around (650-750) °C, or incomplete, probably due to an inhomogeneous heat distribution inside the kiln, that didn't permit the completion of decomposition and formation reactions. Nevertheless, decarbonation of calcite may extend to (1000-1100) °C for calcite rich systems and/or in case of coarse mineral grains and rapid heating rate [44]. Again, calcite can probably be of secondary origin, in agreement with what observed for AZR3 fragment by PIXE and ion-microbeam analyses, due to conservation circumstances such as burial conditions in humid soil, because of the gradual interactions of hygroscopic lime (CaO)-moisture and calcium hydroxide [Ca(OH)2]-atmospheric CO2. Hence, the whole observations suggest a firing temperature surely higher that 850°C and in particular in the range (850-950) °C for all the investigated samples, despite the presence of calcite.
As far as the grains coming from the different pigmented surfaces are concerned, the analysis was aimed at the identification of pigmenting agents used for decoration. Figure 14 shows the photomicrograph, in white reflected light, of sample AZR3, relative to the black pigmented area, together with the corresponding average SR-FTIR spectrum. The collected average absorption spectrum reveals several peaks, that, based on comparison with database and literature [29] could be attributed to burnt umber (features at ∼1032 cm −1 , ∼1458 cm −1 , ∼3393 cm −1 , and ∼3625 cm −1 ), a deep brown color pigment obtained by heating raw umber, a natural clay pigment consisting of iron oxide, manganese oxide and aluminum oxide (Fe2O3 + MnO2 + nH2O + Si + AlO3), mixed with some organic binder that seems to be a fatty acid (features in the ∼1300 ÷ 1700 cm −1 range, at ∼2854 cm −1 and ∼2923 cm −1 ).
As far as sample AZR5 is concerned, the average SR-FTIR spectrum collected (data not shown) on the dark yellow pigmented area unfortunately only gave evidence of a vitreous matrix, hindering the identification of the pigment.
In Figure 15, we report the photomicrograph, in white reflected light, of sample AZR7, relative to the yellowish pigmented area, together with the corresponding SR-FTIR average spectrum. The collected average absorption spectrum reveals several peaks, that, based on comparison with database and literature [29] could be attributed to burnt umber (features at~1032 cm −1 ,~1458 cm −1 , 3393 cm −1 , and~3625 cm −1 ), a deep brown color pigment obtained by heating raw umber, a natural clay pigment consisting of iron oxide, manganese oxide and aluminum oxide (Fe 2 O 3 + MnO 2 + nH 2 O + Si + AlO 3 ), mixed with some organic binder that seems to be a fatty acid (features in the~1300 ÷ 1700 cm −1 range, at~2854 cm −1 and~2923 cm −1 ).
As far as sample AZR5 is concerned, the average SR-FTIR spectrum collected (data not shown) on the dark yellow pigmented area unfortunately only gave evidence of a vitreous matrix, hindering the identification of the pigment.
In Figure 15, we report the photomicrograph, in white reflected light, of sample AZR7, relative to the yellowish pigmented area, together with the corresponding SR-FTIR average spectrum. Based on the literature [45], the collected SR-FTIR average spectrum can be mainly ascribed to limonite, a yellow earth pigment known since antiquity, composed of a mixture of various iron oxide and hydroxide minerals, the most common among them being goethite, α-FeOOH and lepidocrocite, and γ-FeOOH. Figure 16a reports the photomicrograph of sample AZR1, relative to the blue pigmented area. The corresponding SR-FTIR average spectrum is reported in Figure 16b. Once again, the comparison with database and literature [29] has allowed us to identify as pigmenting agent the cuprorivaite (CaCuSi4O10) mineral (features at ∼1019 cm −1 , ∼1044 cm −1 , ∼1168 cm −1 , and ∼1207 cm −1 ). This pigment is a compound produced since the ancient Egypt time by means of a high temperature (>800 °C) synthesis starting from siliceous sand, a copper compound, calcium carbonate and sodium carbonate as a flux. It appears bounded together with some organic binders of proteinaceous origin (features in the ∼1400 ÷ 1700 cm −1 range, at ∼2846 cm −1 , ∼2926 cm −1 , and ∼3392 cm −1 ).
The photomicrograph relative to the area without pigment of sample AZR1 is shown in Figure  17a. The corresponding SR-FTIR average spectrum is reported in Figure 17b. Based on the literature [45], the collected SR-FTIR average spectrum can be mainly ascribed to limonite, a yellow earth pigment known since antiquity, composed of a mixture of various iron oxide and hydroxide minerals, the most common among them being goethite, α-FeOOH and lepidocrocite, and γ-FeOOH. Figure 16a reports the photomicrograph of sample AZR1, relative to the blue pigmented area. The corresponding SR-FTIR average spectrum is reported in Figure 16b. Based on the literature [45], the collected SR-FTIR average spectrum can be mainly ascribed to limonite, a yellow earth pigment known since antiquity, composed of a mixture of various iron oxide and hydroxide minerals, the most common among them being goethite, α-FeOOH and lepidocrocite, and γ-FeOOH. Figure 16a reports the photomicrograph of sample AZR1, relative to the blue pigmented area. The corresponding SR-FTIR average spectrum is reported in Figure 16b. Once again, the comparison with database and literature [29] has allowed us to identify as pigmenting agent the cuprorivaite (CaCuSi4O10) mineral (features at ∼1019 cm −1 , ∼1044 cm −1 , ∼1168 cm −1 , and ∼1207 cm −1 ). This pigment is a compound produced since the ancient Egypt time by means of a high temperature (>800 °C) synthesis starting from siliceous sand, a copper compound, calcium carbonate and sodium carbonate as a flux. It appears bounded together with some organic binders of proteinaceous origin (features in the ∼1400 ÷ 1700 cm −1 range, at ∼2846 cm −1 , ∼2926 cm −1 , and ∼3392 cm −1 ).
The photomicrograph relative to the area without pigment of sample AZR1 is shown in Figure  17a. The corresponding SR-FTIR average spectrum is reported in Figure 17b. Once again, the comparison with database and literature [29] has allowed us to identify as pigmenting agent the cuprorivaite (CaCuSi 4 O 10 ) mineral (features at~1019 cm −1 ,~1044 cm −1 , 1168 cm −1 , and~1207 cm −1 ). This pigment is a compound produced since the ancient Egypt time by means of a high temperature (>800 • C) synthesis starting from siliceous sand, a copper compound, calcium carbonate and sodium carbonate as a flux. It appears bounded together with some organic binders of proteinaceous origin (features in the~1400 ÷ 1700 cm −1 range, at~2846 cm −1 , 2926 cm −1 , and~3392 cm −1 ).
It is finally worth remarking that the present work is one of the first attempts to validate SR-based FTIR technique as micro-destructive powerful tool for achieving a detailed compositional characterization of potteries of historical-artistic interest. It could be potentially applied to all those times in which conventional methods fall down for several reasons.