First, the outcomes achieved by the single techniques are presented; then, a full overview of the analytical investigation has been proposed in the “Discussion” paragraph. Due to their complementarities and vibrational spectroscopy nature, Raman and FTIR are presented in the same paragraph.

X-ray diffraction

X-ray diffraction patterns have been analyzed both qualitatively, for the identification of the crystalline phases, and semi-quantitatively for estimating their relative ratio percentages. The latter analysis has been carried out assigning hundred per cent value to the intensity of the most intense peak, and calculating the relative intensity percentage of the other peaks. The results have been plotted in the ternary diagram (Fig. 3), where a selection of the most significant crystalline phases is proposed on the basis of the following considerations: a noticeable number of samples show a high content of quartz (SiO2); phyllosilicates (talc—Mg3Si4O10(OH)2, phlogopite—KMg3(Si3Al)O10(F,OH)2 and sericite KAl2(AlSi3O10)(OH)2 clay minerals) are a further representative group of minerals; the non-silicatic group exhibits relevant amount of carbonates (calcite CaCO3, dolomite CaMg(CO3)2, magnesite (MgCO3)) and sulphates (gypsum CaSO4·2(H2O)). For a sake of completeness, other minor amount of silicates as feldspars (orthoclase KAISi3O8, albite NaAlSi3O8, plagioclase (Na,Ca)(Si,Al)4O8) and mullite (Al6Si2O13) are added to the quartz values.

Fig. 3
figure 3

Ternary diagram representing XRD data. The sample labels are indicated close to the dots

As shown in Fig. 3, the main outcome is the presence of a major cluster of samples (1–4, 6–7, 9, 11) with a high amount of one or more minerals included in the quartz-feldspars-mullite group; considering also that these samples show the presence of phyllosilicates (Fig. 4 and Additional file 1: Table S2), it is reasonable to ascribe this cluster to terracotta materials; the presence of hematite (Fe2O3) in two samples further supports that. Actually, a deeper look allows noticing that sample 1 differs from the others because mullite is present (Fig. 4 and Additional file 1: Fig. S2). This mineral is meaningful for exploring the terracotta firing process; it is well known [9] that sericite-illite clays used for producing terracotta wares undergo mineralogical changes at specific heating temperatures. First, clays lose “free moisture”, mechanically retained or adsorbed water, in the range between 50 and 200 °C. At higher temperature, in the range 400–800 °C, a dehydroxylation of the clay minerals occurs, due to the water elimination of structural OH groups of the layered structures. Sericite clay minerals starts changing into the dehydroxylated phase [KAl2(AlSiO3)O11] at about 500 °C; however, many factors influence this process and literature reports different temperature ranges [10, 11], indicating that the maximum range is 850–1000 °C. The increase of firing temperature induces a breakdown of the dehydroxylated sericite-illite phase, with a consequent formation of mullite at about 1050 °C. Based on this thermal behavior, the firing temperature of the above samples can be inferred: tassels number 3, 4, 6, 7, 9 and 11, showing a relevant amount of sericites and phlogopite, have been fired below 900–1000 °C whereas sample 1, which exhibits mullite, over 1050 °C; however, the absence of quartz polymorphs in sample 1 points out that it does not exceed 1200 °C [12, 13].

Fig. 4
figure 4

XRD spectra of selected samples, representative of the entire sample set. Q: quartz, F: feldspars, Mu: mullite, S: sericites, C: calcite, D: dolomite, G: gypsum, Ma: magnesite, T: talc

It is worth highlighting that samples seem to derive from the same clay paste made by a non-calcareous clay, as demonstrated by the lack of the firing transformation minerals of calcareous clays as gehlenite and diopside, and the presence of mullite; moreover, the same source is also confirmed by the presence of sericite-illite clay minerals in all samples (except for sample 1 that has been fired at higher temperature, with a consequent collapse of sericite-illite crystalline structure). Based on these results, the low quantity of calcite detected in some samples of this terracotta cluster (4, 6, and 9) cannot be attributed to the terracotta original composition, but it could point out to burial conditions to which the horse could have been subjected; during the burial time, a secondary crystallization of calcite may have occurred. A further explanation of calcite source can be found in the residual of applied finishing layers or of conservation products made with an inorganic matrix.

A distinction is made for samples 5, 8 and 10 which deserve independent discussion due to their particular position in the ternary diagram. XRD spectrum of samples 5 exhibits high background and significant degree of noise, indicating the presence of amorphous compounds which cannot be identified with this technique. Over this background, the main characteristic peaks of talc, magnesite and dolomite can be distinguished (Fig. 4 and Additional file 1: Fig. S2). No reasonable hypothesis can be formulated about the nature of this tassel based on XRD data only.

Sample 8 comprises a high relative content of carbonates and sulphates, despite the main mineral is quartz, and little amount of sericite clay minerals (Fig. 4 and Additional file 1: Fig. S2). The main mineral of sample 10 is calcite; dolomite, quartz and sericites are also detected (Fig. 4 and Additional file 1: Fig. S2). Looking at the mineralogical composition of these three tassels (samples 5, 8 and 10), it is apparent that they cannot be included in the terracotta group; as for 8 and 10, their relative high amount of carbonates (and sulphates in sample 8) points out a stucco material, prepared with a carbonate/sulphate matrix and silicates as aggregate.

Nitratine (NaNO3), detected in five samples, attests a certain degree of decay, which is further studied through the following analyses, in particular with ionic chromatography.

Vibrational Spectroscopy: Raman and FTIR

Raman spectra acquired on the same powders confirmed XRD outcomes and provided additional data. Due to its high Raman scattering cross section, anatase (TiO2) has been detected in almost all the samples (Fig. 5a, b), whereas in XRD spectra just a very weak, and in some cases negligible peak, is present. In art objects, anatase could be ascribed to two different sources: it can be associated to silicates or oxides as accessory mineral, or it indicates the presence of a finishing layer made by titanium white (rutile and/or anatase), a white pigment widely used since nearly 1920.

Fig 5
figure 5

Selected and representative Raman spectra where a hematite (223, 291, 401 cm−1) and anatase (149 cm−1), b anatase (149, 392, 511, 632 cm−1), c anhydrite (1017 cm−1), d anhydrite (1017 cm−1) and gypsum (1009 cm−1) were detected. c, d spectra are acquired in two different areas of the sample 8 powder

One of the main advantages of micro-Raman includes the possibility to observe the powdered samples with a microscope objective and accurately select the grains to be analyzed, correlating their color to a specific composition. This is the reason why hematite (Fe2O3) has been frequently identified by Raman (Fig. 5a), especially in terracotta samples, thanks to its red color. Only in some XRD spectra a weak hematite peak is visible, confirming that micro-Raman is the technique of choice when colored particles are present in a mixture in low quantity. Hematite is a frequent mineral in terracotta materials, demonstrating the transformation of iron phases in oxidizing firing conditions. It provides the typical red color to the matrix, and tiny amount is enough to give this coloration. However, the employment of red ochre (hematite) in finishing layers cannot be completely excluded.

XRD results revealed that sample 8 has the highest relative amount of gypsum. Micro-Raman analysis of this sample allowed noticing the presence of anhydrite (CaSO4) in mixture with gypsum (Fig. 5c, d). Anhydrite is the anhydrous calcium sulphate form, not easily detectable by XRD due to severe peaks overlapping. Its presence could demonstrate a decay of the stucco as gypsum loses its water molecules at temperatures above 30 °C-40°C and moderate R.H. (30–40%), with a consequent crystallization into anhydrite [14]. Alternatively, anhydrite could indicate a calcination of the gypsum rock at high temperatures; this process is aimed at increasing the mechanical characteristics of the stucco, as anhydrite needs less kneading water [15].

The contribution of the characteristic peaks of the main minerals present in the Raman spectra of the samples has been plotted in a ternary diagram to perform a semi-quantitative analysis of the data. The ternary diagram is based on the Raman bands for quartz at 464 cm−1, gypsum at 1009 cm−1 and calcite at 1087 cm−1. The reported values represent the integral under the curve of the peak and was normalized to the sum of all three peaks, or spectral region in the absence of the peak. The peak for anhydrite (1017 cm−1) was also added to the gypsum peak. As it can be observed in Fig. 6, sample 1 (the reference from the body of the horse) showed the highest quartz concentration as expected and as seen in XRD data. Most of the tassels (samples 2–4, 6–9, 11) clustered at high concentration of quartz, while samples 5, 8 and 10 were outliers. Sample 10 had a high value for the calcite axis and sample 5 and 8 instead had a high value for the gypsum and anhydrite axis.

Fig. 6
figure 6

Ternary diagram representing Raman data. The sample labels are indicated close to the dots

Information about organic compounds have been achieved by Raman in sample 5. As mentioned above, XRD spectrum suggested the presence of an amorphous compound; some characteristic bands of organic compounds have been observed in the Raman spectrum, as reported in Fig. 7. An interesting result has been achieved coupling Raman and FTIR data: besides talc (3677 cm−1, 1019 cm−1 and 670 cm−1 in FTIR spectrum) and low amount of dolomite, a synthetic polymer is unequivocally identified through the bands at 1726 cm−1, 1494 cm−1, 1271 cm−1, 1122 cm−1 in the FITR spectrum and 1000 cm−1, 1045 cm−1, 1605 cm−1 and 1732 cm−1 in the Raman spectrum, suggesting the presence of a polyester resin and/or a phthalate.

Fig. 7
figure 7

Raman and FTIR spectra acquired on sample 5. FTIR spectrum shows the vibrational pattern of both talc (light blue wavenumbers) and polyester resin and/or a phthalate (dark blue wavenumbers). Raman spectrum shows the band of polyester resin and/or a phthalate

FTIR analyses have been focused on the identification of organic components within the sample powders, through the extractions with dimethyl ketone. In Fig. 8 four representative FTIR spectra of compounds detected in the samples after the solvent extraction are reported; cellulose nitrate (CN) is clearly identified by the characteristic vibrations at 1656 cm−1, at 1280 cm−1 and ~ 840 cm−1 of nitrate groups (νasNO2, νasNO2 and νO-NO2, respectively; FTIR spectrum of sample 8 in Fig. 8). The cellulosic vibrational envelope between 1220–840 cm−1, with the strong COC stretching vibration at 1077 cm−1, confirms the detection of cellulose nitrate in several samples (2,-4,6–8,10–11) [16, 17].

Fig. 8
figure 8

FTIR spectra collected in transmission on diamond anvil cell of the solid residue obtained after extraction with dimethyl ketone of samples 8, 2, 3 and 6

Moreover, other organic substances are in mixture with cellulose nitrate; in sample 6, an acrylic resin is inferred by the strong band at 1730 cm−1 (νC = Oester), by the vibrational modes at 2959 cm−1, 2926 cm−1, 2874 cm−1, 2854 cm−1 (νCH3 and νCH2) as well as by the characteristic bands in the fingerprint region at 1453 cm−1, 1389 cm−1 (in-plane δas and δsy of C–H), at 1240 cm−1, 1027 cm−1 (νC–Oester), 1151 cm−1 (out-of-plane δ C–H) and 749 cm−1rocking–CH2–) [18].

In samples 1, 3, 4, 7, 10 and 11 a weak absorbance band at 1730 cm−1 is ascribable to an acrylic resin in very low amount. On the other side, this peak could also be attributed to camphor, a plasticizer commonly used to stabilize CN, albeit the C = O stretching vibration of camphor is in general at higher wavenumbers [16]. Therefore, in samples 1 and 9, where cellulose nitrate has not been detected, this band has to be ascribed to an acrylic resin; in the other samples the presence of an acrylic resin in mixture with cellulose nitrate cannot be excluded.

Proteins, in very low amount, could be present in samples 3, 4, 7 and 10 as suggested by the absorption bands at 3344 cm−1, 3077 cm−1, 1656 cm−1 and 1537 cm−1 attributed to the νN-H, amide I and amide II vibrational modes [19, 20]. A further organic substance, having a peak at 1713 cm−1 is identified in samples 2, 8 and 11 (see FTIR spectrum of sample 2 in Fig. 8). This absorption band is due to the C=O stretching and is compatible with an organic acid, most likely released by degraded organic polymers in the samples. In addition, the presence of nitrates (837, 1371 cm−1) and OH stretching broad bands (between 3650–3200 cm−1) after the solvent extraction of all the samples containing cellulose nitrate can be correlated to the decay of CN via hydrolysis [21, 22]. In all the spectra of extracted samples, the strong bands at 2959, 2926, 2874, 2854 cm−1 are due to the symmetric and asymmetric stretching of CH2 and CH3 groups. Considering the sharp shape of these bands and their position, they could be attributed to an aliphatic compound, most likely a paraffin or a wax.

The inorganic phases identified by FTIR perfectly fit with XRD data, identifying a main group composed by samples 1, 2, 4, 6, 7, 11, dominated by the vibrational modes originating from quartz and other hydrated silicates (S2) and three samples (10, 8 and 5) showing different features. Sample 10 is significantly different since its FTIR spectrum mainly shows the vibrational bands of calcite; moreover, nitratine NaNO3 (1385 cm−1, 837 cm−1) is unequivocally detected. According with XRD data, gypsum is one of the most relevant phases in sample 8, in mixture with quartz.

Ionic chromatography

The samples showed a high variability of the solubility in water. The mass fraction of water-soluble ions (WSI), with respect to the total mass of sample, ranges from 1.0% (sample 1) to 34% (sample 8).

The ionic composition (sum of anions and cations) as percentage of each sample, is shown in Fig. 9.

Fig. 9
figure 9

Percentage distribution of WSI in the samples (the columns show colours as reported in the legend on the right). The percentage of WSI with respect to the total mass is reported on top of the figure

The prevailing anion was SO42 − followed by NO3. On average, SO42− and NO3accounted for 60% and 2.9%, respectively, of the total ions mass. The prevailing cation was Ca2+ followed by K+. On average, Ca2+ and K+ accounted for 29% and 1.0%, respectively, of total ions mass.

Good correlations (R2 = 0.844) between anion and cation µEquivalents (µEq) (Additional file 1: Fig. S3) were found for all samples except for the samples 1, 3 and 10; deficiency of cations in sample 3, deficiency of anions in sample 1 and 10. In Additional file 1: Fig. S4 a strong correlation (R2 = 0.998) is obtained removing samples 1, 3 and 10.

The scatter plots of SO42− versus Ca2+, SO42− versus CO32−, NO3versus Cl and NO3 versus Na+ are depicted in Fig. 10. Strong correlation between SO42− and Ca2+ was found (R2 = 0.974) (Fig. 10a) suggesting the presence of CaSO4 (most likely gypsum) in all samples, with the exception of sample 10. The high sensitivity of ionic chromatography is demonstrated by the detection of calcium sulphate even in samples where XRD does not show any evidence.

Fig. 10
figure 10

Scatter plots of different ions: a SO42− vs Ca2+ b SO42− vs CO32−; c NO3 vs Cl and d NO3 vs Na+

The relation of sulphate and carbonate shows that in some samples (green circle in Fig. 10b) high value of sulphate corresponds to low value of carbonate. This could be explained considering the degradation of calcite and the consequent formation of gypsum [23], despite here any analytical evidence has been found for supporting this hypothesis.

A good correlation was observed (Fig. 10c) for nitrate and chloride (R2 = 0.858), whereas weak correlation was observed for nitrate and sodium (Fig. 10d). In two samples (sample 3 and 10) the ratio NO3/Na+ was higher (11 and 8.6, respectively) with respect to other samples (average ratio 1.5). Sample 3 is the only one with an excess of anions; however, in this sample sulphate was almost totally neutralized by calcium (strong correlation sulphate vs calcium), and nitrate was 11% by weight; considering that nitrate cellulose (CN) was observed in almost all samples by FTIR, the hydrolysis and subsequently degradation of CN [16, 24] could be suggested with a consequent enhancement of solubility of nitrate.

Anion deficiency was observed in samples 1 and 10 which is more likely due to the fact that HCO3 and organic acid are not detected by IC; moreover, a release of HNO2, HNO3, and CH3COOH as a consequence of degradation of CN [16] cannot be excluded.

The presence of oxalate in all samples could indicate the mineralization of organic compounds as well as the CN nitrate scission [16].

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit


This article is autogenerated using RSS feeds and has not been created or edited by OA JF.

Click here for Source link (