Archaeometric studies on a Pompeian blue glass fragment from Regio I, Insula 14 for the characterization of glassmaking technology

A Pompeian glass sample found in Reg. I, Insula 14, during the 1950’s Pompeii excavation was examined by Raman and Fourier transformed infrared spectroscopy, scanning electron microscopy, and inductively coupled plasma mass spectrometry. The analyzed specimen was selected based on its intense blue color and its well-preserved aspect. The purpose of the work was the chemical characterization of Pompeii’s glass in correlation to the actual knowledge of Roman glassmaking technology from the Mediterranean area. The results suggested that the Pompeii’s glass was a soda-lime-silica glass, but with a higher calcium content that, given the low content of lead, was used to stabilize the glass. The sample was in origin produced most likely as non-decolorized primary raw materials from eastern Mediterranean sites. Moreover, the intense blue color was related to the use mainly of cobalt, present in a weighty amount, and likely used as important coloring agents in the ancient secondary glass-making workshop.

even though a low amount of these additives contributed to poor chemical durability of the glass while a higher amount rendered the glass prone to devitri cation [15]. The presence of alumina (Al 2 O 3 ) or magnesia (MgO) prevents the devitri cation of the glass [16].
Regarding Pompeii's glass manufacturers, many studies have been carried following Pompeii's excavations. Vallotto and Verità, 2002 [17], showed no great differences in the content of sodium oxide. In particular, the ratio between silicon and natron did not show great variability. Several Pompeian glasses showed similarity with the sand from the River Belus, thus suggesting the use by Pompeian workshops of primary glass row coming from the Middle East. Nevertheless, it cannot be excluded primary row glass production from other sites (perhaps the Volturno river or the provinces of Spain and Gaul) [1]. Also, hundreds of Pompeian glass nds, classi ed as 'game counters', have been identi ed. These manufacturers were also described by Pliny the Helder as products of a melting recycling procedure leading to transparent, opaque, or widely colored objects, thus representing an example of glass production activities during the Roman epoch. Several studies on these nds were mainly nalized for the identi cation of coloring or opacifying compounds [18][19][20][21]. Moreover, Pompeii's excavations revealed the presence of glass production workshops probably dated before the Vesuvius eruption in 79 A.D. [12,22].
Archaeometric methodology applied to cultural heritage is essential to obtain evidence on materials, production techniques, and habits of ancient people [23]. In particular, the multidisciplinary approach allows solving archaeometry problems regarding glass production through Roman times. The present work aims to provide a further contribution to the knowledge of the materials and execution techniques used in Roman glass-making. Therefore, using different analytical approaches, we tried to determine the chemical composition of a Pompeian's blue glass fragment that, as an archaeological nd, was quite rare and characterized by excellent durability and a good state of conservation. Besides, we tried to highlight the technology used for its production.

Materials And Methods
Light microscopy and X-ray diffractometry The specimen was observed using a Nikon Eclipse L 150 re ected light microscope.
X-ray diffractometric analyzes (XRD) were carried out using a Mini ex Rigaku X-ray diffractometer with a Cobaltum tube, operating conditions 30 kV and 15 mA.

Scanning Electron Microscope
Textural and semi-quantitative chemical analyses were performed by using a scanning electron microscope (SEM) JEOL-JSM 5310, coupled with energy dispersive X-Ray spectroscopy (EDS). The setup operated at 15 kV primary beam voltage, 50-100 mA lament current, variable spot size, 20mm WD, and 40 s net acquisition in realtime. The apparatus was equipped with an Oxford Instruments Microanalysis unit and an INCA X-act detector using Energy software with XPP matrix correction scheme and Pulse Pile-up correction. Data were processed with the INCA software, version 4.08 [24].
Back-scattered Electrons (BSE) imaging and semi-quantitative chemical analyses were performed by pressing the glass to a at surface and then coated with graphite. The sample has been placed at the same height as the cobalt standard used for routine calibration. Twenty analytical points were collected for each area and natural materials were used as standards.
Raman and FT-IR spectroscopy Raman spectra were recorded using a confocal Raman microscope (NRS-3100, Jasco Applied Sciences, Halifax, Canada). The 514 nm line of an air-cooled Ar + laser (Melles-Griot) was injected into an integrated Olympus microscope and focused to a spot diameter of approximately 2 μm (100x or 20x objective), with a laser power of 4 mW at the sample. The spectral resolution was 6 cm −1 . Raman spectra were recorded at three separate spots on each paint powder to evaluate the heterogeneity. A holographic notch lter was used to reject the excitation laser line. Raman scattering was collected by a Peltier-cooled charge-coupled device photon detector (DU401BVI, Andor Technology, Belfast, Northern Ireland). A complete data set was collected in 100 s.
A small piece of glass (about 2.5 x 2 mm) was deposited on a 3-mm ZnS window and analyzed by Fourier transformed infrared (FT-IR) spectroscopy with a Nicolet 5700 equipped with a microscope Continum (Thermo, West Palm Beach, FL. USA). Re ection spectra (200 acquisitions) were collected using the microscope focusing windows set at 50x50 mm. Spectra were analyzed by using the Omnic software. Peak assignment was further evaluated based on the data library [25].
Inductively coupled plasma mass spectrometry A fragment of the sample was grounded in a percussion apparatus used in geology. The sample resulted perfectly pulverized and preserved its blue color. The sample was then analyzed by Inductively coupled plasma mass spectrometry (ICP-MS).
In a rst experiment (ICP-MS a ), 100 mg of ground sample was subjected to mineralization in 2.5 ml of a mixture made of 1 part of HF (49% w/w) and three parts of HCl (37% w/w). Sample treatment was carried out inapposite containers in a microwave oven for 24 hours. The solubilized sample was then subjected to the ICP-MS analysis using Argon ux in an Agilent Technology 7506 apparatus (Santa Clara. California. USA).
A subsequent experiment (ICP-MS b ) was carried out on about 50 mg of pulverized glass to detect trace elements (TE) and rare earth elements (REE). The total amount of the sample (46.8 mg) was split into two sub-aliquots (25.7 and 21.1 mg) and completely solubilized by using a reaction mix made of 6 ml HCl. 2 ml HNO3. and 2 ml of HBF 4 (obtained by adding 30 g of boric acid to 100 ml of HF). Reaction vessels were placed into a microwave apparatus (Milestone Ethos-Easy supplied by FKV S.r.l.. Italy). Acid digestion was carried out according to the following three steps: 1) temperature ramp from 25 to 220 °C for 20 min at 1600 watt. 2) 220 °C for 5 min at 1600 watt. 3) temperature ramp from 220 to 25 °C for 40 min at 0 watts. The samples were then quantitatively recovered, brought to 50.0 mL with ultrapure water in disposable polypropylene falcon, and analyzed using the Nexion 2000 (PerkinElmer, Waltham, USA) inductively coupled plasma mass spectrometer (ICP-MS) equipped with a concentric nebulizer (Meinhard Associates, Golden, USA). A cyclonic spray chamber (Glass Expansion Inc., West Melbourne, Australia) and a quartz torch with a quartz injector tube (2 mm internal diameter) were used. To eliminate isobaric interferences, the kinetic energy discrimination (KED) system was used with helium (99.9999%. high purity) at 4.8 mL/min (high ow) for the determination of Fe. Cr. and V; at 3.7 mL/min (low ow) for the determination of Ni, Ca, As, Se, Co, Zn, Mn, and Cu. Standard mode (without any support gas) was employed for the determination of Li, Be, Mo, Ag, Sr, Sb, Sn, Ba, Cd, Hg, Tl, and Pb and for all the REE. A solution of Bi, Rh, Ga, and Re (approximately 100 ng/mL) was added on-line as internal standard by using a speci c seven-lined mixing valve. To state, the concentration levels of each element were carried out as a preventive semi-quantitative analysis using a multi-element standard. Quantitative determination was carried out using the "internal additions" method in the mineralized solution (previously diluted 1:1 with ultrapure water to avoid the processing of a very acidic solution) through the use of calibration curves at four levels of spiking: for Li, Be, Ag, Cd, Tl, at 0.04 -0.20 -1.0 -4.0 ng/ml; for V, Cr, Ni, Zn, As, Se, Mo, Sn, Sb, and Pb at 0.2 -1.0 -5.0 -20 ng/mL; for Ba. Sr and Co at 1.0 -5.0 -25 -100 ng/mL; for Ca, Al, Mn, Fe, and Cu at 10 -50 -250 -1000 ng/mL. For the 16 REE elements (in detail 14 REE with the addition of U and Th) was used the "internal additions" method at 4 -20 -100 -400 pg/ml. Mg and Zr were determined only by a semi-quantitative method. The correlation coe cients (R 2 ) of standard calibration curves for all the trace elements were always higher than 0.999, showing a good linear relationship throughout the selected ranges of concentrations. Four mineralization blanks were carried out (the same reaction mix. without ceramic powder) and the mean concentration was subtracted for each element. The TE and REE concentrations were evaluated as the mean of both measurements. Good repeatability (less than 10%) was obtained for all the analytes, except for Cu, probably due to the uneven distribution of Cu salts in the ceramic matrix.

Results And Discussion
Archaeological contest and glass sample characterization The glass sample analyzed, a fragment of intense blue color ( Fig. 1), was found in an amphora containing fragmented and intact glass manufactures in Pompeii's excavation, Reg. I, Insula 14, Casa 14 (Originally numbered as Reg. II, Insula 14, and successively, during 1950s excavations, changed to Reg. I, Insula 14). Insula 14 is located in the eastern area of the Regio I of Pompeii, in the median sector of the insulae gravitating on the eastern side on via di Nocera and on the northern side on via di Castricio that determines the prevailing orientation of the housing units. The rst information about insula 14 dates back to 1954, the period in which Amedeo Maiuri began an important excavation season to bring to light the entire southeastern sector of the city. In the rst phase, the investigations were limited to freeing the southern front between insulae 13 and 14, and only in 1957, the southeast corner was reached, identifying a thermopolium pertaining to the current number 15. After the excavation and consolidation of the wall hills that emerged, the research activity was interrupted and resumed in 1984 [26]. The archaeological period of the glass nds was attributable to the earthquake described by Tacitus and Seneca that seriously damaged Pompeii in A.D. 62. However, according to some views, the A.D. 62 earthquake (de ned as terminus post quem) was not a single event but other seismic activity, occurred over a certain number of years (Neronian and early Flavian periods) [27], generated stratigraphic sequence due to the subsequent demolition and rebuilding thus, suggesting that not all the glass nds have been precisely contemporaneous. All excavated specimens should have been preserved from subsequent potential deterioration caused by the eruption of Vesuvius in 79 A.D. (terminus ante quem) [21]. Dr. Piccioli, C., an o cial of the former Archaeological Superintendence of Naples and Caserta (SANC) selected the distinctive sample that was considered, according to archaeological caution, well preserved and of great signi cance, because its intense color that appeared identical to that of other intact glass manufactures. The artifact (about 2x3 cm) was carefully handled to avoid additional contamination and softly cleaned with a brush and wet bibula paper to remove dust deposits. The fragment was then stored in a preserved area to avoid further environmental deterioration. The sample did not show any opacity. The glass color, according to the Munsell notation, was organoleptically corresponding to a saturated and intense color [28]. Re ected light microscopy observation revealed some technical properties such as the absence of bubbles and the re nement in cooking which explained the durability of the material. The glass surface appeared non-homogeneous highlighting forms of yellow, white, and dark blue pitting alterations most likely attributable to the chromophoric elements responsible for the blue color (Fig. 2) whereas, no crystalline phases were observed by XRD (not shown), thus suggesting the absence of devitri cation and excellent quality of the glass also in consideration of the elapsed time.

Raman and FT-IR spectroscopy
Raman spectrum of the sample, reported in Fig. 3a, highlighted the presence of two major peaks at 1090 and 584 cm -1 with and two well-de ned components at 945 and 995 cm -1 . This signature corresponded to common limebased glass (typically having a composition with about 10 to 15% Na 2 O, and about 8 to 15% CaO). In some cases, only one shoulder was observed at 950 or 995 cm -1 [29]. The two major signatures are associated with the Si-O bending (~550 cm -1 ) components of SiO 4 entities of the more or less polymerized (Si-O)n framework, and Si-O stretching (~1090 cm -1 ) [30]. The feature at 773 cm -1 is usually assigned to the νQ 0 mode of isolated notbridged SiO 4 entities [31]. The position of the maximum of the SiO 4 bending and stretching bands are reported in Table 1. The maximum of the SiO 4 bending and stretching bands in the general database was determined from the Raman characterization of hundreds of different types of glassy silicate whose elemental compositions were determined by classical methods, thus allowing the identi cation of different types of glass. [32]. Studies made by Colomban et al., 2006, highlighted families of glasses based on the relationship between the Raman peak area ratio (A500/A1000), de ned as polymerization index (I p ), of envelopes and wavenumbers of the different Si-O stretching components. The empirical relationship between I p , glass composition, and the processing temperature was rather well documented [33]. According to this classi cation, the I p value calculated from Raman spectra collected in a different area of the sample (I p = 0.6 ± 0.05) would correspond to a family of silicate-based glasses characterized by an intermediate ratio between ux components (Na 2 O + K 2 O + CaO) with a very low content of PbO and most likely processed at medium temperature. Regarding the blue color and opaci ers, Raman features did not suggest either the detectable amount of lazurite (Na,Ca) 8 (SO 4 ,S,Cl) 2 (AlSiO 4 ) 6 [34] or Ca 2 Sb 2 O 7 (no 672 cm -1 bands) [35].
The FT-IR spectra of the sample (Fig. 3b) Table 2.
Scanning Electron Microscope BSE in SEM imaging showed non-homogeneity of the sample (Fig 4a). The semi-quantitative analysis was characterized by a great variability (high S.D. and CV%) because observations were made on different areas of the glass surface. Areas of the sample showed the presence of CoO up to a concentration of about 6.0 wt% ( Fig.  4b and 4c, whitish zones) whereas, in other areas, the CoO content ranged from zero to about 1.0 wt%. PbO ranged from zero to about 6.0 wt%. Ti and Fe were also detected as well as minerals attributable to the group of zeolites likely formed by the alteration of glass [41]. Table 3 shows the average composition of the glass. The sample appeared as a soda-lime-silica glass with the average concentration of SiO 2 , Na 2 O, and CaO of 61.71 wt%, 1.44 wt%, and 5.16 wt% respectively, although the average Na 2 O concentration resulted lower compared to that reported in the literature data on glasses of the period [42] whereas, a higher average concentration of MnO, FeO, CoO, and PbO was observed. The latter data suggested that cobalt was most likely the key chromophoric element responsible for the sample blue color. [43][44][45][46][47][48].
Inductively coupled plasma mass spectrometry Because ICP-MS measurements of silicon at m/z 28 suffer from numerous spectral interferences that could include C, O, and N (the latter most likely coming from nitric acid), we rst analyzed the presence of trace elements (TE) in the Pompeian glass powder after its digestion in HF and HCl in a 1:3 ratio (ICP-MS a ). We expected a Si concentration above detection limits thus, it was not necessary to pre-concentrating the sample.
Silicon does not require signi cant amounts of strong acid and low levels (< ~10 ppm) are soluble and stable in water. Moreover, the ICP-MS used allowed Si measurements in the range of a few ppm. The ICP mass analytical results are summarized in Table 4. In particular, the sample showed a relatively high content of Si, Na, and Ca and a lower content of Fe, Al, Co, Mn, Cu, Sb, Pb, and K. The amount of these elements, converted in the corresponding oxides, highlighted a composition similar to that observed in several Roman glasses [20]. In fact, the percentage (w%) of SiO 2 , Na 2 O, CaO, Al 2 O 3 , K 2 O, MgO, FeO, MnO, PbO, and CuO were 43.8, 5.6, 5.4, 1.0, 0.66, 0.28, 0.7, 0.24, 0.09, and 0.08, respectively. The MgO and K 2 O compositions were less than 1.5%. This data suggested that natron was the primary alkali ux for this glass [49][50][51]. It is also worth noting that the amount of Sb observed (0.99 w%) was not su cient as an antimony-based opaci er as instead observed for other Roman and Pompeian glasses [42,52].
ICP-MS was used was also performed after-treatment of the sample with HCl, HNO 3 , and HBF 4 at a 2:1:1 ratio (v/v) in a temperature ramp (ICP-MS b ). The results are reported in Table 4. Although with this procedure it was not possible to detect Si, K, and Na, also here was observed a moderate-high content of Fe, Al, Ca, Co, Mn, Cu, and Mg. The weight percentage (w%) of the corresponding oxides FeO, Al 2 O 3 , CaO, CoO, MnO, CuO, and MgO were 0.79, 2.74, 8.06, 0.20, 0.52, 0.68, and 0.65, respectively [53]. The CaO content evaluated by ICP-MS a and ICP-MS b (5.4 w% and 8.05 w%, respectively) was slightly higher compared to that reported for Pompeian glasses by [42] (average 7.215%) and according to the literature data on a glass of the period [43][44][45][46][47][48]. This value could be related to the percentage of sodium present in the natron to ux the silica [6, 54] and the higher quantity of lime was used to stabilize the glass thus, suggesting that the sample represented a specialized production, perhaps using a plant-ash component, during the 1st century A.D.
The relationship between the composition of Al2O3 (1.0 w% by ICP-MSa and 2.74 w% by ICP-MSb) and that of CaO showed to be very close to that reported for Pompeii glasses within the area of Roman Western European sites and in the Mediterranean area in the 1st -3rd century A.D. [55]. As suggested, these values could be due to the employment for the glass productions of similar raw materials along with the Empire and most likely from the Middle-East region [39,44,[56][57][58].
The relatively high content of FeO (0.7 w% by ICP-MS a and 0.79 wt% by ICP-MS b ) as already reported by [42] can be found in blue-colored Roman glasses. However, its amount might be depending upon the MnO concentration.
In this case, the manganese oxide concentration (0.24 w% by ICP-MS a and 0.52 wt% by ICP-MS b ) was within natural limits, thus suggesting the use of iron-containing raw material that was most likely not subjected to the decoloring procedure. Decolorized glasses generally show MnO concentration > of 0.5 wt% probably due to the addition of manganese as pyrolusite (MnO2). The latter was particularly widespread in the Roman period to neutralize the color due to the iron oxides naturally present in the primary raw materials [44,59,60,61]. These ndings support the hypothesis that the Pompeian glass could have been produced from the sands from the Middle-East region [62].
Also, copper and cobalt, contained in the sample at a concentration of 5467 ppm and 1615 ppm, respectively (0.20 w% and 0.68 w% by ICP-MS) were important coloring agents in the ancient glass-making workshop [18]. For instance, copper might produce blue color depending on its interaction with iron and on some level with manganese and lead. However, deep blue glass showed signi cant amounts of copper and cobalt in the order of 1930 ppm and 1453 ppm [63]. Therefore, the deep blue color of the sample, besides the iron present in the raw material, might be due to the presence of copper and cobalt probably added as 2Co 2 This compound was often used for the production of Roman blue glass [63]. Therefore, as also clearly shown by BSE imaging (Fig. 4b and 4c), the blue color of the glass sample was essentially due to cobalt probably used in Pompeian secondary furnaces for glasses manufacture production [42].
ICP-MS b was also used to determine the content of rare earth elements (REE). The results, reported in Table 4, showed that neodymium (Nd) was the most preponderant rare earth element present in the sample with a concentration of 13.629 ppm. This element belongs to the light rare earth elements (LREE) of the lanthanide series and its concentration is in the range of the concentration of Nd in silica-based, non-carbonaceous sediments and sedimentary rocks that generally is in the order of 5-50 ppm [64]. Neodymium content in glass components such as shell and limestone as well as natron is much lower (around 0.5-10 ppm, and 20-40 ppb, respectively [64-66]. These ndings suggest that Roman glasses were originated from heavy minerals or a fraction of non-quartz minerals of the silica-based raw material [46]. Under this aspect, sands from the Campanian beaches by the Garigliano and Volturno Rivers were likely not used in this case [67] since this area all contained more Nd, even up to 296 ppm [68]. Moreover, these sands contained high percentages of heavy minerals, resulting in high Fe 2 O 3 and Al 2 O 3 levels, making them unsuitable for glass production [46, 69].

Conclusions
In this study, by applying a combination of Raman and FTIR spectroscopy, SEM, and ICP-MS, the chemical composition of a Pompeian's glass blue fragment has been determined, thus representing a possible contribution to the archaeological knowledge on Pompeian's glass manufactory. The sample appeared as a re ned glass and was most likely obtained from secondary raw materials. By evaluating the composition of the sample, we tried to de ne possible areas where suitable sand raw materials would have been available.
The glass sample analyzed was a soda-lime-silica glass containing a slightly higher CaO content most likely used to stabilize the glass. Moreover, as suggested by the Ip value, the sample was a lower lead-based silicate most likely processed at medium temperature. Furthermore, the amount of Al2O3 and CaO suggested the employment of similar raw materials along with the Empire and most likely that from the Middle-East region. The Nd content (ppm) of the blue sample, excluded the use of the sand of Campanian beaches for primary raw material.
The FeO content was within natural limits and closer for other Roman glasses thus indicating the use of ironcontaining raw material that was not subjected to the decoloring procedure.
The deep blue color was most likely due to the cobalt, present in a substantial amount, and possibly used as an important coloring agent in the secondary glass-making workshop.
These results suggested the presence of the primary glass production industry and a possible Pompeian secondary workshop for the production of glass manufactures during the rst century AD.

Declarations
Availability of data and materials The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.