The archaeological background and research objectives

The glass vessels selected for archaeometric study come from two recently investigated archaeological contexts in Rome (Fig. 1).

Fig. 1
figure 1

Map of Rome (detailed view) with the location of the two archaeological contexts investigated

Full size image

The former (PNE) is the complex known as “Baths of Helagabalus”, situated on the north-east slope of the Palatine Hill (Saguì 2013; Saguì et al. 2014; Saguì and Cante 2015). The site has been occupied from the early Iron Age to the Middle Ages. During the Imperial age, it was characterized by the construction of two large buildings, which follow one another in a short interval of time. Dated to the Hadrianic period (117–138 AD), the first one comprises a series of utilitarian rooms (horreum), overlooking the road connecting the Colosseum with the Forum. The Hadrianic complex was completely destroyed in the Severan period (end of the second-early third century AD), in order to build a new and different structure, with rooms arranged around a central rectangular courtyard. The glass samples (PNE 1-10, 12; 20-21) belong to forms dating to the last decades of the second century AD and probably used within the Hadrianic horreum (Lepri 2013).

The second archaeological context is located in the southern sector of the Esquiline hill, in the area corresponding to the ancient Horti Lamiani and the present Piazza Vittorio (PV). The site was named after L. Aelius Lamia, a consul living under the reign of the emperor Augustus, responsible for the creation of these gardens. The area became part of the imperial ownership probably under Tiberius and certainly from the reign of Caligula, who built his urban villa there. The horti instead became a private property of the Emperor possibly under Alexander Severus (222–235 AD) (Cima and Talamo 2008; Barbera 2013; Alagia 2014).

The most recent investigations carried out in this area under the direction of Dr. Serlorenzi (Soprintendenza Speciale per i Beni Archeologici di Roma) yielded an important archaeological sequence from the late Republican period to the twentieth century AD. The main discovery is represented by a new sector belonging to this complex, centred around a large hall of the Middle Imperial age (Barbera et al. 2010). The glass analysed here (PV 13-19, 22) comes from late second century AD deposits, found in an open area corresponding to the archaeological sector named “Aula Cortile Grande” and related to the building activities of the hall.

Based on current knowledge of colourless glass, this research is aimed at characterising glass vessels with two main objectives: (1) the material characterisation and (2) the provenance assessment. In the first case, the nature of network formers, fluxes and stabilising agents is a matter for investigation. The identification of decolouring agents is further examined, in order to understand whether it was their addition or the use of a pure white sand which affected the final aspect of glass vessels. These results will be of further help in provenance determination, the latter being addressed by comparing the bulk chemical compositions of the sample set to the known reference groups on colourless glasses.

Materials

A set of 21 glass vessels (Table 1; Figs. 2 and 3) was selected for archaeometric investigation. The morphological types are known from other contexts of Rome and Ostia dating to the Middle Imperial age, a period in which most vessels seem to be made of colourless glass.

Table 1 Sample list
Full size table
Fig. 2
figure 2

Drawings of the specimens analysed

Full size image
Fig. 3
figure 3

Drawings of the specimens analysed

Full size image

The first group of fragments is representative of the most frequent types in both contexts: cups and bowls of common use sharing many technical features. All colourless (PNE 1-4, 6, 8-10, 12; PV 13-17, 19) and aqua-blue (PNE 5, 7; PV 18) specimens are free-blown glass with very thin walls; the rim is fire-worked, often with a rough upper part of the outer surface perhaps caused by a wheel, and triangular-shaped in some specimens more or less sharp-edged. In Rome and Ostia, the presence of these forms from the late second to the first half of the third century AD seems highly significant (Lepri 2013, pp. 142–143). For instance, many similar and coeval vessels unearthed at the archaeological site of Vigna Barberini (Palatine Hill, very close to the Baths of Helagabalus) were recently published (Foy 2014, in particular p. 21 f.fol.).

These types are also very common in some Late Antonine (160–190 AD) contexts in Ostia (the north-east area of the Terme del Nuotatore), whose glass findings still remain unpublished. They are characterised by several dimensional variants, which are consistent with the existence of sets made of these types of vessels.

Their presence in Gallia Narbonensis (Foy 2014, ibid.) makes the provenance issue even more complex, although the important role played by the Italic workshops is beyond any doubt.

In this regard, further data will be provided by the study of glass collections from Vigna Barberini and from Gallia Narbonensis, which are currently being investigated by Foy (2014, p. 13, note 1 and p. 24), within the broader project on the first to fourth century AD Roman colourless glass.

The second group of samples investigated here includes fragments of colourless goblets type Isings 86, characterised by a weathered milky surface (PNE 20-21; PV 22). In particular, two of them are vessels decorated by circular-shaped appliqués, made by stamping the decoration on a drop of ductile glass (i.e. not yet cooled). One appliqué (PNE 20) depicts a male bust in profile and a similar one, with a bearded male profile, was found in the same context. These items probably belonged to the same vessel (Lepri 2013, pp. 144–148 and p. 159, fig. 5.4-5). The third goblet (PV 22) is decorated with rounded-shaped appliqués, reproducing a shell in negative.

Both decorations are typical of vessels dating from the late second to the third centuries AD. The former decoration is generally considered to be derived from or inspired by coins, in fact, the presence of male and female profiles, surrounded by a crown of small points or by a raised edge, sometimes also accompanied by letters, recalls the representation of several members of the Antonine’s and Severans’ dynasties on coins of that period. Specimens with this kind of decoration are known from the Italian peninsula, from Gallia, Germania Inferior and from collections, although the majority of finds seems to be concentrated in Rome.

Also, the shell decoration belongs to the same repertoire and shows morphological and technical variants. The appliqués of the investigated goblet can be directly referred to those decorating a colourless flask, placed in a second century AD burial, within the Roman cemetery of Noli (Savona, Italy) (Pastorino 2007).

Since colourless glass is generally considered to be used for high-quality tableware, the present work is also an attempt to investigate the possible differences between decorated and plain ware with respect to their chemical composition, among vessels made by the same technique of the free blowing.

Experimental

Scanning electron microscopy (SEM-EDS)

Commonly used for imaging samples, scanning electron microscopy (SEM) was used for textural observations, mainly performed in backscattered electrons. The instrument was a Philips XL 30 SEM equipped with an EDAX-DX4 energy dispersive spectrometer (EDS) operating at 20 kV. For chemical microanalyses, a variety of natural phases (albite, almandine, biotite, chlorite, Cr-diopside, diopside, kadeite, kaersutite, olivine, plagioclase, pyrope, rhodonite, sanidine) and synthetic glass materials (NIST 1831) was used as primary and quality control standards. For SEM observations, a small fragment of glass was cut, mounted in resin, polished and carbon coated.

Electron micro probe analysis (EMPA)

The major and minor element concentrations were estimated in all samples by electron microprobe analysis (CNR, Firenze, Italy). The instrument was a JEOL Superprobe JXA-8600 (University of Florence, working with the following operative settings: 15 kV, beam current at 15 μA, beam diameter 1–5 μm. A variety of natural phases (bustamite, stibnite, cuprite, galena, metallic cobalt and Sn) and synthetic glass materials (NIST 1831) was used as primary and quality control standards. PAP software was used for correction. Precision was within 1 % for major elements, about 3–4 % for minor elements and about 8 % for trace elements. Accuracy was ≤ 1 % for SiO2, Na2O, CaO, MgO, K2O, and MnO; equal to 2.5 % for Al2O3; 3.4 % for Fe2O3; below 15 % for Cu, Pb and Sb. Generally 5–7-point microanalyses were performed per sample, and mean and standard deviations were calculated.

Laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS)

The trace element content of glasses was determined by laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS at CNR, Pavia, Italy). The instrument combines an ablation microbeam based on a Nd/YAG laser source (Brilliant, Quantel) operating at 266 nm (for details, see Tiepolo et al. 2003), and a quadrupole ICP-MS (Drc-e, Perkin Elmer). Thirty-four masses from 7Li to 238U were acquired; the laser was operated at 10 Hz of repetition rate, the power on the sample was 1.5 mW and spot size was set at 40 μm. Accuracy was assessed on the USGS BCR-2 reference glass (analysed as an unknown in each analytical run) and was better than 20 % at the sub ppm level. Data reduction was carried out with the software package GLITTER (van Achterbergh et al. 2001) and using NIST SRM 610 and 29Si as external and internal standards, respectively. The obtained values were normalised against the average composition of the weathered upper continental crust (Kamber et al. 2005).

Results

From a textural point of view, all PNE and PV samples are exceptionally homogeneous, never showing relics or newly formed phases. Compositional bands are absent as well. The external surfaces sometimes show a thin alteration layer, mainly characterised by the loss of Na2O and K2O and the increase of SiO2 and Al2O3 with respect to the bulk composition.

The major and minor element concentrations ascribe the investigated samples to the soda-lime-silica glass type (Table 2). The contents of the oxides are rather similar among colourless and aqua blue glass; the only exception being represented by SiO2, whose amounts vary between 68 and 78 wt%. SiO2 and Al2O3 contents (ranging between 2.4 and 3.1 wt%) suggest the use of a siliceous sand as the network former. CaO ranges between 5.1 and 6.3 wt%, being the major stabilizing agent. The MgO and K2O contents are rather low, sometimes even lower than those generally indicated for the use of natron as a flux. In any case, the use of natron is ascertained for both colourless and aqua samples. Excluding sample no. 18, the Fe2O3 contents range between 0.38 and 0.85 wt%, indicating thereby that this component has not been deliberately added. In colourless glass, MnO amounts are always lower than those of Fe2O3 and range between 0.05 and 0.47 wt%, being below 0.2 wt% in 4 samples only. Contrary to the low amounts of Fe2O3 and MnO, Sb2O3 contents are relatively high in all samples, ranging from 0.21 to 0.37 wt% (i.e. from 1754 to 3091 ppm).

Table 2 Major and minor element content (wt%) in colourless and aqua blue samples, determined by EMPA
Full size table

Minor and trace element contents (Table 3) are overall low in both colourless and aqua blue samples, except for Pb, which is generally around or above 100 ppm (except for samples 20–22). The strong positive correlation between Hf and Zr is rather constant around a 0.025 average value. Similarly, the contents of Rb are correlated to those of K (K/Rb ratio = average value of 0.044). The geochemical behavior of Y and that of the heavy REEs (Gd, Tb, Dy, Ho, Er, Tm, Yb and Lu) is very similar and averagely opposed to that of light REEs. These features are particularly evident in the aqua blue sample no. 7, which shows the highest values of Be, B, Cr, Zn, Rb, Zr, Nb, Cs, Ba, Hf and Ta, and the lowest values of La, Ce, Pr, Nd, Gd, Tb, Dy, Ho, Er, Yb and U. Lastly, the colourless sample no. 20 may be distinguished from the rest of the group, showing the highest values of Tb and the lowest values of Li, V, Co, Ni, Rb, Sr, Zr, Nb, Cs, Ba, Eu and Th.

Table 3 Trace element contents (ppm) in colourless and aqua blue samples, determined by LA-ICP-MS
Full size table

Discussion

Based on major element contents, both colourless and aqua blue glass would be derived from the mixing of sand and natron together. The high contents of Pb in colourless glass may indicate the use of recycled materials but it can be also related to the addition of antimony (Paynter 2006; Jackson and Paynter 2015).

As for decolouring agents, the fact that both Mn and Sb are present seems to suggest that their combined action has been effective in neutering the colouring effect of iron. On the other hand, the unnecessary presence of both components may provide an indication of the recycling of previously decolourised cullets.

As for provenance, the present collection has been compared with several known reference groups and especially with a total of 792 measurements performed on colourless glass and available in literature. From the reference collection, naturally colourless glass was excluded, as well as those analyses where manganese and antimony contents were not both provided by the authors. For the scope of comparison, the collected materials have been divided into six main groups, based on their MnO and Sb contents:

  1. (1)

    RC-Sb. Roman colourless glass decolourised by Sb, including items dated between the first century BC and the third century AD, with MnO < 0.2 wt% and Sb > 500 ppm;

  2. (2)

    RC-Mn. Roman colourless glass decolourised by Mn, including items dated between the first century BC and the third century AD, with Mn > 0.2 wt% and Sb < 20 ppm;

  3. (3)

    RC-MnSb. Roman colourless glass with both Mn and Sb, including items dated between the first century BC and the third century AD, with Mn > 0.2 wt% and Sb > 20 ppm;

  4. (4)

    LAC-Sb. Late Antique colourless glass decolourised by Sb, including items dated between the fourth and the sixth centuries AD, with MnO < 0.2 wt% and Sb > 500 ppm;

  5. (5)

    LAC-Mn. Late Antique colourless glass decolourised by Mn, including items dated between the fourth and the sixth centuries AD, with Mn > 0.2 wt% and Sb < 20 ppm;

  6. (6)

    LAC-MnSb. Late Antique colourless glass with both Mn and Sb, including items dated between the fourth and the sixth centuries AD, with Mn > 0.2 wt% and Sb > 20 ppm.

When the chronologies indicated by authors were covering broader time spans than those established for the Roman and the Late Antique period, the oldest chronology has been taken into account. For instance, a sample dated from the third to the fifth century AD is included in the Roman groups; based on the same criterion, a sample dated from the fourth to the seventh century is assigned to the Late Antique groups. A detailed discussion of the results obtained from this collection of bibliographic data is beyond the scope of this article and is therefore postponed to a forthcoming paper; however, mean and average values of these six groups are provided in Table 4, together with other descriptive features such as fluxes, findsites, colours, references and analytical techniques. As regards the latter, it is worth pointing out that different analytical techniques, carried out in different laboratories, can introduce errors which cannot be estimated; however, data appear consistent and outliers are very rare as it appears clear from Figs. 4, 5, 6, 7 and 8.

Table 4 Mean and standard deviation value of the 6 colourless glass groups used here for comparison, based on 792 analysed samples
Full size table
Fig. 4
figure 4

SiO2-Na2O binary diagrams showing the composition of the investigated samples (black dots) compared to that of reference groups (blue dot for Roman glass, red dot for Late Antique glass). The Levantine and HIMT fields have been drawn based on Gliozzo et al. (2013). For the colourless glass groups, both the 95 % and the 50 % prediction ellipses have been drawn

Full size image
Fig. 5
figure 5

Al2O3-CaO binary diagrams showing the composition of the investigated samples (black dots) compared to that of reference groups (blue dot for Roman glass, red dot for Late Antique glass). The Levantine and HIMT fields have been drawn based on Gliozzo et al. (2013). For the colourless glass groups, both the 95 % and the 50 % prediction ellipses have been drawn

Full size image
Fig. 6
figure 6

MgO-K2O binary diagrams showing the composition of the investigated samples (black dots) compared to that of reference groups (blue dot for Roman glass, red dot for Late Antique glass). For the colourless glass groups, both the 95 % and the 50 % prediction ellipses have been drawn

Full size image
Fig. 7
figure 7

Fe2O3-MnO binary diagrams showing the composition of the investigated samples (black dots) compared to that of the RC-Sb, LAC-Sb, RC-MnSb and LAC-MnSb reference groups (blue, red, yellow and gray dots). For the colourless glass groups, both the 95 % and the 50 % prediction ellipses have been drawn

Full size image
Fig. 8
figure 8

Fe2O3-Sb binary diagrams showing the composition of the investigated samples (black dots) compared to that of the RC-Sb, LAC-Sb, RC-MnSb and LAC-MnSb reference groups (blue dot for Roman glass, red dot for Late Antique glass). For the colourless glass groups, both the 95 % and the 50 % prediction ellipses have been drawn

Full size image

The binary diagram SiO2-Na2O provided in Fig. 4 shows that all PNE and PV samples are comparable to the RC-Sb, RC-Mn/Sb, LAC-Sb and LAC-Mn/Sb groups, while they are clearly different from Mn-decolourised glass groups (RC-Mn and LAC-Mn). The RC-Sb and the LAC-Sb groups are very similar to each other as are the RC-Mn/Sb with the LAC-Mn/Sb. Conversely, the RC-Mn and the LAC-Mn do not seem closely correlated; in fact, the former is SiO2 richer and Na2O poorer than the latter.

Despite the fact that they are not colourless glass groups, the composition of the Levantine and the HIMT glass have been compared to the investigated collection, but common features of some significance were not found. In this regard, it is worth remembering that the Levantine group is defined by 9 glass collections (Brill 1988; Freestone et al. 2000; Foy et al. 2003; Schibille et al. 2008; Foster and Jackson 2009), which include materials of Syrian–Palestinian origin, dated between the first and the eighth century AD. Conversely, the later HIMT group is defined by five glass collections (Mirti et al. 1993; Foy et al. 2003; Foster and Jackson 2009; Freestone 1994), which includes materials mainly dated between the fourth and the seventh century AD, likely originating from northern Sinai and Egypt (Freestone et al. 2005; Nenna 2014). For the sake of clarity, the average composition of Levantine, HIMT and colourless glass groups are provided in Table 5.

Table 5 Mean and standard deviation values of the Levantine and HIMT glass groups [n = stands for number of analyses]
Full size table

Al2O3 contents (Fig. 5) in PNE and PV samples are higher than those of both RC-Sb and LAC-Sb glass, but comparable to all other colourless groups. Conversely, CaO contents (Fig. 5) are perfectly compatible with the Sb-decoloured groups (RC-Sb, LAC-Sb) as well as with glass where both manganese and antimony are present (RC-Mn/Sb and LAC-Mn/Sb). Levantine glass is clearly separate from both PNE and PV samples which are instead constantly included in the HIMT field..

In addition to clearly indicating the use of natron for all PNE-PV samples, the MgO-K2O binary diagram provided in Fig. 6 further shows that the vast majority of the colourless glass considered for comparison (768 samples over a total of 792) is natron-based. It is interesting to observe that all Sb-decoloured glass, i.e. both the present collection and the reference data, concentrates just outside the bottom-left corner of the natron field while Mn-decoloured glass overall shows higher levels of K2O and, to a lesser extent, higher MgO contents.

MnO and Fe2O3 contents in PNE-PV samples never exceed 1 wt%, except for aqua blue sample no. 18. In Fe2O3-MnO binary diagrams (Fig. 7), almost all investigated samples lie just above the RC/LAC-Sb groups (which have MnO contents fixed at below 0.2 wt%), overlapping the broader fields of the RC/LAC-MnSb groups. Similarly, in binary diagram Fe2O3-Sb (Fig. 8), the PNE-PV samples overlap the RC/LAC-MnSb group fields while the correspondence with the RC/LAC-Sb fields is only partial.

Summing up, the compositional comparison between the PNE-PV samples and the colourless glass available in literature shows that the majority of samples investigated here are similar to the Roman and Late Antique glass groups containing both antimony and manganese and, to a lesser extent, to those decoloured by antimony only. A small group of vessels (19–22) showed instead lower amounts of MnO (below 0.2 wt%) and could be compared to the RC/LAC-Sb groups, especially in relation to SiO2, Na2O, MgO, K2O and MnO amounts.

With respect to the Levantine and the later HIMT coloured glass groups, a comparison based on major element composition does not seem conclusive. In PNE-PV samples, SiO2 and TiO2 contents are similar to those of the Levantine glass, Na2O, Al2O3 and CaO are similar to those of the HIMT glass, Fe2O3 contents can be compared sometimes to the Levantine while some other times to the HIMT glass, while MnO contents are even lower than the average value provided for the Levantine group.

Comparing minor and trace element data to the average values of both Levantine and HIMT glass (as reported by Schibille 2011b), it is possible to observe a greater similarity of the investigated samples with the Levantine glass and a neat distinction in relation to HIMT glass, mainly based on Zr and Ba values (Fig. 9). Personal data further show that only Hf values are in the range of the Levantine materials while greatly lower if compared to the HIMT glass. Although the depletion in rare earth and trace elements generally indicates the use of a mature sand where heavy minerals are quantitatively very low, the possibility that the REE pattern is “altered” by contaminants introduced via antimony must be also taken into account.

Fig. 9
figure 9

REE pattern of the investigated samples and reference materials. All values have been normalised to the upper continental crust (Kamber et al. 2005). Levantine and HIMT glass patterns from Schibille (2011b)

Full size image

Conclusions

The sample set is rather homogeneous both texturally and compositionally, even though the few aqua blue samples here investigated showed slightly higher iron contents, while a colourless (no. 20) and an aqua blue sample (no. 7) were slightly different from the rest of the group, based on lower and higher trace element contents, respectively.

The majority of the investigated colourless samples were found to be closely comparable to Roman and Late Antique colourless glass in which both antimony and manganese are present, here temporarily named RC-MnSb and LAC-MnSb glass groups; the only exceptions being their averagely higher Al2O3 contents with respect to these wider glass groups.

Three goblets type Isings 86 (PNE 20-21 and PV 22) and one cup (PV 19) showed MnO amounts lower than 0.2 wt% and also the overall values obtained for the other major element contents were similar to those of both RC-Sb and LAC-Sb groups.

As far as provenance, it has been noted that among major elements, some (esp. SiO2, TiO2 and MnO) were more similar to the Levantine glass while others (esp. Na2O, Al2O3 and CaO) were compatible with the later HIMT reference group. However, the REE pattern seems to rule out the possibility of a HIMT and thus Egyptian provenance in favour of a Levantine one. Assuming thus a Levantine origin of the sand, we could hypothesise that a greater amount of Na2O and thus of the natron flux was used for the production of this SiO2-rich colourless glass. The higher Al2O3 contents could possibly be explained by the use of a different (and perhaps geographically close) sand. In that case, the CaO contents may either be typical of this ad hoc sand or intentionally added, based on technological constraints: an increase of vitrifying and fluxing agents corresponding to a relative increase of the stabilizing agent.

Lastly, despite an overall analysis of Roman colourless glass is beyond the scope of this research, it is definitely helpful to observe that the RC-Sb, LAC-Sb, RC-MnSb and LAC-Mn/Sb groups include mostly vessels and tesserae but also windows, moils, chips, rods, trails, chunks and lumps. Although further investigation will be required, it does not seem that colourless glass was preferentially used for a specific category of objects or rather for a glass production addressing a specific level of society. Also, among the investigated vessel collection, the glass composition of uncommon and high quality items such as the three goblets type Isings 86 (PNE 20-21, PV 22) is closely comparable to that of a cup of common use (PV 19).