Introduction

The Roman theatre is one of the most important buildings of the Roman town of Nora, located in the Gulf of Cagliari (south-western Sardinia, Figs. 1 and 2). Nora was founded by the Phoenicians around the mid-eighth century BC, although there is evidence of earlier settlements (AA.VV. 2000; Lugliè 2009b; Wilson 1980). The theatre has a semi-circular isodomic front with a W–NW axis (Fig. 3) and was built during the Augustan or Giulio-Claudio period (first century AD; Bejor 1999; Ghiotto 2004). According to the use in Roman times (Adam 2006; Cagnana 2000; Giuliani Cairoli 2006), a variety of geomaterials (sandstones, conglomerates, volcanic rocks, marbles, bricks, etc.) and different kind of mortars (i.e., Roman concrete, bedding and jointing mortars of ashlars and bricks, plasters) were employed to construct the theatre. In the mortars, several different raw materials (i.e., quartz-feldspathic sands, fine and coarse volcanic aggregate) were used, according to the mortar function and the different sectors of the building: structure-wall, tribunalia vaults, wall of external niches, foundation of cavea tiers, and stage inner wall.

Fig. 1
figure 1

Geological map of Sardinia with localization of the Roman Nora village and other important Punic-Roman city from the south to the north of the island, i.e., Karales (today Cagliari), Forum Traiani (Fordongianus), Turris Lybisonis (Porto Torres); from Columbu and Garau (2017). Legend of patterns and colours refers to lithologies: white, recent alluvial sediments; light gray, Oligo-Miocene volcanics; dark gray, Plio-Pleistocene volcanics; gray stippled, Miocene marine sediments; gray crosses, Palaeozoic crystalline basement and Mesozoic formations. Red continuous and dashed lines, faults

Full size image
Fig. 2
figure 2

Nora Roman theatre: (a) overview of Nora archaeological site with the theatre evidenced by red ellipse; (b) southwest view of the hyposcenium space under the stage (palcoscenium) and east-side tribunalia sector (see Fig. 3); (c) southeast view of orchestra, west-side of cavea tiers and (down) part of hyposcenium; (d) view of east-side tribunalia sector (with recent cement rebuilt-consolidation under the vault); (e) detail of east-side cavea sector with greyish sandstone ashlars, original purplish volcanic ashlars and concrete of the tier foundation; (f) original mosaic decoration of orchestra floor

Full size image
Fig. 3
figure 3

Plan of the Nora Roman theatre. Legend of sectors: (a) cavea (auditorium); (b) orchestra; (c) wall of stage; (d) hyposcenium (space under the palcoscenium); (e) front scene; (f) portico behind the scene; (g) west archway entrance; (h) access ladder at the west tribunalia; (i) west tribunalia; (l) parasceni; (m) external niches; (n) enclosure; (o) via (corridor atop the cavea); (p) scalaria; (q) euripus (underground channel for water drainage); (r) dolia; (s) small pillars under the palcoscenium (from Columbu and Garau 2017, modified)

Full size image

The volcanic glasses show characteristics similar to obsidian facies and not to natural pozzolan normally used in the Roman period for making the mortars. The use of these pozzolanic glasses, which at the outset does not show to share a local origin, is a novelty because, to date, they have never been found in the ancient mortars neither of the Nora archaeological site nor of other roman sites in the island. Given the wide use of obsidian supplied by the Sardinian Mt. Arci source for the production of tools and artefacts from the Neolithic to the late Bronze Age, significant considerations about its use, origin and exploitation can be made.

The archaeometric investigations on both the geochemical and petrographic features and origin of raw materials used in the ancient mortars are fundamental for understanding the ancient routes used by the Romans for the transport of stone materials and for obtaining information on the interconnection of Roman settlements and residential sites present in Sardinia. Moreover, these studies are useful to define the technologies and behaviours for the construction of ancient buildings in different historic times and to address the conservative and restorative works (Adriano et al. 2009; Alvarez et al. 2000; Antonelli et al. 2014b; Bertorino et al. 2002; Bianchini et al. 2004; Bultrini et al. 2006; Columbu 2017, 2018; Columbu et al. 2014a, b, 2015a; b, 2017a; b, 2018a; b; c; Columbu and Verdiani 2014; De Luca et al. 2013, 2015; Franzini et al. 2000; Gutiérrez et al. 2016; Lapuente 2014a, b; Lapuente et al. 2012; Lezzerini et al. 2016, 2018a; b ; Miriello et al. 2010, 2015; Maravelaki-Kalaitzaki et al. 2003; Moropoulou et al. 2000, 2004; Ramacciotti et al. 2018; Riccardi et al. 1998; Smith and Smith 2009; Stanislao et al. 2011; Verdiani and Columbu 2010, 2012; Vola et al. 2011).

To identify their geological provenance, the geochemical data of the glass samples were compared with those of new and literature data of acid volcanics (only the rhyolites and alkali-rhyolites were selected) for south and central Sardinia, where one may find similar lithologies: the island of St. Antioco and Mt. Arci, respectively. In the last of these areas, singled out as possible sources, despite more than 15 years of continuous and systematic surveys performed all around the massif (Lugliè 2003, 2013), there are no clues yet of quarrying nor of reduction activity during the Roman period, but striking evidence from Neolithic to Bronze Age processing of obsidian.

More than 250 rock samples from both the outcrops (Mt. Arci and St. Antioco areas) and volcanic glasses of mortars from the main sectors of the theatre were analysed. The results of chemical analysis together with the petrographic characterisation were used in order to (a) geochemical classify the volcanic glasses from the mortars of Nora theatre, according to different diagrams; (b) define compositional inhomogeneity internal to the same glass samples; (c) identify framing of glasses in the magmatic series, analysing the geochemical trend of trace elements and the rare earth pattern; and (d) identify the geographical origin of these glasses employed in the mortars, to understand their supply in Sardinia in the Roman period.

Materials and analytical methods

Fifty pozzolanic volcanic glass fragments of aggregate among 48 selected and representative samples of mortars were analysed and studied. The mortar samples come from the following sectors of the theatre: tribunalia (9 samples, divided into 7 concretes of two vaults, 2 bedding mortars of ashlar structure); cavea (14 jointing/foundation mortars of volcanic and sandstone ashlars); external niches (16 samples, divided into 9 bedding mortars of outer brick-walls, 3 concrete mortars of inner structure-wall, 2 concrete of basement of via, corridor around the external side of the cavea, 1 concrete of only one vault overlying the niche); and pulpitum (7 bedding mortars and 4 samples of arriccio plasters of the ribalta wall and pillar of hyposcenium).

The mineralogical and petrographic analysis of volcanic rocks and mortars was performed on thin sections under the polarising microscope (Zeiss photomicroscope Pol II).

The major chemical elements of volcanic glass from the mortars were analysed in thin section under an electron microprobe with a Cambridge S360 scanning electron microscope, equipped with an energy-dispersive spectrometer Link QX2000, Pentafet detector and IBM 686 computer equipped with appropriate software for the acquisition of scanned images. Microanalyses were collected at 15 kV using a 3 μA beam current and a 25 μm spot size.

The analysis of the trace elements (including the full pattern of rare earth) of volcanic glasses included in the mortars was carried out with inductively coupled mass spectrometry combined with laser ablation as a sampling system. It has necessitated an ad hoc sample preparation: some volcanic glass fragments isolated from each mortar sample were embedded in two-component epoxy resin (RenLam M-1, viscosity 1300 mPa s at 25 °C) of cylindrical shape. The obtained test specimen was gradually treated with abrasives of silicon carbide and alumina powders to bring to the surface and polishing the embedded glass fragments. The diameter of the laser beam used is of 40 μm, with a frequency of 10 Hz and a fluence of about 15 J/cm2. Data reduction was performed using software “Glitter” (Van Achterbergh et al. 1999). As external standards, synthetic glass NIST 610 and BCR-2 basalt have been used, while as a variable, internal standard 44Ca was used. In this method, the intensity signals of the elements depend both on the concentration changes and the mass ablated in subsequent spot. To make independent analysis by mass variations, the internal standard technique with variable concentrations should be used: the concentration of the analyte is a linear function of the intensity ratio of the analyte in the unknown sample relative to that of the reference standard for the same element, with a correction linearly dependent on the ratio of concentration of the same element (internal standard) content in the unknown sample with respect to a fixed concentration of the same element in the reference sample. Consequently, the knowledge for each point analysis of the concentration of internal standard assumes great importance. The glasses, which have been preliminarily subjected to analysis by electron microprobe, were given an average concentration for Ca element of 01.34 ± 0.13 at 1σ. The content 29Si% was used as a variation index (see the diagrams of Fig. 13a–c). Nevertheless, clearly the concentrations in 29Si% (average 49.14%) obtained by this method have a high standard deviation (± 16:24 at 1σ) in relation to the variability of the concentration of calcium in each fragment and are consistently higher than those obtained in the electron microprobe.

The chemical composition of volcanic rock samples from St. Antioco and (a part) from Mt. Arci was determined with a spectrometer in X-ray fluorescence Philips PW1400 with a Rh tube to analyse the major elements and some trace elements (Rb, Sr, Pb, Zn, Y, Nb, Zr), and with a W tube to analyse Ni, Cr, Ba, V, La and Ce. Data reduction of major elements was performed by the method proposed by Franzini et al. (1975). Data reduction of trace elements was performed by the method of Criss (1977), modified. The measurement accuracy is ± 1% for SiO2, TiO2, Al2O3, Fe2O3, CaO, K2O and MnO and ± 4% for MgO, Na2O and P2O5. The detection limits are about 3 ppm to 3σ for most of the elements; the accuracy of trace elements is ± 2 ÷ 3% to 1000 ppm; ± 5 ÷ 10% at 100 ppm and ± 10 ÷ 20% to 10 ppm. The weight loss for calcination (L.o.I., Loss on Ignition) was determined by calculating the loss in wt.% at 1100 °C, while the FeO was determined by volumetric titration with KMnO4 10 N in acid solution.

The analysis of the major elements of obsidians and perlites from Mt. Arci was performed by electron probe microanalysis using Wavelength Dispersive Spectrometry with a device ARL-SEMQ, using standards such as silicates and natural oxides; the data reduction was performed with the ZAF method (Colby 1971). Detection limits of ~ 100–300 ppm are readily attained. During a typical multi-element analysis of a few minutes duration, 1σ precision of 0.3–1.5% relative is normally attained for major elements (i.e., those present at concentrations > 1% by weight). If desired, detection limits and precision may both be improved by increasing counting times, by increasing beam current and/or by assigning multiple spectrometers to a single element.

Characteristics of mortar aggregate

At the macroscopic observation, the mortars of the theatre (Fig. 4) show a reddish-grey-black and white/beige-grey aggregate (belonging to volcanic rock fragments and sialic mineral sands, respectively) with variable size and amounts, according to the different mortar functions in the sectors.

Fig. 4
figure 4

Mortar samples from different sectors of Nora theatre. (a) Bedding mortar of Tirrenian sandstone ashlars of theatre structure with a flattened rock fragment of local Oligo-Miocenic volcanic stone; (b) bedding mortar of tier volcanic ashlars belonging to “Su Casteddu” outcrops (see Melis and Columbu 2000; Columbu and Garau 2017); (c) mortar sample from Roman concrete of cavea foundation; (d) entrapment mortar of caementium (Roman concrete) from east-side tribunalia vault; (e) entrapment mortar from Roman concrete of cavea foundation; (f) joint-mortar taken between the volcanic ashlars of cavea tiers

Full size image

The main compositional characteristics of the aggregate determined by microscopic modal analysis are reported in Table 1, along with the binder/aggregate ratio (B/A) expressed in vol.% (by modal analysis) and in wt.% (after mechanical disintegration and acid attack of binder). The mortars show a greatest variability of vol.% B/A, due to the small size of the thin section that did not include the coarse fragments observed macroscopically, which probably do not reflect the real mixing proportions set by the manufacturer. B/A ratio calculated using the wt.% data is much closer to those recommended by ancient sources (0.3–0.5 vol.%; Vitruvio Pollione 1521).

Table 1 Composition defined by polarized microscope analysis on thin sections of mortar samples taken from tribunalia, cavea, structure-walls and vault of niches, hyposcenium and pulpitum (stage) walls (from Columbu and Garau 2017, modified). The binder/aggregate ratio (B/A as wt.% after dissolution of binder and as vol.% by modal analysis) has also been reported
Full size table

The mortar aggregate consists mainly of quartz, orthoclase (Fig. 5e), microcline, plagioclase, black-grey volcanic glasses (Fig. 5a, b) and fragments of purplish volcanic rocks (Fig. 5c) with a local origin (according to Columbu and Garau 2017). Subordinately, crystalline rocks from the Palaeozoic basement (Fig. 5f), occasional mafic crystal-clasts (as pyroxene, Fig. 5d, biotite and opaque oxides), rare bioclasts and fragments of carbonate rocks (e.g., marble, limestone) were also used in the mortars. Not local volcanic glasses were used as aggregate but also with pozzolanic function, as highlighted by the reaction edges with the binder (Figs. 5a and 6b). Then, cocciopesto has also been used as pozzolanic material but overall in very low percentages (Table 1). The binder consists mainly of calcite, often present as spherules (Fig. 6d), where occasionally there are immersed lumps of bad-carbonated lime (Fig. 6c).

Fig. 5
figure 5

Photograph details of mortar aggregate on polarized microscope. (a, b) Plain polars: fragment of volcanic glass with evident vacuolar (and fluidal) structure and borders of reaction with the binder; (c) cross Nicol: fragment of local Oligo-Miocenic dacitic volcanic rock with binder-reaction borders; (d) plain polars: orthopyroxene crystal; (e) cross Nicol: altered orthoclase crystals; (f) cross Nicol: rounded fragment of meta-sandstone from Palaeozoic crystalline basement

Full size image
Fig. 6
figure 6

Photograph details of mortars on electronic microscope (SEM). (a) Binder-reaction borders between volcanic aggregate and binder; (b) binder-reaction borders between a fragment of volcanic glass and the binder; (c) lump of bad-carbonated lime; (d) spherules of calcite

Full size image

The local volcanic rocks belong to the Sarroch-Pula volcanism, related to the Late Eocene–Miocene Sardinian magmatic activity that occurred between 38 and 15 Ma (Advokaat et al. 2014; Antonelli et al. 2014a; Beccaluva et al. 1985, 1989, 1994, 2005a, b, 2011; Cherchi et al. 2008; Columbu et al. 2011; Lustrino et al. 2004, 2009, 2011, 2013). Considering the medium-coarse size, it is probably that the purplish volcanic rocks were used mainly as (inert-) aggregate, although they also show pozzolanic characteristics (Figs. 5c and 6a) (see Türkmenoglu and Tankut 2002).

The different distribution of the main components in the aggregate has been showed in the ternary diagram of Fig. 7 where the modal percentages of quartz (Qz), feldspar (Fds) and volcanic glass (V) were reported. The volcanic glasses were used in almost all samples in varying proportions, according to the function of mortar in the theatre; they are less present in the plasters while they are more abundant in the structural concretes of cavea foundation and of the tribunalia vaults, where in some cases represent about 90% of the aggregate.

Fig. 7
figure 7

Aggregate compositional distribution of quartz (Qz), feldspar (Fds) and volcanic glass (V) in the mortars from different sectors of the theatre (from Columbu and Garau 2017, modified)

Full size image

The sandy component is widely used in the bedding mortars of the wall bricks and in the plasters (Fig. 7), while it is found less frequently in the concretes of the tribunalia vaults and variously present in the cavea samples and external niches. It shows an almost constant ratio of quartz and feldspar (2:1, Fig. 7) in almost all mortar samples thus highlighting a unique sampling point of these sands (Columbu and Garau 2017).

Regarding the particle size, according to the Folk (1968) classification, the aggregate is made from sand, slightly gravelled sands, gravelly sands and, in some cases, sandy gravel, while according to the Wentworth (1922) classification, it is mostly composed of large and medium sands. The aggregate shows a moderately selected sorting (according to Folk 1954).

By analysis under a reflection microscope, the volcanic glass makes up the majority of the aggregate fraction below 2000 μm, along with the more rare andesitic-dacitic rocks aggregate from local volcanic outcrops. The sand fraction is mostly represented by quartz, feldspar, Palaeozoic rock fragments, Oligo-Miocenic volcanic rocks, volcanic glass and rare cocciopesto.

Results

Petrographic features of volcanic rocks

In the mortars and concretes of the theatre, two different kind of volcanic stones were employed: (1) medium-welded rocks, mainly used as medium-coarse aggregate and (2) grey-black obsidian glasses with no local origin, used as aggregate and also as pozzolanic material.

(1) Medium-welded rocks

By macroscopic observation, the first rocks, characterised by chromatisms varying from grey-reddish to purplish-brown, show an evident self-clastic structure typical of a volcanic autobreccia, with sub-centimetric lava-clasts. By microscopic analysis, these volcanic rocks show a hypo-crystalline porphyritic structure (with variable porphyritic index between 5 and 8) for phenocrysts of early opaque, dominant plagioclase, pyroxene and hornblende. Due to the squat form, the opaque minerals are formed, presumably, by titan-magnetite or magnetite.

On the base of volcanological and petrographical features, these rocks show a similarity with the dacitic rocks of territory around the Nora village (e.g., “Perdu Pranu” outcrop, NE to the site) and especially with the volcanic stones used for the tiers of the theatre (cavea, see Fig. 3) belonging to the ancient quarry of “Su Casteddu” (Melis and Columbu 2000; Previato 2016; Columbu and Garau 2017; Columbu 2018). The latter is a volcanic structure (of which today essentially remains just at the neck) located about 1.5 km north-west from the Nora site. In fact, the outcrop rock shows the same characteristics already observed in the volcanic-coarse aggregates, characterised by the chaotic presence of large lava-clasts (usually sharp-edged) and smaller lava-clasts with rounded contour, with maximum dimensions of 1 cm immersed in a glassy matrix with a lower degree of welding, compared to the clasts.

(2) Grey-black obsidian glasses

The grey-black obsidian glasses show a hyaline structure with rare phenocrysts of plagioclase and biotite. These glasses are characterised by pearlitic fractures or from a vacuolar and/or fluidal texture. In some cases, there are typical devitrification structures as spherulites (in agreement with Lofgren 1971). These volcanic fragments are always characterised by the presence of reaction rims between the glass and the binder of mortar (Fig. 6b).

Geochemical characteristics of obsidian glasses

Analysis of major elements and rock classification

Table 2a, b shows the results of the chemical analysis of the glasses used in the mortars of Roman theatre, where the analytical values of major elements and the C.I.P.W. norm are reported (according to Cross et al. 1903). Table 3a, b shows the results of chemical analysis on the volcanics (perlites and obsidians) of Mt. Arci (Fig. 1). Table 4 shows the analytical values of the volcanics from St. Antioco area (Fig. 1).

Table 2 Chemical analysis of volcanic glasses from the theatre mortars, where reported the rock classification (according to De La Roche et al. 1980), wt.% of major elements and C.I.P.W. norm according to Cross et al. (1903)
Full size table
Table 3 Chemical analysis of selected volcanic rocks from Monte Arci (central-western Sardinia), where reported the rock classification (according to De La Roche et al. 1980), wt% of major elements and C.I.P.W. norm according to Cross et al. (1903)
Full size table
Table 4 Chemical analysis of selected volcanic rocks from St. Antioco area (Sulcis, south-western Sardinia), where reported the rock classification (according to De La Roche et al. 1980), wt.% of major elements and C.I.P.W. norm according to Cross et al. (1903)
Full size table

Table 5 shows a summary of the rock classification of the glasses from the theatre mortars and volcanic samples from Mt. Arci and St. Antioco, according to the diagrams of Middlemost (1975) (Fig. 8) and De La Roche et al. (1980) (Fig. 9). According to the latter diagram, the glasses of the mortars are classified as rhyolites. The samples of perlite and obsidian from Mt. Arci are classified as alkali-rhyolites and (as transition products) to the rhyolites. The acidic volcanics of St. Antioco constitute a separate group with respect to both the mortar glasses and the samples from the Mt. Arci outcrops.

Table 5 Distribution of analysed volcanic samples within sodium (Na), potassium (K) and high in potassium (HK) series, according to Middlemost (1975) and De La Roche et al. (1980) rock classification
Full size table
Fig. 8
figure 8

Na2O vs. K2O wt.% classification diagram of Middlemost (1975) between the high-potash, potash and soda volcanic series, where plotted the volcanic samples from the outcrops and aggregate glasses from the mortars of the Nora theatre. Abbreviations of the legend: K-Rhy, potash rhyolite; K-alRhy, potash alkali-rhyolite

Full size image
Fig. 9
figure 9

R1 vs. R2 classification diagram of De La Roche et al. (1980), where plotted the volcanic samples from the outcrops and aggregate glasses from the mortars of the Nora theatre. Symbols as legend in Fig. 8

Full size image

According to the TAS (Total Alkali Silica) diagram of Le Maitre et al. (2002); Fig. 10), most of the samples fall within the field of the rhyolites, while only some perlites from Mt. Arci and some volcanics from St. Antioco fall within the trachy-dacite and dacite fields.

Fig. 10
figure 10

Total alkali-silica diagram [(Na2O + K2O) vs. SiO2 wt.%] of Le Maitre et al. (2002), where plotted the volcanic samples from the outcrops and aggregate glasses from the mortars of the Nora theatre. Symbols as legend in Fig. 8

Full size image

All samples fall under the dashed discriminant line of Irvine and Baragar (1971) between the alkaline and subalkaline series, which have been overlapped to the TAS diagram in Fig. 10.

Also, the diagram of De La Roche et al. (Fig. 9) shows the subalkaline character of the volcanics which, falling between the line of the critical plane of unsaturation (and away from it) and the abscissa axis, are strongly supersaturated.

In the classification diagram of Peccerillo and Taylor (1976); Fig. 11), most of the samples fall within the field of shoshonitic series and subordinately of K-high series. Almost all samples are classified as rhyolites, except some perlites of Mt. Arci and Sant’Antioco, which are classified as trachytes of shoshonitic series.

Fig. 11
figure 11

K2O vs. SiO2 wt.% classification diagram of Peccerillo and Taylor (1976) where plotted the volcanic samples from the outcrops and aggregate glasses from the mortars of the Nora theatre. Note: it was not possible to plot all analyses, because some samples have higher values of SiO2 to 73%. Symbols as legend in Fig. 8

Full size image

Figure 12 reports the variation diagrams of major elements versus the differentiation index (D.I.) of Thornton and Tuttle (1960). Regarding the volcanic glasses of the mortars, the typical trend of common magmatic series, with a decidedly positive correlation between SiO2 and D.I. is observed. Even the K2O and Na2O are positively correlated with D.I., although weakly. A negative correlation exists, however, between Al2O3 and P2O5, typical of the evolved rocks which splits the apatite. Other oxides, such as MgO and TiO2, are quite dispersed.

Fig. 12
figure 12

Variation diagrams: major elements (wt.%) vs. differentiation index (D.I.) of Thornton and Tuttle (1960) (where D.I. = normative Q + Ab + Or + Ne + Kp + Lc) for the volcanic samples from the outcrops and aggregate glasses from the mortars of the Nora theatre

Full size image

Almost always, similar trends are even noted for the elements of obsidian and perlite samples from Mt. Arci, except for Al2O3 and CaO that remain almost constant. All diagrams of Fig. 12 show a different behaviour of the mortar glasses (that form a distinct group) compared to the samples from St. Antioco and Mt. Arci. Even among samples from Mt. Arci, it is possible to distinguish two different groups: the perlites and the obsidians. In particular, the perlite samples are characterised by higher values of Al2O3, CaO and K2O and from lower values of P2O5 and Na2O.

The behaviour of K2O and Na2O has already been highlighted by Cioni et al. (2001), according to which the interaction between the anhydrous volcanic glasses (obsidian) and the meteoric waters lead to the formation of hydrate glass (perlite) with leaching of Na2O and relative enrichment in K2O.

Instead, the samples of the Perdas Urias outcrops (eastern side of Mt. Arci) form a group almost always superimposed on the mortar glass samples, as already noted in the above-mentioned diagrams, showing a certain similarity geochemistry with them.

Analysis of trace elements

Tables 6 and 7a–e show the results of the chemical analysis of trace elements of the obsidian glasses from the mortars, where also some major elements are reported. The analytical values are plotted in the diagrams of Fig. 13a–c versus 29Si (expressed in % for graphic reasons). The patterns are comparable with those of common magmatic series, highlighting the validity of the 29Si% as variation index.

Table 6 Chemical analysis of three plagioclase phenocrystals from the aggregate of theatre mortars
Full size table
Table 7 Chemical analysis of trace (and some major) elements of volcanic glasses from the theatre mortars. The values are reported in ppm
Full size table
Fig. 13
figure 13figure 13

(a–c) Variation diagrams: selected major and trace elements (ppm) vs. 29Si% for the volcanic samples from the outcrops and aggregate glasses from the mortars of the Nora theatre

Full size image

There was a positive correlation between 29Si% and 23Na, 66Zn, 49Ti, 85Rb, 133Cs, 137Ba, 93Nb, 139La, 208Pb and 238U, while there is a negative correlation with the 88Sr; the values of the other elements are dispersed.

The analytical values of some elements (Table 7a–e) are far higher than the average of the same element in the other analysed points; these values, which may depend on the presence of phases within the micro volume of vaporised sample by the laser, or by a lack of homogeneity at the microscopic level in the distribution of the elements, were regarded as outsiders and therefore excluded from the variation diagrams.

The trends of some rare earth elements normalized to chondrites (factors taken from Anders and Grevesse 1989) are shown in Fig. 14a. They are characterised by a moderate variability (minimum 157.47 ± 9:49 ppm/chondrite for Ce; maximum: 13.66 ± 3.2 ppm/chondrite for Eu), indicating that they belong to a single magmatic series. Furthermore, while the light rare earths have parallel trends that do not intersect between them, the trends of the heavy rare earths are less correlated and tend to interbreed. Table 8 shows the correlation matrix for the rare earths showed in Fig. 14a; the maximum correlation for light rare earths coarsely tends to decrease for heavy rare earths, as highlighted by the variation diagrams vs. 29Si% in Fig. 13a–c.

Fig. 14
figure 14

(a, b) Geochemical characteristics of the mortar glasses: (a) pattern of selected rare earths normalized to chondrite (factors from Anders and Grevesse 1989); (b) spider diagram of selected elements normalized to primitive mantle, according to Wood (1979)

Full size image
Table 8 Correlation matrix for some rare earths of volcanic glasses from the theatre mortars
Full size table

In general, the patterns are characterised by a negative peak in correspondence with europium, which indicates in all probability the fractionation of plagioclase, in good agreement with the petrographic observations and chemical analysis of some plagioclase phenocrystals of mortar glasses; in fact, in the plagioclases, there is a greater europium content with respect to the content of the same element in the glasses (Table 6). The Eu negative peak also indicates the apatite fractionation (Cox et al. 1979), according to the trend of P2O5 vs. D.I. (Fig. 12).

The pattern of the rare earths is characteristic of the final stages of the magmatic series and it is comparable with the pattern already observed in Sardinia for the dacitic and comenditic rocks from Sulcis area (Morra et al. 1994) and for dacites and rhyolites of Mt. Arci (Beccaluva et al. 1984).

Similarly, the pattern of trace elements normalized to primitive mantle according to Wood 1979 (Fig. 14b) shows that the whole sequence analysis has a regular distribution, with a not very wide range of variation (minimum 5.69 ± 0.31 ppm/primitive mantle for Sr; maximum 13.22 ± 2.48 ppm/primitive mantle for Hf); only some samples deviate on the performance more generally in correspondence of the tantalum and niobium, hafnium and zirconium. The pattern is characterised by an enrichment of the lithophile elements with wide ionic radius (LILE). Some elements, such as Sr and Ba, show the negative peaks, surely due to the fractionation of plagioclase and K-feldspar. In fact, as can be seen from Tables 6 and 7, the value of Sr (in ppm) in plagioclase is much higher than the value of Sr present in the glass. It also shows an impoverishment compared to the primordial mantle, due probably to the fractionation of iron and titanium oxides.

Provenance of pozzolanic glasses

Considering that in the areas adjacent to Nora’s site there are no rock outcrops with similar geochemical-petrographic characteristics, in order to identify the sources of supply, the composition geochemistry of these glasses with new and literature analytical data of similar volcanic rocks (rhyolites/alkali-rhyolites) from Mt. Arci and St. Antioco areas was compared, using the linear discriminant analysis.

Discriminant analysis using major elements

The discriminant analysis (performed using the Statistical Mac program) was applied to subdivide the groups defined a priori that are represented by the glasses used as aggregate and pozzolan in the theatre mortars, the volcanic samples (obsidians and perlites) from Mt. Arci and the volcanics coming from the St. Antioco.

In detail, 232 analyses (divided as in Table 9) were considered. The variables chosen for the discrimination of the groups are represented by the following major elements: SiO2, TiO2, Al2O3, FeOT, MnO, MgO, CaO, Na2O, K2O and P2O5. Among them, those that are found to be significant (based on the discriminant analysis) are the following: CaO, FeO, Al2O3 and P2O5.

Table 9 Subdivision of 232 chemical analyses in the groups identified in advance to make the discriminant analysis on the basis of the major elements
Full size table

The samples will be classified as belonging to the group that has the higher score (Table 10a).

Table 10 (a) Classificative functions for the discriminant analysis on the basis of the major elements. (b) Summary of the classification of sample groups
Full size table

The classification score of a sample for a group, for example from the Nora theatre (TN), is calculated as follows:

$$ \mathrm{Score}\left(\mathrm{TN}\right)=-295,945+\mathrm{CaO}+{\mathrm{FeO}}_{\mathrm{T}}\ast 16,546\ast 63,419+{\mathrm{P}}_2{\mathrm{O}}_5\ast \left(191,734\right)\ast 37784+{\mathrm{Al}}_2{\mathrm{O}}_3. $$

Table 10b shows the summary diagram of the classification results, according to the classification score groups. All analyses of the mortar glasses are properly classified as belonging to the “Theatre” group. Also, all analyses of samples from “Mt. Arci” are properly classified. For “St. Antioco” samples, 92.86% of the total analysis was correctly classified, and single analysis was attributed to the theatre group and a second one to the Mt. Arci group.

Table 11a shows the canonical functions, which are equal to the number of groups minus one. For each sample, each variable must be multiplied by the coefficient shown in the table according to the following scheme:

$$ \mathrm{Root}\ 1=3.9207+\mathrm{CaO}+{\mathrm{FeO}}_{\mathrm{T}}\ast 4.52979\ast \left(-0.79973\right)+{\mathrm{Al}}_2{\mathrm{O}}_3+{\mathrm{P}}_2{\mathrm{O}}_5\ast 10.70594-0.54622. $$
Table 11 (a) Coefficients of canonical functions (root 1 and root 2) for the discriminant analysis on the basis of the major elements. (b) Equations (as y = ax + b) of dividing straight lines (1, 2, 3)
Full size table

In the diagram in Fig. 15 are projected the points of the canonical score R1 and R2, calculated according to the functions given in Table 11a.

Fig. 15
figure 15

Discriminant diagram (root 1 vs. root 2) on the basis of the major elements of volcanic glasses from the mortars and volcanic samples from Mt. Arci and St. Antioco areas. Abbreviations as the legend in Fig. 8

Full size image

In Table 11b, the three equations (with y = ax + b) of straight lines dividing between the three groups are reported.

Observing the discriminating diagram of Fig. 15, the volcanic glasses used in the mortars constitute a distinct group compared with those of other samples from Mt. Arci and St. Antioco. The analysed samples from Perdas Urias outcrops (belonging to the Mt. Arci area), though falling within the field of St. Antioco group (Fig. 15), are very close to the analysis of mortar glasses.

Discriminant analysis using trace elements

The discriminant analysis was applied using the trace elements of the mortar glasses, obsidians of Mt. Arci and volcanics of St. Antioco.

One hundred and sixty-three analyses were considered, among which 112 were related to mortar glasses, 37 obsidians to Mt. Arci and 12 to St. Antioco. The variables chosen for the discrimination of the groups are represented by the following elements: Ti, V, Cr, Zn, Rb, Sr, Y, Zr, Nb, Cs, Ba, La, Ce and Pb. Among them (based on discriminant analysis), those that perform most significant results are as follows: Sr, Ce, V, Ba, Rb and Y. Similarly to the case of major elements, each sample is classified as belonging to the group that shows the higher score, calculated using the coefficients in Table 12a.

Table 12 (a) Classificative functions for the discriminant analysis on the basis of the trace elements
Full size table

Table 12b shows a summary diagram of the classification results, according to the classification score groups. In the case of the mortar glasses, obsidians of Mt. Arci and volcanics of St. Antioco, the classification is correct in 100% of cases.

Table 13a shows the canonical functions, which, also in this case, are equal to two (number of groups minus one). For each sample, each variable must be multiplied by the coefficient shown, according to the scheme of previous paragraph.

Table 13 (a) Coefficients of canonical functions (root 1 and root 2) for the discriminant analysis on the basis of the trace elements. (b) Equations (as y = ax + b) of dividing straight lines (1, 2, 3)
Full size table

In Fig. 16 are projected the points of the canonical score R1 and R2. In Table 13b, the three equations (with y = ax + b) of straight lines dividing between the three groups are reported.

Fig. 16
figure 16

Discriminant diagram (root 1 vs. root 2) on the basis of the trace elements of volcanic glasses from the mortars and volcanic samples from Mt. Arci

Full size image

The analysis of samples from the mortars of the theatre, from Mt. Arci and St. Antioco, constitutes separate fields, while the analysis of samples from Perdas Urias falls under the volcanic glasses from the mortars of the theatre with a good overlap.

Discussion of results

Inside the Nora Roman theatre, a great amount of obsidian glasses together with quartz/feldspar sands is mainly used as aggregate in the hydraulic mortars. These glasses have been also used as pozzolan materials, although their characteristics are different with respect to the natural pozzolans normally used in the Roman period. Thus, the volcanic glass gives good hydraulic characteristics to the mortars, as shown by the constant presence of reaction edges with the binder (Fig. 6b) with consequent chemical exchanges: a decrease of Si, Al and K and a simultaneous increase of Ca and volatiles compared with glass (Columbu and Garau 2017). Given the good pozzolanic characteristics of the volcanic glass, it is conceivable that the Romans voluntarily used this type of material in place of cocciopesto (scarcely present in the mortars or absent in some samples) to make as much resistant mortar from the structure of the theatre, an open work subject to weathering processes.

The use of these volcanic glasses in the archaeological site of Nora, as well as in other Sardinian Roman monuments, at the moment was not known. Differently from the other components of the aggregate, these obsidian rocks do not belong to the vicinity of the Nora site, because there are no outcrops bearing these petro-volcanological characteristics. However, their wide use in the theatre mortars may suggest either a local availability or a not very distant origin. The geochemical comparison of data suggests that the mortar glasses have a source from the volcanic complex of Mt. Arci (central-western Sardinia; Fig. 1) where there are volcanic rocks with similar characteristics. As showed by statistical step-wise linear discriminant analysis, the samples from the locality of Perdas Urias (east of Mt. Arci volcanic complex) show a geochemical similarity (especially on the basis of trace elements) with the volcanic glasses from the mortars, suggesting a provenance from Perdas Urias of the obsidian glasses used in the mortars of Nora theatre. The Mt. Arci obsidian deposits are well-known both in the archaeological and geological literature, (Barca et al. 2007; De Francesco et al. 2008, 2011; Freund 2014; Freund and Batist 2014; Le Bourdonnec et al. 2006, 2010, 2015; Léa 2012; Lilliu 1988; Lugliè 2003, 2009a, 2010; Lugliè 2013; Lugliè et al. 2006, 2007, 2008, 2011; Macciotta et al. 2004; Mackey and Warren 1983; Marchi et al. 2005; Tykot 1996, 1997, 2002; Tykot et al. 2008). The obsidian rocks (emphatically called the “black gold” of the prehistoric period) is an important lithic raw material that has been used since the Early Neolithic with a high diffusion in many Neolithic to Chalcolithic sites in Sardinia and in a part of the Western Mediterranean area. This type of volcanic material was commonly reduced to produce chipped tools (e.g., cutting tools such as axes and arrowheads).

From an archaeological point of view, two possible hypotheses on their procurement in the Roman period can be made: (a) the obsidian was already present in the vicinity of the site of Nora (e.g., as a waste or residue of previous processing of the material) and (b) it was extracted from an ancient quarry of Mt. Arci outcrops, which was well-known for the high presence of obsidians.

Considering the widespread Sardinian road network in the Roman time (Fig. 1; Mastino 2005) and their intensive use in the theatre of Nora, it is conceivable to imagine a procurement of these obsidians from the Mt. Arci area. In fact, in agreement with the archaeological literature, the road connecting the Roman villages of Karales (today Cagliari city, south Sardinia; Fig. 1) and Turris Lybisonis (today Porto Torres, NW Sardinia; Fig. 1) passed near the eastward side of Monte Arci volcanic complex. Moreover, in the middle of this ancient road stood the archaeological site of Forum Traiani (today Fordongianus village; Fig. 1) and the Roman village of Mulargia (Fig. 1), both well-known for its thermal baths and for its ancient quarries of ignimbrite rocks, respectively. Mulargia is a very important Roman site for the production of Roman millstones, exported in various parts of the Mediterranean (e.g., North Africa, Antonelli et al. 2014a) and for the existence of ancient commercial routes of different kinds of stone. Moreover, many Roman public buildings, villae and rural settlements are known in close proximity to the many obsidian sources scattered along the western, southern and north-western flanks of the Monte Arci volcanic massif. For these reasons, it is not unlikely to assume a gathering and transportation activity, even if occasional, of obsidians along the Roman road network.

However, in the frame of a long-term archaeological project on Monte Arci obsidians carried out since 2002 (Lugliè 2003, 2013), including systematic and careful surveys along the north-eastern flanks of the mounting around the so-named Perdas Urias outcrop, no evidence was found thus far to support the local exploitation of this raw material during later historic times. In spite of this intense archaeological exploration of the direct supply area, the possibility of an intended gathering activity of obsidian by Romans still remains in the state of an argumentum ex silentio.

Therefore, the first hypothesis on the obsidian origin seems to be by far more likely, because of the unicity of this case and due to the presence of some prehistoric settlement in the Nora area or its surroundings as in the S’Abuleu region (Migaleddu 1996) and, possibly, in the same site of Nora (Lugliè 2009b). In fact, during the last prehistory in these two settlements, the obsidian reduction of shaped cores transported from the Monte Arci workshops was a daily activity to produce a plentiful of artefacts and specialized tools. Moreover, obsidian is commonly found in the form of finished tools in Nora territory even in later Nuragic times; in fact, indeed, all around the Roman Nora town, many Nuragic sites are scattered in the plain and on the hills, among which stands out the complex tower and village of Antigori (near to the Sarroch city). This is a large site occupied from the 14th to 8th centuries B.C. (Russell 2010; Balmuth 1992) with castle-like structures, heavy multi-towered walls and associated villages, which Webster (1996) labeled as one of 14 known Class III settlements in Sardinia. Like in this last case, reoccupation of Bronze age Nuraghi in Roman times for civil and ritual purposes is a common behaviour that for sure could have put in contact Roman people with abundant obsidian elements.

Conclusions

The research allowed us to define the geochemical characteristics and the probable provenance of obsidian glasses used in the ancient mortars of the Roman Nora theatre.

This kind of volcanic glass, together with a quartz-feldspar sandy and subordinately local Oligo-Miocenic calc-alkaline volcanics and Palaeozoic rocks, was used as mortar aggregate for the different sectors of the theatre (i.e., tribunalia vaults, cavea foundation, vaults and inner walls of the external niches, brick and stone walls). Considering its low bulk density, it is probable that it was used also to lighten some structures (e.g., concrete of tribunalia vaults), together with sub-decimetric fragments of other local stones (i.e., volcanic rocks and sandstones), which has also low bulk density due to their high porosity. Given its chemical–physical features, characterised by an amorphous state reacting with the binder, it was also used as pozzolana, conferring hydraulic properties and high mechanical strengths to the mortars and concretes of the monument, especially in the structural parts where it has been intensively used.

Given the good hydraulic features and their massive use in the theatre, it is probable that the Romans intentionally used the volcanic glasses in place of cocciopesto, to make the mortars much resistant to weathering. Moreover, the unusual presence of volcanic fine aggregate in the plasters of the open-air cavea sector of the theatre confirms the intention of Roman constructors to improve the hydraulic characteristics of the mortars and their resistance to decay processes. It is not accident that the theatre, an open construction designed without roofing, is also one of the best preserved buildings in the Nora site.

The use of these obsidian glasses as pozzolana today represents a novelty in the production of Roman mortars in the archaeological site of Nora, especially considering that in the area surrounding Nora, the obsidian is not available as a raw material.

The results of discriminant analysis highlights that the volcanic glasses show a geochemical similarity with the volcanics of Perdas Urias outcrops, indicating a provenance from Mt. Arci volcanic complex, a well-known source as early as the Neolithic period (Lilliu 1988; Lugliè 2009a). Since an archaeological evidence for the obsidian procurement directly from the source in Roman times is still lacking, this hypothesis on the origin of glasses opens new scenarios from a historical and cultural point of view, raising up new and interesting issues.

From a technical-constructive point of view, it opens up further interesting research topics to understand if the use of obsidian rocks as aggregate in the mortars of a theatre (a) was a local experiment to make the hydraulic mortars of theatre or (b) extends to other buildings of the Nora site and/or in other Sardinian Roman settlements. Considering that currently there is no archaeological evidence about both the use of these obsidian glasses in other Sardinian Roman sites and about the presence of any Roman raw material procurement in the source area, at the moment it is unlikely to suppose that the Romans supplied obsidians directly from Mt. Arci. Assuming the first case, Romans probably used shattered and discarded artefacts from a local pre-existing “source” of obsidian, consisting in an earlier production activity during Neolithic times in the Nora area. In fact, the obsidian processing wastes are usually abundant in these prehistoric sites and well known in the archaeological literature, so it is easy to suppose a later exploitation of crushed earlier obsidian artefacts as temper for the mortars to be used in a single building like the theatre.

In any case, the detection of these glasses inside the mortars of the theatre raises a strong interest in the development of new archaeological investigation and further geochemical and petrographic studies either to check the possible use of Mt. Arci obsidians in other Roman Sardinian sites or, more likely, to find further evidence of their prehistoric storage and processing in the Nora area.