Scholarly article on topic 'Colour and Chemistry of the Glass Finds in the Roman Villa of Treignes, Belgium'

Colour and Chemistry of the Glass Finds in the Roman Villa of Treignes, Belgium Academic research paper on "History and archaeology"

CC BY-NC-ND
0
0
Share paper
Academic journal
Procedia Chemistry
OECD Field of science
Keywords
{Glass / Redox / Iron / Colour / CIELAB / "UV Absorption Edge" / Roman / "Material Culture Studies"}

Abstract of research paper on History and archaeology, author of scientific article — Andrea Ceglia, Wendy Meulebroeck, Peter Cosyns, Karin Nys, Herman Terryn, et al.

Abstract Optical spectroscopy allows the identification of ionic species and, under certain conditions, the quantification of Fe + and Fe3+. The ratio of the oxidation states of iron gives an insight into the technological aspects of production. Moreover from the transmission spectra it is possible to calculate the CIE Lab colour coordinates and the UV absorption edge. The latter parameter is strongly related to the presence of heavy elements because they disrupt the silica polymer network. The optical parameter highlights differences in the sample population allowing the definition of subgroups. A comparison between colour coordinates, iron redox ratios, UV absorption edge and the chemical composition is presented. The results provide important information about the proportion between different compositional groups available from the archaeological site and underline the potentiality of UV-Vis-NIR absorption spectroscopy as a first-step screening method for large sets of archaeological or historical glass fragments. The present case-study demonstrates the results of optical spectroscopy on a selection of 16 late Roman “naturally” coloured glass fragments from the Roman villa complex ‘les Bruyeres’ in Treignes (Belgium).

Academic research paper on topic "Colour and Chemistry of the Glass Finds in the Roman Villa of Treignes, Belgium"

Available online at www.sciencedirect.com

SciVerse ScienceDirect Procedía

Chemistry

Procedía Chemistry 8 (2013) 55 - 64

Youth in Conservation of Cultural Heritage, YOCOCU 2012

Colour and Chemistry of the glass finds in the Roman villa of

Treignes, Belgium

Andrea Cegliaa,b,c*, Wendy Meulebroecka, Peter Cosynsc, Karin Nysc, Herman Terrynb,

Hugo Thienponta

aDepartment of Applied Physics and Photonics, Brussels Photonics Team B-PHOT, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels,

Belgium

bDepartment of Electrochemical and Surface Engineering, Materials and Chemistry, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels,

Belgium

cDepartment of Art Sciences and Archaeology, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels, Belgium

Abstract

Optical spectroscopy allows the identification of ionic species and, under certain conditions, the quantification of Fe2+ and Fe3+. The ratio of the oxidation states of iron gives an insight into the technological aspects of production. Moreover from the transmission spectra it is possible to calculate the CIE Lab colour coordinates and the UV absorption edge. The latter parameter is strongly related to the presence of heavy elements because they disrupt the silica polymer network. The optical parameter highlights differences in the sample population allowing the definition of subgroups. A comparison between colour coordinates, iron redox ratios, UV absorption edge and the chemical composition is presented. The results provide important information about the proportion between different compositional groups available from the archaeological site and underline the potentiality of UV-Vis-NIR absorption spectroscopy as a first-step screening method for large sets of archaeological or historical glass fragments. The present case-study demonstrates the results of optical spectroscopy on a selection of 16 late Roman "naturally" coloured glass fragments from the Roman villa complex 'les Bruyères' in Treignes (Belgium).

© 2013 The Authors. Published by Elsevier B.V.

Selection aad peer-review under responsibility of the IA-CS (Italian Association of (Conservation Scientists) and University of Antwerp Keywords: Glass; Redox; Iron; Colour; CIELAB; UV Absorption Edge; Roman; Material Culture Studies

* Corresponding author.

E-mail address: aceglia@vub.ac.be or and.ceglia@gmail.com.

ELSEVIER

1876-6196 © 2013 The Authors. Published by Elsevier B.V.

Selection and peer-review under responsibility of the IA-CS (Italian Association of Conservation Scientists) and University of Antwerp doi:10.1016/j.proche.2013.03.008

1. Introduction

Due to the growing interest in material culture studies to better understand the consumption pattern in the ancient world, the necessity for archaeometric research on glass from archaeological and historical contexts becomes more preponderant. The obtained results can be linked to the recorded external features in order to match the archaeometric and archaeological issues. The investigation of the manufacturing techniques and the type of raw materials used show to be increasingly decisive in determining the provenance of the glass. The ongoing research on late Roman and early Byzantine glass focuses on the benefit of UV-Vis-NIR and Raman spectroscopy to the study of late antique glass production and consumption [1-10]. In the first place as an initial in-situ evaluation of the bulk material from an excavation preceding a well-defined selection of samples for an ex-situ chemical analysis.

It is widely known and accepted that in the first millennium AD, the economy was highly structured and centralized. The production and distribution of glass does not make an exception. Large primary glass workshops on the Syro-Palestinian coast and in the Egyptian Nile-delta were able to make up to several tons of raw glass, even though other production centres located elsewhere than Egypt or the Syro-Palestinian coast must have existed [11-12]. However the south-eastern Mediterranean is a strategic location in proximity of the raw materials: calcareous sand and mineral soda. Sand is the source of silica, the glass former oxide, while the calcareous portion - like shells - provides stability to the material. However, the melting temperature of sand was difficult to be reached with ancient furnace technologies, hence, other substances, called flux, were added to reduce the needed melting temperature. This capability was found in Natron, a mixture of sodium salts obtained from evaporitic lakes, mainly located in Egypt [13].

The primary produced raw glass were subsequently crushed into large irregular chunks to ship it throughout the entire Roman empire to supply the secondary workshops where the raw glass got remelted and shaped into market ready objects such as vessels, jewellery and window panes.

It is possible to improve this generalized and simplified picture on glass production in antiquity through more intense chemical analyses, as much of the progress of ancient glass research has been made possible thanks to the archaeometrical approach. The most common techniques employed to define the chemical composition of archaeological glass are SEM-EDX [14], EPMA [15], LA-ICP-MS [16-17] and XRF [18], but in the last decade, other techniques such as Raman spectroscopy [1-2, 7, 19-21], PIXE/PIGE [22-23] and X-rays absorption spectroscopy [24-25] have started to be more and more used for the study of historical vitreous materials. Though it is not frequently used, UV-Vis-NIR absorption spectroscopy is an analytical method that has the potentiality to be very beneficial for the study of ancient glass [3-6, 8-10, 26-28]. The application of UV-Vis-NIR brings several advantages in the study of ancient glass. First of all, as some of the methods mentioned above may be costly and destructive, a limited number of glass samples are generally selected for ex-situ lab analysis. This is not the case with UV-Vis-NIR spectroscopy, which, similarly to Raman and XRF, can be used on the entire object prior to sampling and, if necessary or compulsory, in situ without the necessity to sample the artefact. Although chemical techniques, such as SEM-EDX and ICP-MS, are able to determine the elemental composition, glass colour is affected by more parameters than the sole chemical composition. Besides the total amount of colourant, amount of non-bridging oxygens, oxidation state and oxygen/sulphur coordination play a crucial role in glass colouring by Mn and Fe. Optical analysis can offer an insight in colour chemistry, redox properties. Furthermore it is a fast technique and thanks to the availability of portable instrumentation the technique can be applied for in situ glass research. The applicability in situ makes optical absorption spectroscopy appropriate for a first-line analysis of large, unexplored glass collections. However, on the limitations side, it must be taken into account that in order to collect valid measurements it is preferable to clean the surface of the sample to remove crusts, dirt and corrosion.

2. Experimental

The glasses analyzed in this paper are selected among the late Roman findings of the villa 'les Bruyères' located in Treignes, Belgium dating from the late 3rd to the beginning of the 5th century AD [29]. The villa complex demonstrates a building history in two stages as generally recognized in the villas in northern Gaul: a first phase dates back from the mid to second half of the 1st century, when the first large villa structures got constructed and lasts until the destruction of the villa complex by the end of the third quarter of the 3 rd century AD; a second phase started around the end of the 3 rd or the start of the 4th century AD when the villa complex was rebuilt and remained in use until a final destruction by the start of the 5th century AD. The glass fragments recovered have a wide date range from the late 1st century AD up to the very beginning of the 5th century AD. Nevertheless, most glass material is to be dated to the second phase when the remains of the former villa complex got levelled and intensely rebuilt. This observation is supported by the coins retrieved from the villa [30].

A total of 16 glass fragments were analyzed. They were selected from window and vessel pieces ranging from pale green to pale blue hues, with only one piece having a stronger yellow-green tinge. The thickness of the material varied from less than 1 mm to 4 mm.

Optical transmission spectra are recorded with a high-performance optical spectrum analyzer (SPECTRO 320D, Instrument Systems) between 300 and 1600 nm with a spectral resolution of 1.5 nm. The scan rate is set at 10 ms/nm and the final spectrum is the average of three acquisitions. Two lamps are used as light source: a 30-W deuterium lamp for the UV region (emitting between 190 nm and 400 nm) and a 20-W halogen lamp for the VIS-NIR (emitting between 350 and 1700 nm) region. The light is guided from the sources towards the sample by means of an optical fibre bundle transmitting light in the spectral range between 300-2200 nm. The beam is focused onto the sample through a plano-convex fused silica lens, producing a spot size of about 2 mm on the glass surface. The transmitted light is received in an integrating sphere of 30 mm diameter (Avantes) connected to an optical fibre, which guides the light towards the entrance slit of the optical spectrum analyzer. The minimum sample size required for the analysis is 5x5 mm. The optical resolution of the instrument is given as uncertainty on the spectral position. However, other aspects, such as the stability of the source, play an important role on the quality of the data. Hence, two series of repeatability measurements were carried out using modern cobalt-coloured glass: first, a series of 11 measurements were captured on the same session day; second, 11 measurements were performed on different days. The standard deviations for the cobalt peaks in the visible region were lower than 1.4 nm [3].

From the spectra several measurands can be calculated that are useful for the classification of the glasses. First of all the ferrous-ferric redox ratio can be quantified using the formula defined by Bamford [31]:

OD1050 and OD380 are the optical densities correspondingly at 1050 and 380 nm. Hereafter the ratio of the redox state of iron is reported as percentage of ferrous iron with respect to the total amount of iron (Fe2+/Fetot). This value is calculated using the following conversion formula:

Fe2+ „ ,„JOD1050- 0.03 6Ï

Fe2+/Fetot is preferred because it is expressed in percentage that allows an easier comparison than decimal numbers.

Another parameter that can be extrapolated is the UV absorption edge (UVAE). This is the wavelength at which a certain percentage of the incident light is transmitted through an arbitrary thickness of the sample. Although Scholze [32] recommends using a transmittance of 10% or 50% for 5 mm thick material, in this work we will use transmittance of 20% and spectra are normalized to 2 mm thickness (Figure 1). The choice of these values is dictated by the need to avoid specific absorption bands. Normalizing to 5 mm of thickness, the absorption edge would fall in the region between 400-440 nm where there are the absorption bands of ferric iron. However, additional research has to be undertaken to decide which set of values is the best to use for historical glass.

The last parameter used in this paper is the calculation of the colour with the CIE La*b* system, where L stands for lightness ranging from 0 for opaque materials to 100 for air, while a* and b* are the colour-opponent dimensions. Changing from negative to positive values of a* the colour moves from green to red, whereas the colour coordinate b* represents blue hues when it is negative and yellow when it is positive. A concordance list between the qualitative results from absorption spectroscopy needs correlation with the quantitative measurements from SEM-EDX analysis.

To determine the major chemical composition of the artefacts a JEOL JSM-6400 (JEOL LTD) Scanning Electron Microscope coupled with Energy Dispersive X-rays spectroscopy was employed. To analyze properly the bulk composition, the glass samples were inserted in epoxy resin, polished, and carbon coated; all the measurements were carried out on the freshly made cross-section area. All the spectra were collected under the same experimental conditions. To perform standard based quantitative analysis the beam current must be the same for the unknowns and the standards. Unfortunately our system misses a Faraday cup. Hence, to overcome this problem we measured the absorbed current on pure metallic copper to stabilize the beam current. The voltage was set at 20 kV to excite all the elements present in the glass samples and emitted X-rays were recorded for 100 seconds of live time. The spectra were analyzed using the NIST released software DTSA-II. Pure elements and oxides were used to build a library of reference spectra, while Corning glass A, Corning glass B, NIST SRM1830 and SRM620 were employed as standards for quantification.

3. Results and discussion

Table 1 reports the chemical data obtained by means of SEM-EDX analysis and the parameters extrapolated from the optical spectra, respectively the iron redox ratio Fe2+/Fetot, the colour coordinates L, a* and b*, and the UV absorption edge.

All the glasses analyzed have a soda-lime-silica base composition, though they can be divided in three main chemical clusters according to the relative proportion of the main components and the iron redox ratio obtained by optical spectra: Group 1, Group 2 and HIMT. Two samples are only tentatively assigned to Group 1 and Group 2. Figure 1 shows the transmittance spectra of each group.

Group 1 consists of pale blue glasses with about 18 wt% of Na2O and 6.5 wt% of CaO. Despite the glasses belonging to this group have about 0.5 wt% of MnO and 0.7 wt% of Fe2O3; they are highly reduced, having approximately 40% of ferrous ions. The effect on the optical spectra is a strong absorption at about 1050 nm.

Group 2 is made of pale green glasses which are poorer in sodium oxide, about 16 wt%, and richer in lime, roughly 8.2 wt%. This group differs from the previous one also because of the much higher manganese content ranging from 1.4 to 1.8 wt% of MnO. As a consequence the fragments of this group are more oxidized of the glasses of Group 1. On the basis of the manganese content and Fe2+/Fetot, Group 2 can be further divided in Group 2a and Group 2b. The former has less than 10% of iron in the ferrous state and 1.8 wt% MnO, while the latter

group has higher ferrous content and about 1.5 wt% of MnO. The optical spectra of glasses of Group 2 are characterized by a higher contribution of the ferric bands (380-450 nm) respect than Group 1.

Though there is some dissimilarity, it seems reasonable to identify Group 1 with the Roman blue-green glass produced in the 1st-3rd century AD [11] and Group 2 with the Levantine 1 group made on the Syro-Palestinian coast in the late Roman-early Byzantine period [14].

There are two samples, TR1s5 and TR2s7, which are deviant from the two main groups. Based on the initial visual inspection they were assigned to Group 2 because of their very pale green hue. However, chemical analysis shows that they have some intermediate features between the two groups. The amount of Na2O and Al2O3 suggests that they belong to Group 1 and Group 2 respectively, the amount of calcium and manganese oxides are midway between the two groups. Also their in-between Fe2+/Fetot supports this conclusion. They may be the product of glass recycling.

Finally only one fragment has been assigned to the HIMT group. The name, given by Freestone [14], is an acronym for High Iron, Manganese and Titanium oxides. This glass type is apparently already circulating from the late 3rd century AD [33] has been found in several other archaeological sites dated after the half of the 4th century AD. It is believed that this glass was produced in Egypt but clear archaeological evidences are still missing for the moment. The HIMT sample has a very low Fe2+/Fetot, with only 2% of iron in its reduced state. However, because of the high amount of total iron, its absorption is still easily detected by optical spectroscopy. Summing the intense charge transfer bands in the UV tail out the violet/blue part of the visible spectrum, the resulting hue is yellow green typical of this glass type.

The colour of Roman glass is generally ranging from yellowish to greenish to bluish hues with all the possible intermediate shades. Despite the wide range of colours, in all these possible combinations the transition metal ion responsible of the colour is always iron. This element is introduced into the batch as an impurity of the sand. Iron can exist in glass in two oxidation states, Fe2+ and Fe3+. Both species can actually exist in tetrahedral and octahedral coordination, complicating the readability of optical spectra. However, it is assumed that because of the relative chemical homogeneity of Roman glass and the use of very similar furnace technologies throughout the whole empire, the fraction of octahedral and tetrahedral remains unchanged for all the glasses [27].

The ferrous specie absorbs very strongly in the infrared region, at around 1050 nm. Tailing into the visible, Fe2+ can be at the basis of several colours from blue to red. Fe3+ give rise to several weak absorption bands in the range 380-450 nm [34]. Moreover both ionic species absorb intensively the UV light because of charge transfer transitions with the ligands, namely non-bridging oxygen (NBOs). The ratio between the two redox species and the total amount of iron define the final colour of the glass. The iron redox ratio is linked to the chemistry of the glass, to the fabrication technology and oxygen fugacity.

Because the chemical composition of glass affects the redox condition experienced by iron ions, it is then a straightforward consequence that the colour is influenced too. Iron is oxidized by other elements present in the batch such as antimony, chromium and manganese [31]. In the case of the glasses of the Roman villa of Treignes, manganese is the main oxidizing agent.

Figure 2 is the plot of Fe2+/Fetot against the content of manganese oxide. Excluding the HIMT glass, for which the strong oxidation cannot be explained only in terms of manganese content, all the glass samples show a linear correlation between these two parameters. It is clear that the higher is the amount of manganese added to the batch the higher is the portion of iron that is converted to the weakly coloured ferric state according to the following chemical redox reaction:

Fe 2 {blue) + Mn +(p urpiee Fe 3 (yellowish) {colorlesS

The b* colour coordinate is strongly affected by the oxidation state of iron. This parameter explains the contribution from blue to yellow hues, respectively for negative values and positive values. The glasses belonging

to Group 1, with lower MnO2 and higher Fe2+/Fetot ratio, have negative values of the colour coordinate b*, reflecting their blue hue. On the contrary, Group 2 shows more yellow positive values of b*. Samples TR1s5 and TR2s7, which have intermediate values of manganese, lie between the two groups in figure 2. The redox reaction between manganese and iron cannot fully explain the strong oxidation in HIMT glass. It is possible that different fabrication conditions were applied, or the other reducing elements not detected with SEM-EDX may give origin to a higher yield of ferric iron.

Another useful parameter that can be extrapolated from the optical spectra is the UV absorption edge. Figure 3 compares the UVAE and the sum of manganese and iron oxides. These two parameters have a positive correlation. Although the UV transmittance of glass is affected by several factors, such as melting techniques and temperature, the presence of heavy elements is the main cause. Network modifiers increase the number of non-bridging oxygen atoms which are responsible for the absorption in the near UV region [32, 35]. In the glasses under study, manganese and iron are the most abundant; therefore elements such as lead, zinc, titanium, cobalt, antimony, chromium, copper etc. are neglected. Nevertheless, the presence of these elements may become important when glass is recycled, for their enrichment due to contamination with coloured glass during the recycling process [3, 14].

4. Conclusions

Research on the application and on the benefit of UV-Vis-NIR on archaeometrical glass research is still ongoing. Additional tests are needed to investigate all the variables that can affect optical spectra. Therefore, case studies like the villa of Treignes are useful for testing the efficiency of the methodology. This paper provided evidence of possible applications of optical spectroscopy beyond the identification of colouring agents.

From optical spectra of glass it is possible to extrapolate several parameters such as the iron redox ratio, the UV absorption edge and the colour coordinates. The iron redox ratio can be related to the base chemistry of the glass or to the presence of redox agents such as manganese. Also the UV absorption edge is linked to glass chemistry as it is affected by the presence of transition metal ions, particularly iron. Certain metal impurities also contribute to the absorption of UV-Vis-NIR, influencing the overall spectral shape. The resultant of all this effects is a specific glass colour.

The analysis of UV-Vis-NIR spectra of the fragments of the villa of Treignes revealed three distinct groups plus two deviant fragments with intermediate redox characteristics. Chemical analysis by means of SEM-EDX confirmed the differences identified by optical analysis. The three groups may be associated with known chemical compositions of ancient times, namely Roman blue-green, Levantine and HIMT glass. These three compositional groups were in use in different periods. Their identification may help in dating the context of an excavation.

Furthermore, UV-Vis-NIR absorption spectroscopy can be a valid help in the study of archaeological glass material prior to sampling for chemical analysis. Very often, the sample scheme is based on a simple evaluation of the colour by naked eye. However, this may be misleading because the thickness of fragments influences strongly the perceived colour. Moreover, small spectral differences such as the one between Group 2a and Group 2b are impossible to be detected.

5. Acknowledgements

The research leading to these results has received funding from the European Union Seventh Framework Programme FP7/2007-2013 under grant agreement n° 265010. The authors are also thankful to the fruitful

contribution of an anonymous reviewer. Our sincere gratitude goes to Pierre Cattelain from the Musée du Malgré-Tout, Treignes Jean-Marc Doyen from the Centre de Recherche Archéologique (CReA -ULB), Brussels for granting us permission to study and sample the glass material from the Roman villa "les Bruyères" at Treignes.

6. References

[1] Baert K, Meulebroeck W, Wouters H, Cosyns P, Nys K, Thienpont H, Terryn H, Using Raman spectroscopy as a tool for the detection of iron in glass J Raman Spectrosc 2011;42:1789-1795.

[2] Baert K, Meulebroeck W, Wouters H, Ceglia A, Nys K, Thienpont H, Terryn H, Raman spectroscopy as a rapid screening method for ancient plain window glass J Raman Spectrosc 2011;42:1055-1061.

[3] Ceglia A, Meulebroeck W, Baert K, Wouters H, Nys K, Thienpont H, Terryn H, Cobalt absorption bands for the differentiation of historical Na and Ca/K rich glass Surf Interface Anal 2012;44:21^-226.

[4] Meulebroeck W, Baert K, Wouters H, Cosyns P, Ceglia A, Cagno S, Janssens K, Nys K, Terryn H, Thienpont H, The identification of chromophores in ancient glass by the use of UV-VIS-NIR spectroscopy Proc SPIE 2012;7726:77260D.

[5] Meulebroeck W, Cosyns P, Baert K, Wouters H, Cagno S, Janssens K, Terryn H, Nys K, Thienpont H, Optical Spectroscopy as a Rapid and Low-cost Tool for the First-Line Analysis of Glass Artefacts: A step-by-step plan for Roman Green Glass JArchaeol Sci 2011;38: 23872398.

[6] Meulebroeck W, Wouters H, Baert K, Ceglia A, Nys K, Terryn H, Thienpont H, Optical spectroscopy applied to the analysis of medieval and post-medieval plain flat glass fragments excavated in Belgium Proc SPIE 2010;7726:77261E.

[7] Baert K, Meulebroeck W, Ceglia A, Cosyns P, Wouters H, Nys K, Thienpont H, Terryn H, The potential of Raman Spectroscopy in glass studies, Proc. SPIE 8422IASHG 2012;8422:842207.

[8] Ceglia A, Meulebroeck W, Baert K, Wouters H, Nys K, Terryn H, Thienpont H, Using optical spectroscopy to characterize the material of a 16th c. stained glass window, Proc. SPIE 8422 IASHG 2012;8422:84220A.

[9] Meulebroeck W, Baert K, Ceglia A, Cosyns P, Wouters H, Nys K, Thienpont H, Terryn H, The potential of UV-VIS-NIR absorption spectroscopy in glass studies, Proc. SPIE 8422 IASHG 2012;8422:842208.

[10] Cosyns P, Meulebroeck W, Thienpont H, Nys K, Potential prospects in archaeological research by using optical spectroscopy through a black glass ocular, Proc. SPIE8422 IASHG 2012;8422:842209.

[11] Jackson CM, Wager ECW, Joyner L, Day PM, Booth CA, Kilikoglou V Roman Glass-making at Coppergate York? Analytical evidence for the nature of production Archaeometry 2003;45:45-466

[12] Degryse P, Schneider J, Pliny the Elder and Sr-Nd isotopes: tracing the provenance of raw materials for Roman glass production J

Archaeol Sci 2008;35:1993-2000.

[13] Shortland A.J., Evaporites of the Wadi Natrun: Seasonal and annual variation and its implication for ancient exploitation, Archaeometry 2004;46:497-516.

[14] Freestone IC, Ponting M, Hughes MJ, The origins of Byzantine glass from Maroni Petrera Cyprus Archaeometry 2002;44:257-272.

[15] Schibille N, Marii F, Rehren T, Characterization and provenance of late antique window glass from the Petra Church in Jordan

Archaeometry 2008;50:627-642.

[16] Cagno S, Mendera M, Jeffries T, Janssens K, Raw materials for medieval to post-medieval Tuscan glassmaking: new insight from LA-ICP-MS analyses J Archaeol Sci 2010;37:3030-3036.

[17] Jackson C, Making colourless glass in the roman period Archaeometry 2005;47:763-780.

[18] Kato N, Nakai I, Shindo Y, Transitions in Islamic plant-ash glass vessels: on-site chemical analyses conducted at the Raya/al-Tur area on the Sinai Peninsula in Egypt J Archaeol Sci 2010;37:1381-1395.

[19] Oikonomou A, Triantafyllidis P, Beltsios K, Zacharias N, Karakassides M, Raman structural study of ancient glass artefacts from the island of Rhodes JNon-Cryst Solids 2008;354:768-772.

[20] Ricciardi P, Colomban P, Tournié A, Milande V, Nondestructive on-site identification of ancient glasses: genuine artefacts embellished pieces or forgeries? J Raman Spectrosc 2009;40:604-617.

[21] Colomban P, Polymerization degree and Raman identification of ancient glasses used for jewelry ceramic enamels and mosaics J Non-Cryst Solids 2003;323:180-187.

[22] Smit Z, Jezersek D, Knific T, Istenic J, PIXE-PIGE analysis of Carolingian period glass from Slovenia Nucl Instrum Meth B 2009;267:121-124.

[23] Weber g, Strivay D, Martinot L, Garnir HP, Use of PIXE-PIGE under variable incident angle for ancient glass corrosion measurements, Nuc. Instr and Meth in Phys Res B 2002;189;350-357

[24] Quartieri S, Riccardi MP, Messiga B, Boscherini F, The ancient glass production of the Medieval Val Gargassa glasshouse: Fe and Mn XANES study, JNon-Cryst Solids 2005:351;3013-3022

[25] Klysubun W, Thongkam Y, Pongkrapan S, Won-in K, T-Thienprasert J, Dararutana P, XAS study on copper red in ancient glass beads from Thailand, AnalBioanal Chem 2011;399;3033-3040.

[26] Schreurs JWH, Brill RH, Iron and sulfur related colors in ancient glasses Archaeometry 1984;26:199-209.

[27] Bingham P, Jackson C, Roman blue-green bottle glass: chemical-optical analysis and high temperature viscosity modelling J Archaeol Sci 2008;35:302-309.

[28] Green LR, Hart AF, Colour and chemical composition in ancient glass: an examination of some Roman and Wealden glass by means of ultraviolet-visible-infra-red spectrometry and electron microprobe analysis J Archaeol Sci 1987;14:271-282.

[29] Doyen, J.-M., 'Villa romaine à Treignes' in Cahen-Delhaye, A., De Lichtervelde, C., Leuxe, F. (eds.), Archéologie en Wallonie 19801985. Découvertes des cercles archéologiques. Namur, Fédération des Archéologues de Wallonie, 1987:266-271.

[30] Doyen, J.-M., Les monnaies antiques de la villa de Treignes (prov. de Namur, Belgique) : étude préliminaire, 2007. (e-paper accessible at www. cultura-ftp. com).

[31] Bamford CR Colour generation and control in glass Elsevier Amsterdam; 1977

[32] Scholze H, Glass Nature, Structure and Properties, first ed. Springer-Verlag, Germany; 1991

[33] Mirti P, Casoli A, Appolonia L, Scientific analysis of Roman glass from Augusta Praetoria, Archaeometry 1993;35:225-240.

[34] Weyl WA Coloured glasses Dawson's; 1959

[35] Novatski A, Steimacher A, Medina AN, Bento AC, Baesso ML, Andrade LHC, Lima SM, Guyot Y, Boulon G, Relations among nonbridging oxygen, optical properties, optical basicity, and color center formation in CaO--MgO aluminosilicate glasses J. Appl. Phys. 2008;104:094910

Table 1. Chemical values of the samples analyzed. All oxides are in wt% Mn+Fe and Mn/Fe are intended in oxides. Fe2+/Fetot is in percentage; L a* and b* are without dimensions, while the UV Absorption Edge (UVAE) is expressed in nm. Group 1 are pale blue fragments, Group 2a and Group 2b are pale green, HIMT is one yellow-green fragment. Both glasses Group 1? and Group 2? have a very pale blue-green tinge.

Sample Group Na2O MgO A^O3 SiO2 SO3 K2O CaO TiO2 MnO Fe2O3 Mn+Fe Mn/Fe Fe2+/Fetot L a* b* UVAE

TR1s1 Group 1 17.4 0.8 2.9 68.4 0.2 0.7 6.7 <0.1 0.6 0.7 1.4 0.8 44% 95.8 -7.0 -1.8 339.6

TR1s2 Group 1 17.9 0.8 2.7 68.5 0.2 0.7 6.6 <0.1 0.5 0.8 1.3 0.6 44% 95.8 -7.9 -2.4 341.0

TR1s3 Group 1 18.3 0.7 2.4 68.8 0.2 0.5 5.8 <0.1 0.2 0.5 0.8 0.4 42% 94.9 -7.1 -2.2 339.1

TR1s4 Group 1 18.1 0.8 2.5 68.8 0.2 0.6 6.4 <0.1 0.4 0.6 1.0 0.7 32% 96.0 -4.8 -0.6 337.9

TR1s6 Group 1 18.0 0.8 2.7 69.1 0.2 0.7 6.5 <0.1 0.4 0.7 1.2 0.6 43% 94.5 -7.1 -2.2 339.5

TR1s7 Group 1 17.6 0.8 2.7 66.9 0.2 0.7 6.4 <0.1 0.5 0.8 1.3 0.6 38% 92.9 -8.2 -2.9 342.4

TR1s8 Group 1 17.8 0.8 2.8 68.9 0.2 0.7 6.7 <0.1 0.6 0.7 1.3 0.8 40% 96.0 -7.6 -2.3 338.8

TR1s5 Group 1? 17.7 0.9 2.8 68.0 0.2 0.4 7.4 <0.1 0.9 0.5 1.4 1.7 21% 98.6 -4.3 1.8 339.7

TR2s7 Group 2? 16.4 0.9 3.5 69.3 0.1 0.4 7.8 <0.1 1.2 0.7 1.9 1.7 20% 96.5 -5.9 2.5 343.0

TR2s2 Group 2a 16.1 0.9 3.4 67.9 <0.1 0.4 8.2 <0.1 1.8 0.9 2.7 2.0 8% 96.7 -5.7 7.3 351.7

TR2s3 Group 2a 15.9 0.9 3.4 68.1 0.1 0.4 8.3 <0.1 1.8 0.9 2.7 2.0 9% 96.9 -5.7 7.5 352.1

TR2s4 Group 2a 15.9 0.9 3.5 68.2 0.1 0.4 8.3 <0.1 1.8 0.8 2.6 2.1 9% 98.0 -5.7 7.3 351.2

TR2s5 Group 2b 15.8 0.9 3.5 68.9 0.1 0.4 8.3 <0.1 1.4 0.7 2.1 2.0 12% 97.5 -5.0 4.4 344.9

TR2s6 Group 2b 15.9 0.8 3.6 68.7 <0.1 0.4 8.5 <0.1 1.5 0.7 2.2 2.1 13% 98.0 -5.2 4.5 345.5

TR2s9 Group 2b 15.7 0.9 3.5 69.0 0.1 0.4 8.4 <0.1 1.5 0.6 2.1 2.5 17% 97.0 -4.6 2.4 340.4

TR2s10 HIMT 19.0 1.1 2.7 65.1 0.3 0.5 6.0 0.2 1.5 2.2 3.7 0.7 2% 94.9 -9.4 24.6 392.6

400 600 800 1000 1200 1400 1600 400 600 600 1000 1200 1400 1600

Wavelength (nm) Wavelength (nm)

Figure 1. UV-Vis-NIR spectra of the different groups described in the text. For sake of clarity the spectra are divided in two graphs. The UV absorption band and the position of the absorption of Fe2+ and Fe3+ are reported. Group 1 shows a strong absorption due to Fe2+, while from Group 2 the three bands due to Fe3+ are more pronounced. The glass belonging to HIMT exhibits a very different spectral shape,

dominated by the charge-transfer bands in the UV.

Figure 2. Correlation between Fe2+/Fetot and manganese oxide content. A correlation line is shown for all the glasses of Group 1 and Group 2 excluding the HIMT glass whose behaviour is deviant.

Figure 3. Correlation between the UV absorption edge and the amount of Mn and Fe oxides. The higher the amount of the sum of the two transition metal ions the stronger is the red shift of the UV absorption Edge. Again HIMT glass has a deviant behaviour.