The purpose of this paper is to understand why manganese containing Roman glass could be purple or colourless in spite of having very similar chemical compositions. The strategy followed to tackle this question consists in the production of glass with the same chemical composition as Roman glass whereby various production parameters were controlled and systematically analysed. It is shown that redox and colour of glass is more likely to have been managed through internal control through the choice of raw materials and the addition of organic matter. The main difference between ancient and modern glass production relies on the lower melting temperature of Roman furnaces, so that sulphate would have played a less important role in the redox determination. 1 Secondly, a colour compensation occurs between the blue reduced iron and the purple oxidised manganese. This causes a general 'greying' of the glass where no wavelength is absorbed more than others, but the

Roman glass is studied here by means of optical absorption spectroscopy, in order to provide an objective method to quantitatively evaluate colour. The dataset is composed of 165 soda-lime-silicate glass samples from various western European sites, mainly dated from the 1 st to 4 th century AD, and containing variable amounts of iron, manganese and/or antimony. Iron redox ratios and colour coordinates (based on the CIELab colour system) are determined and put in relation with the thickness of samples and their manganese/antimony contents. Results reveal thickness as a crucial parameter when discussing glass hues, thus leading to a differentiation between the 'intrinsic' and perceived' colour of glass objects (i.e., the colour of the object with the thickness normalised to 1 mm, and that with its original thickness, respectively). Apart from HIMT and purple glass, the concentration of ferrous iron appears to be correlated with a*-a colourimetric parameter determining how green the glass is. Significant relations of antimony/manganese contents versus iron redox and glass colour are also considered, resulting in quantitative arguments to entitle antimony-decoloured glass as the most oxidised and colourless glass.

The purpose of this paper is to develop a means of quantifying glass recycling and to discuss the ‘anachronistic’ chemical composition of medieval blue window glass. This method relies on a new numerical method using kernel density estimates and is based on a database of published glass chemical compositions. It seeks to reveal when, to what extent and why blue tesserae were recycled for the production of French and English blue glass. First, it is suggested that blue glass had an ‘anachronistic’ chemical composition only before the 13th century. Second, the ‘anachronistic’ chemical composition of 12th‐century blue glass comes from the recycling of both blue tesserae and non‐coloured glass. Finally, this recycling was motivated by the scarcity of cobalt sources until mines were found in the 13th century.

This paper presents the analysis of decoloured and naturally coloured glass from well-dated contexts in the southwest corner of the Roman fort at Oudenburg (Belgium) ranging from the late second to the early fifth century AD. The aim is threefold. First, provide comparative material in the study of glass consumption from the northwestern provinces of the Roman Empire. Secondly, evaluate possible diachronic shifts in the applied decolourizing agent to produce colourless glass as to assess potential correlations between glass production recipes, provenance and chrono-typology. Finally, provide an added value to the research of glass recycling and mixing in the Roman imperial period. Nine subgroups are distinguished based on their chemical composition determined by LA-ICP-MS: Sb-only, two groups of Mn-only, four groups of mixed Mn-Sb, HIMT and one glass without any decolouring agent. The Sb-decoloured glass is used in the earliest phases and can be attributed to an Egyptian provenance. The two subgroups of Mn-glass likely come from different provenances: one from Egypt and the other later one from the Levant. Most of the glass shows high marks of mixing based on high trace elements concentrations and the simultaneous presence of antimony and manganese. Inhomogeneous mixing of manganese and antimony was also detected through μXRF. One Mn-Sb subgroup likely comes from mixing antimony glass with HIMT. The obtained results help better recognise the shifts in applied glass recipes throughout the Roman imperial period and improve our understanding about the mixing and recycling of glass to supply a Roman military camp.