Comparison of different kinds of models to simulate the alteration of medieval stained glass as a function of climate and pollution

A large number of stained glass windows were installed from the 13^th century. During the Middle-Ages, most of glass pieces have a Si-K-Ca composition with a relatively low SiO₂ content, but high content of K₂O and CaO.  This chemical composition means that medieval stained glass deteriorates during environmental exposure, from climate and environmental pollution. These alterations are manifested in the form of an alteration layer and secondary phases (mainly gypsum or syngenite). The alteration layer is generally depleted in K and Ca, but rich in Si, Al and Fe. Its thickness varies up to 300 µm after 6 or 7 centuries of alteration. In order to reconstruct the alteration history and predict the deterioration of stained glass windows in the future, it is necessary to determine alteration rates as a function of the climate and environmental parameters.

Several methodologies can be used to achieve this. First, short-term exposures or laboratory experiments can assess the first stages of the alteration and short-term kinetics. From these results, dose-response functions (DRF) were established for sheltered and unsheltered rain conditions. They correlate relevant environmental factors (temperature, rain quantity, rain pH, relative humidity, SO₂ concentration) with the response of the materials in terms of alteration layer thickness. The second methodology consists in laboratory experiments that aim at parametrizing kinetic laws as a function of specific parameters (temperature, pH of rain, and relative humidity). These kinetic parameters do not directly consider pollution, but they can be extrapolated over long periods and can be inputs to geochemical models. In this study, we have compared both methodologies to simulate the alteration of a model stained glass at different European sites (using data from the ICP-Materials program). Both models give good results, but the geochemical model tends to underestimate the alteration at polluted sites. This indicates that the pollution via the concentration in SO₂ for example should be included to improve the model.