Visitors to European churches often comment on the colours of stained-glass windows, but few realise that those colours are the product of chemistry as much as of art. A medieval window is a composite of many thin panels of glass, each of which has been coloured through the deliberate introduction of trace metals during manufacture. The craft of producing these colours was developed over centuries, lost for long periods and reconstructed fragment by fragment in modern laboratories.

Glass is at its most basic a mixture of silica (sand), soda (sodium carbonate) and lime (calcium carbonate), melted together at around 1,100 degrees Celsius. In its pure form, the resulting material is transparent and nearly colourless, and it is the addition of small quantities of other substances that produces the visible spectrum of coloured glasses. Cobalt oxide, for instance, gives a deep blue at concentrations of less than one percent; copper produces green or red depending on the precise conditions of melting; gold, dissolved in tiny quantities, produces a rich ruby; iron, the most common impurity, gives greens and browns that medieval glass-makers sometimes struggled to avoid.

The conditions of melting mattered as much as the added metal. The red colour produced by copper, in particular, depends on a narrow range of temperatures and on control of the atmosphere inside the furnace. Too much oxygen and the copper forms green compounds; too little and the glass remains colourless. Medieval workshops used a careful balance of wood fuel and ventilation to maintain the reducing conditions that favoured red. Because the sensitivity to process was so high, red glass often shows bands of different shades in a single pane, the record of small fluctuations in the furnace during manufacture.

Manganese had another role. Iron impurities in the raw sand inevitably tinted the glass green, and this was unwelcome in cases where a neutral window was desired. Adding a small quantity of manganese oxide reversed the tint and produced a nearly colourless glass, provided the proportions were right. Too much manganese, however, turned the glass purple after long exposure to sunlight - a phenomenon sometimes visible today in antique glassware whose original maker intended it to be clear.

These facts were largely lost during the later Middle Ages, when stained-glass production declined along with the major cathedral workshops. Much of the technical knowledge survived only in fragmentary notebooks. Chemist Dr. Beatrice Ostrowska, working at the University of Krakow, has spent two decades reconstructing medieval recipes by systematically varying the composition of experimental batches and matching the optical properties of the results to those of surviving windows. Her work, published over many papers, has shown that the range of colours available to a medieval glass-maker was narrower than sometimes claimed, and that much of the visual richness of a thirteenth-century window was achieved by combining relatively few base colours in clever designs rather than by exotic additives.

Not all of the colour comes from the glass itself. Medieval makers also painted detail on the outside of the finished panes using a glass-based paint called grisaille. Grisaille contained finely ground glass, an iron or copper pigment, and a binder; once applied, the pane was returned to the furnace, where the paint fused into the surface. In this way, facial features and folds of drapery could be rendered on what would otherwise have been flat areas of colour. Modern conservators have to work carefully around these painted layers, which can be damaged by well-intentioned cleaning.

The restoration of stained glass is now a substantial specialism. Many of the great windows in Europe have been dismantled at least once since the Second World War, cleaned, repaired and reinstalled behind protective secondary glazing that reduces temperature fluctuations and the impact of polluted rain. Dr. Ostrowska's work has been important for this effort, because replacement panels made with a similar palette and composition to the original age more like it, producing a less visible repair. The colours of old glass, it turns out, are not simply old; they are still being understood.