The colours were rarely anything beyond shades of green and blue. These were produced by compounds of copper; the blue is especially free from iron, which even in traces produces a green tint. The blue if exposed to damp fades white; the green changes to brown, owing to the decomposition of green silicate of iron and the production of brown oxide of iron. This decomposition may go on beneath an unbroken polished face of glaze, changing the glaze to brown. The shades of blue and green were all experimentally produced in modern times by Dr Russell, F.R.S., who succeeded in exactly copying the purple blue, full blue, light blue and French blue, and the green-blues and full greens in more than a hundred tints. The method was indicated by the half-baked pans of colour found at Tell-el-Amarna. Quartz rock pebbles had been collected, and served for the floor of the glazing furnaces. After many heatings which cracked them they were pounded into fine chips. These were mixed with lime and potash and some carbonate of copper. The mixture was roasted in pans, and the exact shade depended on the degree of roasting. The mass was half fused and became pasty; it was then kneaded and toasted gradually, sampling the colour until the exact tint was reached. A porous mass of frit of uniform colour results. This was then ground up in water, and made into a blue or green paint, which was either used with a flux to glaze objects in a furnace, or was used with gum or white of egg as a wet paint for frescoes. The ovens were small, about two or three feet across; cylindrical pots were set upside down and a fire lighted between them, and the pans of colour rested on the bottom edges of the pots. In Roman times the glazing furnaces were about eight feet square and deep, with an open arch to windward half way up. The vases and dishes were stacked in the furnace upon cylinder pots, and the successive dishes in the piles were kept apart by cones of pottery nearly an inch high. The failure of a furnace-load has revealed the system; by too long heating the glaze soaked through the porous body, and it all settled down and partly fell to pieces. The other colours used were: for the red a body mixed with haematite and covered with a transparent glaze; bright yellow, the composition of which is unknown; violet in various depths, from a faint tinge on the white lotus petals to a deep strong colour, probably made by copper blue and one of the purples; purple in various strengths from a rich bright tint upon white to a black purple for designs upon blue, all produced by manganese; occasionally purple-blue made with cobalt; dead white, which was doubtless produced by tin as at present.