In 1556 Georgius Agricola wrote, in his <i>De Re Metallica</i>, about the "colour of fumes" from various types of ores. This may be the earliest record available in which the color of a flame-excited metal sample was used to identify a substance. If so, it represents the start of flame spectroscopy as an analytical tool.
W. G. Schrenk, "Historical Development of Flame Excitation Sources for Analytical Spectroscopy," Appl. Spectrosc. 40, xix-xxviii (1986)