The delicacy of spectrum analysis may be imagined from the fact that a quantity of sodium amounting to less than the _two-millionth_ of a grain can be detected by its means. Indeed, it has taught us that sodium in one form or other exists almost everywhere. This mode of analysis is only serviceable to indicate the composition of any salt or other substance, the quantities of the different elements found by its use having no influence on the appearances brought out by the prism. Thus, a substance which has only been contaminated with sodium from being handled by warm fingers, will show the yellow bands as strongly as if it contained a large proportion of that metal. For ordinary experiments in spectrum analysis the apparatus used is very simple. It consists of a tube with a fine slit at one end, and a convex lens at the other, for concentrating the light from the coloured flame upon the centre of the prism. After the light passes through the prism, it is examined by a small telescope of low magnifying power. The lamp used may be either a spirit-lamp or a colourless gas flame into which the substance to be examined is introduced upon a platinum wire. We now come to another very important discovery, made by means of our prism and narrow slit—the determination of the composition of the photosphere or mass of luminous vapour surrounding the body of the sun. A simple experiment will show how this brilliant discovery was arrived at. The light of a candle or other flame containing incandescent _solid_ matter is passed through the spectroscope, and is found to decompose into a continuous spectrum, uninterrupted by dark lines. Between the light and the slit a spirit-lamp is placed, but no difference in the appearance of the spectrum is perceived. Introduce, however, the smallest portion of a soda salt into the non-luminous flame of the second-lamp, and a broad black line is immediately seen, crossing the middle of the yellow portion of the band of colour. Remove the sodium flame and the band disappears; but do the same with the lamp producing the spectrum, and the spectrum of course disappears, and the dark band caused by the sodium flame is changed to the yellow line produced by that metal. The same experiments may be tried with potassium, strontium, and other metals; and we shall always find that when a coloured flame is introduced between an incandescent solid and its continuous spectrum, it produces a series of black lines corresponding to the substances by which it is coloured. Thallium, in like manner, would give a black band in the middle of the green, and indium a similar one in the indigo. (Fig. 6, Frontispiece).