Thus, when Sir Norman Lockyer entered on that branch of inquiry in 1872, fourteen substances were recognised as common to the earth and sun. Early in 1878 he was able to increase the list provisionally to thirty-three,[677] all except hydrogen metals. This rapid success was due to his adoption of the test of _length_ in lieu of that of _strength_ in the comparison of lines. He measured their relative significance, in other words, rather by their persistence through a wide range of temperature, than by their brilliancy at any one temperature. The distinction was easily drawn. Photographs of the electric arc, in which any given metal had been volatilised, showed some of the rays emitted by it stretching across the axis of the light to a considerable distance on either side, while many others clung more or less closely to its central hottest core. The former "long lines," regarded as certainly representative, were those primarily sought in the solar spectrum; while the attendant "short lines," often, in point of fact, due to foreign admixtures, were set aside as likely to be misleading.[678] The criterion is a valuable one, and its employment has greatly helped to quicken the progress of solar chemistry. Carbon was the first non-metallic element discovered in the sun. Messrs. Trowbridge and Hutchins of Harvard College concluded in 1887,[679] on the ground of certain spectral coincidences, that this protean substance is vaporised in the solar atmosphere at a temperature approximately that of the voltaic arc. Partial evidence to the same effect had earlier been alleged by Lockyer, as well as by Liveing and Dewar; and the case was rendered tolerably complete by photographs taken by Kayser and Runge in 1889.[680] It was by Professor Rowland shown to be irresistible. Two hundred carbon-lines were, through his comparisons, sifted out from sunlight, and it contains others significant of the presence of silicon--a related substance, and one as important to rock-building on the earth, as carbon is to the maintenance of life. The general result of Rowland's labours was the establishment among solar materials, not only of these two out of the fourteen metalloids, or non-metallic substances, but of thirty-three metals, including silver and tin. Gold, mercury, bismuth, antimony, and arsenic were discarded from the catalogue; platinum and uranium, with six other metals, remained doubtful; while iron was recorded as crowding the spectrum with over two thousand obscure rays.[681] Gallium-absorption was detected in it by Hartley and Ramage in 1889.[682] Dr. Henry Draper[683] announced, in 1877, his imagined discovery, in the solar spectrum, of eighteen especially brilliant spaces corresponding to oxygen-emissions. But the agreement proved, when put to the test of very high dispersion, to be wholly illusory.[684] Nor has it yet been found possible to identify, in analysed sunlight, any significant _bright_ beams.[685]