These experiments are of the greatest practical importance. At 600° acetylene begins to polymerize--i. e., to form more complex hydrocarbons, which are liquid, or solid, at ordinary temperatures. Probably in the generator acetylene is first given off so rapidly that the heat does not act on it, but as decomposition advances into the center of the mass of carbide, the acetylene generated has to pass through the external layers, which, as shown, may be at high temperatures, above that at which acetylene decomposes; thus a considerable amount of gas is lost, and the tar formed may distill into the generator and tubes, clogging the tubes. A more serious evil is the deterioration in the illuminating quality of the gas. Samples of the gas were taken as the maximum temperature was approached, and analyzed with this average result: Acetylene, seventy per cent; other hydrocarbons, eleven per cent; hydrogen, nineteen per cent. This reduces the illuminating value from two hundred and forty to one hundred and twenty-six candles. The hydrocarbons consist largely of benzene, which requires three times as much air for complete combustion as acetylene does. The best possible acetylene burner smokes when the acetylene contains benzene. [Illustration: Type I. TYPE OF GENERATOR.[6]] [Illustration: Type II, _A_. Type II, _B_. Type III. TYPES OF GENERATORS.[6]] [Footnote 6: NOTE.--We are indebted to the courtesy of the Progressive Age for cuts showing types of generators.] At first sight these experiments would seem absolutely to condemn generators of class II, yet the fact remains that some excellent generators are of this type. Under certain conditions excessive overheating may be avoided. The rising bell shown in II, _B_ should be discarded. Generators in which the water rises from below, and slowly attacks the carbide, can be made safe if the water is never driven back from the carbide, and the carbide is in separated layers as in II, _A_. Under these conditions the water is always in excess at the point where it attacks the carbide, so that the evaporation, by rendering heat latent, keeps the temperature down, the temperature of the melting point of tin, 228° C., being rarely reached in good generators where these conditions are met. Undoubtedly the best generators, and the only ones which from a scientific point of view should be employed, are those of class III, in which carbide falls into an excess of water. In such generators it is impossible to get a temperature higher than the boiling point of water, 100° C., while with a properly arranged tank the temperature never exceeds that of the air by more than a few degrees. Under these conditions the absence of polymerization and the washing of the nascent and finely divided bubbles of gas by the limewater in the generator yield acetylene of a degree of purity unapproached by any other form of generator.