We have seen how they freeze either as eutectic alone, as crystals of austenite with eutectic or as crystals of cementite (Fe_{3}C) and eutectic, depending upon the original composition of the molten alloy. At and just below the temperatures represented by the line ED, this undoubtedly represents the situation. What happens to the alloys from this temperature down to normal depends upon conditions. Just what occurs and the mechanism of it is not definitely known except in the practical way. Certain it is, the “precipitation” of free carbon is necessary for cast irons which are to be serviceable for usual purposes. This may occur with consequent softening during the first cooling or they may be cast as “hard iron” and softened afterward. In Chapter XI we said that silicon was a “softener” as its presence brought about precipitation of the carbon as graphite throughout the cast iron, thereby softening it both by reason of the presence of the soft flakes of graphite and because it leaves so little of the carbon in the “combined” or hardening condition. So silicon is a ready means of bringing about decomposition of the higher temperature structures as the alloys cool. The speed of the cooling also exerts a very powerful influence in determining the amount of graphite which will separate. Other conditions being equal, the slower the cooling, the greater the decomposition with resulting graphite. Swift cooling, even such as results from the dumping of castings from the molds while at nearly white or high red heat results in insufficient graphite and otherwise harder metal. Cooling of very hot castings on a cold floor or in a current of cool air has considerable hardening effect even when the composition of the alloy would otherwise give very soft and machinable metal. In an extreme and very interesting case a few very hard cast iron flanges were each day found among the many thousands of habitually soft castings regularly produced. Each day two or three expensive “taps” were ruined by attempting the impossible—the machining of these pieces of hardened iron, which, on the outside, looked just like all the rest. It was soon discovered that two or three mold dumpers each noon and evening were warming water for “wash-up” by dropping into the pails a flange or two which were still white-hot after dumping from the molds. The men were innocent of any intention of harm but their warm water cost several hundred dollars before their method of producing it was discovered. Silicon and rate of cooling are the two most powerful influences but presence of certain elements other than silicon also influences to some extent the degree of hardness. However, while silicon has a strong softening effect, manganese and sulphur have an opposite or hardening tendency. On this account the amounts of these latter elements which can be used or allowed must be strictly limited.