The interpretation of the curves given above is that due essentially to Roozeboom, who concluded from the experimental data that at temperatures below 1000° the stable systems are martensite and cementite, or ferrite and cementite, graphite being labile. It has, however, been pointed out, more especially by E. Heyn,[311] that this is not in harmony with the facts of metallurgy, which show that graphite is undoubtedly formed on slow cooling, and more especially when small quantities of silicon are present in the iron.[312] While, therefore, the relationships represented by Fig. 75 are obtained under certain conditions (especially when manganese is present), Heyn considers that all the curves in that figure, except ACB, represent _metastable_ systems--systems, therefore, akin to supercooled liquids. Rapid cooling will favour the production of the metastable systems containing cementite, and therefore give rise to relationships represented by Fig. 75; whereas slow cooling will lead to the stable system ferrite and graphite. Presence of silicon tends to prevent, presence of manganese tends to assist, the production of the metastable systems. Although this view put forward by Heyn has not been conclusively proved, it must be said that there is much evidence in its favour. Further investigation is, however, required before a final decision as to the interpretation of the curves can be reached. Determination of the Composition of Compounds, without Analysis.--Since the equilibrium between a solid and a liquid phase depends not only on the composition of the liquid (solution) but also on that of the solid, it is necessary {229} to determine the composition of the latter. In some cases this is easily effected by separating the solid from the liquid phase and analyzing it. In other cases, however, this method is inapplicable, or is accompanied by difficulties, due either to the fact that the solid phase undergoes decomposition (_e.g._ when it contains a volatile constituent), or to the difficulty of completely separating the mother liquor; as, for example, in the case of alloys. In all such cases, therefore, recourse must be had to other methods. In the first place, synthetic methods may be employed.[313] In this case we start with a solution of the two components, to which a third substance is added, which, however, does not enter into the solid phase.[314] We will assume that the initial solution contains _x_ gm. of A and _y_ gm. of B to 1 gm. of C. After the solution has been cooled down to such a temperature that solid substance separates out, a portion of the liquid phase is removed with a pipette and analyzed. If, now, the composition of the solution is such that there are _x'_ gm. of A and _y'_ gm. of B to 1 gm. of C., then the composition of the solid phase is _x_ - _x'_ gm. of A and _y_ - _y'_ gm. of B. When _x_ = _x'_, the solid phase is pure B; when _y_ = _y'_, the solid phase is pure A.