POLYMORPHISM AT HIGH PRESSURES. Ill 



gested that it might be abnormal in other respects, and might possibly 

 show other modifications. It does as a matter of fact have at least 

 SL\ modifications, the complication of its phase diagram being equalled 

 or surpassed only by water and ammonium nitrate. It is unfortunate 

 that it was not possible to make measurements with as great accuracy 

 as was possible for many other substances. In fact the uncertainty 

 in many of the measurements of the change of volume is so great that 

 I have not attempted to estimate from them the most probable value 

 of the change of volume or to calculate the latent heats. There can 

 be no question, however, that the existence of the transition curves in 

 approximately the situations indicated has been established. It was 

 a matter of some difficulty, involving considerable time, to be sure of 

 as much as this. More than a month was spent in getting the data 

 given below; to obtain accurate values of the changes of volume and 

 the latent heats in addition would have taken many times longer, if 

 indeed it would have been possible at all. In the present state of our 

 knowledge of polymorphism it did not seem that all this effort on a 

 single substance would have been worth while. 



The phase diagram is shown in Figure 11, the curves for A^ in 

 Figure 12, the latent heat and the change of internal energy in Figure 

 13, and the numerical values in Table VII. The melting curve indi- 

 cated in Figure 11 is taken from the data of Hulett.^^ 



It will give an idea of the various difficulties and assist in forming an 

 estimate of the accuracy of the work to describe the curves somewhat 

 in detail. On the III-IV curve the reaction is very slow and the 

 limits of indifference wide. This does not affect the accuracy of the 

 change of volume, for which fairly satisfactory values were obtained, 

 but may affect the equilibrium values, and so the slope and the latent 

 heat. The III-IV curve is, however, determined with considerably 

 greater accuracy than any of the others. On the II-V and the IV-V 

 curves it is difficult to measure either the equilibrium or the Av values 

 because of the very small change of volume and the slight separation 

 of the curves. Let us suppose that we are trying to get a point on the 

 IV-V curve, lowering pressure from above. The pressure steps have 

 to be made very small, because of the small change of volume, and so 

 the measurements consume much time. Furthermore, after the re- 

 action has once started, one cannot release the pressure far enough 

 to ensure the completion of the reaction without overstepping the 

 II-V line, and so getting the reaction II-V mixed up in the effect; no 

 amount of time will give a good value for Av. It does no good to 

 attempt to ensure the completion of the reaction in the narrow region 



