VELOCITY OF POLYMORPHIC CHANGES BETWEEN SOLIDS. 77 



pressure is due to a reversal of direction of transition at exactly the 

 same svu'faces which we had been previously observing. This was 

 tried on several occasions, and no difference found between experi- 

 ments under these and the usual conditions; the transition velocity 

 with falling pressure was in all cases greater. It is therefore, very 

 probable that in most cases only one nucleus was formed. But one 

 nucleus would give rise to two surfaces, travelling in opposite direc- 

 tions along the cylindrical container. The limiting accelerations 

 listed above were usually observed when between 50 and 90% of the 

 substance had been transformed. If the nucleus were formed at 

 random throughout the mass, this limiting acceleration would some- 

 times be observed with one and sometimes with two surfaces. But if 

 the original nucleus were always formed near one end of the tube, 

 one of the surfaces of separation would have run to the end of the tube 

 and disappeared, so that the final observations would always be on 

 one surface only. The regularity of the results makes it almost cer- 

 tain that this is what happened It is, in any event, to be expected 

 that nuclei will be formed near the ends in preference to the central 

 portions, because the first eftect of change of pressure would be felt 

 near the open ends. 



It was not possible to detect any connection whatever between 

 transition acceleration and the total quantity of the substance that 

 had been transformed. The data were of course at hand from which 

 the total fraction of transformed material corresponding to each limit- 

 ing acceleration could be determined. These fractions were written 

 down against every point of all the curves plotted, and showed an 

 utterly random distribution. This of course is what would be expected 

 if the measurements are actually of the rate of advance of a single 

 surface; the rate of this surface will not involve its position in the 

 cylinch'ical tube. 



Although the data of this paper are almost entirely concerned with 

 the surface growth, some comment is called for on phenomena con- 

 nected with nucleus formation. In some cases the observed pressure 

 change did not initially assume a high velocity and decrease regularly, 

 but at first increased, passed through a maximum, and then decreased. 

 This is shown in Figure 2 for AgNQs. It is without doubt due to the 

 fact that the surface of the freshly formed nucleus has not yet attained 

 the maximum extent allowed by the container; the total rate of trans- 

 formation of one phase into the other is less because the small extent 

 of the surface at which growth occurs more than compensates for the 

 rapid linear advance of the surface. This phenomenon is probably 



