298 
At this period the crystallization ended. 
30 1780 
35 1780 
On the surface of the plate some drops of distilled water were 
now added, and the dish was placed again on the waterbath, so’ 
that the crystals melted in the water of crystallization. The mea- 
surements were repeated in the same manner as before. 
In the moment when the second crystallization started, the acti- 
vity of the solution was 1530; when it was finished the activity 
of the plate was 2850. 
After the third erystallization the activity. was 2940. 
The fourth and fifth and sixth crystallization did not cause 
a further inerease of the activity. 
This maximum activity then decayed with the time and reached 
the value 935 after three days, and disregarding small irregular 
oscillations, remained constant at this value through many weeks. 
Similar experiments were repeatedly made and gave exactly 
the same qualitative results. 
These experiments show that the ß activity of uranium nitrate 
is very considerably augmented by the actual process of erystalli- 
zation and it will be proved later that the decay of activity, noted 
immediately after crystallization, is due to the loss of this excess 
of activity produced by erystallization. 
The explanation of the increase of activity at the moment of 
crystallization is very simple. We know that all the ß activity of 
uranium proceeds not from the uranium itself, but from UrX. But 
UrX is so readily soluble in water that it is possible, as we have 
seen, to separate UrX from uranium by fractional erystallization. 
If, as is usually the case, the hot uranium solution starts to erystal- 
lize from the bottom of the dish, first of all uranium itself ery- 
stallizes and UrX is pushed in the direetion of the surface. When 
the whole mass is solidified we get a plate which contains on the 
surface an excess of UrX and in the lower layers a defieit of this 
substance. The 8 rays which come from the UrX, present near 
the surface, emerge with little absorption in the mass uranium 
itself, and thus the ß activity must be larger than when UrX is 
uniformly distributed throughout the plate. In the same way we 
can explain the steady growth of activity during the actual pro- 
