162 THE ROYAL SOCIETY OF CANADA 
on account of the more uniform field of the former. In the latter 
vessels, while volume recombination might be prevented close to the 
rod, it might still be effective nearer the case. 
The second part B of the curve represents that condition in which 
the field is sufficient to eliminate volume recombination, but only 
partially prevents combination of the rest-atoms with the ions in the 
dense column of ions formed by the rest-atoms on recoil. This col- | 
umnar recombination is much harder to eliminate than the volume 
recombination on account of the density of the ions within the column, 
and particularly so when the columns are formed parallel to the field. 
The greater difficulty of preventing columnar recombination is shown 
by the smaller slope, which appears to be about 3% per 1000 volts. 
By the time a potential gradient of not over 1,000 volts per cm. 
is reached, however, the columnar recombination seems to be practi- 
cally eliminated. (It is to be regretted that it was not possible with 
the facilities available to obtain potential gradients between 150 and 
1,000 volts per cm.) The curve then makes another bend and the 
part C is entered upon. This part of the curve appears to have a 
slope of about 0-3% per 1,000 volts, only 1/10 of the slope of part B. 
Only two points on this portion of the curve fall within the limits of 
Fig. 2, but this curve C, if produced, would pass through the remaining 
points given in Table I. 
The following possible physical explanation can be offered for the 
third part of the curve. The radium rest-atom on recoil produces a 
dense column of ions by collision with the molecules of the gas and is 
rapidly brought to rest, whereupon it can come under the influence 
of the field. Now just as the rest-atom is at the end of its recoil 
path it may not succeed in completely disengaging itself from the last 
molecule which it ionizes, but may remain within the influence of this 
particular molecule. Under these circumstances the problem of 
preventing recombination is no longer the statistical one of preventing 
recombination with some one of the dense group of ions forming the 
column but with one particular ion separated from the rest-atom by a 
distance of a smaller order of magnitude than even the distances 
between the ions in the column. That is, instead of depending on the 
density of the ions as in the cases of volume and columnar ionization 
the problem depends upon: the distance between the rest-atom and 
one particular ion; and an electric field of a different order of magni- 
tude would be required to effect the separation. This would explain 
the smaller slope of the third part of the curve. The fraction of the 
rest-atoms which find themselves within this sphere of influence 
naturally depends upon the nature of the last molecule with which 
the rest-atom comes in contact. One would expect that the larger 
