27 



leaving the surface). Although the radium in this case was 

 nearly free from its radio-active descendants, yet a small trace 

 must have remained. The effect would be to make the value 

 of the current a little too large at all times, but especially 

 when the absorbing sheet was so thick as to stop almost en- 

 tirely the a jDarticles from the radium itself ; so that the last 

 readings of the last column would be too high, which is the 

 case. A more important explanation of the want of concord- 

 ance of the first and last results with the rest seems to be 

 that for some reason the first layers of aluminium foil which 

 are laid over the material cut down the radiation more than 

 they ought to do. This may in part at least be a consequence 

 of the experimental arrangements. The aluminium foil 

 cannot be made to lie very flat on the surface of the material, 

 on account of its flimsy nature, and must have a little air 

 space underneath it. >Tow, the air next the material is the 

 seat of a relatively large amount of ionisation. Thus, the 

 first layer may have an exaggerated importance. Another 

 partial cause may arise from the fact that the first layer or 

 two must cut off the easily absorbed radiation from the radio- 

 active surface which has been shown to exist by J. J. Thom- 

 son and by Rutherford. I am not aware of any measure- 

 ment of the amount of ionisation due to this radiation. If 

 in this experiment only 4 per cent, of the whole ionisation 

 current, when the material is uncovered, is supposed due to 

 this cause; and if the foil whose pd ='000317 cuts off three- 

 quarters of it, and the next addition of foil the remainder, 

 then the figures in the last column become, in order, 488, 

 487, 487, 496, 486, 490, 502, 514. Thus the existence of a 

 small quantity of radiation of this kind would explain the 

 present discrepancies in the experiment. It will be seen later 

 that a similar effect occurs with thorium. It is not so notice- 

 able in the uranium experiment, as will be found on turning 

 back to the table of results. Still, the first result is rather 

 smaller than those which follow, and a separate measurement 

 made with a very thin layer for which pc? = '000133 gave a 

 value for the full range equal to "00426, which is much smaller 

 than the rest. 



It should be mentioned that the first and last readings 

 are more liable to error than the others, since the ends of the 

 curve are used in obtaining them. 



On the whole, therefore, the radium measurements are 

 liable to certain small errors whose magnitude can hardly be 

 estimated as yet. But they are small, and they tend to bal- 

 ance each other, so that for our present purpose we may safely 

 assume the mean of the results of the last column, viz., 

 '00466, to be the pd of that sheet of aluminium, which can 



