Thermions emitted by the Alkali Sulphates. 983 



specimens of the metal. By assuming the ions from plati- 

 num to have the average value (25- 5) previously found, it 

 was possible to calculate the values of e/m and m for the salts 

 investigated without making use of the linear dimensions of 

 the apparatus. If we could rely on the constancy of e/m for 

 the positive ions from platinum, we could in this way get rid 

 of the uncertainty arising from the difficulty of knowing the 

 distance between the hot strip and the plates accurately. It 

 is probable, however, that the value of e/m for the positive 

 ions from platinum is only approximately constant, so that 

 this method can only be regarded as a check on the other. 

 The disagreement between the two methods was not much 

 greater than the errors of experiment. 



It has already been pointed out that the method is liable to 

 an error on account of the uncertainty as to the position of 

 the strip when heated. On account of the expansion and 

 consequent curvature of the strip, this is not the same as 

 when it is cold. This uncertainty was avoided in the former 

 experiments by using the apparatus to measure the known 

 value of e/m for the negative electrons. Assuming that the 

 measurements would be liable to the same error whatever the 

 sign or mass of ions, this enables us to determine a correction 

 to be applied to the positive ions, 



In the present investigation this procedure has not been 

 followed. The distances z which have been measured are 

 those between the back of the strip and the front of the plates 

 in the neighbourhood of the strip. The distances between 

 the plane from which the ions are emitted and the slit-planes 

 will, therefore, be less than this by the thickness of the 

 platinum plus the thickness of the layer of salt. The former 

 was *002 cm. and the latter about as much. The platinum 

 strip was arranged so that it curved away from the plates 

 when heated. It is believed that the increase of the distance 

 due to the curvature more than counterbalances the effect of 

 the neglected thickness. The effect of these errors in the 

 distance is probably to make the values of m deduced from 

 the observations too low. On the other hand, there is an 

 intrinsic error, due to the apparatus not satisfying the theoreti- 

 cal conditions, which tends to make m too high. It is believed 

 that these effects just about balance one another when the 

 value of z as determined above is substituted in the formula. 

 At any rate, it is clear that however considerations of this 

 kind may affect the absolute values of e/m and m, they can 

 have very little influence on the relative values of these 

 quantities. 



