558 Prof. Richardson and Mr. Hulbirt on the 



With one or two exceptions the numbers in the last column 

 do not differ from one another by more than the error of 

 observation. There is a considerable error in these measure- 

 ments arising from the curvature of the strip when heated. 

 Although the strips were very short (*5 cm. in length) the 

 effect of this on z is important since the value of e\m involves 

 the fourth power of z. An attempt was made to allow for 

 this by direct observations of the displacement of the strips 

 when heated, and they were always arranged so that they 

 curved in towards the plates. 



The mean value of all the numbers in the last column 

 counting platinum only once and equal to 25*35 is 25*3. It 

 is probable that the value for iron is affected by the presence 

 of some of the substance which gives rise to the ionization 

 with very high values of e\m. If we omit iron from the 

 average we find the mean value of w/H = 25*7. This number 

 is very close to the values found for carbon and platinum in 

 the previous paper, so that the greater part of the discussion 

 there applies equally to the present results. The case against 

 the view that this ionization is due to one or more of the 

 gases whose molecular weights are about 30 (0 2 , N 2 , and CO) 

 is strengthened by the present experiments. It is probable 

 that traces of carbon monoxide are always present when a 

 metal is heated in a vacuum in the kind of apparatus used in 

 the present experiments, but it is difficult to see why practi- 

 cally all the ionization should be due to this gas in the case 

 of all the substances investigated. It is known that when a 

 great many metals are heated the bulk of the gas given off 

 is hydrogen, and one of the authors* has shown that when 

 this gas escapes from the platinum a considerable amount of 

 it is in the form of positive ions ; whereas in none of the 

 present experiments were any ions detected for which the 

 value of e\m approached that of hydrogen. Moreover, the 

 metal most likely to emit carbon monoxide in quantity is 

 iron, and this is the xme metal for which the value of #/m 

 deviated most widely from that corresponding to carbon 

 monoxide. 



It is perhaps of interest to remark that a great difference 

 was observed in the amount of gas given off by the various 

 substances used. Platinum and palladium gave off most, 

 gold gave a much smaller quantity, whilst the amount given 

 off by the other materials was inappreciable. 



The mean value of m/H is very near to the atomic weight 



* 0. W. Richardson, Phil. Trans. A. ccvii. p. 1 (1906). 



