240 Research Staff of the G. E. C, London, on tl 



1 obtained for a was a mere coincidence (and a very unfor- 

 tunate one, for it suggested many false hypotheses), due to 

 the particular choice of the length of the tube joining the 

 vessel to the trap. If a shorter tube was used higher values 

 of a were obtained ; if a longer tube was used (or the vessel 

 sealed off completely) lower values were obtained. 



10. This may not be a complete account of the matter, 

 but it was clear that it would be very complex so loug as 

 the reverse reaction was possible, and that the only hope of 

 obtaining simple results lay in removing the products of the 

 reaction very rapidly. For this purpose it was proposed 

 to plunge the whole discharge vessel into liquid air. But, 

 before proceeding to these measurements, it may be men- 

 tioned that similar experiments to those described in par. 8 

 were also made with the vessel 1 of par. 2, in which the 

 anode was a fine wire and the grid the walls of the vessel. 

 In this vessel a curve for a. Avas obtained generally similar to 

 that of tig. 5, and reaching the same maximum (for the 

 length of the tube connecting the vessel to the cooled trap 

 was the same), but beginning to fall away from the maximum 

 at considerably lower pressures (about 0'02 mm.). This 

 difference is in accordance with expectation, for the paths of 

 the ions in the gas will be longer at the same pressure, owing 

 to the greater distance between the electrodes. But the 

 experiments were not so complete. 



The new T vessel (vessel 3) was much smaller. It consisted 

 of a cylinder 7 cm. long and 2*5 cm. in diameter immersed 

 in liquid air. It had a straight central filament surrounded 

 by a spiral grid 3 cm. in diameter; the anode was the 

 silvered wall of the vessel. The dimensions of the vessel 

 were so small that the values of a, began to fall off at about 

 O'll mm., and most of the experiments were made above 

 that pressure. 



The results were disappointing. In a long series of 

 measurements between p = - 25 and 0*15 mm., values of a 

 were obtained ranging between 3'5 and 1*5 ; they were all 

 higher than those found before, but some very much too 

 high and some still too low. The high values occurred with 

 large currents and small potentials, the low with small 

 currents. This feature of the high values suggests that they 

 may be due to recombination of ions in the gas, so that i g is 

 less than i p . For if positive ions are formed which react 

 with neutral molecules to give new positive ions (C0 2 or C) 

 which recombine in the gas, these ions will contribute to 

 — p, but will add nothing to i g . 



Support could be obtained for the suggestion. If the 



