346 BELL SYSTEM TECHNICAL JOURNAL 



2^Pi state is known to be so short (of the order of 10"'^ second) that 

 under the actual conditions of some of the experiments an atom would 

 not often meet two quanta in such quick succession that at the advent 

 of the second it would still be in the 2^P] state into which the first had 

 put it. In other words, the number of 2^P] atoms in the gas at any 

 moment is too small. 



This last would be a serious obstacle to any theory, but for the fact 

 that the mercury atom possesses another stationary state slightly 

 below the 2^Pi, the which is metastable. This is the 2^Pq state, of 4.7 

 equivalent volts; its mean duration may amount to something like 

 a hundredth of a second. Now collisions of mercury atoms in the 2^Pi 

 state with atoms of certain other kinds, argon notably, may cause the 

 former to pass over into 2^Po- This is an instance of "collisions of the 

 second kind." When mercury-vapor is mixed with a much larger 

 quantity of argon and is illuminated with 2537 light, the number of 

 2^Po atoms is at any moment much greater than the number of 2^Pi 

 atoms would be, if the argon were absent ; further, it is proportional to 

 the amount of argon. 



Now F. G. Houtermans found that the rate of ionization, in mercury 

 mixed with argon and irradiated by 2537, is proportional to the amount 

 of argon. Therefore, in one stage of its progress from a normal atom 

 to an ion, the mercury atom must be in the 2^Po state. It enters this 

 state from the 2^Pi because of a collision with an argon atom. How 

 does it leave? by absorption of a second 4.9-volt quantum? Two of 

 the considerations of the last paragraph but one speak against this 

 idea, and Houtermans thinks that the 2^Po atom collides with another 

 which is in the 2^Pi state, and there is an interaction — this would be 

 another sort of "collision of the second kind" — in which one of them 

 adds to its store of energy all or most of what the other possesses. So 

 it arrives within an equivalent volt or so of the state of ionization; if 

 one were to take over all the energy of excitation of the other, it would 

 have 4.7 + 4.9 = 9.6 equivalent volts, out of the 10.4 required. Still 

 a third step seems to be essential. 



The reader may have wondered that I have as yet said nothing 

 about the dependence of ionization on intensity of light, for evidently 

 the former should increase as the cube of the latter if the process is a 

 three-stage one as I have sketched. The matter has been tested by 

 experiment; the answer was unexpected, for the ionization varies as the 

 square of the light — as though the process were of two stages.^ We 



* This simple result was obtained only over certain ranges of temperature and 

 pressure of the vapor, but these were precisely the ranges where both are low, and we 

 should expect the result to be most reliable and least subject to confusion by secondary 

 effects. As the pressure rises so does the exponent n in the relation ionization = 



{intensity)'^. 



