26 Messrs. Richardson, Nicol, and Parnell on the 
higher power varying from ‘5 to °53. If we reduced the 
above experiments by assuming that Q ap over the whole 
range, the variations in column VI. would all be reduced to 
within the error of experiment. But as was pointed out on 
page 15, it seems more reasonable to suppose that the 
slight deviation from the square-root law is due to some 
uncorrected error. There are many possible causes for such 
an effect: for instance, the variation of the convection 
currents in the hydrogen inside the platinum tube might 
cause its temperature to vary slightly with the pressure when 
the temperature outside was kept constant. The mean of all 
the 34 values of log;Cu./Ay? given in column IV. is 7-934, 
whence the geometric mean of all the values of Cy/A,* 1s 
8:59x10-7. We thus obtain for the complete formula for 
the mass of hydrogen per second diffusing through each square 
centimetre of the platinum disk thickness d maintained at 
temperature @ when the gradient of the square root of the 
pressure between its faces is 72/d, ; 
9°125 
G2 600% 10,7 viene 
The constancy of the numbers in the preceding tables shows 
that this formula holds with considerable accuracy over a 
range of temperature from 576° C, to 1176° C., and over a 
range of pressure from 76:0 ems. tol mm. There are signs 
of deviation at low temperatures and at the extremes of 
pressure, but these may be due to unavoidable experimental 
errors. 
The quantity A which occurs in the above formulz is the 
partition coefiicient for undissociated hydrogen H,, and is to 
be carefully distinguished from the quantity which would be 
measured by an experimental determination of the solubility 
of hydrogen in platinum. It is, however, easy to obtain a 
formula for the ratio between the total internal and external 
concentrations of hydrogen inside and outside platinum on 
the view developed in this paper. If we neglect the concen- 
tration of the external dissociated hydrogen compared with 
that of the undissociated, we have, in the notation previously 
adopted : 
aN | a ik je see k,Co : 
C= ZC, and ¢=(k0C)) = ( Ah 
whence the total internal concentration reckoned in gramme- 
molecules per c.c. 
