305 
in which C is a constant. As an approximation for vapours we may 
write p=ao-+ bo? in which a and 0 are functions of 7. These 
and all other quantities occurring in the present calculation were 
expressed in absolute measure (the C.G.S. system was chosen). If 
T be given as a function of x, equation (1) can at once be integrated. 
As a further simplification for this integration we shall regard 5 
as negligible on account of the smallness of 9” compared with ag. 
If the pressure difference between the ends of the capillary is small, 
deviations from Boryre's law may, to the same extent, be allowed 
for. For further information on this point I may refer to my disser- 
tation. 
It may be further remarked that we may differentiate between 
three different portions of the capillary. The first part projects above 
the cryostat, and has throughout its whole length the same tempe- 
rature, that of its surroundings (room temperature); for the pressure 
at the ‘upper end of this portion we shall write p, and for the 
pressure at the lower end p,. In the second part of the capillary 
the temperature changes from the room temperature to that of the 
cryostat bath. The pressure at the upper end of this part is p,, and 
for the pressure at the lower end we shall write p,. The third 
portion of the capillary is wholly within the cryostat bath, and over 
its whole length has the temperature of the bath. p, is the pressure 
at the upper end, and we shall write p, for the pressure at the 
lower end. 
With the object above indicated of not only calculating for the 
particular case discussed in $ 3, but also of obtaining simple formulae 
applicable to analogous cases I have endeavoured to find a simple 
form for the function expressing the temperature of the middle portion 
in terms of the length; in order that four terms in this would 
suffice I have imagined a sudden change in the temperature at the 
junction of the second and third portions of the capillary, in other 
words I assume that at that point the temperature changes rapidly 
over a length which is large compared with the diameter of the 
capillary but is still small compared with its length. 
The calculation is therefore made for a temperature distribution 
other than that which actually exists, but, as will be seen, the 
difference between the two cases does not affect the result. 
The temperature distribution over that portion of the capillary in 
which the temperature is variable is thus represented by 
a=q+1,T + mT? + nT. ded Ae es (2) 
In the experiment further discussed in § 3 the temperature change 
