barus.] VISCOSITY OF GASES. 289 
which temperature r the reduction to zero was made. F(6") has the 
above signification. 
F(6")= 
Unfortunately, the measurement of 6" in all these experiments was 
made as shown in the diagram, Fig. 47, and hence these values are not 
so good as those in Tables 86 to 89 above (helix very compact), r is 
directly read off on the thermometer T in the receiver B y in Fig. 44, 
and the correction may be applied either to t" or to the final tf'. 
To measure V, I filled the receiver with mercury and weighed the 
volume of liquid contained. Similarly B was computed from the weight 
of the mercury thread which fills the capillary tube. But this opera- 
tion in case of an opaque tube is very unsatisfactory. My plan was to 
seal the capillary helix (previously dried and weighed with care) into a 
glass tube with mastic, and then allow a current of mercury to pass 
through both. When all air was expelled I stopped up one end of the 
capillary with soft wax, then removed the glass tube, cleansed the helix, 
and weighed again. The weights so obtained differ enough to make 
errors of 5 per cent, and 10 per cent, in jK 4 possible. Hence I put no 
stress on the absolute data, but consider them in their relative bearing 
only. For each such relative series B is constant. Perhaps volumetric 
methods might have led to better results (v. page 213), but I did not 
apply them. 
To calculate if the formula on page 253 was employed in the usual 
way. This presupposes a knowledge of r/ for the cold ends, correspond- 
ing to each pressure observed. It is necessary, therefore, first to con- 
struct r/ as a function of pressure graphically, in order that the proper 
value may be inserted into the correction member of the formula for rf'. 
The error due to cold ends here in question bears an intimate relation 
to the Navier effect; and this method of correcting for it presupposes 
that both the cold part and the hot part of the capillary has two ends. 
Hence it would seem that as there are in all but two ends, instead of 
four, to the capillary tube experimented upon, that the correction 
applied is not valid. I think, however, the following results show that 
the departure from Poiseuille- Meyer's law observed with metallic capil- 
lary tubes is largely due to unavoidable roughness of the internal sur- 
face (i. e., of the walls of the capillary canal), and hence the correction 
is warranted. 
Bull. 54 19 (943) 
