kttus.] VISCOSITY OF GASES. 273 
f variation with greater certainty. Hence ordinary care was taken in 
rying the gases, air being passed through a tube of Ca01 2 and hydro- 
en both through Oa01 2 and concentrated H 2 S0 4 . But neither was 
he hydrogen purified of sulphide* or air or other attendant gases, nor 
ras air purified of carbon dioxide. I desisted from these special pre- 
autions because above red heat platinum is pervious to the hydrogen 
ases of the Bunsen flame, so that in non-enameled platinum capillaries 
he purity of the transpiring gas could not be vouched for, even in the 
ase of a gas originally pure. Again, at the temperatures (500° to 
,300°) within which my data chiefly apply slightly moist air and dry air, 
ure and gaseously impure hydrogen, are probably equally perfect gases, 
leuce there occur in my results two values for the zero-viscosity of 
lydrogen, 140: 10 b and 129: 10 G , respectively, which correspond to gases 
aken from different gasometers at different times. The variations of 
for air are smaller, ranging from 247 : 10 6 to 252 : 10 6 . The variable 
haracter of ?j made it necessary to make the low-temperature measure- 
lent before and after each series of high-temperature measurement 
iade. Indeed, I did not anticipate such large fluctuations in ^ , and 
espite the precautious taken the discrepancies here in question have 
ifoduced no trifling distortions in the high temperature results which 
)llow. This I shall soon have occasion to show. When the gases 
orced through the capillaries are urged forward by an advancing sur- 
ice of mercury (as in the above experiments) traces of mercury vapor 
rill also pass along with the gas, but the tension of mercury vapor at 
10° is only 0.002 cm to 0.004 cm . Hence this discrepancy is nil. 
Finally, Table 90 contains values for 7/ derived both for P=125 cm and 
r P=8S cm . Curiously enough, the value of ?/ for low pressures (P) is 
ecidedly the greater, being about 5 per cent, greater for hydrogen and 
bout 2 per cent, greater for air. This result belongs also to the dis- 
ussion of the next section. Here it is sufficient to note that if meas- 
rements were made at smaller values of P than the ones customary 
v" 
125 cm ), the relative values — where ?/' is a high-temperature viscosity, 
rould not be larger in value than those admitted into the above tables, 
n other words, so far as the present evidence goes, the zero- viscosity 
o has not been increased by the relatively large values of P (125 cm ) em- 
ri" 
ployed throughout the course of the work, and hence — can not be 
too small. 
Viscosity at high temperatures, kinetic inferences. — In order to pro- 
3eed with the discussion of the high-temperature viscosities, Tables 83 
to 89 may be consulted, the values of Tables 81 and 82 being in error 
n the sense already indicated. If all the values of F{6") be constructed 
is functions of 6", the graphic representation Fig. 48 will show that the 
ndividual data group themselves in a band or pathway, of which the 
Auction {1+ad") is so nearly the axis that it is at once justifiable to 
ywept it as such. Hence, even if there were no ulterior reasons for ac- 
Bull. 54—18 (927) 
