270 MEASUREMENT OF HIGH TEMPERATURES. [bull. 54.' 
coincide with (l+ad")i almost perfectly. In general therefore the re- 
sults of earlier investigations (Table 81-85) are emphatically corrobo- 
rated. Having finished the measurements for air, I passed hydrogen 1 
through the same apparatus. The results, near 1,000° (Table 87), coin-j 
cide almost perfectly with the curve (1+ ad")*. The results above 1,200° 
are abnormally high, because platinum is pervious to hydrogen. I was; 
not a little surprised therefore on finding the results ac 500°, although 
very good when compared amongst themselves, much too large to ac-! 
cord with the other data. The error is larger than can be referred to 
anything short of an undiscovered accident. I suspect the hydrogen 
of these experiments in some unforeseen way to have been contaminated' 
with either air or moisture, since the error is constant for all the obser-j 
vations. 
Not being able to discover the cause of the discrepancy, however, 1\ 
resolved to repeat the work with the apparatus adjusted anew. These] 
results are given in Table 88. The results agree much more closely • 
with the exponential (l+a6"% showing the corresponding data of thei 
former series in error. The somewhat high values of ¥(6") are the re-1 
suit of the low value of r/ , to which these data are referred. The diffi- 
culties met with in operating with hydrogen induced me to investigated 
another series for air, which series is given in the last table, 89. The] 
results fall slightly above (l-\-ad")l at 500° ; slightly below the expo- 
nential at 1,000°; above it at 1,200° ; but the accordance throughout is- 
satisfactory. (See Fig. 48.) 
The large number of data for 6 Y and 2 , which the Tables 86-89 con-] 
tain, show that the thermal discrepancy has been reduced to low limits, 
and that its effect is apt to be relatively large in case of low values of] 
0. To properly jacket the helix is a practical problem of great diffi- 
culty, because accidental bending or twisting of the thin capillary tubej 
is apt to produce fine longitudinal fissures, or variations of bore, or to] 
introduce other sources of uncertainty or error. Mica insulations be- j 
come friable and can not be thoroughly relied upon after prolonged j 
heating. Hence all the manipulations must be effected with great] 
care, and therefore consume much time. For these reasons I believe 
that to improve the results further it will be expedient to introduce aj 
radical change of method, such as I will describe below. 
As a whole, the above tables conclusively indicate that many residual 
errors are the result of variations in the composition of the gases. Hence, 1 
before proceeding to a general discussion of the above data, I shall in- 1 
sert a tabular view of the successive values of the viscosity of the gases I 
at zero degrees centigrade. This is given in Table 90, in which the | 
temperatures 6, at which the quantities 
V 
!+4 
(924) 
