, 
230 MEASUREMENT OF HIGH TEMPERATURES. [bull. 
ness of about 0.1 cc in a series of measurements is quite attainable. Ot 
the other hand the determination of-*, by inserting a good mercury 
thermometer into the re-entrant tube of the porcelain bulb while thj 
apparatus is in place in the revolving muffle, is less satisfactory than! 
is convenient,* but in a series of three or four such measurements! 
mean error greater than 0.5° C. is improbable. It is understood thj 
it has not been my object in this paper to push measurements of v an 
t to the extreme fineness desirable; but it is quite clear that this maj 
be done by water calibration and immersion in water. 
Greater difficulty is encountered as we approach the main data 1 
measurement, Fi, T u R, Since F x must be measured throughout win 
an accuracy of about 0.5 CC , it is obvious at once that stages of practical)] 
constant temperature, or of very slow heating and cooling, are necessad 
for measurement. This is attainable in the revolving muffle very pet 
fectly by proper manipulation of the' graduated faucet for gas influl 
and by closing all apertures of the furnace during cooliug. With dm 
care, however, the measurement of variable V x to 0.05 cc will remain I 
problem of nice observation, particularly so since the total range o 
variation of V\ is nearly 235 cc . The difficulty is increased, inasmuch ai 
Ti must be known to 0.1° C; hence it is not only desirable to jackej 
the tube B G (Fig. 40) with a current of circulating water, but smal 
thermometers adjusted tvithin the calibrated tube in such a way as t< 
indicate the temperature of the air inclosed are a desideratum. Fortu 
nately the hot air which emerges from the air thermometer may be conj 
pletely cooled after passing through the metallic capillary tubes. I 
appears, moreover, that 150 cm of platinum capillary tube of the tiimel 
sions given in Table 42 are not seriously detrimental in producing sligh; 
differences of pressure in the bulb and manometer; for if Vi be con 
structed as a function of T (thermo-electric) by aid of the data in Table! 
58 to 66 it will be seen that V\ obtained during the stage of increasing 
temperatures (i. e., with the gas on) is quite the same as the correspond 
ing V } obtained during the stage of decreasing temperatures (i. e., during 
cooling), caeteris paribus. This is a crucial test, which reflects favora 
bly on the method as a whole. Fortunately the value of U, or the bar 
ometer, is sufficiently given if measured to 0.01 cm . This is not 011I3 
easily done, but an interval of pressure of 0.01 cm is larger than the fric 
tion of the gas passing through the capillary tubes can maintain for anj 
interval of time as long as that of an observation. 
The magnitudes of the corrective term 8 have already been referrec 
to. It is seen that an error as large as 30 per cent, in v' is withoul 
serious effect on the result. In like manner $ itself is sufficiently 
known to 50 per cent. It is this fortunate result which makes the use 
of a compensator a questionable desideratum ; and if it be borne ir 
mind that the distribution of internal fissures is usually such that twfl 
stems identical externally are by no means so internally, ijb is probable 
that here no advantage is derived from the compensator. This rem art 
(884) 
