^us.1 PORCELAIN AIR THERMOMETRY. 107 
lie air thermometer during the intermediate measurements must be 
erfectly tight. Since the error decreases proportionally to h , the 
dditional accuracy of greater zero tensions does not compensate the 
urtful effect of high internal pressures at high temperature. Hence 
)w pressures are preferable. Regarding H, it appears that rise or fall 
f temperature must not be so rapid that the retardation due to flow 
f gas through the capillary tubes maintain greater differences of press- 
re than 0.03 cm to 0.08 ,m of mercury. A good cathetometer presup- 
osed, it is not difficult to measure both h and H with the accuracy here 
ailed for. 
Under most favorable circumstances the error of Tis as large as the 
rror of a, a result which equation (15) approximately shows. The 
alue chosen, 0.003665, is RegnaultV constant- volume value, and has 
een found experimentally for the interval 0° to 100°. The use of the 
ime coefficient for temperature indefinitely high and for all tensions 2 
; to some extent arbitrary. The error thence resulting may be esti- 
tated at as much as oue-half of one per cent. The convenience with 
inch the constant pressure method is available for measurements 
ith gas differing widely in normal density is one of its most valuable 
atures. The desideratum of an elliptic revolving muffle for the com- 
irison of the gas thermometer data of different gases, when the tem- 
srature of the same environment is measured, has been suggested. 
The unusually small coefficient of cubical expansion /?=0.00001(5 to 
000017, which MM. Deville and Troost found for the porcelain of Bay- 
ix, makes the necessary accuracy of the coefficient ft sufficiently at- 
Inable. The table shows that even in extreme cases, T= 1,500°, an 
ror of 3 per cent, in ft is not serious, while the expansion coefficients 
' the metal and glass parts of the air thermometer need not be dis- 
iguished from ft, because these parts are almost negligible here. This 
ceptionally small value of/? is, however, only admissible in the case 
thermometers which have frequently been heated. In the case of 
sw thermometers these desirable qualities are vitiated by the occur- 
nce of permanent expansion for each heating. In comparison with 
this permanent expansion (permanent diminution of density) is un- 
jrtunately enormously large, aggregating in the first six heatings, for 
tance, as much as 1.5 per cent. Whenever temperature increases 
o rapidly the temperature, and hence also the volume of the bulb, is 
ger than corresponds to the mean temperature of the gas. The re- 
rse of this takes place on cooling. 
Conformably with the numerical results on page 197, Table 47 shows 
at — may be affected bv an error of almost twice its own magnitude 
v 
thout seriously bearing on T, even in unfavorable cases. The impor- 
aceof — , however, increases rapidly as T increases, and must in un- 
v ' 
1 Regnault: Memoires de l'Insfr., vol. 21, 1847, p. 91 ; ibid., p. 110. 
2 Cf. Literary Digest, pp. 36-38. 
(851) 
