236 
MEASUREMENT OF HIGH TEMPERATURES. 
[bull. 54. 
aged in believing the non-inglazed thermometer bulb nearly as avail- 
able and safe for high temperature measurement as the inglazed form. 
Coefficient of heat expansion of porcelain. — It is necessary to state 
that in the present work no special measurements for ft have as yet 
been made. The value taken is that of Deville and Troost, wbich, for 
the porcelain in hand, is possibly too low. The probable effect of cor- 
recting ft will therefore be an increase of T, since increments of both 
ft and T have like signs. Not wishing at present to redetermine ft I 
made corroborative tests in the following manner: 
In Table 66 results' are given from which approximate values of the 
coefficient of expansion of porcelain may be computed. Let the ma- 
nometer volume be changed when the temperature of the bulb is nearly 
constant. Then if 
H lf V u T u siuAR 2n V 2i T 2 
are two successive readings, it follows that 
H 2 Y, 
[/(T)+^(J/(T0) 
f(T 2 ) 
E l Y l 
ATi 
H, — Hi 
where T is the temperature and v the zero volume of the bulb. It is 
expedient to make H 2 the barometric height for the day, so the T can 
at once be computed by the ordinary formula, ft may then be com- 
puted for/(T). By making the measurements for T (i. e., H u 7i, T x ) 
alternate with the measurements for ft (i. e., H 2 , V 2 , T 2 ) in time series, 
the value of T, which corresponds to the time at which the measurements 
for ft are made, may be accurately determined by graphic interpolation. 
The following results are computed from the data already given in Table 
66, Bulb II, Series V : 
Table 79. — Coefficient of heat expansion of porcelain. 
Time. 
T 
P 
h. 
11 
m. 
45 
o 
568 
11 
12 
53 
03 
(564) 
560 
} 
0. 000022 
12 
12 
06 
25 
(564) 
567 
} 
0. 000026 
2 
34 
978 
2 
2 
50 
58 
(993) 
1002 
1 
0. 000037 
3 
3 
10 
16 
1004 
1006 
! 
0. 000027 
In view of the fact that the quantities on which ft ultimately depends 
are of the same order of magnitude as the stem error J2, this method 
of determining ft can not be looked to for very close results. Indeed, if 
(890) 
