MEASUREMENT OF TEMPERATURE. 
1G7 
If the bulb itself cannot be used, a mercury thermometer is made from the same 
tube as the bulb. 
Since, however, this method ceases to be applicable above the boiling point of 
mercury, and is moreover exceedingly liable to inaccuracy, except between 0° and 
100 ° C., it was decided to measure the variations of the linear coefficient and assume 
that the cubical coefficient varied in the same way. 
The method adopted was exceedingly simple, and is, shortly, as follows :— 
Linear Expansion of Hard Glass. 
A glass tube is heated in a thick iron tube in a long gas furnace, the variations of 
its length are observed by means of reading microscopes, and the mean temperature 
of the portion heated is given by observing the resistance of a standard platinum 
wire extending down the axis of the tube, assuming that when the temperature is 
steady the mean temperature of the wire is the same as that of the portion of the 
tube it occupies. 
Experiments by this method were made on a piece of hard glass tube, the same as 
that used in making the air thermometers No. 3 and No. 5. 
The ends of the tube were drawn down to 5 millims. diameter, to diminish the conduc¬ 
tion of heat, and to ensure that the parts exposed should be nearly at the temperature 
of the atmosphere; copper leads were fused on to the ends of a piece of the standard 
platinum wire, 69 - 6 centims. long, which was then inserted down the axis of the 
glass tube. 
The glass tube was heated by being placed in an iron gas-pipe, 67 centims. long, 
in a gas furnace of the same length, supported on a board with levelling screws for 
focussing. The copper platinum junctions were exposed to the air temperature at 
equal distances, about 1 cm. beyond the ends of the iron tube, and, being screened 
from the heat by bright tinned plate, no appreciable thermo-electric effect was 
produced, as the junctions were very slightly, and nearly equally, warmed. 
A length of about 63 cm. of the glass tube was left in its original state ; the drawn- 
down ends projected about 16 cm. beyond the iron tube, and were practically at the 
air temperature for their whole length beyond the copper platinum junctions; the 
variations in length of these portions of the tube may be neglected in comparison 
with other possible errors; it is, therefore, assumed that the mean temperature of the 
platinum wire is the mean temperature of the part of the tube it occupies, whose 
length is variable. If the coefficient of expansion of glass were accurately constant, 
or if the temperature of the whole portion heated were uniform, this assumption 
would give mathematically correct results. Since both these conditions are nearly 
satisfied, the corrections to be applied for the variations at either end are negligible. 
We take, therefore, the length of the part of the glass tube, whose variations we are 
measuring, to be the distance between the ends of the platinum wire, and its mean 
temperature to be that of the platinum wire. 
