DEEP-SEA EXPLOEATION. 
331 
To accurately reproduce a standard boiling-point temperature it is necessary that 
the escai)ing steam be snbjecteel to a pressure equivalent to the pressure of a vertical 
(iolumn of pure mercury 700 millimeters high, when its temperature is at the freezing 
point, and under a gravity equal to tliat at sea level and at latitude 45°. 
The teiiq)erature interval between the freezing and boiling points on any thermo- 
metric scale is, therefore, established by two definite x)liysical phenomena. The 
subdivision of the interval into small and convenient units is i)urely arbitrary and 
need not be considered here. 
The errors to which the ordinary mercury in glass thermometers are subject are 
the result of three wholly ditfei'ent causes. The sources of errors are: 
(1) The inequalities in the diameter of the bore of the thermometer from point to 
point along the stem. Tubes can not be produced of absolutely uniform diameter, and 
where the diameter is large, unless compensated for in the graduations, the tenq»era- 
ture indicated will tend to be too low, and too higli where the bore is narrow. We may 
include in this source of error any accidental irregularity in the scale of graduations. 
(2) The second source of error is found in the character of the expansion of the 
mercury itself. The amount of increase in volume for a small increase in temperature 
is not exactly the same at low and high temperatures, and the error due to this cause 
is further modified by the irregular expansion of the glass envelope. No two kinds 
of glass will expand in quite the same way. 
(3) A third source of error arises from a small, protracted, and gradual shrinkage 
of the glass in the bulb of the thermometer. Different varieties of glass exhibit 
marked differences of behavior in this respect. A theianometer heated to the boiling 
or other high temperature to-day, and which registers correctly when tested at the 
freezing point will, a month hence, when again tested at the freezing point, often be 
found to indicate one or two tenths of a degree too high. 
It results from the second source of error mentioned above, that, even if a 
mercury in glass thermometer be constructed Avith the greatest care, its freezing and 
boiling points fixed with extreme precision, and its scale of graduations adapted to 
perfectly compensate for inequalities in the bore, the instrument Avill yet fail to 
indicate temperatures correctly, because of the variable rate of ex})ansion of mercury 
combined with the unknown and more or less irregular behavior of the glass envelope. 
It is necessary to have recourse to some other standard for comparison. For- 
tunately, the expansion of dry air, Avhen free from carbonic acid, or better, of ]uire, 
dry hydrogen, is found to be almost perfectly regular over a very wide range of 
temperatures and is, therefore, capable of being a true index of temperature. The air 
thermometer, then, is the standard instrument commonly adopted for the measurement 
of temperature. Its use, however, being a matter of considerable complication, the 
errors of suitable high-grade mercurial thermometers are determined once for all 
with great care by extended comparisons with the air thermometer, Avhereupou the 
mercurials are available for convenient and freqirent use as working substandards 
representing the air-thermometer scale. 
The difference between temperatures on the air thermometer scale and a normal 
mercury in glass thermometer may amount to nearly two-tenths of a degree at ordinary 
temperatures and is much greater at temperatures above the boiling point. 
It is advisable in all cases that the graduations of mercurial and otlier similar 
thermometers be etched on the stem and not engraved on a separate strip of wood or 
metal, as is often the case. 
