Ferdinand II of Tuscany made the first sealed thermometer using wine spirit in 1641.
This was the first device that we would recognise as a thermometer and was a major
advance: a liquid sensor sealed against atmospheric pressure variations and evaporation,
and degrees of temperature permanently marked on the stem. Subsequent developments
of thermometers focused on improved methods of manufacture, especially on the choice
of thermometric fluid and glass, and the method of fixing the scale.
By the early eighteenth century most liquid-in-glass thermometers were stable and
had a reproducible scale. Typically the scales were marked using the temperatures of
fixed points, such as melting snow, body temperature and boiling water (see Figure 1.3).
The scale was then divided into a convenient number of steps or degrees. These scales
are clearly ordinal scales with fixed points, and with the temperatures in between
determined by interpolation using the expansion of the thermometric fluid, usually
mercury. The familiar scales of Fahrenheit and Celsius are good examples of these
types of scales. Elegant experiments involving the mixing of known volumes of hot
and cold mercury, or hot and cold wine spirit, enabled experimenters to establish that
mercury is a more linear thermometric fluid than spirit, but that both are non-linear to
some extent.
In the late eighteenth century Gay-Lussac and Charles, building on the ingenious
work of Amontons 100 years earlier, were both able to demonstrate that the thermal
expansion coefficients of different gases were almost identical. Although Amontons had
suggested that the linear expansion of gases with temperature implied that only one
fixed point was required to calibrate a thermometer (i.e. to establish a metric scale), and
a number of experiments determined values for absolute zero (−273.15°C), the suggestion was not adopted. Chappuis, working at the BIPM, refined gas thermometry further.
He had been charged with the responsibility of calibrating a set of mercury-in-glass
thermometers by gas thermometry. During a series of remarkable studies comparing
temperatures determined using different gases he showed that the gas thermometer
did in fact have a small gas species dependence, and that a scale based on hydrogen,
although not ideal, was probably accurate to better than 0.01°C. In 1889 the Conf´ erence
G´ en´ erale des Poids et Mesures (CGPM), at its first meeting, adopted the first official
temperature scale, the ‘normal hydrogen’ scale. Because of the known dependence of
the scale on the non-ideal properties of hydrogen the initial filling pressure of the thermometer was also specified. This scale was still not a metric scale; instead the scale
was defined by fixing the interval between the ice point and steam point to be 100°
C.
The scale was an ordinal approximation to an interval scale based on two fixed points
and an almost linear interpolating instrument.
The first proposals to use the variation of electrical resistance to measure temperature
came from Davy in 1821 and later Siemens in 1861. However, it took the elaborate
experiments and refinements in the construction of platinum resistance thermometers
by Callendar to get the resistance thermometer accepted. By comparing the platinum
resistance thermometer with the gas thermometer, Callendar and others were able
to show that platinum has a parabolic or quadratic characteristic. In 1899 Callendar
proposed a temperature scale that would be more practical than the ‘normal hydrogen’
scale, based on three fixed points: the ice point, the steam point and the melting point
of sulphur. The scale was defined by fixing the interval between the ice point and steam
point to be 100°C, with the sulphur point defined to be 444.5°C as determined by a gas
thermometer calibrated at the other two points. Callendar’s proposal, again an ordinal