Celsius temperature scale
Thermoscopes and thermometers (i.e., a thermoscope provided with a scale) are very old inventions, which can be dated to the 16th and 17thcenturies. The ordinary thermometer of today with a spherical glass bulb attached to a thin glass tube can be traced to Florence, Italy, at the middle of the 17th century. The amount of heat (the temperature) could be read by help of markings on the thermometer stem forming, e.g., a 100-degree or a 50-degree thermometer owing to the number of markings. As thermometer liquids one usually used mercury or 'spirit-of-wine'. Because of the large difference in expansion coefficient between the two liquids a mercury thermometer must have a more narrow glass tube than the spirit-of-wine thermometer.
Through the years thermometers became more common, and appeared in several European countries. Thus, at the beginning of the 18th century therefore a number of different temperature scales were in use. It is important to realize that thermometers were mainly used in meteorology, in horticulture, and sometimes also for indoor purpose (if you could afford to buy one). This implies a very limited temperature domain of practical interest, namely a temperature region from about plus forty to minus twenty in modern Celsius degrees.
A temperature scale must be based on one or two well defined temperatures, called fixed points. For those fixed points it was natural to choose temperatures within the domain of practical interest. One natural fixed point could be the human body temperature. Of course the freezing and boiling points of water were used, although the latter lied outside the region of interest. Temperatures of local origin were also used, such as the temperature of the observatory cellar in Paris. This choice was convenient since it was constant through the whole year (and still is today at 12 °C), while the freezing point of water called for wintertime. One English temperature scale had a fixed point at 'the highest temperature in sunshine in London'!
Anders Celsius is the first scientist to perform and publish careful experiments leading to the definition of a truly international temperature scale on scientific grounds. Already as a student in Uppsala in the early 1720s he assisted professor Erik Burman in meteorology observations, and got acquainted with several thermometers with different scales. In doing so he perhaps realized the necessity of a reliable thermometer scale.
Celsius became a professor of astronomy at Uppsala University in 1730. He devoted much time to determine two fixed points, i.e., the freezing and boiling points of water, which could define a scale of universal use. After years of experiments he published in 1742 a detailed report on the determination of the two fixed points. In the essay he starts with a discussion of two ways of defining a temperature scale. The first method is to use only one fixed point. The scale is then based on the thermal expansion of the specific thermometer liquid, such as 'spirit-of-wine' or mercury. This method was, e.g., used by Réaumur who chose zero at the freezing point of water as the single fixed point in his original spirit-of-wine thermometer, where each degree equalled the expansion of one thousandth of the volume of the liquid. He adjusted the dilution of the spirit-of-wine in order to place the boiling point of water at 80 degrees. Celsius definitely preferred to use two fixed points, and then divide the temperature interval between the fixed points in a suitable number of degrees (i.e. 100 degrees). This explains his choice of title of his report: Observationer om twänne beständiga Grader på en Thermometer ('Observations of two well-defined degrees on a thermometer'). As fixed points he chose the temperature when water starts freezing and the temperature of boiling water.
Freezing point of water
Celsius observed that 'the same degree of cold' that is needed for making water freeze should reappear when the ice melts to water. The freezing point should be read when the thermometer had been immersed in 'cloggy' snow for at least half an hour. He made different kinds of experiments; he even read the same freezing point temperature, when he placed the thermometer in a bucket with snow into a hot stove! He also investigated the possible influence of barometric pressure on the freezing point (in analogy with Fahrenheit's discovery of the change in boiling point of water with the atmospheric pressure). Naturally, he found no such effect on the freezing point.
In the 18th century it was not obvious that the temperature of melting ice is independent of the geographic latitude. Celsius took part in the French expedition under Maupertuis to North Sweden in 1736-37 (in order to determine the length of a latitude degree). With one and the same Réaumur thermometer Celsius measured the same freezing point of water in Torneå as he had observed in Uppsala and Paris. Celsius concluded that the freezing point of water is well suited as a fixed point on a thermometer scale.
Boiling point of water
The determination of the boiling point called for a long series of experiments. He placed the thermometer deep in the boiling water in a teakettle and took a reading when the water had been boiling hard for several minutes. For the experiments Celsius used water from different sources, such as 'snow-water', water from the Fyris river and water from three different wells. One well had bad water not suitable for tea. Despite this poor water he concluded that water of different origin can be used for the determination of the fixed point.

Celsius was well aware of the observations made by Fahrenheit in Amsterdam, indicating that the temperature of boiling water is dependent of 'the height of mercury in the barometer'. Celsius decided to observe carefully the boiling point at different barometer readings. He marked an arbitrary point on the thermometer stem and measured the position of the mercury level downwards from this mark when the water was boiling. In the diagram his thermometer readings are plotted versus the barometric pressure. The thermometer readings are given in 'Gran', i.e. one hundredth of a Swedish decimal inch, which equalled 29.69 mm. The readings are given with an accuracy close to 1 Gran, i.e., 0.3 mm! The height of the mercury column of the barometer is given in Swedish inches.
Celsius found that a change in atmospheric pressure of one inch on a mercury barometer causes a change in the boiling point of 8 Gran on his thermometer scale (as can be checked in the diagram). He divided the distance between the boiling and freezing points, which was 792 Gran on the thermometer stem, in one hundred degrees. The size of one degree was then approximately 8 Gran. Celsius noticed that the change of one inch in barometric pressure approximately equalled a change of one degree in the boiling point.
Celsius' results are remarkably accurate in the light of modern thermodynamics. A straight line in the diagram indicates the present relation between atmospheric pressure and the boiling point of water. Anders Celsius certainly made very accurate measurements in spite of rather primitive experimental conditions.
Celsius concluded that the temperature of the boiling point of water is reliable as a fixed point only in combination with a defined atmospheric pressure. As a standard Celsius proposed the pressure of 25.3 inches of mercury, which he considered to be the mean value of the atmospheric pressure. This equals 751.2 mm Hg or 1001 hectopascal (millibar). Celsius further gave practical instructions for a correct marking of the fixed point if the barometric pressure deviates from the standard pressure of 25.3 inches of mercury. He marked the fixed point of boiling water with the figure zero on his thermometer scale. The freezing point he marked with 100.
The position of the degree zero on temperature scales (as well as the direction of the scale and the size of the degrees) has varied in the course of time. The French Réaumur scale had zero at the freezing point. For other scales one tried to place zero outside the ordinary temperature region, thus avoiding the mixture of positive and negative numbers. Hence, the zero could be placed at a reasonably low temperature, hopefully below temperatures in wintertime. This method was used by the Danish astronomer Ole Römer and then adopted by Fahrenheit. The French astronomer Joseph-Nicolas Delisle used a thermometer with zero at the boiling point and 150 at the freezing point, thus creating a reversed scale with increasing numbers for decreasing temperatures, and avoiding negative numbers. In 1737 Delisle sent two thermometers to Celsius. One of them has survived till today, and is under display at Museum Gustavianum at Uppsala University. Celsius was well accustomed to the Réaumur thermometer, but when measuring the boiling point he used a Delisle thermometer instead of the more bulky Réaumur thermometer. Celsius then followed Delisle in setting the zero at the boiling point of water. The freezing point of water, marked with the figure 100 by Celsius, corresponds to 151.8 degrees on the Delisle-thermometer and not 150 as indicated by Delisle.
The reversed temperature scale was not as awkward as we think today and it served its purpose at that time. The change to our modern direct scale was inevitable, however, and there is no sense in trying to give the credit to any single person. In fact the inversion of the scale was hardly any great intellectual achievement. In the meteorology journals at Uppsala (after the death of Celsius in 1744) thermometers with the direct scale appeared under different names, such as 'Celsius Novum', 'Ekström' and 'Strömer'. Mårten Strömer was Celsius' successor as professor of astronomy, and Daniel Ekström was a very skilled instrument maker in Stockholm. Linnaeus became acquainted with the Fahrenheit thermometer in Holland. In Uppsala he of course learned about Celsius experiments, including the essay. In 1744 he ordered a 100-degree Celsius thermometer from Ekström with direct scale and zero at the freezing point to be used in his greenhouse.
Obviously Celsius felt the importance of creating a reliable temperature scale as is expressed with his own words (in translation): 'then one can be sure that several such thermometers placed in the same air will always show the same degree, and for instance a thermometer made in Paris will at the same heat show the same height as a thermometer made in Uppsala.'
Centigrade thermometers with a scale increasing with temperature and based on the freezing and boiling points of water had in fact also independently appeared in Switzerland and France in the early 1740s. However, the significance of Celsius' achievement lies in his careful determination of the two fundamental fixed points, especially his definition of the boiling point in relation to the atmospheric pressure. Celsius' centigrade scale was very soon adopted by the scientific community.