THE THERMOMETER. 



139 



part of the tube will be occupied exclusively by the vapor of alcohol, which 

 will be raised by the heat. The pressure of this will prevent the remaining 

 spirit from boiling ; and, the increase of temperature not being limited by ebul- 

 lition, the liquid will continue to be indefinitely dilated. The indications of 

 such a thermometer, however, at a higher temperature, are not, like those of 

 mercury, equable. The scale, therefore, if intended to indicate equal varia- 

 tions of temperature, should not be resolved into equal divisions, but should be 

 divided experimentally by comparison with a mercurial thermometer. The 

 cause of this has been already explained in our chapter on the dilatation of 

 liquids. As we approach the boiling point, the rate of their dilatation sensibly 

 increases, so that equal changes of temperature would correspond to increasing 

 divisions on the scale. 



It is of the most extreme importance, in the construction of mercurial ther- 

 mometers, that the fixed terms of melting ice and of boiling water, which are, 

 in fact, the foundation of the accuracy of the instrument, should be determined 

 with great care, and should be rendered independent of all causes which could 

 produce accidental variation in them. 



In determining the freezing point, care should be taken not to confound the 

 temperature of melting ice with the temperature at which water begins to freeze. 

 It will be explained hereafter that, under certain circumstances, water may be 

 cooled considerably below the temperature of melting ice before it becomes 

 solid ; and, consequently, the temperature at which it freezes or solidifies can- 

 not be considered as fixed. 



The temperature, however, at which ice or snow melts is constantly the 

 same, provided the water of which the snow or ice is formed be perfectly pure. 

 If this water, however, hold salts in solution, it will freeze at lower tempera- 

 tures, and, consequently, it will melt at lower temperatures. Rain-water or 

 pure snow, when melted, will, however, always give the lower term of the 

 thermometric scale, without any liability to error. 



The determination of the higher term of the scale is, however, attended with 

 more difficulty, and with more numerous causes of variation. It is, in the first 

 place, necessary that the water should be pure and free from all admixture with 

 foreign substances. Thus, water charged with salts will boil at temperatures 

 different from pure water. It is necessary, therefore, that the water with 

 which the experiment is made should be either rain-water or distilled water. 



There is, however, another cause, which more constantly affects the temper- 

 ature at which water boils. It appears, as may be elsewhere shown, that the 

 pressure exerted on the surface of the water, whether of the atmosphere or 

 from condensed or rarefied air, will affect its boiling temperature. If this tem- 

 perature be increased, the water will receive a higher temperature before it 

 will boil ; and if it be diminished, it will, on the other hand, boil at a lower 

 temperature. Thus, water in an exhausted receiver will boil at a much lower 

 temperature than when exposed to the atmosphere. These circumstances may 

 be more fully detailed in another lecture ; but, for the present, it will be suf- 

 ficient to allude to them, in order to explain why the pressure of the atmo- 

 sphere must be attended to in determining the boiling point on a thermometric 

 scale. The barometer, from day to day, and from hour to hour, is subject to 

 fluctuation, and a corresponding change takes place in the pressure of the at- 

 mosphere ; consequently, although this variation, being small, cannot affect the 

 temperature at which water boils to any considerable extent, yet it does affect 

 it so much as to render it an object of important calculation in determining an 

 element such as that now under consideration, upon which the accuracy of all 

 thermometric indications must depend. To determine this fixed temperature, 

 therefore, it will be necessary, either to recur to some phenomena not affected 



