116 Prof. D. Mendeleeff on the Variation in the 



probably the conflicting evidence of the observers has its 

 source in the errors accompanying the determination of the 

 linear expansion of solids by the method of comparison. 



4. No less doubtful are the determinations of the expan- 

 sion of glass, carried out between 0° and 100°, by means of 

 mercury, because the true mean coefficient of expansion of 

 mercury adopted by different observers varies greatly, and it 

 is impossible to say at present to what extent the actual value 

 differs from those taken *. After the determinations of 

 Dulong and Petit, this mean coefficient of expansion of 

 mercury was taken as 5^-= 0*00018018 ; Regnault deduced 

 the value 0*00018153 from his classical researches. By 

 applying different methods of interpolation to the determina- 

 tions of Regnault (expressing the result multiplied by 10 s ), 

 Bosscha (1874) found 18211; Galton (1873), 18181; Wullner 

 (1874), 18252; Mendeleeff (1875), 18210 + 7 f; Levy (1881), 

 18207; and Broch (1885), 18216. The chief cause of this 

 discrepancy in the results lies in the circumstance that 

 Regnault, out of 135 determinations, only made 32 for 

 temperatures below 100° ; his cold column of mercury during 

 the experiments was not at 0°, but had a temperature from 

 + 10 to + 18°, and separate experiments present differences 

 amounting to the discrepancies above stated. Without fresh 



* Tlie method pursued up to now since the time of Dulong- consists in 

 determining the true expansion of mercury, then from it that of glass, and, 

 knowing the latter, from the apparent expansion the true expansion of 

 water and of other liquids. The great difference (7-fold) between the coeffi- 

 cients of expansion of glass and mercury constitutes the weak point of this 

 method,because the expansion of mercury must be ascertained with a degree 

 of precision scarcely attainable in experiments. But the most perceptible 

 want in this method lies in the fact that the classical researches of 

 Regnault afford very little material for an accurate judgment of the 

 expansion of mercury between 0° and 100°, where questions of expansion 

 are mainly concentrated and where observations are most within reach. 

 The great necessity for new determinations of the true expansion of 

 mercury from 0° to 100° has already repeatedly been made manifest. I 

 unite my voice to that of many others ; but I will add that when we have 

 reliable figures expressing the expansion of water it will then be easier 

 to obtain them for mercury also, for experiments with water, between 0° 

 and 100°, are easier and more convenient than with mercury, and conse- 

 quently their precision may be greater. In a word, the determination 

 of the true expansion of water is a question urgently requiring solution. 



t I think it will not be superfluous here to draw attention to the fact 

 that in the determinations of the expansion of mercury from 0° to 100° 

 (according to Regnault) there, without doubt, exists an error, attaining +7 

 or 8 millionths of the volume ; and therefore it is necessary to recognize 

 this possible error in the calculations and to determine it, which I have 

 endeavoured to do. The calculations of Levy and Broch, made subse- 

 quently to mine, justified my conclusion, since the difference between 

 their result and mine does not exceed +7 millionths. 



