Variation in the Density of Water with the 'Temperature. 99 



those obtained by Matthiessen, we find that he fixes the maxi- 

 mum point of his curve at an alloy containing 37 per cent, of 

 tin. This number agrees pretty closely with the formula AuSn 

 if we take tin as having au atomic weight of 11 8, and gold as 

 having an atomic weight of 197. The formula, then, AuSn, 

 the maximum point of the electric conductivity-curve, and 

 the point where the E.M.F. abruptly rises, lie between the 

 alloys containing 36 per cent, and 38 per cent, respectively of tin. 

 This method, then, confirms the existence of one of Matthiessen's 

 supposed compounds, but has not confirmed the existence of 

 the other two. It agrees, however, with the results obtained 

 with the copper-tin alloys, and taken in conjunction with 

 them shows that the point of maximum conductivity is the 

 point where a compound exists. This does not seem an un- 

 reasonable result. Compounds of the formula An 4 Sn and 

 An 2 Sn 5 are not very probable. Furthermore, the most common 

 form for the curve of electric conductivity to take in the case 

 of metals not combining is a (J* If we regard the compound 

 AuSn as practically a new metal of fairly good conductivity, 

 we should expect two regions of depression between AuSn 

 and gold, and between AuSn and tin to make the results corre- 

 spond to those usually obtained. But we may derive a further 

 conclusion from the E.M.F. measurements in the case of these 

 alloys. The change in E.M.F. in passing over the compound 

 amounts to '25 volt. Now this is a measure of the energy 

 absorbed in the cell in decomposing the compound and may, 

 therefore, be taken as an approximate measure of the heat of 

 formation of the compound itself. 



XI. The Variation in the Density of Water with the 

 Temperature. By D. Mendeleeff*. 



THE expansion of water with change of temperature, 

 although presenting great scientific interest, both for 

 the comprehension of the action of heat upon matter and also 

 for many experimental investigations, cannot yet be considered 

 as elucidated in its theoretical aspect nor sufficiently worked 

 out experimentally. It will be seen from Tables I. and II. that 

 in the best existing determinations, and after the introduction 

 of possible corrections t, we encounter discrepancies which, 



* Translated by G. Kamensky. Communicated by Prof. Crum Brown, 

 F.B.S. 



t For instance, possible corrections for the value of the true coefficient 

 of expansion of mercury, for the readings of the mercury thermometer, 

 for weighing in air, for the increase of the expansion of the vessel with a 

 rise of temperature, &c< Further on I consider certain of these correc- 



H2 



