260 PETTEBSSON, OK WATER AND ICE. 



This table shows, that the apparent dilatation of the mer- 

 cury in the dilatometer was equal [= 27. . . millimetres = 

 0.006242 cc] for every centigrade degree observed on the scales 

 of the Geissler thermometers Gjj & Guj. The variations [O.s 

 unto 0.7 millimetre] are due to the difficulty of rnaintaining a 

 steady temperature in the mercury vessel for a sufficiently 

 long time. I estimate the unavoidable oscillation of tempera- 

 ture at abont 0\o2 C for temperatures above — 10° C. For 

 lower temperatures (as for example — 16° to — 17° C) the error 

 is somewhat greater. 



This speaks strongly in favör of the Geissler thermo- 

 meters. They are in fact excellently graduated after the prin- 

 ciple of the mercury thermometers. I therefore, upon reflection, 

 thought it advisable, to apply this principle exclusivel}^ to all 

 nieasurements of temperature given in this paper. The num- 

 bers therefore refer to the indications of correctly calibrated 

 mercury thermometers. Besides, all measurements of tempera- 

 tures of ice or water, hitherto published in hydrographic re- 

 searches, are made with mercury thermometers, and the fol- 

 lowing results would be incapable of comparison with those 

 of my predecessors, if I were to apply another standard of 

 temperature. Still I think it to be a serious inconvenience, 

 that science nominally proclaims one standard of temperature 

 [the dilatation of dry air] hut practically applies another [the 

 dilatation of mercury]. 



The preceding experiment also serves another purpose. 

 For the following determinations it was necessary to know 

 the coefficient of absolute dilatation of the glass reservoir of 

 the dilatometer. This number is easily calculated from the 

 last column of the table 1 series a & b. If 



q denotes the coefficient of absolute dilatation of mercury, 

 ft » » » >/ apparent >- » x 



g » » » » cubic dilatation of glass, 



then 



g = q — i^- 



J^ being a constant (see last column of table 1, a & b), the 

 value of g depends upon what number is substituted for q. 

 Here another great inconvenience arises from the uncertainty 

 of the absolute aoefficient of expansion of mercury at 0° and 

 lower temperatures. According to Regnaulfs determinations 



q = 0.00017905 at 0°. 

 According to Wullner's recalculation of Regnault's experi- 

 ments, it ought to be 



