BRIDGMAN. — THERMODYNAMIC PROPERTIES OF WATER. 301 



the above data is 0.0024. Tliis leaves a decrease of 0.0127 to he 

 accounted for by the increase of pressure of 2G40 kgni. which gives 

 a mean compressibiUty over this range of 0.0000048, a Uttle more 

 than one third of the compressibiUty of the H([uid over the same 

 range. 



The direct determination of the compressibihty of the ice was made 

 by two different methods. One of these M'as the same as that used 

 roughly in the preceding paper, that is by finding the difference of 

 the slope of the curves plotting piston displacement against pressure 

 above and below the transition point to the solid. The values obtained 

 in the preceding paper for this were very rough. In these determina- 

 tions the cylinder was very much more carefully seasoned, and the 

 readings were made with all the precautions which had been sug- 

 gested by all the experience of this paper. Two determinations of 

 this quantity were made at 0° and also two determinations at 20°. 

 The two values for the difference of compressibility differed by 2.5% 

 at 0° and by 0.7% at 20°. The value found for the difference was 

 0.0000087 at 0° and 0.0000067 at 20°. Combining with the values 

 given already for the compressibility of the liquid, this gives for the 

 compressibility of ice VI O.O549 at 0° and 6360 kgm., and O.O543 at 20° 

 and 9000 kgm. Mean O.O546. 



The second method for determining the compressibility was exactly 

 the same as that for finding the same quantity for the liquid, com- 

 paring the displacements when the apparatus was filled with ice and 

 kerosene with those when the ice was replaced by bessemer steel. 

 This determination was made over a wider pressure range, to find if 

 possible the variation of compressibility with pressure. No variation 

 with pressure could be found over a range of 4500 kgm. at 0° and 3300 

 kgm. at 20°. The absolute values do not agree with those found 

 by the two other methods, however, the figures being O.O53I at 0° 

 and O.O535 at 20°. The cause of the discrepancy is not clear, but is 

 probably connected in some way with the hysteresis of the cylinder. 

 The hysteresis was not regular for these small pressure ranges, being 

 at times almost negligible, and again being as large as for almost the 

 entire pressure range from atmospheric pressure to the maximum. 

 There seems little question but that the greater weight is to be attached 

 to the values found by the first two methods. This third determina- 

 tion does show, however, that the variation of the compressibility 

 with pressure and temperature over this range is so small as to be 

 beyond the accuracy of these measurements. In selecting the best 

 probable value for the compressibility the only weight that will be 



