( '^^1 ) 



The (lata about 7',^/ could be represented very accnratelv by a 

 quadratic function. 



We may expect that for values of ,v outside the range of obser- 

 vation these formulae will show important deviations from the 

 experiment, because errors in the observations for .v = 0.1 and 

 ,v = 0.2 pass over increased into the values for the other range. This 

 holds especially for T^k, pxk and z^t, which are exposed to so much 

 more sources of errors than the other quantities. 



§ 8. From the formulae given above we find : 

 « = — 0.6563 , ,i = — 1.0871, 



y— __ _^M ^ = 0.5422% 



Vic V dx Jx^O 



while u — ^ = 0.4308. That this value does not agree better with 

 the value of y found directly, while in ^ 6 we found that for the 



two mixtures the relation was properly satisfied, is due to 



Txk 



the representation by the above-mentioned fiuictions, which may be 

 expected to show just at the limits the largest deviations with regard 

 to the difïerential quotients. 



With a view to this the agreements between — - — ; — and 



2k\ dx y^:=o 



1 rdi\r\ _ ^ 1 fd.px,i\ , 1 (dp,; . 



I and between — — — and — — — must be 



Tjc V d.v Jx = o pk V dcv Jx — o Pk\ dx Jx — o 



considered as satisfactory. 



1 fdr,.,,c 



The values of « and /i derived irom the \alues of —- — - — 



ll\ dx /x=:0 



and — ( — - ) by means of the formulae {2<i) and (2/>) of 



pk \ dx y,=o 

 Comm. No. 75 and the values found in this Comm. IV (p. 577) 



for the coefficients C\, ( "T" ) «^"«^ C, \^-y) o^'^'^i'^'i"? there, and 



1 fdTr,\ , 1 fdpxr\ 



the values of it and .i derived from — --— and — —— 



I k\ dx y,.^ „ .//, V dx Jj.^Q 



(as according to Comm. No. 81, Pi'oc. Oct. 1902, p. 350 the sanu' 

 formulae hold for these quantities) have been combined with the 

 values derived from the critical points K in Table XXVI, where the 

 first column indicates from which point the data have been derived. 



1) Gomp. Verschaffelt, Comm. N". 81, Pioc. Oct. 190i2, p. 32.J. 



