268 Wadsworth — Determination of Specific 



that the minimum quantity of water should always be used, 

 so far as effects of errors in t are concerned. From (3) we get 

 directly the percentage variation of s for a given error in t, 

 when 6 — t is known. 



E. g. Suppose 6 — t — 10°, then A^ = 0*1 At . s. Now if we 

 assume the maximum allowable error to be 1 in 1,000, or *1 per 

 cent, 



A t s = 0-001 s. . and At = 0*01 



or the temperature t must be read to *01 of a degree. If 

 0—t=5°, the temperature t must be read to 1-200 of a degree. 

 On the other hand, if At=^° (and we cannot read much closer 

 than this with a thermometer) 



A t s = 0-002 . s for (0 — t) = 10°, an error of \ per cent. 



A,s = 0*004. s " " = 5° " " " nearly J per cent. 



(B) Error in T. 



Errors in T may arise in three ways : 



1st. By error in reading T (the only error which affects t). 



2d. By the fact that the actual temperature of the body may 

 not be that of the atmosphere which surrounds and is heating 

 it, as indicated by the thermometer placed therein. This is 

 avoided by keeping it surrounded for a long time with an 

 atmosphere maintained at a constant temperature (one and one- 

 half to four hours or more, according to the thermal conduc- 

 tivity of the body . in question). Theoretically it would 

 require an infinite time for complete equalization of its tem- 

 perature with that of the surrounding air. The time required 

 for equalization will be very considerably diminished by mak- 

 ing the surface of the body as large as possible compared with 

 its volume, or by using thin sheets of the metal or a ball of 

 wire in place of a solid mass. This will be of advantage also 

 in diminishing the time required for the hot body to impart 

 its heat to the water. 



3d. By the loss of heat during the time while the body is 



being transferred from the heating chamber to the calorimeter, 



which is of course diminished by making the time as short as 



possible, and by protecting it as well as possible during 



transfer. 



ds 

 The error of s due to error in T, or — = A T «9, will be 



ds_ A _ (M + W) c (6— t)w \ 



ST" rS ~~ MT-.0)P' " 1 u . 



c_ (M + W) (fl-Q [ ^ a) 



~ w.~ (T-ey- J 



