mv&.] DEGREE OF CONSTANT TEMPERATURE. 77 
rere partially destroyed by these accidents, necessitating their replace- 
ment by No. 19. The constants and thermo-electrics of the new couple 
>eing different from the old, it is clear that the subsequent values for 
F zn are no longer immediately comparable with the preceding values for 
r zn . To make them as nearly as possible comparable, however, the fol- 
owiug method was pursued : The values for e 20 , corresponding to Nos. 
9, 20, 23, and obtained on the 2d of February, on the 24th of April, and 
n the 12th, 26th, and 30th of May, were averaged, and this mean value- 
rs assumed to correspond with the mean values of T zn given by No. 
8 on the same days, respectively. A glance at the tables below shows 
hat on these days one of the series of elements, Nos. 19, 20, 23, and one 
f the series of elements, Nos. 17, 18, 22, were simultaneously compared, 
^rom these data the constants of the former set (Nos. 19, 20, 23) and a 
raphic representation were investigated ; from this finally we took the 
alues of T zn given in Table 5. In this way the break in the results is 
educed to the least value possible under the circumstances. 
It is curious that in the subsequent work we were not able to obtain 
eries of results as satisfactorily constant as in the earlier experiments, 
'o speculate on the causes for discrepancy is of course futile, and the 
iter data subserve no other purpose than that of comparing the thermo- 
lectric behavior of the couples simultaneously calibrated. 
INFERENCES RELATIVE TO LOW PERCENTAGE ALLOYS. 
^eduction of data. — In view of the insufficient degree of constancy 
bservable in the above results as a whole, it is necessary to resort to 
ii artifice by which all thermo-electric forces may be referred to a fixed 
iterval of temperature, T—t. For the lower limit of this interval we 
ilected 20°, a temperature as near the mean value of £as practicable; 
>r the upper limit 930°, the assumed value of the boiling point of zinc, 
hen the reduction to the lower limit has the value 
e w -e=(t-20) {a+b(t+20) J>; 
ad the reduction to the higher limit the value 
e 930 -e=:(930-T) (a+b (930+T)). 
The method of correction was therefore a quadratic interpolation by 
hich the thermo-electric interval is rectified at each end, and thus re- 
ticed to the uniform temperature interval. The constants a and b 
ere carefully redetermined in a final calibration, so that the sole re- 
aining difficulty in the equation is the choice of T. Fortunately it is 
lly the variations of T with which the above equation is concerned, 
id this may be obtained either by linear reduction of the therino- 
ectric datum T s to 930°, or we may calculate the constants for each 
ement throughout the interval 0° to 930°, and then use the T ZQ so 
)tained. The first method is less accurate than the second without 
sing insufficiently accurate. At the same time the first method is so 
uch more expeditious that we applied it. 
(731) 
