148 
MEASUREMENT OF HIGH TEMPERATURES. 
[HULL. 54. 
Ibese wires, B, (7, D, E, F, were now fused thoroughly together aud 
kept melted before the blow-pipe for fifteen minutes ; thereupon drawn, 
annealed, and softened. The new wire on being tested gave the follow- 
ing results : 
Table 38. — Electrical tests for purity. 
No. 
Galvanics. 
No. 
Tbermo-electrics. 
r t 
t 
?xlO fi 
«t 
t 
T 
e 
(B+ C+ D+ 
E+F) 
786 
23 
1964 
15.4 
-(B+O+D 
25 
358 
-241 
The new wire was now again purified by fusing for fifteen minutes ; 
then drawing and annealing. The results obtained were : 
Table 39. — Electrical tests for purity. 
No. 
Galvanics. 
No. 
Thermo-electrios. 
rt 
t 
q x 10 r ' 
St 
t 
T 
e 
(B' + O+D 1 
+ E'+F') 
504 
23 
2781 
14.0 
-{B'+C' + D' 
-\-E'j-F') + 
25 
358 
-170 
The direct result for e in case of the couple — (7+ A agrees well with 
the result which may be compounded from Tables 35 and 37. If now 
we compare the mean values of s t from Table 37 (17.3) and the values 
in Tables 38 and 39 (15.4, 14.0), it appears that the decrement or volatili- 
zation of the impurity is greatest during the first minutes of the fusion. 
Hence, since it was customary to heat the buttons beyond the fusing 
point for several minutes before alloying, and since, moreover, the range 
of specific resistances due to alloying is enormously large, it is probable 
that impurities in the platinum have not distorted the data for resistance 
and density to a serious extent. 
The same degree of reliance can not be placed on the thermo-electric 
data. Indeed, the behavior of the fused sample here is curious. In 
Table 36, for instance, the values of e oscillate, and are actually greater 
after the third fusion than originally, whereas the values of s t for the 
same wire, under the same circumstances, decrease with the utmost 
regularity. The mean values of Table 37, as compared with the values 
in Tables 38 and 39, show a similar irregularity of decrease. It follows 
therefore that the results for ax 10 3 must be regarded as distorted by 
an arbitrary constant, which remains nearly the same for a single set of 
alloys, but which varies from one set to another by an amplitude, the 
maximum total value of which can not be safely put less than 1,000. 
In other words, the error of ax 10 3 from series to series may be as large 
as ±500. It is only the more powerful thermo-electric combinations in 
the above, i. e., alloys ot gold, cobalt, silver, manganese, chromium, iron, 
(802) 
