*A&u&.l CALIBRATION OF ELECTRICAL PYROMETERS. 115 
constants then calculated from these data and used for trial extrapo- 
lations. It appears from this comparison that the results of these trials 
are almost perfectly coincident. In case of No. 17 the difference at 
1,500° is only 4°; in case of No. 18, 14°; in case of No. 22, 9°. Now, 
when it is remembered that the results of Chapter 1 are obtained by 
the large vapor apparatus there described and that the results of this 
chapter are obtained by the small practical forms, the coincidence of 
the results of this exceedingly severe test is most remarkable and 
gratifying. In Table 16, moreover, the relatively small differences be- 
tween the parenthetical results for observed and calculated e, is further 
evidence in favor of the statement just made. 
If, however, we include the boiling point of zinc in the present com- 
parisons and derive the constants by the method of least squares, the 
relatively large differences thus obtained at once show that the quad- 
ratic equation assumed is no longer applicable, so far as the results in 
hand are concerned. The loci corresponding to these new constants 
differ at 1,500° by amounts as follows: In case of No. 17, by 126°; in 
case of No. 18, by 115°; in case of No. 22, by 119°; in case of No. 35, 
by 126°; in case of No. 36, by 130°. Similarly the differences between 
observed and calculated e in Table 16 (without parentheses), differ by 
large values. To obtain a notion of the nature of this difference it is 
well to insert the observed result for zinc in the chart, Figs. 21 and 22. 
When this is done it appears that the position of the zinc point bears 
no observable relations to the positions of the sulphur, mercury, aniline, 
and water points. There are three causes for this large discrepancy to 
be considered: 1. Either the relation between temperature and electro- 
motive force in case of platinum-iridium elements is circumflexed or 
anomalous between 500° and 1,000°; or the accepted value of the boil- 
ing point of zinc, 930°, is too large by about 75°; or the boiling point 
of zinc in the crucible calibration has not been reached. The last of 
these suppositions is easily disproved, both by the fact that during 
ebullition zinc dust escapes from the top of the crucible, and that after 
breaking the cold crucible the evidences of ebullition are apparent in 
drops of zinc scattered against the walls and solidified there after cool- 
ing. Again, the temperature on the outside of the crucible must have 
been at least 500° above the boiling point of zinc. Finally there is au 
almost complete coincidence between the zinc data for the large cruci- 
bles, in Chapter I, and the present data for small crucibles, as will 
presently be shown more at length. Hence the source of the zinc dis- 
crepancy in question is to be referred to one of two causes: either the 
platinum-iridium thermo-couple shows a circumflex like anomaly in the 
relation between electromotive force and temperature between 500° 
and 1,000°, or the values heretofore assumed for the boiling point of 
zinc {ca. 930°) are too high b 75°. To decide between these two diffi- 
culties, the importance of which is here regarded merely from the 
(769) 
