METHODS OF PYBOMETBY. 
: selenium vapor ma i re- 
fractory glass, and recommends for that purpose tin pert 
frere by, which is nearly rigid at this '■ "ire. Other i 
vant - of Troost 1 on the permeability of platinum to b 
and of silver to oxygen at high temperature have been adverted to. 
Berthelot 2 points out the occurrence of unstable platinum hydrides. 
The value of the boiling point of zinc, to which the later research 
Deville and Troost had given a value compatible with that o: 
rei. was soon to be further fixed in position by the research oi 
Using a triple jacketed boiling point apparatus of enameled iron, he 
found by Devilie and Troost's methods that zinc boils at 930°, thus 
giving further warrant to the data of Becquerel and Deville and 
Troost. In view of the accordance of these data, 
temperature measurement may be regarded as solved with some i 
racy as far as 1,500°. The greater share of the credit for this res; 
undoubtedly due to Deville and Troost. notwithstanding their unfortu- 
nate begiuning and the fact that they allowed the subject to slumber 
in their hands for so many years. Violle refers to the problem of mere 
high temperature measurement as being one of great simplicity, and 
finds his main difficulty in the construction of constant temperature 
apparatus. My experience is the reverse of this. It is not very 
difficult to get the zinc point; but it is difficult to obtain thoroughly 
accordant values for it when (liferent bulbs are used. Violle, who used 
but a single bulb (so far as I have been able to make out), obtains val- 
ues which are almost identical, but which really apply only to the par- 
ticular bulb in hand. The error possible in measuring the constants of 
the bulb is one of a very serious kind, and in case of a single bu] 
remains arbitrarily, fixed. The data of Deville and Troost, which were 
obtained by using a large assortment of bulbs, bear evidence to this. 
The differences between their later results are by no means insignifi- 
cant, and these observers were most scrupulous in perfecting their 
methods, even to the fine points of experimental detail. Becquerel, in 
using divers thermometer bulbs, encountered the same wide limits of 
error. Regarding Becquerel's later and very low values, moreover, it 
is probable that the criticism of Deville and Troost applies. Becque- 
rel's boiling-points apparatus was imperfect. In the case of so large 
an object as the air-thermometer bulb, at so high a temperature as the 
boiling point of zinc, its data can not be regarded as identical with the 
temperature of the vapor unless it be in actual contact with it. Evi- 
dence bearing on all these points will be repeated in Chapter IV. 
Eegarding Deville and Troost's experiments on the coefficient of ex- 
pansion of porcelain, a short critical remark relative to the occurrence 
of permanent dilatation is in place here. When a porcelain rod is sus- 
1 Troost: Ibid., vol. 98, 1884, p. L427. 
-Berthelot: Ann. ck. et phys., 5th series, voL30,1883, p. 530. 
3 Violle: C. R.. vol.94, 188?, p. 720. 
(689) 
