TRANSACTIONS OF SECTION B. 461 
due, probably, to the removal of small quantities of bases derived from more 
complicated radicals than ethyl. 
The whole of the sample of triethylamine was now fractionated with great care, 
and the portion boiling between 90° and 91° C, selected for a repetition of the pro- 
cess; the middle portion from this second distillation, boiling between 90:2° and 
90-4° C., was again separated into three fractions by a new distillation, and the 
molecular weights of the hydrobromates of the respective samples determined. 
The results are given in the following table :— 
Weight of | Molecular 
Silver in Weight of Remarks 
vacuo j\N(C,H;);HBr 
Experi- | _ Weight of 
ment | Salt in vacuo 
L. 7-06272 | 418778 | 182-025 | First sample boiling 90°-91° C. 
2 3. 99 fs Second fraction of I. boiling 
II. 64418 38199 18201 1 50-32-9042. g 
tm. | 15-46765 | 9:18495 | 181-756 es Pay of I. refraction- 
“Middle and largest fraction of 
| II. refractionated. 
f Highest boiling point portion 
| of II. refractionated. 
EV 11°95685 70902 182:012 
NG 13°9522 82664 1827166 
Another sample was precipitated as acid ferrocyanide, and on conversion into 
the hydrobromate its equivalent was found to be 181°752. 
The results of the analyses clearly prove that the sample of base is not homo- 
geneous, the presence of higher and lower bases being clearly revealed. At the same 
time the middle and largest portion of the last fractionation has a molecular weight 
which may provisionally be accepted as that of pure triethylamine. If the mole- 
cular weight of the hydrobromate be 182-012, then we have for triethylammonium, 
102-061, and subtracting from this the value of ammonium, 18-074 (Stas), similarly 
found, we get for the hydrocarbon, C,H,,, the value 83°987. Now, Dumas and 
others have shown that carbon is 12-005, if oxygen =16 ; hence the number 83-987 
necessitates hydrogen being less than unity, instead of greater, as is usually acknow- 
ledged, when 0 =16 is the standard accepted. 
7. The Length of the Prismatic Spectrum as a Test of Chemical Purity. 
By Dr. J. H. Gravstonz, F.R.S. 
The specific refraction of any chemical substance is a very constant property, 
and may be used to determine the purity of anyspecimen. Landoltin fact showed 
how a mixture of two known substances, such as ethylic and methylic alcohol, in 
unknown proportions might be analysed by means of it. The object of the present 
communication is to point out that specific dispersion, i.e. the length of the spec- 
trum divided by the density, is in many cases a still more delicate test of purity. 
The absolute determination of a refractive index is subject to many sources of 
error; but the almost simultaneous measurement of the two extreme lines of a 
prismatic spectrum can be effected with great comparative accuracy, and in the 
author’s observations the error of dispersion probably rarely exceeds + 0:0002. It 
was found that different specimens of benzene, bisulphide of carbon, hydrocarbons 
from essential oils, cymene, &c., often differed from one another by five or ten times 
the above amount, while the differences of refraction scarcely exceeded those of 
probable experimental error. That impurities otherwise unsuspected may reveal 
themselves by their effect on the length of the spectrum is apparent from the fact 
that, whereas the specific refraction of organic bodies for the line-A varies only 
from ‘410 to about ‘570, the specific dispersion from A to H varies from ‘0162 to at 
least ‘0632. Itis the aromatic compounds, or still more such bodies as naphthalene, 
where the number of carbon atoms is in excess of that of the other elements, that 
