108 FIELD CoLUMBIAN MusEUuM—GEOoLoGy, VOL. III. 
siderably exceeds sulphur. It probably occurs combined with nickel- 
iron as phosphide. Sulphur, though evident byits presence in many 
meteorites as troilite, does not appear in large amounts in the analyses, 
and does not seem to be so important or constant an ingredient as 
phosphorus. Carbon is probably more frequent in occurrence than 
analyses usually show, since of twenty-eight iron meteorites investi- 
gated by Cohen for carbon all but one showed appreciable percentages, 
ranging from .19 per cent to .012 per cent.* The silicon reported in 
the analyses is doubtless in some cases to be referred to silicate grains, 
but in other cases may be free or combined with the iron as a silicide. 
The analyses make plain the incompleteness of much of the work 
which has been done hitherto. There can be little doubt that com- 
plete analyses of iron meteorites should always show iron, nickel, 
cobalt, copper, and phosphorus, and: in most cases sulphur, carbon, 
and silicon. When considerable differences occur in the analyses of 
the same meteorite, as, for instance, 2 per cent of nickel reported in 
Burlington by Rockwell and nearly 9 per cent by Shepard, the difference 
is probably not to be regarded as due to the meteorite, but to the 
analyses. Ina substance made up of different alloys and accessory 
minerals as are the iron meteorites, especially the octahedrites, there 
can be no question that portions from different parts of the meteorite 
would of necessity show unlike composition. How wide these varia- 
tions might legitimately be it is difficult to say, but some causes of 
error may be suggested. One of these is imperfect sampling. The 
proper method to secure material for mass analyses of an iron meteor- 
ite, especially if of octahedral structure, is to use dust obtained by 
boring. A mixture of the constituents of the meteorite is thus obtained 
which insures a better representation of its composition than is possible 
when only a fragment broken from some part of the surface is used. 
Such a fragment may contain an excess of taenite, or be largely com- 
posed of some accessory mineral so as to be far from representing the 
true constitution of the meteorite. Yet the larger number of analyses 
of iron meteorites have probably been made with fragments of this char- 
acter, and the wonder is, not that they show so much variation, but that 
they do not show more. Meteorites also doubtless vary in their homo- 
geneity, as shown especially by Canyon Diablo, in one portion of which 
Moissan found 2.89 per cent of nickel, and in another, only one centi- 
meter distant, 5.06 per cent. In another piece of Canyon Diablo 
two analyses made by the same analyst of material obtained at dis- 
tances of one centimeter showed 1.17 per cent and 7.11 per cent of 
* Meteoritenkunde, Heft II., p. 243. 
