20 Penfield — Interpretation of Mineral Analyses. 



the specific gravity is lowered to a trifling extent, it gives one 

 not only great confidence in the purity of the material, but, 

 also, it enables the investigator to present data which others 

 can make use of in judging the character of the work. 



It has been the author's privilege during the past twenty- 

 five years to make many analyses of minerals, and to superin- 

 tend the making of many more in the Sheffield Mineralogical 

 Laboratory ; also to discuss the analytical results and derive 

 therefrom the chemical formulas of minerals, and this occasion 

 will be taken to call attention to certain features which are 

 regarded as most important in mineralogical investigations. 

 In the first place, the utmost pains should be taken to secure 

 pure materia], and, if the results are to be published, the char- 

 acter of the material should be described so that its degree of 

 purity can be judged by others. Secondly, if an analysis pre- 

 sents any especially difficult features, the method of analysis 

 should be carefully described, and it is in almost all cases well 

 to give at least some brief outline of the analytical methods 

 employed. Then, too, when material is abundant, it is advis- 

 able to make analyses in duplicate, and to give all of the deter- 

 minations, together with the averages. Thus the investigator 

 has from beginning to end the satisfaction of a control over all 

 determinations, and. if agreements are close, others can form 

 some estimate concerning the care with which the work was 

 executed. There are those who apparently entertain the belief 

 that closely agreeing duplicate determinations indicate great 

 accuracy in analytical work, but that is not necessarily the 

 case, for in some analytical methods there is a tendency for 

 results to come too high, in others too low, and thus duplicate 

 determinations, made under like conditions, either with faulty 

 methods, or with good methods improperly executed, may be 

 uniformly high or uniformly low, agreeing with one another, 

 and yet varying considerably from the truth. Still two closely 

 agreeing determinations carry with them a certain weight 

 which cannot be ignored. Thirdly, with each analysis, the 

 quotients obtained by dividing the several constituents by their 

 molecular or atomic weights, as the case demands, should be 

 given, and from the quotients thus obtained the ratio between 

 the several constituents should be determined. The ratio 

 ought not to be given simply rounded out to the nearest whole 

 numbers, but, taking the quotieat of the most characteristic or 

 best determined constituent as unity, the ratio should be given 

 to the second place of decimals. It is safe to assume that the 

 close approximation of a ratio to whole numbers constitutes 

 the strongest argument that can be advanced in support of the 

 excellence of an analysis and the correctness of the derived 

 formula. It will seldom happen that a ratio approximates to 

 whole numbers merely as a matter of accident. Provided the 



