4: Allen, etc. — Diopside and its Relations 



readings when separately corrected then seldom varied -J- , 

 which was about thq accuracj^ with which any two working ele- 

 ments could be compared. The variations in the results, some- 

 times amounting to 2°, even with the- more sharply melting 

 substances (the end members, the eutectics, and diopside), were 

 therefore due to uncertainty in locating the melting point, and 

 not to the temperature measurements. 



In the latter part of the work, charges of only 2 J grams 

 were used, in very small crucibles of pure platinum, in which 

 the naked thermoelements could safely be immersed. Under 

 these conditions and with proper precautions, the variations 

 rarely exceeded \° . The small crucibles could also be more 

 easily moved in the hot furnace, chilled in cold water, etc., and 

 a wider range in the conditions of crystallization obtained. 

 Furthermore, the thermal lag of the diminutive charges was 

 very small, and complex thermal phenomena could be analyzed 

 or separated with much greater certainty than with larger cru- 

 cibles. The small crucibles were mainly used in investigating 

 those portions of the field in which the interpretation of the 

 phenomena had proved particularly difficult. They were used, 

 however, in a sufficient number of cases to afford a thorough 

 control of the earlier determinations made with 25 grams, and 

 showed that these were accurate within their own accidental 

 errors. 



The agreement obtained between observations of the same 

 point in both the earlier and later work is well illustrated in 

 the following table : 



Table I. 



Agreement of Observations in Earlier and Later Work. 



Temperatures in microvolts. About 13 microvolts to 1°. 



70#MgSiO 3 : 30#CaSiO 3 . 28$ MgSi0 3 ; ?2$CaSi0 3 . 



25 gr. charge. 2£ gr. charge. 



June, 1905 14,147 Oct. 1907. 13,772 



130 776 



110 777 



138 779 



148 



127 



Mean 14,133 13,776. 



Average deviation, 11 mv. 2 mv. 



Extreme " 37 mv. 7 mv. 



These particular instances were selected for illustration as 

 being, each in its class, the one showing the largest number of 

 strictly comparable observations. Each one is typical, and 





