Alumina with Silica, Lime and Magnesia. 299 



This arbitrary calibration of the instrument was made with 

 the help of certain fixed points. The melting temperatures 

 of diopside and anorthite are now definitely known, and we 

 are sure that the melting point of platinum is constant, though 

 there is some doubt at to its exact temperature in degrees. 

 We have assumed it to be 1750°. Diopside melts at 1395°, 

 according to the values of Allen and White,* provisionally 

 corrected by recent unpublished comparisons with the gas 

 thermometer by Day and Sosman. On the same curve 

 extrapolated, anorthite melts at 1542°. By calibrating the 

 instrument in terms of these fixed points, the relative tempera- 

 tures of our scale are definitely established. The absolute 

 temperatures may shift with more accurate determinations of 

 these fixed points. 



The accuracy of the individual observations is determined 

 by (1) the sharpness of the flash of the tack, and (2) by the 

 delicacy with which the lamp filament can be matched up 

 against the tack surface. In this work we found the points 

 at which the filament was distinctly brighter or distinctly 

 darker than the tack surface to lie 8 millivolts apart. The 

 actual adjustment was ±2 millivolts, so that the maximum 

 error of observation was ±20° and the probable error rbl0° 

 or less. The actual variations in the melting point observed 

 for platinum were ±2 millivolts or 10°. At this temperature, 

 therefore, the variation due to the material melted falls 

 within the limits of error of the pyrometric system. The 

 same is true of the anorthite and sillimanite. We feel 

 justified in saying that the method is reliable, though rough 

 and of limited usefulness compared with the thermoelement 

 below 1600°. Several improvements have been suggested by 

 this work and are now being tested. 



Preliminary studies for locating the approximate position 

 of maxima and minima were made in a 4:0 mm iridium tube 

 furnace (Nernst type) by placing small amounts of the finely 

 powdered and thoroughly combined charge on an iridium tray 

 and exposing for a definite period at constant temperature. 

 This temperature is then increased by degrees until the 

 minima appear, after which the temperature is stepped up 

 until the maximum is located. With sharply melting mix- 

 tures, and where there is a reasonably large difference of 

 temperature betvveen the maxima and minima, the compositions 

 can be quite definitely determined. If, however, the eutectic 

 is viscous, like that between Al 2 Si0 5 and Si0 2 , the minima 

 can be located only approximately by this method. Similarly, 

 where the compositions of the maximum and minimum are 

 *This Journal (4), xxvii, 1, 1909. 



