W. P. White — Melting Point Determination. 471 



size of thermoelement wire did not affect it perceptibly. The 

 freezing point here lay below the melting point as determined 

 from the curve. Hence in this case, the slope of the npper 

 end of the curve is due to the temperature measurement 

 and causes the melting point to come a little high. Presum- 

 ably the same thing is true of the silicates giving curves like 

 A of fig. 4. In fact, the results obtained by adding different 

 amounts of impurity to one of these silicates are much more 

 consistent with each other and with equation (1) if the melt- 

 ing points are taken a degree or so lower than indicated by 

 the curves. 



Sodium sulphatef (melting point 885°±1°)* was then tried. 

 The resulting curve is C of fig. 4. The whole melting inter- 

 val was '1° in one charge, *3° in another, which probably 

 became slightly contaminated in handling. For a temperature 

 so high, this is a remarkably sharp curve. Taken with the pre- 

 ceding it indicates that the false rise at the end of the melting 

 may be avoided, and that a clue to its cause may be found in 

 the physical differences between sodium sulphate and sodium 

 chloride at their melting points. A further investigation of 

 the problem seems to require apparatus not now at hand. 

 Meanwhile, the systematic error for the melting point deter- 

 mination in a limpid silicate has been reduced to a degree or 

 so, and the accidental variation to half a degree, and this 

 accuracy is obtained with exceedingly simple and convenient 

 apparatus. The error in the true melting point of these 

 silicates due to impurity is (in our case) quite as great as (and 

 opposite to) the experimental error. An attempt to realize 

 greater accuracy, probably attainable only through more elab- 

 orate experimental arrangements, has not seemed necessary at 

 present.;}; 



The addition of impurity to the sodium sulphate of fig. 4, 

 G, gave the typical oblique curve of fig. 3. 



*This is a new determination. 



fFrom J. T. Baker Chemical Company, who gave the following analysis: 

 Ca, "005 per cent ; Mg, none ; Fe, '0006 per cent ; CI, trace. 



Jin the work on Diopside, etc. (this Journal, xxvii, 8, 1909), the melting 

 points of the eutectic in different mixtures of pseudowollastonite and diop- 

 side are not quite the same, as by theory they should be, but rise with the 

 percentage of eutectic present, varying about 6°. The variation, though far 

 within the limits of error of most previous determinations at this tempera- 

 ture, is consistently greater than the accidental variations of our results. 

 Two causes for it are indicated by the present article : (1) Where the per- 

 centage of eutectic is larger and the time of melting, therefore, longer, the 

 false rise at the end of the melting curve is greater. These results are there- 

 fore a little too high. (2) Where the amount of eutectic is small the total 

 impurity present bears to it a larger ratio, and there is thus a tendency for 

 the melting point lowering to be increased — these results are thus a little 

 too low. It is probable that both causes are operative. 



