4:66 W. P. White — Melting Point Determination. 



For naphthaline, the time lag was about three times that for 

 diopside. For metals, the conductivity is so great that distor- 

 tion of the kind just considered is probably always negligible 

 in practice. 



i. The above seems to justify the following practical con- 

 clusions : (1) For determinations of single points, the essen- 

 tial thing is to reduce the interval of accelerated heat supply. 

 This calls for small crucibles or slow rates. And when they 

 are used, the difficulties arising from the regular temperature 

 gradient across the charge can be rendered negligible compared 

 to other sources of error. The requirements of a given case 

 depend on the accuracy desired and on the purity of the sub- 

 stance, and can be determined well enough for all practical 

 purposes from the length of the melting interval in conjunc- 

 tion with the approximation just given for the relation of time 

 lag to radius in different substances. (2) For determining the 

 form of the whole melting curve, the time lag must be kept 

 small, and this again calls for the narrow crucible but not the 

 slow rate, for a reason given in the following paper (p. 485). 

 (3) Whether rate or radius, or both, shall be diminished will 

 depend mainly upon the apparatus and conditions at the dis- 

 posal of the investigator. The whole time required to com- 

 plete the melting varies nearly as V, the rate of heating. The 

 distortion varies as R 2 Y, but Y varies as G/R, where G is the 

 external gradient, hence the distortion varies as GR also. 

 And, therefore, for the same distortion by diminishing R, Y 

 and G are both increased. So if R is made, say, one-third, the 

 run may be finished in one-ninth the time, and the effect of 

 variation or uncertainty in the furnace temperature also 

 diminished three times, with little increase in distortion. 



Y. Irregular variations in heat distribution. — The preced- 

 ing section has dealt with an ideal case. It assumes a charge in 

 which the heat travels uniformly toward an infinitesimal ther- 

 moelement situated exactly at the center.. Of course these 

 conditions are never completely realized. 



Sources of irregularity may lie in the furnace, in the charge 

 itself, or in imperfect centering of the thermoelement. Aside 

 from these, there is nearly always a tendency for melting to start 

 at one end of the charge, either top or bottom. In any case, the 

 difficulty comes mainly from the fact that the melted area 

 reaches some parts of the thermoelement before others, so that 

 the temperature of the junction depends on different por- 

 tions of the charge, some of which are melted while others are 

 not. This converts the ideally sharp break (X, fig. 3) into a 

 gradual rise, of uncertain interpretation. To avoid this diffi- 

 culty, two things are necessary. The thermoelement itself should 

 conduct heat as little as possible, and the melting should 



