98 REPORTS ON INVESTIGATIONS AND PROJECTS. 



in the charge are inferred from the changes in its temperature. The tem- 

 perature of the charge thus depends upon two things: first, the furnace 

 temperature, and second, the heat efTects in the charge. The manipulation 

 consists (after the furnace is once loaded) in observing (usually once a 

 minute) the temperature rise, regulating the furnace current accordingly, 

 and observing, either directly or indirectly, the temperature difference of 

 furnace and charge. 



The interpretation of the results may, in theory, be made very simple. 



The temperature gradient, G, between charge and furnace, serves as the 

 measure of heat-flow to the charge. The flow actually is proportional to G, 

 and to the heat-transmittance, F, of the space between furnace and charge. 

 The exact values of both G and F are usually unknown, and, as already 

 stated, may vary with time, temperature, rate of heating, etc., so that accu- 

 rate heat determinations by means of them, though possible, are more diffi- 

 cult than has sometimes been realized. But in the detection of a small heat 

 effect we have merely to determine the change in G caused by the addition of 

 the effect to the heat which is required to change the temperature of the 

 charge. For instance, a silicate charge (2 grams) of specific heat 0.3, heated 

 10° per minute, lagged 3° behind the furnace (G' = 3°). Hence 3° in G 

 corresponds to 3 calories per minute. If an inversion absorbing 3 calories 

 and extending over 100° should occur, G would be increased o.i, or 0.3° for 

 10 minutes; if the inversion should take place in one minute, G would be 

 doubled, or increased 3°, etc. The detection of small heat-effects is easier: 

 (i) The larger G is, per calorie per minute, (2) the freer G is from other 

 variations, (3) the quicker the inversion occurs; it is only sluggish inversions 

 whose detection gives any trouble. 



G increases with the furnace-rate ; hence a rapid rate is of the first im- 

 portance. G also increases with the diameter of the charge, but can be made 

 steadier in the case of a small charge, and the advantage of the small charge 

 appears to be greater on the whole. Fluctuations in the furnace rate cause 

 variations in G; these are partly eliminated by measuring G, not between 

 charge and furnace, but between the charge and another body ("neutral 

 body" or "dead body") closely resembling it. For the best results the other 

 body is used, and the furnace is regulated as carefully as possible, besides. 

 Runs are also made with both bodies alike, to eliminate effects due to the 

 thermo-elements or to the lack of symmetry between the two bodies, and, of 

 course, repetitions eliminate accidental irregularities. The apparatus now 

 used is small, two platinum crucibles holding i c.c. each, 3 mm. apart, and 

 surrounded by a wider porcelain tube to increase uniformity of temperature. 

 A differential thermo-element is also used, which gives directly the tempera- 

 ture difference of the two bodies at any instant. This not only increases 

 accuracy, but saves much tedious computation, necessary when the tempera- 

 tures are read separately. A complete platinum inclosure shields the whole 

 system from leakage currents out of the furnace-coil. In one set of deter- 

 minations made on different days with the same set-up, conditions were 

 reproduced over a 300° interval with a maximum variation of from 0.03° to 

 0.06° in the different determinations. This was with a silicate charge and 

 indicates that i calorie distributed over 100° could often be detected. But 

 this would be a more difficult case than has yet been found in practice. In 

 one case it was observed that the first heating in each day, which, of course, 

 occurred immediately after the furnace had been cold, gave results differing 

 by 0.3° from later heatings, although the furnace was cooled 300° between 

 all the heatings. This condition reproduced itself at 0.1° on successive days 



