486 White — Melting Point Methods at High Temperatures. 



actually occurs. This error is diminished, as shown in the 

 preceding article,* by reducing the diameter of the charge.f 



All these effects conspire to make " radiation " calorimetry 

 somewhat analogous to some of the rapid volumetric methods 

 of analytical chemistry. Under strictly constant conditions 

 results "can be reproduced with great uniformity, but their 

 actual accuracy is dependent, first, on a calibration by some 

 absolute method, and then on the subsequent absence of varia- 

 tion in numerous details, the importance of some of which may 

 easily be overlooked. The absolute method thus far available 

 in calorimetry at high temperatures is the method of mixtures, 

 where the body under investigation is dropped from the 

 furnace into a calorimeter 4 Where that can be used, the 

 radiation method will probably depend on rapidity or con- 

 venience for its usefulness, but hardly seems likely to furnish 

 the final, most accurate values. 



With most silicates, however, the dropping method cannot 

 be used, since their inversions and changes of state will not 

 occur normally with a rapidly falling temperature, and often 

 not even with a very slow fall. This makes the radiation 

 method more necessary, but at the same time deprives it of 

 independent calibration in the very region where that is most 

 needed, i.e., at the high temperatures where most silicates 

 melt. The rapid increase of X and consequent tendency 

 toward smaller values of G- with rising temperature greatly 

 increase the disturbing effect of small variations in tempera- 

 ture distribution. A calorimetric problem of some difficulty is 

 thus presented, whose consideration does not belong here. It 

 has only seemed desirable to call attention to the present 

 limitations, as well as the possibilities, of such quasi-calori- 

 metric determinations as can be made with simple melting- 

 point apparatus. 



As an illustration of these : numerous determinations, with 

 2*5 gram charges, of the latent heat of various silicates melting 

 above 1300° agreed to 3 per cent, but the systematic error may 

 be 15 per cent, so that it has not seemed desirable to publish 

 the results.§ 



*• Pages 462, 466. The difference between center and outside is not only 

 smaller, but relatively more constant (p. 464) in the small charge. Its effect 

 almost disappears where G is kept constant. 



f Still greater accuracy has been obtained by making the crucible wall 

 itself the junction, and then measuring the temperature difference differen- 

 tially. Results are then consistent at least to 0'5 microvolt, or about '03°, 

 above 1400°. 



\ Plato, loc. cit. Goodwin and Kalmus, Phys. Rev., xxviii, 1, 1909. 

 J. A. Harker, Phil. Mag., x, 430, 1905, 1906. W. P. "White, this Journal, 

 xxviii, 334, 1909. A full set of references in this subject might be hard to 

 make, and does not seem needed here. 



§ If, however, the order of magnitude is of interest, it may be taken as 

 106 ± 15 calories for diopside, about 350 times (numerically) the specific heat 

 of the solid just before melting. 



