504 Thermal Conductivity of some Solid Insulators. 



thermal conductivity also may be great. Paalliorn * made 

 measurements with several glasses, including numbers 5, 23, 

 and 27, and obtained for these three the values 2*267, 1*610, 

 and 1*409 respectively. These are much lower than Focke's 

 values, and the reason for the variation is discussed by 

 Winkelmann f , who does not, however, decide which values 

 are the more trustworthy. Focke's coefficients are employed 

 because he used the glasses of which the elasticities are also 

 known. It does not matter which values of k and E are 

 selected, the relation & = Ep is not even approximately valid. 

 The complexity of the substance cannot be adduced as the 

 explanation of the discrepancy, as a very simple glass, 

 number 20, the composition of which is 20 per cent. Si0 2 , 

 80 per cent. PbCX, shows the largest deviation from the 

 empirical relation. It will be observed that kfEp is small 

 for the heavy glasses — for example, numbers 20 and 33; but 

 it has not been found possible to postulate a value of n such 

 that k = Ep n . A survey of the other physical constants of 

 these glasses,' such as the coefficient of expansion and the 

 specific heat, has been made in the hope that some empirical 

 relation might be discovered as suggested by Prof. Thornton. 

 The employment of the volume elasticity, calculated from 

 the given values of Young's modulus and the values of 

 Poisson's ratio obtained by Straubel, makes the deviations 

 from the equation k = ~Eip even mere pronounced ; and no 

 other connexion between the constants could be observed. 

 Also, it does not seem probable that the variation of k/Fip is 

 due to any particular constituent oxide,, except that the 

 presence of lead tends to make this quotient small, numbers 

 20, 24, 26, and 33 all containing large quantities of lead and 

 being the only glasses in the list to do so. It may be that 

 P 2 5 has a similar effect, as numbers 27 and 31 both contain 

 more than 50 per cent, of this oxide, and in both cases the 

 value of the quotient is below the average. In general, 

 however, the compositions are so varied that it is not possible 

 to attribute an abnormal influence to any particular oxide. 



It may be added that, although the empirical relation 

 suggested by Prof. Thornton does not hold good, the state- 

 ment that the thermal conductivity increases with the elasticity 

 and with the density seems to be approximately correct, 

 except for the heavy glasses, and may be of use as a rouo-h 

 guide when seeking for a heat insulator or for a good con- 

 ducting o-lass. 



Manchester, 



12th June, 1920. 



* Paalhorn, Dissertation, Jena, 1894. 



+ Winkelmann, Ann. d. Phys. 1899 ? lxvii. p. 794. 



