GENERAL DISCUSSION OF RESULTS. 209 



The question then arose whether these same relations would hold for a solvent 

 with a very large viscosity. 1 Fortunately there is one such solvent, glycerol, which 

 is well adapted to this work. It has very high viscosity, is an excellent solvent, and 

 has a fairly high dielectric constant, which means that it is a good dissociating agent. 

 Further, it is a fairly strongly associated solvent, which also indicates that it would 

 have considerable dissociating power. 



An examination of the literature showed that very little had been done on the 

 physical chemistry of glycerol. A few measurements of the conductivities of certain 

 salts in glycerol had been made, but had not been carried out at all systematically. 

 A similar condition was found to exist in reference to the measurements of the vis- 

 cosities of solutions in glycerol as the solvent. 



A number of details had to be carefully observed in working with such a viscous 

 solvent as glycerol. These have already been discussed at sufficient length. 



The salts studied in this first investigation in glycerol as a solvent were lithium 

 bromide, potassium iodide, and cobalt chloride. These were studied in glycerol, 

 water, methyl alcohol, ethyl alcohol, and in mixtures of glycerol with these solvents. 



The curves show that the conductivities of these salts in the mixed solvents do 

 not obey the law of averages. There is a marked sagging of the curves. There is 

 no minimum in the conductivity curves, as was found with mixtures of alcohol and 

 water. With cobalt chloride, results were obtained analogous in all essential par- 

 ticulars to those found for lithium bromide. Cobalt chloride has a greater conduc- 

 tivity than lithium bromide in glycerol. This is what would be expected, since the 

 former is a ternary and the second a binary electrolyte. The conductivity of cobalt 

 chloride in ethyl alcohol is, however, apparently abnormally low. This is due to 

 polymerization of the cobalt chloride by the alcohol, as was shown by the molecular 

 weight determination by the boiling-point method. The results with potassium 

 iodide arc similar to those obtained with lithium bromide. 



Our work was carried out at different temperatures, so that the temperature coef- 

 ficients of conductivity could be calculated. The most striking feature of the 

 conductivities of salts in glycerol as a solvent is the enormous magnitude of the tem- 

 perature coefficients. This amounts to more than 10 per cent per degree between 

 25 and 35, and to more than 8.5 per cent between 25 and 45. This is by far 

 the largest temperature coefficient of conductivity that has ever been observed in 

 any solvent. We shall see that it is closely related to the temperature coefficients 

 of fluidity in this same solvent. The enormous increase in conductivity with rise in 

 temperature is, therefore, due largely to the rapid decrease in the viscosity of glycerol 

 as the temperature is raised. 



The temperature coefficients of conductivity in the mixtures of glycerol with the 

 other solvents, like the conductivities themselves, do not in any case obey the law 

 of averages. They are always less than the average. 



Glycerol is therefore a solvent which, with other solvents, gives a mixture whose 

 properties are not additive, and in this respect glycerol resembles water. It has- 

 been shown by earlier work' that when water and the alcohols are mixed each dimin- 

 ishes the association of the other. A similar result manifests itself when glycerol 

 is mixed with water or with the alcohols. Glycerol has the power of diminishing 



'Carnegie Institution of Washington Publication Xo. 80. 



