Azobenzene is also extremely abnormal; it gives values for the molecular weight somewhat smaller than those with benzil and phenanthrene at the same molecular concentration, and the rate of increase is at first slightly less, then slightly greater, than linear. The last four determinations (63.2 to 85.9 per cent.) are exactly linear, and when produced to zero concentration, give the normal molecular weight with considerable accuracy.

Hexadecyl alcohol is about twice as abnormal as in benzene.

The view is very generally held that organic solutes should show less tendency to association in alcohol than in benzene, since the conductivity of electrolytes in benzene is extremely small, whilst in alcohol their conductivity is considerable, and it has been assumed that a substance with dissociating power for electrolytes would not favour the formation of complex molecules. If we admit this view, it might be argued that the abnormalities in alcoholic solution are not due to association. The fact that three substances of such different constitution as azobenzene, phenanthrene, and benzil should all be highly abnormal and to almost the same extent over a considerable range of concentration, whilst in benzene, in concentrated solution, they behave so differently, also suggests that the abnormality is not due to the solute, per se, but to some specific action of the solvent. It therefore seems advisable to consider the possible ways in which the solvent could exert an influence on the apparent molecular weight.

That associated liquids, such as alcohol, have an abnormally low vapour pressure has been well established (Vernon, Chem. News, 1891, 64, 54; Young, Phil. Mag., 1892, [v], 34, 507; T., 1893, 63, 1251; van't Hoff, "Lectures," Part 3, 50). It is also known that, in solution, alcohol is non-associated at sufficient dilution, whether in an inert or in an associating solvent.

If, therefore, we add to alcohol a miscible, inert, non-volatile liquid, the vapour pressure of the alcohol will not be that corresponding with its molar fraction, but will be larger by an amount which is a function of the dissociation of the complex molecules of the alcohol. An abnormally small lowering of the vapour pressure will therefore be obtained. The effect of the smaller lowering of the vapour pressure will be to increase p' and decrease p-p' in gMp' and both these influences will increase the apparent G(p-p')' molecular weight. With such an inert solute, in a solvent the liquid molecules of which are associated, but not the vapour molecules, we should expect to get values for the molecular weight of the solute, which point to the normal molecular weight at extreme dilution, and increase with concentration at a rate which is a

m =

function of the dissociation of the complex molecules of the solvent. Hexadecyl alcohol might be expected to be such an inert solute with respect to alcohol, since it appears to be a general rule that substances of constitution similar to that of the solvent give normal molecular weights in that solvent. The departure of hexadecyl alcohol from Raoult's law may therefore be attributed to the progressive dissociation of the alcohol as it is diluted with the hexadecyl alcohol, and the much greater abnormality of the other substances must be sought in some other influence.

When there is combination between the solute and solvent, it is evident that part of the solvent is removed from the solution and no longer functions as solvent. G is therefore decreased, and at the same time p' will be decreased owing to the increased concentration of the solute complex, and p-p' increased. The decrease of G and p' and the increase of p-p' would all act in the direction of decreasing the molecular weight. It is further evident that the influence of such combination will become very great as the concentration approaches n=N, provided the compound or complex formed does not exert a decided vapour pressure.

After the experiments described in this paper had been carried out, the apparatus was used by Mr. B. C. Burt, who determined the vapour pressure of concentrated sulphuric acid solutions (T., 1904, 85, 1345). His results, when suitably plotted (Fig. 7), show in a most striking manner the influence of combination between solute and solvent in molecular weight determinations in concentrated solutions.

Burt determined the vapour pressure of solutions ranging from 25 to 95 per cent., and calculated the molecular weights by Raoult's formula, using Regnault's values for the vapour pressure of water.

The experimental figures at 100° were used when available; at the higher concentrations, it was necessary to take results at higher temperatures. As the influence of temperature at these high concentrations is very small, the form of the curve cannot be seriously affected. The curve of molecular weight against percentage concentration shows the influence of n = N most clearly. There is first a rapid, then a slower fall until n = N is reached, after which the molecular weight is practically constant and almost zero.

Since, in a solution in which the solute is practically non-volatile and the solvent readily volatile, any chemical compound formed must have a much smaller vapour pressure than that of the solvent, it is evident that, in such solution, combination between solvent and solute cannot occur if the normal or a high molecular weight is indicated at or near equimolecular concentration.

Since neither the dissociation of the solvent nor combination

between solute and solvent will explain the great abnormality of phenanthrene, benzophenone, and azobenzene in alcohol, it can only be supposed that the abnormality is due to association, and it follows that alcohols and the associating solvents" exert an associating influence on each other.

Alcohol was first used as an ebullioscopic solvent by Beckmann (Zeitsch. physikal. Chem., 1890, 6, 454), and from his results in this and other solvents, he drew the conclusion that solvents like the alcohols, organic acids, and phenols, containing hydroxyl groups, are dissociating solvents of the water type,' " and consequently complex molecules are not formed in them. Nernst lent support to Beckmann's hypothesis, by the theory that the higher

FIG. 7.

Apparent molecular weight of sulphuric acid (1) against molecular
concentration, (2) against percentage concentration.

the dielectric constant of a solvent, the greater will be its power to split up double molecules.

It is of interest to examine the experimental results on which Beckmann founded his theory. These are shown graphically in Fig. 8 for alcoholic solutions, molecular weights being plotted against concentration. Carbamide and racemic acid have not been included, the former because it was only examined in dilute solution, the latter because the values are irregular.

A glance at the curves will show that, with the exception of the acids, all the substances show a much more rapid increase of molecular weight than is shown by "normal" substances in benzene or naphthalene. Benzil, phenyl benzoate, ethyl benzoate,

and naphthalene, substances belonging to the "associating type of solvent, are very abnormal; borneol, acetophenoneoxime, and acetanilide, belonging to the "water type," are less abnormal, and the curves of the last three substances are approximately parallel. If the view already brought forward, that the abnormality of hydroxy-compounds in alcohol is due to the dissociation of the

alcohol complex, is correct, these curves should be parallel for equal molecular concentrations.

Succinic and tartaric acids show a decreasing molecular weight with concentration; evidently this is due to ester formation and combination of the water formed with the solvent. The normal molecular weights with salicylic acid must be attributed to a balance between the abnormality due to the dissociation of the solvent and that due to ester formation.

VOL. CXIII.

S

Beckmann's results therefore lend strong support to the deduction that there is mutual association between alcohols and substances of the hydrocarbon type.

Beckmann's classification into two types is still useful, however, and it is proposed to call the "associating solvent" type of substance Carbotype, and the "water type" of substance Oxytype, without, however, assuming any sharp line of demarcation between the two classes.

Numerous instances have been published of the comparatively normal behaviour of hydroxylated compounds and anilides in ethyl and methyl alcohols, but very few instances of association in alcohol, apart from those in Beckmann's paper, appear to have been observed. The following have been found (Peddle and Turner, T., 1911, 109, 696):

m-Dinitrobenzene, 2.6-5.5, 1/100 gram-mols, per 100 c.c., m'=181-211, m=168.

Diphenylamine, 3.79-5.99, 1/100 gram-mols. per 100 c.c.,. m'=190-206, m=169.

Phenanthrene, 1.94-11.99 grams per 100 grams of solvent, m'=183-226 (Behrend, Zeitsch. physikal. Chem., 1892, 10, 279). Numerous salts, inorganic and organic, have been examined ebullioscopically in alcohol, and many of these show association as well as the influence of combination with the solvent (Turner and Pollard, T., 1914, 105, 1751).

The close similarity in behaviour between the alcohols and phenols in solvents of the hydrocarbon type suggests that phenols and carbotype substances should also show mutual association. Robertson (T., 1906, 89, 567) has shown that the hydrocarbons are highly associated in phenol solution, also that carbon tetrachloride is as abnormal as the hydrocarbons. Ethylene dibromide and bromoform are also abnormal, but to a less extent. The mutual association of the abnormal phenols and substances of the carbotype group must therefore be considered to be a general phenomenon.

The degree of abnormality of the substituted phenols follows definite laws as regards the nature and position of the substituting group, and the anilides follow similar laws. There seems little doubt, therefore, that a carbotype solute will give abnormal molecular weights in an anilide, and that mutual association takes place in this case also. Anilides do not seem to have been used as cryoscopic solvents, although formanilide (m. p. 46°) would be a convenient substance to use.

If mutual association takes place with acids and oximes in a carbotype solvent, it is obvious that no direct evidence of this