only on the nature and on the actual valency of that atom, but also on the nature of the other atom, or atoms, between which and the given atom there is direct action and reaction. On the other hand, the work of Brühl has rendered it probable, that, in isomeric carbon compounds, the influence exerted by a polyvalent atom on the 'molecular refraction' is to a large extent dependent on the actual valency of that atom, i.e. on the number, rather than on the nature, of other atoms on which it acts directly'.

90. Much is to be expected from researches into the phenomena which occupy the border-land between chemistry and physics. If the knowledge chemists already have of the structure of molecules, meagre though that knowledge be, can be supplemented by definite dynamical conceptions, obtainable in part by the methods of thermal chemistry, then we may hope that chemistry will enter on a new stage of advance as a branch of the science of matter and motion. It seems to me that a most important step will be made if the bond theory of valency is generally abandoned; with it will go all those quasi-dynamical expressions, the offspring of loose and slipshod ways of thinking, which have gathered round that strange anomaly, a unit of affinity,' employed as a variable standard for measuring nothing.

If it be decided that by the valency of an atom is meant the maximum number of other atoms between which and the given atom there is, so far as we know, direct action and reaction in any molecule, then we are put in possession of a definite conception which may be applied to actually occurring phenomena, and the application of which will gradually lead to more precise knowledge regarding the distribution of the atomic interactions in various molecules. But at the same time that we are classifying molecules in accordance with the valencies of their constituent atoms and the distribution of the mutual actions of these atoms, in a word, in accordance with their structure, we are also becoming more impressed with the inadequacy of this classification; we see a vast field

1 See post, chap. IV. par. 139.

opening for investigation, we see that measurements of losses or gains of energy are required, and that determinations of many physical constants are called for. We begin, I think, to perceive that this knowledge, when gained, will supplement and not supplant that which is already possessed by us, and that it will do this by leading to an exact knowledge of the way in which the variations in the energies of molecules are connected with changes in the configuration and motion of the atoms which constitute these molecules.

91. Admitting that the definition of valency given by Lossen furnishes a better working hypothesis than any other, it becomes necessary to inquire whether this definition and the conceptions which arise from it can be applied to all known cases of isomerism.

In the article ISOMERISM, by Dr Armstrong, in Watts's Dictionary, supplt. III., will be found an account of most of those compounds which are regarded as presenting phenomena inexplicable in terms of the usually adopted theory of isomerism.

If the facts in this article are classified, it will, I think, be apparent (1) that structural formulæ in keeping with reactions may be assigned to some of the isomeric compounds mentioned-e. g. to the acrylic acids, to maleic and fumaric acid, and to the acids obtained by heating citric acid-provided we cease to regard the conventional method of expressing valency by one or more straight lines, as affording any quantitative measurements, even relative measurements, of atomic interactions; (2) that other cases of unexplained isomerisme.g. the dinitrochlorobenzenes, the nitrotetrabromobenzenes, and probably many of the terpenes1-are almost certainly illustrations of modifications in properties being correlated with variations in mutual actions between groups of molecules rather than between the atoms constituting each molecule; and (3) that the remaining cases are true residual phenomena,

1 That optical properties are not always dependent on the structure of the molecule is shewn by the ease with which optically active amylic alcohol and valeric acid are converted into the inactive alcohol, and acid, without change of chemical properties. See Armstrong and Groves, loc. cit. p. 449.

at present inexplicable in terms of the prevailing theory but not therefore contradictory of that theory.

The theory of valency can only be applied to the molecules of gases: if this is remembered the so-called exceptional cases of isomerism appear no longer exceptional, they are simply assertions of the fact that our theory is partial.

Among the more important apparently inexplicable phenomena of isomerism are those presented by the terpenes, hydrobenzoin, dulcitol, the lactic and the tartaric acids, and the acids obtained by Perkin from coumarin'.

16

Chemists are quite undecided as to the structural formula to be assigned to any molecule the composition of which is represented by the empirical formula C1H1; there is no need to bring within the theory hypothetical molecules which perhaps have no existence. We do not certainly know the molecular weights of hydrobenzoin and isohydrobenzoin; but assuming the generally employed formulæ to be molecular, it seems necessary to suppose that the molecule of one of these compounds contains two OH groups in direct combination with the same atom of carbon: our actual knowledge of the connections between molecular structure and stability is not advanced sufficiently for us to deny the possibility, although we may assert the improbability, of such a formula2. We know too little of the reactions of mannitol and dulcitol to enable us to decide between the possible structural formulæ of these compounds. Lactic and sarcolactic acids are extremely unstable; even in aqueous solutions they readily undergo changes; intermolecular actions are probably frequently occurring in any mass of either of these compounds. Tartaric and lævotartaric acids may have different molecular weights; the differences in physical properties which they exhibit are differences between solids, and we know almost nothing of the laws regulating molecular phenomena in liquid and solid bodies. Perkin has described two series.

1 See the article by Armstrong, loc. cit.

2 It seems however very probable that these two compounds belong to the class of 'physically isomeric' bodies; see Zincke, Annalen 198. 191.

3 C. S. Journal. Trans. for 1881. 409.

of acids derived from coumarin, having the same elementary composition but differing in physical properties; any one acid is generally changeable into its isomeride by the action of heat, the process being in many cases reversible by the same agency. But these changes are accompanied by processes of hydration and dehydration, and would appear to consist in great part of intermolecular actions, and not only in interactions of the atoms constituting the molecules. But of intermolecular actions we know as yet very little.

92. The properties of any mass of a solid or liquid compound seem to be conditioned, not only by the valencies of the atoms which constitute the molecule of that compound, by the distribution of the atomic actions within the molecule, and by the relatively large or small quantity of energy associated with this atomic configuration, but also by actions and reactions between groups of molecules, the parts of which groups hold together during more or less wide variations in the physical conditions to which the compound may be subjected1.

The hypothesis that groups of molecules, marked by fairly definite properties, may exist, and may mutually act and react, enables us to give a partial explanation of various facts concerning liquid and solid compounds which cannot, apparently, be so well explained by any other molecular hypothesis that has been suggested.

The phenomena presented by calcium carbonate are typical of those to be explained by the help of the hypothesis in question. Calcspar and arragonite are composed of calcium, carbon and oxygen united in the proportions expressed by the formula CaCO,; arragonite crystallises in rhombic forms,

1 According to Henry [Ber. 4. 604], the two chlorobromopropanes,

are identical in physical properties: if this is really so, we have an instance of identity of physical properties accompanied by a slight difference in the distribution of atomic interactions within the molecule.

calcspar in hexagonal forms; arragonite is harder, and specifically heavier than calcspar, nor is it so easily acted on by various reagents, e.g. hydrochloric or acetic acid, or ammonium chloride or nitrate; when powdered arragonite is heated to redness it is changed into calcspar, but the reverse change is not accomplished by the same agency; both modifications of calcium carbonate can be produced at low temperatures, but there is a certain temperature above which only calcspar is formed. Calcspar and arragonite thus exhibit identity of elementary composition (perhaps the same molecular weight) accompanied by differences of physical, and to some extent of chemical properties. Another typical instance is afforded by antimonious iodide, SbI,; this compound crystallises in redcoloured hexagonal forms, which when heated to 114° are suddenly changed into a mass of yellow-coloured orthorhombic crystals, the original external form of the mass being however preserved1. Several carbon compounds (apparently all belonging to the class of benzenoid compounds) exist in more than one form, each modification being characterised by a definite melting point and generally also by a special crystalline form. Thus chlorodinitrobenzene-C,H,CI(NO) [1:2:4] -is said to form monoclinic crystals which melt at 36°, and also rhombic crystals which melt at 39°. Anthracene, C1H10, crystallises in monoclinic plates melting at 213° which are easily oxidised by the action of nitric acid to anthraquinone (CHO); when a solution, in benzene, of anthracene is exposed to sunlight small prismatic crystals separate, melting at 244°, having the composition CH, but not acted on by nitric acid, and not oxidised to anthraquinone by chromic acid. A very remarkable instance of the phenomenon under consideration is presented by the derivative of diphenyl to which the formula (C,H,BrNHCOCH;), is assigned. This compound melts at 195°; if the melted substance is cooled quickly and again heated its melting point is now 99°; but if heating is continued the liquid again solidifies at 125-130°, and the solid thus obtained melts once more at 195°. Finally

1 J. P. Cooke, Proc. Amer. Acad. of Arts and Sci. [2] 5. 72.

2 See Armstrong and Groves, loc. cit. p. 199.