Dalton proposed to adopt as the unit of atomic weight the weight of the lightest atom, namely, that of hydrogen. Taking, for example, water as one substance containing hydrogen, its percentage composition by weight is—

If the smallest portion of water capable of free existence contains one atom of hydrogen and one of oxygen, then placing the weight of an atom of hydrogen as unity, the weight of an atom of oxygen is eight times as great. And although we do not know the absolute weight of any single atom, we are justified in supposing that an atom of oxygen is eight times as heavy as an atom of hydrogen. But have we any right to make the assumption that a molecule of water contains one atom of each element? Dalton came to the conclusion that this supposition was a justifiable one; but there are strong reasons against it. We have already seen that Cavendish discovered approximately, and that Gay-Lussac and Humboldt determined accurately, that when hydrogen and oxygen unite to form water, two volumes of the former combine with one of the latter. Now it appears improbable on the face of it that any given volume of hydrogen should contain only half as many particles as an equal volume of oxygen; and it is still more improbable, when we take into consideration (1) Boyle’s discovery that if the pressure on a gas be increased, the volume of the gas, whatever it may be, diminishes in like proportion; and (2) Gay-Lussac’s and Dalton’s discovery, that all gases, when equally raised in temperature, expand equally. It would be very remarkable if one gas, containing twice as many particles in unit volume as another, should show exactly similar behaviour towards pressure and temperature. Hence it appeared not unreasonable to suppose that the composition of water was expressed by one particle of oxygen in union with two particles of hydrogen. (The word “particle” is here used in the meaning of “small portion”; such particles may be molecules or they may be atoms.)

When steam is formed by the union of hydrogen with oxygen, it has a volume equal not to the sum of the volumes of the hydrogen and the oxygen, but to two-thirds of the sum, or equal to that of the hydrogen alone, or twice that of the oxygen. And as steam, like hydrogen and oxygen, follows Boyle’s and Gay-Lussac’s laws, it must be supposed that in the steam there are as many particles as in the hydrogen from which it was formed. But the particles of steam must necessarily be more complex than those of the hydrogen, inasmuch as the steam contains oxygen as well as hydrogen.

These difficulties may, however, be easily overcome by the following supposition, which was first formulated by Avogadro in 1811. The ordinary particles of hydrogen and of oxygen are complex, each containing at least two atoms, or smaller particles, which usually exist in combination with each other, or with atoms of some other element. Two volumes of hydrogen, therefore, contain twice as many particles as one volume of oxygen; to such particles the name “molecules” is now universally applied. And as these molecules are themselves each made up of two smaller particles, now termed “atoms,” there exist in two volumes of hydrogen twice as many atoms as in one volume of oxygen. On combination, the atoms in the molecules of hydrogen and oxygen rearrange themselves, so that two atoms of hydrogen and one atom of oxygen combine to form a molecule of water-vapour, containing three atoms. The steam now contains as many molecules as did the hydrogen before combination; but whereas the molecules of hydrogen originally consisted of two atoms each, the molecules of steam contain three atoms. It is this which causes the contraction from three volumes to two when hydrogen and oxygen molecules exchange partners in forming water molecules.

Of course the difficulty would meet with an equally good explanation if it were supposed that the hydrogen molecules and the oxygen molecules each contained four atoms, or eight atoms; but there is no need to increase the complexity of the molecule, and the assumption that these molecules are “diatomic” completely serves the purpose. The composition of water is therefore believed to be two atoms of hydrogen in combination with one atom of oxygen; and when hydrogen and oxygen unite to form water, a transaction similar to an exchange of partners is supposed to occur; the atoms of hydrogen and oxygen are imagined to leave their partners of like kind, and to rearrange themselves so that groups of atoms, or molecules, each containing two atoms of hydrogen and one of oxygen, are formed. To such an arrangement the formula H2O is applied, while ordinary hydrogen molecules may be represented as H2, and molecules of oxygen as O2.

It has been shown already ([p. 150]) how Lord Rayleigh obtained the number 15·882 for the density of oxygen compared with that of hydrogen. To determine the atomic weights of elements, the usual process has been to analyse their oxides, for only a few elements form compounds with hydrogen. Thus the analysis of copper oxide yields the numbers—

And as no compound of copper and hydrogen is known which lends itself to analysis, the atomic weight of copper is necessarily referred to that of oxygen. If the atomic weight of hydrogen be taken as unity, that of oxygen, from Lord Rayleigh’s determination, must be 15·882, because, in comparing the weights of equal volumes of the gases, a comparison is made of the weights of equal numbers of molecules; and as it is reasonable to suppose that each molecule of hydrogen and of oxygen contains two atoms, the number 15·882 represents the weight of an atom of oxygen compared with that of an atom of hydrogen taken as 1. But this number has not been regarded as sufficiently established by experiment. Other observers (for the importance of this ratio has been acknowledged since the beginning of the century) have obtained results differing from that given above, although not to any great extent. And as it is a matter of indifference what basis or standard be taken for atomic weights, which represent only relative numbers, it is common to accept the atomic weight of oxygen as 16, in which case that of hydrogen, if Lord Rayleigh’s determination of its density be regarded as accurate, would be 1·0074. Hence if we place the atomic weight of oxygen as 16, that of copper would be 63·34. And as with copper, so with most other elements. It is very seldom that the atomic weight of an element has been directly compared with that of hydrogen; it is, in fact, almost always ascertained by analysis of its chloride, bromide, or oxide; and the atomic weights of chlorine and bromine have been very carefully compared with that of oxygen. There is, besides, another convenience in accepting 16 as the atomic weight of oxygen: it is that many atomic weights are then represented by whole numbers instead of by fractions; thus, sulphur has the atomic weight 32, if oxygen be made 16, whereas, if it were 15·882, the atomic weight of sulphur would be 31·764, a number much more difficult to remember.

We see then that it is convenient to refer the density of argon to oxygen taken as 16. The density obtained by Professor Ramsay in February 1895, using a globe of small capacity (only 160 cubic centimetres), was 19·94; exactly the same result was given by Lord Rayleigh’s experiments in June 1895 on argon prepared by means of the electric discharge, with a balloon of much greater capacity, which held over two litres of gas. Now as a molecule of oxygen consists of two atoms, the weight of a molecule is twice the atomic weight, or 32; and as a given volume of argon must contain as many molecules as the same volume of oxygen, the weight of a molecule of argon must be twice 19·94, or 39·88.