It is to be observed that Berzelius in this early table made use of Avogadro’s principle in connection with elements forming gaseous compounds, and thus obtained correct formulas and atomic weights in such cases, but that in many instances his atomic weights and those now accepted bear the relation of simple multiples to one another, because he had then no means of deciding upon the formulas of many compounds except the rule of assumed simplicity. For example, the two oxides of iron now considered to be FeO and Fe2O3 he regarded as FeO2 and FeO3, knowing as he did that the ratio of oxygen in them was 2 to 3, and believing that a single atom of iron in each was the simplest view of the case, so that as the consequence of these formulas the atomic weight of iron was then considered to be practically twice as great in its relation to oxygen as at present.
These old atomic weights of Berzelius, used with the corresponding formulas, were just as serviceable for calculating compositions and analytical factors as though the correct multiples had been selected. As time went on, the true multiples were gradually found from considerations of atomic heats, isomorphism, vapor densities, the periodic law, and so on, and suitable changes were made in the chemical formulas.
Berzelius used 100 parts of oxygen as the basis of his atomic weights, a practice which was generally followed for several decades. Dalton, however, had originally used hydrogen as unity as the basis, and this plan finally came into use everywhere, as it seemed to be more logical and convenient, because hydrogen has the smallest atomic weight, and also because the atomic weights of a number of common elements appeared to be exact multiples of that of hydrogen, thus giving simpler numbers for use in calculations.
Within a few years a slight change has been made by the adoption of oxygen as exactly 16 as the basis, which gives hydrogen the value of 1·008.
As early as 1815, Prout, an English physician, had advanced the view that hydrogen is the primordial substance of all the elements, and consequently that the atomic weights are all exact multiples of that of hydrogen. This hypothesis has been one of the incentives to investigations upon atomic weights, for it has been found that these constants in the cases of a considerable number of the elements are very close to whole numbers when based upon hydrogen as unity, or even still closer when based upon oxygen as 16.
With our present knowledge Prout’s hypothesis may be regarded as disproved for nearly all the elements whose atomic weights have been accurately determined, but the close or even exact agreement with it in a few cases is still worthy of consideration. There is an interesting letter from Berzelius to B. Silliman, Jr., in the Journal (48, 369, 1845) in which Berzelius considers the theory entirely disproved.
For a long time entire reliance was placed upon the atomic weights obtained by Berzelius, but it came to be observed that the calculation of carbon from carbon dioxide appeared to give high results in certain cases, so that doubt arose as to the accuracy of Berzelius’s work. Consequently in 1840 Dumas, assisted by his pupil Stas, made a new determination of the atomic weight of carbon, and found that the number obtained by Berzelius, 12·12, was slightly too large. Subsequently Dumas determined more than twenty other atomic weights, but this great amount of work did not bring about any considerable improvement, for it appears that Dumas did not greatly excel Berzelius in accuracy, and that the latter had made one of his most noticeable errors in connection with carbon.
Soon after assisting Dumas in the work upon carbon, Stas began his very extensive and accurate, independent determinations, leading to the publication of a book in 1867 describing his work. Stas made many improvements in methods by the use of great care in purifying the substances employed, and especially by using large quantities of material in his determinations, thus diminishing the proportional errors in weighing. His results, which dealt with most of the common elements, were accepted with much confidence by chemists everywhere.
Stas reached the conclusion that there could be no real foundation for Prout’s hypothesis, since so many of his atomic weights varied from whole numbers, and this opinion has been generally accepted.
The first accurate atomic weight determination published in the Journal was that by Mallett on lithium (22, 349, 1856; 28, 349, 1859), showing a result almost identical with that accepted at the present time. Johnson and Allen’s determination (35, 94, 1863) on the rare element cæsium was carried out with extraordinary accuracy. Lee, working with Wolcott Gibbs, made good determinations on nickel and cobalt (2, 44, 1871). The work of Cooke on antimony (15, 41, 107, 1878) was excellent.