ing them. On the other hand, I have rigidly excluded atomic weights calculated from analyses never designed so to be used. Any chemist, upon whose experiments we could rely, would proceed in a very different manner in making an atomic weight determination, from that which he would select for an ordinary analysis, and to put his credit at stake by calculating atomic weights from analyses not designed for this use is alike unfair to him and to the scientific public, which is asked to receive as an atomic weight determination what really is not such.

The purpose of this paper is distinctly not critical, and the remarks I have added to, or inserted in, the digest are simply explanatory. I have, however, frequently mentioned criticisms which have appeared in literature when they seemed pertinent.

As for the accuracy with which the digests have been made, I may state that the preponderating importance of this point has been constantly before my mind. In the effort to crowd the maximum amount of information into the fewest words, I have had occasion to refer to most of the papers digested a number of times, and at long intervals. I have always taken advantage of such occasions, as well as those on which I have met with a reprint, translation or abstract of a determination, to verify the rough draughts of my digests. Only in a couple of instances have I thus discovered a trifling error. On the other hand, I have been able to detect and point out numerous misprints and miscalculations in the original sources. While, therefore, I cannot hope entirely to have escaped error in the thousands of values I have copied, and the almost equal number of calculations I have made, I have strong hopes that the accuracy of this digest will be found at least on a par with that of the original papers.

When, as is the case with provoking frequency, chemists have given their analytical data, but have omitted to state the atomic weights, or other constants, assumed in calculating their results, I have recalculated their data with accepted constants, which I have in each case stated. I have also, in many instances, recalculated determinations of importance,

in which constants varying considerably from those now received were assumed. I have further reduced the determinations originally given in terms of O= 100, or of O = 15.96, to 0 = 16. No confusion, however, will be found between the numbers for which the original investigators are responsible and my own. All values which I have calculated are in italics, or, with my explanations, enclosed in square brackets. The only arithmetical operation I have permitted myself to perform without these indications is a multiplication or division by two; and even in such cases it will usually appear from the digest itself that this operation has been performed.

The abbreviations of the literary references are essentially those adopted in the Royal Society's Catalogue of Scientific Papers. The first reference in each case is to the source upon which I have depended. When two references are necessary, they are connected by the word and. When my authority is not the original source, that to which it is accredited in my authority is also mentioned.

In conclusion, I shall be grateful to any one who, by drawing my attention to omissions or mistakes, will assist me in perfecting a labor which has occupied all my available time for twenty months.

BERKELEY, CAL., April, 1878.

POSTSCRIPT.

In preparing the following paper, I designed making it preliminary to a discussion of the various determinations and of the value to be assigned to each, and in this work I had already made some progress. After presenting this paper to the Institution, however, I learned that Prof. F. W. Clarke had been for some time engaged on a similar undertaking, and to him I gladly resigned the discussion of the data here compiled. The two papers will appear in the same form, and may be regarded as complementary.

G. F. B.

ATOMIC WEIGHT DETERMINATIONS.

ALUMINIUM.

The specific heat of aluminium, as determined by Regnault and by Kopp, and the vapor density of volatile compounds, as determined by Deville and Troost and by Odling, indicate that the atomic weight of this element is about 27.5. (Gmelin-Kraut, Handbuch der Chemie, 1, 39; and L. Meyer, Moderne Theorien der Chemie, 50.)

J. J. BERZELIUS: 27.267 (0 = 16).

100 parts of anhydrous aluminic sulphate decomposed by heat, gave 29.934 parts of oxide. Preparation not described. Number of experiments, probably 1. In Berzelius' Lehrbuch these data are calculated for S 200.75, and give Al= 170.9 (0 = 100,) or 27.344 (O=16.) [If S = 32, the data give Al = 27.267.] (Poggend. Ann., 8, 1826, 187.*)

=

T. THOMSON: 30 (0 = 16).†

Thomson found, probably from analysis of the sulphate, (see appendix,) that 125 Al = 100 O. Thomson supposed aluminic oxide to be a protoxide. [If it is a sesqui-oxide, the data give Al at 30.] (Thomson's System of Chemistry, 7th ed., 1, 1831, 454.)

W. W. MATHER: 20.55 (016).

According to this chemist 0.646 grammes of chloride, prepared according to Woehler, gave 2.055 grammes argen

*This article by Berzelius, which contains the particulars of a large part of his earlier atomic weight determinations, will be referred to frequently in the course of this paper. It is unfortunately full of misprints, all of which are, by no means, corrected in the table of errata at the end of the volume. The correctly printed values of the atomic weights discussed in it are to be found in Poggend. Ann., 10, 1827, 339.

+It must be remarked, in justice to Dr. Thomson, that his atomic weight determinations are, properly speaking, of a different nature from those of other chemists. So thoroughly persuaded was he of the truth of Prout's hypothesis, (that the atomic weights of the elements are all exact multiples of that of hydrogen,) that his experiments were directed merely towards ascertaining which multiple, in any case, was to be adopted.

tic chloride, and 0.2975 aluminic oxide. (Silliman's Amer. Journ., 27 1835, 138, 241.) Berzelius points out the inconsistency of these data. (Berzelius' Jahresbericht, 15, 1835, 138.)

C. TISSIER: 27.12 (0 = 16).

Determined by dissolving aluminium in chlorhydric acid, evaporating to dryness with excess of nitric acid and decomposing the nitrate by heat. The aluminium employed contained 0.135 per cent. sodium. 1.935 of this metal gave 3.645 oxide. [If Na 23, these data give Al=27.12.] The metal was prepared by heating aluminic fluoride with purified sodium in a graphite crucible. (Paris Comptes Rend., 46, 1858, 1105.)

=

Determined by six experiments on the titration of aluminic chloride with argentic nitrate. The mean result was Al= 13.723 (0-8); extreme difference 0.09. The aluminic chloride, which had been prepared on a large scale, was purified by sublimation over iron-filings and over aluminium filings, and by a third sublimation in a current of hydrogen over aluminium filings, after which it was melted. Experiments on the oxidation of aluminium were found unsatisfactory on account of the difficulty of obtaining the metal pure. They gave Al at from 13.74 to 13.89. Dumas takes Ag = 108; Cl=35.5. (Ann. de Chim. et de Phys., (3,) 55, 1859, 151.)

W. ODLING: 27.5 (016).

Determined from the vapor density of aluminium methide and ethide at 220° and upwards. (Phil. Mag., (4,) 29, 1865, 316.)

- ISNARD: 27 (016).

Pure aluminium dissolved in chlorhydric acid, evaporated and heated to redness, gives 17 of its weight in oxide. (Paris Comptes Rend., 66, 1868, 508.)

Pelouze and Fremy give 27.357 (0 = 16); 170.98 (0 = 100,) for the atomic weight of aluminium, and assert that this value is derived from the composition of potash-alum, but they give no authority for the value. The experiments were made by precipitation with barium chloride.

de Chimie, 3d ed., 1, 50.)

(Traité