of acid driven off by heat, the residue weighed, the potassic sulphate leached out with water, and the residue again weighed. Marignac does not pretend that the determination is accurate. The results gave from 45.01 to 45.48. He thinks that some potassic sulphate may have escaped solution, and therefore takes the minimum. K = 39, S = 16. According to Marignac, Deville also found the atomic weight of zirconium somewhat higher than Berzelius by analysis of the chloride with which he determined the vapor density. (Annal. de Chim. et de Phys., (3,) 60, 1860, 257.) APPENDIX. DETERMINATIONS BY T. THOMSON. In Thomson's Annals of Philosophy, volumes 16 and 17, 1820-21, Thomson published a series of papers descriptive of experiments undertaken for the purpose of verifying Prout's hypothesis. His method consisted in mixing reagents in what he considered equivalent proportions, and after precipitation examining portions of the supernatant liquid for an excess of each of the salts supposed to neutralize one another. In all except four cases, either the salt analyzed was a sulphate and the precipitant barium chloride, or the determination was dependent upon such an analysis; yet although Thomson took barium = 70, in no instance was he able to detect either barium or sulphuric acid in the residual solution when the quantity of the re-agents corresponded to the atomic weights which he adopts. Comparison of his results with those reached by more accurate experimenters will make this exact neutralization appear impossible, nor were his contemporaries able to repeat his experiments successfully. Thomson's determinations are, as such, utterly valueless, yet as they were for many years extensively accepted in English and American scientific literature they are inserted here for reference. In the following table Thomson's numbers are multiplied, when necessary, for the sake of comparison with the values now accepted. Antimony at THOMSON FURTHER DETERMINED 132 by oxidation. 12 from analysis of borax. 200 by conversion of the oxide into chloride. 116 by oxidation with nitric acid. REDUCTION OF WEIGHINGS TO VACUUM. In discussing the analyses recorded in the foregoing pages, or in reconciling atomic weight determinations by various chemists, it may be found convenient to employ the following table. The maximum error involved is less than 0.01 per cent. or 0.1 milligram per gram. GRAM WEIGHTS BEING OF BRASS, FRACTIONS OF PLATINUM. For substances the sp. gr. of which exceeds 6.1; no correction is necessary. For substances the sp. gr. of which is less than 6.1:— To correct the entire grams; multiply their number by the correction in the table opposite the sp. gr. of the substance, found in the first column, and add the product to the observed weight. To correct the fractions of a gram, multiply the correction opposite the sp. gr. of the substance, found in the third column of the table, by the first two decimal figures of the observed weight, if the sp. gr. of the substance is less than 3, and by the first decimal only, if the sp. gr. exceeds 3, and add the product to the observed weight. ALL WEIGHTS USED BEING OF PLATINUM. For substances the sp. gr. of which exceeds 7.8, no correction is necessary. For substances the sp. gr. of which is less than 7.8:Multiply the correction opposite the sp. gr. of the substance, found in the third column, by the number of grams, tenths and hundredths observed, if the sp. gr. falls short of 3, or by the number of grams and tenths, if the sp. gr. exceeds 3, and add the product to the observed weight. The table shows within what limits it is necessary to know the sp. gr. |