ff MEMCAL SCHOOL SMITHSONIAN MISCELLANEOUS COLLECTIONS - '075 THE CONSTANTS OF NATURE PART V A RECALCULATION OF THE ATOMIC WEIGHTS BY FRANK WIGGLESWORTH CLARKE Chief Chemist of the U. S. Geological Survey NEW EDITION, REVISED AND ENLARGED CITY OF WASHINGTON PUBLISHED BY THE SMITHSONIAN INSTITUTION 1897 JUDD & DETWEILER, PRINTERS WASHINGTON, D. C. ADVERTISEMENT. The present publication is one of a series devoted to the discussion and more precise determination of various u Constants of Nature; " and forms the Fifth contribution to that subject published by this Institution. The First number of the series, embracing tables of u Specific Gravi- ties " and of Melting and Boiling Points of Bodies, prepared by the same author, Prof. F. W. Clarke, was published in 1873. The Fourth part of the series, comprising a complete digest of the various "Atomic Weight " determinations of the chemical elements published since 1814, com- mencing with the well-known " Table of Equivalents " by Wollaston (given in the Philosophical Transactions for that year), compiled by Mr. George F. Becker, was published by the Institution in 1880. The present work comprises a very full discussion and recalculation of the "Atomic Weights" from all the existing data, and the assignment of the most probable value to each of the elements. The first edition of this work was published in 1882, and this new edition, revised and enlarged by Professor Clarke, contains new informa- tion accumulated during the past fifteen years. S. P. LANGLEY, Secretary of the Smithsonian Institution. WASHINGTON, January, 1897. 13393 TABLE OF CONTENTS PAGE. Introduction , i Formulae for the Calculation of Probable Error 7 1 . Oxygen 8 2. Silver, Potassium, Sodium, Chlorine, Bromine, and Iodine 34 3. Nitrogen 58 4. Carbon 72 5. Sulphur 80 6. Lithium 84 7. Rubidium v 87 8. Caesium 89 9. Copper 91 10. Gold 101 1 1 . Calcium no 12. Strontium 1 13 13. Barium 1 18 14. Lead 127 15. Glucinum 132 16. Magnesium 135 1 7. Zinc 146 18. Cadmium 156 19. Mercury ., 166 20. Boron 171 21. Aluminum J76 22. Gallium 181 23. Indium 182 24. Thallium 184 25. Silicon 188 26. Titanium 190 27. Germanium 195 28. Zirconium , 196 29. Tin 199 30. Thorium 204 31. Phosphorus 209 32. Vanadium 211 33. Arsenic 213 34. Antimony 4 216 35. Bismuth 229 36. Columbium 234 37. Tantalum 236 38. Chromium 238 39. Molybdenum 250 40. Tungsten 255 41 . Uranium 263 42. Selenium 268 43. Tellurium 271 44. Fluorine ...... 277 (v) VI TABLE OF CONTENTS. PAGE. 45. Manganese 282 46. Iron . 287 47. Nickel and Cobalt 291 48. Ruthenium 311 49. Rhodium , 313 50. Palladium 315 51. Osmium , 322 52. Iridium 325 53. Platinum 327 54. Cerium 335 55. Lanthanum.. 344 56. The Didymiums 351 57. Scandium 354 58. Yttrium 355 59. Samarium, Gadolinium, "Erbium, and Ytterbium . 359 60. Terbium, Thulium, Holmium, Dysprosium, etc , . 362 61. Argon and Helium 363 Table of Atomic Weights 364 Index . . 3 6 7 A RECALCULATION" OF THE ATOMIC WEIGHTS. BY FRANK WIGGLESWORTH CLARKE. INTRODUCTION. In the autumn of 1877 the writer began collecting data relative to determinations of atomic weight, with the purpose of preparing a com- plete resume of the entire subject, and of recalculating all the estima- tions. The work was fairly under way, the material was collected and partly discussed, when I received from the Smithsonian Institution a manuscript by Professor George F. Becker, entitled " Atomic Weight Determinations: a Digest of the Investigations Published since 1814." This manuscript, which has since been issued as Part IV of the " Con- stants of Nature," covered much of the ground contemplated in my own undertaking. It brought together all the evidence, presenting it clearly and thoroughly in compact form ; in short, that portion of the task could not well be improved upon. Accordingly, I decided to limit my own labors to a critical recalculation of the data ; to combine all the figures upon a common mathematical basis, and to omit everything which could as well be found in Professor Becker's " Digest." In due time my work was completed, and early in 1882 it was pub- lished. About a year later Meyer and Seubert's recalculation appeared, to be followed later still by the less elaborate discussions of Sebelien and of Ostwald. All of these works differed from one another in various essential particulars, presenting the subject from different points of view, and with different methods of calculation. Each one, therefore, has its own special points of merit, and, in a sense, reinforces the others. At the same time, the scientific activity which they represent shows how widespread was the interest in the subject of atomic weights, and how fundamentally important these constants undoubtedly are. The immediate effect of all these publications was to render manifest the imperfections of many of the data, and to point out most emphatic- ally in what directions new work needed to be done. Consequently, there has been since 1884 an extraordinary activity in the determination of atomic weights, and a great mass of new material has accumulated. The assimilation of this material, and its combination with the old data, is the object of the present volume. (1) 2 THE ATOMIC WEIGHTS. At the very beginning of my work, certain fundamental questions con- fronted me. Should I treat the investigations of different individuals separately, or should I combine. similar data together in a manner irre- spective of persons ? For example, ought I, in estimating the atomic weight of silver, to take Stas' work by itself, Marignac's work by itself, and so. on, and then average the results together; or should I rather combine all series of figures relating to the composition of potassium chlorate into one mean value, and all the data concerning the composi- tion of silver chloride into another mean, and, finally, compute from such general means the constant sought to be established ? The latter plan was finally adopted ; in fact, it was rendered necessary by the method of least squares, which, in a special, limited form, was chosen as the best method of dealing with the problem. The mode of discussion and combination of results was briefly as follows. The formula employed are given in another chapter. I began with the ratio between oxygen and hydrogen ; in other words, with the atomic weight of oxygen referred to hydrogen as unity. Each series of experiments was taken by itself, its arithmetical mean was found, and the probable error of that mean was computed. Then the several means were combined according to the appropriate formula, each receiving a weight dependent upon its probable error. The general mean thus estab- lished was taken as the most probable value for the atomic weight of oxygen, and, at the same time, its probable error was mathematically assigned. Next in order came a group of elements which were best discussed together, namely, silver, chlorine, potassium, sodium, bromine, and iodine. For these elements there were data from many experimenters. All similar figures were first reduced to common standards, and then the means of individual series were combined into general means. Thus all the data w r ere condensed into nineteen ratios, from which several independent values for the atomic weight of each element could be computed. The probable errors of these values, however, all involved the probable error of the atomic weight of oxygen, and were, therefore, higher than they would have been had the latter element not entered into consideration. Here, then, we have suggested a chief peculiarity of this whole revision. The atomic weight of each element involves the probable errors of all the other elements to which it is directly or indirectly referred. Accordingly, an atomic weight determined by refer- ence to elements whose atomic weights have been defectively ascertained will receive a high probable error, and its weight, when combined with other values, will be relatively low. For example, an atomic weight ascertained by direct comparison with hydrogen will, other things being equal, have a lower probable error than one which is referred to hydro- gen through the intervention of oxygen ; and a metal whose equivalent involves only the probable error of oxygen should be more exactly INTRODUCTION. 3 / known than one which depends upon the errors of silver and chlorine. These points will appear more clearly evident in the subsequent actual discussions. But although the discussion of atomic weights is ostensibly mathe- matical, it cannot be purely so. Chemical considerations are necessarily involved at every turn. In assigning weights to mean values I have been, for the most part, rigidly guided by mathematical rules ; but in some cases I have been compelled to reject altogether series of data which were mathematically excellent, but chemically worthless because of constant errors. In certain instances there were grave doubts as to whether particular figures should be included or rejected in the calcula- tion of means, there having been legitimate reasons for either procedure. Probably many chemists would differ with me upon such points of judg- ment. In fact, it is doubtful whether any two chemists, working inde- pendently, would handle all the data in precisely the same way, or combine them so as to produce exactly the same final results. Neither would any two mathematicians follow identical rules or reach identical conclusions. In calculating the atomic weight of any element those values are assigned to other elements which have been determined in previous chapters. Hence a variation in the order of discussion might lead to slight differences in the final results. As a matter of course the data herein combined are of very unequal value. In many series of experiments the weighings have been reduced to a vacuum standard ; but in most cases chemists have neglected this correction altogether. In a majority of instances the errors thus intro- duced are slight ; nevertheless they exist, and -interfere more or less with all attempts at a theoretical consideration of the results. Necessarily, this work omits many details relative to experimental methods, and particulars as to the arrangement of special forms of appa- ratus. For such details original memoirs must be consulted. Their in- clusion here would have rendered the work unwarrantably bulky. There is such a thing as over-exhaustiveness of treatment, which is equally objectionable with under-thoroughness. Of course, none of the results reached in this revision can be consid- ered as final. Every one of them is liable to repeated corrections. To my mind the real value of the work, great or little, lies in another direc- tion. The data have been brought together and reduced to common standards, and for each series of figures the probable error has been de- termined. Thus far, however much my methods of combination may be criticised, I feel that my labors will have been useful. The ground is cleared, in a measure, for future experimenters; it is possible to see more distinctly what remains to be done ; some clues are furnished as to the relative merits of different series of results. On the mathematical side my method of recalculation has obvious deficiencies. It is special, rather than general, and at some future time, when a sufficiently large mass of evidence has accumulated, it must 4 THE ATOMIC WEIGHTS. give way to a more thorough mode of treatment. For example, the ratio Ag 2 : BaBr 2 has been used for computing the atomic weight of barium, the atomic weights of silver and bromine being supposed to be known. But these atomic weights are subject to small errors, and they are super- imposed upon that of the ratio itself in the process of calculation. Ob- viously, the ratio should contribute to our knowledge of all three of the atomic weights involved in it, its error being distributed into three parts instead of appearing in one only. The errors may be in part compensa- tory ; but that is not certainly known. Suppose now that for every element we had a goodly number of atomic weight ratios, connecting it with at least a dozen other elements, and all measured with reasonable accuracy. These hundreds of ratios could then be treated as equations of observation, reduced to linear form, and combined by the general method of least squares into normal equations. All errors would thus be distributed, never becoming cumulative ; and the normal equations, solved once for all, would give the atomic weights of all the elements simultaneously. The process would be laborious but the result would be the closest possible approach to accuracy. The data as yet are inadequate, although some small groups of ratios may be handled in that way ; but in time the method is sure to be applied, and indeed to be the only general method applicable. Even if every ratio was subject to some small constant error, this, balanced against the similar errors of other ratios, would become accidental or unsystematic with reference to the entire mass of material, and would practically vanish from the final means. Concerning this subject of constant and accidental errors, a word may be said here. My own method of discussion eliminates the latter, which are removable by ordinary averaging ; but the constant errors, vicious and untractable, remain, at least partially. Still, where many ratios are considered, even the systematic errors may in part compensate each other, and do less harm than might be expected. They have, moreover, a peculiarity which deserves some attention. In the discussion of instrumental observations, the systematic errors are commonly constant, both as to direction and as to magnitude. They are therefore independent of the accidental errors, and computation of means leaves them untouched. But in the measurement of chemical ratios the constant errors are most frequently due to an impurity in one of the materials investigated. If different samples of a substance are studied, although all may contain the same impurity, they are not likely to contain it in the same amount ; and so the values found for the ratio will vary. In other words, such errors may be constant in direction but variable in magnitude. That variation appears in the probable error computed for the series of observations, diminishes its weight when com- bined with other series, and so, in part, corrects itself. It is not removed from the result, but it is self-mitigated. The constant errors familiar to the physicist and astronomer are obviously of a different order. INTRODUCTION. 5 That all methods of averaging are open to objections, I am of course perfectly aware. I also know the doubts which attach to all questions of probable error, and to all combinations of data which depend upon them. I have, however, preferred to face these objections and to recog- nize these doubts rather than to adopt any arbitrary scheme which per- mits of a loose selection of data. After all, the use of probable error as a means of weighting is but a means of weighting, and perhaps more justifiable than any other method of attaining the same result. When observations are weighted empirically that is, by individual judg- ment far greater dangers arise. Almost unconsciously, the work of a famous man is given greater weight than that of some obscure chemist, although the latter may ultimately prove to be the best. But the prob- able error of a series of measurements is not affected by the glamor of great names; and the weight which it assigns to the observations is at least as safe as any other. In the long run, I believe it assigns weight more accurately, and therefore I have trusted to its indications, not as if it were a mathematical fetish, but regarding it as a safe guide, even though sometimes fallible. In Meyer and Seubert's recalculation, weights are assigned in quite a novel manner. In each series of experiments the maximum and mini- mum results are given, but instead of the mean there is a value deduced from the sum of the weighings that is, each experiment is weighted proportionally to the mass of the material handled in it. For this method I am unable to find any complete justification. Of course, the errors due to the operations of weighing become proportionally smaller as the quantity of material increases, but these errors, with modern apparatus, are relatively unimportant. The real errors in atomic weight determinations are much larger than these, and due to different causes. Hence an experiment upon ten grammes of material may be a little better than one made upon five grammes, but it is by no means necessarily twice as good. The ordinary mean of a series of observations, with its measure of concordance, the probable error, is a better value than one obtained in the manner just described. If only errors of weighing were to be considered, Meyer and Seubert's summation method would be valid, but in the presence of other and greater errors it seems to have but little real pertinency to the problem at hand. In addition to the usual periodicals, the following works have been freely used by me in the preparation of this volume: BERZELIUS, J. J. Lehrbuch der Chemie. 5 Auflage. Dritter Band. SS. 1147-1231. 1845. VAN GEUNS, W. A. J. Prceve eener Geschiedenis van de ^Equivalent- getallen der Scheikundige Grondstoffen en van hare Soortelijke Gewigten in Gasvorm, voornamelijk in Betrekking tot de vier Grondstoffen der Bewerktuigde Natuur. Amsterdam, 1853. O THE ATOMIC WEIGHTS. MULDER, E. Historisch-Kritisch Overzigt van de Bepalingen der JEquiv- alent-Gewigten van 13 Eenvoudige Ligchamen. Utrecht, 1853. MULDER, L. Historisch-Kritisch Overzigt van de Bepalingen der JLquiv- alent-Gewigten van 24 Metalen. Utrecht, 1853. OUDEMANS, A. C., Jr. Historisch-Kritisch Overzigt van de Bepaling der ^Equivalent-Gewigten. van Twee en Twintig Metalen. Leiden, 1853. STAS, J. S. Untersuchungen iiber die Gesetze der Chemischen Propor- tionen iiber die Atomgewichte und ihre gegenseitigen Verhalt- nisse. Uebersetzt von Dr. L. Aronstein. Leipzig, 1867. See also his " Oeuvres Completes," 3 vols., published at Bruxelles in 1894. MEYER, L., and SEUBERT, K. Die Atomgewichte der Elemente, aus den Originalzahlen neu berechnet. Leipzig, 1883. SEBELIEN, J. Beitrage zur Geschichte der Atomgewichte. Braunschweig y 1884. OSTWALD, W. Lehrbuch der allgemeinen Chemie. Zweite Aufl. I Band. SS. 18-138. Leipzig, 1891. The four Dutch monographs above cited are especially valuable. They represent a revision of all atomic weight data down to 1853, as divided between four writers. For the sake of completeness the peculiar volume by Hinrichs * must also be cited, although the methods and criticisms embodied in it have not been generally endorsed. Hinrichs' point of view is so radically different from mine that I have been unable to make use of his discus- sions. His objections to the researches of Stas seem to be quite un- founded ; and the rejoinders by Spring and by Van der Plaats are suffi- ciently thorough. * The True Atomic Weight of the Chemical Elements and the Unity of Matter. St. I^ouis, 1894. Compare Spring, Chem. Zeitung, Feb. 22, 1893, and Van der Plaats, Compt. Rend., 116, 1362. See also a paper by Vogel, with adverse criticisms by Spring and L,. Henry, in Bull. Acad. Bruxelles, (3), 26, 469. INTRODUCTION. FORMULAE FOR THE CALCULATION OF PROBABLE ERROR. The formula for the probable error of an arithmetical mean, familiar to all physicists, is as follows : Here n represents the number of observations or experiments in the series, and S the sum of the squares of the variations of the individual results from the mean. In combining several arithmetical means, representing several series, into one general mean, each receives a weight inversely proportional to the square of its probable error. Let A, B, C, etc., be such means, and a, 6, c their probable errors respectively. Then the general mean is de- termined by the formula : A JL + __. (2.) u = ^'-^'^- For the probable error of this general mean we have : In the calculation of atomic and molecular weights the following formulae are used : Taking, as before, capital letters to represent known quantities, and small letters for their probable errors respectively, we have for the probable error of the sum or difference of two quantities, A and B : For the product of A multiplied by B the probable error is (5.) e = For the product of three quantities, ABC : T> For a quotient, -T' the probable error becomes (7.) 8 THE ATOMIC WEIGHTS. Given a proportion, A : B : : C : x, the probable error of the fourth term is as follows : This formula is used in nearly every atomic weight calculation, and is, therefore, exceptionally important. Rarely a more complicated case arises in a proportion of this kind : In this proportion the unknown quantity occurs in two terms. Its probable error is found by this expression, and is always large : (9.) When several independent values have been calculated for an atomic weight they are treated like means, and combined according to formulae (2) and (3). Each final result is, therefore, to be regarded as the general or weighted mean of all trustworthy determinations. This method of combination is not theoretically perfect, but it seems to be the one most available in practice. OXYGEN. The ratio between oxygen and hydrogen is the foundation upon which the entire system of atomic weights is sustained. Hence, the accuracy of its determination has, from the beginning, been recognized as of ex- treme importance. A trifling error here may become cumulative when repeated through a moderate series of other ratios. But few of the elements have, so far, been compared directly with th unit, hydrogen ; practically all of them are referred to it through the intervention of oxygen, and therefore the ratio in question requires discussion before any other can be profitably considered. Leaving out of account the earliest researches, which now have only historical value, the first determinations to be noted are those of Dulong and Berzelius,* who, like some of their successors, effected the synthesis of water over heated oxide of copper. The essential features of the method are in all cases the same. Hydrogen gas is passed over the hot oxide, and the water thus formed is collected and weighed. From this weight and the loss of weight which the oxide undergoes, the exact com- * Thomson's Annals of Philosophy, July, 1821, p. 50. OXYGEN. 9 position of water is readily calculated. Dulong and Berzelius made but three experiments, with the following results for the percentages of oxygen and hydrogen in water : O. H. 88.942 11.058 88.809 11.191 88.954 11.046 From these figures we get, for the atomic weight of oxygen, the values 16.124 15-863 16. 106 .Mean, 16.031, db .057. As the weighings were not reduced to a vacuum, this correction was afterwards applied by Clark,* who showed that these syntheses really make = 15.894 ; or, in Berzelian terms, if = 100, H = 12.583. The value 15.894, dz .057 we may therefore take as the true result of Dulong and Berzelius' experiments, a result curiously close to that reached in the latest and best researches. In 1842. Dumas f published his elaborate investigation upon the com- position of water. The first point was to get pure hydrogen. This gas, evolved from zinc and sulphuric acid, might contain oxides of nitrogen ? sulphur dioxide, hydrosulphuric acid, and arsenic hydride. These im- purities were removed in a series of wash bottles; the H 2 S by a solution of lead nitrate, the H 3 As by silver sulphate, and the others by caustic potash. Finally, the gas was dried by passing through sulphuric acid, or, in some of the experiments, over phosphorus pentoxide. The copper oxide was thoroughly dried, and the bulb containing it was weighed. By a current of dry hydrogen all the air was expelled from the apparatus, and then, for ten or twelve hours, the oxide of copper was heated to dull redness in a constant stream of the gas. The reduced copper was allowed to cool in an atmosphere of hydrogen. The weighings were made with the bulbs exhausted of air. The following table gives the results : Column A contains the symbol of the drying substance ; B gives the weight of the bulb and copper oxide ; C, the weight of bulb and reduced copper ; D, the weight of the vessel used for collecting the water ; E, the same, plus the water ; F, the weight of oxygen ; G, the weight of water formed ; H, the crude equivalent of H when O = 10,000 ; I, the equiva- lent of H, corrected for the air contained in the sulphuric acid employed. This correction is not explained, and seems to be questionable. * Philosophical Magazine, 3d series, 20, 341. fCompt. Rend., 14, 537. 10 THE ATOMIC WEIGHTS. vO O N ' O vO O coo -" ON O oo oo N N <-> r^o ON OO tvON^j-vocoONVOOO* ON O ^f NO* M' CN vo ^- Oo' h vo vo vo **t" vo ^" ^t* vo vo vo vo T}- vo vo ^* vo O *i NO N CO NO O CO *tf- N CO tv *- OO NO cOvoOO O ON oc| O* NO' NO' rf- O oo" O '-' co tv. 06 vo o iv, tv od VO T ' vo vo vo vo vo vo vo vo vo vo vo vo vo lO vo Tf NNNNNNNNNNN tv. vo co rj- O tv OQ CONO OOO N ONO tv, O OO vo tv M OVOTJ-NO rj- tv, CNI 00 N tv,OO ONNO tv,ONvotv,tv. 00 ONOO ONO HHNO vocoO noo covocOTj-i-iNO T^-N' COTJ-VOO^ON"-" tv,oO* ONtv.O\Oo' CO *-* 00* NO" ^f HH N NNOOO ThcovoNO VOVONONO VONO ^cococo ON?NJ vOTj-rfi-i w tv.i-.OOOO ONOOO COOiN COlv. IV.NO ONQNO tvoO cocoOOO ONCOOOOONO COM -" co^O covooooo OCOvorv,ooo ^t-tv^H woo CO O' O' rv NO* CO rf vo O i-< N ON N M \o* NO" rj- M* o" t-i C>4 W vorv^-co^-NO VOVOVONO vovococococo T}-N TJ-COMNO N r^NO lv.O O covoONO VOIV.ON COCONO M ONOOC T}- ri- i-i i- O tvio cotvcoco NO ^^ tv, co ci c^ ^- c^ c^ ^ ON t~v 01 ^h co O r^ oo vo vo 1-1 N* CO* CO NO oo" NO" o' ON co OO* N ON CO o' NO* ^ M* ON -< NO ^J" tvi ** tv. t^, Q\ ON co ON vo ON C4 c^ C^l c^ M Tj-voTl-ONONONOOOQOO tv.ONONtv.tv-i ONONONON tvi tv. ** O *"* ON Tf" ^- O vo W tv. ^" t*^ C*J O tVi CJ C^ O C^4 i^ ON co vo O CN) NO ON co OO tv. o^ NO ^" NO NO OO M tvi-i ro "fONONO OOO ^coO tvNOOONOOO o'oo o\Tftv,tv,o\T*-c This series appeared in the earlier paper, but with an error which is here corrected. Second Series. Electrolytic Hydrogen, Dried by Phosphorus Pentoxide. Wt. of H. Wt. of O. At. Wt. O. .5044 4.0095 15-898 6325 5-0385 15-932 .6349 5-5i7 15-913 .55 6 4 4.4175 15-879 7335 5-8224 15.876 5.3181 15.885 Mean, 15.8971,^.0064. Third Series. Electrolytic Hydrogen, Dried by Passage Through a Tube Packed with Sodium Wire. Wt. of H. Wt. of O. At. Wt. O. .9323 7.4077 15-891 9952 7-9045 15 885 .3268 2.5977 15.898 .7907 6.2798 15.884 .7762 6.1671 15.891 1.1221 8.9131 15-887 Mean, 15.8893, i .0014 At the end of this series it was found that the hydrogen contained a trace of water, estimated to be equivalent to an excess of three milli- grammes in the total h}^drogen of the six experiments. Correcting for this, the mean becomes = 15.899. Fourth Series. Electrolytic Hydrogen, Dried over Freshly Sublimed Phos- phorus Pentoxide. Wt. of H. Wt. of O. At. Wt. O. 1.0444 8.3017 15-898 .7704 6.1233 15.896 .8231 6.5421 15.896 .8872 7.0490 15.890 9993 7-9403 15-892 1.1910 9.4595 15.885 Mean, 15.8929, .0013 OXYGEN. 19 The mean of all the twenty-four determinations, taken as one series, with the correction to the third series included, is = 15.8966, .0017. In sum, there were consumed 18.5983 grammes of hydrogen and 147.8145 of oxygen ; whence = 15.8955. Dittmar and Henderson,* who effected the synthesis of water over copper oxide by what was essentially the old method, begin their memoir with an exhaustive criticism of the work done by Dumas and by Erd- mann and Marchand. They show, as I have already mentioned, that hydrogen dried by sulphuric acid becomes contaminated with sulphur dioxide, and also that a gas passed over calcium chloride may still retain as much as one milligramme of water per litre. Fused caustic potash they found to dry a gas quite completely. In their first series of syntheses, Dittmar and Henderson generated their hydrogen from zinc and acid, sometimes hydrochloric and some- times sulphuric, and dried it by passage, first through cotton wool, then through vitrioled pumice, then over red-hot metallic copper to remove oxygen. In later experiments it first traversed a column of fragments of caustic soda to remove antimony derived from the zinc. The oxide of copper used was prepared by heating chemically pure copper clip- pings in a muffle, and was practically free from .sulphur. In weighing the several portions of apparatus it was tared with somewhat lighter similar pieces of as nearly as possible the same displacement. The re- sults of this series of experiments, which are vitiated by the presence, unsuspected at first, of sulphur dioxide in the hydrogen, are stated in values of H when = 16, but in the following table .have been recalcu- lated to the usual unit : Wt. of Water. Wt. of O. At. Wt. O. 4.7980 4.26195 I5-9 01 7.55025 6.71315 16.039 6.2372 5-53935 15.875 11.29325 10.03585 J 5-963 11.6728 10.3715 I5-940 11.8433 10.5256 15.976 11.7317 10.4243 15-947 19.2404 17.0926 15.916 20.83435 18.5234 16.031 17.40235 15.4598 I5-9I7 19.2631 i7."485 x 5-934 Mean, 15.949, =b .0103. Reducing to a vacuum, this becomes 15.843, while a correction for the sulphur dioxide estimated to be present in the hydrogen brings the value * Proc. Roy. Soc. Glasgow, 22, 33. Communicated Dec. 17, 1890. 20 THE ATOMIC WEIGHTS. up again to 15.865. Still another correction is suggested, namely, that as the reduced copper in the combustion tube, before weighing, was ex- posed to a long-continued current of dry air, it may have taken up traces of oxygen chemically, thereby increasing its weight. As this correction, however, is quantitatively uncertain, it may be neglected here, and the result of this series will be taken as = 15.865, .0103. Its weight, relatively to some other series of experiments, is evidently small. In their second and final series Dittmar and Henderson dried their hydrogen, after deoxidation by red-hot copper, over caustic potash and subsequently phosphorus pentoxide. The copper oxide and copper of the combustion tube were both weighed in vacuo. The results were as follows, vacuum weights being given : Wt. Water. Wt. O. At. Wt. O. 19.2057 17-0530 15.843 19-5211 17-3342 [15-853] 19.4672 17.2882 15.868 22.9272 20.3540 15.820 23.0080 20.4421 [15.934] 23.4951 26.8639 15.859 23.5612 20.9226 [15-859] 23.7542 ^ 21.0957 15.870 23.6568 21.8994 15.884 23.6179 21.8593 15.848 24.6021 21.8499 15.878 24.3047 21.5788 15.832 23.6172 20.9709 15.849 Mean, 15.861,^.0052. The authors reject the three bracketed determinations, because of irregularities in the course of the experiments. The mean of the ten remaining determinations is 15.855, .0044. Both means, however, have to be corrected for the minute trace of hydrogen occluded by the reduced copper. This correction, experimentally measured, amounts to -|- .006. Hence the mean of all the experiments in the series becomes 15.867, .0052, and of the ten accepted experiments, 15.861, .0044. The authors themselves select out seven experiments, giving a corrected mean of 15.866, which they regard as the best value. Taking all their evidence, their two series combine thus : First series 15.865, .0103 Second series 15.867, .0052 General mean 15.8667, .0046 Leduc,* who also effected the synthesis of water over copper oxide, * Compt. Rend., 115, 41. 1892. OXYGEN. 21 following Dumas' method with slight modifications, gives the results of only two experiments, as follows : Wt. Water. Wt. O. At. Wt. O. 22.1632 19.6844 15.882 19.7403 17.5323 15-880 Mean, 15.881 These experiments we may arbitrarily assign equal weight with two in Dittmar and Henderson's later series, when the result becomes 15.881, =b .0132, the value to be accepted. Leduc states that his copper oxide, which was reduced at as low a temperature as possible, was pre- pared by heating clippings of electrolytic copper in a stream of oxygen. To E. W. Morley * we owe the first complete quantitative syntheses of water, in which both gases were weighed separately, and afterwards in combination. The hydrogen was weighed in palladium, as was done by Keiser, and the oxygen was weighed in compensated globes, after the manner of Regnault. The globes were contained in an artificial " cave," to protect them from moisture and from changes of temperature; being so arranged that they could be weighed by the method of reversals with- out opening either the " cave " or the balance case. For each weighing of hydrogen about 600 grammes of palladium were employed. After weighing, the gases were burnt by means of electric sparks in a suitable apparatus, from which the unburned residue could be withdrawn for examination. Finally, the apparatus containing the water produced was closed by fusion and also weighed. Rubber joints were avoided in the construction of the apparatus, and the connections were continuous throughout. The weights are as follows : H taken O taken. H^O formed. 3-2645 25.9*76 29.1788 3.2559 25-8531 29.1052 3.8193 30.3210 34-1389 3-8450 3o.5 2 94 Lost 3.8382 30.4700 34.3151 3.8523 30.5818 34.4327 3.8298 30-40 1 3 34.2284 3.8286 30.3966 34.2261 3.8225 30-3497 34 1742 3.8220 30.3479 34-1743 3.7637 29.8865 33.6540 3.8211 30-3429 34.1559 * " On the Density of Oxygen and Hydrogen, and on the Ratio of their Atomic Weights," by Edward W. Morley. Smithsonian Contributions to Knowledge, 1895, 4to, xi + .117 pp., 40 cuts. Abstract in Am. Chem. Journ., 17, 267 (gravimetric), and Ztschr. Phys. Chem., 17, 87 (gaseous densi- ties) ; also note in Am. Chem. Journ., 17, 396. Preliminary notice in Proc. Amer. Association, 1891, p. 185. 22 THE ATOMIC WEIGHTS. Hence we have H : O Ratio H-.H^O Ratio. 15-878 17.877 15.881 I7 .8 7 8 15-878 17.873 15.880 15.877 17.881 15-877 17-876 15.877 17-875 15-878 17.879 15-879 17.881 15-881 17.883 15.881 17.883 15-882 17.878 Mean, 15.8792, .00032 Mean, 17.8785, .00066 Combined, these data give : From ratio H 2 : O . O 15.8792, .00032 " H 2 :H 2 0^15.8785,^.00066 General mean O 15.8790, =b .00028 For details, Morley's fall paper must be consulted. No abstract can do justice to the remarkable work therein recorded. Two other series of determinations, by Julius Thomsen, remain to be noticed. In the earlier paper * he determined the ratio between HC1 and NH 3 , and thence, using Stas' values for Cl and N, fixed by reference to = 16, computed the ratio H : 0. This method was so indirect as to be of little importance, and gave for the atomic weight of oxygen approxi- mately the round number 16. I shall use the data farther on in cal- culating the atomic weight of nitrogen. The paper has been sufficiently criticised by Meyer and Seubert,f who have discussed its sources of error. In Thomsen's later paper J a method of determination is described which is, like the preceding, quite novel, but more direct. First, alu- minum, in weighed quantities, was dissolved in, caustic potash solution. In one set of experiments the apparatus was so constructed that the hydrogen evolved was dried and then expelled. The loss of weight of the apparatus gave the weight of the hydrogen so liberated. In the second set of experiments the hydrogen passed into a combustion chamber in which it was burned with oxygen, the water being retained. The increase in weight of this apparatus gave the weight of oxygen so taken up. The two series, reduced to the standard of a unit weight of aluminum, gave the ratio between oxygen and hydrogen. *Zeitsch. Physikal. Chem., 13, 398. 1894. fBer., 27, 2770. I Zeitsch. Anorg. Chem., :r, 14. 1895. OXYGEN. 23 The results of the two series, reduced to a vacuum and stated as ratios, are as follows : First. Second. Weight of H Weight of O Weight of Al' Weight of Al* o.iuSo 0.88788 0.11175 0.88799 0.11194 0.88774 0.11205 0.88779 0.11189 0.88785 o.i i 200 0.88789 0.11194 0.88798 0.11175 0.88787 0.11190 0.88773 0.11182 0.88798 0.11204 0.88785 o.i 1 202 0.11204 0.88787,^0.000018 0.11179 0.11178 O.I 1202 0.11188 0.11186 0.11185 o.i 1 190 0.11187 1 0.11190, 0.000015 Dividing the mean of the second column by the mean of the first, we have for the equivalent of oxygen : 0.88787, 0.000018 0.11190, 0.000015 Hence == 15.8690, 0.0022. = 7-9345, 0.0011 The details of the investigation are somewhat complicated, and involve various corrections which need not -be considered here. The result as stated includes all corrections and is evidently good. The ratios, how- ever, cannot be reversed and used for measuring the atomic weight of aluminum, because the metal employed was not absolutely pure. We have now before us, representing syntheses of water, thirteen series, as follows : Dulong and Berzelius O = 15.894, .057 Dumas .. . 15.9607, .0070 Erdmann and Marchand l 5-975, .0113 Thomsen, 1870 15.91, .0113 Cooke and Richards 15.869, .0020 Reiser, 1887 15.864, .015 1888 15.9514, .0011 24 THE ATOMIC WKIGHTS. Rayleigh 15.89, .009 Noyes 15.8966,^.0017 Dittmar and Henderson 15.8667, .0046 Leduc 15.881, d= .0132 M.orley 15.8790, .00028 Thomson, 1895 15.8690, .0022 General mean O = 15.8837, .00026 Rejecting Keiser 1 5. 8796, .00027 If we reject all except the determinations of Cooke and Richards, Ray- leigh, Noyes, Dittmar and Henderson, Leduc, Thomsen, and Moiiey, the general mean of these becomes 15.8794, .00027. From this it is evi- dent that Reiser's determinations alone, among the higher values for 0. carry any appreciable weight : and it also seems clear that the rounded- off number, O == 15.88, .0003, cannot be very far from the truth; at least so far as the synthetic evidence goes. In discussing the relative densities of oxygen and hydrogen gases we need consider only the more modern determinations, beginning with those of Dumas and Boussingault. As the older work has some his- torical value, I may in passing just cite its results. For the density of hydrogen we have .0769, Lavoisier; .0693, Thomson; .092, Cavendish; .0732, Biot and Arago ; .0688, Dulong and Berzelius. For oxygen there are the following determinations:' 1.087, Fourcroy, Vauquelin, and Se- guin; 1.103, Kirwan; 1.128, Davy; 1,088, Allen and Pepys ; 1.1036, Biot and Arago; 1.1117, Thomson; 1.1056, De Saussure; 1.1026, Dulong and Berzelius; 1.106, Buff; 1.1052, Wrede.* In 1841 Dumas and Boussingault f published their determinations of gaseous densities. For hydrogen they obtained values ranging from .0691 to .0695 ; but beyond this mere statement they give no details. For oxygen three determinations were made, with the following results : '.1055 1.1058 Mean, 1.10567, .00006 If we take the two extreme values given above for hydrogen, and re- gard them as the entire series, they give us a mean of .0693, .00013. This mean hydrogen value, combined with the mean for oxygen, gives for the latter, when H = 1, the density ratio 15.9538, .031. Regnault's researches, published four years later, J were much more * For Wrede's work, see Berzelius' Jahresbericht for 1843. For Dulong and Berzelius, see the paper already cited. All the other determinations are taken from Gmelin's Handbook, Caven- dish edition, v. i, p. 279. f Compt. Rend., 12, 1005. Compare also with Dumas, Compt. Rend., 14, 537. J Compt. Rend., 20, 975. OXYGEN. 25 elaborately executed. Indeed, they have long stood among the classics of physical science, and it is only recently that they have been sup- planted by other measurements. For hydrogen three determinations of density gave the following results : .06923 .06932 .06924 Mean, .069263, .000019 For oxygen four determinations were made, but in the first one the gas was contaminated by traces of hydrogen, and the value obtained, 1.10525, was, therefore, rejected by Regnault as too low. The other three are as follows : 1.10561 1.10564 1.10565 Mean, 1.105633, .000008 Now, combining the hydrogen and oxygen series, we have the ratio H : : : 1 : 15.9628, .0044. According to Le Conte,* Regnault's reduc- tions contain slight numerical errors, which, corrected, give for the density of oxygen, 1.105612, and for hydrogen, .069269. Ratio, 1 : 15.9611. A much weightier correction to Regnault's data has already been in- dicated in the discussion of Cooke and Richards' work. He assumed that the globes in which the gases were weighed underwent no changes of volume, but Agamennone,f and after him, but independently,! Lord Rayleigh showed that an exhausted vessel was perceptibly compressed by atmospheric pressure. Hence its volume when empty was less than its volume when filled with gas. Crafts, having access to Regnault's original apparatus, has determined the magnitude of the correction indi- cated^ Unfortunately, the globe actually used by Regnault had been destroyed, but another globe of the same lot was available. With this the amount of shrinkage during exhaustion was measured, and Reg- iiault's densities were thereby changed to 1.10562 for oxygen, and .06949 for hydrogen. Corrected ratio, 1 : 15.9105. Doubtless Dumas and Boussingault's data are subject to a similar correction, and if we assume that it is proportionally the same in amount, the ratio derived from their experiments becomes 1 : 15.9015. In the same paper, that which contained the discovery of this correc- tion, Lord Rayleigh gives a short series of measurements of his own. * Private communication. See also Phil. Mag. (4), 27, 29, 1864, and Smithsonian Report, 1878, p. 428. f Atti Rendiconti Acad. I^incei, 1885. t Proc. Roy. Soc., 43, 356. Feb., 1888. g Conipt. Rend., 106, 1662. 26 THE ATOMIC WEIGHTS. His hydrogen was prepared from zinc and sulphuric acid, and was puri- fied by passage over liquid potash, then through powdered mercuric chloride, and then through pulverized solid potash. It was dried by means of phosphorus pentoxide. His oxygen was derived partly from potassium chlorate, and partly from the mixed chlorates of sodium and potassium. Equal volumes of the two gases weighed as follows : H. .15811 .15807 .15798 I579 2 O. 2.5186, 4; .00061* Mean, .15802, 000029. Corrected for shrinkage of the exhausted globe these become H, 0.15860 ; O, 2.5192. Hence the ratio 1 : 15.884, .0048. In 1892 Rayleigh published a much more elaborate determination of this ratio. f The gases were prepared electroly tically from caustic potash , and dried by means of solid potash and phosphorus pentoxide. The hydrogen was previously passed over hot copper. The experiments, stated like the previous series, are in five groups ; two for oxygen and three for hydrogen; but for present purposes the similar sets may be regarded as equal in weight, and so discussable together. The weights of equal volumes are as follows : H. O. ( -15807 2.5182 1 _. 15816 2.5173 First set .15811 2.5172 First set. Mean, .15808 I .15803 2.5193 Mean, 2.51785. .15801 2.5174 L -15809 2.5177 f .15800 2.5183 ' Second set '*& 2Q 2 '5 l68 Second set. Mean, .15797 .15792 25172 .15788 2.5181 Mean, 2.5172. .15783 2.5156 r .15807 .15801 Mean, 2.5176, .00019. .15817 Third set j .1579 Mean, .15804 .15810 .15798 .15802 1.15807 Mean, .15804, .000019. * Arbitrarily assigned the probable error of a single experiment in Rayleigh's paper of 1892. tProc. Roy. Soc., 50, 448, Feb. 18, 1892. OXYGEN. 27 These weights with various corrections relative to temperatures and pressures, and also for the compression of the exhausted globe, ulti- mately become for H, .158531 ; and for 0, 2.51777. Hence the ratio 1 : 15.882, HZ .0023. For details relative to corrections the original memoir should be consulted. In his paper " On a new method of determining gas densities," * Cooke gives three measurements for hydrogen, referred to air as unity. They are : .06957 .06951 .06966 Mean, .06958, ih .000029 Combining this with Regnault's density for oxygen, as corrected by Crafts, 1.10562, .000008, we get the ratio H : : : 1 : 15.890,* .0067. Leduc, working by Regnault's method, somewhat modified, and cor- recting for shrinkage of exhausted globes, gives the following densities : t H. O. .06947 1.10501 .06949 1.10516 .06947 Mean, .06948, =b .00006745 The two oxygen measurements are the extremes of three, the mean being 1.10506, .0000337. Hence the ratio 1 : 15.905, .0154. The first two hydrogen determinations were made with gas produced by the electrolysis of caustic potash, while the third sample was derived from zinc and sulphuric acid. The oxygen was electrolytic. Both gases were passed over red-hot platinum sponge, and dried by phosphorus pentoxide. Much more elaborate determinations of the two gaseous densities are those made by Morley. J For oxygen he gives three series of data ; two with oxygen from potassium chlorate, and one with gas partly from the same source and partly electrolytic. In the first series, temperature and pressure were measured with a mercurial thermometer and a mano- barometer. In the second series they were not determined for each experiment, but were fixed by comparison with a standard volume of hydrogen by means of a differential manometer. In the third series the gas was kept at the temperature of melting ice, and the mano-barometer * Proc. Amer. Acad., 24, 202. 1889. Also Am. Chem. Journ., 11, 509. fCompt. Rend., 113, 186. 1891. I Paper already cited, under the gravimetric portion of this chapter. 28 THE ATOMIC WEIGHTS. alone was read. The results for the weight in grammes, at latitude 45' of one litre of oxygen are as follows : First Series. Second Series. .42864 [.42952 .42849 .42900 .42838 .42863 .42900 .42853 .42907 .42858 .42887 .42873 .42871 .42913 .42872 .42905 .42883 .42896 .42880 Mean, .42875, .000051 .42874 Corrected,* .42879, zh .OOOO5I .42878 .42872 .42859 . ] .42851 Third Series. .42920 .42860 .42906 .42957 .42910 .42930 .42945 .42932 .42908 .42910 4295 1 .42933 .42905 .42914 .42849 .42894 .000048 .42886 000048 Mean, .42912, zfc .000048 Corrected, .42917, .000048 Mean, 1.42882, Corrected, 1.42887, General mean of all three series, 1.42896, .000028. Morley himself, for experimental reasons, prefers the last series, and gives it double weight, getting a mean density of 1.42900. The differ- ence between this mean and that given above is insignificant with ref- erence to the atomic weight problem. In the case of hydrogen, Morley 's determinations fall into two groups, but in both the gas was prepared by the electrolysis of pure dilute sul- phuric acid, and was most elaborately purified. In the first group there are two series of measurements. Of these, the first involved the reading of temperature and pressure by means of a mercurial thermometer and mano-barometer. In the second series, the gas was delivered into the weighing globes after occlusion in palladium ; it was then kept at the temperature of melting ice, and only the syphon barometer was read. In this group the hydrogen was possibly contaminated with mercurial vapor, and the results are discarded by Morley in his final summing up. For present purposes, however, it is unnecessary to reject them, for they have confirmatory value, and do not appreciably affect the final mean. The weight of one litre of hydrogen at 45 latitude, as found in these two sets of determinations, is as follows : * Correction applied by Morley to all his series, for a slight error, standard metre bar. , in the length of his OXYGEN. 29 First Series. Second Series. .089904 .089977 .089936 .089894 .089945 .089987 .089993 .089948 .089974 .089951 .089941 .089960 .089979 .090018 .089936 .089909 .089904 .089953 .089863 .089974 .089878 .089922 .089920 .090093 .089990 .090007 .089926 .089899 .089928 .089974 .089900 Mean, .089934, .000007 .089869 Corrected, .089938, .000007 .090144 .089984 Mean, .089967, .000011 Corrected, .089970, d= .000011 In the second group of experiments, the hydrogen was weighed in palladium before transfer to the calibrated globe ; and in weighing, the palladium tube was tared by a similar apparatus of nearly equal volume and weight. After transfer, which was effected without the intervention of stopcocks, the volume and pressure of the gas were taken at the temperature of melting ice. A preliminary set of measurements was made, followed by three regular series ; of these, the first and second were with the same apparatus, and are different only in point of time, a vacation falling between them. The last series was with a different apparatus. The data are as follows, with the means as usual : Preliminary. Third Series. Fourth Series. Fifth Series. .089946 .089874 .089972 .089861 .089915 .089891 .089877 .089877 .089881 .089886 .089867 .089870 .089901 .089866 .089916 .089867 .089945 .089911 .089770 .089839 .089856 .089846 .089874 Mean, .089918, .089912 .089864 .0000271 .089872 Mean, .089875, .089883 Corrected, .089921 =b .0000187 .089830 Mean, .089883, Corrected, .089880 .089877 .0000049 .089851 Corrected, .089886 Mean, .089863, rb .0000034 Corrected, .089866 30 THE ATOMIC WEIGHTS. Now, rejecting nothing, we may combine all the series into a general mean, giving the weight of one litre of hydrogen as follows : First series 089938, .000007 Second series 089970, .00001 1 Preliminary series, second method 089921, .0000271 Third series 089886, zfc .0000049 Fourth " 089880, .0000187 Fifth " 089866, .0000034 General mean 089897, zfc .0000025 Rejecting the first three 089872, .0000028 This last mean value for hydrogen will be used in succeeding chapters of this work for reducing volumes of the gas to weights. Combining the general mean of all with the value found for the weight of a litre of oxygen, 1.42896, .000028, we get for the ratio H : 0, O = I5 8955, .0005 If we take only the second mean for H, excluding the first three series, we have O = 15.9001, .0005 This value is undoubtedly nearest the truth, and is preferable to all other determinations of this ratio. Its probable error, however, is given too low ; for some of the oxygen weighings involved reductions for tem- perature and pressure. These reductions involve, again, the coefficient of expansion of the gas, and its probable error should be included. Since, however, that factor has been disregarded elsewhere, it would be an over- refinement of calculation to include it here. In a memoir of this kind it is impossible to do full justice to so elab- orate an investigation as that of Morley. The details are so numerous, the corrections so thorough, the methods for overcoming difficulties so ingenious, that many pages would be needed in order to present any- thing like a satisfactory abstract. Hardly more than the actual results can be cited here; for all else the original memoir must be consulted. Still more recently, by a novel method, J. Thomsen has measured the two densities in question.* In his gravimetric research, already cited, he ascertained the weights of hydrogen and of oxygen equivalent to a unit weight of aluminum. In his later paper he describes a method of measuring the corresponding volumes of both gases during the same reactions. Then, having already the weights of the gases, the volume- weight ratio, or density, is in each case easily computable. From 1.0171 to 2.3932 grammes of aluminum were used in each experiment. Omit- ting details, the volume of hydrogen in litres, equivalent to one gramme of the metal, is as follows : *Zeitschr. Anorg. Chern., 12, 4. 1896. OXYGEN. 31 .24297 243Q3 .24286 .24271 .24283 .24260 243*4 .24294 Mean, 1.24289, .00004 The weight of hydrogen evolved from one gramme of aluminum was found in Thomsen's gravimetric research to be 0.11190, zb .000015. Hence the weight of one litre at 0, 760 mm., and 10.6 meters above sea level at Copenhagen is : .090032, .000012; or at sea level in latitude 45, .089947, dh .000012 gramme. The data for oxygen are given in somewhat different form, namely, for the volume of one gramme of the gas at 0, 760, and at Copenhagen. The values are. in litres : .69902 .69923 .69912 .69917 .69903 .69900 .69901 .69921 .69901 .69922 Mean, .69910, .00002 At sea level in latitude 45, .69976, .00002 Hence one litre weighs 1.42906, .00004 grammes. Dividing this by the weight found for hydrogen, 0.089947, .000012 we have for the ratio H : 0, 15.8878, .0022. The density ratios, H : 0, now combine as follows : Dumas and Boussingault, corrected 15.9015, d= .031 Regnault, corrected 15.9105, =b .0044 Rayleigh, 1888 15.884, .0048 " 1892 15.882, .0023 Cooke , 15.890, .0067 Leduc i5-95 - OI 54 Morley, including all the data ., . . 15.8955, .0005 Thomsen 15.8878, .0022 General mean 15.8948, =h .00048 32 THE ATOMIC WEIGHTS. If we reject all of Morley's data for the density of hydrogen except his third, fourth, and fifth series, the mean becomes O = 15.8991, .00048. In either case Morley's data vastly outweigh all others. If oxygen and hydrogen were perfect gases, uniting by volume to form water exactly in the ratio of one to two, then the density of the first in terms of the second would also express its atomic weight. But in fact, the two gases vary from Boyle's law in opposite directions, and the true composition of water by volume diverges from the theoretical ratio to a measurable extent. Hence, in order to deduce the atomic weight of oxygen from its density, a small correction must be applied to the latter? dependent upon the amount of this divergence. Until recently, our knowledge of the volumetric composition of water rested entirely upon the determinations made by Humboldt and Gay-Lussac* early in this century, which gave a ratio between H and of a little less than 2:1, but their data need no farther consideration here. In 1887 Scott t published his first series of experiments, 21 in number, finding as the most probable result a value for the ratio of 1.994 : 1. In March, 1888, J he gave four more determinations, ranging from 1.9962 to 1.998:1; and later in the same year another four, with values from 1.995 to 2.001. In 1893, || however, by the use of improved apparatus, he was able to show that his previous work was vitiated by errors, and to give a series of measurements of far greater value. Of these, twelve were especially good, being made with hydrogen from palladium hydride, and with oxygen from silver oxide. In mean the value found is 2.00245, .00007, with a range from 2.0017 to 2.0030. In 1891 an elaborate paper by Morley^fl appeared, in which twenty concordant determinations of the volumetric ratio gave a mean value of 2.00023, .000015. These measurements were made in eudiometer tubes, and were afterwards practically discarded by the author. In his later and larger paper,** however, he redetermined the ratio from the density of the mixed electrolytic gases, and found it to be, after applying all corrections, 2.00274. The probable error, roughly estimated, is .00005. Morley also reduces Scott's determinations, which were made at the tem- perature of the laboratory, to 0, when the value becomes 2.00285. The mean value of both series may therefore be put at 2.0028, .00004, with sufficient accuracy for present purposes. Leduc's ft single determination, * Journ. de Phys., 60, 129. tProc. Roy. Soc., 42, 396. I Nature, 37, 439. g British Assoc. Report, 1888, 631. I! Proc. Roy. Soc., 53, 130. In full in Philosophical Transactions, 184, 543. 1893. ^ Amer. Journ. Sci. (3), 46, 220, and 276. ** Already cited with reference to syntheses of water. ft Compt. Rend., 115, 311. 1892. OXYGEN. 33 based upon the density of the mixed gases obtained by the electrolysis of water, gave 2.0037 ; but Morley shows that some corrections were neglected. This determination, therefore, may be left out of account. Now, including all data, we have a mean value for the density ratio : (A.) H :O: : I : 15.8948, .00048; or, omitting Morley's rejected series, (B.) H :O: : I : 15.8991, .00048. Correcting these by the volume ratio, 2.0028, .00004, the final result for the atomic weight of oxygen as determined by gaseous densities becomes : From A O 15.8726, =b .00058 From B O = 15.8769, .00058 Combining these with the result obtained from the syntheses of water, rejecting nothing, we have By synthesis of water O = 15.8837, .00026 By gaseous densities O = 15.8726, .00058 General mean O = 15.8821, .00024 If we reject Reiser's Work under the first heading, and omit Morley's defective hydrogen series under the second, we get By synthesis of water O 15.8796, .00027 By gaseous densities O = 15.8769, d= .00058 General mean O = 15.8794, .00025 Morley, discussing his own data, gets a final value of O = 15.8790, .00026, a result sensibly identical with the second of the means given above. These results cannot be far from the truth ; and accordingly, rounding off the last decimals, the value = 15.879, .0003, will be used in computation throughout this work. NOTE. A useful " short bibliography " upon the composition of water, by T. C. Warrington, may be found in the Chemical News, vol. 73, pp. 137, 145, 156, 170, and 184. 34 THE ATOMIC WEIGHTS. SILVER, POTASSIUM, SODIUM, CHLORINE, BROMINE, AND IODINE. The atomic weights of these six elements depend upon each other to so great an extent that they can hardly be considered independently. Indeed, chlorine, potassium, and silver have always been mutually de- termined. From the ratio between silver and chlorine, the ratio between silver and potassium chloride, and the composition of potassium chlo- rate, these three atomic weights were first accurately fixed. Similar ratios, more recently worked out by Stas and others, have rendered it desirable to include bromine, iodine, and sodium in the same general discussion. Several methods of determination will be left altogether out of account. For example, in 1842 Marignac* sought to fix the atomic weight of chlorine by estimating the quantity of water formed when hydrochloric acid gas is passed over heated oxide of copper. His results were wholly inaccurate, and need no further mention here. A little later Laurent f redetermined the same constant from the analysis of a chlorinated de- rivative of naphthalene. This method did not admit of extreme accu- racy, and it presupposed a knowledge, of the atomic weight of carbon ; hence it may be properly disregarded. Maumene's J analyses of the oxalate and acetate of silver gave good results for the atomic weight of that metal; but they also depend for their value upon our knowledge of carbon, and will, therefore, be discussed farther on with reference to that element. Hardin's work also, relating to the nitrate, acetate, and benzoate of silver, will be found in the chapters upon nitrogen and carbon. Let us now consider the ratios upon which we must rely for ascertain- ing the atomic weights of the six elements in question. After we have properly arranged our data we may then discuss their meaning. First in order we may conveniently take up the percentage of potassium chlo- ride obtainable from the chlorate. The first reliable series of experiments to determine this percentage was made by Berzelius. || All the earlier estimations were vitiated by the fact that when potassium chlorate is ignited under ordinary circum- stances a little solid material is mechanically carried away with the oxygen gas. Minute portions of the substance may even be actually volatilized. These sources of loss were avoided by Berzelius, who de- vised means for collecting and weighing this trace of potassium chloride. *Compt. Rend., 14, 570. Also, Journ. f. Prakt. Chetn., 26, 304. tConipt. Rend., 14, 456. Journ. f. Prakt. Chem., 26, 307. t Ann. d. Chim et d. Phys. (3), 18, 41. 1846. g Journ. Arner. Chem. Soc. 18, 990. 1896. j| Poggend. Annalen, 8, i. 1826. SILVER, POTASSIUM, ETC. 35 All the successors of Berzelius in this work have benefited by his exam- ple, although for the methods by which loss has been prevented we must refer to the original papers of the several investigators. In short, then, Berzelius ignited potassium chlorate, and determined the percentage of chloride which remained. Four experiments gave the following results : 60.854 60.850 60.850 60.851 Mean, 60.851, .0006 The next series was made by Penny,* in England, who worked after a somewhat different method. He treated potassium chlorate with strong hydrochloric acid in a weighed flask, evaporated to dryness over a sand bath, and then found the weight of the chloride thus obtained. His results are as follows, in six trials : 60.825 60.822 60.815 60.820 60.823 60.830 Mean, 60.8225, .0014 In 1842 Pelouze f made three estimations by the ignition of the chlo- rate, with these results : 60.843 60.857 60.830 Mean, 60.843, -53 Marignac, in 1842, J worked with several different recrystallizations of the commercial chlorate. He ignited the salt, with the usual precau- tions for collecting the material carried off mechanically, and also exam- ined the gas which was evolved. He found that the oxygen from 50 grammes of chlorate contained chlorine enough to form .003 gramme of silver chloride. Here are the percentages found by Marignac : In chlorate once crystallized 60.845 In chlorate once crystallized 60.835 In chlorate twice crystallized 60.833 In chlorate twice crystallized 60.844 In chlorate three times crystallized 60.839 In chlorate four times crystallized 60.839 Mean, 60.8392, .0013 * Phil. Transactions, 1839, p. 20. f Compt. Rend., 15, 959. I Ann. d. Chera. u. Pharm., 44, 18. 36 THE ATOMIC WEIGHTS. In the same paper Marignac describes a similar series of experiments made upon potassium perchlorate, KC10 4 . In three experiments it was found that the salt was not quite free from chlorate, and in three more it contained traces of iron. A single determination upon very pure material gave 46.187 per cent, of oxygen and 53.813 of residue. In 1845 two series of experiments were published by Gerhardt. * The first, made in the usual way, gave these results : 60.871 60.881 60.875 Mean, 60.8757, .0020 In the second series the oxygen was passed through a weighed tube containing moist cotton, and another filled with pumice stone and sul- phuric acid. Particles were thus collected which in the earlier series escaped. From these experiments we get 60.947 60.947 60.952 Mean, 60.9487, .0011 These last results were afterwards sharply criticised by Marignac,f and their value seriously questioned. The next series, in order of time, is due to Maumene.J This chemist supposed that particles of chlorate, mechanically carried away, might continue to exist as chlorate, undecomposed ; and hence that all previous series of experiments might give too high a value to the residual chloride. In his determinations, therefore, the ignition tube, after expulsion of the oxygen, was uniformly heated in all its parts. Here are his percentages of residue : 60.788 60.790 60.793 60.791 60.785 60.795 60.795 Mean, 60.791, .0009 The question which most naturally arises in connection with these re- sults is, whether portions of chloride may not have been volatilized, and css\ I /^a"f so lost * Compt. Rend., 21, 1280. } Supp. Bibl. Univ. de Geneve, Vol. I. I Ann. d. Chim. et d. Phys. (3), 18, 71. 1846. SILVER, POTASSIUM, ETC. 37 Closely following Maumene's paper, there is a short note by Faget,* giving certain mean results. According to this chemist, when potassium chlorate is ignited slowly, we get 60.847 per cent, of residue. When the ignition is rapid, we get 60.942. As no detailed experiments are given, these figures can have 110 part in our discussion. Last of all we have two series determined by Stas.f In the first series are the results obtained by igniting the chlorate. In the second series the chlorate was reduced by strong hydrochloric acid, after the method followed by Penny : First Series. 60.8380 60.8395 60.8440 60.8473 60.8450 Mean, 60.84276, dr .OOI2 Second Series. 6o.8t;o 60.853 60.844 Mean, 60.849, .0017 In these experiments every conceivable precaution was taken to avoid error and insure accuracy. All weighings were reduced to^ a vacuum standard ; from 70 to 142 grammes of chlorate were used in each experi- ment; and the chlorine carried away with the oxygen in the first series- was absorbed by finely divided silver and estimated. It is difficult to see how any error could have occurred. Now, to combine these different series of experiments. Berzelius, mean result 60. 85 1 , dr .0006 Penny, " 60.8225, dr .0014 Pelouze, " 60.843, .053 Marignac, " 60.8392, dr .0013 Gerhardt, 1st " 60.8757, dr .0020 " 2d V 60.9487, dr .0011 Maumene, " 60.791, dr .0009 .Stas, 1st " 60.8428, dr .0012 " 2d " 60.849, .0017 General mean from all nine series, representing forty experiments 60.846, db .00038 This value is exactly that which Stas deduced from both of his own series combined, and gives great emphasis to his wonderfully accurate * Ann. d. Chim. et d. Phys. (3), 18, 80. 1846. f See Aronstein's translation, p. 24Q. 38 THE ATOMIC WEIGHTS. work. It also finely illustrates the compensation of errors which occurs in combining the figures of different experimenters. Similar analyses of silver chlorate have been made by Marignac and by Stas. Marignac's data are as follows : * The third column gives the percentage of in AgC10 3 : 24.5 10 grin. AgClO 3 gave 18.3616 AgCl. 25 103 25.809 " 19-3345 " 25.086 30.306 22.7072 " 25.074 28.358 21.2453 " 25.082 . 28.287 " 21.1833 " 25.113 57.170 " 42.8366 " 25.072 Mean, 25.088, zfc .0044 Stas f found the following percentages in two experiments only : 25,081 25.078 Mean, 25.0795, H= .0010 Combined with Marignac's mean this gives a general mean of 25,080, .0010 ; that is, Marignac's series practically vanishes. For the direct ratio between silver and chlorine there are seven avail- able series of experiments. Here, as in many other ratios, the first reliable work was done by Berzelius. J He made three estimations, using each time twenty grammes of pure silver. This was dissolved in nitric acid. In the first experiment the silver chloride was precipitated and collected on a filter. In the second and third experiments the solution was mixed with h} T drochloric acid in a flask, evaporated to dry ness, and the residue then fused and weighed without transfer. One hundred parts of silver formed of chloride : 132.700 132.780 132.790 Mean, 132.757, .019 Turner's work closely resembles that of Berzelius. Silver was dis- solved in nitric acid and precipitated as chloride. In experiments one, two, and three the mixture was evaporated and the residue fused. In experiment four the chloride was collected on a filter. A fifth experi- ment was made, but has been rejected as worthless. The results were as follows : In a third column I put the quantity of AgCl proportional to 100 parts of Ag. *Bitjl. Univ. de Gen6ve, 46, 356. 1843. f Aronstein's translation, p. 214. I Thomson's Annals of Philosophy, 1820, v. 15, 89. g Phil. Transactions, 1829,291. SILVER, POTASSIUM, ETC. 39 28.407 grains Ag gave 37.737 41.917 " 55-678 40.006 " 53.143 30.922 " 41.070 132.844 132.829 '32.837 132.818 Mean, 132.832, ^ .0038 The same general method of dissolving silver in nitric acid, precipi- tating, evaporating, and fusing without transfer of material was also adopted by Penny. * His results for 100 parts of silver are as follows, in parts of chloride : 132.836 132.840 132.830 132.840 132.840 132.830 132.838 Mean, 132.8363, .0012 In 1842 Marignacf found that 100 parts of silver formed 132.74 of chloride, but gave no available details. Later, $ in another series of de- terminations, he is more explicit, and gives the following data. The weighings were reduced to a vacuum standard : 79.853 grm. Ag gave 106.080 AgCl. 69.905 " 92.864 " 64.905 " 86.210 " 92.362 " 122.693 " 99.653 " 132.383 " Ratio, 132.844 132-843 132.825 132.839 132.844 Mean, 132.839, .0024 The above series all represent the synthesis of silver chloride. Mau- mene made analyses of the compound, reducing it to metal in a current of hydrogen. His experiments make 100 parts of silver equivalent to chloride : 132.734 132-754 132.724 132.729 132.741 By Dumas Mean, 132.7364, =b .0077 we have the following estimations : 9.954 Ag gave 13.227 AgCl. Ratio, 132.882 19.976 26.542 132.869 Mean, 132.8755, .0044 *Phil. Transactions, 1839, 28. iAnn. Chetn. Pharm., 44, 21. I See Berzelius' I^ehrbuch, sth Ed., Vol. 3, pp. 1192, 1193. J Ann. d. Chim. et d. Phys. (3), 18, 49. 1846. || Ann. Chem. Pharm., 113, 21. 1860. 40 THE ATOMIC WEIGHTS. Finally, there are seven determinations by Stas,* made with his usual accuracy and with every precaution against error. In the first, second, and third, silver was heated in chlorine gas, and the synthesis of silver chloride thus effected directly. In the fourth and fifth silver was dis- solved in nitric acid, and the chloride thrown down by passing hydro- chloric acid gas over the surface of the solution. The whole was then evaporated in the same vessel, and the chloride fused, first in an atmos- phere of hydrochloric acid, and then in a stream of air. The sixth syn- thesis was similar to these, only the nitric solution was precipitated by hydrochloric acid in slight excess, and the chloride thrown down was washed by repeated decantation. All the decanted liquids were after- wards evaporated to dryness, and the trace of chloride thus recovered was estimated in addition to the main mass. The latter was fused in an atmosphere of HC1. The seventh experiment was like the sixth, only ammonium chloride was used instead of hydrochloric acid. From 98.3 to 399.7 grammes of silver were used in each experiment, the operations were performed chiefly in the dark, and all weighings were reduced to vacuum. In every case the chloride obtained was beautifully white. The following are the results in chloride for 100 of silver: 132.841 132.843 132.843 132.849 132.846 132.848 122.8417 Mean, 132.8445, .0008 We may now combine the means of these seven series, representing in all thirty-three experiments. One hundred parts of silver are equivalent to chlorine, as follows : Berzelius 3 2 -757, .0190 Turner 32.832, .0038 Penny 32.8363, .0012 Marignac , 32.839, =b .0024 Maumene ' 32.7364, .0077 Dumas 3 2 -8755, =t .0044 Stas 32.8445, dr .0008 General mean 32.8418, .0006 Here, again, we have a fine example of the evident compensation of errors among different series of experiments. We have also another tribute to the accuracy of Stas, since this general mean varies from the mean of his results only within the limits of his own variations. *Aronstein's translation, p. 171. SILVER, POTASSIUM, ETC. 41 The ratio between silver and potassium chloride, or, in other words, the weight of silver in nitric acid solution which can be precipitated by a known weight of KC1, has been fixed by Marignac and by Stas. Ma- rignac,* reducing all weighings to vacuum, obtained these results. In the third column I give the weight of KC1 proportional to 100 parts ofAg: i 4-7 2 3 g rm - Ag = 3.2626 KG. 69.067 22.725 " 15.001 " 69.050 21.759 " I 5- 2 8 " 69.066 21.909 " 15.131 " 69.063 22.032 " 15.216 " 69.063 25.122 " 17.350 " 69.063 Mean, 69.062, 0017 The work of Stas falls into several series, widely separated in point of time. His earlier experiments f upon this ratio may be divided into two sets, as follows : In the first set the silver was slightly impure, but the impurity was of known quantity, and corrections could therefore be applied. In the second series pure silver was employed. The potassium chloride was from several different sources, and in every case was puri- fied with the utmost care. From 10.3 to 32.4 grammes of silver were taken in each experiment, and the weighings were reduced to vacuum. The method of operation was, in brief, as follows : A definite weight of potassium chloride was taken, and the exact quantity of silver necessary, according to Prout's hypothesis, to balance it was also weighed out. The metal, with suitable precautions, was dissolved in nitric acid, and the solution mixed with, that of the chloride. After double decomposition the trifling excess of silver remaining in the liquid was determined by titration with a normal solution of potassium chloride. One hundred parts of silver required the following of KC1 : First Series. 69.105 69.104 69.103 69.104 69. IO2 Mean, 69.1036, d= .0003 Second Series. 69.105 69.099 69.107 69.103 69. 103 69.105 69.104 *See Berzelius' I^ehrbuch, sth Ed., Vol. 3, pp. 1192-3. fAronstein's translation, pp. 250-257. 42 THE ATOMIC WEIGHTS. 69.099 69.1034 69.104 69.103 69.102 69.104 69.104 69.105 69.103 69.101 69.105 Mean, 69.1033, =b .0003 In these determinations Stas did not take into account the slight solu- bility of precipitated silver chloride in the menstrua employed in the experiments. Accordingly, in 1882* he published a. new series, in which by two methods he remeasured the ratio, guarding against the indicated error, and finding the following values : 69.1198 69.11965 69.121 69.123 Mean, 69.1209, .0003 Corrected for a minute trace of silica contained in the potassium chloride, this mean becomes 69.11903, . 0003. f Still later, in order to establish the absolute constancy of the ratio in question, Stas made yet another series of determinations,^ in which he employed potassium chloride prepared from four different sources. One lot of silver was used throughout. The values obtained were as- follows : 69.1227 69.1236 69.1234 69.1244 69.1235 69.1228 69.1222 69.1211 69.1219 69.1249 69.1238 69.1225 69.1211 * Memoires Acad. Roy. de Beige, t. 43. 1882. fSee Van der Plaats, Ann. Chim. Phys. (6), 7, 15. I Oeuvres Posthumes, edited by W, Spring. SILVER, POTASSIUM, ETC. 43 A series was also begun in which one sample of potassium chloride was to be balanced against silver from various sources, but only one result is given, namely, 69.1240. This, with the previous series, gives a mean of 69.1230, .0002. Five series of determinations are now at hand for the ratio Ag : KC1. They combine as follows : Marignac 69.062, .0017 Stas, ist series 69. 1036, .0003 " 2d " 69.1033, .0003 " 3d " ...., 69. 1190, rb .0003 " 4th " 69.1230, .0002 General mean 69.1143, d= .00013 The difference between the highest and the lowest of Stas' series cor- responds to a difference of 0.021 in the atomic weight of potassium. The rejection of the earlier work might be quite justifiable, but would exert a very slight influence upon our final result. The quantity of silver chloride which can be formed from a known weight of potassium chloride has also been determined by Berzelius, Marignac and Maumene. Berzelius * found that 100 parts of KC1 were equivalent to 194.2 of AgCl ; a value which, corrected for weighings in air, becomes 192.32. This experiment will not be included in our dis- cussion. In 1842 Marignac f published two determinations, with these results from 100 KC1 : 192.33 192-34 Mean, corrected for weighing in air, 192.26, .003 In 1846 Marignac I published another set of results, as follows. The weighings were reduced to vacuum, The usual ratio is in the third column : 17.034 grm. KC1 gave 32.761 AgCl. 192.327 I4-427 27.749 " 192.341 15.028 " 28.910 " 192.374 15.131 29.102 " 192.334 15.216 " 29.271 " 192.370 Mean, 192.349, .006 Three estimations of the same ratio were also made by Maumene as follows : *Poggend. Annal., 8, i. 1826. f Ann. Chem. Pharm., 44, 21, 1842. t Berzelius' I^ehrbuch, sth E}d., Vol. 3, pp. 1192, 1193. Ann. d. Chim. et d. Phys. (3), 18, 41. 1846. 44 THE ATOMIC WEIGHTS. 10.700 grm. KC1 gave 20.627 AffCl. 192.776 10.5195 " 20.273 " 192.716 8.587 " 16.556 " 192.803 Mean, 192.765, .017 The three series of ten experiments in all foot up thus: Marignac, 1842 192.260, .003 1846 192.349, .006 Maumene 192 765, .017 General mean 192.294, .0029 These figures show clearly that the ratio which they represent is not of very high importance. It might be rejected altogether without im- propriety, and is only retained for the sake of completeness. It will obviously receive but little weight in our final discussion. In estimating the atomic weight of bromine the earlier experiments of Balard, Berzelius, Liebig, and Lowig may all be rejected. Their results were all far too low, probably because chlorine was present as an im- purity in the materials employed. Wallace's determinations, based upon the analysis of arsenic tribromide, are tolerably good, but need not be considered here. In the present state of our knowledge, Wallace's analyses are better fitted for fixing the atomic weight of arsenic, and will, therefore, be discussed with reference to that element. The ratios with which we now have to deal are closely similar to those involving chlorine. In the first place, there are the analyses of silver bromate by Stas.* In two careful experiments he found in this salt the following percentages of oxygen : 20.351 20.347 Mean, 20.349, .0014 There are also four analyses of potassium bromate by Marignac. f The salt was heated, and the percentage loss of oxygen determined. The residual bromide was feebly alkaline. We cannot place much reliance upon this series. The results are as follows : 28.7016 28.6496 28.6050 28.7460 Mean, 2^.6755, .0207 *Aronstein's translation, pp. 200-206. fSee E. Mulder's Overzigt, p. 117; or Berzelius' Jahresbericht, 24, 72. SILVER, POTASSIUM, ETC. 45 When silver bromide is heated in chlorine gas, silver chloride is formed. In 1860 Dumas* employed this method for estimating the atomic weight of bromine. His results are as follows. In the third column I give the weight of AgBr equivalent to 100 parts of AgCl : 2.028 grm. AgBr gave 1.547 AgCl. 131.092 4.237 " 3. 2 35 " i3 -974 5.769 4-403 " 131.024 Mean, 131.030, .023 This series is evidently of but little value. The two ratios upon which, in connection with Stas' analyses of silver bromate, the atomic weight of bromine chiefly depends, are those which connect silver with the latter element directly and silver with potassium bromide. Marignac,f to effect the synthesis of silver bromide, dissolved the metal in nitric acid, precipitated the solution with potassium bromide, washed, dried, fused, and weighed the product. The following quanti- ties of bromine were found proportional to 100 parts of silver : Mean, reduced to a vacuum standard, 74.077, dr .003 Much more elaborate determinations of this ratio are due to Stas.J In one experiment a known weight of silver was converted into nitrate, and precipitated in the same vessel by pure hydrobromic acid. The resulting bromide was washed thoroughly, dried, and weighed. In four other estimations the silver was converted into sulphate. Then a known quantity of pure bromine, as nearly as possible the exact amount neces- sary to precipitate the silver, was transformed into hydrobromic acid. This was added to the dilute solution of the sulphate, and, after precip- itation was complete, the minute trace of an excess of silver in the clear supernatant fluid was determined. All weighings were reduced to a vacuum. From these experiments, taking both series as one, we get the following quantities of bromine corresponding to 100 parts of silver: 74.0830 74.0790 74.0795 74.0805 74.0830 Mean, 74.081, db .0006 *Ann. Chem. Phartn., 113, 20. f E. Mulder's Overzigt, p. 116. Berzelius' Jahresbericht, 24, 7; I Aronstein's translation, pp. 154-170. 46 THE ATOMIC WEIGHTS. In his paper on the atomic weight of cadmium,* Huntington gives three syntheses and three analyses of silver bromide. The data are as follows, with the usual ratio given in the last column : 1.4852 grm. Ag gave 2.5855 AgBr. 74.084 1.4080 2.4510 " 74-077 1.4449 " 2.5150 " 74.060 4.1450 grm. AgBr gave 2.3817 Ag. 74-35 1.8172 " 1.0437 " 74-i" 4.9601 2.84 9 7 74.057 Mean, 74.071, .0072 Similar synthetic data are also given by Richards, incidentally to his work on copper.f There are two sets of three experiments each, which can here be treated as one series, thus : :.H235 grm. Ag gave 1.93630 AgBr. 74-73 2-74335 " 74-044 3.77170 " 74.076 " 1.68205 " 74.053 " 1.6789 " 74.069 1.6779 " 74-074 Mean, 74.065, .0035 Another set of data by Richards appears in his research upon the atomic weight of barium ; J in which BaBr 2 was balanced against silver, and the AgBr was also weighed. Richards gives from these data the percentage of Ag in AgBr, which figures are easily restated in the usual form as follows: Percentage. Ratio, 57.460 74.034 57-455 74.049 57-447 74 073 57-445 74-074 57.448 74-070 57.442 74-089 57.451 74.061 57-455 74-049 57-443 74.086 57-445 74-074 57-445 74-074 Mean, 74.067, rb .0034 The same ratio can also be computed indirectly from Cooke's experi- ments upon SbBr 3 , Huntington's on CdBr 2 , Thorpe's on TiBr 4 , and * Proc. Amer. Acad., 1881. fProc. Amer. Acad., 25, pp. 199, 210, 211. 1890. I Proc. Amer. Acad., vol. 28. 1893. SILVER, POTASSIUM, ETC. 47 Thorpe and Laurie's on gold. The values so obtained all confirm the results already given, varying within their limits, but having probable errors so high that their use would not affect the final mean. The latter is obtained as follows : Marignac 74.077, .0030 Stas 74.o8i, .0006 Huntington 74-O7 1 , =b .0072 Richards, 1st series 74.065, .0035 " 2d " 74.067,^.0034 General mean. ... 74.080, .00057 In this case again, as in so many others, Stas' work alone appears at the end, the remaining data having only corroborative value. The ratio between silver and potassium bromide was first accurately determined by Marignac.* I give, with his weighings, the quantity of KBr proportional to 100 parts of Ag : 2.131 grm. Ag = 2.351 KBr. 110.324 2.559 " 2.823 " 110.316 2.447 2.700 " 110.339 3.025 " 3.336 " 110.283 3-946 4.353 " 110.314 11.569 " 12.763 " 110.321 20.120 " 22.191 " 110.293 Mean, corrected for weighing in air, 110.343, , .005 Stas,f working in essentially the same manner, as when he fixed the ratio between potassium chloride and silver, obtained the following results : 110.361 110.360 110.360 110.342 110.346 110.338 110.360 110.336 110.344 110.332 110.343 110.357 110.334 "0.335 Mean, 110.3463, .0020 Combining this with Marignac's mean result, 110.343, .005, we get a general mean of 110.3459, .0019. *Berzelius' Jahresbericht, 24, 72. f- Aronstein's translation, pp. 334-347. 48 THE ATOMIC WEIGHTS. The ratios upon which we must depend for the atomic weight of iodine are exactly parallel to those used for the determination of bromine. To begin with, the percentage of oxygen in potassium iodate has been determined by Millon.* In three experiments he found : 22.46 22.49 22.47 Mean, 22.473, .5 Millon also estimated the oxygen in silver iodate, getting the follow- ing percentages : 17.05 17.03 17.06 Mean, 17.047, .005 The analysis of silver iodate has also been performed with extreme care by Stas.f From 76 to 157 grammes were used in each experiment, the weights being reduced to a vacuum standard. As the salt could not be prepared in an absolutely anhydrous condition, the water expelled in each analysis was accurately estimated and the necessary corrections ap- plied. In two of the experiments the iodate was decomposed by heat, and the oxygen given off was fixed upon a weighed quantity of copper heated to redness. Thus the actual weights, both of the oxygen and the residual iodide, were obtained. In a third experiment the iodate was reduced to iodide by a solution of sulphurous acid, and the oxygen was estimated only by difference. In the three percentages of oxygen given below, the result of this analysis conies last. The figures for oxygen are as follows : 16.976 16.972 16.9761 Mean, 16.9747, d= .0009 This, combined with Millon's series above cited, gives us a general mean of 16.9771, .0009. The ratio between silver and potassium iodide seems to have been de- termined only by Marignac.J and without remarkable accuracy. In five experiments 100 parts of silver were found equivalent to potassium iodide as follows : *Ann. Chim. Phys. (3), 9, 400. 1843. fAronstein's translation, pp. 170-200. I Berzelius' I^ehrbuch, 5th ed., 3, 1196. SILVER, POTASSIUM, ETC. 49 1.616 grm. Ag = 2.483X1. Ratio, 153.651 2.503 " 3.846 " " 153.665 3.427 5.268 " " 153.720 2.141 3.290 " " 153-667 10.821 16.642 " " 153.794 Mean, 153.6994, d= .0178 The synthesis of silver iodide has been effected by both Marignac and Stas. Marignac, in the paper above cited, gives these weighings. In the last column I add the ratio between iodine and 100 parts of silver: 15.000 grm. Ag gave 31.625 Agl. 117.500 14-79 " 3 2 .I70 " H7.5I 2 18.545 " 40.339 " H7.5I9 Mean, corrected for weighing in air, 117.5335, .0036 Stas* in his experiments worked after two methods, which gave, how- ever, results concordant with each other and with those of Marignac. In the first series of experiments Stas converted a known weight of silver into nitrate, and then precipitated with pure hydriodicacid. The iodide thus thrown down was washed, dried, and weighed without trans- fer. By this method 100 parts of silver were found to require of iodine : 117.529 117-536 Mean, 117.5325, .0024 In the second series a complete synthesis of silver iodide from known weights of iodine and metal was performed. The iodine was dissolved in a solution of ammonium sulphite, and thus converted into ammonium iodide. The silver was transformed into sulphate and the two solutions were mixed. When the precipitate of silver iodide was completely de- posited the supernatant liquid was titrated for the trifling excess of iodine which it always contained. As the two elements were weighed out in the ratio of 127 to 108, while the atomic weight of iodine is probably a little under 127, this excess is easily explained. From these experiments two sets of values were deduced ; one from the weights of silver and iodine actually employed, the other from the quantity of iodide of silver col- lected. From the first set we have of iodine for 100 parts of silver : "7-5390 117.5380 117- 53' 8 117.5420 117.5300 Mean, 117.5373, db .0015 : Aronstein's translation, pp. 136, 152. 50 THE ATOMIC WEIGHTS. From the weight of silver iodide actually collected we get as follows. For experiment number three in the above column there is no equivalent here: 117.529 117.531 117-539 117-538 ii7-53 Mean, H7-5334, d= .0014 Now, combining these several sets of results, we have the following general mean : Marignac H7-5335, .3 6 Stas, ist series ii7-53 2 5, - OO2 4 " 2d " "7-5373, .ooi5 " 3d " II7-5334, =t .0014 General mean "7-5345, .0009 One other comparatively unimportant iodine ratio remains for us to notice. Silver iodide, heated in a stream of chlorine, becomes converted into chloride ; and the ratio between these two salts has been thus deter- mined by Berzelius and by Dumas. From Berzelius * we have the following data. In the third column I give the ratio between Agl and 100 parts of AgCl : 5.000 grm. Agl gave 3.062 AgCl. 163.292 12.212 " 7-4755 " 163.360 Mean, 163.326, .023 Dumas' f results were as follows: 3.520 grm. Agl gave 2.149 AgCl. 163. 793 7.011 " 4.281 " 163.770 Mean, 163.782, .008 General mean from the combination of both series, 163.733, .0076. For sodium there are but four ratios of any value for present purposes. The early work of Berzelius we may disregard entirely, and confine ourselves to the consideration of the results obtained by" Penny, Pelouze, Dumas, and Stas, together with a single ratio measured incidentally by Earn say and Aston. The percentage of oxygen in sodium chlorate has been determined only by PennyJ, who used the same method which he applied to the potassium salt. Four experiments gave the following results : * Ann. Chim. Phys. (2), 40, 430. 1829. t Ann. Chem. Pharm., 113, 28. 1860. J Phil. Transactions, 1839, p. 25. SILVER, POTASSIUM, ETC. 51 Mean, 45.0705, d= .0029. The ratio between silver and sodium chloride has been fixed by Pe- louze, Dumas, and Stas. Pelouze * dissolved a weighed quantity of silver in nitric acid, and then titrated with sodium chloride. Equivalent to 100 parts of silver he found of chloride : 54.158 54.125 54.139 Mean, 54.141, .0063 By Dumas f we have seven experiments, with results as follows. The third column gives the ratio between 100 of silver and NaCl : 2.0535 grm. NaCl = 3-788 grm. Ag. 54-2H 2.169 4.0095 " 54.097 4-3554 8.0425 " 54.155 6.509 12.0140 " 54.178 6.413 11-8375 " 54.175 2.1746 4.012 " 54.202 5- "3 " 9-434 " 54.187 Mean, 54.172, .0096 Stas,J applying the method used in establishing the similar ratio for potassium chloride, and working with salt from six different sources, found of sodium chloride equivalent to 100 parts of silver : 54.2093 54.2088 54.2070 54-2070 54.2070 54.2060 54.2076 54.2081 54-2083 54.2089 Mean, 54.2078, .0002 As in the case of the corresponding ratio for potassium chloride, these data needed to be checked by others which took into account the solu- *Cotnpt. Rend., 20, 1047. 1845. t Ann. Chem. Pharm.. 113. 31. 1860. J Aronstein's translation, p. 274. 52 THE ATOMIC WEIGHTS. bility of silver chloride. Such data are given in Stas' paper of 1882,* and four results are as follows : 54.2065 54.20676 54.2091 54-2054 Mean, 54.20694, db .00045 Corrected for a trace of silica in the sodium chloride, this mean becomes 54.2046, it .O0045.t Combining all four series, we have for the NaCl equivalent to 100 parts of Ag Pelouze 54- HI, .0063 Dumas 54- 1 7 2 , .0096 Stas, early series 54.2078, .0002 Stas, late " 54.2046,^.00045 General mean 54.2071, .00018 Here the work of Stas is of such superior excellence that the other de- terminations might be completely rejected without appreciably affecting our final results. In their research upon the atomic weight of boron, Ramsay and Aston J converted borax into sodium chloride. In the latter the chlorine was afterwards estimated gravimetrically by weighing as silver chloride on a Gooch filter. Hence the ratio, AgCl : NaCl : : 100 : x, as follows : 3.0761 grm. NaCl gave 7.5259 AgCl. Ratio, 40.874 2.7700 6.7794 " " 40.859 2.8930 " 7.0804 " " 40-859 2.7360 " 6.6960 " 40.860 1.9187 " 46931 " " 40.863 Mean, 40.867, .0033 Finally, for the ratios between silver and sodium bromide we have one set of measurements by Stas. The bromide was prepared by saturating Na. 2 C0 3 with HBr. The NaBr proportional to 100 parts of silver was 95.4420 95-4383 95.4426 95-4392 Mean, 95.4405, .0007 We have now before us the data for computing, with greater or less accuracy, the atomic weights of the six elements under discussion. In *Mmoires Acad. Roy. de Beige., 43. 1882. fSee Van der Plaats, Ann. Chim. Phys. (6), 7, 16. 1886. % Chem. News, 66, 92. 1892. I Memoires Acad. Roy. Beige., 43. 1882. SILVER, POTASSIUM, ETC. 53 all there are nineteen ratios, involving about two hundred and fifty separate experiments. These ratios may now be tabulated and num- bered for reference, it being understood that the probable error in each case is that of the last term in the proportion. (i.) Percentage of O in KC1O 3 . . ... 39.154, .00038 (2.) " " KBrO 3 28.6755,^.0207 (3-) " KI O 3 22.473, .0050 (4.) NaClO 3 45.0705, .0029 (5.) AgClO 3 25.080, d= .0010 (6.) " " AgBrO s 20.349, .0014 (7-) " " AgI0 3 16.9771, .0009 (8.) Ag : NaCl : : ioo : 54.2071, .00018 (9.) Ag : NaBr : : 100 : 95.4405, .0007 (10.) Ag : KC1 : : 100 : 69.1143, .00013 (li.) Ag : KHr : : 100 : 110.3459, .0019 (12.) Ag : KI : : 100 : J53- 6 994, .0178 ( ! 3-) Ag : Cl : : loo : 32.8418, .0006 (14.) Ag : Br : : 100 : 74.080, .00057 (IS.) Ag : I : : ioo : 117.5345, .0009 (l6.) AgCl : NaCl : : IOO : 40.867, .0033 (17.) KC1 : AgCl : : ioo : 192.294, db .0029 (18.) AgCl : AgBr : : ioo : 131.030, .023 (19.) AgCl : Agl : : ioo : 163.733, .0076 Now, from ratios 1 to 7, inclusive, we can at once, by applying the known atomic weight of oxygen, deduce the molecular weights of seven haloid salts. Let us consider the first calculation somewhat in detail. Potassium chlorate yields 39.154 per cent, of oxygen and 60.846 per cent, of residual chloride. For each of these quantities the probable error is .00038. The atomic weight of oxygen is 15.879, dz .0003, so that the value for three atoms becomes 47.637, .0009. We have now the following simple proportion : 39.154 : 60.846 : : 47-637 : *, whence the molecular weight of potassium chloride becomes = 74.029. The probable error being known for the first, second, and third term of this proportion, we can easily find that of the fourth term by the formula given in our introduction. It is dz .0073. By this method we obtain the following series of values, which may conveniently be num- bered consecutively with the foregoing ratios : (20) KC1, from (i) = 74.029, .0073 (21) KBr, " (2) = 118.487, .0923 (22) KI, " (3) = 164.337, .0382 (23) NaCl, " (4) = 58.057, .0050 (24) AgCl, " (5) = 142.303, .0066 ( 2 5) AgBr, " (6) = 186.463, .0137 (26) Agl, " (7) =. 232.959, .0134 54 THE ATOMIC WEIGHTS. With the help of these molecular weights, we are now able to com- pute seven independent values for the atomic weight of silver. First, from (10) and (20) Ag 107.1 1 1, db .0106 Second, " (u) " (21) " = 107.378,^.0837 Third, " (12) " (22) " = 106.921,^.0278 Fourth, " ( 8 ) " (23) " = 107. 102, .0092 Fifth, " (13) " (24) " = 107.122, .0050 Sixth, " (14) " (25) " =107.113, dr. 0079 Seventh, " (15) " (26) " = 107.091, dr .0062 General mean Ag = 107. 108, dr .0031 It is noticeable that five of these values agree very well. The second and third, however, diverge widely from the average, but in opposite directions ; they have, moreover, high probable errors, and consequently little weight. Of these two, one represents little and the other none of Stas' work. Their trifling influence upon our final results becomes curiously apparent in the series of silver values given a little further along. When we consider closely, in all of its bearings, any one of the values just given, we shall see that for certain purposes it must be excluded from our general mean. For example, the first is derived partly from the ratio between silver and potassium chloride. From this ratio, the atomic weight of one substance being known, we can deduce that of the other. We have already used it in ascertaining the atomic weight of silver, and the value thus obtained is included in our general mean. But if from it we are to determine the molecular weight of potassium chloride, we must use a silver value derived from other sources only, or we should be assuming a part of our result in advance. In other words, we must now use a general mean for silver from which this ratio with reference to silver has been rejected. Hence the following series of silver values, which are lettered for reference : A. General mean from all eight 107.108, dr .0031 B. " excluding the first 107.108, dr .0032 C. " " second 107.107, .0031 D. " third 107.1 IO, rfc .0031 E. " " fourth 107. 109, dr .0033 F. " " fifth 107.099, dr .0039 G. " sixth 107.106, dr .0034 H. " seventh .... 107.113, dr .0036 We are now in a position to determine more closely the molecular weights of the haloid salts which we have already been considering. For silver chloride, still employing the formula for the probable error of the last term of a proportion, we get the following values : SILVER, POTASSIUM, ETC. 55 From (5) AgCl 142.303, .0066 From (13) and (F) " = 142.276, .0052 From ( 1 6) " (23) " == 142.063, .0168 From (17) " (20) " = 142.353,^.0156 From ( 1 8) " (25) " = 142.306, =b .0271 From (19) " (26) " = 142.278, =b .0105 General mean AgCl = 142. 277, .0036 The third of these values is certainly too low, and although it reduces the atomic weight of chlorine by only 0.01, it ought to be rejected. The general mean of the other five values is AgCl = 142.287, .0037. Sub- tracting from this the atomic weight of silver, 107.108, .0031, we have for the atomic weight of chlorine 1 = 35.179, .0048. For silver bromide three ratios are available: From (6) AgBr = 186.463, dr .0137 From (14) and (G) " = 186.450, .0050 From ( 1 8) " (24) " = 186.459,^.0339 General mean AgBr= 186.452, .0054 Hence, applying the atomic weight of silver as before Br = 79.344, d= .0062. For silver iodide we have From (7) ' Agl = 232.950, rh .0134 From (15) and (H) . " = 233.008, .0079 From (19) " (24) " =^232.997,^.0153 General mean Agl = 232.996, rb .0062 Hence, 1= 125.888, rb .0069. For the molecular weight of sodium chloride three values appear, as follows : From (4) NaCl = 58.057, .0050 From (8) and (E) " = 58.061, .0018 From (16) " AgCl " := 58.148, .0049 General mean NaCl = 58.069, rh .0016 Rejecting the third value, which corresponds to the rejected value for AgCl and throws out ratio (16) entirely, the mean becomes NaCl = 58.060, dz .0017 From (9) and (A) NaBr = 102.224, .0031 56 , THE ATOMIC WEIGHTS. Deducting from these molecular weights the values already found for Cl and Br,two measurements of the atomic weight of sodium are obtained, thus: From NaCl Na = 22.881, .0051 FromNaBr.. . " = 22.880, .01 12 General mean Na = 22.881, 0046 The rejection of ratio (16) in connection with the atomic weights of sodium and chlorine is fully justified by the fact that the data which it represents were never intended for use in such computations. They were obtained incidentally in connection with work upon boron, and their consideration here may have some bearing later upon the discussion of the last-named element. For potassium, the ratios available give molecular weights for the chloride, bromide, and iodide. For the chloride, From (i) KC1 = 74.029, db .0073 From ( 10) and (B) " = 74.027,^.0022 From (17) " (24) " = 74.003, .0049 General mean KC1 = 74.025, d= .0019 For the bromide we have From (2) KBr = 118.487, .0923 From (n ) and (C) " = 118.188, .0073 General mean -. . . KBr = 118.200, .0073 And for the iodide ( From (3) KI = 164.337, .0382 From (12) and (D) " = 164.627, =!= .0052 General mean KI = 164.622, .0051 Combining these values with those found for chlorine, bromine, and iodine, we have three values for the atomic weight of potassium, as fol- lows : From KC1 K = 38.846, .0078 From KBr "= 38.856, .0096 From KI " =38.734, .0086 General mean K = 38.817, .0051 To sum up, the six atomic weights, under discussion may be tabulated as follows, both for the standard chosen, and with O = 16 as the base of the system : SILVER, POTASSIUM, ETC. t 57 H=i. <9=i6. Ag , 107.108, .0031 107.924 K. 38.817,^.0051 39.112 Na 22.881, .0046 23.048 Cl 35.179, .0048 35-447 Br 79.344,^.0062 79-949 I 125.888,^.0069 126.847 It must be remembered that tbese values represent the summing up of work done by many investigators. Stas' ratios, taken by themselves, give various results, according to the method of combining them. This computation has been made by Stas himself, with his older determina- tions, and more recently by Ostwald,* Van der Plaats,f and Thomsen, J all with the standard of 16. By Van der Plaats two sets of results are given : one with Stas' ratios assigned equal weight (A), and the other with each ratio given weight inversely proportional to the square of its mean error (B). The results of these several computations may well be tabulated in comparison with the values obtained in my own general discussion, thus : Clarke. Stas. Ostwald. V. der P., A. V.derP.,B. Thomsen. Ag 107.924 107.930 107.9376 107.9202 107.9244 107.9299 39-H2 39-*37 39- I 3 61 39- T 4i4 39-HO3, 39- I 57 23.048 23.043 23.0575 23.0453 23.0443 ' 23.0543 d 35-447 35-457 35-45 2 9 35-45 16 35-45 6 5 35-4494 Br 79-949 79 95 2 79-96^8 79-94Q7 79-9548 79.95 10 I 126.847 126.850 126.8640 126.8445 126.8494 126.8556 The agreement between the new values and the others is highly satis- factory, and gives a strong emphasis to the magnificent accuracy of Stas' determinations. No severer test could be applied to them. *Lehrbuch der allgemeinen Chemie, i, 41. 1885. tCompt. Rend., 116, 1362. 1893. t Zeitsch. Physikal. Chem., 13, 726. 1894. 58 THE ATOMIC WEIGHTS. NITROGEN. The atomic weight of nitrogen has been determined from the density of the gas, and from a considerable variety of purely chemical ratios. Upon the density of nitrogen a great many experiments have been made. In early times this constant was determined by Biot and Arago, Thomson, Dulong and Berzelius, Lavoisier, and others. But all of these investigations may be disregarded as of insufficient accuracy ; and, as in the case of oxygen, we need consider only the results obtained by Dumas and Boussingault, by Regnault, and by recent investigators. Taking air as unity, Dumas and Boussingault* found the density of nitrogen to be .970 .972 974 Mean, .972, .00078 For hydrogen, as was seen in our discussion of the atomic weight of oxygen, the same investigators found a mean of .0693, .00013. Upon combining this with the above nitrogen mean, we find for the atomic weight of the latter element, N = 14.026, .0295. By Regnault f much closer work was done. He found the density of nitrogen to be as follows : .97148 .97H8 97154 .97155 .97108 .97108 Mean, .97137, d= .000062 For hydrogen, Regnault's mean value is .069263, .000019. Hence, combining as before, N = 14.0244 .0039. Both of the preceding values are affected by a correction for the dif- ference in volume between the weighing globes when full and when empty. This correction, in the case of Regnault's data, has been meas- ured by Crafts,J who gives .06949 for the density of H, and .97138 for N. Corrected ratio, N = 13.9787. If we assume the same proportional cor- rection for the determination by Dumas and Boussingault, that becomes N = 13.9771. *Compt. Rend., 12, 1005. 1841. f Compt. Rend., 20, 975. 1845. I Compt. Rend., 106, 1664. NITROGEN. 59 Von Jolly,* working with electrolytic oxygen and with nitrogen pre- pared by passing air over hot copper, but not with hydrogen, compared the weights of equal volumes of the two gases, with results as follows : Oxygen. Nitrogen. .442470 1.269609 .442579 .269389 .442489 .269307 .442570 .269449 442571 .269515 .442562 .269443 .442478 . .269478 Mean, 1.442545, .000013 Mean, 1.269455, . 000024 The ratio, when O = 16, is N = 14.0802, .0003. Corrected by Ray- leigh, the ratio between the weights becomes 14.0805. If = 15.879, dz .0003, the final value for N, deducible from Von Jolly's data, is N = 13.974, .0004. The next determination in order of time is Leduc's.f He made nine measurements of the density of nitrogen, giving a mean of .97203, with extremes of .9719 and .9721; but he neglects to cite the intermediate values. Taking the three figures given as representative, and assuming a fair distribution of the other values between the indicated limits, the probable error of the mean is not far from 0.00002. For hydrogen he found .06948, .00006745. The ratio between the two densities gives N = 13.9901, .0138. Lord Rayleigh,^ preparing nitrogen by passing air over hot copper, and weighing in a standard globe, obtained the following weights : 2.31035 2.31026 2.31024 2.31012 2.31027 Mean, 2.31025, 000025 With corrections for temperature, shrinkage of the globe when ex- hausted, etc., this becomes 2.30883, as against 2.37512 for the same volume of air. Hence the density of N = .97209, .00001. His former work on hydrogen gives .06960, .0000084, for the density of that gas. The ratio is N = 13.9678, .0017. The foregoing data, however, all apply to nitrogen derived from the atmosphere. In a later memoir Rayleigh found that nitrogen from * Poggend. Annalen (2), 6, 529-530. 1879. fCompt. Rend., 113, 186. 1891. j Proc. Roy. Soc., 53, 134. 1894. I Chem. News, 69, 231. 1894. 60 THE ATOMIC WEIGHTS. chemical sources, such as oxides of nitrogen, ammonium nitrate, etc., was perceptibly lighter ; and not long afterwards the discrepancy was explained by the astonishing discovery of argon. The densities given, therefore, are all too high, and unavailable for any discussion of atomic weight. As, however, the reductions had been completed in nearly all . their details before the existence of argon was announced, they may be allowed to remain here as part of the record. Summing up, the ratios found between hydrogen and atmospheric u nitrogen " are as follows : Dumas and Boussingault, corrected 1 3.977 Regnault, " 13-979 Von Jolly, " ij-974 Leduc, " 13.990 Rayleigh, " 13.968 Perhaps at some future time, when the density of argon is accurately known and its amount in the atmosphere has been precisely determined, these figures may be so corrected as to be useful for atomic weight calcu- lations. In discussing the more purely chemical ratios for establishing the atomic weight of nitrogen, we may ignore, for the present, the researches of Berzelius and of Anderson. These chemists experimented chiefly upon lead nitrate, and their work is consequently now of greater value for fixing the atomic weight of lead. Their results will be duly consid- ered in the proper connection further on. The ratio between ammonium chloride and silver has been determined by Pelouze, by Marignac, and by Stas. The method of working is essen- tially that adopted in the similar experiments with the chlorides of sodium and potassium. For the ammonium chloride equivalent to 100 parts of silver, Pelouze* found : 49-556 49-5<7 Mean, 49.5365, .013 Marignac f obtained the following results. The usual ratio for 100 parts of silver is given also : 8.063 g rm - Ag = 3.992 grm. NH 4 C1. 49.510 9.402 4-656 " 49-521 10.339 " 5.120 " 49-521 12.497 " 6.191 " 49.540 "337 " 5-6i7 " 49.546 11.307 5-595 49-483 4.326 2.143 49.538 Mean, 49.523, -0055 *Compt. Rend., 20. 1047. 1845. t Berzelius' Lehrbuch, sth ed., vol. 3, 1184, 1185. NITROGEN. 61 But neither of these series can for a moment compare with that of Stas. * He used from 12.5 to 80 grammes of silver in each experiment^ reduced his weighings to a vacuum standard, and adopted a great variety of precautions to insure accuracy. He found for every 100 parts of silver the following quantities of NH 4 C1 : V 49.600 49.599 49-597 49.598 49-597 49-593 49-597 49-5974 49.602 49-597 49598 49-592 Mean, 49-5973, .0005 In this work, as with the similar ratios for potassium and sodium chloride, the solubility of silver chloride was not guarded against so fully as is needful. Accordingly Stas published a new series of determina- tions in 1882,f carefully checked in this particular, with the subjoined values for the ratio : 49.60001 49-59999 49-599 49.600 49.597 Mean, 49-S99 2 , .00039 Combining all four series, we have Pelouze 49-5365, =b .013 Marignac 49-5 2 3> -OQ55 Stas, early series 49'5973, d= .0005 Stas, later " 49.5992, .00039 General mean 49-5983, .00031 In the paper last cited Stas also gives a similar series of determinations for the ratio Ag : NH 4 Br : : 100 : x. The results are as follows, with re- duction to vacuum : * Aronstein's translation, pp. 56-58. fMemoires Acad. Roy. de Beige., 43. 1882. 62 THE ATOMIC WEIGHTS, 90.831 90.831 90.8297 90.823 90.8317 90.8311 90.832 Mean, 90.8299, .0008 The quantity of silver nitrate which can be formed from a known weight of metallic silver has been determined by Penny, by Marignac, and by Stas. Penny * dissolved silver in nitric acid in a flask, evapo- rated to dryness without transfer, and weighed. One hundred parts of silver thus gave of nitrate : 157.430 157-437 157-458 157.440 157.43 157-455 Mean, 157.4417, .0033 Marignac'sf results were as follows. In the third column they are reduced to the common standard of 100 parts of silver : 68.987 grm. Ag gave 108.608 grm. AgNO 3 . 1 57. 433 57.844 " 9 I -47 I57.40I 66.436 " 104.592 " 157.433 70.340 110.718 157.404 200.000 " 3*4.894 " 157.447 Mean, 157.4236, .0061 Stas,t employing from 77 to 405 grammes of silver in each experiment, made two different series of determinations at two different times. The silver was dissolved with all the usual precautions against loss and against impurity, and the resulting nitrate was weighed, first after long drying without fusion, just below its melting point ; and again, fused. Between the fused and the unfused salt there was in every case a slight difference in weight, the latter giving a maximum and the former a minimum value. In Stas' first series there are eight experiments; but the seventh he himself rejects as inexact. The values obtained for the nitrate from 100 * Phil. Trans., 1839. fBerzelius' I^ehrbuch, sth ed., 3, pp. 1184, 1185. t Aronstein's translation, pp. 305 and 315. NITKOGEN. 63 parts of silver are given below in two columns, representing the two con- ditions in which the salt was weighed. The general mean given at the end I have deduced from the means of the two columns considered separately : Unfused. Fused. IS7-492 157.474 157-510 157.481 157-485 157-477 157.476 i57-47i 157.478 157-47 T57.47I 157.463 157.488 157-469 Mean, 157.4857 Mean, 157.472 General mean, 157.474, .0014 In the later series there are but two experiments, as follows : Unfused. Fused. 157.4964 I57-488 157.4940 i57-48o Mean, 157.4952 Mean, 157.484 General mean, 157.486, .0003 The reverse ratio, namely, the amount of silver obtainable from a weighed quantity of nitrate, has been determined electrolytically by Hardin.* The data obtained, however, are reducible to the same form as in the preceding series, and all are properly combinable together. Pure silver was dissolved in pure aqueous nitric acid, and the crystal- line salt thus formed was dried, fused, and used for the determinations. The silver nitrate, mixed with an excess of pure potassium cyanide solu- tion, was electrolyzed in a platinum dish. The results obtained, reduced to vacuum weights, were as follows : .31202 AgNO 3 gave .19812 Ag. Ratio, 157.490 .47832 .30370 " 157.498 .56742 .36030 " " 157.485 .57728 .36655 " " 157.490 .69409 .44075 " " 157.479 .86367 .54843 " " 157.479 .868u " -SS^o " " 157.466 .93716 .59508 " 157.485 1.06170 .67412 " " 157.494 i 1.19849 " .76104 " J< 157-477 Mean, 157.484, .0020 * Journ. Amer. Chem. Soc., 18,995. 1896. 64 THE ATOMIC WEIGHTS. Now, to combine all five sets of results : Penny 157-4417, -33 Marignac 1 57-4236, .0061 Stas, ist series 157.4740, .0014 Stas, 2d " 157.4860, =h .0003 Hardin 157.484, .0020 General mean 157-479, .0003 For the direct ratio between silver nitrate and silver chloride there are two series of estimations. A weighed quantity of nitrate is easily con- verted into chloride, and the weight of the latter ascertained. In two experiments Turner* found of chloride from 100 parts of nitrate : 84-357 84.389 Mean, 84.373, i.on Penny ,t in five determinations, found the following percentages: 84-370 84.388 84.377 84.367 84-370 Mean, 84.3744, d= .0025 The general mean from both series is 84.3743, .0025. The ratio directly connecting silver nitrate with ammonium chloride has been determined only by Stas. J The usual method of working was followed, namely, nearly equivalent quantities of the two salts were weighed out, the solutions mixed, and the slight excess of one estimated by titration. In four experiments 100 parts of silver nitrate were found equivalent to chloride of ammonium, as follows: 3L489 3L490 31-487 31.486 Mean, 31.488, .0006 I The similar ratio between potassium chloride and silver nitrate- has been determined by both Marignac and Stas. *Phil. Trans., 1833, 537. fPhil. Trans., 1839. jAronstein's translation, p. 309. NITROGEN. 65 Marignac* gives the following weights. I add the quantity of KC1 proportional to 100 parts of AgN0 3 : , 1.849 grm. KC1 4.218 grm. AgNO 3 . 43.836 2.473 " 5.640 " 43-848 3-3I7 7.565 43.847 2.926 " 6.670 " 43.868 6.191 " 14.110 " 43.877 4.351 " 9.918 " 43-870 Mean, 43.858, .0044 Stas' f results are given in three series, representing silver nitrate from three different sources. In the third series the nitrate was weighed in vacuo, while for the other series this correction was applied in the usual way. For the KC1 equivalent to 100 parts of AgN0 3 Stas found : First Series. 43-878 43.875 43-875 43-874 Mean, 43.8755, =h .0005. Second Series. 43-864 43.869 43-876 Mean, 43.8697, .0023 Third Series. 43-894 43-878 43.885 Mean, 43.8857, .0031 i Combining all four series we have : Marignac 43.858, .0044 Stas, ist series 43-8755, rfc .o5 Stas, 2d " 43.8697, .0023 Stas, 3d " 43-8857, .0031 General mean 43.8715, =h .0004 There have also been determined by Penny, by Stas, and by Hibbs a series of ratios connecting the alkaline chlorides and chlorates with the corresponding nitrates. One of these, relating to the lithium salts, will be studied farther on with reference to that metal. *Berzelius' L,e'urbuch, sth ed., 3d vol., 1184, 1185. t Aronstein's translation, p. 308. 66 THE ATOMIC WEIGHTS. The general method of working upon these ratios is due to Penny. * Applied to the ratio between the chloride and nitrate of potassium, it is as follows : A weighed quantity of the chloride is introduced into a flask which is placed upon its side and connected with a receiver. An excess of pure nitric acid is added, and the transformation is gradually brought about by the aid of heat. Then, upon evaporating to dryness over a sand bath, the nitrate is brought into weighable form. The liquid in the receiver is also evaporated, and the trace of solid matter which had been mechanically carried over is recovered and also taken into account. In another series of experiments the nitrate was taken, and by pure hy- drochloric acid converted into chloride, the process being the same. In the following columns of figures I have reduced both series to one stand- ard, namely, so as to express the number of parts of nitrate correspond- ing to 100 of chloride : First Series. KCl treated with !35- 6 39 I35-637 135-640 135.635 135-630 135.640 135-630 Mean, 135.636, .0011 Second Series. KNO^ treated with HCl. 135.628 135-635 135-630 135-641 135 630 135.635 135-630 Mean, 135.633, .0011 Stas' f results are as follows : 135.643 135-638 135.647 135-649 135.640 1 35 -645 135.655 Mean, 135.6453, .0014 *Phil. Trans., 1839. t Aronstein's translation, p. 270. NITROGEN. 67 These figures by Stas represent weighings in the air. Reduced to a vacuum standard, this mean becomes 135.6423. The determinations made by Hibbs* differ slightly in method from those of Penny and Stas. He converted the nitrate into the chloride by heating in a stream of gaseous hydrochloric acid. His results were as follows, vacuum weights being given Weight KNO Z Weight KCl. .11090 .08177 .14871 .10965 .21067 .15533 .23360 .17225 .24284 .17903 Now, combining, we have : Ratio. 135-624 135.622 135.627 135.620 135.642 Mean, 135.627, =h .0026 Penny, ist series J35-636, .001 1 Penny, 2d " i35- 6 33> .0011 Stas I35. 6 423, .0014 Hibbs 135.627, .0026 General mean 135.636, .0007 By the same general process Penny f determined how much potassium nitrate could be formed from 100 parts of chlorate. He found as follows : 82.505 82.497 82.498 82.500 Mean, 82.500, .0012 For 100 parts of sodium chlorate he found of nitrate : 79.875 79-882 79.890 Mean, 79.8823, 4= .0029 For the ratio between the chloride and nitrate of sodium Penny made two sets of estimations, as in the case of potassium salts. The subjoined figures give the amount of nitrate equivalent to 100 parts of chloride : * Thesis for Doctor's degree, University of Pennsylvania, 1896. Work done under the direction of Professor E. F. Smith. fPhil. Trans., 1839. 68 THE ATOMIC WEIGHTS. First Series. NaCl treated with I45-4T5 145.408 145.420 145.424 145.410 145.418 145.420 Mean, 145.4164, .0015 Second Series. NaNO z treated with HCL I45-4I9 I45-39 1 145.412 145.415 145-412 145.412 Mean, 145.410, .0026 Stas* gives the following series : 145-453 145.468 145-465 145.469 145-443 Mean, after reducing to vacuum standard, 145.4526, .0030 Hibbs't data, obtained by the method employed in the case of the potassium compounds, are as follows, vacuum weights being stated : Weight NaNOy Weight NaCl. Ratio. .01550 .01066 i45-43 .20976 .14426 I45.4 4 .26229 .18038 145. 410 .66645 .458 2 9 145.429 .93718 .64456 H5-399 Mean, 145.407, .0026 Combining, we have as follows : Penny, 1st series 145.4164, .0015 Penny, 2d " 145.410, .0026 Stas 145.4526, .0030 Hibbs 145.407, : .0026 General mean 145.418, .0012 * Aronstein's translation, p. 278. t Thesis, University of Pennsylvania, 1896. NITROGEN. 69 Julius Thomsen, * for the purpose of fixing indirectly the ratio H : O, has made a valuable series of determinations of the ratio HC1:NH 3 , which may properly be used toward establishing the atomic weight of nitrogen. First, pure, dry, gaseous hydrochloric acid is passed into a weighed absorption apparatus containing pure distilled water. After noting the increase in weight, pure ammonia gas is passed in until a very slight excess is present, and the apparatus is weighed again. The excess of NH 3 , which is always minute, is measured by titration with standard hydrochloric acid. In weighing, the apparatus is tared by one of similar form, arid containing about the same amount of water. Three series of determinations were made, differing only in the size of the absorption apparatus ; so that for present purposes the three may be taken as one. Thomsen considers them separately, and so gives greatest weight to the ex- periments involving the largest masses of material. I give his weighings, TT/tj and also, as computed by him, the ratio ^T. First series. . Second series. HCl. Nt? Ratio. 5.1624 2.4120 2.1403 39425 1.8409 2.1416 4.6544 2.1739 2.1411 3.9840 1.8609 2.1409 5.3295 2.4898 2.1406 4-2517 1.9863 2.1405 4.8287 2.2550 2.1414 6.4377 3.0068 2.1411 4.1804 1.9528 2.1407 5-3 6 3 2.35 2 3 2.1410 4.6408 2.1685 2.1411 11.8418 5-5302 2.14130 14.3018 6.6808 2.14073 12.1502 5.6759 2.14067 H-5443 5.3927 2.14073 12.3617 5-7733 2.14118 19-3455 9.0360 2.14094 19.4578 9.0890 2.14081 Third series.. Mean of all, 2.14093, .000053 Reduced to vacuo, 2.1394 From the sums of the weights Thomsen finds the ratio to be 2.14087, or 2.13934 in vacuo. From this, using Ostwald's reductions of Stas' data for the atomic weights of N and Cl, he finds the atomic weight of H = 0.99946, when O == 16. We have now, apart from the determinations of gaseous density, eleven ratios, representing one hundred and sixty-four experiments, from which * Zeitsch. Physikal. Chem., 13, 398. 1894. 70 THE ATOMIC WEIGHTS. to calculate the atomic weight of nitrogen. Let us first collect and num- ber these ratios : (i.) Ag : AgNO 3 : : ioo : 157-479, .o3 (2.) AgNO 3 : AgCl : : ioo : 84-3743, - OO2 5 (3.) AgNO 3 : KC1 : : ioo : 43- 8 7i5> .0004 (4.) AgNO 3 : NH 4 C1 : : ioo : 31.488, .0006 (5.) Ag : NH 4 C1 : : ioo : 49.5983, dr .00031 (6.) Ag : NH 4 Br : : ioo : 90.8299, .0008 (7.) KC1 : KNO 3 : : ioo : 135.636, .0007 (8.) KC1O 3 : KN0 3 : : ioo : 82.500, .0012 (9.) NaCl : NaNO 3 : : ioo : 145 418, .001 1 (10.) NaClO 3 : NaNO 3 : : ioo : 79.8823, .0029 (n.) NH 3 : HC1 : : i.oo : 2.1394, d= .000053 From these ratios we are now able to deduce the molecular weight of ammonium chloride, ammonium bromide, and three nitrates. For these calculations we must use the already ascertained atomic weights of oxy- gen, silver, chlorine, bromine, sodium and potassium, and the molecular weights of sodium chloride, potassium chloride, and silver chloride. The following are the antecedent values to be employed : Ag = 107.108, d= .0031 K = 38.817, =b .0051 Na 22.881, .0046 Cl = 35.179, =b .0048 Br = 79.344, .0062 O 3 = 47.637, -0009 AgCl 142.287, .0037 KC1 = 74.025, .0019 NaCl = 58.060, .0017 Now, from ratio number five we get the molecular weight of NH 4 C1 = 53.124, .0016, and N = 13.945, .0051. From ratio number six, NH 4 Br = 97.286, .0029, and N = 13.942, .0077. From ratio number eleven, NH 3 = 16.911, .0048, and N = 13.911, .0048. From ratio number four, which involves an expression of the type A : B : : C + x : D + x, an independent value is deducible, N = 13.935, .0073. For the molecular weight of silver nitrate there are three values, namely : From (i) AgNO 3 == 168.673, .0049 From (2) " = 168.634, .0066 From (3) " = 168.731, .0046 General mean A gNO 3 = 168.690, .0030 Hence N= 13.945, .0044. NITROGEN. 71 The molecular weight of potassium nitrate is twice calculable, as follows : From (7) KNO 3 = 100.405, .0026 From (8) " 100.371, .0059 General mean. . KNO 3 = 100.401, .0024 Hence N = 13.947, .0057. And for sodium nitrate we have : From (9) NaNO 3 = 84.430, .0026 From (.10) " = 84.433, .0053 General mean NaNO 3 = 84.431, .0023 Hence N = 13.913, .0052. There are now seven estimates of the atomic weight of nitrogen, to be combined by means of the usual formula. 1. From NH 4 C1 N = 13.945, .0051 2. " NH 4 Br " = 13.942, =h .0077 3. " ratio (4) " = 13.935, .0073 4- " " (n) " = 13.911, .0048 5. " AgNO 3 "... " = 13.945,^.0044 6. " KNO 3 " = 13.947, .0057 7. " NaNO 3 " = 13.913, .0052 General mean N = 13.935, .0021 If oxygen is 16, this becomes 14.041. From Stas' data alone, Stas finds 14.044 ; Ostwald, 14.0410 ; Van der Plaats, 14.0421 (A), and 14.0519 (B) ; and Thomsen, 14.0396. The new value, representing all available data, falls between these limits of variation. 72 THE ATOMIC WEIGHTS. CARBON. Although there is a large mass of material relating to the atomic weight of carbon, much of it may be summarily set aside as having no value for present purposes. The density of carbon dioxide, which has been scrupulously determined by many investigators,* leads to no safe esti- mate of the constant under consideration. The numerous analyses of hydrocarbons, like the analyses of naphthalene by Mitscherlich, Wosk- resensky, Fownes, and Dumas, give results scarcely more satisfactory. In short, all the work done upon the atomic weight of carbon before the year 1840 may be safely rejected as unsuited to the present requirements of exact science. As for methods of estimation we need consider but four, as follows : First. The analysis of organic salts of silver. Second. The determination of the weight of carbon dioxide formed by the combustion of a known weight of carbon. Third. The method of Stas, by the combustion of carbon monoxide. Fourth. From the density of carbon monoxide. The first of these methods, which is probably the least accurate, was employed by Liebig and Redtenbacher f in 1840. They worked with the acetate, tartrate, racemate, and malate of silver, making five ignitions of each salt, and determining the percentage of metal. From one to nine grammes of material were used in each experiment. In the acetate the following percentages of silver were found : 64.615 64.624 64.623 64.614 64.610 Mean, 64.6172, .0018 After applying corrections for weighing in air, this mean becomes 64.6065. In the tartrate the silver came out as follows : 59.297 59-299 59-287 59-293 59-293 Mean, 59 2938, .0014 Or, reduced to a vacuum, 59.2806 * Notably by Lavoisier, Biot and Arago, De Sauss'ure, Dulong and Berzelius, Buff, Von Wrede, Regnault, and Marchand. For details, Van Geun's monograph may be consulted, f Ann. Chem. Pharm., 38, 137. Mem. Chem. Soc., i, 9. Phil. Mag. (3), 19, 210. CARBON. 73 In the racemate we have : 59.290 59.292 59-287 59.283 59.284 Mean, 59.2872, .0012 Or, corrected, 59.2769 And from the malate : 61.996 61.972 62.015 62.059 62.011 Mean, 62.0106, zb .0096 Or, corrected, 62.0016 Now, applying to these mean results the atomic weights already found for oxygen and silver, we get the following values for carbon : From the acetate C = 1 1-959, .0021 From the tartrate " 11.967, .0019 From the racemate " = 11.973, =h .0017 From the malate " = 11.972, .0098 Now these results, although remarkably concordant, are by no means unimpeachable. They involve two possible sources of constant error, namely, impurity of material and the volatility of the silver. These objections have both been raised by Stas, who found that the silver tar- trate, prepared as Liebig and Redtenbacher prepared it, always carried traces of the nitrate, and that he, by the ignition of that salt, could not get results at all agreeing with theirs. In the case of the acetate a similar impurity would lower the percentage of silver, and thus both sources of error would reinforce each other and make the atomic weight of carbon come out too high. With the three other salts the two sources of error act in opposite directions, although the volatility of the silver is probably far greater in its influence than the impurity. Even if we had no other data relating to the atomic weight of carbon, it would be clear from these facts that the results obtained by Liebig and Redtenbacher must be decidedly in excess of the true figure. Strecker, * however, discussed the data given by Liebig and Redten- bacher by the method of least squares, using the Berzeliaii scale, and assuming H = 12.51. Thus treated, they gave C = 75.415, and Ag = 1348.79 ; or, with =16, C = 12.066 and Ag = 107.903. These values *Ann. Chem. Pharm., 59, 280. 1846. 74 THE ATOMIC WEIGHTS. of course would change somewhat upon adoption of the modern ratio between and H. Observations upon silver acetate, like those of Liebig and Redtenbacher, were also made by Marignac.* The salt was prepared by dissolving silver carbonate in acetic acid, and repeatedly recrystallizing. Two ex- periments gave as follows : 3-3359 g rm acetate gave 2.1561 Ag. 64.633 per cent. 3.0527 " J -97 2 7 " 64.621 " Mean, 64.627, .0040 Reduced to a vacuum, this becomes 64.609. In a second series, conducted with special precautions to avoid me- chanical loss by spurting, Marignac found: 24.717 grm. acetate gave 15.983 Ag. 64.665 per cent. 21.202 " 13.709 " 64.661 " 31.734 " 20.521 " 64.666 " Mean, 64.664, .0010 Or, reduced to a vacuum, 64.646 Other experiments, comparable with the preceding series, have recently been published by Hardin, f who sought to redetermine the atomic weight of silver. Silver acetate and silver benzoate, carefully purified, were subjected to electrolysis in a platinum dish, and the percentage of silver so determined. For the acetate, using vacuum weights, he gives the following data, the percentage column being added by myself: .32470 grm. acetate gave .20987 Ag. 64.635 per cent. .40566 " .26223 " 64.643 " .52736 " .34086 " 64.635 .60300 "' .38976 " 64.637 " .67235 .43455 64.631 " .72452 " .46830 "' 64.636 " .78232 " .50563 " 64.632 " .79804 " .51590 " 64.646 .92101 " .59532 " 64.638 ". 1.02495 " .66250 " 64.637 " Mean, 64.637, .0011 Combining this series with those of the earlier investigators we have : Liebig and Redtenbacher 64.6065, .0018 Marignac, 1st series 64.609, .0040 Marignac, 2d " 64.646, .0010 Hardin 64.637, .001 1 General mean 64.636, .0007 *Ann. Chem. Pharm., 59, 287. 1846. t Journ. Amer. Chem. Soc., 18, 990. 1896. CARBOX. 75 With silver benzoate, C 7 H 5 Ag0 2 , Karelin's results are as follows : .40858 grm. benzoate gave .19255 Ag. 47. 127 per cent. .46674 " -21999 " 47.133 " .48419 " .22815 " 47.120 " .62432 .29418 " 47.120 " .66496 " .3!34Q " 47-I3I " .75853 -35745 " 47.i 2 4 " .76918 .3 62 47 " 47.124 " .81254 " .38286 " 47.H9 " .95673 " .45079 " 47."8 " 1.00840 " .47526 " 47- I 3 " Mean, 47.125, .0012 A different method of dealing with organic silver salts was adopted by Maumene,* in 1846, for the purpose of establishing by reference to carbon the atomic weight of silver. We will simply reverse his results and apply them to the atomic weight of carbon. He effected the com- bustion of the acetate and the oxalate of silver, and, by weighing both the residual metal and the carbon dioxide formed, he fixed the ratio between these two substances. In the case of the acetate his weighings show that for every gramme of metallic silver the weights of CO 2 were produced which are shown in the third column : 8.083 grm. Ag= 6.585 grm. CO 2 . -8147 11.215 " 9-J35 " -8136 I4.35 1 " H.6935 " -8148 9.030 7.358 " .8148 20.227 " 16.475 " .8145 Mean, .81448 The oxalate of silver, ignited by itself, decomposes too violently to give good results ; and for this reason it was not used by Liebig and Redtenbacher. Maumene, however, found that when the salt was mixed with sand the combustion could be tranquilly effected. The oxalate employed, however, with the exception of the sample represented in the last experiment of the series, contained traces of nitrate, so that these results involve slight errors. For each gramme of silver the appended weights of C0 2 were obtained : 14,299 grm. Ag. = 5.835 grm. CO 2 . .4081 17.754 7.217 4059 ".550 4.703 " .4072 10.771 4-387 4073 8.674 3-533 4073 "4355 4.658 4073 Mean, .40718 *Ann. Chim. Phys. (3), 18, 41. 1846. 76 THE ATOMIC WEIGHTS. New*, one of these salts being formed by a bivalent and the other by a univalent acid, we have to reduce both to a common standard. Doing this, we have the following results for the ratio between the atomic weight of silver and the molecular weight of CO 2 ; if Ag = 1.00 : From the acetate CO 2 = .40724, .000076 From the oxalate . . " .40718, =b .000185 General mean CO 2 = .40723, .000071 Here the slight error due to the impurity of the oxalate becomes of such trifling weight that it practically vanishes. As has already been said, the volatility of silver renders all the fore- going results more or less uncertain. Far better figures are furnished by the combustion of carbon directly, as carried out by Dumas and Stas * in 1840 and by Erdmann and Marchandf in 1841. In both investiga- tions weighed quantities of diamond, of natural graphite, and of artificial graphite were burned in oxygen, and the amount of dioxide produced was estimated by the usual methods. The graphite employed was puri- fied with extreme care by treatment with strong nitric acid and by fusion with caustic alkali. I have reduced all the published weighings to a common standard, so as to show in the third column the amount of oxygen which combines with a unit weight (say one gramme) of carbon. Taking Dumas and Stas' results first in order, we have from natural graphite : i.ooo grm. C gave 3.671 grm. CO 2 . 2.6710 .998 " 3.660 " 2.6673 .994 " 3- 6 45 " 2.6670 i. 216 4.461 " 2.6686 1.471 " 5-395 " 2.6676 Mean, 2.6683, =t - oo 5 With artificial graphite : .992 grm. C gave 3.642 grin. CO 2 . 2.6714 .998 " 3.662 " 2.6682 1. 660 " 6.085 " 2.6654 1.465 " 5-365 " 2.6744 Mean, 2.66985, .0013 And with diamond : .708 grm. C gave 2.598 grm. CO 2 . 2.6695 .864 3.1675 " 2.6661 1.219 4.465 " 2.6628 1.232 " 4.519 " 2.6680 1.375 " 5.041 " 2.6662 Mean, 5.6665 .0007 * Compt. Rend., 11, 991-1008. Ann. Chira. Phys. (3), i, i. f Jour, f Prakt. Chem., 23, 159. CARBON. 7/ Erdmann and Marchand's figures for natural graphite give the follow- ing results : J-537 6 g rm - g av e 5.6367 grm. CO 2 . 2.6659 1.6494 " 6.0384 " 2.6609 I-4505 " 5.31575 " 2.6647 In one experiment 1.8935 grm. of artificial graphite gave 6.9355 grm. CO 2 . Ratio for 0, 2.6628. This, combined with the foregoing series, gives a mean of 2.6636, .0007. With the diamond they found : .8052 grm. gave 2.9467 grm. CO 2 . 2.6596 1.0858 " 3-9875 " 2.6632 1.3557 " 4.9659 " 2.6629 1-6305 " 5-97945 " 2.6673 .7500 " 2.7490 " 2.6653 Mean, 2.6637, .0009 In more recent years the ratio under consideration has been carefully redetermined by Roscoe, by Friedel, and by Van der Plaats. Roscoe* made use of transparent Cape diamonds, and in a sixth experiment he burned carbonado. The combustions were effected in a platinum boat, contained in a tube of glazed Berlin porcelain ; and in each case the ash was weighed and its weight deducted from that of the diamond. The results were as follows, with the ratios stated as in the preceding series : 1.2820 grm. C gave 4.7006 CO 2 . 2.6666 1.1254 " 4.1245 " 2.6649 1.5287 " 5.6050 " 2.6665 .7112 " 2.6070 " 2.6656 1.3842 " 5.0765 " 2.6675 .4091 " J-4978 " 2.6612 Mean, 2.6654, .0006 Friedel's work,f also upon Cape diamond, was in all essential par- ticulars like Roscoe's. The data, after deduction of ash, were as follows : .4705 grm. C gave 1.7208 CO 2 . 2.6628 .8616 " 3.1577 " 2.6640 Mean, 2.6634, .0004 By Van der Plaats J we have six experiments, numbers one to three on graphite, numbers four and five on sugar charcoal, and number six on charcoal made from purified filter paper. Each variety of carbon was submitted to elaborate processes of purification, and all weights were *Ann. Chini. Phys. (5), 26, 136. Zeit. Anal. Chem., 22, 306. 1883. Compt. Rend., 94, 1180. 1882. fBull. Soc. Chim., 42, 100, 1884. % Compt. Rend., 100, 52. 1885. 78 THE ATOMIC WEIGHTS. reduced to vacuum standards. The data, with ash deducted, are sub- joined : 1. 5.1217 g rm - c g ave 18.7780 CO 2 . 2.6664 2. 9.0532 " 33- I 93i " 2.6664 3. 13.0285 " 47.7661 " 2.6663 4. 11.7352 " 43.0210 " 2.6660 5. 19.1335 " 7o.i33 6 " 2.6655 6. 4.4017 16.1-352 " 2.6657 Mean, 2.6660, =fc .0001 This combines with the previous series thus : Dumas and Stas, first set 2.6683, .0005 Dumas and Stas, second set 2.66985, .0013 Dumas and Stas, third set 2.6665, .0007 Erdmann and Marchand, first set 2.6636, .0007 Erdmann and Marchand, second set 2.6637, .0009 Roscoe 2.6654, d= .0006 Friedel 2.6634, .0004 Van der Plaats 2.6660, .0001 General mean 2.6659, .0001 Another very exact method for determining the atomic weight of car- bon was employed by Stas* in 1849. Carefully purified carbon mo- noxide was passed over a known weight of copper oxide at a red heat, and both the residual metal and the carbon dioxide formed were weighed. The weighings were reduced to a vacuum standard, and in each experi- ment a quantity of copper oxide was taken representing from eight to twenty-four grammes of oxygen. The method, as will at once be seen, is in all essential features similar to that usually employed for determin- ing the composition of water. The figures in the third column, deduced from the weights given by Stas, represent the quantity of carbon mo- noxide corresponding to one gramme of oxygen : 9.265 grm. O = 25.483 CO 2 . .75046 8.327 " 22.900 " .75010 13.9438 " 38.35 1 " .75040 11.6124 " 3L935 " .75oo8 18.763 " 51.6055 " .75039 19.581 " 53-8465 " -74994 22.515 " 61.926 " .75043 24.360 " 67.003 " -7553 Mean, 1.75029, db .00005 For the density of carbon monoxide the determinations made by Leducf are available. The globe used contained 2.9440 grm. of air. *Bull. Acad. Bruxelles, 1849 (*), 31. fCompt. Rend., 115, 1072. 1893. CARBON. 79 Filled with CO, it held the following weights, which give the accom- panying densities : Wt. CO. Density. 2.8470 -96705 2.8468 .96698 2.8469 .96702 Mean, .96702, .000015 Combining this density with Leduc's determination of the density of hydrogen, 0.6948, .00006745, it gives for the atomic weight of carbon : .C =^.11.957, .0270. Leduc himself combines the data with the density of oxygen, taken as 1.10503, and finds = 11.913. In either case, however, the probable error of the result is so high that it can carry little weight in the final combination. For carbon, including all the foregoing series, we now have the sub- joined ratios : (i.) Per cent. Ag in silver acetate 64.636, .0007 (2.) " " tartrate.... 59.2806, =b .0014 (3.) " " racemate.. 59.2769,1^.0012 (4.) " malate .... 62.0016, ".0096 (5.) " benzoate... 47.125, HT .0012 (6.) Ag : CO 2 : : i.oo : 0.40723, .000071 (7.) C : O 2 : : i.oo : 2.6659, .0001 (8.) O : CO : : i.oo : 1.75029, .00005 (9.) Density of CO (air = i), 0.96702, d= .000015 Now, computing with = 15.879, .0003, and Ag = 107.108, .0031, we get nine values for the atomic weight of carbon, as follows : From (i) C= 11.921, .0012 From (2) " 11.967, .0019 From (3) "= 11.973, .0017- From (4) " = 11.972, .0098 From (5) ..."== 11.917, .0008 From (6) " = 11.860, .0077 From (7) " 11.913, .0006 From (8) " = 11.914, .0010 From (9) " = 11.957, db .0270 General mean C = 11.920, .0004 If = 16, this becomes C = 12.011. 80 THE ATOMIC WEIGHTS. SULPHUR. The atomic weight of sulphur has been determined hy means of four ratios connecting it with silver, chlorine, oxygen, sodium ancl carbon. Other ratios have also been considered, but they are hardly applicable here. The earlier results of Berzelius w r ere wholly inaccurate, and his later experiments upon the synthesis of lead sulphate will be used in discussing the atomic weight of lead. Erdmann and Marchand deter- mined the amount of calcium sulphate which could be formed from a known weight of pure Iceland spar; and later they made analyses of cinnabar, in order to fix the value of sulphur by reference to calcium and to mercury. Their results will be applied in this discussion toward ascer- taining the atomic weights of the metals just named. First in order let us take up the composition of silver sulphide, as directly determined by Dumas, Stas, and Cooke. Dumas'* experiments were made with sulphur which had been thrice distilled and twice crys- tallized from carbon disulphide. A known weight of silver was heated in a tube in the vapor of the sulphur, the excess of the latter was distilled away in a current of carbon dioxide, and the resulting silver sulphide was weighed. I subjoin Dumas' weighings, and also the quantity of Ag 2 S proportional to 100 parts of Ag, as deduced from them : 9-9393 g rm - Ag= 1.473 s - Ratio, 114.820 9.962 1.4755 " " 114.811 30.637 4.546 " " 114.838 30.936 4.586 " " 114.824 3 .720 4.554 " " 114.824 Mean, 114.8234, rb .0029 Dumas used from ten to thirty grammes of silver in each experiment. Stas, f however, in his woi& employed from sixty to two hundred and fifty grammes at a time. Three of Stas' determinations were made by Dumas' method, while in the other two the sulphur was replaced by pure sulphuretted hydrogen. In all cases the excess of sulphur was expelled by carbon dioxide, purified with scrupulous care. Impurities in the dioxide may cause serious error. The five results come out as follows for 100 parts of silver : 114.854 114.853 114.854 114.851 114.849 Mean, 114.8522, .00x37 *Ann. Chem. Pharm., 113, 24. 1860. t Aronstein's translation, p. 179. SULPHUR. 81 The experiments made by Professor Cooke* with reference to this ratio were only incidental to his elaborate researches upon the atomic weight of antimony. They are interesting, however, for two reasons : they serve to illustrate the volatility of silver, and they represent, not syntheses, but reductions of the sulphide by hydrogen. Cooke gives three series of results. In the first the silver sulphide was long heated to full redness in a current of hydrogen. Highly concordant and at the same time plainly erroneous figures were obtained, the error being eventually traced to the fact that some of the reduced silver, although not heated to its melting point, was actually volatilized and lost. The second series, from reductions at low redness, are decidedly better. In the third series the sulphide was fully reduced below a visible red heat. Rejecting the first series, we have from Cooke's figures in the other two the subjoined quan- tities of sulphide corresponding to 100 parts of silver : 7.5411 grm. Ag 2 S lost .9773 grm. S. Ratio, 114.889 5.0364 .6524 " " 114.882 2.5815 -3345 " " 114.886 2.6130 .3387 " " 114.892 2.5724 .3334 " " 114.891 Mean, 114.888, .0012 I - I 357 S rm - A S2$ lost -1465 S. Ratio, 114.810 1.2936 .1670 " " 114.823 Mean, 114.8165, db .0044 Now, combining all four series, we get the following results : Dumas 1 14.8234, .0029 Stas '.'-.. 114.8522,^.0007 Cooke's 2d 114.888, .0012 Cooke's 3d 1 14.8165, d= .0044 General mean 1 14.8581, d= .0006 Here again we encounter a curious and instructive compensation of errors, and another evidence of the accuracy of Stas. The percentage of silver in silver sulphate has been determined by Struve and by Stas. Struve t reduced the sulphate by heating in a cur- rent of hydrogen, and obtained these results : 5.1860 grm. Ag ? SO 4 gave 3.5910 grm. Ag. 69.244 per cent. 6.0543 4.1922 " 69.243 8.6465 " 5.9858 " 69.228 " 11.6460 8.0608 " 69.215 " 9.1090 6.3045 " 69.212 " 9.0669 " 62778 " 69.239 " Mean * Proc. Ainer. Acad. of Arts of Sciences, vol. 12. 1877. f Ann. Chem. Pharm., 80, 203. 1^51. 82 THE ATOMIC WEIGHTS. Stas,* working by essentially the same method, with from 56 to 83 grammes of sulphate at a time, found these percentages : 69.200 69.197 69.204 69.209 69.207 69.202 Mean, 69.203, .0012 Combining this mean with that from Struve's series, we get a general mean of 69.205, 0011. The third sulphur ratio with which we have now to deal is one of minor importance. When silver chloride is heated in a current of sul- phuretted hydrogen the sulphide is formed. This reaction was applied by Berzelius f to determining the atomic weight of sulphur. He gives the results of four experiments ; but the fourth varies so widely from the others that I have rejected it. I have reason to believe that the varia- tion is due, not to error in experiment, but to error in printing ; never- theless, as I am unable to track out the cause of the mistake, I must exclude the figures involving it entirely from our discussion. The three available experiments, however, give the following results : The last column contains the ratio of silver sulphide to 100 parts of chloride. 6.6075 grm. AgCl gave 5.715 grm. Ag. 2 S. 86.478 9.2323 " 7-98325 " 86.471 10.1775 " 8.80075 " 86.472 Mean, 86.4737, db .0015 We have also a single determination of this value by Svanberg and Struve.J: After converting the chloride into sulphide they dissolved the latter in nitric acid. A trifling residue of chloride, which had been enclosed in sulphide, and so protected against change, was left undis- solved. Hence a slight constant error probably affects this whole ratio. The experiment of Svanberg and Struve gave 86.472 per cent, of silver sulphide derived from 100 of chloride. If we assign this figure equal weight with the results of Berzelius, and combine, we get a general mean of 86.4733, .0011. The work done by Richards relative to the atomic weight of sulphur is of a different order from any of the preceding determinations. Sodium carbonate was converted into sodium sulphate, fixing the ratio Na 2 CO s : Na a S0 4 : : 100 : x. The data are as follows, with vacuum weights : * Aronstein's translation, pp. 214-218. f Berzelius' Lehrbuch, sth ed., vol. 3, p. 1187. t Journ. Prakt. Chem., 44, 320. 1848. I Proc. Amer. Acad., 26, 268. 1891. SULPHUR. 83 Na 2 CO 3 . Na 2 SO. Ratio. 1.29930 I.74H3 134.005 3.18620 4.26790 133-950 1.01750 1.36330 133.985 2.07680 2.78260 I 33-985 1.22427 1.63994 I33-95 2 1.77953 2.38465 134.005 2.04412 2.73920 134.004 3.06140 4.10220 I33.997 Mean, 133.985, .0055 The available ratios for sulphur are now as follows : (l.) Ag 2 : Ag. 2 S : : loo : 114.8581, .0006 (2.) Per cent. Ag in Ag 2 SO 4 , 69.205, dz .oou (3.) 2 AgCl : Ag 2 S : : 100 : 86.4733, -O 011 (4.) Na 2 C0 3 : Na 2 SO 4 : : 100 : 133.985, =fc -OO55 From these ratios, four values for the atomic weight of sulphur are deducible. Calculating with O = 15.879, rt .0003 Ag = 107.108, .0031 Cl -== 35.179, .0048 Na = 22.88l, .0046 C = II.92O, rb .0004 AgCl = 142.287, .0037, we have : From (i) S = 31.828, =b .0016 From (2) " = 31.806, zb .0048 From (3) " = 31.864, i .0086 From (4) " = 31.835,^1.0191 General mean S = 31.828, .0015 If = 16, S = 32.070. From Stas' ratios alone, Stas found 32.074; Ostwald, 32.0626; Van der Plaats, (A) 32.0576, (B) 32.0590, and Thorn- sen, 32.0606. Here again Stas' determinations far outweigh all others. 84 THE ATOMIC WEIGHTS. LITHIUM. The earlier determinations of the atomic weight of lithium by Arfved- son, Stromeyer, C. G. Gmelin, and Kralovanzky were all erroneous, because of the presence of sodium compounds in the material employed. The results of Berzelius, Hagen, and Hermann were also incorrect, and need no further notice here. The only investigations which we need to consider are those of Mallet, Diehl, Troost, Stas, and Dittmar. Mallet's experiments* were conducted upon lithium chloride, which had been purified as completely as possible. In two trials the chloride was precipitated by nitrate of silver, which was collected upon a filter and estimated in the ordinary way. The figures in the third column represent the LiCl proportional to 100 parts of AgCl : 7.1885 grm. LiCl gave 24.3086 grm. AgCl. 29.606 8.5947 " 29.0621 29.574 In a third experiment the LiCl was titrated with a standard solution of silver. 3.9942 grm. LiCl balanced 10.1702 grm. Ag, equivalent to 13.511 grm. AgCl. Hence 100 AgCl = 29.563 LiCl. Mean of all three experiments, 29.581, .0087. Diehl.f whose paper begins with a good resume of all the earlier determinations, describes experiments made with lithium carbonate. This salt, which was spectroscopically pure, was dried at 130 before weighing. It was then placed in an apparatus from which the carbon dioxide generated by the action of pure sulphuric acid upon it could be expelled, and the loss of weight determined. From this loss the follow- ing percentages of C0 2 in Li 2 C0 3 were determined : 59.422 59.404 59.440 59.401 Mean, 59.417, .006 Diehl's investigation was quickly followed by a confirmation from Troost.J This chemist, in an earlier paper, had sought to fix the atomic weight of lithium by an analysis of the sulphate, and had found a value not far from 6.5, thus confirming the results of Berzelius and of Hagen, who had employed the same method. But Diehl showed that the BaS0 4 precipitated from Li. 2 S0 4 always retained traces of Li, which were recog- * Silliman's Amer. Journal, November, 1856. Chem. Gazette, 15, 7. f Ann. Chem. Pharm., 121, 93. JZeit. Anal. Chem., i, 402. I Annales d. Chim. et d. Phys., 51, 108. LITHIUM. 85 nizable by spectral analysis, and which accounted for the error. In the later paper Troost made use of the chloride and the carbonate of lithium, both spectroscopically pure. The carbonate was strongly ignited with pure quartz powder, thus losing carbon dioxide, which loss was easily estimated. The subjoined results were obtained : .97ogrm. Li 2 CO 3 lost .577 grm. CO 2 . 59-485 per cent. 1.782 " 1.059 " 59.427 " Mean, 59.456, .020 The lithium chloride employed by Troost was heated in a stream of dry hydrochloric acid gas, of which the excess, after cooling, was ex- pelled by a current of dry air. The salt was weighed in the same tube in which the foregoing operations had been performed, and the chlorine was then estimated as silver chloride. The usual ratio between LiCl and 100 parts of AgCl is given in the third column : 1.309 grm. LiCl gave 4.420 grm. AgCl. 29 615 2.750 " 9.300 " 29.570 Mean, 29.5925, .0145 This, combined with Mallet's mean, 29.581, .0087, gives a general mean of 59.584, .0075. Next in order is the work of Stas,* which was executed with his usual wonderful accuracy. In three titrations, in which all the weights were reduced to a vacuum standard, the following quantities of LiCl balanced 100 parts of pure silver : 39.356 -39-357 39-361 Mean, 39.358, .001 In a second series of experiments, intended for determining the atomic weight of nitrogen, LiCl was converted into LiN0 3 . The method was that employed for a similar purpose with the chlorides of sodium and of potassium. One hundred parts of LiCl gave of LiN0 3 : 162.588 162.600 162.598 Mean, 162.5953, dr .0025 The determinations of Dittmarf resemble those of Diehl; but the lithium carbonate used was dehydrated by fusion in an atmosphere of carbon dioxide. The carbonate was treated with sulphuric acid, and * Aroiistein's translation, 279-302. t Trans. Roy. Soc. Edinburgh, 35, II, 429. 1889. 86 THE ATOMIC WEIGHTS. the C0 2 was collected and weighed in an absorption apparatus, which was tared by a similar apparatus after the method of Regnault. The following percentages of CO 2 in Li 2 C0 3 were found : 59.601 59.645 59.529 rejected. 59.655 59.683 59.604 59.517 59.663 60.143 rejected. 59-794 59-584 Mean of all, 59.674 Rejecting the two experiments which Dittmar regards as untrust- worthy, the mean of the remaining nine becomes 59.638, .0173. This combines with the work of Diehl and Troost, as follows : Diehl 59.417, =b .0060 Troost 59.456, .0200 Dittmar 59.638, db .0173 General mean 59.442, =b .0054 Dittmar's determinations give a much lower value for the atomic weight of lithium than any of the others, and therefore seem to be ques- tionable. As, however, they carry little weight in the general combina- tion, it is not necessary to speculate upon their possible sources of error. The ratios for lithium are now as follows': (l.) AgCl : LiCl : : 100 : 29.584, .0075 (2.) Ag : LiCl : : 100 : 39.358, .001. (3.) LiCl : LiNO 3 : : 100 : 162.5953, .0025 (4.) Per cent, of CO 2 in Li 2 CO 3 , 59.442, .0054 And the data to use in their reduction are O -- 15.879, .0003 N 13.935, =t .0015 Ag 107.108, .0031 C = 11.920, db .0004 Cl == 35-179, .0048 AgCl= 142.287, .0037 These factors give two values for the molecular weight of lithium chloride, thus : From (i) LiCl = 42.0942, .01 10 From (2) =42.1556, .0016 General mean LiCl = 42. 1542, .0016 RUBIDIUM. 87 For lithium itself there are three values : From molecular weight LiCl Li = 6.9752, .0051 From (3^ " 6.9855, .0129 From (4) " 6.9628, d= .0077 General mean Li r= 6.9729, .0040 If 16, Li =- 7.026. From Stas' ratios, Stas found Li = 7.022 ; Ost- wald, 7.0303; Van cler Plaats (A), 7.0273; (B), 7.0235; and Thomsen, 7.0307. RUBIDIUM. i The atomic weight of rubidium has been determined by Bunsen, Pic- card, Godeffroy, and Hey cock from analyses of the chloride and bromide. Bunsen,* employing ordinary gravimetric methods, estimated the ratio between AgCl and RbCl. His rubidium chloride was purified by frac- tional crystallization of the chloroplatinate. He obtained the following results, to which, in a third column, I add the ratio between RbCl and 100 parts of AgCl : One grm. RbCl gave 1.1873 grm. AgCl. 84.225 1.1873 " 84.225 1.1850 " 84.388 I. 1880 " 84.175 Mean, 84.253, db .031 The work of Piccardf was similar to that of Bunsen. In weighing, the crucible containing the silver chloride was balanced by a precisely similar crucible, in order to avoid the correction for displacement of air. The filter was burned separately from the AgCl, as usual ; but the small amount of material adhering to the ash was reckoned as metallic silver. The rubidium chloride was purified by Bunsen's method. The results, expressed according to the foregoing standard, are as follows : I - 1 S&7 S rm - RbCl= 1.372 AgCl -f- .0019 Ag. 84.300 1.4055 " 1.6632 " .0030 " 84.303 i. ooi " 1.1850 " .0024 " 84.245 " 1-7934 " .0018 " 84.313 Mean, 84.290, d= .0105 Godeffroy, J starting with material containing both rubidium and *Zeit. Anal. Chem., i, 136. Poggend. Annal., 113, 339. 1861. f Journ. fur Prakt. Chem., 86, 454. 1862. Zeit. Anal. Chem., i, 518. I Ann. Chem. Pharm., 181, 185. 1876. 88 THE ATOMIC WEIGHTS. caesium, separated the two metals by fractional crystallization of their alums, and obtained salts of each spectroscopicalty pure. The nitric acid employed was tested for chlorine and found to be free from that impurity, and the weights used were especially verified. In two of his analyses of RbCl the AgCl was handled by the ordinary process of nitra- tion. In the other two it was washed by decantation, dried, and weighed in a glass dish. The usual ratio is appended in the third column : 1.4055 grm. RbCl gave 1.6665 g rm - AgCl. 84.338 1.8096 " 2.1461 84320 2.2473 " 2.665 " 84.326 2.273 " 2.6946 " 84.354 Mean, 84.3345, .0051 Combining the three series, we get the following result : Bunsen ................. 84.253, .031 Rb = 84.7O2 Piccard ................. 84.290, .0105 " 1=84.754 Godeffroy ............... 84.3345, dz .0051 " =84.817 General mean 84.324, =fc .0045 Heycock* worked by two methods, but unfortunately his results are given only in abstract, without details. First, silver solution was added in slight deficiency to a solution of rubidium chloride, and the excess of the latter was measured by titration. The mean of seven experiments gave Ag : RbCl : : 107.93 : 120.801 Hence Rb = 84.702. Two similar experiments with the bromide gave Ag : RbBr : : 107.93 : 165.437 Ag : RbBr : : 107.93 : 1 ^>S'34- 2 Mean, 165.3895, .0320 There are now three ratios for the metal rubidium, as follows : (i.) AgCl : RbCl : : loo : 84.324, .0045 (2.) Ag : RbCl : : 107.93 : 120.801 (3.) Ag : RbBr : : 107.93 ' l6 5-3 8 95> -3 2 To reduce these ratios we have Ag = 107.108, zb .0031 Br = 79.344, .0062 C1 = 35- T 79, -0048 AgCl = 142.287, zh .0037 * British Association Report, 1882, p. 499. CAESIUM. 89 For the molecular weight of RbCl, two values are calculable : From (i) RbCl= 119.981, + .0109 From (2) " 119.881, .0218 General mean RbCl = 119.961, =b .0097 To the value from ratio (2) I have arbitrarily assigned a weight rep- resented by the probable error as written above. The data for system- atic weighting are deficient, and no other course of procedure seemed advisable. From RbCl Rb = 84.782, .0109 From RbBr, ratio (3) " 84.786, .0329 General mean Rb =r 84.783, .0103 If = 16 Rb 85.429. CESIUM. The atomic weight of caesium, like that of rubidium, has been deter- mined from the analysis of the chloride. The earliest determination, by Bunsen,* was incorrect, because of impurity in the material employed. In 1863 Johnson and Allen published their results.f Their material was extracted from the lepidolite of Hebron, Maine, and the caesium was separated from the rubidium as bitartrate. From the pure caesium bitartrate caesium chloride was prepared, and in this the chlorine was estimated as silver chloride by the usual gravimetric method. Reducing their results to the convenient standard adopted in preceding chapters, we have, in a third column, the quantities of CsCl equivalent to 100 parts of AgCl : I-837 1 g rm - CsCl gave 1.5634 grm. AgCl. ' 117.507 2.1295 " i. Si n " n7-5 8 2.7018 " 2.2992 " 117-511 1.56165 " 1.3302 " ( "7-399 Mean, 117.499, .025 Shortly after the results of Johnson and Allen appeared a new series of estimations was published by Bunsen. J His caesium chloride was purified by repeated crystallizations of the chloroplatinate, and the ordi- *Zeit. Anal. Chem., i, 137. f Atner. Journ. Sci. and Arts (2), 35, 94. J Poggend. Annalen, 119, i. 1863. 90 THE ATOMIC WEIGHTS, nary gravimetric process was employed. The following results represent, respectively, material thrice, four times, and five times purified : 1.3835 grm. CsCl gave 1.1781 grm. AgCl. Ratio, 117.435 1.3682 " 1.1644 " " 117.503 1.2478 " 1.0623 98 THE ATOMIC WEIGHTS. alent proportions. In the first paper the following results were obtained. The third column gives the value of x in the ratio Cu : Ag. 2 : : 100 : x. Cu Taken. Ag Found. Ratio. .53875 I-8292 339.5 2 7 .56190 1.9076 339-49 1 1.00220 3.4016 339.414 1.30135 4.4173 339.440 99 s 7o 3.39035 339-477 1.02050 3.4646 ' 339.500 Mean, 339.475, =h .0114 In the second paper Richards states that the silver of the fifth experi- ment, which had been dried at 150, as were also the others, still retained water, to the extent of four-tenths milligramme in two grammes. If we assume this correction to be fairly uniform, as the concordance of the series indicates, and apply it throughout, the mean value for the ratio then becomes 339.408, .0114. This procedure, however, leaves the ratio in some uncertainty, and accordingly some new determinations were made, in which the silver, collected in a Gooch crucible, was heated to incipient redness before final weighing. Copper from two distinct sources was taken, and three experiments were carried out upon one sample to two with the other. Treating both sets as one series, the results were as follows : Cu Taken. Ag found. Ratio. .7576o 2.5713 339.40 .95040 3-2256 339-39 75993 2.5794 339-42 1.02060 3-4640 339-42 .90460 3.0701 339-39 Mean, 339.404, d= .0046 a value practically identical with the corrected mean of the previous determinations, and w 7 ith that found in the later experiments upon copper bromide. In various electrical investigations the same ratio, the electrochemical equivalent of copper, has been repeatedly measured, and the later results of Lord Rayleigh and Mrs. Sidgewick,* Gray,f Shaw, % and Vanni may properly be included in this discussion. As the data are somewhat dif- ferently stated, I have reduced them all to the common standard adopted above. Gray gives 'two sets of measurements, one made with large and the other w.ith^syiigill; metallic plates : ' ' T ' r ''',* Phil. 7>an r % British A3soc. Report, 1886. Abstract in Phil. Mag. (5), 23, 138. T ' r ' r r ^A?fn' der Phys. (Wiedemanu's) (2), 44, 214. COPPER. 99 Rayleigh and S. 340.483 340.832 340.367 Gray i. 341.297 34L4I3 340.815 340.252 339-905 341.064 340.832 341.297 341.064 34L4I3 Gray 2. 340.252 339.674 340.020 339.905 339-674 339-328 340.136 340.136 340.136 340.020 340.020 340.136 ' Shaiu. 339-68 340.05 339.84 339-71 340.04 339-94 340.35 339.82 340.09 339.84 339.90 339.98 340.H 340.56 339-82 340.56r, .0935 340.935, . 1072 339-953, .0521 Vanni. 340.483 340.600 340.367 340.252 340.600 340.136 340.406, .0520 The lack of sharp concordance in these data and the consequently high probable errors seem to indicate a distinct superiority of the purely chemical method of determination over that adopted by the physicist. The eight distinct series now combine as follows : Richards, first series corrected 339-48, .0114 Richards, second series ... 339.404, .0046 Richards, CuBr 2 series . 339.392, =b .0108 Rayleigh and Sidgewick 340.561, d= .0935 Gray, with large plates 34-935, ^ . 1072 Gray, with small plates 339-953, .0521 Shaw . .. 339.983, =b .0411 Vanni 340.406, .0520 General mean 339-41 1, .0039 If we combine Richards' three series into a general mean separately, we get 339.402, .0040. Hence the other determinations, having high >robable errors, practically vanish from the result, and it is a matter of idifference whether they are retained or rejected. We now have the following ratios from which to compute the atomic .'eight of copper : (i.) Percentage of Cu in CuO 79-8355, .0010 (2.) (3-) (4.) (5.) Cu (6.) Cu of Cu in CuSO 4 39-795, =fc -0036 of Cu in CuSO 4 , 5H 2 O. . 25.451, .0011 of CuO in CuSO 4 49-8i6, ^-.0017 Na,CO, Na 2 SO 4 : 100 IOO 166.838, .0035 223.525, =fc .0098 (7.) BaSO 4 : Cu : : 93- 28 9 : 25-448- (8.) 2AgBr : Cu : : IOO : 16.924, .0007 (9- : Ag 2 .0039 100 THE ATOMIC WEIGHTS. Reducing these ratios with the subjoined data : O - = 15.879, .0003 Na _ 22.881, .0046 Ag 107.108, .0031 Ha = 136.392, =h .0086 S = 31.828, .0015 AgBr = 186.452, =b .0054 C = 11.920, d- .0004 We have nine values for the atomic weight of .copper. Since ratio (7) depends upon one experiment only, it is necessary to assign the value derived from it arbitrary weight. This will be taken as indicated by a probable error double that of the next highest, obtained from ratio (.2). The values then are as follows : From (i) ......................... Cu = 62.869, d= .0034 From (2) .......................... " = 63.022, db .0070 From (3) .......................... <( = 63.070, .0030 From (4) ......................... " =63.003, .0042 From (5) .......................... " =63.127, d= .0051 From (6).. , ....................... " = 63.128, .0050 From (7) ................ .......... " 63.215, .0140 From (8) .............. ........... " = 63. 1 10, .0032 From (9) .......................... " =63. 114, .0020 General mean ................ Cu 63.070, d= .0012 If O = 16, Cu = 63.550. If we include Hampe's analyses of copper sulphate, which gave Cu = 62.839, .0035, the general mean becomes Cu 63.046, .0011. The foregoing means, however, are significant only as showing the effect and weight of the older data upon the newer determinations of Richards. The seventh of the individual values is also interesting, for the reason that the experiment upon which it depends was published by Richards previous to his investigation of the atomic weight of barium. With the old value for Ba, 137, it gives a value for copper in close agree- ment with Richards' other determinations. With the new value for barium it becomes discordant, although its weight is so low that it pro- duces no appreciable effect upon the final mean. Rejecting values 1 to 4, inclusive, the remaining five values give a gen- eral mean of Cu ==63.119, rfc .0015. If = 16, this becomes 63600, and in the light of all the evidence these figures are to be preferred. If, again, we combine with this mean the results of Richards' work on the oxide and sulphate of copper, the final value becomes , \\ \\ '', ; Cu = 63.108, .0013, with = 16 f $8$$$.'% This departs but little from the previous mean '.value', 'bu^it'i'H'el'uclee' 'data which render it, in all probability, a trifle too low, l^h'p/v^Hie Cu = 63.119 will be regarded as the best. GOLD. 101 GOLD. Among the early estimates of the atomic weight of gold the only ones worthy of consideration are those of Berzelius and Levol. The earliest method adopted by Berzelius* was that of precipitating a solution of gold chloride by means of a weighed quantity of metallic mercury. The weight of gold thus thrown down gave the ratio between the atomic weights of the two metals. In the single experiment which Berzelius publishes, 142.9 parts of Hg precipitated 93.55 of Au. Hence if Hg = 200, Au = 196.397. In a later investigation f Berzelius resorted to the analysis of potassio- auric chloride, 2KC1. A uCl 3 . Weighed quantities of this salt were ignited in hydrogen ; the resulting gold and potassium chloride were separated by means of water, and both were collected and estimated. The loss of weight upon ignition was, of course, chlorine. As the salt could not be perfectly dried without loss of chlorine, the atomic weight under inves- tigation must be determined by the ratio between the KC1 and the Au. If we reduce to a common standard, and compare with 100 parts of KC1, the equivalent amounts of gold will be those which I give in the last of the subjoined columns : 4.1445 grm. K 2 AuCl 5 gave .8185 grm. KC1 and 2.159 g rm - Au. 263.775 2.2495 .44425 " i.7 2 " 26 3- 8l 5 5 1300 " 1.01375 " 2.67225 " 263.600 3.4130 " .674 " 1.77725 " 263.687 4.19975 .8295 " 2.188 263.773 Mean, 263.730, .026 Still a third series of experiments by Berzelius $ may be included here. In order to establish the atomic weight of phosphorus he em- ployed that substance to precipitate gold from a solution of gold chloride in excess. Between the weight of phosphorus taken and the weight of gold obtained it was easy to fix a ratio. Since the atomic weight of phosphorus has been better established by other methods, we may properly reverse this ratio and apply it to our discussion of gold. 100 parts of P precipitate the quantities of Au given in the third column : .829 grm. P precipitated 8.714 grm. Au. 1051.15 .754 " 7-93 " 1051.73 Mean, 1051.44, d= .196 Hence if P = 31, Au = 195.568. * Poggend. Annalen, 8, 177. f Lehrbuch, 5 Aufl., 3, 1212. J Lehrbuch, 5 Aufl., 3, 1188. 102 THE ATOMIC WEIGHTS. Level's * estimation of the atomic weight under consideration can hardly have much value. A weighed quantity of gold was converted in a flask into AuCl 3 . This was reduced by a stream of sulphur dioxide,. and the resulting sulphuric acid was determined as BaS0 4 . One gramme of gold gave 1,782 grin. BaS0 4 . Hence Au = 195.06. All these values may be neglected as worthless, except that derived from Berzelius' K 2 AuCl 5 series. In 1886 Kriissf published the first of the recent determinations of the atomic w r eight under consideration, several distinct methods being re- corded. First, in a solution of pure auric chloride the gold was pre- cipitated by means of aqueous sulphurous acid. In the filtrate from the gold the chlorine was thrown down as silver chloride, and thus the ratio Au : 3 AgCl was measured. I subjoin Kriiss' weights, together with a third column giving the gold equivalent to 1QO parts of silver chloride: Au. AgCl. Ratio. 7.72076 16.84737 45-828 5.68290 12.40425 45.814 3.24773 7.08667 45-828 4.49167 980475 45-8ii 3-47949 7-59300 45-825 3.26836 7 13132 45-832 5.16181 11.26524 45.821 4.86044 10.60431 45.834 Mean, 45.824, .0020 The remainder of Kruss' determinations were made with potassium auribromide, KAuBr 4 , and with this salt several ratios were measured. The salt was prepared from pure materials, repeatedly recrystallized under precautions to exclude access of atmospheric dust, and dried over phosphorus pentoxide. First, its percentage of gold was determined, sometimes by reduction with sulphurous acid, sometimes by heating in a stream of hydrogen. For this ratio, the weights and percentages are as follows, the experiments being numbered for further reference, and the reducing agent being indicated. KAuBr. Au. Per cent. i. SO, 10.64821 3-77753 35476 2 - S0 2 4.71974 1.67330 35-453 3- H 7-05762 2.50122 35-440 4. H 4-49558 1-59434 35-465 5- SO, 8.72302 3-09448 35-475 6. SO 2 7.66932 2.71860 35-448 7- SO, 7.15498 2.53695 35.457 8 - H 12.26334 4-34997 35-471 9- II 7-10342 2.51919 35-465 Mean, 35.461, .0028 - , , * Ann. Chim. Phys. (3), 30, 355. 1850. "t Untersuchungen uber das Atomgewicht des Goldes. Mi'mchen, 1886. 112 pp., Svo. GOLD. 103 In five of the foregoing experiments the reductions were effected with sulphurous acid ; and in these, after filtering off the gold, the bromine was thrown down and weighed as silver bromide. This, in comparison with the gold, gives the ratio Au : 4AgBr : : 100 : x. Au. Ratio. i 3-77753 H.39542 381.080 2 1.67330 6.37952 381.254 5 3- 9448 ".78993 380.999 6 2.71860 10.35902 381.042 7 2.53695 9.66117 380.731 Mean, 381.021, . .057 Hence Au : AgBr : : 100 : 95.255, .0142. In the remaining experiments, Nos. 3, 4, 8, and 9, the KAuBr 4 was reduced in a stream of hydrogen, the loss of weight, Br 3 , being noted. In the residue the gold was determined, as noted above, and the KBr was also collected and weighed. The weights were as follows : Au. Loss, Br z . KBr. 3 2.50122 3-04422 1.51090 4 1-59434 1-93937 -96243 8 4-34997 5- 2 93 l6 2.62700 9 2.51919 3-06534 L52I53 From these data we obtain two more ratios, viz., Au : Br 3 : : 100 : SB, and Au : KBr : : 100 : x, thus : Au : Br z . Au : KBr. 3 121.710 60.405 4 121.641 60.365 8 121.683 60.391 9 121.680 60.398 Mean, 121.678, .0100 Mean, 60.390, .0059 From all the ratios, taken together, Krtiss deduces a final value of Au = 197.13, if = 16. It is obviously possible to derive still other ratios from the results given, but to do so would be to depart unneces- sarily from the author's methods as stated by himself. Thorpe and Laurie, * whose work appeared shortly after that of Kruss, also made use of the salt KAuBr 4 , but, on account of difficulty in drying it without change, they did not weigh it directly. After proving the con- stancy in it of the ratio Au : KBr, even after repeated crystallizations, they adopted the following method : The unweighed salt was heated with gradual increase of temperature, up to about 160, for several hours, and afterwards more strongly over a small Bunsen flame. This was done in a porcelain crucible, tared by another in weighing, which latter was treated in precisely the same way. The residue, KBr -f Au, was weighed, the KBr dissolved out, and the gold then weighed separately. The * Journ. Chein. Soc., 51, 565. 1887. 104 THE ATOMIC WEIGHTS. weightjof KBr was taken by difference. The ratio Au:KBr: : 100 : x appears in a third column. An. KBr. Ratio. 6.19001 3-73440 60.329 4.76957 2.87715 60.32? 4.14050 2.49822 60.336 3.60344 2.17440 60.342 3.67963 2.21978 60.326 4-57757 2.76195 60.337 5-36659 3-23821 60.326 5.16406 3. 11533 60.327 Mean, 60.331, .0016 This mean combines with Krtiss' thus: Kriiss 60.390, .0059 Thorpe and Laurie 60.331, d= .0016 General mean 60.338, d= .0015 The potassium bromide of the previous experiments was next titrated with a solution of pure silver by Stas' method, the operation being performed in red light. Thus we get the following data for the ratio Ag : Au : : 100 : :c, using the weights of gold already obtained : Ag. Au. Ratio. 3.38451 6.19001 182.893 2.60896 4.76957 182.813 2.28830 4.18266 182.786 2.26415 4.14050 182.868 1.97147 3-60344 182.775 2.01292 3-67963 182.801 2.50334 4-57757 182.863 2.93608 5.36659 182.780 2.82401 5.16406 182.865 Mean, 182.827, .0101 Finally, in eight of these experiments, the silver bromide formed during titration was collected and weighed, giving values for the ratio Au: AgBr: 100 : x, as follows : An. AgBr. 6.19001 5.89199 4.76957 4-54261 4.18266 3.98288 4.14050 3-94309 3-60344 3-43 i5 3.67963 3-50207 4.57757 4.35736 5.36659 5-11045 Ratio. 95.186 95.242 95-224 95.232 , 95.i9i 95-175 95-189 95.227 Mean, 95.208, .0061 Kriiss found, 95.255, dr .0142 General mean, 95.222, .0056 GOLD. 105 From the second and third of the ratios measured by Thorpe and Laurie an independent value for the ratio Ag : Br may be computed. It becomes 100 : 74.072, which agrees closely with the determinations made by Stas and Marignac. Similarly, the ratios Ag : KBr and AgBr : KBr may be calculated, giving additional checks upon the accuracy of the manipulation, though not upon the purity of the original material studied. Thorpe and Laurie suggest objections to the work done by Kriiss, on the ground that the salt KAuBr 4 cannot be completely dried without loss of bromine. This suggestion led to a controversy between them and Kriiss, which in effect was briefly as follows : First, Kriiss* urges that the potassium auribromide ordinarily contains traces of free gold, not belonging to the salt, produced by the reducing action of dust particles taken up from the air. He applies a correction for this supposed free gold to the determinations made by Thorpe and Laurie, and thus brings their results into harmony with his own. To this argument Thorpe and Laurie f reply, somewhat in detail, stating that the error indicated was guarded against by them, and that they had dissolved quantities of from eight to nineteen grammes of the auri- bromide without a trace of free gold becoming visible. A final note in defense of his own work was published by Kriiss a little later. J In 1889 an elaborate set of determinations of this constant was pub- lished by Mallet, whose experiments are classified into seven distinct series. First, a neutral solution of auric chloride was prepared, which was weighed off in two approximately equal portions. In one of these the gold was precipitated by pure sulphurous acid, collected, washed, dried, ignited in a Sprengel vacuum, and weighed. To the second por- tion a solution containing a known weight of pure silver was added. After filtering, with all due precautions, the silver remaining in the fil- trate was determined by titration with a weighed solution of pure hydro- bromic acid. We have thus a weight of gold, and the weight of silver needed to precipitate the three atoms of chlorine combined with it; in other words, the ratio Ag 3 : Au : : 100 : x. All weights in this and the subsequent series are reduced to vacuum standards, and all weighings were made against corresponding tares. Au. Ag y Ratio. 7.6075 12.4875 60.921 8.4212 13.8280 60.900 6.9407 ir -3973 60.898 3.3682 5.5286 60.923 2.8244 4.6371 60.909 Mean, 60.910, .0034 Hence Ag : Au : : 100 : 182.730, .0102. *Ber. Deutsch. Chem. Gesell., 20, 2365. 1887. fBerichte, 20, 3036, and Journ. Chem. Soc., 51, 866. 1887. t Berichte, 21, 126. 1888. % Philosophical Transactions, 180, 395. 1889. 106 THE ATOMIC WEIGHTS. The second series of determina.tions was essentially like the first, ex- cept that auric bromide was taken instead of the chloride. The ratio measured, Ag 3 : Au, is precisely the same as before. Results as follows : Au. Ag y Ratio. 8.2345 13-5149 60.929 7.6901 12.6251 60.911 105233 17.2666 60.945 2.7498 4.5141 60.916 3.5620 5-8471 60.919 3.9081 6.4129 60.941 Mean, 60.927, .0038 Hence Ag : Au : : 100 : 182.781, .0114. In the third series of experiments the salt KAuBr 4 was taken, purified by five recrystallizations. The solution of this was weighed out into nearly equal parts, the gold being measured as in the two preceding series in one portion, and the bromine thrown down by a standard silver solution as before. This gives the ratio Ag 4 : Au : : 100 : x. Au. Ag. Ratio. 5.7048 12.4851 45. 6 93 7.9612 I7.4I93 45- 6 93 2 -4455 5.35!3 45- 6 99 4.1632 9-"53 45- 6 73 Mean, 45.689, .0040 . Hence Ag : Au : : 100 : 182.756, .0160. The fifth series of determinations, which for present purposes naturally precedes the fourth, was electrolytic in character, gold and silver being simultaneously precipitated by the same current. The gold was in solu- tion as potassium auro-cyanide, and the silver in the form of potassium silver cyanide. The equivalent weights of the two metals, thrown down in the same time, were as follows, giving directly the ratio Ag : Au : : 100 : x. Au. Ag. Ratio. 5.2721 2.8849 182.748 6.3088 3.4487 182.933 4.2770 2.3393 182.832 3-5 I2 3 1.9223 182.713 3.6804 2.0132 182.814 Mean, 182.808, .0256 This mean may be combined with the preceding means, and also with the determination of the same ratio by Thorpe and Laurie, thus : Thorpe and Laurie 182.827, .0.101 Mallet, chloride series 182.730, .0102 Mallet, bromide series 182.781, .01 14 Mallet, KAuBr 4 series 182 756, .0160 Mallet, electrolytic 182.808, .0256 General mean 182.778, =h .0055 GOLD. 107 Iii Mallet's fourth series a radically new method was employed. Tri- m ethyl-ammonium aurichloride, N(CH 3 ) 3 HAuCl 4 , was decomposed ly heat, and the residual gold was determined. In order to avoid loss by spattering, the salt was heated in a crucible under a layer of fine siliceous sand of known weight. Several crops of crystals of the salt were studied, as a check against impurities, but all gave concordant values. Salt. Residual Au. Percent. A u. 14-9072 7-3754 49-475 15.5263 7.6831 49-484 10.4523 5-1712 49-474 6.5912 3.2603 49.464 5-5744 2.7579 49-474 Mean, 49.474, .0021 In his sixth and seventh series Mallet seeks to establish, by direct measurement, the ratio between hydrogen and gold. In their experi- mental details his methods are somewhat elaborate, and only the pro- cesses, in the most general way, can be indicated here. First, gold was precipitated electrolytically from a solution of potassium aurocyanide, and its weight was compared with that of the amount of hydrogen simul- taneously liberated in a voltameter by the same current in the same time. The hydrogen was measured, and its weight was then computed from its density. The volumes are given, of course, at and 760 mm. Wt. Au. Vol. H, cc. Wt. H. 4.0472 228.64 . 2 5483 4.0226 227.03 .0204046 4.0955 231.55 , .0208103 These data, with the weight of one litre of hydrogen taken as 0.89872 gramme, give the subjoined values in the ratio H : Au : : 1 : x. 196.960 197-151 196.805 Mean, 196.972, =b .0675 In the-last series of experiments a known quantity of metallic zinc was dissolved in dilute sulphuric acid, and the amount of hydrogen evolved was measured. Then a solution of pure auric chloride or'bromide was treated with a definite weight of the same zinc, and the quantity of gold thrown down was determined. The zinc itself was purified by practical distillation in a Sprengel vacuum. From these data the ratio H 3 : Au was computed by direct comparison of the weight of gold and that of the liberated hydrogen. The results were as follows : 108 THE ATOMIC WEIGHTS. Wt. Au. 10.3512 8.2525 8.1004 3-2913 3.4835 3.6421 Vol. H, cc. 1756.10 1400.38 1374.87 558.64 590-93 Wt. H. .157824 125857 .123565 .050206 .053109 055551 Hence for the ratio H 3 : Au : : 1 : x we have : 65-587 65-571 65.557 65.556 65o93 65.563 Mean, 65.571, .00436 And H : Au : : 1 : 196.713, .0131. This, combined with the value found in the preceding series, gives a general mean of 196.722, .0129. The ratios available for gold are now as follows : (l.) 2KC1 : Au : : 100 : 263.730, .026 ( 2 -) 3^gCl : Au : : 100 : 45.824, ; .0020 (3.) KAuBr 4 : Au : : 100 : 35.461, db .0028 (4.) Au : AgBr : : 100 : 95.222, .0056 (5.) Au : Br 3 : (6.) Au : KBr (7.) Ag : Au : (8. (9.) H : Au : IOO : 121.678, db .OIOO : : TOO : 60.338, d= 0015 100 : 182.778, d= -0055 loo : 49.474, .0021 196.722, dr .0129 For the reduction of these ratios the antecedent data are : Ag= 107.108, zb .0031 Cl == 35.179, d= .0048 Br = 79-344, .0062 K = 38.817, d= .0051 N = J 3.935, .0021 C = 11.920, dr .0004 AgCl 142.287, .0037 AgBr = 186.452, d= .0054 KC1 = 74.025, .0019 KBr = 118.200, .0073 Hence for the atomic weight of gold we have nine values : From (i) Au = 195.226, .0193 From (2) From (3) From (4) From (5) From (6) From (7) From (8) From (9) 195.605, d= .0099 = I95-7 11 , .0224 = 195.808, d= .0126 = 195.624, .0222 = T 95- 8 96, db .0131 = 195.770, db .0082 = 196.238, =h .0224 = 196.722, .0129 General mean Au = 195.850, dz .0044 If = 16, this becomes Au = 197.342. GOLD. 109 Of the foregoing values the first one, which is derived from Berzelius' work, should certainly be rejected. So also, apparently, should the eighth and ninth values. Excluding these, values 2 to 7, inclusive, give a gen- eral mean of Au = 195.743, .0049. With = 16, this becomes Au = 197.235. Probably these values are more nearly correct than those which include all the determinations. The ninth value in the list given above represents Mallet's comparisons of gold directly with hydrogen, and is peculiarly instructive. In Mal- let's paper the other determinations are discussed upon the basis of O = 15.96, which brings them more nearly into harmony with the hydro- gen series. The great divergence shown in this recalculation is due to the new value for oxygen, 15.879, and its effect upon the atomic weights of silver, bromine, etc. The former agreement between the several series of gold values was therefore only apparent, and we are now able to see that concordance among determinations maybe only coincidence, and no proof of accuracy. It is probable, furthermore, that direct compari- sons of metals with hydrogen cannot give good measurements of atomic weights, for several reasons. First, it is not possible to be certain that every trace of hydrogen has been collected and measured, and any loss tends to raise the apparent atomic weight of the metal studied ; secondly, the weight of the hydrogen is computed from its volume, and a slight change in the factors used in reduction of the observations may make a considerable difference in the final result. These uncertainties exist in all determinations of atomic weights hitherto made by the hydrogen method. 110 THE ATOMIC WEIGHTS. CALCIUM. For determining the atomic weight of calcium we have sets of experi- ments by Berzelius, Erdmann and Marchand, and Dumas. Salvetat * also has published an estimation, but without the details necessary to enable us to make use of his results. I also find a reference f to some work of Marignac, which, however, seems to have been of but little im- portance. The earlier work of Berzelius was very inexact as regards calcium, and it is not until we come down to the year 1824 that we find any material of decided value. The most important factor in our present discussion is the composi- tion of calcium carbonate, as worked out by Dumas and by Erdmann and Marchand. In 1842 Dumas J made three ignitions of Iceland spar, and determined the percentages of carbon dioxide driven off and of lime remaining. The impurities of the material were also determined, the correction for them applied, and the weighings reduced to a vacuum standard. The per- centage of lime came out as follows : 56.12 56.04 56.06 Mean, 56.073, .016 About this same time Erdmann and Marchand began their researches upon the same subject. Two ignitions of spar, containing .04 per cent, of impurity, gave respectively 56.09 and 56.18 per cent, of residue ; but these results are not exact enough for us to consider further. Four other results obtained with artificial calcium carbonate are more noteworthy. The carbonate was precipitated from a solution of pure calcium chloride by ammonium carbonate, was washed thoroughly with hot water, and dried at a temperature of 180. With this preparation the following residues of lime were obtained : 56.03 55.98 56.00 55-99 Mean, 56.00, .007 It was subsequently shown by Berzelius that calcium carbonate pre- pared by this method retains traces of water even at 200, and that *Compt. Rend., 17, 318. 1843. fSee Oudeman's monograph, p. 51. JCompt. Rend., 14, 537. 1842. g Journ. fur Prakt. Chem., 26, 472. 1842. CALCIUM. Ill minute quantities of chloride are also held by it. These sources of error are, however, in opposite directions, since one would tend to diminish and the other to increase the weight of residue. In the same paper there are also two direct estimations of carbonic acid in pure Iceland spar, which correspond to the following percentages of lime : 56.00 56.02 Mean, 56.01, .007 In a still later paper* the same investigators give another series of results based upon the ignition of Iceland spar. The impurities were carefully estimated, and the percentages of lime are suitably corrected : 4.2134 grm. CaCO 3 gave 2.3594 grm. CaO. 55-997 per cent. 15.1385 " 8.4810 " 56.022 " 23.5503 " 13.1958 " 56-031 " 23.6390. I3-245 6 " 5 6 -3 2 42.0295 23.5533 " 56.044 " 49.7007 " 27.8536 " 56.042 " Mean, 56.028, .0047 Six years later Erdmann and March and f published one more result upon the ignition of calcium carbonate. They found that the compound began giving off carbon dioxide below the temperature at which their previous samples had been dried, or about 200, and that, on the other hand, traces of the dioxide were retained by the lime after ignition. These two errors do not compensate each other, since both tend to raise the percentage of lime. In the one experiment now under consideration these errors were accurately estimated, and the needful corrections were applied to the final result. The percentage of residual lime in this case came out 55.998. This agrees tolerably well with the figures found in the direct estimation of carbonic acid, and, if combined with those two. gives a mean for all three of 56.006, .0043. Combining all these series, we get the following result : Dumas 56.073, .016 Erdmann and Marchand . . 56.006, rb .007 Erdmann and Marchand 56.028, dr .0047 Erdmann and Marchand 56.006, .0043 General mean 56.0198, .0029 For reasons given above, this mean is probably vitiated by a slight instant error, which makes the figure a trifle too high. * Journ. fur Prakt. Cheni., 31, 269. 1844. f Journ. fi'ir Prakt. Chem., 50, 237. 1850. 112 THE ATOMIC WEIGHTS. In the earliest of the three papers by Erdmann and Marchand there is also given a series of determinations of the ratio between calcium car- bonate and sulphate. Pure Iceland spar was carefully converted into calcium sulphate, and the gain in weight noted. One hundred parts of spar gave of sulphate : 136.07 136.06 136.02 136.06 Mean, 136.0525, .0071 In 1843 the atomic weight of calcium was redetermined by Berzelius, * who investigated the ratio between lime and calcium sulphate. The calcium. was first precipitated from a pure solution of nitrate by means of ammonium carbonate, and the thoroughly washed precipitate was dried and strongly ignited in order to obtain lime wholly free from ex- traneous matter. This lime was then, with suitable precautions, treated with sulphuric acid, and the resulting sulphate was weighed. Correction was applied for the trace of solid impurity contained in the acid, but not for the weighing in air. The figures in the last column represent the percentage of weight gained by the lime upon conversion into sulphate : 1.80425 grm. CaO gained 2.56735 grm. 142.295 2.50400 " 3.57050 " 142.592 3.90000 S-SSHO " 142.343 3.04250 " 4.32650 " 142.202 3.45900 " 4- 93 HO " 142.567 Mean, 142.3998, .0518 Last of all we have the ratio between calcium chloride and silver, as determined by Dumas, t Pure calcium chloride was first ignited in a stream of dry hydrochloric acid, and the solution of this salt was after- wards titrated with a silver solution in the usual way. The CaCl 2 pro- portional to 100 parts of Ag is given in a third column : 2.738 grm. CaCl 2 = 5.309 grm. Ag. 51-573 2.436 " 4.731 " 5 '.490 1.859 3-6i7 " 5^396 2.771 5.38*5 " 5L424 2.240 4.3585 " 5 I -394 Mean, 51.4554, .0230 We have now four ratios to compute from, as follows : (i.) Percentage CaO in CaCO 3 , 56.0198, .0029 (2.) CaO : SO 3 : : 100 : 142.3998, .0518 (3.) CaCO 3 : CaSO 4 : : 100 : 136.0525, .0071 (4-) Ag 2 : CaC] 2 : : 100 : 51.4554, .0230 * Journ. fur Prakt. Chem., 31, 263. Ann. Chem. Pharm., 46, 241. t Ann. Chim. Phys. (3), 55, 129. 1859. Ann. Chem. Pharm., 113, 34. STRONTIUM. 113 The antecedent values are O = 15.879, .0003 c= 11.920, .0004 Ag = 107.108, .0031 S =31.828, -0015 Hence the subjoined values for the atomic weight of calcium : From (i) .......................... Ca = 39.757, dz .0048 From (2) ... ...................... " = 39.925, .0203 From (3) ................. ........ " == 39-706, .0204 From (4) ......................... " = 39.868, HE .0503 Mean Ca = 39.764, .0045 If = 16, Ca = 40.067. STRONTIUM. The ratios which fix the atomic weight of strontium resemble in gen- eral terms those relating to barium, only they are fewer in number and represent a smaller amount of work. The early experiments of Stro- meyer,* who measured the volume of CO 2 evolved from a known weight of strontium carbonate, are hardly available for the present discussion. So also we may exclude the determination by Salvetat,f who neglected to publish sufficient details. Taking the ratio between strontium chloride and silver first in order, we have series of figures by Pelouze and by Dumas. Pelouze J employed the volumetric method to be described under barium, and in two ex- periments obtained the subjoined results. In another column I append the ratio between SrCl 2 and 100 parts of silver : 1.480 grm. SrCl 2 = 2.014 grm. Ag. 73-486 2.210 " 3.008 " 73-47 1 Mean, 73.4781, db .0050 Dumas, by the same general method, made sets of experiments with three samples of chloride which had previously been fused in a current of dry hydrochloric acid. His results, expressed in the usual way, are as follows : * Schweigg. Journ., 19, 228. 1816. f Compt. Rend., 17, 318 1843. I Compt. Rend., 20, 1047. 1845. I Ann. Chim. Phys. (3), 55, 29. 1859. Ann Cheat. Pharm., 113, 34. 114 THE ATOMIC WEIGHTS. Series A. 3.137 grm. SrCl 2 = 4.280 grm. Ag. Ratio, 73.2944 1.982 " 2.705 " " 73-27I7 3.041 4.142 " 73.4186 3.099 " 4.219 " 73-4534 Mean , 73-3595 Series B. 3.356 grm. SrCl 2 = 4-574 grm. Ag. Ratio, 73-3713 6.3645 8.667 " 73.4327 7.131 9.712 " 73.4246 Mean, 73.4095 Series C. 7.213 grm. SrC I 2 9.811 grm. Ag. Ratio, 73-5J95 2.206 " 3.006 " " 73.3866 4.268 5.816 " " 73.5529 4.018 " 5-477 " " 73.3613 Mean, 73.455 1 Mean of all as one series, 73.4079, .0170 Combining these data we have : Pelouze 73.478i, rb .0050 Marignac 73.4079, .0170 General mean 73.4725, zb .0048 The foregoing figures apply to anhydrous strontium chloride. The ratio between silver and the crystallized salt, SrCl,.6H,O, has also been determined in two series of experiments by Marignac.* Five grammes of salt were used in each estimation, and, in the second series, the per- centage of water was first determined. The quantities of the salt corre- sponding to 100 parts of silver are given in the last column : Series A. 5 grm. SrCl 2 .6H 2 O =4.0515 grm. Ag. 123.411 4.0495 " 123.472 4.0505 " 123.442 Mean, 123.442 Series B. 5 grm. SrCl. 2 . 6 H 2 O 4.0490 grm. Ag. 123.487 4.0500 " 123.457 4.049 " 123.487 Mean, 123.477 Mean of all as one series, 123.460, .0082 * Journ. fur Prakt. Chem., 74, 216. 1858. STRONTIUM. 115 In the same paper Marignac gives two sets of determinations of the percentage of water in crystallized strontium chloride. The first set, cor- responding to u B " above, is as follows : 40.55 6 40.568 40.566 Mean, 40.563 In the second set ten grammes of salt w.ere taken at a time, and the following percentages were found : 40.58 40.59 40.58 Mean, 40.583 Mean of all as one series, 40.573, .0033 The chloride used in the series of estimations last given was subse- quently employed for ascertaining the ratio between it and the sulphate. Converted directly into sulphate, 100 parts of chloride yield the quanti- ties given in the third column : 5.942 grm. SrCl 2 gave 6.887 grm. SrSO 4 . 5-941 " 6.8855 " 5.942 " 6.884 II5-932 i '5-949 H5.927 Mean, 115.936, .004 Richards.* in his study of strontium bromide, followed pretty much the lines laid down in his work on barium. The properties of the bromide itself were carefully investigated, and its purity established beyond reasonable doubt, and then the two usual ratios were deter- mined. First, the ratio Ag 2 : SrBr 2 : : 100 : x, by titration with standard solutions of silver. For this ratio there are three series of measurements, by varied processes, concerning which full details are given. The data obtained, with weights reduced to a vacuum, are as follows : First Series. Wt. Ag. 1.30755 2.10351 2.23357 5-3684 Wt. 1.49962 2.41225 2.56153 6.15663 Ratio. 114.689 114.677 114.683 114.683 Mean, 114.683 * Proc. Amer. Acad. of Sciences, 1894, p. 369. 116 THE ATOMIC WEIGHTS. Second Series. Wt. Ag. 1.30762 2.10322 4-57502 5.3680 .S434 3-3957 3.9607 4.5750 Wt. i. 49962 2.41225 5.24727 6.15663 Third Series. 2.9172 3.8946 4.5426 5-2473 Ratio. 114.683 114.693 114.694 114.691 Mean, 114.690 114.697 1 14.692 114.692 114.695 Mean, 114.694 Mean of all as one series, 114.689, db .0012 For the ratio, measured gravimetrically, 2AgBr : SrBr 2 : : 100 : x, two series of determinations are given : First Series. Wt. AgBr. 2.4415 2.8561 6.9337 2.27625 3.66140 3-88776 9.34497 Wt. SrBr., i. 6086 1.8817 4.5681 Second Series. 1.49962 2.41225 2.56153 6.15663 Ratio. 65.886 65.884 65.883 Mean, 65.884 65.881 65.883 65.887 65.882 Mean, 65.883 Mean of all as one series, 65.884, .0006 For the atomic weight of strontium we now have the subjoined ratios (i.) Ag 2 : SrCl 2 : : 100 : 73.47 2 5, . (2.) Ag 2 : SrCl 2 .6H 2 O : : 100 : 123.460, dr .0082 (3.) Per cent. H 2 O in SrCl 2 .6H 2 O, 40.573, =b .0033 (4.) SrCl 2 : SrSO 4 : : 100 : 115.936, .0040 (5.) Ag 2 : SrBr 2 : : 100 : 114.689, H= .0012 (6.) 2 AgBr : SrBr 2 : : 100 : 65.884, .0006 The antecedent values are O. = 15.879, .0003 Ag 107.108, .0031 Ci = 35.179, .0048 Br = 79-344, .0062 S : 31.828, db .0015 AgBr 186.452, .0054 STRONTIUM. 117 For the molecular weight of SrCl 2 three estimates are available : From (i) SrCI 2 157.390, =}= .0112 From (2) " = 157.197, .0192 From (3).... " = 157-123, .'57 General mean SrCl 2 = 157.281, .0083 For SrBr 2 there are two values : From (5) SrBr 2 = 245.682, .0076 From (6) " = 245.684, rb .0075 General mean SrBr 2 = 245.683, .0053 Finally, with these intermediate data we obtain three independent measures of the atomic weight of strontium, as follows : From molecular weight SrCl 2 Sr = 86.923, .0127 From molecular weight SrBr 2 " = 86.995, =t - OI 35 From ratio (4) " = 86.434, .081 1 General mean Sr = 86.948, .0092 If = 16, Sr = 87.610. Rejection of the third value, which is worth- less, raises these means by 0.01 only. The second value, 86.995, which represents Richards' work, is undoubtedly the best of the three. 118 THE ATOMIC WEIGHTS. BARIUM. For the atomic weight of barium we have a series of eight ratios, estab- lished by the labors of Berzelius, Turner, Struve, Marignac, Dumas, and Richards. Andrews* and Salvetat,f in their papers upon this subject, gave no details nor weighings, and therefore their work may be properly disregarded. First in order, we may consider the ratio between silver and barium chloride, as determined by Pelouze, Marignac, Dumas, and Richards. Pelouze, J in 1845, made the three subjoined estimations of this ratio, using his well known volumetric method. A quantity of pure silver was dissolved in nitric acid, and the amount of barium chloride needed to precipitate it was carefully ascertained. In the last column I give the quantity of barium chloride proportional to 100 parts of silver: 3.860 grin. BaCl 2 ppt. 4.002 grm. Ag. 96.452 5.790 " 6.003 " 9 6 -452 2.895 " 3-Qoi " 96.468 Mean, 96.4573, .0036 Essentially the same method was adopted by Marignac in 1848. His experiments were made upon four samples of barium chloride, as fol- lows. A, commercial barium chloride, purified by recrystallization from water. B, the same salt, calcined, redissolved in water, the solution saturated with carbonic acid, filtered, and allowed to crystallize. C, the preceding salt, washed with alcohol, and again recrystallized, D, the same, again washed with alcohol. For 100 parts of silver the following quantities of chloride were required, as given in the third column : Ag. BaCL. Ratio. ( 3-4445 3-3190 96.356 ) A. \ 3.748o 3.6110 96.345 \ Mean, 96.354 (6.3446 6.1140 96.362 ) | 4.3660 4.1780 96.356 | B. { 4-8390 4.6625 . 96.352 1 Mean, 96.354 j 6.9200 6.6680 96.358 ~j C. { 5.6230 5-4185 96.363 f- Mean, 96.360 ( 5-8435 5.6300 96.346 1 1 8.5750 8.2650 96.384 1 . ) 4.8225 4.6470 96.361 [ [6.8460 6.5980 96.377 J Mean, 96.360 , .0024 * Chemical Gazette, October, 1852. fCompt. Rend., 17, 318. I Compt. Rend., 20, 1047. Journ. fur Prakt. Chern., 35, 73. g Arch, d. Sci. Phys. etNat., 8, 271. BARIUM. 119 Dumas* employed barium chloride prepared from pure barium nitrate, and took the extra precaution of fusing the salt at a red heat in a current of dry hydrochloric acid gas. Three series of experiments upon three samples of chloride gave the following results : L7585 3.8420 2.1585 4.0162 1.6625 2.4987 3.4468 4.0822 4.2062 4.4564 8.6975 2.2957 4.I372 4.2662 4.4764 5.6397 Ratio. 9 6 -303 1 9 6 .339 [ 96.340 | 96.358 j 96.265^ 96-304 96.306 96.290 96.289 96.271 96.307 96.3 l6> | 96.371 96.303 \ 96.3 2 9 96-372J Mean, 96.333 >. Mean, 96.290 Mean, 96.338 Mean, 96.316, dr .0055 The work done by Richards f was of a much more elaborate kind, for it involved some collateral investigations as to the effect of heat upon barium chloride, etc. Every precaution was taken to secure the spectro- scopic purity of the material, which was prepared from several sources, and similar care was taken with regard to the silver. For details upon these points the original paper must be consulted. As for the titrations, three methods were adopted, and a special study was made with refer- ence to the accurate determination of the end point ; in which particular the investigations of Pelouze, Marignac, and Dumas were at fault. In the first series of determinations, silver was added in excess, and the latter was measured with a standard solution of hydrochloric acid. The end point was ascertained by titrating backward and forward with silver solution and acid, and was taken as the mean between the two apparent end points thus observed. The results of this series, with weights reduced to vacuum standards, were as follows : AS- Bad,. Ratio. 6.1872 5.9717 96.517 5.6580 5-4597 96.495 3-5988 3.4728 96.499 9.4010 9.0726 96.507 .7199 .6950 96.541 Mean, 96.512, d= 0055 *Ann. Chem. Pharm., 113, 22. 1860. Ann. Chim. Phys. (3), 55, 129. 120 THE ATOMIC WEIGHTS. In the second series of experiments a small excess of silver was added as before, and the precipitate of silver chloride was removed by filtra- tion. The filtrate and wash waters were concentrated to small bulk whereupon a trace of silver chloride was obtained and taken into account. The excess of silver remaining was then thrown down as silver bromide, and from the weight of the latter the silver was calculated, and sub- tracted from the original amount. Ag. BaCl T Ratio. 6.59993 6.36974 96.512 5-552 2 9 5-3 6oi 96.539 4.06380 3.92244 96.522 Mean, 96.524, .0054 The third series involved mixing solutions of barium chloride and silver in as nearly as possible equivalent amounts, and then determining the actual quantities of silver and chlorine left unprecipitated. The filtrate and wash waters were divided into two portions, one-half being evaporated with hydrobromic acid and the other with silver nitrate. The small amounts of silver bromide and chloride thus obtained were determined by reduction and the use of Volhard's method : Ag. BaCl v Ratio. 4-4355 4.2815 96.528 2.7440 2.6488 96.531 6.1865 5-97 12 96.520 3 4023 3.2841 96.526 Mean, 96.526, .0035 Two final experiments were carried out by Stas' method, somewhat as in the first series, with variations and greater refinement in the observa- tion of the end point. The results were as follows : Ag. Bad*. Ratio. 6.7342 6.50022 96.525 10.6023 IO - 2 3365 96.523 Mean, 96.524, .0007 A careful study of Richards' paper will show that, although the last two experiments are probably the best, they are not entitled to such pre- ponderance of weight as the " probable error" here computed would give them. I therefore treat Richards' work as I have already done that of Marignac and Dumas, regarding all of his series as one, which gives for the value of the ratio 96.520, .0025. This combines with the previous series thus : BARIUM. 121 Pelouze 96.457, rfc .0036 Marignac 96.360, .0024 Dumas 96.316, db .0055 Richards 96.520, .0025 General mean .................... 96.434, .0015 The ratio between silver and crystallized barium chloride has also been fixed by Marignac.* The usual method was employed, and two series of experiments were made, in the second of which the water of crys- tallization was determined previous to the estimation. Five grammes of chloride were taken in each determination. The following quantities of BaCl 7 .2H 2 O correspond to 100 parts of silver : 113.109") A. J 113.135 V Mean, 113.114 - B. J 113.122 V- Mean, 113.106 (113.060) Mean, 113.110, .0079 The direct ratio between the chlorides of silver and barium has been measured by Berzelius. Turner, and Richards. Berzelius t found of barium chloride proportional to 100 parts of silver chloride 72.432 72.422 Mean, 72.427 Turner J made five experiments, with the following results : 72.754 72.406 72.622 72.664 72.653 Mean, 72.680, .0154 Of these, Turner regards the fourth and fifth as the best ; but for present purposes it is not desirable to so discriminate. Richards' determinations fall into three series, and all are character- ized by their taking into account chloride of silver recovered from the wash waters. In the first series the barium chloride was ignited at low redness in air or nitrogen ; in the second series it was fused in a stream of pure hydrochloric acid ; and in the third series it was not ignited at all. In the last series it was weighed in the crystallized state, and the * Tourn. fur Prakt. Chem., 74, 212. 1858. t Poggend. Annalen, 8, 177. t Phil. Trans., 1829, 291. \ Proc. Amer. Acad., 29, 55, 1893. 122 THE ATOMIC WEIGHTS. amount of anhydrous chloride was computed from the data so obtained. The data, corrected to vacuum standards, are as follows : AgCl. Bad*. Ratio. ( 8.7673 6.3697 72.653 } I 5-1979 3.7765 72.654 A. 1 4.9342 3.5846 72.648 ^ Mean, 72.649 | 2.0765 1.5085 72.646 | U-427I 3.2163 72.650 J 2.09750 1.52384 72-650 ^ B. ^7.37610 5.36010 72.669 V- Mean, 72.6563 5.39906 3-92244 72.650 ) 8.2189 5.97123 72.6524 1 4.5199 3.28410 72.6587} P an ' 72 -' Mean, 72.653, .0014 If we assign Berzelius' work equal weight with that of Turner, the three series representing the ratio 2AgCl : BaCl 2 combine as follows Berzelius 72.427, =b .01 54 Turner 72.680, .0154 Richards 72.653, .0014 General mean 72.650, i .0014 Incidentally to some of his other work, Marignac* determined the percentage of water in crystallized barium chloride. Two sets of three experiments each were made, the first upon five grammes and the socond upon ten grammes of salt. The following are the percentages obtained : f 14.79*0 A. J 14.796 y Mean, 14.795 (14.800) c 14.80 S . B. 1 14.81 C Mean, 14.803 (14-80 ) Mean, 14.799, .0018 The ratio between barium nitrate and barium sulphate has been de- termined only by Turner, f According to his experiments 100 parts of sulphate correspond to the following quantities of nitrate : 112.060 111.990 112.035 Mean, 112.028, .014 For the similar ratio between barium chloride and barium sulphate, there are available determinations by Turner, Berzelius, Struve, Marignac, and Richards. * Journ. fur Prakt. Chem., 74, 312. 1858. fPhil. Trans., 1833. 538. BARIUM. 123 Turner * found that 100 parts of chloride ignited with sulphuric acid gave 112.19 parts of sulphate. By the common method of precipitation and nitration a lower figure was obtained, because of the slight solubility of the sulphate. This point bears directly upon many other atomic weight determinations. Berzelius,f treating barium chloride with sulphuric acid, obtained the following results in BaS0 4 for 100 parts of BaCl 2 : 112.17 112.18 Mean, 112.175 Struve, I in two experiments, found : 112.0912 112.0964 Mean, 1 12.0938 Marignac's three results are as follows : 8.520 grm. BaCI 2 gave 9.543 BaSO 4 . Ratio, 112.007 8.519 9.544 " " 112.032 8.520 " 9-542 " " ui-995 Mean, 112.011, .0071 Richards, in his work on this ratio, regards the results as of slight value, because of the occlusion of the chloride by the sulphate. This source of error he was never able to avoid entirely. Another error in the opposite direction is found in the retention of sulphuric acid b} r the precipitated sulphate. Eight experiments were made in two series, one set by adding sulphuric acid to a strong solution of barium chloride in a platinum crucible, the other by precipitation in the usual way. Rich- ards gives in his published paper only the end results and the mean of his determinations ; the details cited below I owe to his personal kind- ness. The weights are reduced to vacuum standards : Bad.,. BaSO* Ratio. 1.78934 2.0056 112.086 2.07670 2.3274 112.072 1.58311 i.774i 112.064 3.27563 3- 6 7i2 112.076 3.02489 3-393 112.080 3.87091 4.3385 112.080 (3.02489 3-9726 112.076 nd - (3,87091 3.4880 112.085 Mean, 112.077, .0017 * Phil. Trans., 1829, 291. t Poggend. Annalen, 8, 177. 1 Ann. Cheni. Pharm., 80, 204. 1851. g Journ. fi'ir Prakt. Chem., 74, 212. 1858. First. 124 THE ATOMIC WEIGHTS. This mean is subject to a small correction due to loss of chlorine on drying the chloride, which reduces it to 112.073. Omitting Turner's single determination as unimportant, and assigning to the work of Ber- zelius and of Struve equal weight with that of Marignac, the measure- ments of this ratio combine thus : Berzelius 112.175, =t .0071 Struve ii 2.094, =t .7 r Marignac... 112.011,^.0071 Richards 1 12.073, .0017 General mean 112.075, .0016 In an earlier paper than the one previously cited, Richards* studied with great care the ratios connecting barium bromide with silver and silver bromide. The barium bromide was prepared by several distinct processes, its behavior upon dehydration and even upon fusion'was studied, and its specific gravity was determined. The ratio with silver was measured by titration, a solution of hydrobromic acid being used for titrating back. The data are subjoined, with the BaBr 2 equivalent to 100 parts of silver stated : BaBr T Ag. Ratio. 2.28760 1.66074 137.746 3.47120 2.52019 I37-73 6 2.19940 1.59687 I37.73 2 2 -3597i i.7'3 2 3 '37-735 2.94207 2.13584 137-748 1.61191 1.17020 137.747 2.10633 i.5 2 92i 137.740 2.19682 2.11740 137.755 237290 1.72276 137.738 1.84822 L34I75 137.747 5.66647 4.11360 I37.75 3.52670 2.56010 37.756 4-3 l6 90 3- I 343 I37-73 1 3-36635 2.44385 137.748 3.46347 2.51415 137-759 Mean, 137.745, .0015 The silver bromide in most of these determinations, and in some others, was collected and weighed in a Gooch crucible with all necessary pre- cautions. Vacuum standards were used throughout for both ratios. I give in a third column the BaBr 2 equivalent to 100 parts of AgBr : Proc. Amer. Acad., 28. 1893. BARIUM. 125 AgBr. Ratio. 2.28760 2.89026 79-149 3-47120 4.3*635 79.136 3.81086 4.81688 79.133 2.35971 2.98230 79-124 2.94207 3-71809 79-129 2.10633 2.66191 79.128 2.91682 3.68615 79.129 2.37290 2.99868 79.131 1.84822 2.33530 79.143 1.90460 2.40733 79.116 5.66647 7.16120 79.127 3.52670 4.45670 79-133 2.87743 3-63644 79-127 3.46347 4-37669 79.135 Mean, 79.132, .0015 The ratios for barium now sum up as follows: (I.) Ag 2 : BaC) 2 : : 100 : 96.434, .0015 (2.) Ag 2 : BaCl 2 .2H 2 O : : 100 : 113.110, .0079 (3.) 2AgCl : BaG 2 : : 100 : 72.650, =fc .0014 (4.) Per cent, of H 2 O in BaCl 2 .2H 2 O, 14.799, =b .0018 (5.) BaSO 4 : BaN 2 O 6 : : 100 : 112.028, .014 (6.) BaCl 2 : BaSO 4 : : 100 : 112.075, =h .0016 (7.) Ag 2 : BaBr 2 : : 100 : 137-745, .0015 (8.) 2AgBr : BaBr 2 : : 100 : 79.132, .0015 The reduction of these ratios depends upon the subjoined antecedent values : Ag= 107.108, .0031 N = 13.935, =b .0021 Cl = 35.179,^.0048 S == 31.828, .0015 Br = 79.344, .0062 AgCl = 142.287, dz .0037 O = 15.879, .0003 AgBr = 186.452, .0054 With these factors four estimates are obtainable for the molecular weight of barium chloride : From (i) BaCl 2 = 206.577, .0068 From (2) " = 206.542, .0183 From (3) " 206.745, .0067 From (4) " = 205.866, .0257 General mean BaCl 2 = 206.629, .0045 For barium bromide we have : From (7) BaBr 2 295.070, .0091 From (8) " =295.086,^.0102 General mean BaBr 2 = 295.078, .0068 126 THE ATOMIC WEIGHTS. And for barium itself, four values are finally available, thus : From molecular weight BaCl 2 Ba = 136.271, .0106 From molecular weight BaBr. 2 " = 136.390, .0141 From ratio (5) " 135.600, rb .2711 From ratio (6) " = 136.563, .0946 General mean Ba = 136.315, d= .0085 Or, if = 16, Ba = 137.354. In the foregoing computation all the data, good or bad, are included. Some of them, as shown -by the weights, practically vanish ; but others, as in the chloride series, carry an undue influence. A more trustworthy result can be deduced from Richards' experiments alone, which reduce as follows : From Ag 2 : BaCl 2 BaCl 2 = 206.761, .0080 From 2AgCl : BaCl 2 " = 206.754, .0067 General mean BaCl 2 = 206.755, From the bromide, as given above, Ba = 136.390, dz .0141. From the value just found for the chloride, Ba 136.397, .0109. Combining the two values Ba = 136.392, .0086. Or, if = 16, Ba = 137.434. This determination will be adopted in subsequent calculations as the most probable. LEAD. 127 LEAD. For the atomic weight of lead we have to consider experiments made upon the oxide, chloride, nitrate, and sulphate. The researches of Ber- zelius upon the carbonate and various organic salts need not now be considered, nor is it worth while to take into account any work of his done before the year 1818. The results obtained by Dobereiner* and by Longchamp f are also without special present value. For the exact composition of lead oxide we have to depend upon the researches of Berzelius. His experiments were made at different times through quite a number of years ; but were finally summed up in the last edition of his famous *' Lehrbuch." J In general terms his method of experiment was very simple. Perfectly pure lead oxide was heated in a current of hydrogen, and the reduced metal weighed. From his weighings I have calculated the percentages of lead thus found and given them in a third column : Earlier Results. 8.045 g rm - PbO S ave 74675 grm. Pb. 92.8217 per cent. 14.183 " 13.165 " 92.8224 " 10.8645 " 10.084 " 92.8160 " 13.1465 " 12.2045 " 92.8346 " 21.9425 " 20.3695 " 92.8313 " 11.159 " IO -359 " 92.8309 " Latest. 6.6155 6.141 " 92.8275 " 14.487 " 13.448 " 92.8280 " 14.626 <( 13-5775 " 92.8313 " Mean, 92.8271, .0013 For the synthesis of lead sulphate we have data by Berzelius, Turner, and Stas. Berzelius, whose experiments were intended rather to fix the atomic weight of sulphur, dissolved in each estimation ten grammes of pure lead in nitric acid, then treated the resulting nitrate with sul- phuric acid, brought the sulphate thus formed to dryness, and weighed. One hundred parts of metal yield of PbS0 4 : 146.380 146.400 146.440 146.458 Mean, 146.419, .012 * Schweig. Journ., 17, 241. 1816. f Ann. Chim. Phys., 34, 105. 1827. t Bd. 3, s. 1218. I I^ehrbuch, sth ed., 3, 1187. 128 THE ATOMIC WEIGHTS, Turner,* in three similar experiments, found as follows : 146.430 146.398 146.375 Mean, 146.401, .on In these results of Tamer's, absolute weights are implied. The results of Stas' syntheses,t effected after the same general method, but with variations in details, are as follows. Corrections for weighing in air were applied : 146.443 146.427 146.419 146.432 146.421 146.423 Mean, 146.4275, .0024 Combining, we get the subjoined result: Berzelius 146.419, .012 Turner 146.401, .01 1 Stas 146.4275, .0024 General mean 146.4262, .0023 Turner, in the same paper, also gives a series of syntheses of lead sul- phate, in which he starts from the oxide instead of from the metal. One hundred parts of PbO, upon conversion into PbS0 4 , gained weight as follows : 35-84 35-71 35.84 35-75 35-79 35.78 35.92 Mean, 35.804, .018 These figures are not wholly reliable. Numbers one, two, and three represent lead oxide contaminated with traces of nitrate. The oxide of four, five, and six contained traces of minium. Number seven was free from these sources of error, and, therefore, deserves more consideration. The series as a whole undoubtedly gives too low a figure, and this error would tend to slightly raise the atomic weight of lead. *Phil. Trans., 1833, 527-538. t Aronstein's translation, 333. LEAD. 129 Still a third series by Turner establishes the ratio between the nitrate and the sulphate, a known weight of the former being in each experi- ment converted into the latter. One hundred parts of sulphate represent of nitrate: 109.312 109.310 109.300 Mean, 109.307, .002 In all these experiments by Turner the necessary corrections were made for weighing in air. In 1846 Marignac* published two sets of determinations of only moderate value. First, chlorine was conducted over weighed lead, and the amount of chloride so formed was determined. The lead chloride was fused before weighing. The ratio to 100 Pb is given in the last column : 20.506 grm. Pl> gave 27.517 PbCl 2 . 134.190 16.281 " 21.858 " 134.225 25.454 34.H9 " '34.159 Mean, 134.19^ .013 Secondly, lead chloride was precipitated by silver nitrate and the ratio between PbCl, and 2AgCl determined. The third column gives the AgCl formed by 100 parts of PbCl 2 : 12.534 grm. PbCl 2 gave 12.911 AgCl. 103.01 14.052 14.506 " JO3.23 25.533 " 26.399 " 103.39 Mean, 103.21, .0745 For the ratio between lead chloride and silver we have a series of re- sults by Marignac and one experiment by Dumas. There are also un- available data by Turner and by Berzelius. Marignac,t applying the method used in his researches upon barium and strontium, and working with lead chloride which had been dried at 200, obtained these results. The third column gives the ratio between PbCl 2 , and 100 parts of Ag: 4.9975 grm. PbCl 2 = 3.8810 grm. Ag. 128.768 4.9980 " 3.8835 " 128.698 5.0000 3.8835 " 128.750 5.0000 " 3.8860 " 128.667 Mean, 128.721, .016 Dumas, J in his investigations, found that lead chloride retains traces *Aun. Chern. Pharrn., 59, 289; and 290. 1846. t Journ. fiir Prakt. Chem., 74, 218. 1858. I Ann. Chem. Pharm., 113, 35. 1860. 130 THE ATOMIC WEIGHTS. of water even at 250, and is sometimes also contaminated with oxychlo- ride. In one estimation 8.700 grammes PbCl 2 saturated 6.750 of Ag. The chloride contained .009 of impurity ; hence, correcting, Ag : PbCI 2 : : 100 : 128.750. If we assign this figure equal weight with those of Marig- nac, we get as the mean of all 128.7266, .013. The sources of error in- dicated by Dumas, if they are really involved in this mean, would tend slightly to raise the atomic weight of lead. The synthesis of lead nitrate, as carried out by Stas,* gives excellent results. Two series of experiments were made, with from 103 to 2pO grammes of lead in each determination. The metal was dissolved in nitric acid, the solution evaporated to dryness with extreme care, and the nitrate weighed. All weighings were reduced to the vacuum standard. In series A the lead nitrate was dried in an air current at a temperature of about 155. In series B the drying was effected in vacuo, 100 of lead yield of nitrate : A. 159-973 159.975 159.982 159-975 159.968 J59-973 Mean, 159.9743, =fc .0012 159.970 159.964 159-959 I59-965 Mean, 159.9645, .0015 Mean from both series, 159.9704, 2 .0010 There is still another set of experiments upon lead nitrate, originally intended to fix the atomic weight of nitrogen, which may properly be included here. It was carried out by Anderson f in Svanberg's labora- tory, and has also appeared under Svanberg's name. Lead nitrate was carefully ignited, and the residual oxide weighed, with the following results : 5.19485 grm. PbN 2 O 6 gave 3.5017 grm. PbO. 67.4071 per cent. 9.7244 6.5546 " 67.4037 " 9.2181 6.2134 " 67.4044 " 9.6530 6.5057 " 67.3957 Mean, 67.4027, i .0016 * Aronsteiii's translation, 316. t Ann. Chim. Phys. (3), 9, 254. 1843. LEAD. 131 We have now nine ratios from which to compute : (i.) Per cent, of Pb in PbO, 92.8271, .0013 (2.) Per cent of PbO in PbN 2 O 6 , 67.4027, .0016 (3.) Pb : PbSO 4 : : 100 : 146.4262, .0023 (4.) PbO : PbSO 4 : : 100 : 135.804, .0180 (5.) PbSO 4 : PbN 2 O 6 : : 100 : 109.307, .0020 (6.) Pb : PbN 2 O 6 : : iqo : 159.9704, .0010 (7.) Pb : PbC) 2 : : 100 : 134.191, .013 (8.) PbCl 2 : 2AgCl : : 100 : 103.21, .0745 (9.) Ag 2 : PbCl 2 : : 100 : 128.7266, db .0130 To reduce these ratios we must use the following data : O =. 15.879, .0003 s = 31.828, .0015 Ag= 107.108, =b .0031 N 13.935,^.0021 Cl == 35.179, db .0048 AgCl= 142.287, .0037 For the molecular weight of lead oxide we now get three estimates : From (i) PbO = 221.375, d= .0403 From (2) " 221.796, .0132 From (4) " = 221.944, d= .1116 General mean PbO = 221.757, =b .0125 For lead chloride we have From (8) PbCl 2 = 275.723, .1989 From (9) " = 275.753,^.0290 General mean PbO 2 = 275.752, dr .0287 Including these results, six values are calculable for the atomic weight of lead : From molecular weight of PbO Pb = 205.878, dr .0126 From molecular weight of PbCl 2 " = 205.394, .0302 From (3) " = 205.367, dr .0051 From (5) " = 203.352, .0479 From (6) " = 205.341, db .0068 From (7) " = 205.779, .0831 General mean Pb = 205.395, .0038 If = 16, Pb = 206.960. If we reject the first, fourth, and sixth of these values, which are untrustworthy, the remaining second, third, and fifth give a general mean of Pb = 205.358, .0040. If O = 16, this becomes Pb = 206.923. From Stas' ratios alone Stas calculates Pb = 206.918 to 206.934 ; Ostwald finds 206.911 ; Van der Plaats (A), 206.9089, (B), 206.9308, and Thomson 206.9042. The value adopted here repre- sents mainly the work of Stas, and with H = 1 is Pb = 205.358, .0040. 132 THE ATOMIC WEIGHTS. GLUCINUM. Our knowledge of the atomic weight of glucinum is chiefly derived from experiments made upon the sulphate. Leaving out of account the single determination by Berzelius, * we have to consider the data fur- nished by Awdejew, Weeren, Klatzo, Debray, Nilson and Pettersson, and Kriiss and Moraht. Awdejew, f whose determination was the earliest of any value, analyzed the sulphate. The sulphuric acid was thrown down as barium sulphate ; and in the nitrate, from which the excess of barium had been first re- moved, the glucina was precipitated by ammonia. The figures which Awdejew publishes represent the ratio between S0 3 and G10, but not absolute weights. As, however, his calculations were made with S0 3 = 501.165, and Ba probably 855.29, we may add a third column showing how much BaS0 4 is proportional to 100 parts of G10 : SO S . GIO. Ratio. 4457 !4o 921.242 4531 1420 9 2 7.304 7816 2480 9 I 5-93 12880 4065 920 814 Mean, 921.316, LS77 The same method was followed by Weeren and by Klatzo, except that Weeren used ammonium sulphide instead of ammonia for the precipita- tion of the glucina. Weeren J gives the following weights of GIO and BaS0 4 . The ratio is given in a third column, just as with the figures by Awdejew : GIO. BaSO. Ratio. .3163 2.9332 927.031 .2872 2.6377 918 419 .2954 2.7342 9 2 5-592 .5284 4.8823 902.946 Mean, 918.497, =b 3.624 Klatzo's figures are as follows, with the third column added by the writer : GIO. BaSO. Ratio. .2339 2.1520 920.052 .1910 J-7556 919.162 .2673 2.4872 930-49 3585 3-3"5 9 2 3.7io .2800 2.5842 922.989 Mean, 923.281, 1.346 * Poggend. Annal., 8, i. t Poggend. Aiinal., 56, 106. 1842. t Poggend. Aiinal., 92, 124. 1854. g Zeitschr. Anal. Chem., 8, 523. 1869. GLUCINUM. 133 Combining these series into a general mean, we get the subjoined result : Awdejew 921.316, L577 Weeren 9 l8 -497, 3.624 Klatzo 923.281, -_h 1.346 General mean 922.164, dr 0.985 Hence G10 = 25.130, .0269. Debray* analyzed a double oxalate of glucinum and ammonium, G1(NH 4 ) 2 C 4 8 . In this the glucina was estimated by calcination, after first converting the salt into nitrate. The following percentages were found : ii.5 II. 2 ii. 6 Mean, 11.433, d= .081 The carbon was estimated by an organic combustion. I give the weights, and put in a third column the percentages of CO 2 thus obtained : Salt. CO* Per cent. CO V .600 .477 79 500 .603 .478 79.270 .600 .477 79-5 Mean, 79.423, .052 Calculating the ratio between C0 2 and G10, we have for the molecular weight of the latter, G1O = 25.151, .1783. In 1880 the careful determinations of Nilson and Pettersson appeared.f These chemists first attempted to work with the sublimed chloride of glucinum, but abandoned the method upon finding the compound to be contaminated with traces of lime derived from a glass tube. They finally resorted to the crystallized sulphate as the most available salt for their purposes. This compound, upon strong ignition, yields pure glucina. The data are as follows : GISO^H.,O. GIO. Percent. GIO. 3-8014 .5387 2.6092 -3697 > 4. 307 2 .6099 3.0091 .4266 Mean, 14.169, .0023 Kriiss and MorahtJ in their work follow the general method adopted *Ann. Chim. Phys. ($\ 44, 37- l8 55- f Berichte d. Deutsch. Chem. Gesell., 13, 1451. 1880. J Ann. d. Chem., 262, 38. 1891. - -_-_._ v ---: :-:.-: ' ; --- ; : - - ' ;- -55 -':- - --- - ' - : - - - C \\jyx . : - : = -: 5 = ~ : : e I: 0= 1- >- - - ">': Tl :: ... - : . -- :-:,-:. . -" :, :-; .,. .,. 136 THE ATOMIC WEIGHTS. In a later note* Scheerer shows that the barium sulphate of these ex- periments carries down with it magnesium salts in such quantity as to make the atomic weight of magnesium 0.039 too low. The work of Bahr, Jacquelain, Macdonnell, and Marignac, and in part that of Svanberg and Nordenfeldt, also relates to the composition of magnesium sulphate. Jacquelain's experiments were as follows : f Dry magnesium sulphate was prepared by mixing the ordinary hydrous salt to a paste with sul- phuric acid, and calcining the mass in a platinum crucible over a spirit lamp to constant weight and complete neutrality of reaction. This dry sulphate was weighed and intensely ignited three successive times. The weight of the residual MgO having been determined, it was moistened with sulphuric acid and recalcined over a spirit lamp, thus reproducing the original weight of MgS0 4 . Jacquelain's weighings for these two experiments show that 100 parts of MgO correspond to the quantities of MgS0 4 given in the last column : 1.466 grm. MgSO 4 gave .492 grm. MgO. 297.968 .492 " MgO " 1.466 " MgSO 4 . 297.968 Jacquelain also made one estimation of sulphuric acid in the foregoing sulphate as BaS0 4 . His result (1.464 grm. MgS0 4 = 2.838 grm. BaSOJ, reduced to the standard adopted in dealing with Scheerer's experiments, gives for 100 parts of MgS0 4 , 193.852 BaS0 4 . If this figure be given equal weight with a single experiment in Scheerer's series, and combined with the latter, the mean will be 193.700, .0331. This again is subject to the correction pointed out by Scheerer for magnesium salts retained by the barium sulphate, but such a correction determined by Scheerer for a single experiment is only a rough approximation, and hardly worth applying. The determinations published by Macdonnell J are of slight impor- tance, and all depend upon magnesium sulphate. First, the crystallized salt, MgS0 4 .7H 2 0, was dried in vacuo over sulphuric acid and then de- hydrated at a low red heat. The following percentages of water were found : 5^7 51.14 51.26 51.28 5 r - 2 9 Mean, 51.21, .020 *Poggend. Annalen, 70, 407. f Ann. Chim. Phys. (3), 32, 202. J Proc. Royal Irish Acad., 5, 303. British Association Report, 1852, part 2, p. 36. MAGNESIUM. 137 Secondly, anhydrous magnesium sulphate was precipitated with ba- rium chloride. From the weight of the barium sulphate, with S0 3 = 80 and Ba = 137, Macdonnell computes the percentages of S0 3 given below. I calculate them back to the observed ratio in uniformity with Scheerer's work : Per cent. SO,. Ratio, MgSO : 66.67 194.177 66.73 '94-35 1 66.64 194.089 66.65 194.118 66.69 194-239 In another experiment 60.05 grains MgS0 4 gave 116.65 grains BaS0 4 , a ratio of 100 : 194.254. Including this with the preceding figures, they give a mean of 194.205, .027. This, combined with the work of Scheerer and Jacquelain, 193.700, .033, gives a general mean of MgSO 4 : BaSO 4 : : 100 : 194.003, .021. In one final experiment Macdonnell found that 41.44 grains of pure magnesia gave 124.40 grains of MgSO 4 , or 300.193 per cent. Bahr's * work resembles in part that of Jacquelain. This chemist converted pure magnesium oxide into sulphate, and from the increase in weight determined the composition of the latter salt. From his weigh- ings 100 parts of MgO equal the amounts of MgS0 4 given in the third column : 1.6938 grm. MgO gave 5.0157 grm. MgSO 4 . 296.122 2.0459 " 6.0648 " 296.437 1.0784 " 3.I9 2 5 " 296.040 Mean, 296.200, d= .0815 About four years previous to the investigations of Bahr the paper of Svanberg and Nordenfeldtf appeared. These chemists started with the oxalate of magnesium, which was dried at a temperature of from 100 to 105 until it no longer lost weight. The salt then contained two molecules of water, and upon strong ignition it left a residue of MgO. The percentage of MgO in the oxalate comes out as follows : 7.2634 grm. oxalate gave 1.9872 grm. oxide. 27.359 per cent. 6-3795 1.7464 27-375 " 6.3653 1.7418 27-364 " 6.2216 1.7027 27.368 " Mean, 27.3665, zb .OO2 3 * Journ. fur Prakt. Chem., 56, 310. 1852. f Journ. fi'ir Prakt. Chem., 45, 473. 1848. 138 THE ATOMIC WEIGHTS. In three of these experiments the MgO was treated with H 2 S0 4 , and converted, as by Jacquelain and by Bahr in their later researches, into MgS0 4 . One hundred parts of MgO gave of MgSO 4 as follows : 1.9872 grin. MgO gave 5.8995 grm. MgSO 4 . 296.875 1.7464 " 5- x 783 " 296.513 1.7418 " 5.1666 " 296.624 Mean, 296. 67 r, .072 In 1850 the elaborate investigations of Marchand and Scheerer * ap- peared. These chemists undertook to determine the composition of some natural magnesites, and, by applying corrections for impurities, to deduce from their results the sought-for atomic weight. The magnesite chosen for the investigation was, first, a yellow, transparent variety from Snarum ; second, a white opaque mineral from the same locality ; and,, third, a very pure quality from Frankenstein. In each case the im- purities were carefully determined ; but only a part of the details need be cited here. Silica was of course easily corrected for by simple sub- traction from the sum of all of the constituents; but iron and calcium,, when found, having been present in the mineral as carbonates, required the assignment to them of a portion of the carbonic acid. In the atomic weight determinations the mineral was first dried at 300. The loss in weight upon ignition was then carbon dioxide. It was found, however, that even here a correction was necessary. Magnesite, upon drying at 300, loses a trace of C0 2 , and still retains a little water ; on the other hand, a minute quantity of C0 2 remains even after ignition. The C0 2 expelled at 300 amounted in one experiment to .054 per cent. ; that retained after calcination to .055 per cent. Both errors tend in the same direction, and increase the apparent percentage of MgO in the magnesite. On the yellow mineral from Snarum the crude results are as follows, giving percentages of MgO, FeO, and CO 2 after eliminating silica : CO.,. MgO. FeO. 51.8958 47-3278 .7764 51.8798 47-3393 -7809 51.8734 47.3154 -8112 5'-*8 7 5 47.3372 .7753 Mean, 47.3299, .0037 After applying corrections for loss and retention of C0 2 , as previously indicated, the mean results of the foregoing series become CO.,. MoO. FeO. 51.9931 47.2743 -7860 The ratio between the MgO and the C0 2 , after correcting for the iron, will be considered further on. * Journ. fi'ir Prakt. Chem., 50, 385. MAGNESIUM. 139 Of the white magnesite from Snarum but a single analysis was made, which for present purposes may be ignored. Concerning the Franken- stein mineral three series of analyses were executed. In the first series the following results were obtained : 8.996 grm. CO 2 = 8.2245 grm. MgO. 47.760 per cent. MgO. 7-960 " 7.2775 " 47.76i 9-3 2 65 8.529 47.767 7-553 " 6.9095 " 47-775 Mean, 47.766, .0022 This mean, corrected for loss of C0 2 in drying, becomes 47.681. I give series second with corrections applied : 6.8195 S rm - MgCO 3 gave 3.2500 grm. MgO. 47.658 per cent. 11.3061 " - 5-3849 " 47.628 " 9-7375 " 4-635 (( 47-599 " 12.3887 5.9033 47.650 32.4148 15.453 47-674 " 38.8912 18.5366 " 47.663 " 26.5223 12.6445 " 47.675 " Mean, 47.650, d= .0069 The third series was made upon very pure material, so that the cor- rections, although applied, were less influential. The results were as follows : 4.2913 grm. MgCO 3 gave 2.0436 grm. MgO. 47.622 per cent. 27.8286 " 13-2539 " 47.627 " 14.6192 " 6.9692 " 47.672 " 18.3085 '< 8.7237 " 47-648 " Mean, 47.642, =fc .0077 In a supplementary paper* by Scheerer, it was shown that an impor- tant correction to the foregoing data had Been overlooked. Scheerer, re- examining the magnesites in question, discovered in them traces of lime, which had escaped notice in the original analyses. With this correction the two magnesites in question exhibit the following mean composition : Snarum. Frankenstein. C0 2 52.131 52.338 MgO 46.663 47-437 CaO 430 . 225 FeO. 776 100.000 100.000 Correcting for lime and iron, by assigning each its share of C0 2 , the Snarum magnesite gives as the true percentage of magnesia in pure * Ann. d. Chem. und Pharm., no, 240. 140 THE ATOMIC WEIGHTS. magnesium carbonate, the figure 47.624. To this, without serious mis- take, we may assign the weight indicated by the probable error, .0037, the quantity previously deduced from the percentages of MgO given in the unconnected analyses. From the Frankenstein mineral, similarly corrected, the final mean percentage of MgO in MgC0 3 becomes 47.628. This, however, represents three series of analyses, whose combined probable errors may be prop- erly assigned to it. The combination is as follows : dr .OO22 dz .0069 2 .0077 Result, .0020, probable error of the general mean. We may now combine the results obtained from both magnesites: Snarum mineral Per cent. MgO, 47.624, .0037 Frankenstein mineral " 47.628, d= .0020 General mean Per cent. MgO, 47.627, .0018 The next investigation upon the atomic weight of magnesium which we have to consider is that of Dumas. * Pure magnesium chloride was placed in a boat of platinum, and ignited in a stream of dry hydrochloric acid gas. The excess of the latter having been expelled by a current of dry carbon dioxide, the platinum boat, still warm, was placed in a closed vessel and weighed therein. After weighing, the chloride was dissolved and titrated in the usual manner with a solution containing a known quantity of pure silver. The weighings which Dumas reports give, as proportional to 100 parts of silver, the quantities of MgCL 2 stated in the third column : 2.203 8 rm - MgCl 2 = 4.964 grm. Ag. 44.380 2.5215 " 5.678 " 44.408 2.363 5-325 " 44-376 3.994 " 9.012 " 44.319 2.578 5.834 " 44.189 2.872 " 6.502 " 44.i7t 2.080 4Jio " 44.161 2.214 " 5- 2 " 44.262 2.086 " 4.722 " 44.176 1.688 <( 3823 " 44.154 1.342 " 3.031 " 44.276 Mean, 44.261, dz .020 This determination gives a very high value to the atomic weight of magnesium, which is unquestionably wrong. The error, probably, is due to the presence of oxychloride in the magnesium chloride taken, an *Ann. Chem. Pharm., 113, 33. 1860. MAGNESIUM. 141 impurity tending to raise the apparent atomic weight of the metal. Richards 1 and Parker's revision of this ratio is more satisfactory. Marignac, * in 1883, resorted to the old method of determination, de- pending upon the direct ratio between MgO and S0 3 . This ratio he measured both synthetically and analytically. First, magnesia from various sources was converted into sulphate. The MgS0 4 from 100 parts of MgO is given in the third column : MgO. MgSO. Ratio. 4.6620 298.17 4.2025 298.32 4.7480 298.30 4.3855 298.23 4.4060 298.15 4-8530 298.33 4.0740 298.37 5.8390 298.29 5.0600 298.26 5.5715 298.26 Mean, 298.27, .0149 The magnesia for experiments 1 to 5 was prepared by calcination of the nitrate, that of 6 to 8 from the sulphate, and the remaining two from the carbonate. But Richards and Rogers t have shown that magnesia derived from the nitrate always contains occluded gaseous impurity, so that the experiments depending upon its use are somewhat questionable. The results tend to give an atomic weight for magnesium which is pos- sibly too high. Whether the other samples of magnesia are subject to similar objections I cannot say. Marignac's second series was obtained by the calcination of the sul- phate, with results as follows : J .5635 2 . .... .4087 3 .5917 4 47O5 c 4778 6 6267 7 7657 8 .ocyt: Q 606^ 10. . .8680 MgSOv MgO. Ratio. 3-7705 i .2642 298.25 4.7396 i .5884 298.39 3-3830 .1345 298.19 4.7154 .5806 298.33 4.5662 .5302 298.43 4.5640 .5300 298.30 3-2733 .0979 298.14 4.8856 .6378 298.30 5.0092 .6792 298.31 5.3396 .7898 298.33 5.1775 .7352 298.38 5.0126 .6807 298.24 5-0398 .6894 298.32 Mean, 298 30, .0150 * Arch. Sci. Phys. et Nat. (3), 10, 206. f Am. Chem. Journ., 15, 567. 1893. 142 THE ATOMIC WEIGHTS. These data may now be combined with the work of previous investi- gators, giving Macdonnell's one result and Jacquelain's two, each equal weight with a single experiment in Bahr's series: Macdonnell 300.193, .1413 Jacquelain 297.968, -b .0999 Bahr 296.200, .0815 Svanberg and Nordenfeldt 296.671, d= .0720 Marignac, synthetic 298.27, H~ .0149 Marignac, calcination 298.30, dz .0150 General mean 298. 210, .0103 Burton and Vorce,* who published their work on magnesium in 1890, started out with the metal itself, which had been purified by distillation in a Sprengel vacuum. This metal was dissolved in pure nitric acid, and the resulting nitrate was converted into oxide by calcination at a white heat. The oxide was carefully tested for oxides of nitrogen, which were proved to be absent, but occluded gases, the impurity pointed out by Richards and Rogers, were not suspected. This impurity must have been present, and it would tend to lower the apparent atomic weight of magnesium as calculated from the data obtained. The results were as follows, together with the percentage of Mg in MgO : Mg Taken. MgO Formed. Per cent. Mg. .33009 .54766 60.273 .34512 .57 2 5 2 60.281 .26058 .43221 60.290 .28600 .47432 60.297 .30917 .5^273 60.299 .27636 .45 8 53 60.271 .36457 .60475 60.284 .32411 .53746 60.304 .32108 .53263 60.282 .28323 .46988 60.262 Mean, 60.2845, =h .0027 The latest determinations of all are those of Richards and Parker,f who studied magnesium chloride with all the precautions suggested by the most recent researches. The salt itself was not only free from oxy- chloride, but also spectroscopically pure as regards alkaline contamina- tions, and all weighings were reduced to a vacuum standard. The first series of experiments gives the ratio between silver chloride and mag- nesium chloride, and I have reduced the data to the form 2AgCl : MgCl 2 : : 100 : x. The weighings and values for x are subjoined : * Am. Chem. Journ., 12, 219. 1890. fZeitsch. Anorg. Chem., 13, 81. 1896. MAGNESIUM. 143 MgCl*. L 33550 I.5I60I 1.40664 1.25487 AgCl. Ratio. 4.01952 33-225 4.5 6 3 6 9 33-219 3.98528 33.226 4.23297 33-231 3.77670 33 227 Mean, 33.226, .0013 The remaining series of experiments, three in number, relate to the ratio 2Ag : MgCl 2 , which was earlier investigated by Dumas. For the elaborate details of manipulation the original memoir must be consulted. I can give little more than the weights found, and their reduction to the usual form of ratio, Ag 2 : MgCl 2 : : 100 : x : MgCl v 2.78284 2.29360 2-36579 Second Series. Ag. ' 6.30284 5.19560 5.35989 Ratio. 44.152 44.145 44.13 Mean, 44.142, .0043 This series gives slightly higher results than the others, and the authors, for reasons which they assign, discard it : Third Series. MgCl*. 1.99276 1.78870 2.12832 2.51483 2.40672 1.95005 4-05256 4.82174 5.69714 545294 4.41747 Ratio. 44.^31 44.138 44. 140 44.I4I 44- H4 Mean, 44.138, =b .0013 The fourth series, because of the experience gained in the conduct of the preceding determinations, is best of all, and the authors adopt its results in preference to the others : Fourth Series. Ag. Ratio. 4.60855 44-136 4.32841 44.138 4.75635 44.137 4.12447 44.137 4o5 I 5 I 44-138 2.51876 44.138 Mean, 44.137, db .0003 2.03402 1.91048 2.09932 1.82041 1.92065 1.11172 144 THE ATOMIC WEIGHTS. These series combine with that of Dumas as follows : Dumas 44.261, zb .0200 Richards and Parker, second series 44.142, zb .0043 Richards and Parker, third series 44.138, zb .0013 Richards and Parker, fourth series 44. 137, db .0x303 General mean 44. 138, .0003 Here the first two values practically vanish, and the third and fourth series of Richards and Parker appear alone. To sum up, we now have the subjoined ratios, bearing upon the atomic weight of magnesium : (i.) MgSO 4 : BaSO 4 : : IOO : 194.003, .021 (2.) MgO : MgSO 4 : : 100 : 298.210, .0103 (3.) Per cent, of water in MgSO 4 , 7H 3 O, 51.21, .020 (4.) Per cent, of MgO in oxalate, 27.3665, .0023 (5.) Per cent, of MgO in carbonate, 47.627, .0018 (6.) Per cent, of Mg in MgO, 60.2845, .0027 (7.) 2Ag : Mgd 2 : : IOO : 44.138, .0003 (8.) 2AgCl : MgCl 2 : : 100 : 33.226, .0013 To reduce these ratios we have O = 15.879, zb .0003 C . 11.920, zb .0004 Ag=: 107.108, .0031 Ba = 136.392, zb .0086 Cl = 3S> 1 79, -48 AgCl = 142.287, .0037 S = 31.828, zb .0015 For the molecular weight of MgSO 4 , two values are now calculable : From (i) MgSO 4 = 119.450, zb .0137 From (3) " = 119.239,^.0675 General mean MgSO 4 == 119.443, .0135 Hence Mg = 24.099, .0136. For MgO, three values are found : From (2) MgO = 40.091, zb .0023 From (4) " 40.404, dr .0037 From (5) " =: 39 721, rb .0021 General mean MgO = 39.974, =b .0014 Hence Mg = 24.095, .0014. For MgCl 2 there are two values : From (7) MgCl 2 = 94-551, db - OO 32 From (8) " =94.553,^.0044 General mean MgCl 2 = 94.552, .0026 Hence Mg 24.194, .0099. MAGNESIUM. 145 With the aid of these intermediate values, four estimates of the atomic weight of magnesium are available, as follows : From molecular weight of MgSO 4 . ... Mg 24.099, .0136 From molecular weight of MgO " = 24.095, + .0014 From molecular weight of MgCl 2 " = 24. 194, .0099 From ratio (6) " 24. 103, db .0020 General mean Mg = 24. TOO, .001 1 If = 16, this becomes Mg = 24.283. On purely chemical grounds the third of the foregoing values, that derived from magnesium chloride, seems to be the best. I should un- hesitatingly adopt it, rejecting the others, were it not for the fact that it rests upon one compound of magnesium alone, and therefore is not ab- solutely conclusive. It agrees admirably, however, with the sulphate determinations of Marignac, and it is highly probable that it may be fully confirmed later by evidence from other sources. Marignac's data, taken alone, give Mg = 24.197. The fourth series of Richards and Parker, by itself, gives Mg = 24.180. The approximate mean of these, 24.19, may be preferred by many chemists to the general mean derived from all the observations. 10 146 THE ATOMIC WEIGHTS. ZINC. The several determinations of the atomic weight of zinc are by no means closely concordant. The results obtained by Gay-Lussac* and Berzelius f were undoubtedly too low, and may be disregarded here. We need consider only the work done by later investigators. In 1842 Jacquelain published the results of his investigations upon this important constant. J In two experiments a weighed quantity of zinc was converted into nitrate, and that by ignition in & platinum cruci- ble was reduced to oxide. In two other experiments sulphuric acid took the place of nitric. As the zinc contained small quantities of lead and iron, these were estimated, and the necessary corrections applied. From the weights of metal and oxide given by Jacquelain the percent- ages have been calculated : Nitric Series. 9.917 grm. Zn gave 12.3138 grm. ZnO. 80.536 per cent. Zn. 9.809 " 12.1800 " 80.534 " Sulphuric Series. 2 -398 grm. Zn gave 2.978 grm. ZnO. 80.524 3.197 " 3.968 " 80.570 Mean of all four, 80.541, .007 Hence Zn = 65.723. The method adopted by Axel Erdmann is essentially the same as that of Jacquelain, but varies from the latter in certain important details. First, pure zinc oxide was prepared, ignited in a covered crucible with sugar, and then, to complete the reduction, ignited in a porcelain tube in a current of hydrogen. The pure zinc thus obtained was converted into oxide by means of treatment with nitric acid and subsequent igni- tion in a porcelain crucible. Erdmann's figures give us the following percentages of metal in the oxide : 80.247 80.257 80.263 80.274 Mean, 80.260, .0037 Hence Zn = 64.562. * Memoire d'Arceuil, 2, 174. tGilb. Annal., 37, 460. I Compt. Rend., 14, 636. $ Poggend. Annal., 62, 611. Berz. lyehrb., 3, 1219. ZINC. 147 Upon comparing Erdmann's results with those of Jacquelain two points are worth noticing : First, Erdmann worked with purer material than Jacquelain, although the latter applied corrections for the impuri- ties which he knew were present ; secondly, Erdmann calcined his zinc nitrate in a porcelain crucible, while Jacquelain used platinum. In the latter case it has been shown that portions of zinc may become reduced and alloy themselves with the platinum of the crucible ; hence a lower weight of oxide from a given quantity of zinc, a higher percentage of metal, and an increased atomic weight. This source of constant error has undoubtedly affected Jacquelain's experiments, and vitiated his results. In Erdmann's work no such errors seem to be present. Favre * employed two methods of investigation. First, zinc was dis- solved in sulphuric acid, the hydrogen evolved was burned, and the weight of water thus formed was determined. To his weighings I ap- pend the ratio between metallic zinc and 100 parts of water : 25.389 grm. Zn gave 6.928 grm. H 2 O. 366.469 30.369 " 8.297 " 366024 31.776 " 8.671 " 366.463 Mean, 366.319, .088 Hence Zn = 65.494. The second method adopted by Favre was to burn pure zinc oxalate, and to weigh the oxide and carbonic acid thus produced. From the ratio between these two sets of weights the atomic weight of zinc is easily deducible. From Favre 's weighings, if C0 2 = 100, ZnO will be as given in the third column below : 7.796 grm. ZnO = 8.365 grm. CO 2 . 93. 198 7-342 " 7.883 " 93-137 5.2065 " 5.588 " 93-173 Mean, 93.169, .012 Hence Zn = 65,521. Both of these determinations are open to objections. In the water series it was essential that the hydrogen should first be thoroughly dried before combustion, and then that every trace of water formed should be collected. A trivial loss of hydrogen or of water would tend to increase the apparent atomic weight of zinc. In the combustion of the zinc oxalate equally great difficulties are encountered. Here a variety of errors are possible, such as are due, for example, to impurity of material, to imperfect drying of the carbon dioxide, and to incomplete collection of the latter. Indeed a fourth combustion is omitted from the series as given, having been rejected by Favre himself. In this case the oxide formed was contaminated by traces of sulphide. 'Ann. Chim. Phys. (3), 10, 163. 1844. 148 THE ATOMIC WEIGHTS. Baubigny,* in 1883, resorted to the well-known sulphate method. Zinc sulphate, elaborately purified, was dried at 440 to constant weight, and then calcined at a temperature equal to the fusing point of gold. These data were obtained: ZnSO 4 . ZnO. Per cent. ZnO. 6.699 3.377 5 -4io 8.776 4.4245 50-416 Mean, 50.413, .0020 Hence Zn = 64.909. In Marignac's determinations of the atomic weight of zinc, published also in 1883,f there is a peculiar complication. After testing and criti- cising some other methods, he finally decided to study the double salt K 2 ZnCl 4 , which, however, is difficult to obtai n in absolutely definite con- dition. Although the compound was purified by repeated crystalliza- tions, it was found to deliquesce readily, and thereby to undergo partial dissociation, losing chloride of zinc, and leaving the porous layer on the crystalline surfaces richer in potassium. In order to evade this diffi- culty, Marignac placed a large quantity of the salt in a funnel, and col- lected the liquid product of deliquescence as it ran down. In this product he determined chlorine by volumetric titration with a standard solution of silver, and also estimated zinc by precipitation with sodium carbonate, and weighing as oxide. From the data thus obtained equa- tions were formed, giving for each a nalysis an atomic weight of zinc which is independent of the proportion between ZnCl 2 and KC1 in the substance analyzed. The data unfortunately are too bulky for repro- duction here and the calculations are complex ; but the results found for zinc, when Ag = 107.93, Cl 35.457, and K = 39.137, are as follows : 1. One titration Zn 65.22 2. Two titrations 65.37 3. Two titrations ... 65.31 4. Two titrations 65.28 5. One titration 65. 26 Each of these values represents a distinct sample of the deliquesced material, and the number of chlorine determinations is indicated. A second set of determinations was made by the same analytical method directly upon the recrystallized and carefully dried K 2 ZnCl 4 . The values for Zn are as follows : 6. Two titrations N . Zn = 65.28 7. Two titrations 65.39 8. One titration 65.32 * Ccmpt. Rend., 97, 906. 1883. tArch. Sci. Phys. et Nat. (3), 10, 194. ZINC. 149 In order to adapt these data to the uniform scheme of calculation em- ployed in this work, taking into account their probable error and the probable errors of the antecedent values for K, Cl, and Ag, it seems to be best to calculate them back with the atomic weights used by Marignac into the form of the ratio A 4 : K 2 Z nd 4 : : 100 : x. Doing this, and tak- ing each value as many times as there are titrations represented in it that is, giving the results of a double determination twice the weight of a single one we have the following series of data for the ratio in question : From 1 ................................... 66.090 From 2.. 66.124 f 66. no From 3 .................................... { l66.no f 66.104 P rom 4 .................................... < 166.104 From 5 .................................... 66.099 f 66.104 P rom 6 .................................... 4 (. 66.104 f 66.129 From 7 ................................... \ 166.129 From 8 ................................... 66.113 Mean, 66.111, d= .0023 Hence, from Marignac's work, Ag 4 : K 2 ZnCl 4 : : 100 : 66.111, .0023, a ratio which can be discussed along with others at the close of this chapter. During the years between 1883 and 1889, a number of determinations were made of the direct ratio between zinc and hydrogen that is, weighed quantities of zinc were dissolved in acid, the hydrogen evolved was measured, and from its volume, with Regnault's data, the weight of H was computed. First in order are Van der Plaats' determi nations j* whose results, as given by himself, are subjoined. The weights are reduced to a vacuum. Sulphuric acid was the solvent. Zn, grms. H, litres. Zn = 6.6725 1.1424 65.21 9.1271 i.5 6 43 65.14 13.8758 2.3767 65.18 Mean, 65.177, .0137 With the new value for the weight of hydrogen, .089872 gramme per litre, this becomes Zn = 64.980, db .0137. Reynolds and Ramsay made 29 determinations of this ratio.f rejecting, however, all but 5. The weighings were reduced to vacuum, and in each experiment the volume of hydrogen was fixed by the mean of seven or eight readings. The values for Zn are as follows : * Compt. Rend., 100, 52. 1885. f Journ. Chem. Soc., 51, 854. 1887. 150 THE ATOMIC WEIGHTS. 65.5060 65.4766 65.445 65.5522 65.4141 Mean, 65.4787, rh .0161 These values were computed with Regnault's data for the weight of H. Corrected by the new value the mean becomes Zn = 65.280, .0161. A few determinations by Mallet were made incidentally to his work on the atomic weight of gold, and appear in the same paper.* According to these experiments, one gramme of zinc gives 341.8500. H., and Zn = 65.158 341.91 " " 65.146 341-93 " 65.143 342.04 " " 65.122 Mean, 65.142, .0039 In this case the Crafts-Regnault weight of H was taken, one litre .08979 gramme. Corrected, the mean gives Zn = 65.082, .0039. Two other series of determinations of questionable value remain to be noticed before leaving the consideration of the direct H : Zn ratio. They represent really the practice work of students, and are interesting as an illustration of the closeness with which such work can be done. The first series was made in the laboratory of the Johns Hopkins Uni- versity, under the direction of Morse and Keiser,f and contains 51 deter- minations, as follows : 64.68 65.74 65.40 65.26 64.72 64.80 65.32 65.26 65.20 65.20 64.74 64.40 65.60 64.72 65.00 64.60 65.10 64.40 65.00 64.76 65-24 65.68 64.90 64.60 65.38 64.92 64.80 65.06 64.64 65.H 64.84 65.24 64.84 64.88 64.72 64.82 65.00 65.20 64.80 65.08 65.12 64.40 65.06 66.40 64.60 64.74 64.60 64.80 65.12 65.60 64.74 Mean of all, Zn = 64.997, .0328 *Amer. Chern. Journ., 12, 205. 1890. f Amer. Chem. Journ., 6, 347. 1884. ZINC. 151 Corrected for the difference between Regnault's value for H and the new value, this becomes Zn = 64.800, .0328. The second student series was published by Torrey,* who gives 15 determinations, as follows : 65.36 64.96 65.30 64.70 64.92 65.00 64.72 64.78 65.04 64.44 64.80 65.24 65.20 64.92 64.90 Mean, 64.952, dr .0436 Corrected as in the other series, this gives Zn 64.755, .0436. The five corrected means for the ratio H : Zn may now be combined, thus : Van der Plaats 64.980, .0137 Reynolds and Ramsay 65.280, .0161 Mallet 65.082, .0039 Morse and Keiser 64.800, .0328 Torrey 64.755, .0436 General mean 65.079, .0036 Morse and Burton, f in their determinations of the atomic weight of zinc, returned essentially to the old method adopted by Erdmann and by Jacquelain. Their zinc was obtained spectroscopically pure by dis- tillation in a vacuum, and was oxidized by nitric acid which left abso- lutely no residue upon evaporation. The conversion to oxide was effected in a porcelain crucible, which was enclosed in a larger one, and the ignition of the nitrate was carried out in a muffle. In weighing, the crucible was tared by one of nearly equal weight. Results as follows : Wf. Zn. Wt. ZnO. Percent. Zn in ZnO. .11616 1.38972 80.320 .03423 1.28782 80.308 .11628 1.38987 80.315 .05760 1.31681 80.316 .04801 1.30492 80.313 .02957 1.28193 80.318 .09181 1.35944 80.315 [.16413 1-44955 80.305 .07814 1.34248 80.305 .12754 1.40400 80.306 .91112 1.13446 80.310 * Amer. Chem. Journ., 10, 74, 1888. t Anier. Chem. Journ., 10, 311. 1888. 152 THE ATOMIC WEIGHTS. i.iooii 1.36981 1.17038 1^45726 1.03148 1.28436 L05505 I.3I365 Mean, 80.3115, 00084. Combining this mean with the means found by the earlier investigators, we have Jacquelain 80.541, .0070 Erdmann 80.260, .0037 Morse and Burton 80.3115, d= .00084 General mean 80.317, .0008 Morse and Burton verified by experiment the stability of oxide of zinc at the temperatures of ignition, and found that it did not dissociate. They also proved the absence of oxides of nitrogen from the zinc oxide. The investigations of Richards and Rogers,* however, have shown that zinc oxide prepared by ignition of the nitrate always carries gaseous occlusions, so that the atomic weight of zinc computed from the data of Morse and Burton is probably too low. But for that objection, their work would leave little to be desired on the score of accuracy. The determinations made by Gladstone and Hibbard f represent still another process for measuring the atomic weight of zinc. Zinc was dis- solved in a voltameter, and the same current was used to precipitate metallic silver or copper in equivalent amount. The weight of zinc dis- solved, compared with the weight of the other metal thrown down, gives the atomic weight sought for. Two voltameters were used in the experi- ments, giving duplicate estimates for zinc with reference to each weigh- ing of silver or copper. The silver series is as follows, with the ratio Ag 2 : Zn : : 100 : x in the third column : Zn. 4r- Ratio. .7767 2.5589 30.353 .7758 2.5589 30.318 .5927 1.9551 30.316 .5924 1.9551 30.300 .2277 7517 30.291 .2281 .7517 30.345 7452 2.4588 30.307 7475 2.4588 30.401 . .8770 2.9000 30.241 .8784 2.9000 30.290 9341 3.0809 30-3!9 .9347 3.0809 30.339 Mean, 30.318, =b .0077 * Proc. Amer. Acad., 1893, 200. t Journ. Chem. Soc., 55, 443. 1889. ZINC. 153 To the copper series I add the ratio Cu : Zn : : 100 : x. Zn. .7767 .7758 .5927 .5924 .2277 .2281 .8770 .8784 9341 9347 Cu. .7526 .7526 5737 5737 .2209 .2209 .8510 .8510 .9038 .9038 Ratio. 103-13 103.08 I03-3 1 103.26 103.08 103.26 103.05 103.22 103.36 103.42 Mean, 103.22, =fc .0261 Richards and Rogers,* in their investigation of the atomic weight of zinc, studied the anhydrous bromide. This was prepared by solution of zinc oxide in hydrobromic acid, evaporation to dryness, and subse- quent distillation in an atmosphere of carbon dioxide. In some experi- ments, however, the bromide was heated in an atmosphere of nitrogen, mingled with gaseous hydrobromic acid. All water can thus be removed, without formation of oxy bromides. The zinc bromide so obtained was dissolved in water, and precipitated with a solution containing a known amount of silver in the form of nitrate. The silver bromide was weighed on a Gooch crucible, and the ratio 2AgBr: ZnBr 2 thus found. An excess of silver was always used, and in one series of experiments it was estimated by precipitation with hydrobromic acid. Deducting the excess thus found from the original quantity of silver, the amount of the latter proportional to the zinc bromide was found; hence the ratio Ag 2 : ZnBr 2 . The results, with vacuum weights, are as follows : Series A. ZnBr. 2 . AgBr. Ratio. 1.69616 2.82805 59.976 1.98198 3.3 45o 59-978 1.70920 2 84949 59-984 2.35079 3-9'94i 59.978 2.66078 4-4375 l 59.9 6 i Mean, 59.975, .0034 Series B. ZnBr^. Ag. AgBr. Ag Ratio. AgBr Ratio. 2.33882 2. 24063 3.90067 104.382 59-959 1.97142 1.88837 3.28742 104.398 59.969 2.14985 2.05971 3-58539 104.376 59.96i 2.00966 1.92476 3-35074 104 411 59-977 Mean, 104.392, Mean, 59.967, .0054 =b .0027 *Zeitsch. Aiiorg. Chem., 10, i. 1895. 154 THE ATOMIC WEIGHTS. At the end of the same paper, Richards alone gives two more series of determinations made upon zinc bromide prepared by the action of pure bromine upon pure electrolytic zinc. The bromide so obtained was further refined by sublimation or distillation, and dried by heating in a stream of carbon dioxide and gaseous hydrobromic acid. Thus was ensured the absence of basic salts and of water. The weights and results found in the two series were as follows : Series C. ZnBr v . Ag. Ratio. 6.23833 5-9766 104.379 5 26449 5-0436 104.380 9.36283 8.9702 104.377 Mean, 104.379, .0007 Series D. ZnBr. 2 . AgBr. Ratio. 2.65847 4.4335 8 59.962 2.30939 3-85149 59.96i 5.26449 8.77992 59-961 Mean, 59.961, .0004 In some details of manipulation these series differ from those given by Richards and Rogers jointly, but their minutiaB are not essential to the present discussion. Combining these several series, we have For Ag^ : ZnBr^ : : 100 : x. Series E 104.392, .0054 Series C 104.379, .0007 General mean 104.380, .0007 For 2 AgBr : ZnBr^ : : zoo : x. Series A 59-975, .0034 Series B 59.967, =b .0027 Series D.. 59. 961, d= .0004 General mean 59.962, .0004 From the Ag ratio ZnBr 2 = 223.599, .0066 From the AgBr ratio " 223.601, .0066 General mean ZnBr 2 223.600, .0047 And Zn = 64.912, d= .0133 ZINC. 155 For computing the atomic weight of zinc we now have these ratios: (i.) Per cent. Zn in ZnO, 80.317, .0008 (2.) Per cent. ZnO in ZnSO 4 , 50.413, rb .0020 (3.) H 2 O : Zn : : 100 : 366.319, .088 (4.) 2CO 2 : Zn : : 100 : 93.169, rb -OI2 (5.) H : Zn : : I : 65.079, .0036 (6.) Ag 4 : K 2 ZnCl 4 : : 100 : 66. in, .0023 (7.) Ag 2 : Zn : : 100 : 30.318, .0077 (8.) Cu : Zn : : 100 : 103.22, =b .0261 (9.) Ag 2 : ZnBr 2 : : 100 : 104.38, rb .0007 (10.) 2AgBr : ZnBr 2 : : 100 : 59.962, rb .0004 The antecedent atomic weights, with H = 1, are O 15.879, rb .0003 C = 11.920, rb .0004 Cl = 35.179, .0048 S = 31.828, rb .0015 Br = 79.344, rb .0062 Cu 63.119, rb .0015 Ag = 107.108, rb .0031 AgBr = 186.452, rb .0054 K = 38.817, rb .0051 With these data, combining ratios 9 and 10 into one (see preceding paragraphs), we have nine independent values for the atomic weight of zinc, as follows : From (i) Zn = 64.795, d= .0030 From (2) " = 64.909, rb .0073 From (3) " = 65.494, rb .0019 From (4) " =65.521, rb .0115 From (5) " = 65.079, rb .0036 From (6) " = 64.891, rb .0253 From (7) " = 64.947, rb .0166 From (8) " =65.151, .0166 From (9) and (10) " = 64.912, rb .0133 General mean of all Zn = 65.152, .0014 With O = 16 Zn = 65.650 Of these values, Nos. 3 and 4, representing Favre's work, are unques- tionably far wrong. Rejecting them, the general mean of the remaining seven values becomes Zn = 64.912, .OO2I. If = 16, this gives Zn = 65.407. These figures are identical, except as regards the lower probable error, with the result deduced from Rich- ards and Rogers' determinations alone, and they may be taken as satisfactory. 156 THE ATOMIC WEIGHTS. CADMIUM. The earliest determination- of the atomic weight of this metal was by Stromeyer, who found that 100 parts of cadmium united with 14.352 of oxygen.* With our value for the atomic weight of oxygen, these figures make Cd = 110.64. This result has now only a historical interest. The more modern estimates of the atomic weight of cadmium begin with the work of v. Hauer.f He heated pure anhydrous cadmium sul- phate in a stream of dry hydrogen sulphide, and weighed the cadmium sulphide thus obtained. His results were as follows, with the percent- age of CdS in CdS0 4 therefrom deduced : 7.7650 grm. CdSO 4 gave 5.3741 grm. CdS. 69.209 per cent. 6.6086 " 4.5746 " 69.222 " 7-3821 " $.1117 " 69.245 " 6.8377 " 4.7336 " 69.228 8.1956 " 5.6736 " 69.227 " 7.6039 " 5.2634 " 69.220 " 7.1415 4-943 1 69.217 " 5.8245 4.0335 69.251 " 6.8462 4.74I5 69.257 " Mean, 69.231, .0042 LenssenJ worked upon pure cadmium oxalate, handling, however, only small quantities of material. This salt, upon ignition, leaves the following percentages of oxide : .5128 grm. oxalate gave .3281 grm. CdO. 63.982 per cent. .6552 " .4193 " 63.996 " .4017 .2573 64.053 " Mean, 64.010, d= .014 Dumas 1 1 dissolved pure cadmium in hydrochloric acid, evaporated the solution to dryness, and fused the residue in hydrochloric acid gas. The cadmium chloride thus obtained was dissolved in water and titrated with a solution of silver after the usual manner. From Dumas' weigh- ings I calculate the ratio between CdCl 2 and 100 parts of silver : 2-369 grm. CdCl 2 = 2.791 grm. Ag. 84.880 4.540 " 5.348 " 84.892 6.177 " 7-260 " 85.083 2.404 " 2.841 " 84.618 3.5325 " 4.166 " 84.794 4.042 " 4.767 84.791 Mean, 84.843, .026 * See Berz. Lehrbuch. sth Aufl., 3, 1219. t Journ. fiir Prakt. Chem., 72, 350. 1857. t Journ. fi'ir Prakt. Chem., 79, 281. 1860. || Ann. Chem. Pharm., 113, 27. 1860. CADMIUM. 157 Next in order comes Huntington's* work, carried out in the laboratory of J. P. Cooke. Bromide of cadmium was prepared by dissolving the carbonate in hydrobromic acid, and the product, dried at 200, was puri- fied by sublimation in a porcelain tube. Upon the compound thus ob- tained two series of experiments were made. In one series the bromide was dissolved in water, and a quantity of silver not quite sufficient for complete precipitation of the bromine was then added in nitric acid solution. After the precipitate had settled, the supernatant liquid was titrated with a standard solution of silver containing one gramme to the litre. The precipitate was washed by de- cantation, collected by reverse filtration, and weighed. To the weigh- ings I append the ratio between CdBr 2 and 100 parts of silver bromide : 1.5592 grm. CdBr 2 gave 2.1529 grm. AgBr. Ratio, 72.423 * 3.745 6 5- I 7 2 4 " " 7 2 .4i5 2.4267 3.3511 " " 72.415 * 3.6645 5.0590 " 72.435 * 3.7679 5.2016 " " 72.437 2.7938 3- 8 5 8 3 " " 7 2 .4io * i. 9225 2.6552 " " 72.405 3-4473 " 4-7593 " " 72.433 Mean, 72.4216, .0028 The second series was like the first, except that the weight of silver needed to effect precipitation was noted, instead of the weight of silver bromide formed. In the experiments marked with an asterisk, both the amount of silver required and the amount of silver bromide thrown down were determined in one set of weighings. The third column gives the CdBr 2 proportional to 100 parts of silver: * 3. 7456 grm. CdBr. 2 =: 2.9715 grm. Ag. 126.051 5.0270 " 3.9874 " 126.072 * 3.6645 " 2.9073 " 126.045 * 3.7679 " 2.9888 " 126.067 * 1. 9225 " 1.5248 " 126.082 2.9101 " 2.3079 " 126.093 3.6510 " 2.8951 " 126.110 3.9782 " 3.1551 " 126.088 Mean, 126.076, .0052 According to Huntington's own calculations, these experiments fix the ratio between silver, bromine, and cadmium as Ag : Br : Cd : : 108 : 80 112.31. In 1890, Partridge f published determinations of the atomic weight of cadmium, made by three methods, the weighings being reduced to * Proc. Araer. Acad., 1881. t Amer. Journ. Sci. (3), 40, 377. 1890. 158 THE ATOMIC WEIGHTS. vacuum standards throughout. First, Leiissen's method was followed, viz., the ignition of the oxalate, with the subjoined results: CdC.,0,. .09898 .21548 .10711 .17948 .16066 17995 34227 .43154 53510 .41311 CdO. .70299 .77746 .70807 75440 .74327 75471 .85864 .91573 .98197 .80397 Percent. CdO. 63.966 63.962 63.957 63.959 63.959 63.964 63.968 63-970 63.968 63-971 Mean, 63.964, .0010 Second!} 7 , v. Hauer's experiments were repeated, cadmium sulphate being reduced to sulphide by heating in a stream of H 2 S. The following data were obtained : 1.60514 1.55831 1.67190 1.66976 1.40821 1.56290 1.63278 1.58270 1.53873 1.70462 as. Percent. CdS. .11076 69.204 .07834 69.197 .15669 69.185 .15554 69.200 .9745 69.202 .08156 69.205 .12985 69.194 .09524 69.198 .06481 69.201 .17962 69.201 Mean, 69.199, =h .0012 v. Hauer found, 69.231, .0042 General mean, 69.202, .0012 In the third set of determinations cadmium oxalate was transformed to sulphide by heating in H 2 S, giving the ratio CdC 2 4 : CdS : : 100 : x. 1.57092 1.73654 2.19276 1.24337 1.18743 1.54038 1-38905 2.03562 2.03781 1.91840 CdS. 1.13065 1.24979 1.57825 .89492 .85463 1.10858 99974 1.46517 1.46658 1.38075 Per cent CdS. 71.972 71.973 71-974 7L974 71-975 71.968 71.976 71.979 71.970 71.971 Mean, 71.973, =b .0007 CADMIUM. 159 This work of Partridge was presently discussed by Clarke,* with ref- erence to the concordance of the data, and it was shown that the three ratios determined could be discussed algebraically, giving values for the atomic weights of Cd, S, and C, when = 16. These values are Cd= 111.7850 C = 11.9958 S = 32.0002, and are independent of all antecedent values except that assumed for the standard, oxygen. Morse and Jones, f starting out from cadmium purified by fractional distillation in vacuo, adopted two methods for their determinations. First, they effected the synthesis of the oxide from known weights of metal by dissolving the latter in nitric acid, evaporating to dryness, and subsequent ignition of the product. The oxide thus obtained was found to be completely free from oxides of nitrogen. The weighings,-which are given below, were made in tared crucibles. The third column gives the percentage of Cd in CdO. Cd Taken, CdO Found. Per cent. Cd. .77891 2.03288 87.507 .82492 2.08544 87.508 .74688 1.99626 87.507 .57000 1.79418 87.505 .481 2.26820 87.506 .27297 2.59751 87.504 .75695 2.00775 87.508 .70028 1.94305 87.505 .92237 2.19679 87.508 .92081 2.19502 87.508 Mean, 87.5066, rb .00032 The second method employed by Morse and Jones was that of Lenssen with cadmium oxalate. This salt they find to be somewhat hygroscopic, a property against which the operator must be on his guard. The data found are as follows : CdC 2 O t . CdO. Percent. CdO.' 53937 .98526 64.004 .77483 1.13582 63996 .70211 1.08949 64.008 .70238 1.08967 64.004 .74447 1.11651 64.003 Mean, 64.003, .0042 Lorimer and Smith, like Morse and Jones, determined the atomic weight of cadmium by means of the oxide, but by analysis instead of *Am. Chem. Jourii., 13, 34. 1891. t Am. Chem. Journ., 14, 261. 1892. 160 THE ATOMIC WEIGHTS. synthesis. Weighed quantities of oxide were dissolved in potassium cyanide solution, from which metallic cadmium was thrown down elec- trolytically. The weights are reduced to vacuum standards. CdO Taken. Cd Found. Per cent. Cd. .34767 .30418 87.491 .41538 -36352 87.515 1.04698 .91618 87.507 1.04066 .9 1 5 87.493 1.26447 1.10649 87.506 .78493 .68675 87.492 .86707 .75884 87.518 .67175 -58785 87.510 1.44362 1.26329 87.508 Mean, 87.5044, .0023 Mr. Bucher's dissertation* upon the atomic weight of cadmium does not claim to give any final measurements, but rather to discuss the vari- ous methods by which that constant has been determined. Neverthe- less, it gives many data which seem to have positive value, and which are certainly fit for discussion along with those which have preceded this paragraph. Bucher begins with cadmium purified by distillation nine times in vacuo, and from this his various compounds were prepared. His first series of determinations was made by reducing cadmium oxalate to oxide, the oxalate having been dried fifty hours at 150. The reduc- tion was effected by heating in jacketed porcelain crucibles, with various precautions, and the results obtained, reduced to vacuum standards, are as follows : Oxalate. Oxide. Percent. Oxide. .97674 1.26414 63.951 .94912 1.24682 63.968. .96786 1.25886 63.971 .87099 1.19675 . 63.958 3755 -87994 63.972 .33313 .85308 63.991 94450 1.24452 64.002 2.01846 1.29210 64.014 Mean, 63.978, d= .0052 Combining this with the means found by previous experimenters, we have for the percentage of oxide in oxalate Lenssen 64.010, .0140 Partridge 63.962, . .0010 Morse and Jones. 64.003, .0042 Bucher 63.978, .0052 General mean 63.966, .0010 * "An examination of some methods employed in determining the atomic weight of cadmium." Johns Hopkins University doctoral dissertation. By John B. Bucher. Baltimore, 1895. CADMIUM. 161 Bucher's next series of determinations was by Partridge's method the conversion of cadmium oxalate into cadmium sulphide by heating in a stream of sulphuretted hydrogen. The sulphide was finally cooled in a current of dry nitrogen. The vacuum weights and ratios are sub- joined : Oxalate. Sulphide. Percentage. 2.56319 1.84716 72.065 2.18364 I-5734I 72.055 2.11643 1.52462 72.037 3.13105 2.25582 72.047 Mean, 72.051, =b .0127 Partridge found, 71.973, .0007 General mean, 71.974, .0007 Here Bucher's mean practically vanishes. The third method employed by Bucher was that of weighing cadmium chloride, dissolving in water, precipitating with silver nitrate, and weigh- ing the silver chloride found. The cadmium chloride was prepared, partly by solution of cadmium in hydrochloric acid, evaporation to dryness, and sublimation in vacuo; and partly by the direct union of the metal with chlorine. The silver chloride was weighed in a Gooch crucible, with platinum sponge in place of the asbestos. To the vacuum weights I append the ratio 2AgCl : CdCl 2 : : 100 : x. 3.09183 2.26100 1-35729 2.05582 1.89774 3-53 6 7 2.70292 4.24276 3.40200 4.60659 2.40832 2.19144 2.84628 2.56748 2.31003 .25008 .96015 .29787 .94227 .10976 .63080 AgCl. 4.83856 3.53854 2.12431 3.21727 2.97041 5.48473 4.23087 6.63598 5-3 2 3 r 4 7.20386 3.42724 4-45477 4.01651 3.61370 1.95652 3- 6 54i 3-59391 3.03811 1.73547 2.55016 Ratio. 63.900 63.896 63-893 63.899 63.886 63.880 63.886 63.936 63.910 63.946 63.930 63.942 63-893 63923 63.924 63-893 63.944 63.938 63.9'5 63.946 63-949 Mean, 63.916, .0032 Bucher gives a rather full discussion of the presumable errors in this method, which, however, he regards as somewhat compensatory. 11 The 162 THE ATOMIC WEIGHTS. series is followed by a similar one with cadmium bromide, the latter having been sublimed in vacuo. Results as follows : CdBr 2 . AgBr. Ratio. 4.39941 6.07204 7 2 .454 3.18030 4-38831 7 2 .472 3.60336 4.97I5 72.480 4.04240 5-58062 72.453 3.60505 4.97519 72.461 Mean, 72.464, .0035 Huntington found, 72.4216, .0028 General mean, 72.438, .0022 In order to fix a minimum value for the atomic weight of cadmium, Bucher effected the synthesis of the sulphate from the metal. 1.15781 grammes of cadmium gave 2.14776 of sulphate. Hence Cd =.- 111.511. , The sulphate produced was dried at 400, and afterwards examined for free sulphuric acid, giving a correction which was applied to the weighings. The corrected weight is given above. Any impurity in the sulphate would tend to lower the apparent atomic weight of cadmium, and therefore the result is believed by the author to be a minimum. Finally, Bucher examined the oxide method followed by Morse and Jones. The syntheses of oxide were effected in double crucibles, first with both crucibles porcelain, and afterwards with the small inner cruci- ble of platinum. Two experiments were made by the first method, three by the last. Weights and percentages (Cd in CdO) as follows : Cd. CdO. Percentage. {1.26142 1.44144 87.511 .99785 1.14035 87.504 Mean, 87.508 ^1.11321 1.27247 87.484 4 1.02412 1.17054 87.491 (2.80966 3.21152 87.487 Mean, 87.487 Mean of alias one series, 87.495, .0035 The two means given above, representing work done with porcelain and with platinum crucibles, correspond to a difference of about 0.2 in the atomic weight of cadmium. Experiments were made with pure oxide of cadmium by converting it into nitrate and then back to oxide, exactly as in the foregoing syntheses. In each case the oxide obtained at the end of the operation represented an increase in weight, but the increase was greater in platinum than in porcelain. Hence the weigh- ings of cadmium oxide in the foregoing determinations probably are subject to constant errors, and cannot be trusted to fix the atomic weight CADMIUM. 163 of cadmium. Their mean, taken in one series, has really no significance ; but as the computations in this work involve a study of compensation of errors, the data may be combined with their predecessors, as follows : Morse and Jones 87.5066, .00032 Lorimer and Smith 87.5044, rh .0023 Bucher 87.495, .0035 General mean 87.5064, db .0003 This is equivalent to the absolute rejection of Buchers data, and is therefore not wholly fair to them. His work throws doubt upon the validity of the ratio, as determined, altogether. The latest determinations relative to the atomic weight of cadmium are those of Hardin.,* who effected the electrolysis of the chloride and bromide, and also made a direct comparison between cadmium and silver. The aqueous solutions of the salts, mixed with potassium cyanide, were electrolyzed in platinum dishes. The cadmium which served as the starting point for the investigation was purified by distil- lation in hydrogen. All weights are reduced to a vacuum. The data for the chloride series are as follows, with a column added for the per- centage of Cd in CdCl a : Weight CdCl v .43 HO .49165 .71752 .72188 .77264 .81224 .90022 1.02072 1.26322 L52344 Weight Cd. .26422 .30112 43942 .44208 .49742 .55135 .62505 .77365 933*4 Percentage Cd. 61.247 61.247 61.241 61.241 61.245 61.240 61.246 61.236 61.244 61.252 Mean, 61.244, .0010. The results for the bromide, similarly stated, are these: Weight CdBr^. Weight Cd. Percentage Cd. .57745 .23790 41.198 .76412 .31484 41.203 .91835 .37842 41.207 .01460 .41808 41.206 I 574 .474H 41.203 2475 1 51392 41.196 2595 1 .51905 41.210 51805 .62556 41.208 63543 .67378 4i.i99 2.15342 .88722 4 1 . 200 Mean, 41.203, 0010. ' Journ. Amer. Gheni. Soc., 18, 1016. 1896. 164 THE ATOMIC WEIGHTS. The direct comparison of cadmium and silver was effected by the simultaneous electrolysis, in the same current, of double cyanide solu- tions. Silver was thrown down in one platinum dish, and cadmium in another. The process was not altogether satisfactory, and gave diver- gent results, those which are cited below having been selected by Har- din from the mass of data obtained. I have added in a third column the cadmium proportional to 100 parts of silver : Weight Cd. Weight Ag. Ratio. .12624 -24335 5L 8 76 .11032 .21262 51.886 .12720 .24515 51.887 .12616 -2433 1 51-852 .22058 .42520 51-877 Mean, 51.876, d= .0041 For cadmium we now have the following ratios : (I.) Per cent, of Cd in CdO, 87.5064, .0003 (2.) Per cent, of CdO in CdC 2 O 4 , 63.966, .0010 (3.) Per cent, of CdS from CdC 2 O 4 , 71.974, .0007 (4.) Per cent, of CdS from CdSO 4 , 69.202, dz .0012 (5.) Ag 2 : CdCl 2 : : 100 : 84.843, .0260 (6.) 2AgCl : CdCl 2 : : 100 : 63.916, .0032 (7.) Ag 2 : CdBr 2 : : 100 : 126.076, .0052 (8.) 2AgBr : CdBr 2 : : 100 : 72.438, .0022 (9.) Per cent, of Cd in CdG 2 , 61.244, .0010 (10.) Per cent of Cd in CdBr 2 , 41.203, =b .0010 (il.) 2Ag : Cd : : 100 : 51.876, .0041 Bucher's single experiment upon the synthesis of the sulphate, although important and interesting, cannot carry weight enough to warrant its consideration in connection with the other ratios, and is therefore not included. The antecedent values, for use in computation are O = I 5-879, .0003 S = 31.828, =b .0015 Ag = 107.108, d= .0031 C = 11.920, dr .0004 Cl == 35.179, .0048 AgCl = 142.287, ,0037 Br = 79.344, .0062 AgBr = 186.452, .0054 For the molecular weight of cadmium chloride, two values are now deducible : From (5) CdCl 2 = 181.739, .0560 From (6) " 181.888, + .0103 General mean CdCl 2 = 181.883, .0138 Hence Cd = 111.525, .0138. CADMIUM. 165 For cadmium bromide we have From (7) CdBr 2 = 270.073, =b .0136 From (8) " = 270.124,^.0113 General mean CdBr 2 = 270.105, .0087 Hence Cd = 111.417, .0151. For cadmium there are nine independent values, as follows : From (3) Cd = 1 10.793, d= .0081 From (4) " = i 10.890, .0069 From (2) " = 1 1 1.004, db - OO 47 From (11) " = 111.127, -0095 From (9) " = 1 1 1. 183, .0155 From (10) " = 111.202, .0093 From (i). " = 111.227, -0034 From molecular weight CdBr 2 " = 111.417, =b .0151 From molecular weight CdCl 2 ....... ".= 111.525, .0138 General mean Cd = iii.ioo, dz .0022 If 0=16, Cd= 111.947. This result is obviously uncertain. The data are far from being con- clusive, however, and I am therefore inclined to trust the mean rather than any one of the values taken separately. It is quite possible that the highest of all the figures may be nearest the truth, as Bucher's ex- periments seem to indicate ; but until new evidence is obtained it would hardly be wise to make any selection. The mean obtained agrees well with the data of Morse and Jones, Lorimer and Smith, and Hardin. 166 THE ATOMIC WEIGHTS. MERCURY. In dealing with the atomic weight of mercury we may reject the early determinations by Sefstrom* and a large part of the work done by Tur- ner, f The latter chemist, in addition to the data which will be cited below, gives figures to represent the percentage composition of both the chlorides of mercury ; but these results are neither reliable nor in proper shape to be used. First in order we may consider the percentage composition of mercuric oxide, as established by Turner and by Erdmann and Marchand. In both investigations the oxide was decomposed by heat, and the mercury was accurately weighed. Gold leaf served to collect the last traces of mercurial vapor. Turner gives four estimations. Two represent oxide obtained by the ignition of the nitrate, and two are from commercial oxide. In the first two the oxide still contained traces of nitrate, but hardly in weighable proportions. A comparison of the figures from this source with the others is sufficiently conclusive on this point. The third column represents the percentage of mercury in HgO : 144 805 grains Hg = 11.54 grains O. 92.619 per cent. 125.980 " 10.08 " 92.592 " I73-5 61 " 13.82 " 92.625 " 114.294 " 9.101 " 92.620 " Mean, 92.614, db .0050 In the experiments of Erdmann and Marchand J every precaution was taken to ensure accuracy. Their weighings, reduced to a vacuum stand- ard, give the subjoined percentages : 82.0079 grm. HgO gave 75.9347 grm. Hg. 92.594 per cent. 51.0320 47.2538 " 92.597 " 84.4996 " 78.2501 " 92.604 " 44-6283 " 41-3285 " 92.606 " 118.4066 " 109.6408 " 92.597 " Mean, 92.5996, .0015 Hardin's determination of the same ratio, being different in character, will be considered later. With a view to establishing the atomic weight of sulphur, Erdmann and Marchand also made a series of analyses of pure mercuric sulphide. These data are now best available for discussion under mercury. The *Sefstrom. Berz. L,ehrb., 5th ed., 3, 1215. Work done in 1812. fPhil. Trans., 1833, 531-535. J Journ. fur Prakt. Chem., 31, 395. 1844. MERCURY. 167 v sulphide was mixed with pure copper and ignited, mercury distilling over and copper sulphide remaining behind. Gold leaf was used to retain traces of mercurial vapor, and the weighings were reduced to vacuum : 34.3568 grm. HgS gave 29.6207 grm. Hg. 86.215 P er cent - H g. 24.8278 " 21.40295 " 86.206 " 37.2177 " 32.08416 " 86.207 " 80.7641 " 69.6372 " 86.223 " Mean, 86.2127, .0027 For the percentage of mercury in mercuric chloride we have data by Turner, Millon, Svanberg, and Hardin. Turner,* in addition to some precipitations of mercuric chloride by silver nitrate, gives two experi- ments in which the compound was decomposed by pure stannous chloride, and the mercury thus set free was collected and weighed. The results were as follows : 44.782 grains Hg = 15.90 grains CI. 73-798 per cent. 73.09 " 25.97 " 73.784 " Mean, 73.791, .005 Millon f purified mercuric chloride by solution in ether and sublima- tion, and then subjected it to distillation with lime. The mercury was collected as in Erdmann and Marchand's experiments. Percentages of metal as follows : 73-87 73-8i 73-83 73-87 Mean, 73.845, .010 Svanberg, J following the general method of Erdmann and Marchand, made three distillations of mercuric chloride with lime, and got the following results : 12.048 grm. HgC) 2 gave 8.889 grm. Hg. 73.780 per cent. 12.529 " 9-24S 6 " 73-794 " 12.6491 " 9-3363 " 73-8io " Mean, 73.795, .006 The most recent determinations of the atomic weight of mercury are due to Hardin, whose methods were entirely electrolytic. First, pure mercuric oxide was dissolved in dilute, aqueous potassium cyanide, and *Phil. Trans., 1833, 53 I -535- fAnn. Chirn. Phys. (3), 18, 345. 1846. I Journ. fur Prakt. Chem., 45, 472. 1848. I Journ. Amer. Chem. Soc., 18, 1003. '1896. 168 THE ATOMIC WEIGHTS. electrolyzed in a platinum dish. Six determinations are published, out of a larger number, but without reduction of the weights to a vacuum. The data, with a percentage column added, are as follows : Weight HgO. Weight Hg. Per cent. Hg. .26223 .24281 92.594 .23830 .22065 9 2 .593 .23200 .21482 92.595 .14148 .13100 92.593 .29799 .27592 92.594 .19631 .18177 92.593 Mean, 92.594, d= 0003. Various sources of error were detected in these experiments, and the series is therefore rejected by Hardin. It combines with previous series as follows : Turner % 92.614, rfc .0050 Erdmann and Marchand 92.5996, .0015 Hardin 92. 594, .0003 General mean 9 2 -595, rb .0003 Hardin also studied mercuric chloride, bromide, and cyanide, and the direct ratio between mercury and silver, with reduction of weights to a vacuum. Electrolysis was conducted in a platinum dish, as usual. With the chloride and bromide, the solutions were mixed with dilute potassium cyanide. The data for the chloride are as follows, the per- centage column being added by myself: Weight HgCL v Weight Hg. Per cent. Hg. 45932 -33912 54735 -40415 .56002 .41348 .63586 .46941 .64365 .47521 .73281 .54101 .86467 .63840 1.06776 .78825 1.07945 .79685 . 1.51402 1.11780 73-830 Mean, 73.829, .0012 Combining this with the earlier determinations, we hav Turner 73-791, db .0050 Millon 73.845, .0100 Svanberg 73-795, .0060 Hardin 73.829, .001 2 General mean 73.826, d= .001 1 MERCURY. 169 For the bromide Hardin's data are Weight HgBr v .70002 57*42 .77285 .80930 .85342 1.11076 i 17270 1.26186 1.40142 And for the cyanide Weight HgC 2 N 2 . .55776 .63290 .70652 .80241 .65706 .81678 1.07628 1.22615 1.66225 2.11170 Weight Hg. .38892 .3135 .3i75 .42932 .44955 .47416 .61708 .65145 .70107 .77870 Weight Hg. .44252 .50215 .56053 .63663 .52130 .64805 .85392 .97282 1.31880 1.67541 Per cent. Hg. 55-558 55-555 55-563 55-550 55.548 55-56o 55-555 55-55 1 55-559 55-565 Mean, 55.556, .0012 Per cent. Hg. 79-337 79-341 79-337 79-340 79-338 79-342 79-340 79-339 79-338 79-339 Mean, 79.339, .0004 In the last series cited no potassium cyanide was used, but the solution of mercuric cyanide, with the addition of one drop of sulphuric acid, was electrolyzed directly. The direct ratio between silver and mercury was determined by throw- ing down the two metals, simultaneously, in the same electric current. Both metals were taken in double cyanide solution. With Hardin's equivalent weights I give a third column, showing the quantity of mer- cury corresponding to 100 parts of silver. Many experiments were re- jected, and only the following seven are published by the author : Weight Hg. .06126 .06190 .07814 .10361 .15201 .26806 .82808 Weight Ag. .06610 .06680 .08432 .11181 . 1 6402 .28940 .89388 Ratio. 92.678 92.665 92.671 92.666 92.678 92.626 92.639 Mean, 92.660, .0051 170 THE ATOMIC WEIGHTS. We now have six ratios involving the atomic weight of mercury, as follows : (i.) Per cent, of Hg in HgO, 92.595, .0003 (2.) Per cent, of Hg in HgS, 86.2127, =b .0027 (3.) Per cent, of Hg in HgCl a> 73.826, .0011 (4.) Per cent, of Hg in HgBr 2 , 55.556, .0012 (5.) Per cent, of Hg in HgC 2 N 2 , 79.339, .0004 (6.) 2Ag : Hg : : 100 : 92.660, .0051 The calculations involve the following values : O = 15.879, -.0003 Br=r 79.344, .0062 Ag= 107.108, .0031 S =31.828, .0015 Cl = 35.179, .0048 C = 11.920, .0004 N = 13.935, .0021 Hence the values for mercury are From (I) Hg = 198.557, .0084 From (2) " = 199.027, H= .0406 From (3) " = 198.482, .0285 From (4) " = 198.364,^.0170 From (5) 198.568, .0170 From (6) " = 198.493, .0124 General mean Hg = 198.532, db .0059 If 0= 16, Hg = 200.045. But according to Hardin the value derived from the analyses of mer- curic oxide is untrustworthy. Rejecting this, and also the abnormally high -result from the sulphide series, the general mean of the four re- maining values is Hg = 198.491, .0083, or, with = 16, Hg = 200.004. These figures seem to be the best for the atomic weight of mercury. BORON. 171 BORON. In the former edition of this work the data relative to boron were few and unimportant. There was a little work on record by Berzelius and by Laurent, and this was eked out by a discussion of Deville's analyses of boron chloride and bromide. As the latter were not intended for atomic weight determinations they will be omitted from the present re- calculation, which includes the later researches of Hoskyns-Abrahall, Ramsay and Aston, and Rimbach. Berzelius* based his determination upon three concordant estima- tions of the percentage of water in borax. Laurent f made use of two similar estimations, and all five may be properly put in one series, thus : 47-10] 47.10 j- Berzelius. 47-ioj 47- '5 I Laurent. 47-20* Mean, 47.13, .013 In 1892 the posthumous notes of the late Hoskyns-Abrahall were edited and published by Ewan and Hartog. J This chemist especially studied the ratio between boron bromide and silver, and also redeter- mined the percentage of water in crystallized borax. The latter work, which was purely preliminary, although carried out with great care, gave the following results, reduced to vacuum standards : Na^BtPv Per cent. 7.00667 3.69587 47.2069 12.95936 6.82560 47.3308 4.65812 2.45248 47.35 4 4.47208 3-9395 6 47. 2 7 6 3 4.94504 2.60759 47.2686 Mean, 47.2866, db .0171 Two sets of determinations were made with the bromide, which was prepared from boron and bromine directly, freed from excess of the latter by standing over mercury, and finally collected, after distillation, in small, weighed, glass bulbs. It was titrated with a solution of silver after all the usual precautions. The first series of experiments was as follows, with BBr 3 proportional to 100 parts of silver stated as the ratio : *Poggend. Annalen, 8, i. 1826. t Journ. fur Prakt. Chem., 47, 415. 1849. I Journ. Chem. Soc., 61, 650. August, 1892. -v v, .- - ' v, :v - :.-S,v - :x- ; -. . :;-^ \". -- --- = -- --.- = . . : : . . : -- - -' - ' : : : ----- :: ;- L>- 174 THE ATOMIC WEIGHTS. Na^B^O,. AgCl. Ratio. 5.3118 7.5 2 59 70.580 4.7806 6.7794 70.517 4.9907 7.0801 70.489 4.7231 6.6960 70.53 6 3.3138 4-693 1 70.610 Mean, 70.546, .0146 Rimbach * based his determination of the atomic weight of boron upon the fact that boric acid is neutral to methyl orange, and that therefore it is possible to titrate a solution of borax directly with hydrochloric acid. His borax was prepared from carefully purified boric acid and sodium carbonate, and his hydrochloric acid was standardized by a series of precipitations and weighings as silver chloride. It contained 1.84983 per cent, of actual HC1. The borax, dissolved in water, was titrated by means of a weight-burette. I give the weights found in the first and second columns of the following table, and in the third column, calcu- lated by myself, the HC1 proportional to 100 parts of crystallized borax. Rimbach himself computes the percentage of Na 2 O and thence the atomic weight of boron, but the ratio Na 2 B 4 7 .10H 2 : 2HC1 is the ratio actually determined. Na^B^O-f.wH^O. HCl Solution. Ratio. 10.00214 103.1951 19.0853 15.32772 158-1503 19.0864 15.08870 155-7271 19.0917 10.12930 104.5448 19.0922 5.25732 54-2571 19.0908 15.04324 155-2307 19.0883 15-04761 I55-2959 19.0908 10.43409 107.6602 19.0868 5.04713 52.0897 i9-9 I 5 Mean, 19.0893, d= .0006 , Obviously, this error should be increased by the probable errors in- volved in standardizing the acid, but they are too small to be worth considering. The following ratios are now available for boron : (1) Percentage of water in Na 2 B 4 O 7 .ioH 2 O, 47.1756, =h .0066 (2) 3Ag : BBr 3 : : 100 : 77.425, .0017 (3) Na 2 B 4 O 7 : 2NaCl : : 100 : 57-933, .0074 (4) 2AgCl : Na 2 B 4 O 7 : : 100 : 70.546, + .0146 (5) Na 2 B 4 O 7 .ioH 2 O : 2HC1 : : 100 : 19.0893, .0006 * Berichte Deutsch. Chein. Gesell., 26, 164. 1893. BORON. 175 For reduction we have the antecedent atomic and molecular weights O = 15-879, -0003 Na = 22.881, .0046 Ag= 107.108, .0031 NaCl=; 58.060, .0017 Cl = 35.179,^.0048 AgCl= 142.287, rb .0037 Br = 79-344, =b .0062 For the molecular weight of Na 2 B 4 7 we now have From (i) . . . .' Na 2 B 4 O 7 = 200. 198, .0377 From (3) " = 200.439, .0263 From (4) " = 200.756, .0419 From (5) " = 200.260, .0518 General mean Na 2 B 4 O t = 200.421, .0180 Hence B = 10.876, .0051. From ratio (2), B = 10.753, .0207. The two values combined give B =1 10.863, .0050. Or, if = 16, B == 10.946. If we consider ratios (1), (3), (4), and (5) separately, they give the fol- lowing values for B : From (i) B = 10.821 From (3) " = 10.881 From (4) " = 10.960 From (5) " = 10.836 Of these, the second and third involve the data from which, in a previous section of this work, the ratio NaCl : AgCl was computed. In using that ratio for measuring the molecular weights of its component molecules, discordance was noted, which again appears here. The chief uncertainty in it seems to be connected with ratio (4), which is therefore entitled to comparatively little credence, although its rejection is not necessary at this point. In ratio (2), Abrahall's determination, the high probable error of B is due to the also high probable error of 3Br, and it is quite likely that the result is undervalued. The general mean, B = 10.863, .0050, however, can hardly be much out of the way. It is cer- tainly more probable than any one of the individual values. 176 THE ATOMIC WEIGHTS. ALUMINUM. The atomic weight of aluminum has been determined by Berzelius, Mather, Tissier, Dumas, Isnard, Terrell, Mallet, and Baubigny. The early calculations of Davy and of Thomson we may properly disregard. Berzelius' * determination rests upon a single experiment. He ignited 10 grammes of dry aluminum sulphate, A1 2 (S0 4 ) 3 , and obtained 2.9934 grammes of A1 2 3 as residue. Hence Al = 27.103. In 1835 1 Mather published a single analysis of aluminum chloride, from which he sought to fix the atomic weight of the metal. 0.646 grm. of A1CL, gave him 2.056 of AgCl and 0.2975 of A1 2 3 . These figures give worthless values for Al, and are included here only for the sake of com- pleteness. From the ratio between AgCl and A1C1 3 , Al = 28.584. Tissier's J determination, also resting on a single experiment, appeared in 1858. Metallic aluminum, containing .135 per cent, of sodium, was dissolved in hydrochloric acid. The solution was evaporated with nitric acid to expel all chlorine, and the residue was strongly ignited until only alumina remained. 1.935 grm. of Al gave 3.645 grm. of A1. 2 3 . If we correct for the trace of sodium in the aluminum, we have Al = 26.930. Essentially the same method of determination was adopted by Isnard, who, although not next in chronological order, may fittingly be men- tioned here. He found that 9 grm. of aluminum gave 17 grm. of A1. 2 3 . Hence Al = 26.8 In 1858 Dumas, 1 1 in connection with his celebrated revision of the atomic weights, made seven experiments with aluminum chloride. The material was prepared in quantity, sublimed over iron filings, and finally resublimed from metallic aluminum. Each sample used was collected in a small glass tube, after sublimation from aluminum in a stream of dry hydrogen, and hermetically enclosed. Having been weighed in the tube, it was dissolved in water, and the quantity of silver necessary for precipitating the chlorine was determined. Reducing to a common standard, his weighings give the quantities of A1C1 3 stated in the third column, as proportional to 100 parts of silver : 1.8786 grm. Alt .1 3 =4.543 grm. Ag. 41.352 3.021 " 7.292 41.459 Bad. 2.399 5.802 41.348 1.922 " 4.6525 " 41.311 1.697 4.1015 4L375 4-3165 10.448 " 4L3H 6.728 16.265 41.365 *Poggend. Annal., 8, 177. tSilliman's Amer. Journ., 27, 241. J Cotnpt. Rend., 46, 1105. I Compt. Rend., 66, 508. 1868. || Ann. China. Phys. (3), 55, 151. Ann. Cheni. Pharm., 113, 26. ALUMINUM. 177 In the second experiment the A1C1 3 contained traces of iron. Reject- ing this experiment, the remaining six give a mean of 41.344, .007. These data give a value for Al approximating to 27.5, and were for many years regarded as satisfactory. It now seems probable that the chloride contained traces of an oxy-compound, which would tend to raise the atomic weight. In 1879 Terreil * published a new determination of the atomic weight under consideration, based upon a direct comparison of the metal with hydrogen. Metallic aluminum, contained in a tube of hard glass, was heated strongly in a current of dry hydrochloric acid. Hydrogen was set free, and was collected over a strong solution of caustic potash. 0.410 grm. of aluminum thus were found equivalent to 508.2 cc., or .045671 grm. of hydrogen. Hence Al = 26.932. About a year after Terrell's determination appeared, the lower value for aluminum was thoroughly confirmed by J. W. Mallet.f After giving a full resume of the work done by others, exclusive of Isnard, the author describes his own experiments, which may be summarized as follows : Four methods of determination were employed, each one simple and direct, and at the same time independent of the others. First, pure ammonia alum was calcined, and the residue of aluminum oxide was estimated. Second, aluminum bromide was titrated with a standard solution of silver. Third, metallic aluminum was attacked by caustic soda, and the hydrogen evolved was measured. Fourth, hydrogen was set free by aluminum, and weighed as water. Every weight was care- fully verified, the verification being based upon the direct comparison, by J. E. Hilgard, of a kilogramme weight with the standard kilogramme at Washington. The specific gravity of each piece was determined, and also of all materials and vessels used in the weighings. During each weighing both barometer and thermometer were observed, so that every result represents a real weight in vacuo. The ammonium alum used in the first series of experiments was specially prepared, and was absolutely free from ascertainable impuri- ties. The salt was found, however, to lose traces of water at ordinary temperatures a circumstance which tended towards a slight elevation of the apparent atomic weight of aluminum as calculated from the weighings. Two sets of experiments were made with the alum ; one upon a sample air-dried for two hours at 21-25, the other upon mate- rial dried for twenty-four hours at 19-26. These sets, marked A and B respectively, differ slightly, B being the less trustworthy of the two, judged from a chemical standpoint. Mathematically it is the better of the two. Calcination was effected with a great variety of precautions, concerning which the original memoir must be consulted. To Mallet's weighings I append the percentages of A1 2 3 deduced from them : * Bulletin de la Soc. Chimique, 31, 153. f Phil. Trans., 1880, p. 1003. 12 178 THE ATOMIC WEIGHTS. Series A. 8.2144 grm. of the alum gave .9258 grm. A1 2 O 3 . 11.270 per cent. 14.0378 " 1.5825 " 11.273 " 5.6201 " '.6337 " 11.275 " 11.2227 " 1.2657 " 11.278 " 10.8435 " 1. 2216 " 11.266 " Mean, 11.2724, .0014 Series B. 12.1023 grm. of the alum gave 1.3660 grm. A1 2 O 3 . 11.287 per cent. 10.4544 " 1.1796 " 11.283 6.7962 " .7670 " 11.286 " 8.5601 " .9654 " 11.278 4.8992 .5528 " 11.283 " Mean, n.,2834, .0011 Combined, these series give a general mean of 11.2793, . 0008. Hence Al === 26.952. The aluminum bromide used in the second series of experiments was prepared by the direct action of bromine upon the metal. The product was repeatedly distilled, the earlier portions of each distillate being re- jected, until a constant boiling point of 263. 3 at 747 mm. pressure was noted. The last distillation was effected in an atmosphere of pure nitro- gen, in order to avoid the possible formation of oxide or oxy-bromide of aluminum ; and the distillate was collected in three portions, which proved to be sensibly identical. The individual samples of bromide were collected in thin glass tubes, which were hermetically sealed after nearly filling. For the titration pure silver was prepared, and after fusion upon charcoal it was heated in a Sprengel vacuum in order to eliminate occluded gases. This silver was dissolved in specially purified nitric acid, the latter but very slightly in excess. The aluminum bro- mide, weighed in the sealed tube, was dissolved in water, precautions be- ing taken to avoid any loss by splashing or fuming which might result from the violence of the action. To the solution thus obtained the silver solution was added, the silver being something less than a decigramme in deficiency. The remaining amount of silver needed to complete the precipitation of the bromine was added from a burette, in the form of a standard solution containing one milligramme of metal to each cubic centimetre. The final results were as follows, the figures in the third column representing the quantities of bromide proportional to 100 parts of silver. Series A is from the first portion of the last distillate of AlBr 3 ; series B from the second portion, and series C from the third portion : Series A. 6.0024 grm. AlBr 3 = 7.2793 grm. Ag. 82.458 8.6492 10.4897 " 82.454 3.1808' " 3.8573 " 82.462 ALUMINUM. 179 6.9617 grm. AlBr a 11.2041 " 3.7621 5.2842 9.7338 Series B. 8.4429 grm 13.5897 4.5624 6.4085 11.8047 Ag. 82.456 82.445 82.459 82.456 82.457 9-35I5 S rm - 4.4426 5.2750 Series C. AlBr 3 i= 1 1. 3424 grm. Ag. 5.3877 " 6.3975 82.447 82.458 82.454 Mean, 82.455, .001 Hence Al = 26.916. The experiments to determine the amount of hydrogen evolved by the action of caustic soda upon metallic aluminum were conducted with pure metal, specially prepared, and with caustic soda made from sodium. The soda solution was so strong as to scarcely lose a perceptible amount of water by the passage through it of a dry gas at ordinary temperature. As the details of the experiments are somewhat complex, the original memoir must be consulted for them. The following results were obtained, the weight of the hydrogen being calculated from the volume, reckoned at .089872 gramme per litre. Wt. AL 3697 3769 .3620 7579 73*4 7541 Vol. H. 458.8 467.9 449-1 941-5 907.9 936.4 Wt. H. .041234 .042051 .040362 .084614 .081595 .084156 At. Wt. 26.898 26.889 26.907 26.872 26.891 26.882 Mean, 26.890, .0034 'he closing series of experiments was made with larger quantities of aluminum than were used in the foregoing set. The hydrogen, evolved by the action of the caustic alkali, was dried by passing it through two drying tubes containing pumice stone and sulphuric acid, and two others containing asbestos and phosphorus pentoxide. Thence it passed through a combustion tube containing copper oxide heated to redness. A stream of dry nitrogen was employed to sweep the last traces of hy- drogen into the combustion tube, and dry air was afterwards passed through the entire apparatus to reoxidize the surface of reduced copper, and to prevent the retention of occluded hydrogen. The water formed by the oxidation of the hydrogen was collected in three drying tubes. 180 THE ATOMIC WEIGHTS. The results obtained were as follows. The third column gives the amount of water formed from 10 grammes of aluminum. 2.1704 grm. Al gave 2.1661 grm. H 2 O. 9.9802 2.9355 " 2.9292 9-9785 5.2632 " 5- 2 5 62 " 9-9 86 7 Mean, 9.9818, .0017 Hence Al = 26.867. From the last two series of experiments an independent value for the atomic weight of oxygen may be calculated. It becomes O = 15.895. The closeness of this figure to some of the best determinations affords a good indication of the accuracy of Mallet's work. In connection with Mallet's work it is worth noting that Torrey * pub- lished a series of measurements of the H : Al ratio, representing determi- nations made under his direction by elementary students. These meas- urements are thirteen in number, and calculated with Regnault's old value for the weight of hydrogen, range from 26.661 to 27.360, or in mean, 27.049, .323. Corrected by the latest value for the weight of H, this mean becomes 26.967. The result, of course, has only confirmatory significance. By Baubignyf we have only two determinations, based upon the calcination of anhydrous aluminum sulphate, A1 2 (SOJ 3 . 3.6745 grm. salt gave 1.0965 A1 2 O 3 . 29.841 per cent. 2.539 " -7572 " 29.823 " Mean, 29.832, .0061 Hence Al = 26.858. It is clear that the single determinations of Berzelius, Mather, Tissier, Isnard, and Terrell may now be safely left out of account, for the reason that none of them could affect appreciably the final value for Al. The ratios to consider are as follows : (I.) 3Ag : A1C1 3 : : TOO : 41. 344, .0070 (2.) Percentage of A1 2 O 3 in ammonium alum, 11.2793, rb .0008 (3-) 3^g : A113r 3 : : 100 : 82.455, .0010 (4.) H : Al : : I : 26.890, .0034 (5-) A1 2 : 3 H 2 O : : 10 : 9.9818, .0017 (6.) Percentage of A1 2 O 3 in A1 2 (SO 4 ) 3 , 29.832, .0061 The antecedent values are O = 15.879,43.0003 Br= 79-344, .0062 Ag= 107.108, .0031 N= 13.935, d=.oo2i Cl = 35.179, .0048 S == 31.828, .0015 * Am. Chem. Journ., 10, 74. 1888. f Compt. Rend., 97, 1369. 1883. GALLIUM. 181 Hence for aluminum we have From (i) Al = 27.31 1, .0270 From (2) " = 26.952, db .0037 From (3) " = 26.916, .0201 From (4) " = 26.890, rh .0034 From (5) " = 26.867, .0046 From (6) " = 26.858, .0113 General mean Al = 26.906, .0021 With = 16, Al = 27.111. The rejection of Dumas' data only lowers the result to 26.903. GALLIUM. Gallium has been so recently discovered, and obtained in such small quantities, that its atomic weight has not as yet been determined with much precision. The following data were fixed by the discoverer, Lecoq de Boisbaudran : * 3.1044 grammes gallium ammonium alum, upon ignition, left .5885 grm. Ga. 2 O 3 . Hence Ga = 69.595. If = 16, Ga = 70.125. .4481 grammes gallium, converted into nitrate and ignited, gave .6024 grm. Ga 2 O 3 . Hence Ga = 69.171. If O = 16, Ga = 69.698. These values, assigned equal weight, give these means : With H = i, Ga = 69.383. With O = 16, Ga = 69.912 * Journ. Chem. Soc., 1878, p. 646. 182 THE ATOMIC WEIGHTS. INDIUM. Reich and Richter, the discoverers of indium, were also the first to determine its atomic weight.* They dissolved weighed quantities of the metal in nitric acid, precipitated the solution with ammonia, ignited the precipitate, and ascertained its weight. Two experiments were made, as follows : 5 T 35 S rm - indium gave .6243 grm. In 2 O 3 . .699 .8515 Hence, in mean, In = 110.61, if = 16 ; a value known now to he too low. An un weighed quantity of fresh, moist indium sulphide was also dis- solved in nitric acid, yielding, on precipitation, .2105 grm. In 2 O 3 and .542 grm. BaSO 4 . Hence, with BaS0 4 = 233.505, In = 112.03 ; also too low. Soon after the publication of Reich and Richter's paper the subject was taken up by Winkler .f He dissolved indium in nitric acid, evap- orated to dryness, ignited the residue, and weighed the oxide thus obtained. 5574 S rm - I* 1 gave .6817 S rm - In 2 O 3 . .6661 " .8144 " .5011 " .6126 " Hence, in mean, if = 16, In = 107.76 ; a result even lower than the values already cited. In a later paper by Winkler J better results were obtained. Two methods were employed. First, metallic indium was placed in a solu- tion of pure, neutral, sodio-auric chloride, and the amount of gold pre- cipitated was weighed. I give the weighings and, in a third column r the amount of indium proportional to 100 parts of gold : In. Au. Ratio. .4471 grm. .8205 grm. 57-782 .8445 1.4596 " 57.858 Mean, 57.820, .026 Hence, if Au = 195.743, .0049, In = 113.179, .0517. Winkler also repeated his earlier process, converting indium into oxide by solution in nitric acid and ignition of the residue. An ad- * Journ. fur Prakt. Chem., 92, 484. t Journ. fur Prakt. Chem., 94, 8. % Journ. fur Prakt. Chem., 102, 282. INDIUM. 183 ditional experiment, the third as given below, was made after the method of Reich and Richter. The third column gives the percentage of In in In 2 3 : 1.124 g rm - J n gave 1.3616 grm. In 2 O 3 . Per cent., 82.550 1.015 " 1.2291 " " 82.581 .6376 " .7725 82.537 These figures were confirmed by a single experiment of Bunsen's,* published simultaneously with the specific heat determinations which showed that the oxide of indium was In 2 3 , and not InO, as had been previously supposed : 1.0592 grm. In gave 1.2825 grm. In 2 O 3 . Per cent. In, 82.589 For convenience we may add this figure in with Winkler's series, which gives us a mean percentage of In in In 2 s of 82.564, .0082. Hence, if = 15.879, .0003, In = 112.787, .0542. Combining both values, we have From gold series In = 113.179, =b .0517 From oxide series '. (( = 112.787, .0542 General mean In = 1 12.992, .0374 If = 16, In = 113.853. * Poggend. Annal., 141, 28. 184 THE ATOMIC WEIGHTS. THALLIUM. The atomic weight of this interesting metal has been fixed by the re- searches of Lamy, Werther, Hebberling, Crookes, and Lepierre. Lamy and Hebberling investigated the chloride and sulphate ; Wer- ther studied the iodide; Crookes' experiments involved the synthesis of the nitrate. Lepierre's work is still more recent, and is based upon several compounds. Lamy * gives the results of one analysis of thallium sulphate and three of thallium chloride. 3.423 grammes T1 2 S0 4 gave 1.578 grm. BaS0 4 ; whence 100 parts of the latter are equivalent to 216.920 of the former. In the thallium chloride the chlorine was estimated as silver chloride. The following results were obtained. In the third column I give the amount of T1C1 proportional to 100 parts of AgCl : 3.912 grm. T1C1 gave 2.346 grm. AgCl. 166.752 3.000 " 1.8015 u 166.528 3.912 " 2.336 " 167.466 Mean, 166.915, .1905 Hebberling's f work resembles that of Lamy. Reducing his weighings to the standards adopted above, we have from his sulphate series, as equivalent to 100 parts of BaS0 4 , the amounts of T1,S0 4 given in the third column : 1.4195 grm. T1 2 SO 4 gave .6534 grm. BaSO 4 . 217.248 1.1924 " .5507 " 216.524 .8560 " .3957 " 216.325 Mean, 216.699 Including Lamy's single result as of equal weight, we get a mean of 216.754, .1387. From the chloride series we have these results, with the ratio stated as usual : .2984 grm. T1C1 gave .1791 grm. AgCl. 166.611 .5452 " .3 2 78 " 166.321 Mean, 166,465, =b .097 Lamy's mean was 166.915, .1905. Both means combined give a general mean of 166.555, .0865. Werther'sJ determinations of iodine in thallium iodide were made by two methods. In. the first series Til was decomposed by zinc and potas- sium hydroxide, and in the filtrate the iodine was estimated as Agl. *Zeit. Anal. Chem., 2, 211. 1863. f Ann. Chem. Pharm., 134, n. 1865. % Journ. fur Prakt. Chem., 92, 128. 1864. THALLIUM. 185 One hundred parts of Agl correspond to the amounts of Til given in the last column : .720 grm. Til gave .51 grm. Agl. 141.176 2.072 " 1.472 " 140.761 .960 " .679 " 141-384 3 8 5 -273 " 141.026 1.068 " .759 " 140.711 Mean, 141.012, .085 In the second series the thallium iodide was decomposed by ammonia in presence of silver nitrate, and the resulting Agl was weighed. Ex- pressed according to the foregoing standard, the results are as follows : 1.375 grm. Til gave .978 grm. Agl. Ratio, 140.593 1.540 1.095 " " 140.639 1.380 " .981 " " 140.673 Mean, 140.635, db .016 General mean of both series, 140.648, .016. In 1873 Crookes,* the discoverer of thallium, published his final deter- mination of its atomic weight. His method was to effect the synthesis of thallium nitrate from weighed quantities of absolutely pure thallium. No precaution necessary to ensure purity of materials was neglected ; the- balances were constructed especially for the research ; the weights were accurately tested and all their errors ascertained ; weighings were made partly in air and partly in vacuo, but all were reduced to absolute stand- ards ; and unusually large quantities of thallium were employed in each experiment. In short, no effort was spared to attain as nearly as possi- ble absolute precision of results. The details of the investigation are too voluminous, however, to be cited here ; the reader who wishes to become familiar with them must consult the original memoir. Suffice it to say that the research is a model which other chemists will do well to copy. The results of ten experiments by Professor Crookes may be stated as follows. In a final column I give the quantity of nitrate producible from 100 parts of thallium. The weights given are in grains : Thallium. TINO^ + Glass. Glass Vessel. Ratio. 497.972995 1121.851852 472.557319 130.3875 293.193507 i i 11.387014 729.082713 130.393 288 562777 971.214142 594.949719 130.3926 324.963740 1142.569408 718.849078 1 30. 3900 183.790232 1005.779897 766.133831 130.3912 190.842532 997.334615 748.491271 130.3920 195.544324 1022.176679 767.203451 I30-39I5 201.816345 1013.480135 750.332401 130.3897 295.683523 H53.947672 768.403621 130.3908 299.203036 1159.870052 769.734201 130.3917 Mean, 130.3910, .00034 * Phil. Trans., 1873, p. 277. 186 THE ATOMIC WEIGHTS. Lepierre's* determinations were published in 1893, and represented several distinct methods. First, thallous sulphate was subjected to elec- trolysis in presence of an excess of ammonium oxalate, the reduced metal being dried and weighed in an atmosphere of hydrogen. The cor- rected weights, etc., are as follows: J - 8 935 grm. T1 2 SO 4 gave 1.5327 Tl. 80.945 per cent. 2.7243 " 2.2055 " 80.957 2.8112 " 2.2759 " 80.958 " Mean, 80.953, =t .0030 Secondly, weighed quantities of crystallized thallic oxide were con- verted into thallous sulphate by means of sulphurous acid, and the solu- tion was then subjected to electrolysis, as in the preceding series. 3.2216 grm. T1 2 O 8 gave 2.8829 Tl. 89.487 per cent. 2.5417 " 2.2742 " 89.475 Mean, 89.481, =h .0040 In the third set of experiments a definite amount of thallous sulphate or nitrate was fused in a polished silver crucible with ten times its weight of absolutely pure caustic potash. Thallic oxide was thus formed, which, with various precautions, was washed with water and alcohol, and finally weighed in the original crucible. One experiment with the nitrate gave 2.7591 grm. T1NO 3 yields 2.3649 T1 2 O 3 . 85.713 per cent. Two experiments were made with the sulphate, as follows : 3.1012 grm. T1 2 SO 4 gave 2.8056 T1 2 O 3 . 90.468 per cent. 2.3478 " 2.1239 " 90-463 " Mean, 90.465, .0020 Finally, crystallized thallic oxide was reduced by heat in a stream of hydrogen, and the water so formed was collected and weighed. 2.7873 grm. T1 2 O 3 gave .3301 H 2 O. 11.843 P er cent - 3.9871 " .4716 " 11.828 " 4.0213 " .4761 " 11-839 " Mean, 11.837, d= .0029 Iii a supplementary notef Lepierre states that his weights were all reduced to vacuum standards. Some work by Wells and Penfield, J incidentally involving a deter- mination of atomic weight, but primarily intended for another purpose, may also be taken into account. Their question was as to the constancy of thallium itself. The nitrate was repeatedly crystallized, and the last crystallization, with the mother liquor representing the opposite end of * Bull. Soc. Chim. (3), 9, 166. fBull. Soc. Chim. (3), n, 423. 1894. J Amer. Journ. Sci. (3), 47, 466. 1894. THALLIUM. 187 the series, were both converted into chloride. In the latter the chlorine was estimated as silver chloride, which was w r eighed on a Gooch filter, with the results given below, which are sensibly identical. The T1C1 equivalent to 100 parts of AgCl is stated in the last column. TICl. AgCl. Ratio. Crystals 3-9*46 2.3393 167.341 Mother liquor 3-34'5 1.9968 167. 343 Mean, 167.342 The general mean of Lamy's and Hebberling's determinations of this ratio gave 166.555, : .0865. If we arbitrarily assign Wells and Pen- field's mean equal weight with that, we get a new general mean of 166.948, .0610. The ratios to be considered are now as follows : (I.) BaSO 4 : T1 2 SO 4 : : 100 : 216.754, .1387 (2.) AgCl : TICl : : 100 : 166.948, .0610 (3.) Agl : Til : : 100 : 140.648, .016 (4.) Tl : T1NO 3 : : 100 : 130.391, + .00034 (5.) T1 2 S0 4 : T1 2 : : 100 : 80.953, .0030 (6.) T1 2 O 3 : T1 2 : : IOO : 89.481, .0040 ( 7 .) 2T1N0 3 : T1 2 3 : : 100:85.713 (8.) T1 2 SO 4 : T1 2 O 3 : : 100 : 90.465, .0020 (9.) T1 2 O 3 : 3H 2 O : : IOO : 11.837, .0029 And the antecedent data are these : = 15.879, db .0003 N = 13.935, db .0021 Ag= 107.108, =b .0031 S = 31.828, it .0015 Cl = 35.179, =fc .0048 AgCl = 142.287, i .0037 1 = 125.888, .0069 Agl = 232.996, db .0062 Ratio number seven rests upon a single experiment, and the atomic weight of thallium derived from it must therefore be arbitrarily weighted. It has been assumed, therefore, that its probable error is the same as that from number eight. Taking this much for granted, we have nine values for thallium, as given below : From (i) Tl = 203.478, .1610 Fro'm (2) " = 202.366, db .0872 From (3) " = 201.816, .0389 From (4) ' ' = 202. 595, .0117 From (5) " = 202.614, .0330 From (6) " = 202 620, .0775 From (7) " = 202.679, d= .0483 From (8) " = 202.496, .0483 From (9) '* = 202.746, d= .0576 General mean Tl = 202.555, .0098 If = 16, Tl = 204.098. 188 THE ATOMIC WEIGHTS. If we reject the first three values, retaining only those due to the ex- periments of Crookes and Lepierre, we have Tl = 202.605, .0103 If O = 16, this becomes 204.149. This mean exceeds Crookes' deter- mination only by 0.01, and may be regarded as fairly satisfactory. Crookes' ratio evidently outweighs all the others. SILICON. Although Berzelius * attempted to ascertain the atomic weight of silicon, first by converting pure Si into Si0 2 , and later from the analysis of BaSiF 6 , his results were not satisfactory. We need consider only the work of Pelouze, Schiel, Dumas, and Thorpe and Young. Pelouze,f experimenting upon silicon tetrachloride, employed his usual method of titration with a solution containing a known weight of silver. One hundred parts of Ag gave the following equivalencies of SiCl 4 : 39-4325 39.4570 Mean, 39.4447, .0083 Essentially the same method was adopted by Dumas. J Pure SiCl 4 was weighed in a sealed glass bulb, then decomposed by water, and titrated. The results for 100 Ag are given in the third column : 2.899 grm. SiCl 4 = 7.3558 grm. Ag. 39.41 1 1.242 " 3.154 " 39-379 3.221 8.1875 " 39.340 Mean, 39.377, db .014 Dumas' and Pelouze's series combine as follows : Pelouze 39.4447, dr .0083 Dumas 39.377, .014 General mean 39.4265, =fc .0071 Schiel, also studying the chloride of silicon, decomposed it by am- monia. After wanning and long standing it was filtered, and in the * Lehrbuch, 5 Aufl., 3, 1200. f Compt. Rend., 20, 1047. 1845. I Ann. Cheni. Pharm., 113, 31. 1860. Ann. Chem. Pharm., 120,94. SILICON. 189 filtrate the chlorine was estimated as AgCl. One hundred parts of AgCl correspond to the quantities of SiCl 4 given in the last column : .6738 grm. SiCl 4 gave 2.277 g rm - AgCl. 29.592 1.3092 " 4.418 " 29.633 Mean, 29.6125, .0138 Thorpe and Young,* working with silicon bromide, seem to have ob- tained fairly good results. The bromide was perfectly clear and color- less, and boiled constantly at 153. It was weighed, decomposed with water, and evaporated to dryness,the crucible containing it being finally ignited. The crucible was tared by one precisely similar, in which an equal volume of water was also evaporated. Results as follows, with weights at vacuum standards : 9.63007 grm. SiBr 4 gave 1.67070 SiO 2 . 17.349 per cent. 12.36099 " 2.14318 " 17.338 12.98336 2.25244 " 17-349 " 9.02269 " L5 6 542 " I7-350 " 15.38426 " 2.66518 " 17.324 " 9.74550 1.69020 " 17-343 6.19159 " 1.07536 " 17.368 " 9.51204 " 1.65065 " 17.353 " 10.69317 1.85555 " '7-353 " Mean, 17.347, .0027 The ratios now available are (i.) 4Ag : SiCl 4 : : 100 : 39.4265, .0071 (2.) 4AgCl : SiCl 4 : : 100 : 29.6125, =b .0138 (3.) SiBr 4 : SiO 2 : : loo : 17.347, .0027 Reducing these ratios with O = I 5-879, db .0003 Br = 79.344, .0062 Ag= 107.108, .0031 AgCl = 142.287, .0037, Cl =. 35.179, =h .0048 we have the following values for the atomic weight of silicon : From (i) Si = 28.200, .0363 From (2) " = 27.823, .0810 From (3) .... " = 28.187, =b .0122 General mean Si 28.181, .0114 If = 16, Si = 28.395. *Journ. Chem. Soc., 51,576. 1887. 190 THE ATOMIC WEIGHTS. TITANIUM. The earliest determinations of the atomic weight of titanium are due to Heinrich Rose.* In his first investigation he studied the conversion of titanium sulphide into titanic acid, and obtained erroneous results ; later, in 1829, he published his analyses of the chloride, f This compound was purified by repeated rectifications over mercury and over potassium, and was weighed in bulbs of thin glass. These were broken under water in tightly stoppered flasks ; the titanic acid was precipitated by ammo- nia, and the chlorine was estimated as silver chloride. The following results were obtained. In a fourth column I give the Ti0 2 in percentages referred to TiCl 4 as 100, and in a fifth column the quantity of TiCl 4 pro- portional to 100 parts of AgCl : TiCl. TiO T AgCl. Percent. TiO. 2 . AgCl Ratio. .885 grm. .379 grm. 2.661 grm. 42.825 33- 2 S8 2.6365 " 1. 120 " 7.954 " 42.481 33-147 I.7I57 " -73 2 " 5- I 72 " 42.665 33.173 3.0455 " 1.322 " 9.198 " 43.423 33-100 2.4403 " 1.056 ' 7.372 " 43-273 33.102 Mean, 42.933, .121 33.156, .019 If we directly compare the AgCl with the Ti0 2 we shall find 100 parts of the former proportional to the following quantities of the latter : 14.243 14.081 14-153 H.373 14.324 Mean, 14.235, .036 Shortly after the appearance of Rose's paper, MosanderJ published some figures giving the percentage of oxygen in titanium dioxide, from which a value for the atomic weight of titanium was deduced. Although no details are furnished as to experimental methods, and no actual weigh- ings are given, I cite his percentages for whatever they may be worth : 40.814 40.825 40.610 40. 1 80 40.107 40.050 40.780 40.660 39-830 Mean, 40.428 * Gilbert's Annalen, 1823, 67 and 129. t Poggend. Annalen, 15, 145. Berz. I,ehrbuch, 3, 1210. j Berz. Jahresbericht, 10, 108. 1831. TITANIUM. 191 These figures, with O = 15.879, give values for Ti ranging from 46.03 to 47.98; or, in mean, Ti = 46.80. They are not, however, sufficiently explicit to deserve any farther consideration. In 1847 Isidor Pierre made public a series of important determina- tions.* Titanium chloride, free from silicon and from iron, was pre- pared by the action of chlorine upon a mixture of carbon with pure, artificial titanic acid. This chloride was weighed in sealed tubes, these were broken under water, and the resulting hydrochloric acid was titrated with a standard solution of silver after the method of Pelouze. I subjoin Pierre's weighings, and add, in a third column, the ratio of TiCl 4 to 100 parts of silver : TiClt. Afr. Ratio. .8215 grm. 1.84523 gran. 44-5 2 .7740 " i.739 9 " 44-506 7775 " I.746I3 " 44.527 .7160 " 1.61219 " 44412 .8085 " 1.82344 " 44-339 .6325 " 1.42230 " 44.470 8155 " 1-83705 " 44-39.2 .8165 1.83899 " 44.399 .8065 " 1.81965 " 44.322 Mean, 44.432, .0173 It will be seen that the first three of these results agree well with each other and are much higher than the remaining six. The last four ex- periments were made purposely with tubes which had been previously opened, in order to determine the cause of the discrepancy. According to Pierre, the opening of a tube of titanium chloride admits a trace of atmospheric moisture. This causes a deposit of titanic acid near the mouth of the tube, and liberates hydrochloric acid. The latter gas being heavy, a part of it falls back into the tube, so that the remaining chloride is richer in chlorine and poorer in titanium than it should be. Hence, upon titration, too low figures for the atomic weight of titanium are obtained. Pierre accordingly rejects all but the first three of the above estimations. The memoir of Pierre upon the atomic weight of titanium was soon followed by a paper from Demoly, f who obtained much higher results. He also started out from titanic chloride, which was prepared 'from rutile. The latter substance was found to contain 1.8 per cent, of silica ; whence Demoly inferred that the TiCl 4 investigated by Rose and by Pierre might have been contaminated with SiCl 4 , an impurity which would lower the value deduced for the atomic weight under consideration. Accordingly, in order to eliminate all such possible impurities, this process was resorted *Ann. Chim. Phys. (3), 20, 257. t Ann. Chem. Pharm., 72, 214. 1849. 192 THE ATOMIC WEIGHTS. to : the chloride, after rectification over mercury and potassium, was acted upon by dry ammonia, whereupon the compound TiCl 4 .4NH 3 was deposited as a white powder. This was ignited in dry ammonia gas, and the residue, by means of chlorine, was reconverted into titanic chloride, which was again repeatedly rectified over mercury, potassium, and po- tassium amalgam. The product boiled steadily at 135. This chloride, after weighing in a glass bulb, was decomposed by water, the titanic acid was precipitated by ammonia, and the chlorine was estimated in the filtrate as silver chloride. Three analyses were performed, yielding the following results. I give the actual weighings : 1.470 grm. TiCl 4 gave 4.241 grm. AgCl and .565 grm. TiO 2 2.330 " 6.752 .801 " 2.880 " 8.330 " 1.088 " The ".801 " in the last column is certainly a misprint for .901. Assum- ing this correction, the results may be given in three ratios, thus : Per cent. TiOifrom TiClv TiCl: looAgCl. TiO 2 : woAgCL 38.435 34-662 13-322 38669 34. 508 13.344 37.778 34-574 13.061 Mean, 38.294, .180 34-58i, .030 13.242, zfc .061 These three ratios give three widely divergent values for the atomic weight of titanium, ranging from about 36 to more than 56, the latter figure being derived from the ratio between AgCl and TiCl 4 . This value, 56, is assumed by Demoly to be the best, the others being practically ignored. Upon comparing Demoly's figures with those obtained by Rose, certain points of similarity are plainly to be noted. Both sets of results were reached by essentially the same method, and in both the discordance between the percentages of titanic acid and of silver chloride is glaring. This discordance can rationally be accounted for by assuming that the titanic chloride was in neither case absolutely what it purported to be ; that, in brief, it must have contained impurities, such for example as hydrochloric acid, as shown in the experiments of Pierre, or possibly traces of oxy chlorides. Considerations of this kind also throw doubt upon the results attained by Pierre, for he neglected the direct estimation of the titanic acid altogether, thus leaving us without means for correctly judging as to the character of his material. In 1883* Thorpe published a series of experiments upon titanium tetrachloride, determining three distinct ratios and getting sharply con- cordant results. The first ratio, which was essentially like Pierre's, by * Berichte Deutsch. Chem. Gesell., 16, 3014. 1883. TITANIUM. 193 decomposition with water and titration with silver, was in detail as follows : 7VC/4. Ag. 7VC7 4 : iooAg. 2-43275 5.52797 44.008 5-42332 12.32260 44.015 3.59601 8.17461 44.000 3.31222 7.52721 44.003 4.20093 9.54679 44.004 5.68888 12.92686 44.008 5.65346 12.85490 43-979 4.08247 9.28305 43.978 Mean, 43.999, =b .0032 Pierre found, 44.432, d= .0073 General mean, 44.017, db .0031 The second ratio, which involved the weights of TiCl 4 taken in the last five determinations of the preceding series, included the weighing of the silver chloride formed. The TiCl 4 proportional to 100 parts of AgCl is given in a third column : 7Ya 4 . AgCl. Ratio. 3.31222 10.00235 33- "4 4.20093 12.68762 33.111 5.68888 17.17842 33.117 5.65346 17.06703 33- I2 5 4.08247 12.32442 33- I2 5 Mean, 33.118, .0019 Rose found, 33.156, =b .019 Demoly found, 34.581, .030 General mean, 33.123, .0019. In the third series the chloride was decomposed by water, and after evaporation to dry ness the resulting Ti0 2 was strongly ignited. TtC/ t . TiO v Percent. TiO.,. 6.23398 2.62825 42.160 8.96938 3./8335 42.181 10.19853 4.30128 42.176 6.56894 2.77011 42.170 8.99981 3-79575 42.176 8.32885 3-5"58 42.162 Mean, 42.171, .0022 Rose found, 42.933, dr .121 Demoly found, 38.294, .180 General mean, 42.171, .0022 In short, the work of Rose, Pierre, and Demoly practically vanishes. Furthermore, as will be seen later, the three ratios now give closely 13 194 THE ATOMIC WEIGHTS. agreeing values for the atomic weight of titanium. The cross ratio, 4AgCl : Ti0 2 is not directly given by either of Thorpe's series ; but the data furnished by Rose and Demoly combine into a general mean of 4AgCl : Ti0 2 : : 100 : 13.980, .0303. Some two years later Thorpe published his work more in detail,* and added a set of determinations, like those made upon the chloride, in which titanium tetrabromide was studied. Three ratios were measured, as was the case with the chloride. In the first, the bromide was decom- posed by water and titrated with a silver solution. TiBr. Ag. TiBr : rooAg. 2.854735 3-349 2 7 85.235 3.120848 3.66*22 85.241 4-73"i 8 5-5597 85.230 6.969075 8.17645 85.234 6.678099 7.83493 85.234 Mean, 85.235, + .0027 In the four last experiments of the preceding series, the silver bromide formed was weighed. The third column gives the TiBr 4 proportional to 100 parts of AgBr. TiBr^. AgBr, Ratio. 3.120848 6.375391 48.951 4.731118 9-663901 48.957 6.969075 14.227716 48.982 6.678099 I3-639956 48.959 Mean, 48.962, .0049 For the third ratio the bromide was decomposed by water ; and after evaporation with ammonia the residual titanic oxide was ignited and TiO.,. Percent. TiO 2 . 6.969730 1.518722 21.790 8.836783 1.923609 21.768 9.096309 . I-9795J3 21.762 Mean, 21.773, .0062 Ignoring Mosander's work as unavailable, we have the following ratios to consider : (I.) 4Ag : TiCl 4 : : 100 : 44.017, i .0031 (2.) 4AgCl : TiCl 4 : : 100 : 33- I2 3, .0019 (3.) 4AgCl : TiO 2 : : 100 : 13.980, =h .0303 (4.) TiCl 4 : TiO 2 :: 100 : 42.171, .0022 (5.) 4 Ag : TiBr, : : 100 : 85.235, .0027 (6.) 4AgBr : TiBr 4 : : 100 : 48.962, .0049 (7.) TiBr 4 : TiO 2 : : 100 : 21.773, .0062 * Journ. Chem. Soc., Feb., 1885, p. 108, and March, p. 129. GERMANIUM. 195 These are to be computed with O = -- 15.879, + .0003 Br 79.344, zb .0062 Ag = 107. 108, .0031 AgCl = 142.287, .0037 cl = 35- z 79, =b - 48 AgBr = 186.454, =b -54 For the molecular weight of titanium chloride they give two values : From (i) ...................... TiCl 4 == 188.583, .0144 From (2) ...................... " = 188.519, rb .0119 General mean ............. TiCl 4 = 188.545, .0092 For TiBr we have From (5) ...................... TiBr 4 = 365.i74, .0157 From (6) ...................... " = 365.163,^.0380 General mean ............ TiBr 4 = 365. 172, =b .0145 And for the atomic weight of titanium five values are calculable, as follows : From molecular weight of TiCl 4 ...... Ti = 47.829, rb .0213 From molecular weight of TiBr 4 . ..... " =47.796, rb .0260 From (3) .......................... " = 47.809. .1725 From (4) ---- , ..................... " =47.698, .0268 From (7) ....... .................. " = 47-738, .0787 General mean ..... . ........... Ti = 47.786, d= .0138 If = 16, this becomes Ti = 48.150. GERMANIUM. The data relative to the atomic weight of germanium are rather scanty, and are due entirely to the discoverer of the element, Winkler.* The pure tetrachloride was decomposed by sodium carbonate, mixed with a known excess of standard silver solution, and then titrated back with ammonium sulphocyanate. The data given are as follows : Cl Found. Percent. Cl. .1067 .076112 66.177 .1258 .083212 66.146 .2223 .147136 66.188 .2904 .192190 66.182 Mean, 66.173 Hence, with Cl = 35.179, Ge = 71.933. If O = 16, Ge = 72.481. * Journ. fiir Prakt. Chem. (2), 34, 177. 1886. 196 THE ATOMIC WEIGHTS. ZIRCONIUM. The atomic weight of zirconium has been determined by Berzelius, Hermann, Marignac, Weibull, and Bailey. Berzelius* ignited the neutral sulphate, and thus ascertained the ratio in it between the Zr0 2 and the SO 3 . Putting S0 3 at 100, he gives the following proportional quantities of Zr0 2 : 75-84 75-92 75.80 75-74- 75-97 75.85 Mean, 75.853, .023 This gives 43.134, .0142 as the percentage of zirconia in the sulphate. Hermann's t estimate of the atomic weight of zirconium was based upon analyses of the chloride, concerning which he gives no details nor weighings. From sublimed zirconium chloride he finds Zr = 831.8, when = 100; and from two lots of the basic chloride 2ZrOCl 2 ,9H 2 0, Zr = 835.65 and 851.40 respectively. The mean of all three is 839.62 ; whence, with modern formulae and O = 15.879, Zr becomes = 88.882. Marignac's results J were obtained by analyzing the double fluoride of zirconium and potassium. His weights are as follows : i.ooo grm. gave .431 grm. ZrO 2 and .613 grm. K 2 SO 4 . 2.000 " .864 " 1.232 " .654 " .282 " .399 5.000 " 2.169 3-78 " These figures give us three ratios. A, the Zr0 2 from 100 parts of salt; B, the K 2 SO 4 from 100 parts of salt ; and C, the ZrO 2 proportional to 100 parts of K 2 SO, : A. B. C. 43.100 61.300 70-3 10 43.200 61.600 70.130 43.119 61.000 70.677 43.380 61.560 70.468 Mean, 43.200, d= .043 Mean, 61.365, =h .094 Mean, 70.396, 079. Weibull, following Berzelius, ignited the sulphate, and also made a *Poggend. Annal , 4, 126. 1825. t Journ. fi'ir Prakt. Chem., 31, 77. Berz. Jahresb., 25, 147. jAnn. Chim. Phys. (3), 60, 270. 1860. I Lund. Arsskrift, v. 18. i88i-'82. ZIRCONIUM, 197 similar set of experiments with the selenate of zirconium, obtaining re- sults as follows : Sulphate. Zr(SO^ v 1.5499 grm. salt gave .6684 ZrO 2 . 43.126 per cent. 1-5445 " -6665 " 43- r 53 " 2.1683 " .9360 " 43.168 " 1.0840 " .4670 " 43.081 " .7913 " .3422 " 43-321 " .6251 .2695 " 43. 113 " .4704 .2027 " 43. 9i " Mean, 43.150, =fc .0207 Selenate. Zr(SeO^. i. 02 1 2 grm. salt gave .3323 ZrO 2 . 32.540 per cent. .8418 " .2744 " 32.597 " .6035 .1964 " 32.544 " .8793 .2870 " 32.640 " .3089 " .1003 " 3 2 .470 " Mean, 32.558, .0192 Bailey * also ignited the sulphate, after careful investigation of his material, and of the conditions needful to ensure success. He found that the salt was perfectly stable at 400, while every trace of free sulphuric acid was expelled at 350. The chief difficulty in the process arises from the fact that the zirconia produced by the ignition is very light, and easily carried off mechanically, so that the percentage found is likely to be too low. This difficulty was avoided by the use of a double crucible, the outer one retaining particles of zirconia which otherwise might be lost. The results, corrected for buoyancy of the air, are as follows : 2.02357 salt gave .87785 ZrO a . 43-38i per cent. 2.6185 " I.I3S4 " 43-36o " 2.27709 " .98713 " 43-35 " 2.21645 " -96152 " 43-385 " L75358 " .76107 " 43-402 " 1.64065 " .7120 " 43.397 " 2.33255 " 1.01143 " 43.36i " 1.81105 " .78485 " 43-337 " Mean, 43.372, .0056 This, combined with previous determinations, gives Berzelius 43. 134, .0142 Weibull 43. 150, .0207 Bailey 43-37 2 , .0056 General mean 43-3 I 7, .0051 * Proc. Roy. Soc., 46, 74. Chem. News, 60, 32. 198 THE ATOMIC WEIGHTS. For computing the atomic weight of zirconium we now have the sub- joined ratios : (i.) Percentage ZrO 2 in Zr(SO 4 ) 2 , 43.317, .0051 (2.) Percentage ZrO 2 in Zr(SeO 4 ) 2 , 32.558, .0192 (3.) Percentage ZrO 2 from K 2 ZrF 6 , 43.200, .043 (4.) Percentage K 2 SO 4 from K 2 ZrF 6 , 61.365, .094 (5.) K 2 S0 4 : Zr0 2 : : 100 : 70.396, .079 Tlie antecedent atomic weights are O = 15.879, .0003 K = 38.817, dt .0051 S =31.828, .0015 F = 18.912, .0029 Se = 78.419, .0042 With these data we first get three values for the molecular weight of zirconia : From (i) ZrO 2 121.454, .0182 From (2) " = 121.708, .0798 From (5) " = 121.770, .1370 General mean ZrO 2 121.471, .0176 Finally, there are three independent estimates for the atomic weight of zirconium : From molecular weight ZrO 2 Zr = 89.713, d= .0177 From ratio (3) " = 89.437, .2390 From ratio (4) " 90.778, .4326 General mean Zr = 89.716, .0175 If = 16, Zr 90.400. Here the first value alone carries appreciable weight. TIN. 199 TIN. The atomic weight of tin has been determined by means of the oxide, the chloride, the bromide, the sulphide, and the stannichlorides of potas- sium and ammonium. The composition of stannic oxide has been fixed in two \vays : by synthesis from the metal and by reduction in hydrogen. For the first method we may consider the work of Berzelius, Mulder and Vlaanderen, Dumas, Van der Plaats, and Bongartz and Classen. Berzelius * oxidized 100 parts of tin by nitric acid, and found that 127.2 parts of Sn0 2 were formed. The work done by Mulder and Vlaanderen f was done in connection with a long investigation into the composition of Banca tin, which was found to be almost absolutely pure. For the atomic weight determina- tions, however, really pure tin was taken prepared from pure tin oxide. This metal was oxidized by nitric acid, with the following results. 100 parts of tin gave of SnO 2 : 127.56 Mulder. 127.56 Vlaanderen. 1 27.43 Vlaanderen. Mean, 127.517, .029 Dumas J oxidized pure tin by nitric acid in a flask of glass. The re- sulting Sn0 2 was strongly ignited, first in the flask and afterwards in platinum. His weighings, reduced to the. foregoing standard, give for dioxide from 100 parts of tin the amounts stated in the third column : 12.443 grm. Sn gave 15.820 grm. SnO 2 . 127.14 15.976 " 20.301 " 127.07 Mean, 127.105, =b .024 In an investigation later than that previously cited, Vlaanderen found that when tin was oxidized in glass or porcelain vessels, and the resulting oxide ignited in them, traces of nitric acid were retained. When, on the other hand, the oxide was strongly heated in platinum, the latter was perceptibly attacked, so much so as to render the results uncertain. He therefore, in order to fix the atomic weight of tin, reduced the oxide by heating it in a porcelain boat in a stream of hydrogen. Two experiments gave Sn = 118.08, and Sn = 118.24. These, when =s 16, become, if reduced to the above common standard, *Poggend. Annal., 8, 177. t Journ. fur Prakt. Chem., 49, 35. 1849. t Ann. Chem. Pharm., 113, 26. 3 Jahresbericht, 1858, 183. 200 THE ATOMIC WEIGHTS. 127.100 127.064 ^ Mean, 127. 082, =b .012 Van der Plaats * prepared pure stannic oxide from East Indian tin (Banca), and upon the material obtained made two series of experiments J one by reduction and one by oxidation. The results, with vacuum weights, are as follows, the ratio between Sn and Sn0 2 appearing in the third column : Oxidation Series. 9.6756 grm. tin gave 12.2967 SnO 2 . 127.091 12.7356 16.1885 " 127.114 23.4211 " 29.7667 " 127.093 Reduction Series. 5-5 OI 5 g rm - Sn 2 S ave 4.3280 tin. 127. 1 14 4.9760 3-9 T 45 " 127.117 3.8225 " 3.0278 " 127.086 2.9935 " 2.3553 " 127.096 Mean of both series as one, 127.102, .0033 The reductions were effected in a porcelain crucible. Bongartz and Classen f purified tin by electrolysis, and oxidized the electrolytic metal by means of nitric acid. The oxide found was dried over a water-bath, then heated over a weak flame, and finally ignited for several hours in a gas-muffle. Some reduction experiments gave values which were too low. The oxidation series was as follows, with the usual ratio added by me in a third column : Sn. SnO^ Ratio. 2.5673 3- 2 57o 126.865 3.8414 4-8729 126.852 7.3321 9. 2 994 126.831 5.43 6 7 6.8962 126.845 7.3321 9- 2 994 126.831 9.8306 12.4785 126.935 11.2424 14.2665 126.896 5.5719 7.0685 126.860 9.8252 12.4713 126.932 4-3959 5-5795 126.925 6.3400 8.0440 126.877 Mean, 126.877, .0080 - We now have six series of experiments showing the amount of SnO a formed from 100 parts of tin. To Berzelius' single determination may be assigned the weight of one experiment in Mulder and Vlaanderen's series : * Corapt. Rend., 100, 52. 1885. fBerichte Deutsch. Chem. Gesell., 21, 2900. 1888. TIN. 201 Berzelius 127.20x3, .041 Mulder and Vlaanderen 127.517, dr .029 Dumas 127. 105, .024 Vlaanderen 127.082, d= .012. Van der Plaats 127. 102, .0033 Bongartz and Classen 126.877, .0080 General mean , 127.076, dr .0026 Dumas, in the paper previously quoted, also gives the results of some experiments with stannic chloride, SnCl 4 . This was titrated with a solu- tion containing a known weight of silver. From the weighings given, 100 parts of silver correspond to the quantities of SnCl 4 named in the third column : 1.839 grm. SnC! 4 3.054 grm. Ag. 60.216 2.665 4.427 " 60.199 Mean, 60.207, =t '6 Tin tetrabromide and the stannichlorides of potassium and ammonium were all studied by Bongartz and Classen ; who, in each compound, carefully purified, determined the tin electrolytically. The data given are as follows, the percentage columns being added by myself: Taken. 8.5781 9-5850 9.9889 10.4914 16.8620 16.6752 11.1086 10.6356 11.0871 19.5167 Tin Tetrabromide. Sn Found. 2.3270 2.6000 2.7115 2.8445 4.5735 4.5236 2.8840 3.0060 5-2935 Percent. Sn. 27.127 27.126 27.145 27.113 27.123 27.119 27.116 27.113 27.123 27.128 Mean, 27.123, dr .0020 2.5718 2.2464 9-3353 12.1525 12.4223 15.0870 10.4465 18.9377 18.4743 17.6432 Potassium Stannichloride. Sn Found. Per cent. Sn. .7472 29.054 .6524 29.042 2.7100 29.030 3-5285 ^ 29.035 3.6070 29.036 4.3812 29.040 3.0330 29.034 5.5029 29.058 5.3630 29.029 5.1244 29.045 Mean, 29.040, .0021 202 THE ATOMIC WEIGHTS. Ammonium Stannichloride. Am^SnQl^ Sn Found. Per cent. Sn. 1-6448 .5328 3 2 .393 1.8984 .6141 32.347 2.0445 .6620 32.381 2.0654 .6690 32.391 2.0058 .6496 32.386 2.4389 .7895 32.371 4.0970 L3254 32. 35 1 3.4202 1.1078 32.390 3.6588 1.1836 32.349 1.5784 .5108 32-362 7.3248 2.3710 32.37 13.1460 4.2528 32.351 11.9483 3-8650 32.348 18.4747 5.9788 32-362 18.6635 6.0415 32.371 17.8894 5.7923 32.378 Mean, 32.369, .0088 One other method of determination for the atomic weight of tin was employed by Bongartz and Classen. Electrolytic tin was converted into sulphide, and the sulphur so taken up was oxidized by means of hydrogen peroxide, by Classen's method, and weighed as barium sulphate. The results, as given by the authors, are subjoined : Sn Taken. Per cent, of S Gained. 2.6285 53.91 7495 53.87 1.4785 53-94 2.5690 53.94 2.1765 53.85 1.3245 53-88 9897 53.83 2.7160 53-86 Mean, 53.885, =h .0098 This percentage of sulphur, however, was computed from weighings of barium sulphate. What values were assigned to the atomic weights of barium and sulphur is not stated, but as Meyer and Seubert's figures are used for other elements throughout this paper, we may assume that they apply here also. . Putting O = 15.96, S = 31.98, and Ba = 136.86, the 53.885 per cent, of sulphur becomes 392.056, .0713 of BaS0 4 , the compound actually weighed. This gives us the ratio Sn : 2BaSO 4 : : loo : 392.056, d= .0713 as the real result of the experiments, from which, with the later values for Ba, S, and 0, the atomic weight of tin may be calculated. TIN. 203 We now have, for tin, the following available ratios : (l.) Sn : SnO 2 : : loo : 127.076, dr .0026 (2.) 4Ag : SnG 4 : : 100 : 60.207, -0060 (3.) Percentage of tin in SnBr 4> 27.123, .0020 (4.) Percentage of tin in K 2 SnCl 6 , 29.040, .0021. (5.) Percentage of tin in Am 2 SnCI 6 , 32.369, .0088 (6.) Sn : 2BaSO 4 : : 100 : 392.056, .0713 The antecedent values are O = 15.879, .0003 K= 38.817, d= .0051 Ag = 107.108, rb .0031 N = 13.935, .0021 Cl = 35.179, .0048 S = 31.828, .0015 Br = 79.344, dr .0062 Ba = 136.392, .0086 With these, six independent values for Sn are computable, as follows : From (i). Sn 117.292, .0115 From (2) " = 117.230, =h -0331 From (3) " = 1 18.120, .0131 From (4) " = 118.152, d=. 0155 From (5) " = 118.190, .0382 From (6) " = 118.216, .0220 General mean Sn = 1 17.805, .0069 If = 16, Sn = 118.701. If we reject the first two of these values, which include all of the older work, and take only the last four, which represent the concordant results of Bongartz and Classen, the general mean becomes Sn 1 1 8. 150, =b .0089 Or, with O = 16, Sn = 119.050. This mean I regard as having higher probability than the other. A single determination of the atomic weight of tin, made by Schmidt,* ought not to be overlooked, although it was only incidental to his research upon tin sulphide. In one experiment, 0.5243 grm. Sn gave 0.6659 Sn0 2 . Hence, with = 16, Sn = 118.49. This lies about midway between the two sets of values already computed. * Berichte, 27, 2743. 1894. 204 THE ATOMIC WEIGHTS. THORIUM. The atomic weight of thorium has been determined from analyses of the sulphate, oxalate, formate, and acetate, with widely varying results. The earliest figures are due to Berzelius,* who worked with the sulphate, and with the double sulphate of potassium and thorium. The thoria was precipitated by ammonia, and the sulphuric acid was estimated as BaS0 4 . The sulphate gave the following ratios in two experiments. The third column represents the weight of ThO 2 proportional to 100 parts of BaSO, : 6 754 g rm - ThO 2 1.159 grm. BaSO 4 . Ratio, 58.274 1.0515 " 1.832 " 57.396 The double potassium sulphate gave .265 grm. Th0 2 , .156 grin. S0 3 , and .3435 K 2 S0 4 . The S0 3 , with the Berzelian atomic weights, repre- sents .4537 grm. BaS0 4 . Hence 100 BaSO 4 is equivalent to 58.408 Th0 2 . This figure, combined with the two previous values for the same ratio, gives a mean of 58.026, .214. From the ratio between the K 2 S0 4 and the Th0 2 in the double sul- phate, Th0 2 = 266.895. In 1861 new determinations were published by Chydenius.t whose memoir is accessible to me only in an abstract J which gives results with- out details. Thoria is regarded as a monoxide, ThO, and the old equiv- alents (O = 8) are used. The following values are assigned for the molecular weight of ThO, as found from analyses of several salts : From Sulphate. From K. Th. Sulphate. 66.33 67. 2 67.13 67.75 68.03 Mean, 67.252, d= .201 From Acetate. From Formate. From Oxalate. 67.31 68.06 65.87-^ Two results 66.59 67.89 65.95 j by Berlin. 67.27 68.94 65.75 67.06 65.13 68.40 Mean, 68.297, rb .219 6654 65.85 Mean, 67.326, .201 Mean, 65.85, .123 * Poggend. Annal., 16, 398. 1829. Lehrbuch, 3, 1224. t Keraisk undersokning af Thorjord och Thorsalter. Helsingfors, 1861. An academic disser- tation. I Poggeud. Annal., 119, 55. 1863. THORIUM. 205 We may fairly assume that these figures were calculated with = 8, C = 6, and S = 16. Correcting by the values for these elements which have been found in previous chapters, Th0 2 becomes as follows : From sulphate ThO 2 = 267.170, rfc .7950 From acetate " = 267.488, .7950 From formate " = 271.239, .8698 From oxalate " = 261.478, d= .4884 General mean ............. ThO 2 = 265.103, -3394 The single result from the double potassium sulphate is included with the column from the ordinary sulphate, and the influence of the atomic weight of potassium is ignored. Chydenius was soon followed by Marc Delafontaine, whose researches appeared in 1863.* This chemist especially studied thorium sulphate ; partly in its most hydrous form, partly as thrown down by boiling. In Th(S0 4 ) 2 .9H 2 0, the following percentages of Th0 2 were found : 45.08 44.90 45.06 45- 21 45.06 Mean, 45.062, dz .0332 The lower hydrate, 2Th(SO 4 ) 2 .9H 2 0, was more thoroughly investi- gated. The thoria was estimated in two ways : First (A), by precipita- tion as oxalate and subsequent ignition ; second (B), by direct calcination. These percentages of Th0 2 were found : 52.83! 52.72 52.I3J 52.47 52.49 52.53 52-13 52.13 52.43 52.60 52.40 52.96 52.82 Mean, 52.511, .047 In three experiments with this lower hydrate the sulphuric acid was also estimated, being thrown down as barium sulphate after removal of the thoria : *Arch. Sci. Phys. et Nat. (2), 18, 343. 206 THE ATOMIC WEIGHTS. 1.2425 grm. gave .400 SO 3 . (1.1656 grm. BaSO 4 .) 1.138 " .366 " (1.0665 " ) .734 .2306 ( .6720 ) The figures in parentheses are reproduced by myself from Delafon- taine's results, he having calculated his analyses with O = 100, S = 200, and Ba = 857. These data may be reduced to a common standard, so as to represent the quantity of 2Th(S0 4 ) 2 .9H 2 0, equivalent to 100 parts of BaS0 4 . We then have the following results : 106.597 106.704 109.226 Mean, 107.509, .585 Delafontaine was soon followed by Hermann,* who published a single analysis of the lower hydrated sulphate, as follows : Th0 2 52.87 S0 3 32.11 H 2 15.02 IOO.OO Hence, from the ratio between S0 3 and Th0 2 , Th0 2 = 262.286. Prob- ably the S0 3 percentage was loss upon calcination. Both Hermann's results and those of Delafontaine are affected by one serious doubt, namely, as to the true composition of the lower hydrated sulphate. The latest and best evidence seems to establish the fact that it contains four molecules of water instead of four and a half,f a fact which tends to lower the resulting atomic weight of thorium consid- erably. In the final discussion of these data, therefore, the formula Th(S0 4 ) 2 .4H 2 will be adopted. As for Hermann's single analysis, his percentage of Th0 2 , 52.87, may be included in one series with Delafon- taine's, giving a mean of 52.535, .0473. The next determinations to consider are those of Cleve,J whose results, obtained from both the sulphate and the oxalate of thorium, agree ad- mirably. The anhydrous sulphate, calcined, gave the subjoined per- centages of thoria : 62.442 62.477 62.430 62.470 62.357 62.366 Mean, 62.423, .014 * Journ. fur Prakt. Chetn., 93, 114. t See Hillebrand, Bull. 90, U. S. Geol. Survey, p. 29. I K. Sveuska Vet. Akad. Handling., Bd. 2, No. 6, 1874. THORIUM. 207 The oxalate was subjected to a combustion analysis, whereby both thoria and carbonic acid could be estimated. From the direct percentages of these constituents no accurate value can be deduced, there having undoubtedly been moisture in the material studied. From the ratio between C0 2 and Th0 2 , however, good results are attainable. This ratio I put in a fourth column, making the thoria proportional to 100 parts of carbon dioxide : Oxalate. ThO^. CO.,. Ratio. I -7 I 35 S rm - 1.0189 grm. .6736 grm. 151.262' 1.3800 " .8210 " .5433 " 151.114 1.1850 " .7030 " -4650 " 151.183 1.0755 " .6398 " .4240 " 150.896 Mean, 151.114, .053 Iii 1882, Nilson's determinations appeared.* This chemist studied both the anhydrous sulphate, and the salt with nine molecules of water, using the usual calcination method, but guarding especially against the hygroscopic character of the dry Th (SOJ 2 and the calcined Th0 2 . The hydrated sulphate gave results as follows : Percent. ThO.,. 2.0549 .9267 45.097 2.1323 .9615 45-092 3.0017 1.3532 45-081 2.7137 1.2235 45-086 2.6280 1.1849 45.088 1.9479 .8785 45.. 099 Mean, 45.091, .0019 Delafontaine found, 45.062, it .0332 General mean, 45.090, .0019 The anhydrous sulphate gave data as follows : Th(SO^. ThO v Percent. 1.4467 -9013 62.300 1.6970 1.0572 62.298 2.0896 1.3017 62.294 1.5710 .9787 62.298 Mean, 62.297, =b .0009 The last four determinations appear again in a paper published five years later by Kriiss and Nilson,f who, however, give four more made *Ber. Deutsch. Chem. Gesell., 15, 2519. 1882. f Ber. Deutsch. Chem. Gesell., 20, 1665. 1887. 208 THE 'ATOMIC WEIGHTS. upon material obtained from a different source. The new data are sub- joined : Th(SOJ v ThO 2 . Percent. ThO.,. 1.1630 .7245 62.296 .8607 .5362 62.298- 1.5417 .9605 62.301 1.5217 .9479 62.292 Mean, 62.297, .0013 Nilson's series, 62.297, .0009 Cleve found, 62.423, zb .0140 General mean, 62.298, .0007 From Chydenius' work we have four values for the molecular weight of thoria, which, combined as usual, give a general mean of Th0 2 = 265.103, db .3394. We also have the following ratios : (I.) 2BaSO 4 : ThO 2 : : ZOO : 58.026, dz .214 (2.) 2BaSO 4 : Th(SO 4 ) 2 .4H 2 O : : 100 : 107.509, .585 (3.) 4CO 2 : ThO 2 :: 100 : 151.114, .053 (4.) Percentage of ThO 2 in Th(SO 4 ) 2 .9H 2 O, 45.090, =b .0019 (5.) Percentage of ThO 2 in Th(SO 4 ) 2 .4H 2 O, 52.535, .0473 (6.) Percentage of ThO 2 in Th(SO 4 ) 2 .62.298, =h .0007 Reducing with the following data, seven values for the atomic weight of thoria are calculable : O = 15.879, .0003 C = 11.920, .0004 S = 31.828, .0015 Ba = 136.392, .0086 The values for Th0 2 are Chydenius' determinations ThO 2 265.103, -3394 From(i) =268.937, -9919 From (2) " rr= 268.021, 2.7115 From (3) " =264. 1 20, dr .0927 From (4) " =262.641, .0149 From (5) " = 255.061, .3426 From (6) " =262.613, .0081 General mean ThO 2 = 262.626, .0071 Hence Th = 230.868, .0071. If = 16, Th = 232.626. PHOSPHORUS. 209 PHOSPHORUS. The material from which we are to calculate the atomic weight of phosphorus is by no means abundant. Berzelius, in his Lehrbuch,* adduces only his own experiments upon the precipitation of gold by phosphorus, and ignores all the earlier work relating to the composition of the phosphates. These experiments have been considered with refer- ence to gold. Pelouze,t in a single titration of phosphorus trichloride with a stand- ard solution of silver, obtained a wholly erroneous result ; and Jacque- lain, J in his similar experiments, did even worse. Schrdtter's criticism upon Jacquelain sufficiently disposes of the latter. Only the determinations made by Schrotter, Dumas, and Van der Plaats remain to be considered. Schrotter || burned pure amorphous phosphorus in dry oxygen, and weighed the pentoxide thus formed. One gramme of P yielded P 2 3 in the following proportions : 2.28909 2.28783 2.29300 2.28831 2.29040 2.28788 2.28848 2.28856 2.28959 2.28872 Mean, 2.289186, =h .60033 Dumas ^| prepared pure phosphorus trichloride by the action of dry chlorine upon red phosphorus. The portion used in his experiments boiled between 76 and 78. This was titrated with a standard solution of silver in the usual manner. Dumas publishes weights, from which I calculate the figures given in the third column, representing the quantity of trichloride proportional to 100 parts of silver : 1.787 grm. PC1 3 = 4.208 grm. Ag. 42.4667 1.466 " 3.454 " 42.4435 2.056 " 4.844 " 42.4443 2.925 " 6.890 " 42.4528 3.220 7.582 " 42.4690 Mean, 42.4553, d= .0036 *5th ed., 1188. fCompt. Rend., 20, 1047. J Compt. Rend., 33, 693. % Journ. fur Prakt. Cheni., 57, 315. || Journ. fur Prakt. Chera., 53, 435. 1851. 11 Ann. Chem. Pharm., 113, 29. 1860. 14 210 THE ATOMIC WEIGHTS. By Van der Plaats* three methods of determination were adopted, and all weights were reduced to vacuum standards. First, silver was precipitated from a solution of the sulphate by means of phosphorus. The latter had been twice distilled in a current of nitrogen. The silver, before weighing, was heated to redness. The phosphorus equivalent to 100 parts of silver is given in the third column. .9096 grm. P gave 15.8865 Ag. 5-7 2 56 .5832 " 10.1622 " 5.7389 Mean, 5.7322, .0045 The second method consisted in the analysis of silver phosphate ; but the process is not given. Van der Plaats states that it is difficult to be sure of the purity of this salt. 6.6300 grm. Ag 3 PO 4 gave 5.1250 Ag. 77.3 P er cent. 12.7170 " 9.8335 " 77.326 " Mean, 77.313, .0088 In the third set of determinations, yellow phosphorus was oxidized by oxygen at reduced pressure, and the resulting P 2 5 was weighed. 10.8230 grm. P gave 24.7925 P 2 O 5 . Ratio, 2 29072 7.7624 I7-79 J 5 " " 2.29201 As these figures fall within the range of Schrotter's, they maybe aver- aged in with his series, the entire set of twelve determinations giving a mean of 2.28955, .00032. From the following ratios an equal number of values for P may now be computed : (i.) 2P : P 2 O 3 : : l.o : 2.28955, .00032 (2.) 3Ag : PC1 3 : : 100 : 42-4553, -0036 (30 5 A S : p : : I0 : 5-7322, .0045 (4.) Ag 3 PO 4 : 3Ag : : 100 : 77.313, .0088 Starting with = 15.879, .0003, Ag = 107.108, .0031, and Cl = 35.179, .0048, we have From (i) P = 30.784, =fc .0077 From (2) " = 30.882, .0189 From (3) " 30.698, =b .0241 From (4) " = 30.774, .0382 General mean P = 30.789, .0067 If = 16, P = 31.024. The highest of these figures is that from ratio number two, represent- ing the work of Dumas. This is possibly due to the presence of oxy- chloride, in traces, in the trichloride taken. Such an impurity, if present, would tend to raise the apparent atomic weight of phosphorus. *Compt. Rend., 100, 52. 1885. VANADIUM. 211 VANADIUM. Roscoe's determination of the atomic weight of vanadium was the first to have any scientific value. The results obtained by Berzelius * and by Czudnowicz f were unquestionably too high, the error being probably due to the presence of phosphoric acid in the vanadic acid employed. This particular impurity, as Roscoe has shown, prevents the complete reduction of V 2 O 5 to V 2 O 3 by means of hydrogen. All vanadium ores contain small quantities of phosphorus, which can only be detected with ammonium molybdate a reaction unknown in Berzelius' time. Fur- thermore, the complete purification of vanadic acid from all traces of phosphoric acid is a matter of great difficulty, and probably never was accomplished until Roscoe undertook his researches. In his determination of the atomic weight, Roscoe J studied two com- pounds of vanadium, namely, the pentoxide, V 2 O 5 , and the oxychloride, VOC1 3 . The pentoxide, absolutely pure, was reduced to V 2 O 3 by heating in hydrogen, with the following results : 7-7397 g rm - V 2 O 5 gave 6.3827 grm. V 2 O 3 . 17-533 P er cent, of loss. 6.5819 5-4296 " i7-57 5-1895 4-2819 " 17.489 5.0450 4.1614 " 17.515 5. 4296 grm. V 2 O 3 , reoxidized, gave 6. 5814 grm. V 2 O 5 . 17.501 per cent, difference. Mean, 17.509, =b .005 Hence V = 50.993, .0219. Upon the oxychloride, VOC1 3 , two series of experiments were made one volumetric, the other gravimetric. In the volumetric series the com- pound was titrated with solutions containing known weights of silver, which had been purified according to the methods recommended by Stas. Roscoe publishes his weighings, and gives percentages deduced from them ; his figures, reduced to a common standard, make the quan- tities of VOCL given in the third column proportional to 100 parts of silver. He was assisted by two analysts : Analyst A. 2.4322 grm. VOC1 3 = 4.5525 grm. Ag. 53.425 4.6840 " 8.7505 53.528 4.2188 1 1 7.8807 " 53-533 3-949 " 7-3799 53-5'Q 9243 < < 1.7267 53-530 1-4330 " 2.6769 53.532 * Poggend. Annal., 22, 14. 1831. t Poggend. Annal., 120, 17. 1863. t Journ. Chem. Soc., 6, pp. 330 and 344. 1868. 212 THE ATOMIC WEIGHTS. Analyst B. 2.853ogrm. VOCI 3 = 5.2853 grm. Ag. 53-98o 2.1252 " 3-9535 " 53-755 1.4248 " 2.6642 " 53-479 Mean, 53.586, =b .039 The gravimetric series, of course, fixes the ratio between VOC1 3 and AgCl. If we put the latter at 100 parts, the proportion of VOC1 3 is as given in the third column : Analyst A. 1.8521 grm. VOC1 3 gave 4.5932 grm. AgCl. 40.323 .7013 " 1.7303 " 40.53 1 .7486 1.8467 " 40.537 1.4408 3-57I9 " 40.337 9453 2.3399 " 40.399 1.6183 " 4.0282 " 40.174 Analyst B. 2.1936 grm. VOC1 3 gave 5.4039 grm. AgCl. 40.391 2.5054 " 6.2118 " 40.333 Mean, 40.378, b .028 These two series give us two values for the molecular weight of VOC1 3 : From volumetric series . . VOC1 3 = 172.185, rb .1254 From gravimetric series " = 172.358, .1196 General mean VOC1 3 172.277, zfc .0866 Hence V = 50.881, .0877. Combining the two values for V, we have : From VOC1 3 V = 50.881, .0877 From V 2 O 5 " = 50.993, .0219 General mean V = 50.986, .0212 If = 16, V = 51.376. These values are calculated with = 15.879, .0003; Cl = 35.179, .0048; Ag = 107.108, .0031, and AgCl = 142.287, .0037. ARSENIC. 213 ARSENIC. For the determination of the atomic weight of arsenic three compounds have been studied the chloride, the trioxide, and sodium pyroarsenate. The bromide may also be considered, since it was analyzed by Wallace in order to establish the atomic weight of bromine. His series, in the light of more recent knowledge, may properly be inverted, and applied to the determination of arsenic. In 1826 Berzelius * heated arsenic trioxide with sulphur in such a way that only S0 2 could escape. 2.203 grammes of As 2 3 , thus treated, gave a loss of 1.069 of S0 2 . Hence As = 74.460. In 1845 Pelouzef applied his method of titration with known quan- tities of pure silver to the analysis of the trichloride of arsenic, AsCl 3 . Using the old Berzelian atomic weights, and putting Ag = 1349.01 and Cl = 443.2, he found in three experiments for As the values 937.9, 937.1, and 937.4. Hence 100 parts of silver balance the following quantities of AsCl s : 56.029 56.009 56.016 Mean, 56.018, .004 Later, the same method was employed by Dumas, J whose weighings, reduced to the foregoing standard, give the following results : 4.298 grm. AsCl 3 = 7.673 grm. Ag. Ratio, 56.015 5.535 " 9.880 " " 56.022 7.660 " 13.686 " " 55-97 4-680 " 8.358 " " 55-993 Mean, 56.000, -_h .008 The two series of Pelouze and Dumas, combined, give a general mean -of 56.014, .0035, as the amount of AsCl 3 equivalent to 100 parts of silver. Hence As = 74.450, .019, a value closely agreeing with that deduced from the single experiment of Berzelius. The same process of titration with silver was applied by Wallace to the analysis of arsenic tribromide, AsBr 3 . This compound was repeatedly distilled to ensure purity, and was well crystallized. His weighings .show that the quantities of bromide given in the third column are pro- portional to 100 parts of silver : 8.3246 grm. AsBr 3 = 8.58 grm. Ag. 97.023 4.4368 " 4-573 " 97.022 5.098 " 5.257 " 96.970 Mean, 97.005, .012 * Poggend. Annalen, 8, i. fCompt. Rend., 20, 1047. I Ann. Chim. Phys. (3), 55, 174. 1859. I Phil. Mag. (4), 18, 270. 214 THE ATOMIC WEIGHTS. Hence As = 73.668, .0436. Why this value should be so much lower than that from the chloride is unexplained. The volumetric work done by Kessler.* for the purpose of establishing the atomic weights of chromium and of arsenic, is described in the chromium chapter. In that investigation the amount of potassium dichromate required to oxidize 100 parts of As. 2 O 3 to As 2 5 was determined and compared with the quantity of potassium chlorate necessary to pro- duce the same effect. From the molecular weight of KC10 3 , that of K 2 Cr 2 O 7 was then calculable. From the same figures, the molecular weights of KC10 3 and of K 2 Cr 2 being both known, that of As 2 3 may be easily determined. The quan- tities of the other compounds proportional to 100 parts of As 2 3 are as- follows : A- 2 2 <9 7 . KCIO* 98.95 4i.i5 6 98.94 41.116 99.17 41.200 98.98 41-255 99.08 41.201 99.15 41.086 41.199 Mean, 99.045, .028 41.224 41.161 4M93 41.149 41.126 Mean, 41.172, db .009 Another series with the dichromate gave the following figures : 99.08 99.06 99.10 98.97 98.97 Mean, 99.036, .019 Previous series, 99.045, =b .028 General mean, 99.039, =h .016 Other defective series are given to illustrate the partial oxidation of the As 2 3 by the action of the air. From Kessler's data we get two values for the molecular weight of As 2 O 3 , thus : From KC1O 3 series As 2 O 3 = 196.951, .0445 From K 2 Cr 2 O 7 series " = 196.726, db .0562 General mean As 2 O 3 = 196.851, =b .0349 And As = 74.607, .0175. * Poggend Annal., 95, 204. 1855. Also 113, 134. 1861. ARSENIC. 215 The determinations made by Hibbs* are based upon an altogether different process from any of the preceding measurements. Sodium pyroarsenate was heated in gaseous hydrochloric acid, yielding sodium chloride. The latter was perfectly white, completely soluble in water, unfused, and absolutely free from arsenic. The vacuum weights are subjoined, with a column giving the percentage of chloride obtained from the pyroarsenate. Na^As^O^. NaCl. Percentage. .02177 - OI 439 66. 100 .04713 .03"5 ' 66.094 .05795 .03830 66.091 .40801 .26981 66.128 .50466 -33345 66.092 .77538 .51249 66.095 .82897 .54791 66.095 1.19124 .78731 66.092 1.67545 1.10732 66.091 3.22637 2.13267 66. 101 Mean, 66.098, .0030 Hence As = 74.340, .0235. In the calculation of the foregoing values for arsenic, the subjoined atomic weights have been assumed : O ---- 15.879, .0003 K = 38.817, .0051 Ag 107.108, db .0031 Na = 22.881, .0046 Cl =. 35.179, zb .0048 S = 31.828, ib. ooi 5 Br = 79.344, -0062 Cr = 51.742, .0034 To the single determination by Berzelius we may arbitrarily assign a weight equal to that of the result from Wallace's bromide series. The general combination is then as follows : From Berzelius' experiment As = 74.460, .0436 " = 74.45> . OI 9 " = 73.668, .0436 From As 2 O 3 (Kessler) " = 74.607, .0175 From Na 4 As 2 O 7 " = 74.340, db .0235 General mean As 74.440, .0106 If O = 16, As = 75.007. * Doctoral thesis, University of Pennsylvania, 1896. Work done under the direction of Professor E. F. Smith. In the fifth experiment the weight of NaCl is printed .33045. This is evidently a misprint, which I have corrected by comparison with the other data. The rejection of this ex- periment would not affect the final result appreciably. 216 THE ATOMIC WEIGHTS. ANTIMONY. After some earlier, unsatisfactory determinations, Berzelius,* in 1826, published his final estimation of the atomic weight of antimony. He oxidized the metal by means of nitric acid, and found that 100 parts of antimony gave 124.8 of Sb 2 O 4 . Hence, if O 16, Sb = 129.03. The value 129 remained in general acceptance until 1855, when Kessler, f by special volumetric methods, showed that it was certainly much too high. Kessler's results will be considered more fully further along, in connec- tion with a later paper; for present purposes a brief statement of his earlierj conclusions will suffice. Antimony and various compounds of antimony were oxidized partly by potassium dichromate and partly by potassium chlorate, and from the amounts of oxidizing agent required the atomic weight in question was deduced : By oxidation of Sb 2 O 3 from 100 parts of Sb Sb = 123.84 By oxidation of Sb with K 2 Cr 2 O 7 " 123.61 By oxidation of Sb with KC1O 3 + K 2 Cr 2 O 7 " = 123.72 By oxidation of Sb 2 O 3 with KC1O 3 + K 2 Cr 2 O 7 . . . " = 123.80 By oxidation of Sb 2 S s with K 2 Cr 2 O 7 " = 123.58 By oxidation of tartar emetic " = 1 19.80 The figures given are those calculated by Kessler himself. A recalcu- lation with our newer atomic weights for O, K, Cl, Cr, S, and C would yield lower values. It will be seen that five of the estimates agree closely, while one diverges widely from the others. It will be shown hereafter that the concordant values are all vitiated by constant errors, and that the exceptional figure is after all the best. Shortly after the appearance of Kessler's first paper, Schneider J pub- lished some results obtained by the reduction of antimony sulphide in hydrogen. The material chosen was a very pure stibnite from Arnsberg, of which the gangue was only quartz. This was corrected for, and cor- rections were also applied for traces of undecom posed sulphide carried off mechanically by the gas stream, and for traces of sulphur retained by the reduced antimony. The latter sulphur was estimated as barium sulphate. From 3.2 to 10.6 grammes of material were taken in each ex- periment. The final corrected percentages of S in Sb 2 S 3 were as follows : 28.559 28.557 28.501 28.554 28.532 *Poggend. Aimalen, 8, i. tPoggend. Annalen, 95, 215. I Poggend. Annalen, 98, 293. 1856. Preliminary note in Bd. 97. ANTIMONY. 217 28.485 28.492 28.481 Mean, 28.520, db .008 Hence, if S = 32, Sb = 120.3. Immediately after the appearance of Schneider's memoir, Rose* pub- lished the result of a single analysis of antimony trichloride, previously made under his supervision b} 7 Weber. This analysis, if Cl = 35.5, makes Sb = 120.7, a value of no great weight, but in a measure confirmatory of that obtained by Schneider. The next research upon the atomic weight of antimony was that of Dexter,f published in 1857. This chemist, having tried to determine the amount of gold precipitable by a known weight of antimony, and having obtained discordant results, finally resorted to the original method of Berzelius. Antimony, purified with extreme care, was oxidized by nitric acid, and the gain in weight was determined. From 1.5 to 3.3 grammes of metal were used in each experiment. The reduction of the weights to a vacuum standard was neglected as being superfluous. From the data obtained, we get the following percentages of Sb in Sb. 2 4 : 79.268 73.272 79-255 79.266 79-253 79.271 79.264 79.260 79.286 79-274 79.232 79-395 79-379 Mean, 79.283, .009 Hence, if = 16, Sb = 122.46. The determinations of Dumas J were published in 1859. This chemist sought to fix the ratio between silver and antimonious chloride, and ob- tained results for the atomic weight of antimony quite near to those of Dexter. The SbCl 3 was prepared by the action of dry chlorine upon pure antimony; it was distilled several times over antimony powder, and it seemed to be perfectly pure. Known weights of this preparation were added to solutions of tartaric acid in water, and the silver chloride was precipitated without previous removal of the antimony. Here, as * Poggend. Annalen, 98, 455. 1856. t Poggend. Annalen, 100, 363. 1857. I Ann. Chim. Phys. (3), 55, 175. 218 THE ATOMIC WEIGHTS. Cooke has since shown, is a possible source of error, for under such circumstances the crystalline argento-antimoiiious tartrate may also be thrown down and contaminate the chloride of silver. But be that as it may, Dumas' weighings, reduced to a common standard, give as propor- tional to 100 parts of silver, the quantities of SbCl 3 which are stated in the third of the subjoined columns : i.876grm. SbCl 3 = 2.66o grm. Ag. 70.526 4.336 " 6.148 " 70.527 5.065 " 7.175 " 70.592 3-475 4-93 " 70.487 3.767 5.350 70.411 5.910 " 8.393 " 70.416 4.828 " 6.836 " 70.626 Mean, 70.512, .021 Hence, if Ag = 108, and Cl = 35,5, Sb = 122. In 1861 Kessler's second paper * relative to the atomic weight of an- timony appeared. Kessler's methods were somewhat complicated, and for full details the original memoirs must be consulted. A standard solution of potassium dichromate was prepared, containing 6.1466 grammes to the litre. With this, solutions containing known quantities of antimony or of antimony compounds were titrated, the end reaction being adjusted with a standard solution of ferrous chloride. In some cases the titration was preceded by the addition of a definite weight of potassium chlorate, insufficient for complete oxidation ; the dichromate then served to finish the reaction. The object in view was to determine the amount of oxidizing agent, and therefore of oxygen, necessary for the conversion of known quantities of antimonious into antimonic com- pounds. In the later paper Kessler refers to his earlier work, and shows that the values then found for antimony were all too high, except in the case of the series made with tartar emetic. That series he merely states, and subsequently ignores, evidently believing it to be unworthy of further consideration. For the remaining series he points out the sources of error. These need not be rediscussed here, as the discussion would have no value for present purposes ; suffice it to say that in the series repre- senting the oxidation of Sb 2 s with dichromate and chlorate, the ma- terial used was found to be impure. Upon estimating the impurity and correcting for it, the earlier value of Sb = 123.80 becomes Sb = 122.36, according to Kessler's calculations. In the paper now under consideration four series of results are given. The first represents experiments made upon a pure antimony trioxide which had been sublimed, and which consisted of shining colorless needles. This was dissolved, together with some potassium chlorate, in *Poggend. Annalen, 113, 145. 1861. ANTIMONY. 219 hydrochloric acid, and titrated with dichromate solution. Six experi- ments were made, but Kessler rejects the first and second as untrust- worthy. The data for the others are as follows : S 2 <9 3 . KCIO*. K.jCr.jO^ sol. in cc. 1,7888 grm. .4527 grm. 19.200. 1.6523 " .45 6 " 3-9 " 3.2998 " .8806 " 16.5 " 1.3438 " .3492 " 10.2 " From these figures Kessler deduces Sb = 122.16. These data, reduced to a common standard, give the following quanti- ties of oxygen needed to oxidize 100 parts of Sb 2 3 to Sb 2 O 5 . Each cubic centimetre of the K 2 Cr 2 7 solution corresponds to one milligramme of : 10.985 10.939 10.951 10.936 Mean, 10.953, - OO 75 In the second series of experiments pure antimony was dissolved in hydrochloric acid with the aid of an unweighed quantity of potassium chlorate. The solution, containing both antimonious and antimonic compounds, was then reduced entirely to the antimonious condition by means of stannous chloride. The excess of the latter was corrected with a strong hydrochloric acid solution of mercuric chloride, then, after diluting and filtering, a weighed quantity of potassium chlorate was added, and the titration with dichromate was performed as usual. Cal- culated as above, the percentages of oxygen given in the last column correspond to 100 parts of antimony: Sb. KClO y A" 2 O a <9 7 sol. cc. Per cent. O. 1.636 grm. 0.5000 grm. 18.3 13.088 3.0825 " 0.9500 " 30.2 i3-5 4.5652 " 1.4106 " 45.5 13.098 Mean, 13.079, .0096 This series gave Kessler Sb = 122.34. The third and fourth series of experiments were made with pure antimony trichloride, SbCl 3 , prepared by the action of mercuric chloride upon metallic antimony. This preparation, in the third series, was dis- solved in hydrochloric acid, and titrated. In one experiment solid K 2 O 2 7 in weighed amount was added before titration; in the other two estimations KC10 3 was taken as usual. The third column gives the percentages of oxygen corresponding to 100 parts of SbCl 3 . 220 THE ATOMIC WEIGHTS. Per cent. O. 1.8576 grm. SbCl 3 needed .5967 grm. K 2 Cr 2 O 7 and 33.4 cc. sol. 7.0338 1.9118 " .3019 " KC1O 3 " 16.2 " 7.0321 4.1235 " .6801 " KC1O 3 " 23.2 " 7.0222 Mean, 7.0294, .0024 The fourth set of experiments was gravimetric. The solution of Sb01 3 mixed with tartaric acid, was first precipitated by hydrogen sulphide, in order to remove the antimony. The excess of H 2 S was corrected by copper sulphate, and then the chlorine was estimated as silver chloride in the ordinary manner. 100 parts of AgCl correspond to the amounts of SbCl 3 given in the third column. 1.8662 grm. SbCl 3 gave 3.483 grm. AgCl. 53-58o 1.6832 3.141 " 53.588 27437 5-IH5 " 53-677 2.6798 5.0025 " 53.569 5.047 9.411 53.629 3.8975 " 7.2585 " 53.696 Mean, 53.623, =b .015 The volumetric series with'SbC! 3 gave Kessler values for Sb ranging from 121.16 to 121.47. The gravimetric series, on the other hand, yielded results from Sb = 124.12 to 124.67. This discrepancy Kessler rightly attributes to the presence of oxygen in the chloride; and, ingeniously correcting for this error, he deduces from both sets combined the value of Sb = 122.37. The several mean results for antimony agree so fairly w r ith each other, and with the estimates obtained by Dexter and Dumas, that we cannot wonder that Kessler felt satisfied of their general correctness, and of the inaccuracy of the figures published by Schneider. Still, the old series of data obtained by the titration of tartar emetic with dichromate con- tained no evident errors, and was not accounted for. This series,* if we reduce all of Kessler's figures to a single common standard, gives a ratio between K 2 Cr 2 7 and C 4 H 4 KSb0 7 .H 2 0. 100 parts of the former will oxidize of the latter : 336.64 338.01 336.83 337-93 338.59 335 ; 79 Mean, 337.30, .29 From this, if K,Cr 2 7 = 292.271, Sb = 118.024. The newer atomic weights found in other chapters of this work will *Poggend. Annalen, 95, 217. ANTIMONY. 221 be applied to the discussion of all these series further along. It may, however, be properly noted at this point that the probable errors assigned to the percentages of oxygen in three of Kessler's series are too low. These percentages are calculated from the quantities of KC10 3 involved in the several reactions, and their probable errors should be increased with reference to the probable error of the molecular weight of that salt. The necessary calculations would be more laborious than the importance of the figures would warrant, and accordingly, in computing the final general mean for antimony, Kessler's figures will receive somewhat higher weight than they are legitimately entited to. Naturally, the concordant results of Dexter, Kessler, and Dumas led to the general acceptance of the value of 122 for antimony as against the lower figure, 120, of Schneider. Still, in 1871, linger * published the re- sults of a single analysis of Schlippe's salt, Na 3 SbS 4 .9H 2 0. This analysis gave Sb = 119.76. if S = 32 and Na = 23, but no great weight could be attached to the determination. It served, nevertheless, to show that the controversy over the atomic weight of antimony was not finally settled. More than ten years after the appearance of Kessler's second paper the subject of the atomic weight of antimony was again taken up, this time by Professor Cooke. His results appeared in the autumn of 1877 1 and were conclusive in favor of the lower value, approximately 120. For full details the original memoir must be consulted ; only a few of the leading points can be cited here. Schneider analyzed a sulphide of antimony which was already formed. Cooke, reversing the method, effected the synthesis of this compound. Known weights of pure antimony were dissolved in hydrochloric acid containing a little nitric acid. In this solution weighed balls of antimony were boiled until the liquid became colorless ; subsequently the weight of metal lost by the balls was ascertained. To the solution, which now contained only antimonious compounds, tartaric acid was added,* and then, with a supersaturated aqueous sulphhydric acid, antimony trisul- phide was precipitated. The precipitate was collected by an ingenious process of reverse filtration, converted into the black modification by drying at 210, and weighed. After weighing, the Sb. 2 S 3 was dissolved in hydrochloric acid, leaving a carbonaceous residue unacted upon. This was carefully estimated and corrected for. About two grammes of antimony were taken in each experiment and thirteen syntheses were performed. In two of these, however, the antimony trisulphide was weighed only in the red modification, and the results were uncorrected by conversion into the black variety and estimation of the carbonaceous residue. In fact, every such conversion and correction was preceded by a weighing of the red modification of the Sb,S 3 . The mean result of these weighings, if S 32, gave Sb = 119.994. The mean result of the cor- * Archiv. der Pharmacie, 197, 194. Quoted by Cooke. f Proc. Amer. Acad., 5, 13. 222 THE ATOMIC WEIGHTS. reeled syntheses gave Sb = 120.295. In these eleven experiments the following percentages of S in Sb a S 3 were established : 28.57 28.60 28.57 28.43 28.42 28.53 28.50 28.49 28.58 28.50 28.51 Mean, 28.5182, =b .0120 These results, confirmatory of the work of Schneider, were presented to the American Academy in 1876. Still, before publication, Cooke thought it best to repeat the work of Dumas, in order to detect the cause of the old discrepancy between the values Sb = 120 and Sb = 122. Ac- cordingly, various samples of antimony trichloride were taken, and puri- fied by repeated distillations. The final distillate was further subjected to several recrystallizations from the fused state ; or, in one case, from a saturated solution in a bisulphide of carbon. The portions analyzed were dissolved in concentrated aqueous tartaric acid, and precipitated by silver nitrate, many precautions being observed. The silver chloride was collected by reverse filtration, and dried at temperatures from 110 to 120. In one experiment the antimony was first removed by H 2 S. Seventeen experiments were made, giving, if Ag = 108 and Cl = 35.5. a mean value of Sb = 121.94. If we reduce to a common standard, Cooke's analyses give, as proportional to 100 parts of AgCl, the quantities of SbCl s stated in the third column : i.5974grm. SbCl 3 gave 3.0124 grm. AgCl. 53.028 1.2533 " 2.3620 " 53.061 .8876 1.6754 52.978 .8336 i 5 6 74 53-^4 .5326 " i. 0021 " 53-H8 .7270 " i.3 6 9 T " 53- T i 1.2679 " 2.3883 " 53.088 1.9422 3.6646 52.999 1-7702 " 3-3384 " 53.025 2.5030 4-7184 53.048 2.1450 " 4.0410 " 53.081 1.7697 " 3.3281 " 53.175 2-3435 4.4157 53.072 1.3686 " 2.5813 " 53-O2O 1.8638 " 3-5'46 " 53.03 2.0300 " 3.8282 " 53.028 2.4450 " 4.6086 53.053 Mean, 53 066, zfc .0096 ANTIMONY. 223 This mean may be combined with that of Kessler's series, as follows : Kessler ............. ' ................ . ... 53.623, d= .015 Cooke ............ ---- ............ ____ 53.o66, .0096 General mean ................... 53.2311, .008 The results thus obtained with SbCl 3 confirmed Dumas' determination of the atomic weight of antimony as remarkably as the syntheses of Sb 2 S 3 had sustained the work of Schneider. Evidently, in one or the other series a constant error must be hidden, and much time was spent by Cooke in searching for it. It was eventually found that the chloride of antimony invariably contained traces of oxychloride, an impurity which tended to increase the apparent atomic weight of the metal under con- sideration. It was also found, in the course of the investigation, that hydrochloric acid solutions of antimonious compounds oxidize in the air during boiling as rapidly as ferrous compounds, a fact which explains the high values for antimony found by Kessler. In order to render "assurance doubly sure." Professor Cooke also undertook the analysis of the bromide and the iodide of antimony. The bromide, SbBr s , was prepared by adding the finely powdered metal to a solution of bromine in carbon disulphide. It was purified by repeated distillation over pulverized antimony, and by several recrystallizations from bisulphide of carbon. The bromine determinations resemble those of chlorine, and gave, if Ag = 108 and Br = 80, a mean value for anti- mony of Sb = 120. Reduced to a common standard, the fifteen analyses give the subjoined quantities of SbBr 3 proportional to 100 parts of silver bromide : 1.8621 grm. SbBr 3 gave 2.9216 grm. AgBr. .9856 1.8650 1.5330 1.3689 1.2124 .9417 2.5404 1.5269 1.8604 1.7298 3-2838 2.3589 L3323 2.6974 1.5422 2.9268 2.4030 2.1445 1.8991 1.4749 3-9755 2.3905 2.9180 2.7083 5.1398 3.6959 2.0863 4.2285 63-736 63.909 63.721 63.795 63-833 63841 63.848 63.901 63-874 63-756 63.870 63.890 63.825 63-859 63-791 Mean, 63.830, .008 The iodide of antimony was prepared like the bromide, and analyzed in the same way. At first, discordant results were obtained, due to the presence of oxyiodide in the iodide studied. The impurity, however, 224 THE ATOMIC WEIGHTS. was removed by subliming the iodide in an atmosphere of dry carboi dioxide. With this purer material, seven estimations of iodine wei made, giving, if Ag = 108 and I = 127, a value for antimony of Sb = 120. Reduced to a uniform standard, Cooke's weighings give the following quantities of SbI 3 proportional to 100 parts of silver iodide : 1.1877 grm. SbI 3 gave 1.6727 grm. Agl. 71.005 .4610 3.2527 1. 8068 1.5970 2.3201 3496 .6497 2.5389 2.2456 .4927 70.956 71.150 71.165 71.117 71.071 70.956 Mean, 71.060, .023 Although Cooke's work was practically conclusive, as between the rival values for antimony, his results were severely criticised by Kessler,* who evidently had read Cooke's paper in a very careless way. On the other- hand, Schneider published in Poggendorff 's Annalen a friendly review of the new determinations, which so well vindicated his own accuracy. In reply to Kessler, Cooke undertook still another series of experiments with antimony bromide,f and obtained absolute confirmation of his previous results. To a solution of antimony bromide was added a solu- tion containing a known weight of silver not quite sufficient to precipi- tate all the bromine. The excess of the latter was estimated by titration with a normal silver solution. Five analyses gave values for antimony ranging from 119.98 to 120.02, when Ag = 108 and Br = 80. Reduced to a common standard, the weights obtained gave the amounts of SbBr stated in the third column as proportional to 100 parts of silver : 2.5032 grm. SbBr 3 = 2.2528 grm. Ag. 2.0567 " 1.8509 2.6512 " 2.3860 " 3-353 " 2.9749 2.7495 2-4745 111.115 111.119 111.115 111.106 111.113 Mean, 1 11.114, .0014 Schneider^ also, in order to more fully answer Kessler's objections, repeated his work upon the Arnsberg stibmte. This he reduced in hydro- gen as before, correcting scrupulously for impurities. The following percentages of sulphur were found : 28.546 28.534 28.542 Mean, 28 541, db .0024 *Berichte d. Deutsch. Chem. Gesell., 12, 1044. 1879. f Amer. Journ. Sci. and Arts, May, 1880. Berichte, 13, 951. JJourn. fur Prakt. Chem. (2), 22, 131. ANTIMONY 225 These figures confirm his old results, and may be fairly combined with them and with the percentages found by Cooke, as follows : Schneider, early series 28.520, .008 Schneider, late series 28.541, .0024 Cooke 28.5182, .0120 General mean 28.5385, =b .0023 In 1881 Pfeifer * determined electrolytically the direct ratios between silver and antimony, and copper and antimony. With copper the fol- lowing data were obtained : G/ 1.412 grm, 1.902 3.367 Sb = 1.1008 Cu. 1.4832 " 2.6249 " Sb} : : IOO 128.270 128.236 128.272 If Cu = 63.6, Sb = 122.36. With silver he found 5.925 grm. Sb= 15.774 Ag. 6.429 10.116 4 865 4.390 9.587 4.525 17.109 26.972 13.014 11.697 25.611 12.097 Mean, 128.259, .0077 Ag^ : Sb : \ 100 : ,r. 37.562 37-577 37.506 37.383 37-531 37.433 37.406 Mean, 37.485, d= .0198 If Ag = 108, Sb == 121.45. The latter ratio was also determined by Popper, f several years after- wards. The two metals were precipitated simultaneously by the same current ; and in some experiments two portions of antimony were thrown down against one of silver. These are indicated in the subjoined table by suitable bracketing, and the ratio is given in the third column : Sb. Ag. Ratio. 1.4856) 1.4788 / 3-9655 37.463 37.292 2.OI2O | 2.OO74 ) 3.88821 3.8903 5-3649 10.3740 37.503 37.417 37.48o 37.50 4.1885 11.1847 37-455 37-447 * Ann. Chem. Pharm., 209, 161. t Ann. Chem., 233, 153. 15 226 THE ATOMIC WEIGHTS. gfig 37.507 4.2752 j 37.545 5.6860 1 37.460 5.6901 / 37.487 4.4117 11.8014 37.383 4.9999 13.3965 37.322 5.2409 14.0679 37.250 Mean, 37.434, .0149 Pfeifer found, 37.485, .0198 General mean, 37.452, .0119 If Ag = 108, Popper's figures give in mean Sb = 121.3. I am inclined to attach slight importance to these electrolytic data, for the reasons that it would be very difficult to ensure the absolute purity and freedom from occlusions of the antimony as weighed, or to guarantee that no secondary reactions had modified the ratios. The work done by Bongartz * in 1883 was quite different from any of the determinations which had preceded it. Carefully purified 'antimony was weighed as such, and then dissolved in a concentrated solution of potassium sulphide. From this, after strong dilution, antimony trisul- phide was thrown down by means of dilute sulphuric acid. After thorough washing, this sulphide was oxidized by hydrogen peroxide, by Classen's method, and the sulphur in it was weighed as barium sulphate. The ratio measured, therefore, was 2Sb : 3BaS0 4 , and the data were as follows. The BaS0 4 equivalent to 100 parts of Sb is the ratio stated : Sb Taken. BaSO Found. Ratio. 1.4921 4.3325 290.362 .6132 1.7807 290.394 .5388 1.5655 290.553 T.2II8 3.5205 290.518 .9570 2.7800 290.491 .6487 1.8855 290.349 .7280 2. 1 100 289.835 9535 2.7655 290.036 I.O275 2.9800 290.024 .9635 2.7980 290.399 .9255 2.6865 290.275 .7635 2.2175 290.438 Mean, 290.306, .0436 We have now before us the following ratios, good and bad, from which to calculate the atomic weight of antimony. The single results obtained by Weber and by Unger, being unimportant, are not included : * Ber. Deutsch. Chem. Gesell., 16, 1942. 1883. ANTIMONY. 227 (i.) Percentage of S in Sb 2 S 3 , 28.5385, .0023 (2.) Percentage of Sb in Sb 2 O 4 , 79.283, .009 (3.) O needed to oxidize 100 parts SbCJ 3 , 7.0294, .0024 (4.) O needed to oxidize 100 parts Sb 2 O 3 , 10.953, -O75 (5.) O needed to oxidize 100 parts Sb, 13.079, Hh .0096 (6.) K 2 Cr 2 O 7 : tartar emetic : : 100 : 337.30, .29 (7-) A Ss ' SbCl 3 : : 100 : 70.512, .021 (8.) 3AgCl : SbG 3 : : 100 : 53.2311, .008 (9-) A 3 ' SbBr 3 : : loo : 111.114, .0014 (10.) 3AgBr : SbBr 3 : : loo : 63.830, .008 (11.) 3AgI : SbI 3 : : 100 : 71.060, .023 (12.) Cu 3 : Sb 2 : : 100 : 128.259, .0077 ( T 3-) A 3 = Sb : : 100 : 37.452, .0119 (14.) Sb 2 : 3BaSO 4 : : 100 : 290.306, rb .0436 In the reduction of these ratios a considerable number of antecedent atomic weights are required, thus : = 15.879, + .0003 C = 11.920, .0004 Ag = 107.108, .0031 Cu = 63.119, .0015 cl =:: 35-179, .0048 Ba = 136.392, .0086 Br == 79-344, .0062 Cr = 51.742, .0034 1 = 125.888,^.0069 AgCl = 142.287, .0037 K = = 38.817, .0051 AgBr=r 186.452, .0054 S ;= 31.828,^.0015 Agl =232.996,^3.0062 Three of the ratios give the molecular weight of antimony trichloride, and two give corresponding values for the bromide. These values may be combined, as follows : First, for the chloride From (3) SbCl 3 = 225.894, .0771 From (7) , " = 226.572, .0678 From (8) " = 227.223, dr .0347 General mean SbCl 3 = 226.924, .0286 Hence Sb = 121.387, dr .0321. For the bromide we have From (9) r. SbBr 3 357.036, .0113 From ( 10) " = 357.037, .0250 General mean SbBr 3 = 357.036, .0103 Hence Sb = 119.005, .0212. All the data yield eleven values for antimony, which are arranged below in the order of their magnitude : 228 THE ATOMIC WEIGHTS. 1. From tartar emetic, ratio (6) Sb = 118.024, .2827 2. From SbBr 3 " = 119.005, d= .0212 3. From SbI 3 , ratio (i i ) "= 119.037, .1626 4. From Sb 2 S 3 , ratio (i) " = 119.548, .0069 5. From ratio (14) " = 119.737, .0188 6. From ratio (13) " = 120.342, .0384 7. From ratio (4) " = 121.155, .1000 8. From SbCl 3 " =. 121.387, .0321 9. From ratio (5) " 121.408, .0891 10. From ratio (12) " = 121.434, .0078 11. From Sb 2 O 4 , ratio (2) u = 121.542, .0546 General mean Sb = 120.299, zb .0047 If = 16, this becomes Sb = 121.218. Among these figures the discordance is so great that the mathematical combination has no real value. We must base our judgment in this case mainly upon chemical evidence, and this, as shown in the investigations of Cooke and of Schneider, favors a lower rather than a higher value for the atomic weight of antimony. Dumas' work was affected by constant errors which are now known, and Dexter's data are also presumably in the wrong. A general mean of values 2, 3, 4, and 5 gives Sb = 119.521, .0062, or, if = 16, Sb = 120.432. Even now the range of uncertainty is greater than it should be, but none of the four values combined can be accepted exclusively or rejected without more evidence. This result, therefore, should be adopted until new determinations, of a more con- clusive nature, have been made. BISMUTH. 229 BISMUTH. Early in the century the combining weight of bismuth was approxi- mately fixed through the experiments of Lagerhjelm.* Effecting the direct union of bismuth and sulphur, he found that ten parts of the metal yield the following quantities of trisulphide : 12.2520 12.2065 12.2230 12.2465 Mean, 12.2320 Hence Bi = 215 in round numbers, a value now known to be much too high. Lagerhjelm also oxidized bismuth with nitric acid, and, after igni- tion, weighed the trioxide thus formed. Ten parts of metal gave the following quantities of Bi 2 3 : 11.1382 11.1275 Mean, 11.13285 Hence, if = 16, Bi = 211.85, a figure still too high. In 1851 the subject of the atomic weight of bismuth was taken up by Schneider,f who, like Lagerhjelm, studied the oxidation of the metal with nitric acid. The work was executed with a variety of experimental refinements, by means of which every error due to possible loss of mate- rial was carefully avoided. For full details the original paper must be consulted ; there is only room in these pages for the actual results, as follows. The figures represent the percentages of Bi in Bi 2 O 3 : 89.652 89.682 89.644 89.634 89.656 89.666 89-655 89-653 Mean, 89.6552, .0034 Hence, if = 16, Bi = 208.05. Next in order are the results obtained by Dumas. J Bismuth tri- * Annals of Philosophy, 4, 358. 1814. Adopted by Berzelius. t Poggend. Annalen, 82, 303. 1851. I Ann. Chitn. Phys. (3), 55, 176. 1859. 230 THE ATOMIC WEIGHTS. chloride was prepared by the action of dry chlorine upon bismuth, and repeatedly rectified by distillation over bismuth powder. The product was weighed in a closed tube, dissolved in water, and precipitated with sodium carbonate. In the filtrate, after strongly acidulating with nitric acid, the chlorine was precipitated by a known amount of silver. The figures in the third column show the quantities of BiCl 3 proportional to 100 parts of silver : 98.90x3 9 8 -373 98.005 97.829 97.99 6 97.806 97.643 97.712 97.762 3.506 grm. BiC H 3 = 3.545 g rm . Ag. 1.149 " 1.168 i < 1.5965 1.629 " 2.1767 2.225 ( ( 3.081 3-H4 " 2.4158 2.470 it 1.7107 I-75 2 n 3.523 3-6055 i < 5.241 5.36i " - Mean, 98.003, .090 Hence, with Ag = 108 and Cl = 35.5, Bi = 211.03. The first three of the foregoing experiments were made with slightly discolored material. The remaining six percentages give a mean of 97.791, whence, on the same basis as before, Bi = 110.79. Evidently these results are now of slight value, for it is probable that the chloride of bismuth, like the corresponding antimony compound, contained traces of oxy chloride. This assumption fully accounts for the discordance be- tween Dumas' determination and the determinations of Schneider and of still more recent investigators. In 1883 Marignac * took up the subject, attacking the problem by two methods. His point of departure was commercial subnitrate of bismuth, which was purified by re-solution and reprecipitation, and from which he prepared the oxide. First, bismuth trioxide was reduced by heating in hydrogen, beginning with a moderate temperature and closing the operation at redness. The results were as follows, with the percentage of Bi in Bi 2 3 added : 2.6460 grm. Bi. 2 O 3 lo^t 0.2730 grm. O. 89.683 per cent. 6.7057 " .6910 " 89.696 " 3.6649 " .3782 " 89.681 " 5.8024 " .5981 " 89.692 " 5.1205 " .5295 " 89.658 " 5.5640 .5742 " 89.680 " Mean, 89.682, i: .0036 Hence, if = 16, Bi = 208.60. *Arch. Sci. Phys. et Nat. (3), 10, 10. BISMUTH. 231 Marignac's second method of determination was by conversion of the oxide into the sulphate. The oxide was dissolved in nitric acid, and then sulphuric acid was added in slight excess from a graduated tube. The mass was evaporated to dryness with great care, and finally heated over a direct flame until fumes of S0 3 no longer appeared. The third column gives the sulphate formed from 100 parts of oxide : 2.6503 Bi 2 O 3 gave 4.0218 Bi 2 (SO 4 ) 3 . Ratio, 151.749 2.8025 4.2535 " " 151.775 2.710 4.112 " " I5L734 2.813 " 4- 26 7 " " 151.688 2.8750 4.3 62 5 " ". I5I-739 2.7942 " 4-2383 " " 151.682 Mean, 151.728, .0099 Hence, with O = 16 and S = 32.06, Bi = 208.16. This result needs to be studied in the light of Bailey's observation,* that bismuth sulphate has a very narrow range of stability. It loses the last traces of free sulphuric acid at 405, and begins to decompose at 418, so that the foregoing ratio is evidently uncertain. The concordance of the data, however, is favorable to it. The next determination of this atomic weight was by L6we,f who oxidized the metal with nitric acid, and reduced the nitrate to oxide by ignition. Special care was taken to prepare bismuth free from arsenic, and the oxide was fused before weighing. In the paper just quoted Bailey calls attention to the volatility of bismuth oxide, which doubt- less accounts for the low results found in this investigation. The data are as follows : Bi Taken. Bi^O^ Found. Per cent. Bi. 11.309 12.616 89.640 12.2776 !3'694 89.656 Mean, 89.648, .0040 Hence, if = 16, Bi = 207.84. In Classen's J work upon the atomic weight of bismuth, the metal itself was first carefully investigated. Commercial samples, even those which purported to be pure, were found to be contaminated with lead and other impurities, and these were not entirely removable by many successive precipitations as subnitrate. Finally, pure bismuth was ob- tained by an electrolytic process, and this was converted into oxide by means of nitric acid and subsequent ignition to incipient fusion. Results as follows, with the percentage of Bi in Bi 2 O 3 added : * Journ. Chem. Soc., 51, 676. tZeit. Anal. Chem., 22, 498. \ Ber. Deutsch. Chem. Gesell., 23, 938. 1890. 232 THE ATOMIC WEIGHTS. Bi Taken. Bi^O z Found. Per cent. Bi. 25.0667 27.9442 89.703 21.0691 23.4875 89.7035 27.2596 30.3922 89.693 36.5195 40.713^ 89.700 27.9214 3H295 89.6944 32.1188 35-8103 89.692 30.1000 33.5587 89.694 26.4825 59.5257 89.693 19.8008 22.0758 89.695 Mean, 89.696, .0009 Hence, if == 16, Bi = 208.92, or, reduced to vacuum standards, 208.90. Classen's paper was followed by a long controversy between Schneider and Classen,* in which the former upheld the essential accuracy of the work done by Marignac and himself. Schneider had started out with commercial bismuth, and Classen found that the commercial bismuth which he met with was impure. Schneider, by various analyses, showed that other samples of bismuth were so nearly pure that the common modes of purification were adequate ; but Classen replied that the original sample used by Schneider in his atomic weight investigation had not been reexamined. Accordingly, Schneider published a new series of determinations f made by the old method, but with metal which had been scrupulously purified. Results as follows : Bi. Bi^. Percent. Bi. 5.0092 5.5868 89.661 3.6770 4.1016 89.648 7.2493 8.0854 89.659 9.2479 10.3142 89.662 6.0945 6.7979 89.653 12.1588 13.5610 89.660 Mean, 89.657, .0015 Hence with O = 16, Bi = 208.05, a confirmation of the earlier deter- minations. Although the results so far are not final, a combination of the data relative to bismuth oxide is not without interest. 1. Lagerhjelm .......................... 89.865, db .0650 2. Schneider, 185 1 ................. ..... 89.655, =b .0034 3. Marignac ............................ 89.682, .0036 4. Lowe ............................... 89.648, .0040 5. Classen ........................... 89. 696, .0009 6. Schneider, 1894 ...................... 89.657, .0015 General mean 89.681, rb .0007 * Journ. fiir Prakt. Chem. (2), 42, 553 ; 43, 133 ; and 44, 23 and 411. t Journ. fiir Prakt. Chem. (2), 50, 461. 1894. BISMUTH. 233 Omitting the first and fifth means, the other data give a general mean percentage of 89.659, .0012. The ratios now before us are as follows : (I.) Percentage of Hi in Bi 2 O 3 , 89.681, .0007 (2.) Bi 2 O 3 : Bi 2 (SO 4 ) 3 : : 100 : 151.728, .0099 13.) 3Ag : BiCl 3 : : 100 : 98.003, .090 For computation we have O = 15.879, =b .0003 Ag = 107. 108, zh .0031 8=31.828,^.0015 Cl = 35.179, .0048 Hence, reducing the ratios From (i) Bi = 207.003, .0150 From (2) .... " = 206.613, -444 From (3) " = 209.370, .2847 General mean Bi = 206.971, =b .0142 If O = 16, Bi = 208.548. Classen's data alone give Bi = 207.389, or, with = 16, 208.969. Omitting this set of determinations and rejecting Dumas', the remaining data give From Bi 2 O 3 Bi 206.512, .0244 From Bi 2 (SO 4 ) 3 " = 206.613, .0444 General mean Bi = 206.536, .0214 If = 16, this becomes Bi = 208.11. Between this figure and Classen's, future investigation must decide. The confirmation afforded by the sulphate series is in favor of the lower value. 234 THE ATOMIC WEIGHTS. COLUMBIUM.* The atomic weight of this metal has been determined by Rose, Her- mann, Blomstrand, and Marignac. Rosef analyzed a compound which he supposed to be chloride, but which, according to Rammelsberg, J must have been nearly pure oxychloride. If it was chloride, then the widely varying results give approximately Cb = 122 ; if it was oxychloride, the value becomes nearly 94. If it was chloride, it was doubtless contami- nated with tantalum compounds. Hermann's results seem to have no present value, and Blomstrand's || are far from concordant. The latter chemist studied columbium penta- chloride and sodium columbate. In the first case he weighed the colum- bium as columbium pentoxide, and the chlorine as silver chloride, the oxide being determined by several distinct processes. In some cases it was thrown down by water, in others by sulphuric acid, and in still others by sodium carbonate or ammonia jointly with sulphuric acid. The weights given are as follows : Cb.,0,. AgCl. 591 -294 ..... .8085 .401 2.085 633 .317 ..... .195 .0974 .500 .507 .2505 1.302 .9415 -472 2.454 .563 .2796 ..... .9385 .4675 2.465 .4788 .2378 .408 .204 1.067 9065 .45 1 5 Hence the subjoined percentages, and the ratios 5AgCl : CbCl 5 : : 100 : x, and 5 AgCl : Cb 2 O 5 : : 100 : x. Percent. C6 2 <9 5 . AgCl : CbCl,. AgCl : Ct>,0,. 40 788 T"-7 / 49.598 38-777 19.233 50.079 49-949 39.000 19-435 49.408 38.940 19.240 50.135 38.366 19-234 *This name has priority over the more generally accepted " niobium," and therefore deserves preference. fPoggend. Annal., 104, 439. 1858. JPoggend. Annal., 136,353. r86g. I Journ. fiir Prakt. Chem., 68, 73. 1856. | Acta Univ. Lund. 1864. COLUMBIUM. 235 49.662 ...... ...... 49.813 38-073 18.966 49.666 ...... ...... 50.000 3 8 -238 19.119 49.807 Mean, 49.806, zh .045 Mean, 38.566, .108 Mean, 19.205, .043 From these means the atomic weight of columbium may be computed, thus: From 2CbCl 5 : Cb 2 O 5 ........................ Cb 95.397 From CbCl 5 : 5AgCl ........................ ";== 98.477 From 5 AgCl : Cb 2 O 5 ........................ = 96.933, when == 15,879, Ag = 107.108, and Cl = 35.179. The series upon sodium columbate, which salt was decomposed with sulphuric acid, both Cb 2 5 and Na 2 S0 4 being weighed, is too discordant for discussion. The exact nature of the salt studied is not clear, and the data given, when transformed into the ratio Na 2 SO 4 : Cb 2 6 : : 100 : a;, give values for x ranging from 151.65 to 161.20. Further consideration of this series would therefore be useless. It seems highly probable that Blom- strand's materials were not entirely free from tantalum, however, since the atomic weight of columbium derived from his analyses of the chloride are evidently too high. Marignac* made about twenty analyses of the potassium nuoxy colum- bate, CbOF 3 .2KF.H 2 O. 100 parts of this salt give the following percent- ages : Cb 2 O 5 ............ Extremes 44.15 to 44.60 Mean, 44.36 K 2 SO,... ......... 57.60-58.05 H 2 ............. " 5.75 " 5.98 F ................ " 30.62 " 32.22 From the mean percentage of Cb 2 O 5 , Cb = 92.852. If = 16, this becomes 93.56. From the mean between the extremes given for K 2 S0 4 , Cb = 93.192. If = 16, this becomes 93.90. As Beville ami Troost'sf results for the vapor density of the chloride and oxychloride agree fairly well with Cb = 94, we may adopt this value as approximately correct. The mean of the two values computed from Marignac's data is 93.022 when H = 1, and 93.73 when == 16. * Arch. Sci. Phys. Nat. (2), 23. 1865. f Compt. Rend., 56, 891. 1863. 236 THE ATOMIC WEIGHTS. TANTALUM. The results obtained for the atomic weight of this metal by Berzelius,* Rose,f and Hermann J may be fairly left out of account as valueless. These chemists could not have worked with pure preparations, and their data are sufficiently summed up in Becker's " Digest." Blomstrand's determinations, as in the case of columbium, were made upon the pentachloride. His weights are as follows : Ta. 2 Or,. AgCl. .9808 .598 ...... 1.4262 .867 2.906 2.5282 1.5375 5.0105 1.0604 . 6 455 2.156 2.581 i.577 ...... 8767 -534 Hence the subjoined percentages of Ta 2 5 from TaCl 5 , and the ratios SAgCl : TaCl 5 : : 100 : x, and 5AgCl : Ta 2 5 : : 100 : x. Percent. Ta,O 5 . AgCl : TaCl y AgCl : Ta,O- . 60.971 ...... f ...... 60.791 49. 78 29.835 60.814 50.458 30685 60.873 49.297 29.940 60.960 ...... ...... 60.924 ...... Mean, 60.889, .0208 49-6ir, =b .289 30.153, dr .180 From these ratios we get for the atomic weight of tantalum : From per cent. Ta 2 O 5 Ta = 172.342 From 5AgCl : TaCl 5 ; " = 177.055 From 5 AgCl : Ta 2 O 5 " =174.821 These results are too low. Probably Blomstrand's material still con- tained some columbium. In 1866 Marignac's determinations appeared. || He made four analyses of a pure potassium fluotantalate, and four more experiments upon the ammonium salt. The potassium compound, K 2 TaF 7 , was treated with sulphuric acid, and the mixture was then evaporated to dryness. The potassium sulphate was next dissolved out by water, while the residue * Poggend. Annalen, 4, 14. 1825. f Poggend. Annalen, 99, 80. 1856. 1 Journ. fur Prakt. Chem., 70, 193. 1857. g Acta Univ. I^und, 1864 || Arch. Sci. Phys. Nat. (2), 26, 89. 1866. TANTALUM. 237 was ignited and weighed as Ta 2 5 . 100 parts of the salt gave the follow- ing quantities of Ta 2 O 5 and K 2 S0 4 : 56.50 56.75 56.55 56.56 Mean, 56.59, .037 44-37 44-35 44.22 44.24 Mean, 44.295, .026 From these figures, 100 parts of K 2 S0 4 correspond to the subjoined quantities of Ta 2 5 : 127.338 127.960 128.178 127.848 Mean, 127.831, .120 The ammonium salt, (NH 4 ) 2 TaF 7 , ignited with sulphuric acid, gave these percentages of Ta 2 O 5 . The figures are corrected for a trace of K 2 SO 4 which was always present : 63.08 63.24 63.27 63.42 Mean, 63.25, .047 Hence we have four values for Ta : From potassium salt, per cent. Ta 2 O 5 Ta = 182.336 From potassium salt, per cent. K 2 SO 4 " 180.496 From potassium salt, K 2 SO 4 : Ta 2 O 5 " 181.422 From ammonium salt, per cent. Ta 2 O 5 " = 181.559 Average Ta = 181.453 'Or, if = 16, Ta = 182.836. These values are computed with O N = 13.935, and F = 18.912. 15.879, K = 38.817, S = 31.828, 238 THE ATOMIC WEIGHTS. CHROMIUM. Concerning the atomic weight of chromium there has been much dis- cussion, and many experimenters have sought to establish the true value. The earliest work upon it having any importance was that of Berzelius,* in 1818 and 1826, which led to results much in excess of the correct figure. His method consisted in precipitating a known weight of lead nitrate with an alkaline chromate and weighing the lead chro- mate thus produced. The error in his determination arose from the fact that lead chromate, except when thrown down from very dilute solu- tions, carries with it minute quantities of alkaline salts, and so has its apparent weight notably increased. When dilute solutions are used, a trace of the precipitate remains dissolved, and the weight obtained is too low. In neither case is the method trustworthy. In 1844 Berzelius' results were first seriously called in question. The figure for chromium deduced from his experiments was somewhat over 56 ; but Peligot f now showed, by his analyses of chromous acetate and of the chlorides of chromium, that the true number was near 52.5. Unfortunately, Peligot's work, although good, was published with in- sufficient details to be useful here. For chromous acetate he gives the percentages of carbon and hydrogen, but not the actual weights of salt, carbon dioxide, and -water from which they were calculated. His figures vary considerably, moreover enough to show that their mean would carry but little weight when combined with the more explicit data fur- nished by other chemists. Jacquelain's work we may omit entirely. He gives an atomic weight for chromium which is notoriously too low (50.1), and prints none of the numerical details upon which his result rests. The researches which particularly command our attention are those of Berlin, Moberg, Lefort, Wildenstein, Kessler, Siewert, Baubigny, Rawson, and Meineke. Among the papers upon the atomic weight under consideration that by Berlin is one of the most important. His starting point was normal silver chromate; but in one experiment the dichromate Ag. 2 Cr,0 7 was used. These salts, which are easily obtained in a perfectly pure condi- tion, were reduced in a large flask by means of hydrochloric acid and alcohol. The chloride of silver thus formed was washed by decantation, dried, fused, and weighed without transfer. The united washings were supersaturated with ammonia, evaporated to dry ness, and the residue treated with hot water. The resulting chromic oxide was then collected upon a filter, dried, ignited, and weighed. The results were as follows : *Schweigg. Journ., 22, 53, and Poggend. Annal., 8, 22. fCompt. Rend., 19, 609, and 734; 20, 1187 ; 21, 74. I Compt. Rend., 24, 679. 1847. f Journ. fur Prakt. Chem., 37, 509, and 38, 149. 1846. CHROMIUM. 239 4.6680 grm. Ag 2 CrO 4 gave 4.027 grm. AgCl and 1.0754 grm. Cr 2 O 3 . 3.4568 " 2.983 " .7960 2.5060 " 2.1605 " .5770 " 2.1530 " 1.8555 " -4945 4-3335 g rm - Ag 2 Cr 2 O 7 gave 2.8692 i.53 " From these weighings three values are calculable for the atomic weight of chromium. The three ratios upon which these values depend we will consider separately, taking first that between the chromic oxide and the original silver salt. In the four analyses of the normal chromate the percentages of Cr 2 3 deducible from Berlin's weighings are as follows : Mean, 23.014, =fc .on And from the single experiment with Ag 2 Cr 2 7 the percentage of Cr 2 O, was 35.306. For the ratio between Ag 2 Cr0 4 and AgCl, putting the latter at 100, we have for the former : 115-917 115.883 115.992 116.033 Mean, 115.956, rb .023 In the single experiment with dichromate 100 AgCl is formed from 151.035 Ag. 2 Cr 2 O 7 . Finally, for the ratio between AgCl and Cr 2 3 , the five experiments of Berlin give, for 100 parts of the former, the following quantities of the latter : 26.705 26.685 26.707 26.650 26.662 Mean, 26.682, .0076 These results will be discussed, in connection with the work of other investigators, at the end of this chapter. In 1848 the researches of Moberg* appeared. His method simply consisted in the ignition of anhydrous chromic sulphate and of am- monium chrome alum, and the determination of the amount of chromic * Journ. fi'ir Prakt. Cheni., 43, 114. 240 THE ATOMIC WEIGHTS. oxide thus left as residue. In the sulphate, Cr 2 (S0 4 ) 3 , the subjoined per- centages of Cr 2 3 were found. The braces indicate two different sam- ples of material, to which, however, we are justified in ascribing equal value : .542 grm. sulphate gave .212 grm. Cr 2 O 3 . 39.114 per cent. ~\ 1.337 " .523 " 39.117 " .5287 .207 " 39. 153 " 3 1.033 .406 " 39o03 " ) .868 " .341 " 39-286 " Mean, 39.1946, .0280 From the alum, NH 4 .Cr(S0 4 ) 2 .12H 2 0, we have these percentages of O 2 O 3 . The first series represents a salt long dried under a bell jar at a temperature of 18. The crystals taken were clear and transparent, but may possibly have lost traces of water,* which would tend to increase the atomic weight found for chromium. In the second series the salt was carefully dried between folds of filter paper, and results were obtained quite near those of Berlin. Both of these series are discussed together, neither having remarkable value: 1.3185 grm. alum gave .213 grm. Cr 2 O 3 . 1 ^> 1 55 P er cent. .7987 " .129 " 1 6. 151 " 1.0185 " .1645 " 16.151 " 1.0206 .1650 " 16.167 " .8765 .1420 " 16.201 " .7680 " .1242 " 16.172 " 1.6720 " .2707 " 16.190 " .5410 .0875 <( 16.174 1.2010 " .1940 " T 6.i53 " i. ooio " .1620 " 16.184 " .7715 " .1235 " 16.007 1.374 " .2200 " 16.012 " Mean, 16.143, .0125 The determinations made by Lefortf are even less valuable than those by Moberg. This chemist started out from pure barium chromate, which, to thoroughly free it from moisture, had been dried for several hours at 250. The chromate was dissolved in pure nitric acid, the barium thrown down by sulphuric acid, and the precipitate collected upon a filter, dried, ignited, and weighed in the usual manner. The natural objection to the process is that traces of chromium may be carried down with the sul- phate, thus increasing its weight. In fact, Lefort's results are somewhat too high. Calculated from his weighings, 100 parts of BaS0 4 correspond to the amounts of BaCr0 4 given in the third column : * This objection is suggested by Berlin in a note upon I v efort's paper. Journ. fur Prakt. Chem. 71, 191. t Journ. fur Prakt. Chem., 51, 261. 1850. CHROMIUM. 241 1.2615 g rm - BaCrO 4 gave 1.1555 g rm - BaSO 4 . 109.174 1.5895 " L458o " 109.019 2.3255 " 2.1340 108.974 3.0390 2.7855 " 109.101 2.3480 2.1590 " 108.754 1.4230 1.3060 u 108.708 I.I975 1.1005 108.814 3.4580 " 3-1690 " 109.119 2.0130 1.8430 " 109.224 3.5570 " 3-2710 " 108.744 1.6470 " 1.5060 " 109.363 1.8240 1-6725 " 109.058 1.6950 " 1.5560 " 108.933 2.5960 " 2.3870 " 108.756 Mean, 108.9815, .0369 Wildenstein,* in 1853, also made barium chromate the basis of his researches. A known weight of pure barium chloride was precipitated by a neutral alkaline chromate, and the precipitate allowed to settle until the supernatant liquid was perfectly clear. The barium chromate was then collected on a filter, washed with hot water, dried, gently ignited, and weighed. Here again arises the objection that the precipitate may have retained traces of alkaline salts, and again we find deduced an atomic weight which is too high. One hundred parts BaCr0 4 correspond to BaCl 2 as follows : 81.87 81.57 81.80 81.75 81.61 81.66 81.78 81.83 81.52 81.66 81.84 81.80 81.85 81.66 81.70 81.85 81.68 81.57 81.54 81.83 81.66 81.71 81.55 81.63 81.81 81.56 81.86 81.58 81.54 81.67 81.68 81 84 Mean, 81.702, .014 Next in order we have to consider two papers by Kessler, who em- ployed a peculiar volumetric method entirely his own. In brief, he com- pared the oxidizing power of potassium dichromate with that of the chlorate, and from his observations deduced the ratio between the mo- lecular weights of the two salts. t Journ. fiir Prakt. Chem., 59, 27. 16 242 THE ATOMIC WEIGHTS. Iii his earlier paper* the mode of procedure was about as follows: The two salts, weighed out in quantities having approximate chemical equivalency, were placed in two small flasks, and to each was added 100 cc. of a ferrous chloride solution and 30 cc. hydrochloric acid. The ferrous chloride was added in trifling excess, and, when action ceased, the amount unoxidized was determined by titration with a standard solu- tion of dichrpmate. As in each case the quantity of ferrous chloride was the same, it became easy to deduce from the data thus obtained the ratio in question. I have reduced all of his somewhat complicated figures to a simple common standard, and give below the amount of chromate equivalent to 100 of chlorate : 120.118 120.371 120.138 120.096 120.241 120.181 Mean, 120.191, .028 In his later paper f Kessler substituted arsenic trioxide for the iron solution. In one series of experiments the quantity of dichromate needed to oxidize 100 parts of the arsenic trioxide was determined, and in an- other the latter substance was similarly compared with 'the chlorate. The subjoined columns give the quantity of each salt proportional to 100 of As 2 3 : Mean, 99.045, .028 Mean, 41.172, .009 Reducing the later series to the standard of the earlier, the two com- bine as follows : '(l) 2KC1O 3 : K 2 Cr 2 O 7 : : 100 : 120.191, .028 (2) 2KC1O 3 : K 2 Cr 2 O 7 : : 100 : 120.282, .043 General mean ...... 120.216, .0235 *Poggend. Annalen, 95, 208 1855. fPoggend. Annalen, 113, 137. 1861. CHROMIUM. 243 Siewert's determinations, which do not seem to have attracted general attention, were published in 1861.* He, reviewing Berlin's work, found that upon reducing silver chromate with hydrochloric acid and alcohol, the chromic chloride solution always retained traces of silver chloride dissolved in it. These could be precipitated by dilution with water ; but, in Berlin's process, they naturally came down with the chromium hydroxide, making the weight of the latter too high ; hence too large a value for the atomic weight of chromium. In order to find a more cor- rect value Siewert resorted to the analysis of sublimed, violet, chromic chloride. This salt he fused with sodium carbonate and a little nitre, treated the fused mass with water, and precipitated from the resulting solution the chlorine by silver nitrate in presence of nitric acid. The weight of the silver chloride thus obtained, estimated after the usual manner, gave means for calculating the atomic weight of chromium. His figures, reduced to a common standard, give, as proportional to 100 parts of chloride of silver, the quantities of chromic chloride stated in the third of the subjoined columns : .2367 grm. CrCl s gave .6396 grm. AgCl. 37-Oo; .2946 " .7994 3 6 .853 .2593 -7039 36-838 .4935 I -3395 36.842 .5850 " 1.5884 " 36-830 .6511 " 1.76681 " 36.852 .5503 " L4939I " 36.836 Mean, 36.865, .0158 The first of these figures varies so widely from the others that we are justified in rejecting it, in which case the mean becomes 86.842, .0031. Siewert also made two analyses of silver dichromate by the following process. The salt, dried at 120, was dissolved in nitric acid. The silver was then thrown down by hydrochloric acid, and, in the filtrate, chro- mium hydroxide was precipitated by ammonia. Reduced to a uniform standard, we find from his results, corresponding to 100 parts of AgCl, Ag 2 O 2 7 as in the last column : .7866 grm. Ag 2 Cr 2 O 7 gave .52202 AgCl and .2764 Cr 2 O 3 . 150.684 1.089 " .72249 " .3840 " 150.729 Berlin's single determination of this ratio gave 151.035. Taking all three values together as one series, they give a mean of 150.816, .074. Siewert's percentages of Cr. 2 3 obtained from Ag 2 O 2 O r are as follows, calculated from the above weighings : 35-'39 35.262 Mean, 35.2005, .0415 * Zeit. Gesammt. Wissenschaften, 17, 530. 244 THE ATOMIC WEIGHTS. Combining, as before, with Berlin's single result, giving the latter equal weight with one of these, we have a general mean of 35.236, .0335. For the ratio between silver chloride and chromic oxide, Siewert's two analyses of the dichromate come out as follows. For 100 parts of AgCl we have of Cr 2 8 : Mean, 53.049, .068 This figure, reduced to the standard of Berlin's work on the mono- chromate, becomes 26.525, .034. Berlin's mean was 26.682, .0076. The two means, combined, give a general mean of 26.676, .074. By Baubigny * we have only three experiments upon the calcination of anhydrous chromic sulphate, as follows : 1.989 grm. Cr 2 (SO 4 ) 8 gave .7715 grm. Cr. 2 O 3 . 38.788 per cent. 3.958 " 1.535 " 38.782 " 2.6052 1.0115 " 38.826 " Mean, 38:799, .0092 Moberg found for the same ratio the percentage 39.195, .028. The general mean of both series, Moberg's and Baubigny's, is 38.838, .0087. In Rawson's work f ammonium dichromate was the substance studied. Weighed quantities of this salt were dissolved in water, and then reduced by hydrochloric acid and alcohol. After evaporation to dryness the mass was treated with water and ammonia, reevaporated, dried five hours at 140, and finally ignited in a muffle. The residual chromic oxide was bright green, and was tested to verify its purity. The corrected weights are as follows : Am^Cr^O-. Cr. 2 O s . Percent. Cr. 2 O 3 . 1.01275 -61134 60.365 1.08181 .65266 60.330 1.29430 -78090 60.334 1.13966 .68799 60.368 98778 .59595 60.332 1.14319 .68987 60.346 Mean, 60.346, .0046 Latest in time and most elaborate of all, we come to the determinations of the atomic weight of chromium made by Meineke,J who studied the chromate and ammonio-chromate of silver, and also the dichromates of potassium and ammonium. For the latter salt he measured the same ratio that Rawson determined, but by a different method. He precipi- *Compt. Rend., 98, 146. tjourn. Chem. Soc., 55, 213. t Ann. d. Chem., 261, 339. 1891. CHROMIUM. 245 tated its solution with mercurous nitrate, and ignited the precipitate, with the subjoined results. Vacuum weights are given ; Am. 2 Cr. 2 O r Cr 2 O s . Percent. Cr 2 O s . 2.0416 1.2316 60.325 2.1618 1.3040 60.320 2.0823 1.2562 60.328 2.1913 1.3221* 60.335 2.0970 1.2656 60.353 Mean, 60.332, .0037 Rawson found, 60.346, .0046 General mean, 60.337, =b .0029 The chromate of silver, Ag. 2 Cr0 4 , and the ammonio-chromate, Ag,Cr0 4 .4NH 3 , both prepared with all necessary precautions to insure purity, were first treated essentially as in Berlin's experiments, except that the traces of silver chloride held in solution by the chromic chloride were thrown out by sulphuretted hydrogen, estimated, and their amount added to the main portion. Thus the chief error in Berlin's work was avoided. I subjoin the data obtained, with vacuum standards, as usual. All of Meineke's results are so corrected : Ag.CrO,. 2.7826 3.2627 3.6362 4.6781 3-2325 3-9I37 AgCL 2.4047 2.8199 3.1416 4.0414 2.7930 3-3805 .6384 .7480 8338 1.0726 -74H .8976 Hence we have the following ratios, as in the case of Berlin's data : Percent. Cr. 2 O s . looAgCl : Ag^CrO^. looAgCl : 22.943 "5-7I5 26.548 22.926 "5.703 26.526 22.931 115.744 26.602 22.928 115.754 26.601 22.924 "5.736 26.531 22.935 "5-773 26.552 Mean, 22.931, .0019 Berlin, 23.014, =t .0110 Mean, 115.737, .0072 Mean, 26.560, .0093 Berlin, 115.956, db .0230 General mean, 22.934, =h .0018 General mean, 115.760, .0069 With the ammonio-chromate Meineke found as follows : ' AgCL Cr,O,. 4.1518 4.2601 5.9348 2.9724 3.0592 4.2654 794 .8125 1.1317 * Calculated back from Meineke's value for Cr, to replace an evident misprint in the original. 246 THE ATOMIC WEIGHTS. And the ratios become Percent. Cr.,O. A . looAgCl : Salt. woAgCl : Cr.,O 3 . 19.037 139-679 26.591 19.072 139.255 26.559 19.059 i39- I 38 26.532 Mean, 19.059, HZ .0074 Mean, 139.357, .1 109 Mean, 26.561, =h .01 15 The first of these three analyses is rejected by Meineke as suspicious, but for the present I shall allow it to remain. The data in the third column may now be combined with the corresponding figures from the normal chromate, as found by Meineke and his predecessors. Berlin 26.682, .0076 Siewert, from Ag 2 Cr 2 O 7 26.525, .0340 Meineke, from Ag 2 CrO 4 26.560, .0093 Meineke, from Ag 2 CrO 4> 4NH 3 26.561, dr .0115 General mean ..................... 26.620, rb .0052 : Cr 2 O 3 : : 100 : 26.620, .0052 Obviously, this mean is vitiated by the known error in Berlin's work, the ultimate effect of which is yet to be considered. In all four of the salts studied by Meineke he determined volumetric- ally the oxygen in excess of the normal oxides by measuring the amount of iodine liberated in acid solutions. With the silver salts the process was essentially as follows : A weighed quantity of the chromate was dis- solved in weak ammonia, and the solution was precipitated with potas- sium iodide. After the silver iodide had been filtered off, five or six grammes of potassium iodide were added to the filtrate, which was then acidulated with phosphoric acid and a little sulphuric. The liberated iodine was then titrated with sodium thiosulphate solution, which had been standardized by means of pure iodine, prepared by Stas' method, From the iodine thus measured the excessive oxygen was computed, and from that datum the atomic weight of chromium was found. For pres- ent purposes, however, the data may be used more directly, as giving the ratios I 3 : Ag 2 Cr0 4 and I, : Ag 2 Cr0 4 .4NH 3 . Thus treated, the weights are as follows, reduced to a vacuum. Reckoning the salt as 100, the third column gives the percentage of iodine liberated : Ag.fr O. I Set Free. Percentage. .43838 .50251 114.628 .90258 1.03432 H4-595 .89858 1.02980 114.603 .89868 1.03072 T 14.693 Mean, 114.630, .015 CHROMIUM. 247 The next series, obviously, gives the ratio I 3 : Ag 2 CrO 4 .4NH 3 . / Set Free. Percentage * .54356 .51784 95-267 .54856 .5 20 46 94.877 .54926 .52322 95.258 .54906 .52376 95.392 .54466 .5*910 95.307 .54536 .51891 95- 15 Mean, 95.208, =b .0497 In dealing with the two dichromates Meineke used the acid potassium iodate in place of potassium iodide, the chromate and the iodate reacting in the molecular ratio of 2:1. The thiosulphate was standardized by means of the acid iodate, so that we have direct ratios between the latter and the two chromates. The data are as follows, with the amount of iodate proportional to one hundred parts of the dichromate in the third column : Percentage. .25090 .16609 66.198 .25095 .16613 66.200 .25078 .16601 66.197 .24979 .16541 66.220 .24987 .16540 66.192 .24966 16543 66.262 .25015 16559 66.196 .25012 .16559 66.204 .24977 .16546 66.245 .25034 .16572 66.198 .25025 .16567 66.202 .25015 .16568 66.234 Mean, 66.212, .0044 Am. 2 Cr. l0r KHI^O, Percentage. .21457 .16584 77.290 .21465 .16588 77.279 .21464 .16584 77-264 .21416 .'6543 77.246 .21447 .16564 77.232 .21427 16559 77.281 .22196 .17152 77.272 .22194 17151 77.278 .22180 '7139 77-272 Mean, 77.268, .0041 * These figures are not wholly in accord with the percentages of oxygen computed by Meineke. I suspect that there is a misprint among his data as published, probably in the second experi- ment, but I cannot trace it with certainty. 248 THE ATOMIC WEIGHTS. The following ratios are now available for computing the atomic weight of chromium : (i.) Percentage Cr 2 O 3 from Ag 2 CrO 4 , 22.934, .0018 (2.) Percentage Cr 2 O 3 from Ag 2 Cr 2 O 7 , 35.236, =b .0335 (3.) 2AgCl : Ag 2 CrO 4 : : loo : 115.760, rb .0069 (4.) 2AgCl : Ag 2 Cr 2 O 7 : : 100 : 150.816, .074 (5.) 4AgCl : Cr 2 O 3 : : loo : 26.620, rb .0052 (6.) Percentage Cr 2 O 3 in Cr 2 (SO 4 ) 3 , 38.838, =b .0087 (7.) Percentage Cr 2 O 3 in AmCr(SO 4 ) 2 . 12H 2 O, 16.143, .0125 (8.) BaSO 4 : BaCrO 4 : : 100 : 108.9815, .0369 / (9.) BaCrO 4 : BaCl 2 : : 100 : 81.702, .014 (10.) 3AgCl : CrCl 3 : : 100 : 36.842, .0031 (II.) 2KC1O 3 : K 2 Cr 2 O 7 : : 100 : 120.216, rb .0235 (12.) Percentage Cr 2 O 3 in Ag 2 CrO 4 .4NH 3 , 19.059, .0074 (13 ) 2AgCl : Ag 2 CrO 4 .4NH 3 : : 100 : 139.357, d= .1109 (14.) Percentage Cr 2 O 3 in Am 2 Cr 2 O 7 , 60.337, .0029 (15.) Ag 2 CrO 4 : 3! : : 100 : 114.630, .015 (16.) Ag 2 CrO 4 .4NH 3 : 3! : : 100 : 95.208, -O497 (17.) 2K 2 Cr 2 O 7 : KHI 2 O 6 : : 100 : 66.212, =b .0044 (18.) 2Am 2 Cr 2 O 7 : KHI 2 O 6 : : 100 : 77.268, .0041 The antecedent values to use in the reduction are = 15 879, .0003 S = 31.828, rb .0015 Ag = 107. 108, zb .0031 N = 13.935, rb .0021 Cl = 35.179, rb .0048 Ba 136.392, rb .0086 1 = 125.888, rb .0069 AgCl = 142.287, .0037 K = 38.817, .0051 For the molecular weight of Cr a O s , seven values are now calculable, as follows : From (i) ................ ...... Cr 2 O 3 151.120, .0130 From (2) ...................... " = 151.105, rb .1636 From (5) ...................... " = 151.507, .0299 From (6) .................. ---- " = 151.384,^.0341 Prom (7) ..................... " =.- 153.756, .1205 From (12) ..................... " 151.478, .0606 From (14) ..................... " = 151.190, .0110 General mean ............ Cr 2 O 3 = 151.229, .0039 For silver chromate there are two values From (3) .................... Ag 2 CrO 4 = 329.423, .0195 From (15) ................... " =r 329.464, it .0467 General mean .......... Ag 2 CrO 4 = 329.430, rb .0180 And for the ammonio-chromate we have From (13) ............. Ag 2 CrO 4 .4NH 3 = 396-574, - From (16) ............. " = 396.673, rb .2082 General mean Ag. 2 CrO 4 .4NH 3 = 396.647, .1738 CHROMIUM. 249 From (4) Ag 2 Cr 2 O 7 = 429-177, .2109 From (10) CrCl 3 = 157.266, dr .01 13 From (18) Am 2 Cr 2 O 7 250.341, dr .0164 For the molecular weights of K 2 Cr 2 7 and BaCr0 4 there are two esti- mates each, as given below : From (u) K 2 Cr 2 O 7 = 292.433, =b .0189 From (17) " = 292.143, =b .0224 General mean K 2 Cr 2 O 7 = 292.311, dr .0144 From (8) BaCrO 4 = 252.549, ,dr .0966 From (9) " = 253.054, .0377 General mean BaCrO 4 252.985, .0351 Finally, from these molecular weights, eight independent values are obtained for the atomic weight of chromium : From Cr 2 O 3 Cr = 5 1 . 796, dr .0039 From Ag 2 CrO 4 " === 51.698, dr .0191 From Ag 2 CrO 4 , 4NH 3 " =51.175, .1741 From Ag 2 Cr 2 O 7 " = 51.904, dr .1055 From Am 2 Cr 2 O 7 " 51.659, .0085 From K 2 Cr 2 O 7 "= 51.762, .0102 From CrCl 3 " =51.729, dr .0183 From BaCrO 4 " = 53.077, .0362 General mean Cr = 51.778, dr -0032 If = 16, Cr = 52.172. Rejecting the last of the eight values, that from barium chromate, the mean becomes Cr = 51. 767, .0032. Even this result is probably too high, for it includes ratios which are certainly erroneous, and which yet exert appreciable weight. From the ratios which are reasonably concordant a better mean is derivable, as follows : From (l) Cr 51.741, dr .0065 From (2).. " =51.734, db .0818 From (14) " =51.776, .0055 From (3) and (15) " = 51.698, d= .0191 From (4) " =51.904, .1055 From (10) " = 51.729, dr .0183 From (18) <9 6 . Loss hi Weight. Percent. Te. 1.2299 .5471 55.517 1.0175 -45 26 55-5'S 2.5946 I.I549 55-488 Mean, 55.508, .0068 Hence Te = 126.303. Staudenmaier also gives four reductions of Te0 2 to Te, in presence of finely divided silver. The data are as follows : . 7><9. 2 . Loss in Weight. Per cent. Te. .9171 .1839 79.948 i 9721 .3951 79.966 2-4115 -4835 7995 1.0172 .2041 79-935 Mean, 79.950, .0043 Hence Te = 126.636. The last series, giving the percentage of tellurium in the dioxide, com- bines with previous series thus : Berzelius 80.042, .0050 Wills, first series 80.015, d= .0410 Wills, second series 80.028, .0040 Brauner, synthesis 79. 7 r I , .0239 Brauner, analysis 79-93 2 , .534 Staudenmaier 79-95i .004 3 General mean 80.001, =t .0025 The very recent determinations byChikashige* were made by Brauner's method, giving the ratio between silver and TeBr 4 . In all essential par- ticulars the work resembles that of Brauner. except that the tellurium, * Journ. Chetn. Soc., 69, 8Si. 1896. 276 THE ATOMIC WEIGHTS. instead of being extracted from metallic tellurides, was derived from Japanese native sulphur, in which it exists as an impurity. This differ- ence of origin in the material studied gives the chief interest to the investigation. The data are as follows : Ag. Ratio. 4.1812 4.0348 103.628 4.3059 4-1547 103.639 4.59 2 9 4.43 ! 9 103.633 Mean, 103.633, .0023 Brauner found, 103.644, .0018 General mean, 103.640, it .0014 Now, to sum up, the subjoined ratios are available for computing the atomic weight of tellurium : (I.) Percentage Te in TeO 2 , 80.001, =b .0025 (2.) Percentage Te in H 6 TeO 6 , 55.508, .0068 (3.) Percentage TcO 2 in H 6 TeO 6 , 64.440, .0024 (4.) Ag 4 : TeBr 4 : : 100 : 103.640, .0014 (5.) K 2 TeBr g : 6AgBr : : 100 : 164.468, =b .0330 (6.) TeO 2 : O : : 100 : 10.068, =b .0100 To reduce these ratios we have O = 15.879,^.0003 K = 38.817, .0051 Ag = 107.108, .0031 AgBr = 186.452, rb .0054 Br = :: 79-344, . -0062 For the atomic weight of tellurium six values appear, as follows : From (i) Te = 127.040, .0165 From (4) " = 126.650, rb .0302 From (5) " = 126.502,1^.1430 From (2) " =126.303,^.0246 From (3) " = 126.209, zb .0138 From (6) " = 125.960,^.1574 General mean Te = 126.523, rb .0092 If = 16, Te = 127.487. A careful consideration of the foregoing figures, and of the experi- mental methods by which they were obtained, will show that they are not absolutely conclusive with regard to the place of tellurium under the periodic law. The atomic weight of iodine, calculated in a previous chapter, is 125.888. Wills 1 values for Te, rejecting his first series as rela- tively unimportant, range from 125.40 to 127.32 ; that is, some of them fall below the atomic weight of iodine, although none descend quite to the 125 assumed by Mendelejeff. Some of Brauner's data fall even lower; and the same thing is true in FLUORINE. 277 Gooch and Rowland's series, of which the mean gives Te = 125.96, a value very little above that of iodine. In considering the experimental methods, reference may properly be made to the controversy regarding the atomic weight of antimony. It will be seen that Dexter, estimating the latter constant by the conver- sion of the metal into Sb 2 4 , obtained a value approximately of Sb = 122. Dumas, working with SbCl 3 , obtained nearly the same value. Schneider and Cooke, on the other hand, have established an atomic weight for antimony near 120, and Cooke in particular has traced out the constant errors which lurked unsuspected in the work of Dumas. Now in their physical aspects tellurium and antimony are quite similar. The oxida- tion of tellurium to dioxide resembles in many particulars that of anti- mony, and may lead to error in the same way. In each of the six tel- lurium ratios there is still uncertainty, and a positive measurement, free from objections, of the constant in question is yet to be made. FLUORINE. The atomic weight of fluorine has been chiefly determined by one general method, namely, by the conversion of fluorides into sulphates. The work of Christensen, however, is on different lines. Excluding the early results of Davy,* we have to consider first the experiments of Berzelius, Louyet, Dumas, De Luca, and Moissan with reference to the fluorides of calcium, sodium, potassium, barium, and lead. The ratio between calcium fluoride and sulphate has been determined by the five investigators above named, and by one general process. The fluoride is treated with strong sulphuric acid, the resulting sulphate is ignited, and the product weighed. In order to insure complete trans- formation special precautions are necessary, such, for instance, as re- peated treatment with sulphuric acid, and so on. For details like these the original papers must be consulted. The first experiments in chronological order are those of Berzelius,f who operated upon an artificial calcium fluoride. He found, in three experiments, for one part of fluoride the following of sulphate : 1-749 1.750 I-75I Mean, 1.750, .0004 Louyet's researches J were much more elaborate than the foregoing. He began with a remarkably concordant series of results upon fluor spar, * Phil. Trans., 1814, 64. f Poggend. Annalen, 8, i. 1826. I Ann. Chim. Phys. (3), 25, 300. 1849. 278 THE ATOMIC WEIGHTS. in which one gramme of the fluoride yielded from 1.734 to 1.737 of sul- phate. At first he regarded these as accurate, but he soon found that particles of spar had been coated with sulphate, and had therefore escaped action. In the following series this source of error was guarded against. Starting with fluor spar, Louyet found of sulphate as follows: .742 744 745 744 7435 7435 Mean, 1.7437, .0003 A second series, upon artificial fluoride, gave : i.743 1.741 Mean, 1.7417, .0004 Dumas * published but one result for calcium fluoride. .495 grm. gave .864 grm. sulphate, the ratio being 1 : 1.7455. De Lucaf worked with a very pure fluor spar, and published the fol- lowing results. The ratio between CaS0 4 and one gramme of CaF 2 is given in the third column : .9305 grm. CaF 2 gave 1.630 grm. CaSO 4 . .836 " 1.459 " 1.7452 .502 " .8755 " 1.7440 .3985 " .6945 " 1.7428 If we include Dumas' single result with these, we get a mean of 1.7459, .0011. MoissanJ unfortunately gives no details nor weighings, but merely states that four experiments with calcium fluoride gave values for F rang- ing from 19.02 to 19.08. To S he assigned the value 32.074, and probably Ca was taken as 40. With these data his extreme values as given may be calculated back into uniformity with the ratio as stated above, becoming 1-7444 1.7410 Mean, 1.7427 *Ann. Chem. Pharm., 113, 28. t Compt. Rend., 51, 299. 1860. I Compt. Rend , in, 570. 1890. FLUORINE. 279 If we assign this equal weight with Berzelius' series, the data for this ratio combine thus : Berzelius 1.7500, .0004 Louyet, first series 1.7437, .0003 Louyet, second series 1.7417, .0004 De Luca with Dumas 1.7459, .0011 Moissan 1.7427, .0004 General mean 1.7444, .00018 For the ratio between the two sodium salts we have experiments by Dumas, Louyet, and Moissan. According to Louyet, one gramme of NaF gives of Na 2 S0 4 1.686 1.683 1.685 Mean, 1.6847, .0006 The weighings published by Dumas are as follows : .777 grm. NaF give 1.312 grm. Na 2 SO 4 . Ratio, 1.689 1.737 " 2.930 " " 1.687 Mean, 1.688, .0007 Moissan says only that five experiments with sodium fluoride gave . F = 19.04 to 19.08. This was calculated with Na = 23.05 and S'= 32.074. Hence, reckoning backward, the two values give for the standard ratio 1.6873 Mean, 1.6881 Giving this equal weight with Dumas' mean, we have Louyet 1 .6847, =fc .0006 Dumas 1.688, .0007 Moissan 1.6881, .0007 General mean 1 .6867, .00038 Dumas also gives experiments upon potassium fluoride. The quantity of sulphate formed from one gramme of fluoride is given in the last column : 1.483 grm. KF give 2.225 g rm - K 2 SO 4 . 1.5002 1.309 " 1.961 " 1.4981 Mean, 1.499^ .0007 The ratio between barium fluoride and barium sulphate was measured 280 THE ATOMIC WEIGHTS. by Louyet and Moissan. According to Louyet, one gramme of BaF., gives of BaS(\ L33 2 1.331 1.330 Mean, 1.331, =b .0004 Moissan, in five experiments, found F 19.05 to 19.09. Assuming that he put Ba = 137, and S 32.074 as before, these two extremes become 1-3305 Mean, 1.3308 Giving this equal weight with Louyet's mean, we get the subjoined combination : Louyet I-33I, .0004 Moissan 1 .3308, db .0004 General mean i-339> .00028 The experiments with lead fluoride are due to Louyet, and a new method of treatment was adopted. The salt was fused, powdered, dis- solved in nitric acid, and precipitated by dilute sulphuric acid. The evaporation of the fluid and the ignition of the sulphate was then effected without transfer. Five grammes of fluoride were taken in each opera- tion, yielding of sulphate : 6.179 6.178 6.178 Mean, 6.1783, d= .0002 In Christensen's determinations* we find a method adopted which is radically unlike anything in the work of his predecessors. He started out with the salt (NH 4 ) 2 MnF 5 . When this is added to a mixture, in solution, of potassium iodide and hydrochloric acid, iodine is set free, and may be titrated with sodium thiosulphate. One molecule of the salt (as written above), liberates one atom of iodine. In four experi- ments Christensen obtained the following data : 3.1199 grm. Am. 2 MnF 5 gave 2.12748 I. 68.191 per cent. 3.9190 " 2.67020 " 68.135 " 3.5005 " 2.38429 " 68.113 " 1.2727 " .86779 " 68.185 " Mean, 68.156, .0128 * Journ. fiir Prakt. Chem. (2), 35, 541. Christensen assigns to the salt double the formula here given. FLUORINE. 2j31 The ratios from which to compute the atomic weight of fluorine are now (I.) CaF 2 : CaSO 4 : : i.o : 1.7444, .00018 (2.) 2NaF : Na 2 SO 4 : : i.o : 1.6867, .00038 (3.) 2KF : K 2 SO 4 : : i.o : 1.4991, .0007 (4.) BaF 2 : BaSO 4 : : i.o : 1.3309, .00028 (5.) PbF 2 : PbSO 4 : : 5.0 : 6.1783, .0002 (6.) Am 2 MnF 5 : I : : 100 : 68.156, .0128 To reduce them we have l 5-&79, db .0003 K 38.817, dr .0051 S = 31.828, .0015 Ca = 39.764, =h .0045 N = 13.935, =t .0021 Ba = 136.392, zfc .0086 1 125.888, .0069 Pb = 205.358, .0040 Na 22.881, rh .0046 Mn= 54-571, i: .0013 And the values derived for fluorine are as follows: From (i) F= 18.844, d- .0048 From (2) " = 18.948, dr .0108 From (31 " = 18.877, .0276 From (4) " = 18.869, .0192 From (5) " = 18.997, dr .0047 From (6) " = 18.853, .0073 General mean F = 18.912, .0029 If O = 16, F = 19.056. In all probability these values for fluorine average a trifle too high. It is difficult to be certain that a fluoride has been completely converted into sulphate, and an incomplete conversion tends to raise the apparent atomic weight of fluorine. This possible source of error exists in all of the ratios except the last one, but the fair concordance of the results obtained seems to indicate that the uncertainty cannot be very large. 282 THE ATOMIC WEIGHTS. MANGANESE. The earliest experiments of Berzelius* and of Arfvedsonf gave values for Mn ranging between 56 and 57, and therefore need no farther con- sideration here. The first determinations to be noticed are those of Turner J and a later measurement by Berzelius. who both determined gravimetrically the ratio between the chlorides of manganese and silver. The manganese chloride was fused in a current of dry hydrochloric acid, and afterwards precipitated with a silver solution. I give the MnCl 2 equivalent to 100 parts of AgCl in the third column : 4.20775 grm. MnG 2 == 9.575 grm. AgCl. 43-945 \ , _ _ > 3.063 = 6.96912 43-95-' 12.47 grains MnQ 2 = 28.42 grains AgCl. 43.878 Turner. Mean, 43.924, .015 Many years later Dumas || also made the chloride of manganese the starting point of some atomic weight determinations. The salt was fused in a current of hydrochloric acid, and afterwards titrated with a standard solution of silver in the usual way. One hundred parts of Ag are equiva- lent to the quantities of MnCl 2 given in the third column : 3.3672 grm. MnCl 2 = 5.774 grm. Ag. 5 8 -3i7 3.0872 " 5.293 " 58.3 26 2.9671 " 5-0875 " 58.321 1.1244 1.928 " 58.320 1.3134 " 2.251 " 58.321 Mean, 58.321, =h .001 An entirely different method of investigation was followed by von Hauer,^]" who, as in the case of cadmium, ignited the sulphate in a stream of sulphuretted hydrogen, and determined the quantity of sulphide thus formed. I subjoin his weighings, and also the percentage of MnS in MnS0 4 as calculated from them : 4.0626 grm. Mn?O 4 gave 2.3425 grm. MnS. 57-66o per cent. 4.9367 " 2.8442 " 57.613 " 5.2372 " 3- OI 9 2 " 57.649 c< 7.0047 " 4.0347 " 57.600 " 4.9175 " 2.8297 " 57-543 4-8546 " 2.7955 57.585 " 4.9978 2.8799 " 57.625 4 6737 " 2.6934 " 57.629 4.7240 2.7197 " 57.57 2 Mean, 57.608, =fc .008 * Poggend. Anualen, 8, 185. 1826. t Berz. Jahresbericht, 9, 136. 1829. | Trans. Roy. Soc. Ediub., ir, 143. 1831. I Lehrbuch, 5 Aufl., 3. 1224. || Ann. Chem. Pharm., 113, 25. 1860. If Journ. fur Prakt. Chem., 72, 360. 1857. MANGANESE. 283 This method of von Hauer, which seemed to give good results with cadmium, is, according to Schneider,* inapplicable to manganese, for the reason that the sulphide of the latter metal is liable to be contaminated with traces of oxysulphide. Such an impurity would bring the atomic weight out too high. The results of two different processes, one carried out by himself and the other in his laboratory by Rawack, are given by Schneider in this paper. Rawack reduced manganoso-manganic oxide to manganous oxide by ignition in a stream of hydrogen, and weighed the water thus formed. From his weighings I get the values in the third column, which repre- sent the Mn 3 4 equivalent to one gramme of water: 4.149 grm. Mn 3 O 4 gave 0.330 grm. II 2 O. 12.5727 4.649 " .370 " 12.5643 6.8865 .5485 " 12.5552 7.356 " .5855 " 12.5636 8-9445 -7135 " 12.5361 11.584 .9225 " 12.5572 Mean, 12.5582, .0034 Here the most obvious source of error lies in the possible loss of water. Such a loss, however, would increase the apparent atomic weight of manganese ; but we see that the value found is much lower than that^ obtained either by Dumas or von Hauer. Schneider himself effected the combustion of manganous oxalate with oxide of copper. The salt was not absolutely dry, so that it was neces- sary to collect both water and carbon dioxide. Then, upon deducting the weight of water from that of the original material, the weight of anhydrous oxalate was easily ascertained. Subtracting from this the CO ? , we get the weight of Mn. If we put CO 2 = 100, the quantities of manganese equivalent to it will be found in the last column : 1.5075 grm. oxalate gave .306 grm. H 2 O and .7445 grm. CO 2 . 61.3835 2.253 .4555 " *- ll 35 " 61.4291 3.1935 -652 1-5745 " 61.4163 5.073 " 1.028 ." 2.507 " 61.3482 Mean, 61.3943, =b .0122 Up to this point the data give two distinct values for Mn one near 54, the other approximately 55 and with no sure guide to preference between them. The higher value, however, has been confirmed by later testimony. In 1883 Dewar and Scott f published the results of their work upon silver permanganate. This salt is easily obtained pure by recry stall iza- tion, and has the decided advantage of not being hygroscopic. Two sets * Poggend. Annalen, 107, 605. tProc. Roy. Soc., 35, 44. 1883. 284 THE ATOMIC WEIGHTS. of experiments were made. First, the silver permanganate was heated to redness in a glass hulb, first in air, then in hydrogen. Before weigh- ing, the latter gas was replaced by nitrogen. The data are as follows : ^g + MnO. Per cent. Ag + MnO. 5-8696 4.63212 78.917 5-4988 4-3359 1 78.852 7.6735 6.05395 78.894 13-10147 10.31815 78.756 12.5799 {9.9.065 78.782 (9.91435 , 78.811 Mean, 78.835, .0174 The duplication of the last weighing is not explained. In the second series the permanganate was dissolved in dilute nitric acid, reduced by sulphur dioxide, potassium nitrite, or sodium formate, and titrated with potassium bromide. The AgMn0 4 equivalent to 100 KBr appears in the third column. AgMnO. KBr. Ratio. 6.5289 3-42385 190.686 7.5378 3-9553 190.575 6.1008 3.20166 * 90.559 5.74647 3-00677 191.117 6.16593 3. 23602 190.540 5.11329 2.6828 190.596 5.07438 2.66204 190.624 13.4484 7.05602 190.604 12.5799 6.60065 190.588 12.27025 6.43808 190.584 Mean, 190.647, .0361 Vacuum weights are given throughout. To the first series of experi- ments the authors attach little importance, and numbers 1 and 4 of the second series they also regard as questionable. These experiments rep- resent the use of sulphur dioxide as the reducing agent, and were attended by the formation of an insoluble residue, apparently of a sulphide. Ex- cluding them, the remaining eight experiments of the second series give in mean KBr : AgMnO 4 : : 100 : 190.584, db .0062, which will be used for the present calculation. Dewar and Scott also made determinations with manganese chloride and bromide. With the first salt they found Mn = 54.91, and with the second, Mn = 54.97 ; but they give no details. Marignac's work upon the atomic weight of manganese also appeared in 1883.* He prepared the oxid.e, MnO, by ignition of the oxalate and ^Arch. vSci. Phys. et Nat. (3), 10. 21. 1883. MANGANESE. 285 subsequent reduction of the resulting Mn 3 O 4 in hydrogen. The oxide, with various precautions, was then converted into sulphate. The per- centage of MnO in MnS0 4 is appended : 2.6587 grrn. MnO gave 5.6530 MnSO 4 . 47.032 per cent. 2.5185 " 5-3600 " 46.987 " 2.5992 5-5 2 95 " 47.oo6 " 2.8883 6.1450 " 47.002 " Mean, 47.007, + .0025 J. M. Weeren, in 1890,* published determinations made by two meth- ods, the one Marignac's, the other von Hauer's. From manganese sul- phate he threw down the hydrated peroxide electrolytically,and the latter compound was then reduced in hydrogen which had been proved to be free from oxygen. The resulting monoxide was cooled in a stream of purified nitrogen. After the oxide had been treated with sulphuric acid, converted into sulphate, and weighed, a few drops of sulphuric acid and a little sulphurous acid were added to it, after which it was reheated and weighed again. This process was repeated until four successive weigh- ings absolutely agreed. The results of this set of experiments were as follows, with vacuum standards : 15.2349 grm. MnO gave 32.4142 MnSO 4 . 47.005 per cent. 13.9686 " 29.7186 " 47.004 " 13.7471 29.2493 " 47.000 " ^ 15.5222 " 33.0246 " 47.001 " 14.9824 " 3 I -8755 " 47.002 " 14.6784 " 3 -2304 " 47.000 Meanj 47.002, .0006 Marignac's mean, combined with this, hardly affects either the per- centage itself or its probable error. Fortunately, both Marignac and Weeren are completely in agreement as to the ratio, and either set of measurements would be valid without the other. In order, therefore, to give Marignac's work some proper recognition, we can assume a general mean of 47.004, =b .0006, without danger of serious error. The manganese sulphate produced in the foregoing series of experi- ments was used, with many precautions, for the next series carried out by von Hauer's method. It was transferred to a porcelain boat, dried at 260 to avoid errors due to retention of water taken up in the process of transfer, and then heated to constant weight in a stream of hydrogen sulphide. Before weighing, the sulphide was heated to redness in hy- drogen and cooled in the same gas. The results, with vacuum weights, were as follows : * Atom-Gewichtsbestimmung des Mangans. Inaugural Dissertation, Halle, 1890. 286 THE ATOMIC WEIGHTS. 16.0029 g rm - MnSO 4 gave 9.2228 MnS 57.632 per cent. 16.3191 " 9.4048 " 57.631 15.9307 9.1817 " 57-634 " 15-8441 9-131$ " 57.634 " 16.2783 9.3819 " 57.635 " 17.0874 9-8477 " 57.633 " Mean, 57.633, .0004 von Hauer found, 57.608, =b .0080 Hence the general mean is identical with Weeren's to the third deci- mal place, which is unaffected by combination with von Hauer's data. We have now to consider the following ratios for manganese : (i.) 2AgCl : MnCl 2 : : 100 : 41.924, =b .0150 (2.) 2Ag : MnCl 2 : : loo : 58.321, d= .0010 (3.) 1J 2 O : Mn 3 O 4 : : 100 : 1255.82, .340 (4. ) 2L'O 2 : Mn : : 100 : 61.3943, .0122 (5.) AgMnO 4 : Ag -f MnO : : 100 : 78.835, .0174 (6.) KBr : AgMnO 4 : : 100 : 190.584, .0062 (7.) MnSO 4 : MnO : : 100 : 47.004, .0006 (8.) MnSO 4 : MnS : : 100 : 57.633, .0004 Computing with the subjoined preliminary data O 15.879,^.0003 K = 38.817,^.0051 Ag = 107.108, .0031 C = 1 1.920, .0004 Cl = 35.179, dr .0048 S 31.828,^.0015 Br = 79-344, .0062 AgCl = 142.287, -0037 these ratios reduce as follows : First, for the molecular weight of manganese chloride, two values are deducible. From (i) MnCl 2 124.996, d= .0428 From (2) " = 124.933, .0042 General mean MnO 2 124.934, .0042 Hence Mn = 54.576, .0075. For manganese there are seven independent values, as follows : From molecular weight MnCl 2 Mn = 54.576, .0075 From (3) " = 5.3.667, ..0203 From (4) " = 53.633, .0107 From (5) " = 54.450, .1511 From (6) " = 54.572, .0173 From (7) " 54.601, .0018 From (8) " = 54-575, dr .0022 General mean Mn = 54.571, =fc .0013 If = 16, this becomes Mn = 54.987. In this case five of the separate values are well in accord, and the re- jection of the two aberrant values, which have high probable errors, is IRON. 287 not necessary. Their influence is imperceptible. Weeren's marvelously- concordant data seem to receive undue weight, but they are abundantly confirmed by the evidence of other experimenters. In short, the atomic weight of manganese appears to be quite well determined. IRON. The atomic weight of iron has been mainly determined from the com- position of ferric oxide, with some rather scanty data relative to other compounds. Most of the earlier data relative to the percentage of metal and oxygen in ferric oxide we may reject at once, as set aside by later investigations. Among this no longer valuable material there is a series of experiments by Berzelius, another by Dobereiner, and a third by Capitaine. The work done by Stromeyer and by Wackenroder was probably good, but I am unable to find its details. The former found 30.15 per cent, of oxygen in the oxide under consideration, while Wackenroder obtained figures ranging from a minimum of 30.01 to a maximum of 30.38 per cent.* In 1844 Berzelius f published two determinations of the ratio in ques- tion. He oxidized iron by means of nitric acid, and weighed the oxide thus formed. He thus found that when = 100 Fe 350.27 and 350.369. Hence the following percentages of Fe in Fe 2 3 : 70.018 70.022 Mean, 70.020, .0013 About the same time Svanberg and Norlin { published two elaborate series of experiments ; one relating to the synthesis of ferric oxide, the other to its reduction. In the first set pure piano-forte wire was oxidized by nitric acid, and the amount of oxide thus formed was determined. The results were as follows : 1.5257 grm. Fe gave 2.1803 grm. Fe 2 O s . 69.977 per cent. Fe. 2.4051 3-4390 " 69.936 it 2.3212 3-3 r 94 it 69.928 (( 2.32175 3.383 " , 69.968 a 2.2772 3.2550 (t 69.960 " 2.4782 3.5418 " 69.970 " 2.3582 3.3720 < < 69.935 " , Mean, 69.9534, .0050 * For additional details concerning these earlier papers I must refer to Oudemans' mono- graph, pp. 140, 141. t Ann. Chem. Pharm., 30, 432. Berz. Jahresb., 25, 43. I Berzelius' Jahresbericht, 25, 42. 288 THE ATOMIC WEIGHTS. Iii the second series ferric oxide was reduced by ignition in a current of hydrogen, yielding the subjoined percentages of metal : 2.98353 grm. Fe 2 O 3 gave 2.08915 grm. Fe. 70.025 per cent. 2.41515 i.6oro 70.015 " 299175 " 2.09455 " 70.014 " 3.5783 2.505925 70.030 " 4.1922 2.9375 70.072 " 3.1015 " 2.17275 " 70.056 " 2.6886 " 1.88305 " 70.036 " Mean, 70.0354, .0055 It is evident that one or both of these series must be vitiated by con- stant errors, and that these probably arise from impurities in the mate- rials employed. Impurities in the wire taken for the oxidation series could hardly have been altogether avoided, and in the reduction series it is possible that weighable traces of hydrogen may have been retained by the iron. At all events, it is probable that the errors of both series are in contrary directions, and therefore in some measure compensatory. In 1844 there was also published an important paper by Erdmann and Marchand.* These chemists prepared ferric oxide by the ignition of pure ferrous oxalate, and submitted it to reduction in a stream of hydrogen. Two sets of results were obtained with two different samples of ferrous oxalate, prepared by two different methods. For present pur- poses, however, it is not necessary to discuss these sets separately. The percentages of iron in Fe 2 3 are as follows : 70.013 ] 69.962 | 69.979 }-A. 70.030 I 69.977 J 70.044 1 70.015 j-B. 70.055 J Mean, 70.0094, =b .0080 In 1850 Maumene'sf results appeared. He dissolved pure iron wire in aqua regia, precipitated with ammonia, filtered off the precipitate, washed thoroughly, ignited, and weighed, after the usual methods of quantitative analysis. The percentages of Fe in Fe 2 3 are given in the third column : 1.482 grm. Fe gave 2.117 grm. Fe 2 O 3 . 70.005 per cent. 1.452 2.074 " 70.010 " 1.3585 " 1.941 " 69.990 1.420 " 2.0285 " 70.002 " 1.492 2.1315 " 69.998 " 1-554 " 2.22O " 7O.OOO " Mean, 70.0008, =h .0019 * Journ. fiir Prakt. Chem., 33, i. 1844. tCompt. Rend., Oct. 17, 1850. IRON. 289 Two more results, obtained by Rivot* through the reduction of ferric oxide in hydrogen, remain to be noticed. The percentages are : 69.31 69-35 Mean, 69.33, .013 We have thus before us six series of results, which we may now com- bine : Berzelius 70.020, .0013 Erdmann and Marchand 70.0094, =b .0080 Svanberg and Norlin, oxidation 69.9534, .0050 Svanberg and Norlin, reduction 70.0354, .0055 Maumene 70.0008, .0019 Rivot 69.33, - OI 3 General mean 70.0075, .0010 From this we get Fe = 55.596. Dumas' f results, obtained from the chlorides of iron, are of so little weight that they might safely be omitted from our present discussion. For the sake of completeness, however, they must be included. Pure ferrous chloride, ignited in a stream of hydrochloric acid gas, was dissolved in water and titrated with a silver solution in the usual way. One hundred parts of silver are equivalent to the amounts of Fed, given in the third column : 3.677 grm. FeCl. 2 = 6.238 grm. Ag. 58.945 3.924 " =6.675 " 58.787 Mean, 58.866, .053 Ferric chloride, titrated in the same way, gave these results : 1.179 g rm - FeCl 3 = 2.3475 grm. Ag. 50.224 1.242 " =2.471 " 5- 26 3 Mean, 60.2435, .0132 These give us two additional values for Fe, as follows : From FeC! 2 Fe = 55.742 From FeCl s " = 55.907 A series of determinations of the equivalent of iron, made by students by measuring the hydrogen evolved when the metal is dissolved in an acid, was published by Torrey in 1888. J The data have, of course, slight * Ann. Chem. Pharm., 78, 214. 1851. f Ann. Chem. Pharm., 113, 26. 1860. I Am. Chem. Journ., 10, 74. 19 290 THE ATOMIC WEIGHTS. value, but may be considered as being in some measure confirmatory. They are as follows : 56.40 55.6o 55-3* 55.5 6 55.48 55-5 55.86 56.06 56.22 55-So 55-78 55.6o 55.70 55-94 Mean, 55-777, .0532 These values undoubtedly depend on Regnault's value for the weight of hydrogen. Correcting by the later value, as found in the chapter of this work relating to the density ratio H : 0, the mean becomes Fe = 55.608, zh .0532. Here the probable error in the weight of the hydrogen is ignored, as being of no practical significance. The four ratios for iron are now as follows : (i.) Per cent. Fe in Fe 2 O 3 , 70.0075, .0010 (2.) Ag 2 : FeG 2 : : loo : 58.866, .0530 (3.) Ag 3 : FeC) 3 : : 100 : 50.2435, .0132 (4.) H:Fe:: I : 55.608,^.0532 Reducing these with O = 15.879, .0003 Ag = 107.108, .0031 Cl = 35.179, .0048 we have From (i) Fe = 55.596, .0023 From (2) " = 55.742, .1140 From (3) " = 55.907, .0450 From (4) " = 55.608, =b .0532 General mean Fe = 55.597, .0023 If O = 16, then Fe = 56.021. Here all the values are absorbed prac- tically by the first, the other three having no real significance. NICKEL AND COBALT. 291 NICKEL AND COBALT. On account of the close similarity of these metals to each other, their atomic weights, approximately if not actually identical, have received of late years much attention. The first determinations, and the only ones up to 1852, were made by Rothhoff,* each with but a single experiment. For nickel 188 parts of the monoxide were dissolved in hydrochloric acid ; the solution was evaporated to dryness, the residue was dissolved in water, and precipi- tated by silver nitrate. 718.2 parts of silver chloride were thus formed ; whence Ni = 58.613. The same process was applied also to cobalt, 269.2 parts of the oxide being found equivalent to 1029.9 of AgCl ; hence Co = 58.504. These values are so nearly equal that their differences were naturally ascribable to experimental errors. They are, however, entitled to no special weight at present, since it cannot be certain from any evi- dence recorded that the oxide of either metal was absolutely free from traces of the other. In 1852 Erdmann and Marchand f published some results, but with- out details, concerning the atomic weight of nickel. They reduced the oxide by heating in a current of hydrogen, and obtained values ranging from 58.2 to 58.6, when = 16. Their results were not very concordant, and the lowest was probably the best. In 1856, incidentally to other work, Deville J found that 100 parts of pure metallic nickel yielded 262 of sulphate ; whence Ni = 58.854. To none of the foregoing estimations can any importance now be at- tached. The modern discussion of the atomic weights under considera- tion began with the researches of Schneider in 1857. This chemist examined the oxalates of both metals, determining carbon by the com- bustion of the salts with copper oxide in a stream of dry air. The carbon dioxide thus formed was collected as usual in a potash bulb, which, in weighing, was counterpoised by a similar.bulb, so as to eliminate errors due to the hygroscopic character of the glass. The metal in each oxalate was estimated, first by ignition in a stream of dry air, followed by intense heating in hydrogen. Pure nickel or cobalt was left behind in good con- dition for weighing. Four analyses of each oxalate were made, with the results given below. The nickel salt contained three molecules of water, and the cobalt salt two molecules : * Cited by Berzelius. Poggend. Annaleti, 8, 184. 1826. t Journ. fiir Prakt. Chem., 55, 202. 1852. t Ann. Chim. Phys. (3), 46, 182. 1856. t Poggend. Annalen, 101, 387. 1857. 292 THE ATOMIC WEIGHTS. 1.1945 grm. gave .528 grm. CO 2 . 44.203 per cent. 2.5555 " 1.12625 " 44- 72 " 3.199 " 1.408 44.014 " 5.020 " 2.214 44.104 " Mean, 44.098, .027 The following percentages of nickel were found in this salt 29.107 29.082 29.066 29.082 Mean, 29.084, dz .006 rm. gave .781 grm. CO 2 . 47-753 P er cent. 1.107 " .5295 " 47-832 " 2.309 " i.ioi 47-683 " 3.007 1-435 47.722 " Mean, 47-7475, .0213 The following were the percentages found for cobalt : 32-552 32.619 32.528 32.523 Mean, 32.5555, .0149 In a later paper* Schneider also gives some results obtained with a nickel oxalate containing but two molecules of water. This gave him 47.605 per cent, of C0 2 , and the following percentages of nickel : 3 I -4"5 31-4038 Mean, 31.4076, d= .0026 The conclusion at which Schneider arrived was that the atomic weights of cobalt and nickel are not identical, being about 60 and 58 respectively. The percentages given above will be discussed at the end of this chapter in connection with all the other data relative to the constants in ques- tion. The next chemist to take up the discussion of these atomic weights was Marignac, in 1858.f He worked with the chlorides and sulphates *Poggend. Annalen, 107, 616. t Arch, des Sci. Phys et Nat. (nouv. serie), i, 372. 1858. NICKEL AND COBALT. 293 of nickel and cobalt, using various methods, but publishing few details, as he did not consider the determinations final. The sulphates, taken as anhydrous, were calcined to oxides. From the ratio NiS0 4 : NiO, he found Ni = 58.4 to 59.0, and from five measurements of the ratio CoS0 4 : Co, Co = 58.64 to 58.76. If oxygen is taken as 16, these give for the percentages of oxide in sulphate : CoO in CoSOv NiO in 48.267 48.187 48.307 48.387 Mean, 48.287, d= .0135 Mean, 48.287, .0675 The chlorides were dried at 100, but found to retain water; and in most cases were then either fused in a stream of chlorine or of dry, gaseous hydrochloric acid, or else calcined gently with ammonium chloride. The determinations were then made by titration with a standard solution of silver in nitric acid. Three experiments with an- hydrous CoCl, gave Co = 58.72 to 58.84. Three more with CoCl 2 dried at 100 gave Co = 58.84 to 59.02. Three with anhydrous NiCl 2 gave Ni = 58.80 to 59.00. If the calculations were made with Ag = 108 and Cl = 35.5, then these data give as proportional to 100 parts of silver : 60.093 60.185 Mean, 60.139, .0310 , Mean, 60.118, .0192 In one more experiment NiCl. 2 was precipitated with a known quan- tity of silver. The filtrate was calcined, yielding NiO ; hence the ratio ,Ag- 2 : NiO, giving Ni = 59.29. This experiment needs no farther atten- tion. In short, according to Marignac, and contrary to Schneider's views, the two atomic weights are approximately the same. Marignac criticises Schneider's earlier paper, holding that the nickel oxalate may have con- tained some free oxalic acid, and that the cobalt salt was possibly con- taminated with carbonate or with basic compounds. In his later papers Schneider rejects these suggestions as unfounded, and in turn criticises Marignac. The purity of anhydrous NiS0 4 is not easy to guarantee, and, according to Schneider, the anhydrous chlorides of cobalt and nickel are liable to be contaminated with oxides. This is the case even when the chlorides are heated in chlorine, unless the gas is carefully freed from all traces of air and moisture. 294 THE ATOMIC WEIGHTS. Dumas' * determinations of the two atomic weights were made with the chlorides of nickel and cobalt. The pure metals were dissolved in aqua regia, the solutions were repeatedly evaporated to dryness, and the residual chlorides were ignited in dry hydrochloric acid gas. The last two estimations in the nickel series were made upon NiCL 2 formed by heating the spongy metal in pure chlorine. In the third column I give the NiCl 2 or CoCl 2 equivalent to 100 parts of silver : .9123 grm. NiCl 2 = 1.515 grm. Ag. 60.218 2.295 " 3-8ii5 " 60.212 3.290 5.464 " 60.212 1.830 " 3.041 " 60.178 3.001 " 4.987 " 60.176 Mean, 60.1992, .0062 2 352 grm. CoCl 2 = 3.9035 grm. Ag. 60.254 4.210 6.990 " 60.229 3.592 " 5.960 " 60.268 2.492 " 4.1405 " 60.186 4.2295 " 7.0255 " 60.202 Mean, 60.2278, .on These results give values for Co and Ni differing by less than a tenth of a unit ; here, as elsewhere, the figure for Ni being a trifle the lower. Combining these data with Marignac's, we have Agi : NiC^ : : 100 : x. Marignac 60. 139, .0310 Dumas 60.199,^.0062 General mean 60. 194, db .0061 Ag^ : CoCl 2 : : TOO : X, Marignac 60.118, .0192 Dumas 60 228, .0110 General mean 60.200, dr .0095 In 1863 f the idea that nickel and cobalt have equal atomic weights was strengthened by the researches of Russell. He found that the black oxide of cobalt, by intense heating in an atmosphere of carbon dioxide, became converted into a brown monoxide of constant composition. The ordinary oxide of nickel, on the other hand, was shown to be convert- ible into a definite monoxide by simple heating over the blast lamp. The pure oxides of the two metals, thus obtained, were reduced by ignition in hydrogen, and their exact composition thus ascertained. *Ann. Chem. Pharm., 113, 25. 1860. f Journ. Chem. Soc. (2), i, 51. 1863. NICKEL AND COBALT. 295 Several samples of each oxide were taken, yielding the following data. The separate samples are indicated by lettering : Nickel c. D. B. CoO. 2. 1211 2.0241 I 2.1226 I L9947 {3.0628 2.1167 I.77I7 1.7852 1.6878 2.2076 | 2.6851 (2.1, 46I f 3.4038 E. J 2.2778 (2.1837 Ni. 1.6364 .6468 5838 7342 7952 .6761 .79" .6845 .9030 .7179 5788 1.6379 2.0873 Cobalt. Co. 1.6670 L5907 1.6673 1.5678 2.4078' .6638 .3924 .4030 .3264 735 2.1104 1.6868 2.6752 i.79 01 1.7163 Percent. Ni. 78.597 78.584 78.608 78.581 78.589 78.583 78.616 78.590 78.588 78.590 78.594 78.597 78.588 Mean, 78.593, .0018 Percent. Co. 78.591 78.588 78.550 78.598 78.614 78.603 78.591 78.591 78.588 78.592 78.597 78.598 78.595 78.589 78.596 Mean, 78.592, .0023 These percentages are practically identical, and lead to essentially the same mean value for each atomic weight. In a later paper Russell* confirmed the foregoing results by a different process. He dissolved metallic nickel and cobalt in hydrochloric acid and measured the hydrogen evolved. Thus the ratio between the metal and the ultimate standard was fixed without the intervention of any other element. About two-tenths of a gramme of metal, or less, was * Journ. Chem. Soc. (2), 7, 494. 1867. 296 THE ATOMIC WEIGHTS. taken in each experiment. The data obtained were as follows ; the last column giving the weight of hydrogen, computed from its volume, yielded by 100 parts of cobalt or nickel : Wt. Ni. f .0906 .1017 .1990 A. <{ .0997 .1891 .1859 .1838 B. - .1806 .2026 C. .1933 .1890 D. -j .1942 .1781 Nickel. Vol. H in cc. 153-62 172.32 337.o6 168.93 319.86 314.75 311-25 318.75 305.28 333-81 325.93 319.77 328.15 301.09 Cobalt. Vol. H in cc. 321.36 312.95 319-63 328.96 3 2 8.43 329.55 290.17 308.97 318.60 3H.73 3 5-4o Ratio. 3.420 3.418 3-4i6 3.417 3.412 3.415 3.4i6 3.398 3-409 3-404 3.401 3-412 3.408 3-410 Mean, 3.411, .001 Ratio. 3-395 3.398 3-397 3-398 3403 3-401 3-401 3-404 3.405 3.410 3.407 Mean, 3.4017, .0009 The weight of the hydrogen in these determinations was doubtless computed from Regnault's data concerning the density of that gas. Cor- recting by the new value for the weight of a litre of hydrogen, .089872 gramme, the ratios become: For nickel 3-42H, .0010 For cobalt 3.4112, =b .0009 Some time after the publication of Russell's first paper, but before the appearance of his second, some other investigations were made known. NICKEL AND COBALT. 297 Of these the first was by Sommaruga,* whose results, obtained by novel methods, closely confirmed those of Schneider and antagonized those of Dumas, Marignac, and Russell. The atomic weight of nickel Som- maruga deduced from analyses of the nickel potassium sulphate, K 2 Ni(S0 4 ) 2 .6H 2 0, which, dried at 100, has a perfectly definite compo- sition. In this salt the sulphuric acid was determined in the usual way as barium sulphate, a process to which there are obvious objections. In the third column are given the quantities of the nickel salt proportional to 100 parts of BaS0 4 : 0.9798 grm. gave 1.0462 grm. BaSO 4 . 93-653 1.0537 " 1.1251 " 93.654 1.0802 " LI535 " ' 93-645 1.1865 " 1.2669 " 93.654 3.2100 " 3.4277 " 93649 3.2124 " 3.43 3 " 93. 6 48 Mean, 93.6505, rt .001 For cobalt Sommaruga used the purpureocobalt chloride of Gibbs and Genth. This salt, dried at 110, is anhydrous and stable. Heated hotter, CoCl 2 remains. The latter, ignited in hydrogen, yields metallic cobalt. In every experiment the preliminary heating must be carried on cautiously until arnmoniacal fumes no longer appear : .6656 grm. gave .1588 grm. Co. 23.858 per cent. 1.0918 " .2600 " 23.814 " .9058 " .2160 " 23.846 L5895 " .3785 " 23.813 " 2.9167 " .6957 " 23.847 " 1.8390 .4378 " 23.806 " 2.5010 " .5968 " 23.808 Mean, 23.827, .006 Further along this series will be combined with a similar one by Lee. It may here be said that Sommaruga's paper was quickly followed by a critical essay from Schneider,f endorsing the former's work and object- ing to the results of Russell. In 1867 still another new process for the estimation of these atomic weights was put forward by Winkler, J who determined the amount of gold which pure metallic nickel and cobalt could precipitate from a neutral solution of sodio-auric chloride. In order to obtain pure cobalt Winkler prepared purpureocobalt chloride, which, having been four or five times recrystallized, was ignited in hydrogen. His nickel was repeatedly purified by precipitation with sodium hypochlorite. From material thus obtained pure nickel chloride * Sitzungsb. Wien. Akad., 54, 2 Abth., 50. 1866. 1 Poggend. Annalen,/i30, 310. 1 Zeit. Anal. Chem., 6, 18. 1867. 298 THE ATOMIC WEIGHTS. was prepared, which, after sublimation in dry chlorine, was also reduced by hydrogen. One hundred parts of gold are precipitated by the quanti- ties of nickel and cobalt given in the third columns respectively. In the cobalt series I include one experiment by Weselsky, which was published by him in a paper presently to be cited : .4360 grm. nickel precipitated .9648 grm. gold. 45.191 4367 .9666 " 45.179 5189 " I.I457 " 45-29I .6002 " 1.3286 " 45.175 Mean, 45.209, .019 .5890 grm. cobalt precipitated 1.3045 grm. gold. 45.151 3 '47 .6981 " 45.080 5829 1.2913 45- HI Sni 1.1312 " 45.182 .5821 1.2848 " 45.307 559 " 1.241 " 45.044 Weselsky. Mean, 45.151, .025 Weselsky 's paper,* already quoted, relates only to cobalt. He ignited the cobalticyanides of ammonium and of phenylammonium in hydrogen, and from the determinations of cobalt thus made deduced its atomic weight. His results are as follows : 7575 S rm - (NH 4 ) 6 Co a Cy 12 S ave - l66 S rm - Co - 21.914 per cent. 5 J 43 " .113 " 21.972 " Mean, 21.943, .029 .8529 grm. (C 6 H 8 N) 6 Co 2 Cy 12 gave .1010 grm. Co. 11.842 per cent. .6112 " .0723 " 11.829 " .7 J 4 .0850 " 11.905 " .9420 .1120 " 11.890 " Mean, 11.8665, .0124 Next in order is the work done by Lee f in the laboratory of Wolcott Gibbs. Like Weselsky, Lee ignited certain cobalticyanides and also nickelocyanides in hydrogen and determined the residual metal. The double cyanides chosen were those of strychnia and brucia, salts of very high molecular weight, in which the percentages of metal are relatively low. A series of experiments with purpureocobalt chloride was also carried out. In order to avoid admixture of carbon in the metallic resi- dues, the salts were first ignited in air, and then in oxygen. Reduction by hydrogen followed. The salts were in each case covered by a porous septum of earthenware, through which the hydrogen diffused, and which served to prevent the mechanical carrying away of solid particles ; fur- * Ber. d. Deutsch. Chem. Gesell., 2, 592. 1868. t Am. Journ. Sci. and Arts (3), 2, 44. 1871. NICKEL AND COBALT. 299 thermore, heat was applied from above. The results attained were very satisfactory, and assign to nickel and cobalt atomic weights varying from each other by about a unit ; Ni being nearly 58, and Co about 59, when O = 16. The exact figures will appear later. The cobalt results agree remarkably well with those of Weselsky. The following are the data obtained : Brucia nickelocyanide, Ni. A Cy Vi (^C^H^N^O^ & H 6 .10H 2 0. Salt. Ni. Percent. Ni. .3966 .0227 5.724 .5638 .0323 5.729 .4000 .0230 5-75 .3131 -01795 5-733 .4412 .0252 5.712 .4346 .0249 5.729 Mean, 5.7295, .0034 Strychnia nickelocyanide, Ni 9 Cy l2 ( C 2l H^N 2 2 \H 6 .8H 2 0. Salt. Ni. Per cent. AY. .5358 .0354 6.607 .5489 .0363 6.613 .3551 -0234 ' 6.589 4495 - 02 97 6 - 6 7 .2530 .0166 6.561 .1956 .0129 6.595 Mean, 6.595, .005 Brucia cobalticyanide, Co 2 Cy l2 ( C 2Z H 26 N 2 0^> 6 H 6 .20H 2 0. Salt. Co. Percent. Co. .4097 .0154 3.759 3951 .0147 3-720 \ 5456 .0204 3.739 .4402 .0165 3.748 .4644 .0174 3-747 .4027 .0151 3.749 Mean, 3.7437, .0036 Strychnia cobalticyanide, Co. 2 Cy l2 (C 2l H 22 N 2 0,\H 6 .8H. 2 0. Salt. Co. Percent. Co. .4255 -0195 4.583 .4025 .0185 4.596 3733 .0170 4-554 -4535 -0207 4.564 -2753 -0126 4.577 .1429 -0065 4.549 Mean, 4.5705, =b .005 300 THE ATOMIC WEIGHTS. Parpureo-cobalt chloride, C Salt. Co. Percent. Co. 9472 .2233 23.575 .8903 .2100 23.587 .6084 .1435 23.586 .6561 .1547 23.579 .6988 .1647 23.569 .7010 .1653 23.581 Mean, 23.5795, .0019 The last series may be combined with Sommaruga's, thus : Sommaruga 23.817, .006 Lee 23.5795, .0019 General mean 23.6045, .0018 Baubigny's * determinations of the atomic weight of nickel are limited to two experiments upon the calcination of nickel sulphate, and his data are as follows : 6.2605 grm. NiSO 4 gave 3.9225 NiO. 48.279 per cent. 4.4935 " 2.1695 " 48.281 Mean, 48.280 Zimmermann's work, published after his death by Krtiss and Alibe- goff,f was based, like Russell's, upon the reduction of cobalt and nickel oxides in hydrogen. The materials used were purified with great care, and the results were as follows: Nickel 1 mo. Ni, Percent. Ni. 6.0041 4.7179 78.578 6 4562 5-0734 78.582 8.5960 6.7552 78.585 4.7206 3.7096 78.583 8.2120 6.4536 78.587 9-1349 7.1787 78.585 IO.OI56 7.8702 78.579 4.6482 3.6526 78.580 8.9315 7.0184 78.580 10.7144 8.4196 78.582 3.0036 2.3602 78.579 Mean, 78.582, .0006 * Compt. Rend., 97, 951. 1883. f Ann. der Chem., 232, 324. 1886. NICKEL AND COBALT. 301 Cobalt. CoO. Co. Per cent. Co. 6.3947 5-0284 78.634 6.6763 5.2501 78.638 5.6668 4.45 60 78.633 2.9977 2.3573 78.637 8.7446 6.8763 78-635 3.2625 2.5655 78.636 6.3948 5.0282 78.630 8.2156 6.4606 78.638 9.4842 7.458o 78.636 9.9998 7.8630 78.632 Mean, 78.635, .0002 Shortly after the discovery of nickel carbonyl, NiC 4 O 4 , Mond, Langer, and Quincke*made use of it with reference to the atomic weight of nickel. The latter was purified by distillation as nickel carbonyl, then converted into oxide, and that was reduced by hydrogen in the usual way. NiO. Ni. Per cent. Ni. .2414 .1896 78.542 .3186 .2503 78.562 .3391 .2663 78.531 Mean, 78.545, .0061 Schutzenberger's experiments,t published in 1892, were also few in number. First, nickel sulphate, dehydrated at 440, was calcined to oxide. 3.505 grm. NiSO 4 gave 1.690 NiO. 48.217 per cent. 26008 " 1.2561 " 48.297 " Mean, 48.257, .027 Second, nickel oxide was reduced in hydrogen, as follows : 1.6865 grm. NiO gave 1.3245 Ni. 78.535 per cent. 1.2527 " .9838 " 78.533 " Mean, 78.534 Iii one experiment with cobalt oxide, 3.491 grm. gave 2.757 Co, or 78.975 per cent. In view of the many determinations of this ratio by other observers, this single estimation may be neglected. The experi- ments on nickel sulphate, however, should be combined with those of Marignac and Baubigny, giving the latter equal weight with Schutzen- berger's, thus : *Journ. Chem. Soc., 57, 753. 1890. tConipt. Rend., 114, 1149. 1892. 302 THE ATOMIC WEIGHTS. Marignac 48.287, .0675 Baubigny 48.280, .027 Schutzenberger 48.257, .027 General mean. ..... 48.269, .018 From this point on the determination of these atomic weights is com- plicated by the questions raised by Kriiss as to the truly elementary character of nickel and cobalt. If that which has been called nickel really contains an admixture of some other hitherto unknown element, then all the determinations made so far are worthless, and the investiga- tions now to be considered bear directly upon that question. First in order comes Remmler's research upon cobalt.* This chemist, asking whether cobalt is homogeneous, prepared cobaltic hydroxide in large quantity, and made a series of successive ammoniacal extracts from it, twenty-five in all. Each extract represented a fraction, from which, by a long series of operations, cobalt monoxide was prepared, and the latter was reduced in hydrogen after the manner of Russell. The actual deter- minations began with the second fraction, and the data are subjoined, the number of the fraction being given with each experiment : CoO. Co. Percent. Co. 2 09938 .07837 78.859 3 i52i .11814 78.650 4 .22062 .17360 78.687 5 390H .30681 78.647 6 .28820 .22661 78.629 7 3434 .26968 78.615 8 43703 .34321 78.532 9 9H77 .71864 78.560 10 63256 .49661 78.508 ii 32728 .25701 78.529 12 .38042 .29899 78.595 13 16580 .13027 78.571 14 I.OI6O7 79873 78.610 15 I-3I63S 1-03545 78.661 16 91945 .72315 78.650 17 53 IQ o .41773 78.668 , 18 82381 .64728 78.572 19 81139 .63754 78.574 20 76698 .60292 78.610 21 LI3693 .89412 78.643 22 2.OO259 1-57495 78.646 23 1.04629 .82185 78.549 24 48954 .38466 78.576 25 69152 .54326 78.560 Mean, 78.613, .0099 *Zeit. Anorg. Chem., 2, 221. Also more fully in an Inaugural Dissertation, E)rlangen, NICKEL AND COBALT. 303 Considered with reference to the purpose of the investigation, this mean and its probable error have no real significance. But it is very close to the means of other experimenters, and a study of the variations represented by the several fractions seems to indicate fortuity rather than system. Remmler regards his results as indicating lack of homo- geneity in his material ; but it seems more probable that such differences as exist are due to experimental errors and to impurities acquired in the long process of purification to which each fraction was submitted, rather than to any uncertainty regarding the nature of cobalt itself. For either interpretation the data are inconclusive, and I therefore feel justified in treating the mean like other means, and in combining it finally with them. From the same point of view that is, with reference to the supposed heterogeneity of nickel Kruss and Schmidt * carried out a series of frac- tionations of the metal by distillation in a stream of carbon monoxide. Nickel oxide, free from obnoxious impurities, was first reduced. to metal by heating in hydrogen, after which the current of carbon monoxide was allowed to flow. The latter, carrying its small charge of nickel tetra- carbonyl was then passed through a Winkler's absorption apparatus con- taining pure aqua regia, from which, by evaporation, nickel chloride was obtained, and from that, by reduction in hydrogen, the nickel. Ten such fractions were successively prepared and studied ; first, by prepa- ration of NiO and its reduction in hydrogen ; and, secondly, in some cases, by the reoxidation of the reduced metal, so as to give a synthetic value for the ratio Ni : 0. The data obtained are as follows, the successive fractions being numbered : Reduction of NiO. NiO. Ni. Per cent. Ni. J 1 .3722 .2926 78.614 ' 1 .7471 .5870 78.571 2 { .7659 I .7606 .60085 .5961 78.450 78.372 0.0175 .7984 78.467 3. -j 1.2631 .99065 78.430 (1.2582 .9868 78.429 4- -! ' 5I93 .4076 78.490 \ .9200 .7215 78.424 f -4052 .3179 78.455 '* 1 . 6 5 J 8 .5111 78.414 6 I * 5623 4399 78.232 ' 1 .5556 4350 78.294 ( -9831 .7724 78.568 7- -j .9765 .7646 78.300 (. -9639 7557 78.400 *Zeit. Anorg. Chem., 2, 235. 1892. 304 THE ATOMIC WEIGHTS. 2. 3- 5- Ni. .5870 6011 .7988 9913 .9868 .4093 .7216 394 6. .4415 .4350 7752 7- 1 .7667 .7558 4555 445 6 .44415 4423 2508 2467 4538 .4451 .4438 .4272 .2491 .2467 .3904 .3891 Oxidation qf Ni. NiO. 7471 7659 .7606 1.0175 1.2631 1.2582 .5193 .9200 .4052 .6518 5 62 3 555 6 .9831 1 10. ( .3918 1.3891 .9639 5756 .56765 .5663 .5642 .3174 .3H8 .4976 .4961 78.839 78.411 78.368 78.400 78.481 78.367 78.457 78.432 Mean, 78.444, =h .0166 Per cent. Ni. 78.571 78.372 78.359 78.506 78.482 78.429 78.818 78.435 78.825 78.414 78.517 78.294 78.853 78.515 78.411 79-135 78.499 78.43 78.394 79-015 78.367 78.738 78.432 Mean, 78.557, .0319 To these data of Kriiss and Schmidt the remarks already made con- cerning Remmler's work seem also to apply. The variations appear to be fortuitous, and not systematic, although the authors seem to think that they indicate a compositeness in that substance which has been hitherto regarded as elementary nickel. There is doubtless something to be said on both sides of the question ; but if Kriiss and Schmidt are right, all previous atomic weight determinations for cobalt and nickel are invalidated. In view of all the evidence, therefore, I prefer to regard their varying estimations as affected by accidental errors, and to treat their means like others. On this basis, their work combines with previ- NICKEL AND COBALT. 305 ous work as follows, Schulzenberger's measurements of the ratio NiO : Ni being assigned equal weight with those of Mond, Langer, and Quincke : Russell 78.593, .0018 Zimmermann 78.582, .0006 Mond, Langer, and Quincke 78.545, .0061 Schutzenberger 78.534, .0061 Kriiss and Schmidt, reduction series 78.444, .0166 Kriiss and Schmidt, oxidation series 78.557, zh .0319 General mean 78-57, db .0006 In 1889 Winkler * published a short paper concerning the gold method for determining the atomic weights in question, but gave in it no actual measurements. In 1893 f he returned to the problem with a new line of attack, and at the same time he takes occasion to criticise Kriiss and Schmidt somewhat severely. He utterly rejects the notion that either nickel or cobalt contain any hitherto unknown element, and ascribes the peculiar results obtained by Kriiss and Schmidt to impurities derived from the glass apparatus used in their experiments. For his own part he now works with pure nickel and cobalt precipitated electrolytically upon platinum, and avoids the use of glass or porcelain vessels so far as possible. With material thus obtained he operates by two distinct but closely related methods, both starting with the metal, nickel or cobalt, converting it next into neutral chloride, and then measuring the chloride gravimetrically in one process, volumetrically in the other. After precipitation in a platinum dish, the nickel or cobalt is washed with water, rinsed with alcohol and ether, and then weighed. It is next dissolved in pure hydrochloric acid, properly diluted, and by evapora- tion to dryness and long heating to 150 converted into anhydrous chlo- ride. The nickel chloride thus obtained dissolves perfectly in water, but the cobalt salt always gave a slight residue in which the metal was electrolytically determined and allowed for. In the redissolved chloride, by precipitation with silver nitrate, silver chloride is obtained, giving a direct ratio between that compound and the nickel or cobalt originally taken. The gravimetric data are as follows, with the metal equivalent to 100 parts of silver chloride given in a final column : Nickel Ni. Aga. Ratio. .3011 1.4621 20.594 .2242 1.0081 20.605 .5166 2.5108 20.570 .4879 2.3679 20.605 3827 1.8577 20.601 3603 75i7 20.568 Mean, 20.590, .0049 * Ber. Deutsch. Chem. Gesell., 22, 891. fZeit. Anorg. Chem., 4, 10. 1893. 20 306 THE ATOMIC WEIGHTS. Co. .3458 3776 4493 .4488 .2856 .2648 Cobalt. AgCl. Ratio. 1.6596 20.836 1.8105 20.856 2.1521 20.877 2.1520 20.855 1.3683 20.873 1.2768 20.886 Mean, 20.864, .0050 In the volumetric determinations the neutral chloride, prepared as before, was decomposed by means of a slight excess of potassium car- bonate, and in the potassium chloride solution, after removal of the nickel or cobalt, the chlorine was measured by titration by Volhard's method with a standard solution of silver. The amount of silver thus used was comparable with the metal taken. Nickel. Ni. .1812 .1662 .2129 .2232 .5082 1453 Co. .177804 .263538 .245124 .190476 .266706 263538 Af. .6621260 .6079206 7775252 .8162108 .8556645 . 53 r 5 4<> Cobalt. .6418284 .9514642 .8855780 .6866321 .9629146 .9503558 Ratio. 27.366 27.339 27.382 27.346 27.386 27.338 Mean, 27.359, - OO 59 Ratio. 27.702 27.699 27.679 27.741 27.696 27-73 1 Mean, 27.708, .0064 In view of the possibility that the cobalt chloride of the foregoing ex- periments might contain traces of basic salt; Winkler, in a supplement- ary investigation,* checked them by another process. To the electrolytic cobalt, in a platinum dish, he added a quantity of neutral silver sulphate and then water. The cobalt gradually went into solution, and metallic silver was precipitated. The weights were as follows : Co. 2549 .4069 Ag. .9187 1.4691 * Zeit. Anorg. Chem., 4, 462. 1893. NICKEL AND COBALT. 307 On examination of the silver it was found that traces of cobalt were retained less than 0.5 mg. in the first determination and less than 0.2 mg. in the second. Taking these amounts as corrections, the two experi- ments give for the ratios Ag. 2 : Co : : 100 : x the subjoined values : 27.706 27.687 These figures confirm those previously found, and as they fall within the limits of the preceding series, they may fairly be included in it, when all eight values give a mean of 27.705, .0050. Still another method, radically different from all of the foregoing pro- cesses, was adopted by Winkler in 1894.* The metals were thrown down electrolytically upon platinum, and so weighed. Then they were treated with a known excess of a decinormal solution of iodine in potassium iodide, which redissolved them as iodides. The excess of free iodine was then determined by titration with sodium thiosulphate, and in that way the direct ratio between metal and haloid was ascertained. The results were as follows, with the metal proportional to 100 parts of iodine given in the third column : Cobalt. WL Co. Wt. I. 2.128837 2.166750 First series \ .5290 2.254335 2.908399 2.861617 2.209694 Second series.. ^ .5267 2.246037 2.268736 Mean, 23.462, .0027 Nickel. Wt. Ni. Wt. I. Ratio. .5144 2.217494 23.251 .4983 2.148502 23.246 First series.. .. ^ .5265 2.268742 23.260 .6889 2.970709 23.243 .6876 2.965918 23.237 f.5120 2.205627 23.267 Second series. . 1 .5200 2.240107 23.267 (.5246 2.259925 23.267 Mean, 23.255, .0091 In these experiments, as well as in some previous' series, a possible source of error is to be considered in the occlusion of hydrogen by the * Zeitsch. Anorg. Chem., 8, i. 1894. 308 THE ATOMIC WEIGHTS. metals. Accordingly, in a supplementary paper, Winkler* gives the results of some check experiments made with iron, which, however, was not absolutely pure. The conclusion is that the error, if existent, must be very small. In 1895 Hempel and Thiele's work on cobalt appeared. f First, cobalt oxide, prepared from carefully purified materials, was reduced in hydro- gen. The weights of metal and oxygen are subjoined, with the percent- age of cobalt in the oxide deduced from them : Co. O. Percentage. .90068 .24429 78.664 .79159 .21445 78.686 1.31558 .357i6 78.648 Mean, 78.666, .0074 This mean combines with former means as follows : Russell ........... . ..................... 78.592, d= .0023 Zimmermann ............................ 78-635, .0002 Retnmler ............................ .*. , 78.613, .0099 Hempel and Thiele ...................... 78.666, .0074 General mean 78.633, .0002 In their next series of experiments, excluding a rejected series, Hempel and Thiele weighed cobalt, converted it into anhydrous chloride, and noted the gain in weight. In four of the experiments the chloride was afterwards dissolved, precipitated with silver nitrate, and then the silver chloride was weighed. The data are as follows : Co. Cl Taken Up. AgCl. .7010 - 8 453 3138 -3793 .2949 .3562 1.4340 .4691 .5657 2.2812 .5818 .7026 2.8303 .5763 .6947 .5096 .6142 2.4813 From these weights we get two ratios, thus : C7 2 : Co : 100 : X, 2AgCl : Co : : IOO : x. 82.929 20.565 82.731 20.564 82.791 20.556 82.924 20.538 82.807 82.957 Mean, 20.556, .0043 82.970 Mean, 82.873, .0241 * Zeitsch. Anorg. Chem., 8, 291. 1895. fZeitsch. Aiiorg. Chem., n, 73. NICKEL AND COBALT. 309 The second of these ratios was also studied by Winkleiyand the two series combine as follows : Winkler 20.864, =b .0050 Hempel and Thiele.^. 20.556, rb .0043 General mean 20.687, =b - OO 33 Hempel and Thiele apply to it a correction for silver chloride retained in solution, but its amount is small and not altogether certain. For present purposes the correction may be neglected. For the atomic weight of nickel we now have ratios as follows : (I.) Per cent, of Ni in NiC,O 4 .3H 2 O, 29.084, .006 (2.) Per cent, of CO 2 from NiC 2 O 4 .2H 2 O, 44.098, rb .027 (3.) Per cent, of Ni in NiC 2 O 4 .2H 2 O, 31.408, .0026 (4.) Per cent, of CO 2 from NiC 2 O 4 .2H 2 O, 47.605, =h .053 (5.) Per cent, of Ni in brucia nickelocyanide, 5.7295, .0034 (6.) Per cent, of Ni in strychnia nickelocyanide, 6.595, =fc .005 (7.) Per cent, of NiO in NiSO 4 , 48.269, rb .018 (8.) Per cent, of Ni in NiO, 78.570, .0006 (9.) Ag 2 : NiCl 2 : : 100 : 60.194, rb .0061 (10.) 2AgCl : Ni : : 100 : 20.590, rb .0049 (n.) Ag 2 : Ni : : 100 : 27.359, $9 (12.) Au 2 : Ni 3 : : 100 : 45.209, .019 (13.) BaSO 4 : K 2 Ni(SO 4 ) 2 .6H 2 O : : 100 : 93.6505. .001 (14.) Ni : H 2 : : 100 : 3.4211, .001 (15.) I 2 : Ni : : IOO : 23.255, .0091 To the reduction of these ratios the following atomic and molecular weights are applicable : O = 15.879, db .0003 I = 125.888, rb .0069 C = 11.920, rb .0004 K 38.817,1^.0051 N = 13.935, rb .0021 Ba = 136.392, .oo86 S = 31.828, rb. 0035 Au = 195.743, rb .0049 Ag = 107.108, rb .0031 AgCl= 142.287, rb .0037 Cl = 35- T 79, .0048 Since the proportion of water in the oxalates is not an absolutely cer- tain quantity, the data concerning them can be best handled by employ- ing the ratios between carbon dioxide and the metal. Accordingly, ratios (1) and (2) give a single value for Ni, and ratios (3) and (4) another. In all, there are thirteen values for the atomic weight in question : From (i) and (2) Ni = 57.614, rb .0372 From (5) " =57.625, rb. 0343 From (3) and (4) " = 57.635, rb .0644 From (6) " = 57.687, rb .0439 From (8) " = 58.218, rb .0020 From (7) " 58.268, .0428 From (13) " = 58.448, rb .0206 310 THE ATOMIC WEIGHTS. From (14) Ni 58.456, .0316 From (15) , " = 58.551, rb .0231 From (9) " = 58.587, =b .0179 From (10) . . . " = 58.594, .0141 From (u)... . "=58.607,^.0128 From (12.) " = 58.994, zb .0248 General mean Ni = 58.243, .0019 If = 16, this becomes Ni = 58.687. It is quite evident here that ratio (8), which includes the marvelously concordant determinations of Zimmermann, far outweighs all the other data. Whether so excessive a weight can justifiably be assigned to one set of measurements is questionable, but the general mean thus reached is not far from midway between the highest and lowest of the values, and hence it may fairly be entitled to provisional acceptance. No one of the individual values rests upon absolutely conclusive evidence, so that no one can be arbitrarily chosen to the exclusion of the others. Further investigation is evidently necessary. For cobalt we have sixteen ratios, as follows : (i.) Per cent, of Co in CoC 2 O 4 .2H 2 O, 32.5555, .0149 (2.) Per cent, of CO 2 from CoC 2 4 .2H 2 O, 47-7475, =b .0213 (3.) Per cent, of Co in CoO, 78.633, .0002 (4.) Per cent, of Co in purpureocobalt chloride, 23.6045, .0018 (5.) Per cent, of Co in phenylammonium cobalticyanide, 11.8665, .0124 (6.) Per cent, of Co in ammonium cobalticyanide, 21.943, .029 (7.) Per cent, of Co in brucia cobalticyanide, 3.7437, zb .0036 (8.) Per cent, of Co in strychnia cobalticyanide, 4.5705, zb .005 (9.) Per cent, of CoO in CoSO 4 , 48.287, .0135 (10.) Ag 2 : CoCl 2 : : 100 : 60.200, .0095 (n.) 2AgCl : Co : : 100 : 20.687, zb .0033 (12.) Ag 2 : Co : : 100 : 27.705, .0050 (13.) Au 2 : Co 3 : : 100 : 45.151, .025 (14.) Co : H 2 : : 100 : 3.4110, .0009 (15.) T 2 : Co : : 100 : 23.462, .0027 (16.) C1 2 : Co : : 100 : 82.873, .0241 From these, using the atomic weights already cited under nickel, and combining ratios (1) and (2), we get From (16) Co = 58.308, zb .0187 From (9) " =. 58.321, .0288 From (3) " = 58.437, .0014 From (i o) "= 58.600, .0228 From (14) " = 58.630, .0286 From (5) " = 58.639, db .0619 From (8) " = 58.696, =b .0642 From (6) " 58.736, .0808 From (4) " =58.774, .0071 From (7) " = 58.791, .0566 RUTHENIUM. 311 From (i i) Co = 58.870, .0094 From (13) " 58.920, .0327 From (15) " = 59. 7 2 , .0075 From (12) " = 59.349, .0108 From (i) and (2) " = 59-5 62 , .0382 General mean , Co = 58.487, .0013 If = 16, this becomes Co = 58.932. Here again the oxide ratio, because of Zimmermann's work, receives excessive and undue weight. The arithmetical mean of the fifteen values is Co = 58.781. Between this and the weighted general mean the truth probably lies, but the evidence is incomplete, and more determinations are needed. RUTHENIUM. The atomic weight of this metal has been determined by Claus and by Joly. Although Claus* employed several methods, we need only consider his analyses of potassium rutheniochloride, K 2 RuCl 5 . The salt was dried by heating to 200 in chlorine gas, but even then retained a trace of water. The percentage results of the analyses are as follows^ Ru. 2KCI. C/ 3 . 28.96 40.80 30.24 28.48 41.39 30.22 28.91 41.08 30.04 Mean, 28.78 41.09 30.17 Reckoning directly from the percentages, we get the following dis- cordant values for Ru : From percentage of metal Ru = 102.45 l From percentage of KC1 " = 106.778 From percentage of C1 3 " = 96.269 These results are obviously of little importance, especially since the best of them is not in accord with the position of ruthenium in the periodic system. The work of Joly is more satisfactory. f Several com- pounds of ruthenium were analyzed by reduction in a stream of hy- drogen with the following results : * Journ. fur Prakt. Chem., 34, 435. 1845. fCompt. Rend., 108, 946. 312 THE ATOMIC WEIGHTS. First, reduction of Ru0 2 : Ru. Per cent. Ru. 2.1387 1.6267 76.060 2.5846 1.9658 76.058 2.3682 i. 8016 76.075 2.8849 2 - J 939 76.046 Mean, 76.060, rb .0040 Second, reduction of the salt RuCl 3 .NO.H 2 : Per cent. Ru. 39-78 39.66 Mean, 39.72, . 0405 Third, reduction of RuCl 3 .N0.2NH 4 Cl : Per cent. Ru. 29.44 29.47 Mean, 29.455, .0101 Computing with = 15.879, .0003 ; N = 13.935, .0021, and Cl = 35.179, =h .0048, these data give three values for ruthenium, as follows: 1. From RuO 2 Ru = 100.922, .0178 2. From RuCl 3 .NO.H 2 O " = 100.967, . 1 102 3. From RuCl 3 .NO.2AmCl " = 100.868, .0387 General mean Ru = 100.913, .0160 If = 16, Ru=101.682. RHODIUM. 313 RHODIUM. Berzelius * determined the atomic weight of this metal by the analysis of sodium and potassium rhodiochlorides, Na 3 RhCl 6 , and K 2 RhCl 5 . The latter salt was dried by heating in chlorine. The compounds were ana- lyzed by reduction in. hydrogen, after the usual manner. Reduced to percentages, the analyses are as follows : In Na,RhCl 6 . Rh. 3 NaCl. .0026 General mean 50.388, dr .00062 For palladium, ignoring the work of Quintus Icilius, the subjoined ratios are now available : (i.) 2KC1 : Pd : : 100 : 70.865, dr .0150 (2.) Per cent. Pd in Pd(NH 3 Cl),, 50.388, dr .00062 (3.) 2AgCl : Pd(NH 3 Cl) 2 : : 100 : 73.807, dr .0742 (4.) Pd(NH 3 Cl) 2 : PdS : : 100 : 65.651, dr .0051 The antecedent data are Cl = 35.179, i .0048 S = 3 1." 828, +3 .0015 K = 38.817, .0051 AgCl = 142.287, .0037 N = 13.935, dr .C02I 21 THE ATOMIC WEIGHTS. Hence, for the atomic weight of palladium, we have From (i) Pd = 104.874, it .0243 From (2) " 105.858, .0200 From (3) " = 105.808, .2117 From (4) " = 106.550, =b .0491 General mean I'd 105.556, .0147 With O = 16, Pd = 106.364. Taking the values separately, the second is probably the best ; but in view of the work done by Bailey and Lamb on one side, and by Keller and Smith on the other, it cannot be accepted unreservedly. Until the cause of variation in the results is clearly determined, it is better to take the general mean of all the data, as given above. OSMIUM. The atomic weight of this metal has been determined by Berzelius, by Fremy, and by Seubert. Berzelius * analyzed potassium osmichloride, igniting it in hydrogen like the corresponding platinum salt. 1.3165 grammes lost .3805 of chlorine, and the residue consisted of .401 grm. of potassium chloride, with .535 grm. of osmium. Calculating only from the ratio between the Os and the KC1, the data give Os = 197.523. Fremy's determination f is based upon the composition of osmium tetroxide. No details as to weighings or methods are given ; barely the final result is stated. This, if = 16, is Os = 199.648. When the periodic law came into general acceptance, it became clearly evident that both of the foregoing values for osmium must be several units too high. A redetermination was therefore undertaken by Seubert,J who adopted methods based upon that of Berzelius. First, ammonium osmichloride was reduced by heating in a stream of hydrogen. The residual osmium was weighed, and the ammonium chloride and hydro- chloric acid given off were collected in a suitable apparatus, so that the total chlorine could be estimated as silver chloride. The weights were as follows : Am 2 OsCl B . Os. 6AgCl. 1.8403 7996 3.5897 2.0764 .9029 4.0460 2.1501 .9344 . 4.195 2.1345 .9275 4.1614 *Poggend. Annalen, 13, 530. 1828. fCompt. Rend., 19, 468. Journ. fiir Prakt. Chem., 31, 410. 1844. J Bericnte Deutsch. Chem. Gesell., 21, 1839. l888 - OSMIUM. 323 Hence we have for the percentage of osmium and for the osmichloride proportional to 100 parts of AgCl Per cent. Os. AgCl : Salt. 43.446 51.266 43.484 $1.32 43-458 51-254 43-453 5L293 Mean, 51.283, .0099 In a later paper * two more reductions are given, in which only osmium was estimated. Sail. Os. Percent. Os. 2.6687 1.1597 43.45 6 2.6937 1.1706 43-457 These determinations, included with the previous four as one series, give a mean percentage of Os in Am 2 OsCl 6 of 43.459, .0036. Secondly, potassium osmichloride was treated in the same way, but the residue weighed consisted of Os + 2KC1. From this the potassium chloride was dissolved out, recovered by evaporating the solution, and weighed separately. The volatile portion, 4HC1, was also measured by precipitation as silver chloride. In Seubert's first paper these data are given : Os. 2KCI. 4AgCl. 2.5148 ..... .7796 2.9837 2.1138 .8405 .6547 2.5076 Hence, with salt proportional to 100 parts of AgCl in the last column we have Per cent. Os. Per cent. KCl. AgCl : Salt. ...... 31.000 84.091 39.762 3 .973 84.102 Mean, 84.097, .0030 In his second paper Seubert gives fuller data relative to the potassium osmichloride, but treats it somewhat differently. The salt was reduced by a stream of hydrogen as before, but after that the boat containing the Os -{- 2KC1 was transferred to a platinum tube, in which, by prolonged heating in the gas, the potassium chloride was completely volatilized. The determinations of 4C1 as 4 AgCl were omitte \. Two series of data are given, as follows : *Ann. d. Chem., 261, 258. 324 THE ATOMIC WEIGHTS. Os. Percent. Os. 1.1863 .4691 39-543 .9279 -3667 39-5*9 1.0946 .433 39-558 1.6055 .6351 39.558 4495 .1778 39-555 .8646 .3417 39.521 .7024 .2781 39-593 1.2742 .504! 39-562 1.0466 -4H 1 39.566 Mean, 39.553, rb .0052 KfisClv 2KCL Percent. KCl. 2.2032 .6820 3O.955 2.0394 .6312 30.950 2.7596 .8544 30.961 2.4934 .77io 30.922 2.8606 .8843 30.913 2.8668 .5768 30.898 1.2227 .3778 30899 Mean, 30.931 t/ 3 '- C '130.973 , 31.000 Earlier set. ' J Mean of all nine determinations, 30.941, dr .0079 The single percentage of osmium in the earlier memoir is obviously to be rejected. The ratios to examine are now as follows : (i.) Per cent. Os in Am 2 OsC) 6 , 43.459, dr .0036 (2.) 6AgCl : Am 2 OsCl 6 : : loo : 51.283, dr .0099 (3.) 4AgCl : K 2 OsCl 6 : : IOO : 84.097, .0030 (4.) Per cent. Os in K 2 OsCl 6 , 39.553, dr .0052 (5.) Per cent. KCl in K 2 OsCI 6 , 30.951, dr .0079 To reduce these ratios we have Cl = 35.179, db .0048 KCl = 74.025, .0019 K =38.817, rb .0051 AgCl= 142.287, rb .0037 N = 13.935, .0021 Hence there are five independent values for osmium, as follows : From (i) Os = 190.111, rb .0300 From (2) " = 190.870, .0901 From (3) " = 189.928, =b .0371 From (4) " = 188.914, =b .0243 From (5) " = 189.571, =b .0928 General mean Os == 189.546, .0163 If = 16, Os = 190.990. IRIDIUM. 325 These figures serve to fix the place of osmium below iridium in the periodic classification of the elements, but are not concordant enough to be fully satisfactory. More determinations are evidently needed. IRIDIUM. The only early determination of the atomic weight of iridium was made by Berzelius,* who analyzed potassium iridichloride by the same method employed with the platinum and the osmium salts. The result found from a single analysis was not far from Ir = 196.7. This is now known to be too high. I have not, therefore, thought it worth while to recalculate Berzelius' figures, but give his estimation as it is stated in Roscoe and Schorlemmer's " Treatise on Chemistry." In 1878 the matter was taken up by Seubert,f who had at his disposal 150 grammes of pure iridium. From this he prepared the iridichlorides of ammonium and potassium (NH 4 ) 2 IrCl 6 and K 2 IrCl 6 , which salts were made the basis of his determinations. The potassium salt was dried by gentle heating in a stream of dry chlorine. Upon ignition of the ammonium salt in hydrogen, metallic iridium was left behind in white coherent Iamina3. The results obtained were as follows : Ir. Per cent. Jr. 1-3164 .5755 43725 1.7122 .7490 43-745 1.2657 .5536 43-739 1.3676 .5980 43.726 2.6496 1.1586 43-739 2.8576 1.2489 43-705 2.9088 1.2724 43-74 2 Mean, 43-732, .0035 The potassium salt was also analyzed by decomposition in hydrogen with special precautions. In the residue the iridium and the potassium chloride were separated after the usual method, and both were estimated. Eight analyses gave the following weights : KJrCl* C/ 4 , Loss. Ir. KCl. 1.6316 .4779 .6507 5030 2.2544 .6600 .8993 6953 2.1290 .6238 .8488 .6560 1.8632 5457 743 .5745 2.6898 .7878 1.0726 .8291 2-3719 .6952 9459 .7308 2.6092 .7641 1.0406 .8040 2.5249 7395 1.0070 7775 * Poggend. Annalen, 13, 435. 1828. fBer. Deutsch. Chem. Gesell., n, 1767. 1878. 326 THE ATOMIC WEIGHTS. Hence we have the following percentages, reckoned on the original salt: Ir. 2 KCL Cl,. 39.881 30.829 29.290 39.890 30. 842 29.277 39.868 30-813 29.300 39.876 30-835 29.289 39.877 30-825 29.287 39.879 3.8n 29.310 39.882 30.814 29.285 39.883 30.792 29.288 Mean, 39.880, =fc .0015 Mean, 30.820, .0037 Mean, 29.291, =b .0024 Joly * studied derivatives of iridium trichloride. The salts were dried at 120, and reduced in hydrogen. With IrCl 3 .3KC1.3H 2 he found as follows : Salt. Ir. KCl. 1.5950 .5881 .6803 1.6386 -6037 .7000 2.6276 .9689 1.1231 These data, if the weight of the salt itself is considered, give discordant results, but the ratio Ir : 3KC1 : : 100 : x is satisfactory. The values of x are as follows : 115.677 115.952 Mean, 115.848, .0583 The ammonium salt, IrCl 3 .3NH 4 Cl, gave the subjoined data : Wt. of Salt. Wt. of Ir. Per cent. Ir. 1.5772 .6627 42.017 1.6056 .6742 41.990 Mean, 42.003, .0094 To sum up, the ratios available for iridium are these : (i.) Per cent. Ir in Am 2 IrC) 6 , 43.732, .0035 (2.) Per cent. Ir in K 2 IrCl 6 , 39.880, .0015 (3.) Per cent. KCl in K 2 IrC) 6 , 30.820, .0037 (4.) Per cent. C1 4 in K 2 IrCl 6 , 29.291, =b .0024 (5.) Per cent. Ir in Am 3 IrCl 6 , 42.003, .0094 (6.) Ir : 3KC1 : : 100 : 115.848, .0583 The data for computation are O == 15.879, i .0003 N = 13.935, .o 21 Cl = 35.179, .0048 KG] = 74.025, .0019 K =. 38.817, .0051 H = i *Compt. Rend., no, 1131. 1890. PLATINUM. 327 And the six independent values for the atomic weight of iridium be- come From (i) Ir = 191.935, .0300 From (2) " = 191.511, .0221 From (3) " = 191.604, .0485 From (4) " = 191.641, .0622 From (5) " = 191-833, .0641 From (6) , " = 191.695, .0966 General mean Ir = 191.664, .0154 If 0=16, Ir= 193.125. PLATINUM. The earliest work upon the atomic weight of this metal was done by Berzelius,* who reduced platinous chloride and found it to contain 73.3 per cent, of platinum. Hence Pt = 193.155. In a later investigation f he studied potassium chloroplatinate, K 2 PtCl 6 . 6.981 parts of this salt, ignited in hydrogen, lost 2.024 of chlorine. The residue consisted of 2.822 platinum and 2.135 potassium chloride. From these data we may calculate the atomic weight of platinum in four ways : 1. From loss of Cl upon ignition Pt = 196.637 2. From weight of Pt in residue " = 195.897 3. From weight of KC1 in residue " = 195.384 4. From ratio between KCl and Pt " = 195.690 The last of these values is undoubtedly the best, for it is not affected by errors due to the possible presence of moisture in the salt analyzed. The work done by Andrews J is even less satisfactory than the foregoing, partly for the reason that its full details seem never to have been pub- lished. Andrews dried potassium chloroplatinate at 105, and then decomposed it by means of zinc and water. The excess of zinc having been dissolved by treatment with acetic and nitric acids, the platinum was collected upon a filter and weighed, while the chlorine in the filtrate was estimated by Pelouze's method. Three determinations gave as fol- lows for the atomic weight of platinum : Mean, 197.887 Unfortunately, Andrews does not state how his calculations were made. *Poggend. Annalen, 8, 177. 1826. fPoggend. Annaleti, 13, 468. 1828. I British Assoc. Report, 1852. Chera. Gazette, 10, 328 THE ATOMIC WEIGHTS. In 1881 Seubert* published his determinations, basing them upon very pure chloroplatinates of potassium and ammonium. The ammo- nium salt, (NH 4 ) 2 PtCl 6 . was analyzed by heating in a.stream of hydrogen, expelling that gas by a current of carbon dioxide, and weighing the residual metal. In three experiments the hydrochloric acid formed during such a reduction was collected in an absorption apparatus, and estimated by precipitation as silver chloride. Three series of experi- ments are given, representing three distinct preparations, as follows : Series I. Am. 2 PtCl 6 . Pt. Percent. Pt, 2.1266 .9348 43-957 1.7880 .7858 43.948 1.8057 .7938 43-960 2.6876 1.1811 43-946 4 7^74 2.0959 43-963 2.0325 .8935 43.961 Mean, 43.956, =b .002 Series II. Am^PtCl^. Pt. Per cent. Pt. 3- 46o .3363 43-87 1 2.6584 .1663 43-876 2.3334 .0238 43-872 1,9031 .8351 43-88: 3.1476 .3810 43.875 2.7054 .1871 43-889 Mean, 43.876, .001 Another portion of this preparation, recrystallized from water, of 1,4358 grm. gave 0.6311 of platinum, or 43.955 per cent. Series III. Am.PtCl,. Pt. Per cent. Ft. 2.5274 1.11*8 43-99 3.2758 1.4409 43.986 1.9279 .8483 44.001 2.0182 .8884 44.020 1.8873 8303 43-994 2.2270 .9798 43.996 2.4852 1.0936 44.004 2.5362 i.i i 66 44.026 3.0822 I-356I 43 99 s Mean, 44.001, .003 *Ber. Deufcsch. Chem. Gesell., 14, 865. PLATINUM. 329 If these series are treated as independent and combined, giving each a weight as indicated by its probable error, and regarding the single ex- periment with preparation II as equal to one in the first series, we get a mean percentage of 43.907, .0009. On the other hand, if we regard the twenty-two experiments as all of equal weight in one series, the mean percentage of platinum becomes 43.953, .0078. Upon comparing the work with that done later by Halberstadt, the latter mean seems the fairer one to adopt. For the chlorine estimations in the ammonium salt, Seubert gives the subjoined data. I add in the last column the weight of salt proportional to 100 parts of silver chloride. Am^PtCl^. Pt. 6AgCl. Ratio. 2.7054 1.1871 . 5.2226 51.802 2.2748 .9958 4.3758 5L9S6 3.0822 i-356i 5-9496 S'-SoS Mean, 51.864, .041 The potassium salt, K. 2 PtCl 6 , was also analyzed by ignition in hydro- gen, treatment with water, and weighing both the platinum and the potassium chloride. The weights given are as follows : Pt. zKCl. 5.0283 2.0173 i.544o 7.0922 2.8454 2.1793 3.5475 1.4217 1.0890 3-2296 1.2941 .9904 35834 1-4372 i.iooi 4.4232 1.7746 1.3547 4.0993 1.6444 1.2589 4.4139 1.7713 1.3516 Hence we have these percentages, reckoned on the original salt KCl. 30.706 30.728 30.698 30.666 30.700 30.627 30.710 30.621 Mean, 40.107, .005 Mean, 30.682, .009 As with the ammonium salt, three experiments were made upon the potassium compound to determine the amount of chlorine (four atoms in this case) lost upon ignition in hydrogen. In the fourth column I add the amount of K 2 PtCl 6 corresponding to 100 parts of AgCl : 330 THE ATOMIC WEIGHTS. PL *AgCL Ratio. 6.7771 2.7158 7.9725 85.006 3.5834 L4372 4.2270 84.774 4.4139 1.7713 5-2144 84.648 Mean, 84.809, .071 Halberstadt,* like Seubert, studied the chloroplatinates of potassium and ammonium, and also the corresponding double bromides and platinic bromide as well. The metal was estimated partly by reduction in hy- drogen, as usual, and partly by electrolysis. Platinic bromide gave the following results : I. By Reduction in H. PtBr^. Pt. Per cent. Pt. .6396 .2422 37.867 1.7596 .6659 37.844 .9178 .3476 37.873 1.1594 .4388 37.847 1.9608 .7420 37.842 2.0865 .7898 37.853 4.0796 1.5422 37-852 6.8673 2.5985 37-8j9 77. By Electrolysis. 1.2588 .4763 37-837 1-4937 .5649 37-819 Mean of all ten experiments, 37.847, .0033 The ammonium platinbromide, (NH 4 ) 2 PtBr 6 , was prepared in two ways, and five distinct lots were studied. With this salt, as well as with those which follow, the data are given in distinct series, with from one to several experiments in each group, but for present purposes it seems best to consolidate the material and so put it in more manageable form. The percentages of platinum and weights found are as follows : /. By Reduction in H. Pt. Percent. Pt. ' .6272 .1719 27.408 .0438 .2865 27.447 .1724 .3215 27.422 1 .4862 .4076 27.426 .0811 .2966 27.435 . .3383 .3672 27.437 *Ber. Deutsch. Chem. Gesell., 17, 2962. 1884. PLATINUM. 331 PL .2769 .3269 .3611 .6159 .3668 .4899 1.1427 3250 .6591 .6940 .4705 .6316 .8245 I-3329 .4210 5594 5751 Per cent. PL 27.426 27.390 27-393 27.402 27-45* 27.431 27.441 27.460 27.459 27.438 27-439 27-444 27.435 27.430 27.449 27-457 27.465 II. By Electrolysis. .4272 27.409 .4397 27.392 .8569 27.439 .3180 27.386 .7081 27.427 .2809 27.456 .4591 27.418 .4591 27.418 4397 27.392 Mean of all thirty-two experiments, 27.429, .0027 With potassium platinbromide Halberstadt found as follows : f 2.5549 | 2.6323 j 2.93 '5 3-4463 1^4.0081 3-9554 2.0794 2.1735 2.3099 1.4085 2.6166 2.6729 PL .6630 .6831 .7598 .8939 1.0404 1.0266 .5388 .5635 .5986 3645 .6772 .6923 /. By Reduction in H. 2 KBr. .8071 .8318 .9259 1.0895 1-2653 1.2495 .6558 .6849 .7297 .4446 .8279 .8469 Per cent. PL Per cent. KBr. 25.940 25.947 25.910 25-938 25.957 25.954 25.911 25.926 25.914 25.880 25.881 25.900 3L590 31-599 31-584 3L6I3 31.568 31-589 3L538 31.5" 3L590 3L565 31.640 31.684 332 THE ATOMIC WEIGHTS. $: 2 /^r 6 . Pt. sKBr. Percent. Pt. 2. 21 10 5726 .6997 25.898 3.1642 .8188 9983 1.9080 1.6754 4947 4341 .6025 .5286 25.927 25-915 1.3148 3403 .4160 25.882 L5543 .4025 .4911 25-895 By Electrolysis. Per cent. KBr. 3L647 3L550 31-577 3L550 3^.640 31.596 Mean of eighteen experiments, 25.915, .0040 31.591, .0068 For ammonium platinchloride Halberstadt gives the following data : /. By Reduction in H. Pt. .4.662 .6087 .6617 1.0227 .6059 .7638 1.2068 1.4019 2-4035 J-532I Per cent. Pt. 43-964 43.962 43.95 6 43.880 43.906 44.011 43-971 43-984 43.951 9474 1.1069 1.5101 5345 1-6035 1.9271 1.1046 1.4179 //. By Electrolysis. .4161 .4865 .6634 .2347 .7044 .8459 .4858 6233 43.920 43.951 43.930 43-9 10 43-928 43.894 43-979 43-959 Mean of eighteen experiments, 43.943, .0054 Seubert found, 43.953, .0078 General mean, 43.946, .0044 For potassium platinchloride Halberstadt's data are /. By Reduction in H. K.PtCl,. Pt. 2KCI. Percent. Pt. Per cent. KCL f 1.6407 .6574 .5029 40.069 30-651 1 1-9352 {' 7757 5921 40.084 30.600 L5793 .6334 .4836 40.106 30.621 1.6446 6595 .5049 40. 101 . 30.700 1.0225 2.4046 .4102 .9641 3133 .7388 40.117 40.094 30.640 30.724 f 5.8344 2.3412 1.7005 40.127 30.688 (7.1732 2.8776 2.1998 40.116 30.666 PLATINUM. 333 77. By Electrolysis. PL 2KCL Per cent. Pt. Per cent. KCl. 1.2354 .4953 .3792 40.092 30.695 2.5754 1.0318 .7898 40.063 30.667 L0933 .4387 .3355 40.126 30.668 1.3560 .5438 .4167 40.103 30.730 L7345 .6956 .5298 40.104 30.545 2.0054 .8038 .6147 40.081 30.652 2.0666 .8291 .6356 40.117 3O.755 1.2759 .5"8 .3908 40.112 30.629 1.9376 .7763 .5927 40.065 30.589 2.3972 .9608 .7355 40.080 30.681 1.2.7249 1.0929 .8364 40.108 30.691 Mean of nineteen experiments, 40.098, rb .0031 30.663, .0080 Seubert found, 40. 107, .0050 30.682, .0090 General mean ,.40.101, d= .0026 30.671, .0060 The work of Dittmar and M'Arthur* on the atomic weight of platinum is difficult to discuss and essentially unsatisfactory. They investigated potassium platinchloride, and came to the conclusion that it contains traces of hydroxyl replacing chlorine and also hydrogen replacing potassium. It is also liable, they think, to carry small quantities of potassium chloride. In their determinations, which involve corrections indicated by the foregoing considerations, they are not sufficiently ex- plicit, and give none of their actual weighings. They attempt, however, to fix the ratio 2KC1 : Pt, and after a number of discordant, generally high results, they give the following data for the atomic weight of plati- num based upon the assumption that 2KC1 = 149.182 : 195.54 195.48 195.60 195.37 Mean, 195.50, .0330. Dittmar and M'Arthur also discuss Seubert's determinations, seeking to show that the latter also, properly treated, lead to a value nearer to 195.5 than to 195. Seubert at once replied to them,f pointing out that the concordance between his determinations by very different methods (a concordance verified by Halberstadt's investigation) precluded the existence of errors due to impurities such as Dittmar and M'Arthur assumed. * Trans. Roy. Soc. Edinburgh, 33, 561. 1887. tBer. Deutsch. Chem. Gesell., 21, 2179. 1888. 334 THE ATOMIC WEIGHTS. The ratios from which to compute the atomic weight of platinum are now as follows, rejecting the work of Berzelius and of Andrews : (i.) Percentage of Pt in ammonium platinchloride, 43.946, .0044 (2.) Percentage of Pt in ammonium platinbromide, 27.429, db .0027 (3.) Percentage of Pt in potassium platinchloride, 40.101, .0026 (4.) Percentage of Pt in potassium platinbromide, 25.915, .0040 (5.) Percentage of Pt in platinic bromide, 37.847, =b .0033 (6.) Percentage of KC1 in potassium platinchloride, 30.671, .0060 (7.) Percentage of KBr in potassium platinbromide, 31.591, =b .0068 (8.) 6AgCl : Am 2 PtCl 6 : : 100 : 51.864, rb .041 (9.) 4AgCl : K 2 PtCl 6 : : loo : 84.809, .071 (10.) 2KC1 : Pt : : 149.182 : 195.50, dr .033 Computing with the subjoined atomic and molecular weights Cl = 35.179, .0048 KC1 = 74.025, rb .0019 Br = 79.344, .0062 KBr = 118.200, rb .0073 K = 38.817, rb .0051 . AgCl = 142.287, .0037 N = 13.935, .0021 we have the following ten values for platinum : From (i) Pt = 193.603, rb .0336 From (2) "= 193.493, .0248 From (3) " = 193.283, =b .0254 From (4) " = 193.684, db .0344 From (5) " = 193.261, rfc .0248 From (6) " = 193 938, rb .0746 From (7) " = 194-538, =b . 1276 From (8) " = 195.836, rb .3515 From (9) " = 193.980, .4054 From (10) " = 194.017, db .0331 General mean . Pi = 193.443, .0114 If = 16, Pt = 194.917. Of these ten values the first five are obviously the most trustworthy. Their general mean is Pt = 193.414, .0124 ; or, if = 16, Pt = 194.888. This result is preferable to the mean of all, even though the latter varies little from it. The five high values carry very little weight because of their larger probable errors. CERIUM. 335 CERIUM. Although cerium was discovered almost at the beginning of the present century, its atomic weight was not properly determined until after the discovery of lanthanum and didymium by Mosander. In 1842 the in- vestigation was undertaken by Beringer,* who employed several methods. His cerium salts, however, were all rose-colored, and therefore were not wholly free from didymium ; and his results are further affected by a negligence on his part to fully describe his analytical processes. First, a neutral solution of cerium chloride was prepared by dissolving the carbonate in hydrochloric acid. This gave weights of eerie oxide and silver chloride as follows. The third column shows the amount of CeO 2 proportional to 100 parts of AgCl : CeO 2 . AgCl. Ratio. 5755 grm. 1.419 grm. 4O-557 .6715 " 1.6595 " 40.464 1.1300 " 2.786 " 40.560 .5366 " i.33'6 " 40.297 Mean, 40.469, .0415 The analysis of the dry cerium sulphate gave results as follows. In a fourth column I show the amount of Ce0 2 proportional to 100 parts of BaS0 4 : Sulphate. CeO^. BaSO Ratio. 1.379 grm. .8495 grm. 1.711 grm. 49.649 1.276 " .7875 " 1.580 " 49.836 1.246 " .7690 " 1.543 " 49.838 1.553 " .9595 " 1.921 " 49.948 Mean, 49.819, .042 Beringer also gives a single analysis of the formate and the results of one conversion of the sulphide into oxide. -The figures are, however, not valuable enough to cite. The foregoing data involve one variation from Beringer's paper. Where I put Ce0 2 as found he puts Ce 2 O s . The latter is plainly inad- missible, although the atomic weights calculated from it agree curiously well with some other determinations. Obviously, the presence of didym- ium in the salts analyzed tends to raise the apparent atomic weight of cerium. Shortly after Beringer, Hermann f published the results of one experi- ment. 23.532 grm. of anhydrous cerium sulphate gave 29.160 grm. of BaS0 4 . Hence 100 parts of the sulphate correspond to 123.926 of BaS0 4 . *Ann. Chem. Pharm.,42, 134. 1842. t Journ. fur Prakt. Chem., 30, 185. 1843. 336 THE ATOMIC WEIGHTS. In 1848 similar figures were published by Marignac,* who found the following amounts of BaS0 4 proportional to 100 of dry cerium sulphate : Mean, 122.40, .138 If we give Hermann's single result the weight of one experiment in this series, and combine, we get a mean value of 122.856, .130. Still another method was employed by Marignac. A definite mixture was made of solutions of cerium sulphate and barium chloride. To this were added, volumetrically, solutions of each salt successively, until equilibrium was attained. The figures published give maxima and minima for the BaCl 2 proportional to each lot of Ce. 2 (SO 4 ) 3 . In another column, using the mean value for BaCl 2 in each case, I put the ratio between 100 parts of this salt and the equivalent quantity of sulphate. The latter compound was several times recrystallized : BaCl v Ratio. First crystallization ...... ii.ongrm. 11.990 12.050 grm. 91.606 First crystallization. : ____ 13.194 " i4-3 6 5 T 4-425 " 91-657 Second crystallization.. . . 13.961 15.225 15.285 91.518 Second crystallization.. . . 12.627 " 13.761 13.821 " 9L559 Second crystallization.. . . 11.915 " 12.970 13.030 " 91-654 Third crystallization 14.888 < 16.223 16.283 " 91.602 Third crystallization 14.113 " I5.383 I5-423 " 9L755 Fourth crystallization.. . . 13.111 " 14.27014.330 " 91.685 Fourth crystallization 13.970 J 5-223 15.283 " 91.588 Mean, 91.625, .016 Omitting the valueless experiments of Kjerulf,f we come next to the figures published by Bunsen and Jegel J in 1858. From the air-dried sulphate of cerium the metal was precipitated as oxalate, which, ignited, gave Ce0 2 . In the filtrate from the oxalate the sulphuric acid was esti- mated as BaSO 4 : 1.5726 grm. sulphate gave .7899 grm. CeO 2 and 1.6185 S rm - BaSO 4 . 1.6967 " .8504 " 1.7500 " Hence, for 100 parts BaSO 4 , the CeO a is as follows : 48.804 48.575 Mean, 48.689, d= .077 *Arch. Sci. Phys. et Nat. (i), 8, 273. 1848. t Ann. Chem. Pharra., 87, 12. J Ann. Chem. Pharni., 105, 45. 1858. CERIUM. 337 One experiment was also made upon the oxalate : 353 S rm - oxalate gave .1913 CeO 2 and .0506 H 2 O. Hence, in the dry salt, we have 63.261 per cent, of CeO 2 . In each sample of Ce0 2 the excess of oxygen over Ce 2 3 was estimated by an iodometric titration ; but the data thus obtained need not be fur- ther considered. In two papers by Rammelsberg* data are given for the atomic weight of cerium, as follows. In the earlier paper cerium sulphate was analyzed, the cerium being thrown down by caustic potash, and the acid precipi- tated from the nitrate as barium sulphate : .413 grm. Ce 2 (SO 4 ) 3 gave .244 grm. Ce0 2 and .513 grm. BaSO 4 . Hence 100 BaSO 4 = 47.563 Ce0 2 , a value which may be combined with others, thus ; this figure being assigned a weight equal to one experi- ment in Bunsen's series : Beringer ............................... 49.819, .042 Kunsen and Jegel ......................... 48.689, .077 Rammelsberg ..... ....................... 47-5^3> -t- . 108 General mean 49.360, =b .035 It should be noted here that this mean is somewhat arbitrary, since Bunsen and Rammelsberg's cerium salts were undoubtedly freer from didymium than the material studied by Beringer, In his later paper Rammelsberg gives these figures concerning cerium oxalate. One hundred parts gave 10.43 of carbon and 21.73 of water. Hence the dry salt should yield 48.862 per cent, of CO 2 , whence Ce = 137.14. In all of the foregoing experiments the eerie oxide was somewhat col- ored, the tint ranging from one shade to another of light brown according to the amount of didymium present. Still, at the best, a color remained, which was supposed to be characteristic of the oxide itself. In 1868, however, some experiments of Dr. C. Wolff were posthumously made public, which went to show that pure ceroso-ceric oxide is white, and that all samples previously studied were contaminated with some other earth, not necessarily didymium but possibly a new substance, the re- moval of which tended to lower the apparent atomic weight of cerium very perceptibly. Cerium sulphate was recrystallized at least ten times. Even after twenty recrystallizations it still showed spectroscopic traces of didymium. The water contained in each sample of the salt was cautiously estimated, and the cerium was thrown down by boiling concentrated solutions of * Poggend. Annalen, 55, 65 ; 108, 44. t Amer. Journ. Science and Arts (2), 46, 53. 338 THE ATOMIC WEIGHTS. oxalic acid. The resulting oxalate was ignited with great care. I de- duce from the weighings the percentage of Ce0 2 given by the anhydrous sulphate : CeO.^ .76305 grin. .7377 .70665 " Sulphate. 1.4542 grm. 1.4104 " 1.35027 " Water. . 19419 grrn. .1898 " .1820 " Percent. 60.559 60.437 60.487 Mean, 60.494 After the foregoing experiments the sulphate was further purified by solution in nitric acid and pouring into a large quantity of boiling water. The precipitate was converted into sulphate and analyzed as before : Sulphate. Water. CeO.>. L4327 g rm - . 2 733 g rm - -69925 grm. 1.5056 " .2775 " .7405 " 1.44045 " .2710 " .7052 " Per cent. CeO. 2 . 60.311 60.296 60.300 Mean, 60.302 From another purification the following weights were obtained : 1.4684 grm. .1880 grm. .7717 grm. 60.270 per cent. A last purification gave a still lower percentage : t.3756 grm. .1832 grm. .7186 grm. 60.265 per cent. The last oxide was perfectly white, and was spectroscopically free from didymium. In each case the Ce0 2 was titrated iodometrically for its excess of oxygen. It will be noticed that in the successive series of de- terminations the percentage of Ce0 2 steadily and strikingly diminishes to an extent for which no ordinary impurity of didymium can account. The death of Dr. Wolf interrupted the investigation, the results of which were edited and published by Professor F. A. Genth. In the light of more recent evidence, little weight can be given to these observations. All the experiments, taken equally, give a mean percent- age of Ce0 2 from Ce 2 (S0 4 ) 3 of 60.366, .0308. This mean has obviously little or no real significance. The experiments of Wolf attracted little attention, except from Wing,* who partially verified certain aspects of them. This chemist, incidentally to other researches, purified some cerium sulphate after the method of Wolf, and made two similar analyses of it, as follows : Sulphate. Water. CeO. 2 . Percent. CeO. 2 . 1.2885 grm. .1707 grm. .6732 grm. 60.225 1.4090 " .1857 " .7372 " 60.263 Mean, 60 244 * Am. Journ. Sci. (2), 49, 358. 1870. CERIUM. 339 The cerio oxide in this case was perfectly white. The cerium oxalate which yielded it was precipitated boiling by a boiling concentrated solu- tion of oxalic acid. The precipitate stood twenty-four hours before filtering. In 1875 Buehrig's * paper upon the atomic weight of cerium was issued. He first studied the sulphate, which, after eight crystallizations, still retained traces of free sulphuric acid. He found, furthermore, that the salt obstinately retained traces of water, which could not be wholly ex- pelled by heat without partial decomposition of the material. These sources of error probably affect all the previously cited series of experi- ments, although, in the case of Wolf's work, it is doubtful whether they could have influenced the atomic weight of cerium by more than one or two tenths of a unit. Buehrig also found, as Marignac had earlier shown, that upon precipitation of cerium sulphate with barium chloride the barium sulphate invariably carried down traces of cerium. Furthermore, the eerie oxide from the filtrate always contained barium. For these reasons the sulphate was abandoned, and the atomic weight determina- tions of Buehrig were made with air-dried oxalate. This salt was placed in a series of platinum boats in a combustion tube behind copper oxide. It was then burned in a stream of pure, dry oxygen, and the carbonic acid and water were collected after'the usual method. Ten experiments were made; in all of them the above-named products were estimated, and in five analyses the resulting eerie oxide was also weighed. By de- ducting the water found from the weight of the air-dried oxalate, the weight of the anhydrous oxalate is obtained, and the percentages of its constituents are easily determined. In weighing, the articles weighed were always counterpoised with similar materials. The following weights were found : Oxalate. Water. CO 2 . CeO. z . 9.8541 grm. 2.i987grm. 3.6942 grm 9.5368 " 2.1269 " 3-5752 " 9.2956 " 2.0735 " 3.4845 " 10.0495 " 2.2364 " 3-774 " 10.8249 " 2.4145 " 4.0586 " 9.3679 " 2.0907 " 3-5 118 " 4-6150 grm. 9.7646 " 2.1769 " 3.6616 " 4.8133 " 9.9026 " 2.2073 " 3.7139 s " 4-8824 " 9.9376 " 2.2170 " 3.7251 " 4-8971 " 9.5324 " 2.1267 " 3-5735 " 4-6974 " These figures give us the following percentages for C0 2 and Ce0 2 in the anhydrous oxalate : * Journ. fi'ir Prakt. Chem., 120, 222. 1875. 340 THE ATOMIC WEIGHTS, CO. r CeO,. 48.256 48.249 48.248 48.257 48.257 48.258 63417 48.257 63.436 48.262 63.446 48.249 63.429 48.253 63.430 Mean, 48.2546 .001 Mean, 63.4316, =b .0032 These results could not be appreciably affected by combination with the single oxalate experiments of Jegel and of Rarnmelsberg, and the latter may therefore be ignored. Robinson's work, published in 1884,* was based upon pure cerium chloride, prepared by heating dry cerium oxalate in a stream of dry, gaseous hydrochloric acid. This compound was titrated with standard solutions of pure silver, prepared according to Stas, and these were weighed, not measured. In the third column I give the ratio between CeCl 3 and 100 parts of silver : CeCl 3 . Ag. Ratio. 5.5361 7.26630 76.189 6.0791 7-98377 76 172 6.4761 8.50626 76.133 6.98825 9.18029 76.122 6.6873 8.78015 76.164 7.0077 9.20156 76.158 6.9600 9- r 393 76.150 Mean Reduced to a vacuum this becomes 76.167. In a later paper, f Robinson discusses the color of eerie oxide, and criticises the work of Wolf. He shows that the pure oxide is not white, and makes it appear probable that Wolf's materials were contaminated with compounds of lanthanum. He also urges that Wolf's cerium sul- phate could not have been absolutely definite, because of defects in the method by which it was dehydrated. Brauner,J in 1885, investigated cerium sulphate with extreme care, and appears to have obtained material free from all other earths and absolutely homogeneous. The anhydrous salt was calcined with all * Chemical News, 50, 251. Nov. 28, 1884. Proc. Roy. Soc., 37, 150. t Chemical News, 54, 229. 1886. t Sitzungs. Wien. Akad., Bd. 92. July, 1885. CERIUM. 341 necessary precautions, and the data obtained, reduced to a vacuum, were as follows : Ce. 2 (SO\. CeO r Percent. CeO 2 . 2.16769 1.31296 60.5693 2.43030 1.47205 60.5707 2.07820 1.25860 60.5620 2.21206 1.33989 60.5721 1.28448 .77845 60.6043 1.95540 1.18436 60.5687 2.46486 1.49290 60.5673 2.04181 1.23733 6o -5997 2.17714 1.31878 60.5739 2.09138 1.26654 60.5605 2.21401 1.34139 60.5863 2.44947 1.48367 60.5711 2.22977 1.35073 60.5771 2.73662 1.65699 60.5486 2.62614 1.59050 60.5642 1.67544 1.01470 60.5632 1.57655 -95540 60.6007 2.72882 1.65256 60.5600 2.10455 1.27476 60.5716 2 - IO 735 1.27698 60.5965 2-43557 I-475 1 ? 60.5692 3.01369 1.82524 60.5649 4.97694 3.0I37 2 60.5537 Mean, 60 5729, .0021 This mean completely outweighs the work done by Wolf and Wing, so that upon combination the latter practically vanish. Wing's mean is arbitrarily given equal weight with Wolf's, and the combination is as follows : Wolf. 60.366, .0308 Wing 60. 244, =b .0308 Brauner 60.5729, .0021 General mean 60.566, d= .0021 In 1895 several papers upon the cerite earths were published by Schutz- ^nberger.* In the first of these a single determination of atomic weight is given. Pure Ce0. 2 , of a yellowish white color, was converted into sul- phate, which was dried in a current of dry air at 440. This salt, dis- solved in water, was poured into a hot solution of caustic soda, made from sodium, and, after filtration and washing, the filtrate, acidulated with hydrochloric acid, was precipitated with barium chloride. The trace of sulphuric acid retained by the cerium hydroxide was recovered by re-solution and a second precipitation, and added to the main amount. * Compt. Rend., 120, pp. 663, 962, and 1143. 1895. 342 THE ATOMIC WEIGHTS. 100 parts of Ce,(S0 4 ) 3 gave 123.30 of BaS0 4 . This may be assigned equal weight with one experiment in Marignac's series, giving the following combination : Hermann 123 926, .238 Marignac 122.40, .138 Schutzenherger 123.30, .238 General mean ..................... 122.958, . 1 139 Schutzenberger, criticising Brauner's work, claims that the latter was affected by a loss of oxygen during the calcination of the cerium dioxide. In his second and third papers Schutzenberger describes the results obtained upon the fractional crystallization of cerium sulphate. Prepa- rations were thus made yielding oxides of various colors canary yellow, rose, yellowish rose, reddish, and brownish red. These oxides, by syn- thesis of sulphates, the barium-sulphate method, etc., gave varying values for the atomic weight of cerium, ranging from 135.7 to 143.3. Schutzen- berger therefore infers that cerium oxide from cerite contains small quantities of another earth of lower molecular weight ; but the results as given are not sufficiently detailed to be conclusive. The third paper is essentially a continuation of the second, with reference to the didymiums. Schutzenberger's papers were promptly followed by one from Brauner,* who claims priority in the matter of fractio nation, and gives some new data, the latter tending to show that cerium oxide is a mixture of at least two earths. One of these, of a dark salmon color, he ascribes to a new element, " meta-cerium." The other he calls cerium, and gives for it a preliminary atomic weight determination. The pure oxalate, by Gibbs y method, gave 46.934 per cent, of Ce0 2 , and, on titration with potassium permanganate, 29.503 and 29.506 per cent, of C 2 O 3 . Hence Ce = 138.799. In mean, this ratio may be written 3 C 2 3 : 2Ce0 2 : : 29.5045 : 46-934, and to each of its numerical terms we may roughly assign the probable error .001. This is derived from the average of the two titrations, and is altogether arbitrary. The ratios, good and bad, for cerium now are (i.) Ce 2 (SO 4 \ s : 3BaSO 4 : : 100 : 122.958, d= .1139 (2.) 3BaSO 4 : 2CeO 2 : : 100 : 49-36o, .035 (3.) 3 BaCl 2 : Ce 2 (S0 4 ) 3 : : loo : 91.625, .016 (4.) 3AgCl : CeO 2 : : loo : 40.469, .0415 (5.) Percentage CeO 2 from Ce 2 (SO 4 ) 3 , 60.566, .0021 (6.) Percentage CeO 2 from Ce 2 (C 2 O 4 > ) 3 , 63.4316, + .0032 (7.) Percentage CO 2 from Ce 2 (C 2 O 4 ) 3 , 48.2546, =b .001. (8.) 3Ag : CeCl 3 : : IOO : 76.167, .0065 (9-) 3 C 2 3 : 2Ce 2 : : 2 9.5 45, .001 : 46.934, -ooi *Chem. News, 71, 283. CERIUM. 343 To reduce these ratios we have O = I5.879,:t.OOO3 C = II.92O, .OOO4 Cl = 35.179, d= .0048 S = 31.828, zb .0015 Ag = 107.108, dz .0031 Ba = 136.392, .0086 i42.287, .0037 From the ratios, with these intermediate data, we can get two values for the molecular weight of Ce 2 (S0 4 ) 3 , and five for that of Ce0 2 . For cerium sulphate we have From (i) ................... Ce 2 (SO 4 ) 3 = 565.404, . 1670 From (3) ................... " = 568.304, db . 1054 General mean ......... Ce z (SO 4 ) 3 = 567.478, .0891 Hence Ce == 140.723, .0451. For eerie oxide the values are From (2) ...... ' ................. CeO 2 171.577, .1218 From (4) ........... . ........... " = 172.746,^.1772 From (5) ....................... " =r 170.879, .0115 From (6) ....................... " =172.125,^.0177, From (9) ...................... " = 170.557, .0076 General mean ............. CeO 2 = 170.827, .0060 And Ce = 139.069, .0061. For cerium itself, four independent values are now calculable, as follows : From molecular weight of sulphate. . . Ce = 140.723, .0451 From molecular weight of dioxide ... " = 139.069, rb .0061 From ratio (8) .................... " = 139.206, .0263 From ratio (7) ............... ..... " = 140.516, .0047 General mean ............... Ce = 140. 1 13, =b .0036 If = 16, Ce = 141.181. It must be admitted that this combination is of very questionable utility. Its component means vary too widely from each other, and in- volve too many uncertainties. Furthermore, Schutzenberger and Brau- ner both impugn the homogeneity of the supposed element, as it has hitherto been recognized. Even if no " meta-elements " are involved in the discussion, it seems clear, on chemical grounds, that the two lower values are really preferable to the two higher, and that ratio (7) receives excessive weight. The general mean obtained is probably a full unit too high. The value 139.1 is perhaps nearly correct. 344 THE ATOMIC WEIGHTS. LANTHANUM. Leaving out of account the work of Mosander. and the valueless ex- periments of Choubine, we may consider the estimates of the atomic weight of lanthanum which are due to Hermann, Rammelsberg, Marig- nac, Czudnowicz, Holzmann, Zschiesche, Erk, Cleve, Brauner, Bauer, and Bettendorff. From Rammelsberg* we have but one analysis. .700 grm. of lantha- num sulphate gave .883 grm. of barium sulphate. Hence 100 parts of BaS0 4 are equivalent to 79.276 of La 2 (S0 4 ) 3 . Marignac.f working also with the sulphate of lanthanum, employed two methods. First, the salt in solution was mixed with a slight excess of barium chloride. The resulting barium sulphate was filtered off and weighed; but, as it contained some occluded lanthanum compounds, its weight was too high. In the filtrate the excess of barium was estimated, also as sulphate. This last weight of sulphate, deducted from the total sulphate which the whole amount of barium chloride could form, gave the sulphate actually proportional to the lanthanum compound. The following weights are given : , BaCl r ist BaSO,. 2 d BaSO,. 4.346 grm. 4.758 grm. 5.364 grm. .115 grm. 4-733 " 5.178 " 5-848 " .147 " Hence we have the following quantities of La,,(S0 4 ) 3 proportional to 100 parts of BaS0 4 . Column A is deduced from the first BaS0 4 and column B from the second, after the manner above described : A. B. 81.022 80.934 83.281 83.662 Mean, 80.978, .030 From A Mean, 83.471, - La b.I28 n8 4.7 From B . . 147. n A agrees best with other determinations, although, theoretically, it is not so good as B. Marignac's second method, described in the same paper with the forego- ing experiments, consisted in mixing solutions of La.,(S0 4 \. 5 with solutions of BaCl. 2 , titrating one with the other until equilibrium was established. The method has already been described under cerium. The weighings * Poggend. Annalen, 55, 65. t Arch. Sci. Phys. et Nat. (i), n, 29. 1849. LANTHANUM. 345 give maxima and minima for BaCl 2 . In another column I give La.,(S0 4 ) 3 proportional to 100 parts of BaCl 2 , mean weights being taken for the latter : Ratio. 91.004 90.968 91-297 9'.332 9 r -3 62 9L475 91.364 91.615 91.482 Mean, 91.322, .048 Hence La = 140.2. Although not next in chronological order, some still more recent work of Marignac's * may properly be considered here. The salt studied was the sulphate of lanthanum, purified by repeated crystallizations. In two experiments the salt was calcined, and the residual oxide weighed ; in two others the lanthanum was precipitated as oxalate, and converted into oxide by ignition. The following percentages are given for La 2 O 3 : "'5 I By calcination. La.,(SO BaCl v 1 1. 644 g m. 12.765 12.825 12.035 ' I 3-i95 I 3-265 10.690 ' 11.669 11-749 12.750 13.920 14.000 io.757 11.734 11.814 12.672 13-813 13.893 9.246 10.080 10.160 10.292 11.204 1 1.264 10.192 ( ii. in 11.171 57.58 57.50 57-55 jPpt. as oxalate. Mean, 57-5475, - OII 5 The atomic weight determinations of Holzmann f were made by analy- ses of the sulphate and iodate of lanthanum, and the double nitrate of magnesium and lanthanum. In the sulphate experiments the lantha- num was first thrown down as oxalate, which, on ignition, yielded oxide. The sulphuric acid was precipitated as BaSO 4 in the filtrate. 5'57 .3323 .4626 .9663 grm. .6226 " .8669 " 1.1093 grm. .7123 " .9869 " These results are best used by taking the ratio between the BaS0 4 , put at 100, and the La, 2 0. r The figures are then as follows : 46.489 46652 46.873 Mean, 46.671, .075 * Ann. Chim. Phys. (4), 30, 68. 1873. t Journ. fur Prakt. Chem., 75, 321. 1858. 346 THE ATOMIC WEIGHTS. In the analyses of the iodate the lanthanum was thrown down as oxa- late, as before. The iodic acid was also estimated volumetrically, but the figures are hardly available for present discussion. The following percentages of La 2 3 were found : 23-454 23.419 23.468 Mean, 23.447, .0216 The formula of this salt is La 2 (I0 3 ) 6 .3H 2 0. The double nitrate, La 2 (N0 3 ) 6 .3Mg(N0 3 ) 2 .24H 2 0, gave the following analytical data : Salt. H^O. MgO. 53 2 7 g rm - I 569grm. .0417 grm. .1131 grm. 5931 " -1734 " -0467 " .1262 " .S 662 " .1647 " -0442 " .1197 " 3757 " .0297 " .0813 " .3263 <( .0256 l< .0693 " These weighings give the subjoined percentages of La 2 3 : 21.231 21.278 21.141 21.640 21.238 Mean, 21.3056, .058 These data of Holzmann give values for the molecular weight of La 2 s as follows : From sulphate , La 2 O 3 = 322.460 From iodate " 320.726 From magnesian nitrate " = 322.904 Czudnowicz* based his determination of the atomic weight of lantha- num upon one analysis of the air-dried sulphate. The salt contained 22.741 per cent, of water. .598 grm. gave .272 grm. La 2 O 3 and .586 grm. BaSO 4 . The La 2 3 was found by precipitation as oxalate and ignition. The BaSO 4 was thrown down from the filtrate. Reduced to the standards already adopted, these data give for the percentage of La 2 O 3 in the anhy- drous sulphate the figure 58.668. 79.117 parts of the salt are propor- tional to 100 parts of BaSO 4 . * Journ. fur Prakt. Chem., So, 33. 1860. LANTHANUM. 347 Hermann * studied both the sulphate and the carbonate of lanthanum. From the anhydrous sulphate, by precipitation as oxalate and ignition, the following percentages of La 2 O 3 were obtained : Mean, 57.654, .016 The carbonate, dried at 100, gave the following percentages : 68.47 La 2 3 . 27.67 C0 2 . 3.86 H 2 0. Reckoning from the ratio between C0 2 and La 2 O 3 , the molecular weight of the latter becomes 324.254. Zschiesche's f experiments consist of six analyses of lanthanum sul- phate, which salt was dehydrated at 230, and afterwards calcined. I subjoin his percentages, and in a fourth column deduce from them the percentage of La 2 3 in the anhydrous salt : H^O. SO 3 . La-iO^ La z O 3 in Anhydrous Salt. 22.629 33-470 43.909 56.745 22.562 33.306 44.132 5 6 .9 6 4 22.730 33.200 44.070 57-34 22.570 33-333 44.090 56.947 22.610 33.160 44.24 57- I 5 22.630 33-05 1 44-3 10 57.277 Mean, 57.021, .051 Erk J found that .474 grm. of La 2 (S0 4 ) 3 ,by precipitation as oxalate and ignition, gave .2705 grm. of La 2 3 , or 57.068 per cent. .7045 grm. of the sulphate also gave .8815 grm. of BaS0 4 . Hence 100 parts of BaS0 4 are equivalent to 79.921 of La 2 (S0 4 ) 3 . From Cleve we have two separate investigations relative to the atomic weight of lanthanum. In his first series strongly calcined La 2 3 , spec- troscopically pure, was dissolved in nitric acid, and then, by evaporation with sulphuric acid, converted into sulphate : 1.9215 grm. La 2 O 3 gave 3.3365 grm. sulphate. 57-59 P er cent. 2.0570 3.5705 57.6ii it 1.6980 " 2.9445 57.667 < < 2.0840 " 3.6170 57.6i7 1.9565 , " 3.396o 57.612 11 Mean, 57.619, rb .0085 * Journ. fur Prakt. Chem., 82, 396. 1861. t Journ. fur Prakt. Chem., 104, 174. I Jenaisches Zeitschrift, 6, 306. 1871. g K. Svensk. Vet. Akad. Handlingar, Bd. 2, No. 7. 1874. 348 THE ATOMIC WEIGHTS. From the last column, which indicates the percentage of La 2 O 3 in La 2 (S0 4 ) 3 , we get, if SO, =* 80, La = 139.15. In his second paper,* published nine years later, Cleve gives results similarly obtained, but with lanthanum oxide much more completely freed from other earths. The data are as follows, lettered to correspond to different fractions of the material studied : B - ' 8 39 S rm . La 2 O 3 gave 1.4600 sulphate. 57.466 per cent. fi.i86i 2.0643 " 57458 " c I -8993 " L5645 " 57.482 ' | .8685 1.5108 " 57.486 " I .8515 " 1.4817 " 57.468 " D I .6486 " 1.1282 " 57.490 " ' 1 .7329 1.2746 57oOO " E. 1.2477 2.1703 " 57.490 F | 1.1621 2.0217 " 57.481 ' i 1-5749 " 2.7407 " 57-463 " G | 1.3367 2.3248 " 57.497 " .4455 " 2.5146 " 57-484 " Mean, 57.480, dz .0040 Hence with S0 3 = 80, La = 138.22. From Brauner we also have two sets of determinations, both based upon the conversion of pure La 2 O 3 into La 2 (S0 4 ) 3 . In his first paper, Brauner f gives only two syntheses, as follows: 1-75933 g rm - La 2 O 3 gave 3.05707 La a (SO 4 ) 3 . 57-5 6 6 per cent. .92417 " 1.60589 " 57-549 Mean, 57-5575 This mean we may regard as of equal weight with Marignac's, and assign to it the same probable error. In Brauner's second paper J six experiments are given ; but the weights are affected by a misprint in the second determination, which I am un- able to correct. Only five of the syntheses, therefore, are given below. 7850 grm. La 2 O 3 gave 1.3658 La 2 (SO 4 ) 3 . 57-476 per cent. 2.1052 " 3-6633 " 57.467 " i. ooio " I-74H " 57-5 2 5 1.3807 " 2.4021 " 57-479 " 1.5275 " 2.6588 " 57.451 Mean, 57.480, db .0084 Brauner's weighings are all reduced to a vacuum. Both Bauer and Bettendorff made their determinations of the atomic * K. Sveiisk. Vet. Akad. Handlingar, No. 2, 1883. t Journ. Chem. Soc., Feb., 1882, p. 68. I Sitzuugsb. Wien. Akad., June, 18*2, Bd. 86, II Abth. LANTHANUM. 349 weight of lanthanum by the same general method as the preceding Bauer's data * are as follows : .6431 grm. La 2 O 3 gave 1.1171 sulphate. 57.569 per cent. .7825 " 1.3613 " 57.482 " 1.0112 " I.757I " 57-549 " .7325 " 1.2725 " 57.564 " Mean, 57.541, =b .0136 Bettendorff found f- .9146 grm. La 2 O 3 gave 1.5900 sulphate. 57-522 per cent. -9395 " '.6332 " 57.525 " .9133 " I-5877 " 57.523 " 1.0651 1.8515 " 57.526 " Mean, 57.524, .0006 We may now combine the similar means into general means, and de- duce a value for the atomic weight of lanthanum. For the percentage of oxide in sulphate we have estimates as follows. The single experiments of Czudnowicz and of Erk are assigned the probable error and weight of a single experiment in Hermann's series : Czudnowicz 58.668, =b .027 Erk 57.o68, .027 Hermann 57-654, rfc .016 Zschiesche 57-O2I, .051 Marignac 57-5475, .01 15 Cleve, earlier series 57-6i9, .0085 Cleve, later series 57-48o, .0040 Brauner, earlier series 57-5575, =b - OI ! 5 Brauner, later series 57.480, .0084 Bauer 57-541, db .0136 Bettendorff. 57-524, .0006 General mean 57.522, .00059 This result is practically identical with that of Bettendorff, whose work seems to receive excessive weight. The figure, however, cannot be far out of the way. For the quantity of La 2 (S0 4 ) 3 proportional to 100 parts of BaSO 4 . we have five experiments, which may be given equal weight and averaged together : Marignac , 81.022 Marignac 80.934 Rammelsberg 79.276 Czudnowicz 79. 1 1 7 Erk 79.921 Mean, 80.054, .270 * Freiburg Inaugural Dissertation, 1884. i Ann. d. Chem., 256, 168. 350 THE ATOMIC WEIGHTS. Iii all, there are six ratios from which to calculate : (i.) Percentage of La. 2 O 3 in La 2 (SO 4 ) 3 , 57.522, .00059 (2.) 3BaCl. 2 : La 2 (SO 4 ) 3 : : ioo : 91.322, .048 Marignac (3.) 3BaSO 4 : La 2 (SO 4 ) 3 : : ioo : 80.054, .270 (4.) 3BaSO 4 : La 2 O 3 : : ioo : 46.671, .075 Holzmann (5.) Percentage of La 2 O 3 in iodate, 23.447, dr .0216 Holzmann (6.) Percentage of La 2 O 3 in magnesian nitrate, 21.3056, .058 Holzmann Hermann's single experiment on the carbonate is omitted from this scheme as being unimportant. For the reduction of these data we have O= 15.879, zh. 0003 N :: 13.935, .0021 Cl 35.179, .0048 C = 11.920, .0004 I = 125.888, .0069 Mg = 24. ioo, =b .001 1 S = 31.828, -0015 Ba == 136.392, .0086 For lanthanum sulphate two values are obtainable : From (2) La 2 (SO 4 ) 3 = 566.425, .2999 From (3) " = 556.542,^1.8729 General mean Ln 2 (SO 4 ) 3 = 566. 182, rh .2961 Hence La = 140.075, .1481. For the oxide there are four independent values, as follows : From (i) La 2 O 3 = 322.825, .0090 From (4) " = 322.460,^.5215 From (5) " =320.726,^.3159 From (6) " = 322.904, .9107 A glance at these figures shows that the first alone deserves considera- tion, and that a combination of all would vary inappreciably from it. Taking, then, La 2 3 = 322.825, =fc .0090, we get- La = 137.594, =b .0046; or, with = 16, La = 138.642. If we take the concordant results of Cleve's and Brauner's later series, which give the percentage of La 2 3 in La 2 (S0 4 ) 3 as 57.480, then La = 137.316. Possibly this value may be better than the other, but the evi- dence is not conclusive. THE DIDYMIUMS. 351 THE DIDYMIUMS. Leaving Mosander's early experiments out of account, the atomic weight of the so-called u didymium " was determined by Marignac, Her- mann, Zschiesche, Erk, Cleve, Brauner, and Bauer. All of these data now have only historical value, and may be disposed of very briefly. Marignac* determined the ratios between didymium sulphate and barium sulphate, between silver chloride and didymia, and between didymium sulphate and didymium oxide. The other determinations all relate to the sulphate-oxide ratio. Leaving all else out of account, the earlier data for the percentage of Di 2 O 3 in Di 2 (SO 4 ) 3 are as follows. The atomic weight of Di in the last column is based upon SO 3 = 80 : Per cent. Z?/ 2 Oj. At. Wt. Di. Marignac, f five experiments 58.270 ! 43-56 Hermann, J one experiment 58.140 142.67 Zschiesche,^ five experiments .... 57-9 2 6 141.21 Erk, || two experiments 58.090 i4 2 -33 Cleve, ^[ six experiments 58.766 147.02 Brauner,** three experiments 58.681 146.42 The discordance of the determinations is manifest, and yet up to 1883 the elementary nature of didymium seems to have been undoubted. In that year, however, Cleve and Brauner both showed, independently, that the didymia previously studied by them contained samaria, and that source of disturbance was eliminated. In Brauner 's investigation ft the didymium compounds were carefully fractionated, and the determinations of atomic weight were made by synthesis of the sulphate from the oxide in the usual way. Neglecting details, his first series gave results as follows : Per cent. Di^O y At. Wt. 5 8 -5 6 I45-36 58-526 145.50 58.5 145-31 58-515 I45-42 58.53 1 145-53 *Two papers: Arch. Sci. Phys. et Nat. (i), n, 29. 1849. Ann. Chim. Phys. (3), 38, 148. 1853. f Ann. Chim. Phys. (3), 38, 148. 1853. | Journ. fur Prakt. Chem., 82, 367. 1861. \ Journ. fi'ir Prakt. Chem., 107, 74. || Jenaisches Zeitschrift, 6, 306. 1871. f K. Svensk. Vet. Akad. Handl., Bd. 2, No. 8. 1874. ** Berichte, 15, 109. 1882. ft Journ. Chem. Soc., June, 1883. The values given are as computed by Brauner, with O = 16 and S = 32.07. 352 THE ATOMIC WEIGHTS. Another determination, with material refractionated from that used in his investigation of the previous year, gave 58.512 per cent. Di. 2 3 and Di = 145.40. These determinations, although concordant among themselves, are still about a unit lower than those published in 1882, indicating that in the earlier research some earth of higher molecular weight was present. Accordingly, another series of fractionations was carried out, and the several fractions of " didyrnia " obtained gave the following values : Fraction. Per cent. Di^O^. At.Wt.^Di." i 58.355 H4.32 2 58.479 i45- 16 3 5 8 -5o 145-39 4 58.755 i47.io c J 59.071 149.35 ' ' 1 59-086 149.46 The last fraction is evidently near samaria (Sm = 150), and this earth was proved to be present by a study of the absorption spectra of the material investigated. Similar results, but in some respects more explicit, were obtained by Cleve,* who also found that his earlier research had been vitiated by the presence of samaria. He gives two series of syntheses of sulphate from oxide, with two different lots of material, after eliminating samaria, and obtains, computing with S0 3 = 80, values for Di as follows : ' First Series. Per cent. Di 2 O 3 . At. Wt. Di. 58.088 142.31 58.113 142.49 58.047 142-03 58.099 142.39 58.104 142.42 58.098 142.38 58.104 142.42 58.103 142.42 58.070 142.19 58.079 142.25 Second Series. Percent. >/ 2 6> 3 . At. Wt. Di. 58.125 142.57 58-093 H2.35 58.088 142.31 58.111 142.47 58.056 142.10 58.097 142.38 58.057 142.10 In short, the atomic weight of this " didymium " is not far from 142. *Bull. Soc. Chim., 39, 289. 1883. Ofv. K. Vet. Akad. Forhandl., No. 2, 1883. THE DIDYMIUMS. 353 Bauer's little known determinations* were also made by the synthesis of the sulphate. They have corroborative value and are as follows : Per cent. >/ 2 <9 3 . At. Wt. Di. 58.285 I43-56 58.100 142.40 58.133 142.64 58.098 142.38 In 1885 all of the foregoing determinations were practically brushed aside by Auer von Welsbaeh,f who by the most laborious fraction ations proved that the so-called " didymia " was really a mixture of oxides, whose metals he names neodidymium and praseodidymium, names which are now commonly shortened into neodymium and praseodymium. One of these metals gives deep rose-colored salts, the other forms green compounds, and the difference of color is almost as strongly marked as in the cases of cobalt and nickel. Their atomic weights, determined by the sulphate method, are given by Welsbach a Pr = 143.6 Nd = 140.8 No further details as to these determinations are cited, and whether they rest upon = 16, S0 3 = 80, or = 15.96 is uncertain. Fuller deter- minations are evidently needed. * Freiburg Inaugural Dissertation, 1884.. t Monatsh. Chem., 6. 4QO. 1885. 23 354 THE ATOMIC WEIGHTS. SCANDIUM. Clove,* who was the first to make accurate experiments on the atomic weight of this metal, obtained the following data : 1.451 grm. of sulphate, ignited, gave .5293 grm. of Sc 2 3 . .4479 grin, of Sc 2 3 , converted into sulphate, yielded 1.2255 grm. of the latter, which, upon ignition, gave .4479 grin, of Sc 2 3 . Hence, for the percentage of c 2 3 in Sc 2 (S0 4 ) s we have : 36.478 36.556 36.556 Mean, 36.530, .0175 Hence, if SO, = 79.465, Sc = 44.882. Later results are those of Nilson,t who converted scandium oxide into the sulphate. I give in a third column the percentage of oxide in sul- phate : .3379 grm. Sc 2 3 gave .9343 grm. Sc 2 (SO 4 ) 3 . 36.166 per cent. .3015 .8330 36.194 .2998 " .8257 " 36.187 " .3192 " .8823 " 36-178 " Mean, 36.181, db .004 Hence Sc == 43.758. Combining the. two series, we have Cleve 36.530, =b .0175 Nilson 36. 1 8 1 , .0040 General mean 36. 190, .0039 Hence, with SO, = 79.465, .00175, Sc = 43.784, .0085. If = 16, Sc 44.118. As between the two values found, the presumption is in favor of the lower. The most obvious source of error would be the presence in the scandia of earths of higher molecular weight. *Compt. Rend., 89, 419. fCompt. Rend., 91, 118. YTTRIUM. 355 YTTRIUM. All the regular determinations of the atomic weight of yttrium depend upon analyses or syntheses of the sulphate. A series of analyses of the oxalate, however, by Berlin,* is sometimes cited, and the data are as fol- lows. In three experiments upon the salt Yt/C 2 4 ) 3 3H,0 the subjoined percentages of oxide were found : 45-70 45-^5 45-72 Mean, 45.69, dz .0141 Hence with = 15.879 and C = 11.920, Yt == 88.943. Ignoring the early work of Berzelius,f the determinations to be con- sidered are those of Popp, Delafontaine, Bahr and Bunsen, Cleve, and Jones. Popp t evidently worked with material not wholly free from earths of higher molecular weight than yttria. The yttrium sulphate was dehy- drated at 200 ; the sulphuric acid was then estimated as barium sul- phate, and after the excess of barium in the filtrate had been removed the yttrium was thrown down as oxalate and ignited to yield oxide- The following are the weights given by Popp : Sulphate. BaSO. Y^O 3 . H. 2 O. 1.1805 grm. *-3 l 45 g rm - -4742 grni. .255 grm. 1.4295 1.593 " -5745 " -308 " .8455 " .9407 " .3392 " .1825 " 1.045 " 1.1635 " .4195 " .2258 " Eliminating water, these figures give us for the percentages of Yt 2 3 in Yt 2 (SOj 3 the values in column A. In column B I put the quantities of Yt 2 3 proportional to 100 parts of BaS0 4 : A. B. 51.237 36.075 51.226 36.064 51.161 36.058 51-209 36.055 Mean, 51.208, .on Mean, 36.063, .003 From B, Yt = 101.54. The values in A will Be combined with similar data from other experimenters. * Forhandlingar ved de Skaiidinaviske Naturforskeres, 8, 452. 1860. f lyehrbuch, V Aufl., 3, 1225. I Ann. Chem. Pharm., 131, 179. 1804. 356 THE ATOMIC WEIGHTS. In 1865 Delafontaine* published some results obtained from yttrium sulphate, the yttrium being thrown down as oxalate and weighed as oxide. In the fourth column I give the percentages of Yt 2 3 reckoned from the anhydrous sulphate : Sulphate. Yt 2 <9 3 . H.jp. Percent. ] 9545 S rm - .371 grm. .216 grm. 50.237 2.485 " .9585 " .565 " 49.9 22 2.153 " .827 4935 " 49.834 Mean, 49.998, =b .081 In another paper f Delafontaine gives the following percentages of Yt. 2 3 in dry sulphate. The mode of estimation was the same as before : 48.23 48.09 48-37 Mean, 48.23, .055 Bahr and Bunsen, J and likewise Cleve, adopted the method of con- verting dry yttrium oxide into anhydrous sulphate, and noting the gain in weight. Bahr and Bunsen give us the two following results. I add the usual percentage column : Yt. 2 3 . Yt^SO^ Percent. F/ 2 6> 3 . .7266 grm. L4737 grm. 49.34 .7856 " L5956 " 49-235 Mean, 49.2695, .0233 Cleve's first results are published in a joint memoir by Cleve and Hoeglund, and are as follows : Percent. 1. 4060 grm. 2. 8925 grm. 48.608 1.0930 " 2.2515 " 48.545 1.4540 " 2.9895 " 48.637 1.3285 " 2.7320 " 48.627 2.3500 " 4-833 " 48.624 2.5780 " 5.3055 " 48.591 Mean, 48.605, =h .0096 In a later paper Cleve || gives syntheses of yttrium sulphate made with yttria, which was carefully freed from terbia. The weights and percent- ages are as follows : *Ann. Chem. Pharm., 134, 108. 1865. t Arch. Sci. Phys. et Nat. (2), 25, 119. 1866. J Ann. Chem. Pharm., 137, 21. 1866. g K. Svenska Vet. Akad. Handlingar, Bd. i, No. S. 1873. |j K. Svenska Vet. Akad. Handlingar, No. 9, 1882. See also Bull. Soc. Chim., 39, 120. 1883. YTTRIUM. 357 yt. 2 O 3 . y/ 2 (S0 4 ) 3 . Percent. Yt. t O z . .8786 1.8113 48/507 .8363 .7234 48.526 .8906 .8364 48.497 .7102 .4645 48.494 .7372 .5194 48.519 .9724 .0047 48.506 .9308 .9197 48.487 .8341 .7204 48.483 1.0224 2.1073 48.5*7 .9384 i.934i 48.519 9744 2.0093 48.494 1.53*4 3.1586 48.484 Mean, 48.503, .0029 Hence Yt = 88.449. The y ttria studied by Jones* had been purified by Rowland's method- thai is, by precipitation with potassium ferrocyanide and certainly con- tained less than one-half of one per cent, of other rare earths as possible impurities. Two series of determinations were made one by ignition of the sulphate, the other by its synthesis. The results were as follows, with the usual percentage column added : First Series. Syntheses. Yt^Oy Yt^SO^. Percent. Yt 2 O s . .2415 .4984 48.455 .41 12 .8485 48.462 .2238 .4617 48.473 3334 .6879 48.466 .3408 .7033 48.457 .3418 .7049 48.489 .2810 .5798 48.465 .3781 .7803 48.456 4379 .9032 48.483 .4798 .9901 48.460 Mean, 48.467, .0025 Second Series. Analyst. 58. Yt z (SO 4 ) 3 . Yt,0,. Percent. Yt. 2 O 3 . .5906 .2862 48.459 .4918 .2383 48.455 .5579 .2705 48.485 .6430 .3"7 48.478 .6953 .3369 48.454 1.4192 .6880 48.478 .8307 .4027 48.477 .7980 .3869 48.484 .8538 .4*39 48.477 1.1890 .5763 48.469 Mean, 48.472, .0024 * Anier. Chem. Journ., 17, 154. 1895. 358 THE ATOMIC WEIGHTS, From syntheses Yt = 88. 287 From analyses " = 88.309 These data of Jones were briefly criticised by Delafontaine,* who re- gards a lower value as more probable. In a brief rejoinder f Jones defended his own work; but neither the attack nor the reply needs farther consideration here. They are referred to merely as part of the record. For the percentage of yttria in the sulphate we now have eight series of determinations, to be combined in the usual way : Popp 51.208, rb .01 10 Delafontaine, first 49,998, rb .0810 Delafontaine, second 48.230, .0550 Bahr and Bunsen 49.2695, rb .0233 Cleve, earlier 48.605, db .0096 Cleve, later 48.503, .0029 Jones, syntheses 48.467, rb .0025 Jones, analyses 48.472, rb .0024 General mean 48.532, rb .0015 Hence, if = 15.879, .0003, and S = 31.828, .0015, Yt = 88.580, rb .0053. If = 16, Yt = 89.255. If only the four series by Cleve and by Jones are considered, the mean percentage of yttria in the sulphate becomes 48.481. Hence Yt = 88.350, or, with = 16, 89.023. This result is preferable to that derived from all the data, for it throws out determinations which are certainly erroneous. Cleve's early series might also be rejected, but its influence is insignificant. *Chem. News, 71, 243. fChem. News, 71, 305. SAMARIUM, GADOLINIUM. ETC. SAMARIUM, GADOLINIUM, ERBIUM, AND YTTERBIUM. The data relative to the atomic weights of these rare elements are rather scanty, and all depend upon analyses or syntheses of the sul- phates. SAMARIUM. Atomic weight given by Marignac,* without details, as 149.4, and by Brauner,f as 150.7 in maximum. The first regular series of determina- tions was by Cleve, J who effected the synthesis of the sulphate from the oxide. Data as follows : Sm 2 O 3 . Sm.i(SOJ 3 . Per cent. 1.6735 2.8278 59- l8 i.97o6 3.3301 59- '75 I.II22 1.8787 59-201 1.0634 1.7966 S9- 1 9 .8547 1.4440 59.J90 7447 1-2583 59-183 Mean, 59.1865, .0025 Hence Sm = 149.038. Another set of determinations by Bettendorff, after the same general method, gave as follows: Sm.,O. A . Sm^SO^. Per cent. Sm. 2 O 3 . 1.0467 1.7675 59-219 1.0555 1.7818 59. 2 38 1.0195 1.7210 59.225 Mean, 59,227, .0038 Hence Sm = 149.328. v Combining the two series, we have Cleve .................................. 59. 1865, = .0025 Bettendorff ............................. 59.227, .0038 General mean 59.1 99, =h .002 r Hence, if S0 3 = 79.465, .00175, Sm = 149. 127, =b .01 15. If 0=16, Sm = 150.263*. According to Demarcay.|| samaria contains an admixed earth whose properties are yet to be described. * Arch. Sci. Phys. et Nat. (3), 3, 435. 1880. t Journ. Chem. Soc., June, 1883. 1 Journ. Chem. Soc., August, 1883. Conipt. Rend., 97, 94. gAnn. Chem. Pharm., 263, 164. 1891. j| Compt. Rend., 122, 728. 1896. 360 THE ATOMIC WEIGHTS. GADOLINIUM. This element, discovered by Marignac, must not be confounded with the mixture of metals from the gadolinite earths to which Nordenskiold gave the same name. Several determinations of its atomic weight have been made, but Bettendorff's only were published with proper details.* He effected the synthesis of the sulphate from the oxide, and his weights were as follows. The percentage of Gd 2 O 8 in Gd 2 (SOJ 3 is given in the third column : Gd. 2 O. A . Gd^SO^. Percent. G 1.0682 1-7779 60.082 1.0580 1.7611 60.076 1.0796 1.7969 60.081 Mean, 60.080, .0013 Hence, with S0 3 = 79.465, Gd = 155.575. If = 16, Gd 156.761, Boisbaudranf found Gd = 155.33, 156.06, 155.76, and 156.12. The last he considers the best, but gives no details as to antecedent values. He also quotes Marignac, who found Gd 156.75, and Cleve, who found 154.15, 155.28, 155.1, and 154..77. Probably these all depend upon S0 3 = 80. ERBIUM. Since the earth which was formerly regarded as the oxide of this metal is now known to be~a mixture of two or three different oxides, the older determinations of its molecular weight have little more than historical interest. Nevertheless the work done by several investigators may prop- erly be cited, since it sheds some light upon certain important problems. First, Delafontaine's J early investigations may be considered. A sul- phate, regarded as erbium sulphate, gave the following data. An oxalate was thrown down from it, which, upon ignition, gave oxide. The per- centages in the fourth column refer to the anhydrous sulphate. In the last experiment water was not estimated, and I assume for its water the mean percentage of the four preceding experiments : Sulphate. r 2 O 3 . ff. 2 O. Per cent. Er z O 3 . .827 grm. .353 grm. .177 grm. 54. 308 1.0485 " .4475 " .226 " 54.407 .803 " .3415 " .171^ " 54.035 1.232 " .523 " .264" " 54.028 1.1505 " .495 " 54-76o Mean, 54.308, zb .0915 Hence Er = 117.86. * Ann. Chem. Pharm., 270, 376. 1892. t Compt. Rend., in, 409. 1890. J Ann. Chem. Pharm., 134, 108. 1865. ERBIUM, YTTERBIUM, ETC. 361 Bahr and Bunsen * give a series of results, representing successive puri- fications of the earth which was studied. The final result, obtained by the conversion of oxide into sulphate, was as follows : .7870 grm. oxide gave 1.2765 grm. sulphate. 61.653 P er cent, oxide. Hence Er = 167.82. Hoeglund, f following the method of Bahr and Bunsen, gives these results : Er. 2 O,. Er^(SO^. A . Per cent. Er. 2 O 3 . 1 .8760 grm. 3.0360 grm. 61.792 1.7990 " 2.9100 " 61.821 2.8410 " 4-5935 " 61.848 1.2850 " 2.0775 " 61.853 1.1300 " 1.827 " 61.850 .8475 " r -37Q " 61.861 ' Mean, 61.8375, .0063 Hence Er = 169.33. According to Thalen,t spectroscopic evidence shows that the " erbia " studied by Hoeglund \vas largely ytterbia. Humpidge and Burney give data as follows : 1.9596 grm. Er 2 (SO 4 ) 3 gave 1.2147 g rm - Er. 2 O 3 . 61.987 per cent. 1.9011 " 1.1781 " 61.965 " Mean, 61.976, -0074 Hence Er= 170.46. The foregoing data were all published before the composite nature of the supposed erbia was fully recognized. It will be seen, however, that three sets of results were fairly comparable, while Delafontaine evidently studied an earth widely different from that investigated by the others. Since the discovery of ytterbium, some light has been thrown on the matter. The old erbia is a mixture of several earths, to one of which, a rose-colored body, the name erbia is now restricted. For the atomic weight of the true erbium Cleve || gives three determinations, based on syntheses of the sulphate after the usual method. His weights were as follows, with the percentage ratio added : Er. 2 O 3 . Er.i(SO^ y Per cent. Er. 2 O 3 . 1.0692 1.7436 61.321 1.2153 1.9820 61.317 .7850 1.2808 61.290 Mean, 61.309, d= .0068 Hence, with S0 3 = 79.465, Er == 165.059. If =16, Er= 166.316. *Ann. Chem. Pharm., 137, 21. 1866. fK. SvenskaVet. Akad. Handlingar, Bd. i, No. 6. I Wiedemann's Beibliitter, 5, 122. 1881. # Journ. Chem. Soc., Feb., 1879, p. 116. || K. Svensk. Vet. Akad. Handlingar, No. 7, 1880. Abstract in Compt. Rend., 91, 382. 362 THE ATOMIC WEIGHTS. It is not worth while to combine this result with the earlier determi- nations, for they are now worthless. YTTERBIUM. For ytterbium we have one very good set of determinations by Nilson.* The oxide was converted into the sulphate after the usual manner : 1.0063 g rm - 1.0139 " .8509 " .7371 " 1.0005 " .8090 " 1.0059 " Percent. Y&. 2 O 3 . .6186 giro. 62.171 .6314 " 62.149 .3690 " 62.155 .1861 " 62.145 .6099 " 62.147 .3022 " 62.126 .6189 <( 62.134 Mean, 62.147, .0036 Hence, with S0 3 = 79.465, Yb = 171.880. If O = 16, Yb = 173.190. TERBIUM, THULIUM, HOLMIUM, DYSPROSIUM, ETC. For these elements the data are both scanty and vague. Concerning the atomic weights of holmium and dysprosium, practically nothing has been determined. To thulium, Clevef assigns a value of Tm = 170.7, approximately, but with no details as to weighings. Probably the value was computed with S0 3 = 80. For terbium, ignoring older determinations, Lecoq de Boisbaudran has published two separate estimates.]! First, for two preparations, one with a lighter and one with a darker earth, he gives Tb = 161.4 and 163.1 respectively. In his second paper he gives Tb = 159.01 to 159.95. These values probably are all referred to S0 3 = 80. *Compt. Rend., 91, 56. 1880. Berichte, 13, 1430. t Compt. Rend., 91, 329. 1880. J Compt. Rend., 102, 396, and in, 474. ARGON AND HELIUM. 363 ARGON AND HELIUM. The true atomic weights of these remarkable gases are still in doubt, and so far can only be inferred from their specific gravities. For argon, the discoverers, Rayleigh and Ramsay,* give various deter- minations of density, ranging, with hydrogen taken as unity, from 19.48 to 20.6. In an addendum to the same paper, Ramsay alone gives for the density of argon prepared by the magnesium method the mean value of 19.941. In a later communication f Rayleigh gives determinations made with argon prepared by the oxygen method, and puts the density at 19.940. For the density of helium, Ramsay J gets 2.18, while Langlet finds the somewhat lower value 2.00. From one set of physical data both gases appear to be m on atomic, but from other considerations they are supposably diatomic. Upon this question controversy has been most active, and no final settlement has yet been reached. If diatomic, argon and helium have' approximately the atomic weights two and twenty respectively; if monatomic, these values must be doubled. In either case helium is an element lying be- tween hydrogen and lithium, but argon is most difficult to classify. With the atomic weight 20, argon falls in the eighth column of the periodic system between fluorine and sodium, but if it is 40 the position of the gas is anomalous. A slightly lower value would place it between chlorine and potassium, and again in the eighth column of Mendelejeff's table; but for the number 40 no opening can be found. It must be noted that neither gas, so far, has been proved to be abso- lutely homogeneous, and it is quite possible that both may contain ad- mixtures of other things. This consideration has been repeatedly urged by various writers. If argon is monatomic, a small impurity of greater density, say of an unknown element falling between bromine and rubid- ium, would account for the abnormality of its atomic weight, and tend towards the reduction of the latter. If the element is diatomic, its classi- fication is easy enough on the basis of existing data. Its resemblances to nitrogen, as regards density, boiling point, difficulty of liquefaction, etc., lead me personally to favor the lower figure for its atomic weight, and the same considerations may apply to helium also. Until further evidence is furnished, therefore, I shall assume the values two and twenty as approximately true for the atomic weights of helium and argon. * Phil. Trans., 186, pp. 220 to 223, and 238. 1,895. fChem. News, 73, 75. 1896. JJourn. Chem. Soc., 1895, p. 684. \ Zeitsch. Anorg. Chem., 10, 289. 1895. 364 THE ATOMIC WEIGHTS. TABLE OF ATOMIC WEIGHTS. The following table contains the values for the various atomic weights found or adopted in the preceding calculations. As the table is intended for practical use, the figures are given only to the second decimal, the third being rarely, if ever, significant. In most cases even the first deci- mal is uncertain, and in some instances whole units may be in doubt. H = i. 0=16. Aluminum 26.91 27.11 Antimony 11 9-S 2 I2O 43 Argon ? ? Arsenic 74-44 75- 01 Barium 136.39 r 37-43 Bismuth 206. 54 208. 1 1 Boron 10.86 10.95 Bromine 79-34 79-95 Cadmiu'm ni.io lll -95 Caesium l 3 l -&9 132.89 Calcium 39-76 40.07 Carbon..., 11.92 12.01 Cerium I39-IO 140.20 Chlorine 35 .18 35-45 Chromium 5!-74 5 2 - J 4 Cobalt 58.49 58.93 Columbium 93-Q2 93-73 Copper 63. 12 63.60 Erbium 165.06 166.32 Fluorine 18.91 19.06 Gadolinium , 155-57 156.76 Gallium 69.38 69.91 Germanium 7 r -93 72.48 Glucinum 9.01 9.08 Gold 195-74 i97- 2 3 Helium ? ? Hydrogen .... i.ooo 1.008 Indium 112.99 Ir 3-^5 Iodine 125.89 12685 Iridium 191.66 '93- 12 Iron.. 55-6o 56.02 Lanthanum T 37-59 138.64 Lead , 205.36 206.92 Lithium 6.97 7.03 Magnesium 24.10 24.28 Manganese 54-57 54-99 Mercury 198.49 200.00 Molybdenum 95. 26 95-99 Neodymium 139-7 140.80 Nickel 58.24 58.69 TABLE OF ATOMIC WEIGHTS. 365 Nitrogen ........................... r 3-93 1 4>4 Osmium ............... ............ J ^>9-SS I 9-99 Oxygen ........................... 15.88 16.00 Palladium .......................... IO 5-56 106.36 Phosphorus ......................... 30.79 3 J .O2 Platinum ........................... 193-41 l 94-%9 Potassium ................... ....... 38.82 39. 1 1 Praseodymium ................ ..... 142.50 143.60 Rhodium ........................... 102.23 103.01 Rubidium.. ....................... 84.78 85.43 Ruthenium ......................... 100.91 101.68 Samarium .......................... I 49- I 3 150.26 Scandium ..... ...... .............. 43-78 44-12 Selenium ........................... 78.42 79.02 Silicon ............................. 28. 1 8 28.40 Silver .............................. 107. 1 1 107.92 Sodium ............................ 22.88 23.05 Strontium ......................... 86.95 87.61 Sulphur ............................ 31.83 32.07 Tantalum .......................... 181.45 182.84 Tellurium .......................... 126.52 127.49 Terbium ........................... 158.80 160.00 Thallium ................. ........... 202.61 204.15 Thorium .......................... 230.87 232.63 Thulium ........... ................ 169.40 170.70 Tin ..... , ...... ................. 118.15 119.05 Titanium ........................... 47-79 48. 1 5 Tungsten ......................... J83-43 ^4. 83 Uranium ........................... 237.77 239.59 Vanadium ........................ 5-99 5 r -38 Ytterbium ......................... 171.88 ! 73-i9 Yttrium ........ . ................... 88.35 89.02 Zinc. ... ........................... 64.91 65.41 Zirconium .......................... 89. 72 90.40 INDEX TO AUTHORITIES. Agamennone X 4> 2 5 Allen 89 Allen and Pepys 24 AlibegofF 266, 300 Anderson 130 Andrews 1 18, 327 Arago 24, 58, 72 Arfvedson 84, 263, 282 Aston 52, 172 Awdejew , . . . ; 132 Bahr 137 Bahr and Bunsen 356, 361 Bailey 197, 231 Bailey and L,amb. 316 Balard 44 Baubigny 93, 148, 180, 244, 300 Bauer 349, 353 Becker I Beringer 335 Berlin 238,251,355 Bernoulli 257 Berzelius. . 5,8, 24, 34, 38, 43, 44, 50, 58, 72, 82, 84,91, IOI, IIO, 112, 121, 123, 127, 132, 135, 146, 171, 176, iSS, 196, 204, 209, 211, 213, 2l6, 236, 238, 250, 255. 263, 268, 271, 277, 282, 257, 313, 315, 322, 325, 327, 355 Bettendorff 349, 359, 360 Biot and Arago 24, 58, 72 Blomstrand 234, 236 Boisbaudran 181, 360, 362 Bongartz 226 Bongartz and Classen 200, 201 Borch, von 256 Boussingault 24, 58 Brauner. .272, 274, 340, 342, 348, 351, 359 Breed 320 Bucher 1 60 Buehrig ... 339 Buff 24, 72 Bunsen 87, 89, 356, 361 Bunsen and Jegel 336 Burney 361 Burton 151 Burton and Vorce , . . . 142 Capitaine 287 Cavendish 24 Chikashige 275 Choubine 344 Christensen 280 Chydenius 204 Clark 9 Clarke 159 Classen 200, 201, 231, 232 Claus . . 31 r Cleve. . 206, 347, 348, 351, 352, 354, 356' 359, 360, 361, 362 Cleve and Hoeglund 356 Commaille 91 Cooke 27, 8r, J57, 221, 222, 224 Cooke and Richards 13 Crafts 25, 58 Crookes 185 Czudnowicz. 211, 346 Davy 24 Debray 133, 251 Delafoutaine 205, 356, 358, 360 De Luca 278 Demarcay 359 Detnoly 191 De Saussure 24, 72 Desi 262 Deville 291 Deville and Troost 235 Dewar and Scott _ , 283 Dexter 217 Diehl 84 Dittmar 85 Dittmar and Henderson 12, 19 Dittmar and M' Arthur 333 Dobereiner 127, 287 Duloug and Berzelius 8, 24, 58, 72 Dumas. . 9, 39, 45, 50, 51, 72, 80, 91, no^ 112, 113, 119, 129, 140, 156, 176, 188, 199, 201, 209, 213, 217, 229, 251, 256, 269, 278, 279, 282, 289, 294 (367) 368 THE ATOMIC WEIGHTS. Dumas and Boussingault. . 24, 58 Dumas and Stas 76 Ebelmen 264 Ekman and Pettersson 269 Erdmanii 146 Erdmann and Marchand. ..IT, 76, 110, in, 166, 268, 288, 291 Erk 347, 351 Ewan and Hartog 171 Faget 37 Favre 147 Fourcroy 24 Fownes 72 Fremy 322 Friedel 77 Gay-Lussac 32, 135, 146 Genth 338 Gerhardt 36 Gibbs 298, 342 Gladstone and Hibbard . . 152 Ginelin 84 Godeffroy 87, 90 Gooch and Rowland 274 Gray 98 Hagen 84 Halberstadt 330 Hampe 92 Hardin 34, 63, 74, 163, 167 Hartog 171 Hauer, von 156, 271, 283 Hebberling 184 Hempel and Thiele 308 Henry 6 Hermann 84, 196, 206, 234, 236, 335, 347, 351 Heycock 88 Hibbard 152 Hibbs 67, 68, 215 Hinrichs 6 Hoeglund 356, 361 Holzmann 345 Hoskyns-Abrahall 171 Howland 274 Humboldt and Gay-Lussac 32 Humpidge and Burney 361 Huntington 46, 157 Isnard 176 Jacquelain 136, 146, 238 Jegel 336 Johnson 17 Johnson and Allen 89 Jolly 59 Joly 311, 326 Joly and Leidie 319 Jones 159, 3^7, 358 Jorgensen 313 Keiser 15, 150, 316 Keiser and Breed* 320 Keller and Smith 318 Kemp 252 Kessler 214. 216, 218, 224, 241, 242 Kirwan 24 Kjerulf 336 Klatzo 132 Kobbe 314 Kralovanzky 84 Kriiss 102 Kriiss and Alibegoff 266, 300 Kriiss and Moraht 133 Kriiss and Nilson . 207 Kriiss and Schmidt 303 Lagerhjelm 229 Lamb 316 Lamy 184 Langer 301 Langlet 363 Laurent . 34, 171 Laurie 103 Lavoisier 24, 58, 72 Le Conte 25 Leduc 20, 27, 32, 59, 78 Lee 298 Lefort 240 Leidie 319 Lenssen 156 Lepierre 186 Levol 102 Liebig 44 Liebig and Redtenbacher 72 Liechti and Kemp ... 252 Lougchamp ,. . 127, 135 INDEX TO AUTHORITIES. 369 Lorimer and Smith 159 Louyet 277, 279, 280 Lowe 231 Lowig 44 M Maas 252 M'Arthur 333 Macdonnell I3 6 Malaguti 255 Mallet 84, 105, 150, 177 Marchand 1 1 , 72, 76, 1 10, 1 1 1, 166, 256, 263, 268, 288, 291 Marchand aud Scheerer 138 Marignac. . 34, 35, 36, 3 8 , 39, 4i, 43, 44, 45, 47, 48, 49, 60, 62, 65, 74, no, 114, II.5, Il8, 121, 122, 123, 129, 141, 148, 196, 230, 235, 236, 284, 292, 336, 344, 345, 35i, 359, 3 6 o. Mather 176 Maumene 34, 36, 39, 43, 75, 288 Meineke 244 Meyer... , 252 Meyer and Seubert i, 5, 6 Millon 48, 167 Millon and Commaille 91 Mitscherlich 72 Mitscherlich and Nitzsch 268 Moberg 239 Moissan 278, 279, 280 Mond, Langer, and Quincke 301 Morabt 133 Morley 12, 21, 27, 32 Morse and Burton 151 Morse and Jones 159 Morse and Reiser 150 Mosander 190, 335, 344, 351 Mulder 6 Mulder and Vlaanderen 199 N Nilson 207, 354, 362 Nilson and Pettersson 133 Nitzsch 268 Nordenfeldt 137 Nordenskiold 360 Norlin 287 Noyes 16, 17 Ostwald i, 6, 57, 71, 83, 131 Oudemans , 6 24 Parker 142 Partridge 157 Peligot 238, 264, 265 Pelouze 35, 51,60, 113, 118, 188, 209, 213 Penfield 186 Pennington and Smith 258 Penny 35, 39, 5o, 62, 64, 66, 67 Pepys 24 Persoz 257 Petrenko-Kritscbenko 319 Petterssou 133, 269 Pfeifer 225 Piccard 87 Pierre 191 Pollard . . 252 Popp 355 Popper 225 Q Quincke 301 Quintus Icilius 315 Rammelsberg. . . 234, 252, 263, 337, 344 Ramsay 149, 363 Ramsay and Aston 52, 172 Rawack 283 Rawson 244 Rayleigh. . 14, 16, 25, 26, 58, 59, 98, 363 Rayleigb nnd Ramsay 363 Rayleigh and Sidgewick 98 Redtenbacher 72 Regnault 24, 25, 72 Reich and Ricbter 182 Remmler 302 Reynolds and Ramsay 149 Richards 13, 46, 82, 92, 93, 94, 96, 97, 115, 119, 121, 123, 124, 154 Richards and Parker 142 Richards and Rogers 141, 152, 153 Riche 259 Richter . . . , 182 Rimbacb 1 74 Rivot 289 Robinson 340 Rogers 141, 152, 153 Roscoe 77, 211, 257, 262 Rose 190, 217, 234, 236 Rothhoff 291 Russell , 294, 295 370 THE ATOMIC WEIGHTS. Sacc 268 Salvetat 1 10, 1 13, 1 18 Scheerer 135, 136, 138, 139 Scheibler 260 Schiel 188 Schmidt 203, 303 Schneider 216, 224, 229, 232, 255, 258, 282, 291, 292, 297 Schrotter 209 Schutzeuberger 301 , 34 1 Scott 3 2 , 283 Sebelien 1,7 Sef strom ...* 166 Seubert i, 322, 323, 325, 328, 333 Seubert and Kobbe 314 Seubert and Pollard 252 Shaw 98 Shinn 259 Sidgewick 98 Siewert , 243 Smith 159, 258, 318 Smith and Desi 262 Smith and Maas 252 Sommaruga 297 Spring 6 Stas. . 6, 37, 38, 40, 41, 42, 44, 45, 47, 48, 49, 5i, 52, 57, 61, 62, 64, 65, 66, 71, 73, 76, 78, 80, 82, 83, 85, 128, 130, 131 Staudenmaier 274 Strecker 73 Stromeyer 84, 156, 287 Struve 81, 82, 123, 250 Svanberg 130, 167 Svanberg and Nordenfeldt 137 Svanberg and Norlin 287 Svanberg and Struve. 82, 250 Terreil 177 Thalen 361 Thiele 308 Thomsen. . 13, 22, 30, 57, 69, 71, 83, 131 Thomson 24, 58 Thorpe 192 Thorpe and Laurie 103 Thorpe and Young ... 1 89 Tissier 176 Torrey 151, 180,289 Troost 84, 235 Turner.. 38, 64, 121, 122, 123, 128, 166, 167, 282 Unger 221 Van der Plaats. . . 6, 57, 71, 77, 83, 131, 149, 200, 210 Van Geuns . . 5 Vanni 98 Vauquelin 24 Vlaanderen 199 Vogel 6 Vorce 142 W Wackeuroder. ... 287 Waddell 258 Wallace 44, 213 Warrington 33 Weber 217 Weeren 132, 285 Weibull 196 Wells and Penfield 186 Welsbach 353 Wertheim 265 Werther 184 Weselsky 298 * Wildenstein 241 Wills 271 Wing 338 Winkler. . . . 182, 195, 297, 305, 306, 307 Wolf. 337, 340 Woskresensky 72 Wrede 24, 72 Young 189 Zettuow 260 Zimmermann 266, 300 Zschiesche 347, 351 DATE DUE SLIP UNIVERSITY OF CALIFORNIA MEDICAL SCHOOL LIBRARY THIS BOOK IS DUE ON THE LAST DATE STAMPED BELOW 2m-ll,"29