138 ROBINSON AND BRISCOE: XXPI1.-A Redetermination of the Atomic Weight of The Inseparability of the Isotopes by Bromine. Fractional Crystallisation. By PERCY LUCOCK ROBINSON and HENRY VINCEPU'T AIRD ERISCOE. THE only precise experimental evidence in support of the generally accepted view that isotopes are inseparable by fractional crystallis-ation is the attempt by Richards and Hall ( J . Amer. Chem. Xoc., 1917 39 531) to separate thus the isotopic forms present in lead derived from Australian carnotite. Both they and Soddy (J., 1911 99 72; J . Amer. Chem. SOC. 1917 39 1614) refer to the earlier data which need not be quoted here. Richards and Hall obtained no evidence of separation by 904 effective crystallisations but their lead presented a relatively unfavourable case since it contained only about 20y0 of the heavier isotope and the nitrates of the isotopes differed in solubility by but o*7470 (Richards and Schumb J .Anzer. Chem. SOC. 1918, 40,1403). Since Aston's work has disclosed the existence of isotopes of many of the lighter elements it has become clear that cases may be found permitting a more precise test (or proof) of inseparability by fractional crystallisation. The isotopes of boron differ by 10% in atomic weight but the isotope ratio is unfavourable (approxim-ately 20% BlO and 80% Bl1) and an accurate determination of the atomic weight is both difficult and laborious. A prolonged frac-tionation of boric acid was effected but the determinations of atomic weight although they had other unexpected features o€ interest (see following paper) proved valueless for the purpose now under discussion.Bromine in the form of ammonium bromide seems to present the best test-case because (1) bromine has but two isotopes (Br79 and BrSl) present in nearly equal proportions and differing by 2.2y0 in atomic weight (2) hydrogen and nitrogen are " simple " elements and together form less than 20% of the salt (3) the atomic weight of bromine can be determined by direct and trustworthy methods and (4) ammonium bromide crystallises from water in a fashion convenient for a long series of fractionations. Therefore with the double object of making a re-determination of the atomic weight of bromine and of confirming in this more favour-able case the finding of Richards and Hall ammonium bromide was subjected to prolonged fractional crystallisation and the atomic weight of the bromine in the final head and tail fractions was deter-mined by measurement of the ratio Ag AgRr A EEDETERMIXATIOX O F THE ATORlIC WEIGHT O F SROXISE.130 Pu?<Jicatio.rz of Reagmts. T17cter.-Laboratory distilled water already free from halogen, was redistilled from a little caustic soda and potassium perman-ganate in a 10-gallon copper still through a vertical spray-trap packed with glass beads and condensed in pure tin. The first and last portions of the distillate were rejected anif the main fraction was collected and stored (for but a short time) in 'O-litre stoppered resistance glass bottles which had been n-ell cleaned steamed out and kept for months full of distilled water.Many iwphclometric tests were made but on no occasion did the water give any indication of chloride. Ammo?zin.-Ammonia gas liberated by warming the purest commercial ammonia (d = O.SSU) was dissolved in pure water, in an apparatus constructed entirely of rcsistance glass with cz ground-in leading tube. SuZpJzur Dioxide.-The gas used both for the conversicii of bromine to hydrobromic acid and for the precipitation cf silver by Stas's method was aln-ays obtained from the middle fraction of a syphon of the liquid. n'ilric Acid.-Commercial nitric acid containing o d y small traces of halogen was thrice redistilled in an apparatus constrwted of '' Duro " resistance glass with ground joints large head and tail fractions being rejected. The final main fraction tested in the ncyhelometer was free from halogens.It was kept in a resistance glass bottle having c well-ground glass stopper protected by a glass cap. Formic Acid.-The purest obtainable reagent acid was twice redistilled from a fused silica ffask having a ground-in silica con-denser and adapter. Considerable head and tad fractions were rejected in each distillation and the main fraction n-as preserved i s a stoppered silica flask. il~ag?zesin.-~lagiiesiulrz nitrate was thrice precipitated from saturated aqueous solution by excess of nitric acid filtered on porcelain dissolvcd in water and precipitated with ammonia. The hydroxide was washed by decantation filtered dried a t lOO", ignited in air in a platinum dish and finally heated at about 1000" in a current of moist pure hydrogen.Magnesia boats moulded from the pure oxide nioisteried with dilute nitric acid were ignited and used on a silica plate. Hydrochloric Acid-Coinmercial reagent acid was freed from arsenic by treatment with a copper-tin couple and distillation from copper gauze according to the method of Thorne and Jefiers Zi?ic.-A sample of granulated electrolytic zinc kindly supplied ( d ? ~ d y s t 1906 31 101). F" 140 ROBINSON AND BRISCOE : by Messrs. Brunner Mond & Co. Ltd. proved t'o be free from arsenic and was used without further purification. Hydrogen.-This gas prepared from pure hydrochloric acid, containing a little platinic chloride and pure zinc in a Kipp's apparatus passed to a purifying train the several parts of which were sealed together in which it traversed successively a 12 inch column of concentrated potassium hydroxide solution two 12 inch columns packed with broken sticks of pot'assium hydroxide and two 10 inch columns packed with phosphorus pentoxide distributed on glass wool.XiZwer.-Two samples " A " and " B," were used prepared in the same way but at different times and from different samples of commercial silver nitrate. In each case a filtered aqueous solution of silver nitrate (200 g.) was precipitated in a volume of about 20 Iitres with ammonium bisulphite according to t'he method of Stas (Briscoe J. 1915 107 69). The precipitated silver was washed six times with dilute ammonia allowed to stand over-night with ammonia washed six times with water and dried at 100".This silver in 100 g. portions was dissolved in pure nitric acid, precipitated hot with ammonium formate in a bulk of 2 litres, well washed with ammonia and water and dried first a t 100" and finally at 250" in an electrically heated covered beaker. Before weighing the silver was melted on a boat of pure magnesia in an atmosphere of hydrogen in an electrically heated silica tube furnace. The hydrogen issuing from the furnace during melting contained no impurities detectable by Marsh's test or by smell. At the end of each fusion the silver was collected into large buttons by shaking the furnace and these when cool were etched with 1 1 nitric acid well washed with water heated at 250" for 12 hours, and cooled in a desiccator over solid potassium hydroxide.The Determinatioia of Weight. All precise weighings were made on an Oertling balance con-structed specially for this work resembling in many respects the standard type known as " No. 7 S.W.," but with a beam of Firth's 36% nickel-steel having a low coefficient of expansion pan-supports of special design and a separate enclosure for the beam after the principle used by Manley (Phil. Trans. 1910 210 A 387). No proper situation giving constant temperature and freedom from vibration was available for the balance it was used on an ordinary stout wooden table in a room subject to vibration and to consider-able and rapid fluctuations of temperature. Even under these adverse conditions it gave results sufficiently precise and consistent for this work a remarkable performance which is undoubtedl A REDETERMINATION OF THE ATOMIC WEIGKT OF BRONINE.141 attributable to the special features of construction indicated above. The sensitiveness of the balance increased very slightly with the load and during all the mork here described was very nearly 100 scale divisions per milligram. In weighing by the method of oscil-lations the zero could easily be determined to the nearest half-division and the apparent weight thus ascertained with an error not exceeding 0.00001 g. A set of gold-plated brass weights with platinum fractions by Oertling and a gold 5-mg. rider were used and were calibrated to ascertain the relative weights in air on three separate occasions before and during the weighings. The corrections applied to ascertain relative weights in a vacuum were calculated using the density of air at the temperature and pressure observed during the weighing and the densities 10.49 and 6.47 for silver and silver bromide respectively.As relative weights only were desired and the inequality of the arms of the balance was very small and constant all weighings were made directly. All were made in duplicate and many in triplicate in no case did the values for the vacuum weight thus obtained differ by more than 0.00003 g., thus it seems probable that the mean values arc in error by less than this amount. Silver was weighed directly on the balance pan silver bromide was weighed in a stoppered glass bottle against a tare of thc same glass and of closely similar form volume and weight which had been treated in all respects as the bottle.The balance case contained solid caustic potash and a piece of pitchblende. The Fractional Crystallisalioiz of mtio'i/iuitL Brmtiide. As starting material commercial ammonium bromide puriss., was used. Prolonged exposure to the laboratory atmosphere and contact with glass inevitably introduce impurities compared with which those originally present are insignificant ; hence no attempt was made to ascertain the nature and amount of the latter. About 2500 g. of ammonium bromide were dissolved in hot water in such proportion that about one-half crystallised out u n cooling. Each fraction was again fractionated in a similar manner until a series of 30 fractions had been built up. Thereafter the number of fractions was kept constant and fractionation was continued in the usual way by crystallising the whole series taking away the " head '' (most soluble) mother-liquor to form part of a new fraction, transferring each of the other mother-liquors to the crystals of the next higher fraction adding water to the crystals of the " tail " (least soluble) fraction and again crystallising the whole series 142 ROBINSON AND BRISCOE: A detailed scheme of such a fractionation is given by Richards and Hall (loc.cit.). I n all 80 crystallisations of the whole series were thus made * and the total number of crystallisations including those in-volved in the establishment of the series was approximately 2700. Throughout the later part of the fractionation each fraction contained about 80 g .of ammonium bromide of which one-half was transferred a t each crystallisation. The fractions were contained in 200 C.C. conical flasks closed against dust by loose hollow glass stoppers distilled water of good quality was used for all crystal-lisations. At the end of the fractionation the extreme end fract’ions, Nos. 42 and 72 were rejected and the ammonium bromide of Nos. 43,57 and 71 the ‘‘ tail,” middle and “ head ” fractions respectively, was taken for the atomic weight determinations. PuriJication of Bromine for Analysis. The volatility of bromine and its liberation from a bromide by oxidation afford an unexceptionable means of separating it sharply from all elements other than chlorine and iodine. It is therefore with these elements which would tend t o concentrate in the tail and head fractions respectively that the scheme of purification is concerned.At the same time as it was important that the determinations on head and tail fractions should be strictly com-parable both had to receive identical treatment. Bromine is usually purified from the other halogens by applying the facts that it liberates iodine from an iodide and is itself liberated from a bromide by chlorine. Whilst in ordinary analytical work the assumption that these reactions are complete and irreversible holds well enough they are probably not so in fact (see e.g. Schuyten, Chem. Ztg. 1908 32 619). The methods usually adopted for the rigorous purification of bromine for atomic weight work (see e.g., Scott J.1900 97 614; Baxter J . Amer. Chem. Xoc. 1906 28, 1322) reject so great a proportion of the material in head and tail fractions (using “head ” and tail^' here in a chemical sense) as to ensure elimination of the other halogens but were for that reason inapplicable t o the small quantities (50-60 g. of bromine in each fraction) available for purification in this case. Therefore each fraction of ammonium bromide was evaporated t o dryness dried at 160° and weighed. A quantity of pure sodium carbonate about 5% in excess of the calculated amount dissolved in a small quantity of water was added to the bromide and the * The establishment of the series and the first 30 crystallisations thereof were carried out by Mr. T. Ranby whose valuable assistance in this tedious work w e desire t o acknowledge A REDETERMIXATION OF THE ATOMIO WEIGHT OF BROMINE.143 whole was again evaporated to dryness and fused a t about 800" in platinum in an oxidising atmosphere. Thus any organic impurities were destroyed. A solution of about 1 g. of potassium dichromate and 30 C.C. of concentrated sulphuric acid in 170 C.C. cf water was boiled vigorously for 10 minutes to expel any possible trace of halogen cooled and used to dissolve the sodium bromide from the platinum dish. This solution was transferred to the flask A of the distillation apparatus shown in Fig. 1 heated and kept boiling for about 10 minutes whilst the column B jacketed with cold water acted as a reflux condenser so that the free bromine might react with any trace of iodide present.Then the cooling was discontinued and the solution boiled until the bromine had distilled over into a flask containing pure water cooled externally by ice. Finally this flask was removed so that steam passed uncondensed right through the apparatus and carried away any residual traces of free halogen. The quantityof potassium dichromate used was calculated to displace about 2% of the total bromine and the procedure described evidently favoured the elimination therewith of any trace of iodine. After cooling the flask was detached from the apparatus and into it was introduced a further quantity of potassium dichromate ( 3 5 4 5 g.) calculated to displace about 95% of the total bromine originally present. Then A was refitted to the condenser heated very slowly a t fist using B as a reflux condenser and later distilling slowly over into a flask as already described.Thus a main fraction of bromine and bromine water was obtained under conditions favourable to the reaction of any trace of chlorine with the residual bromide in solution. As a check this solution was in each case treated with an excess of potassium dichromate and distilled a third time the liberation of 1-2 g. of bromine afforded satisfactory evidence that excess of bromide had been present during the distillation of the main fraction. Tests made with the largest (middle) fraction showed that digestion with sodium oxalate was unsatisfactory as a means of conversion to sodium bromide and that the quantities available were too small to permit a satisfactory distillation of aqueous hydrobromic acid.Therefore in working up the end fractions of ammonium bromide the main fraction of bromine from the first oxidation was converted directly to hydrobromic a,cid by the use of sulphur dioxide and the acid was subjected to a second fractional oxidation with potassium dichromate as already described. Thus it was possible to obtain in each final main fraction of the second series of oxidations about 90% of the bromine present in the original ammonium bromide. This bromine was collected in a 144 ROBINSON AND BRISCOE : excess of pure dilute aqueous ammonia and the resulting ammoniacal solution of ammonium bromide was transferred to a carefully cleaned, stoppered resistance glass bottle which was kept free from dust in a clean desiccator until required for analysis.Method of Anal@. To avoid the uncertainty attendant on drying and weighing ammonium bromide the ratio chosen for measurement was that of silver to silver bromide. A dilute aqueous solution of silver nitrate, made from an accurately known weight of silver was precipitated with a slight excess of ammonium bromide and the silver bromide was washed collected dried and weighed. FIG. 1. FIG. 2. FIG. 3. FIG. 4. FIG. 6. Platinum Gooch-Munroe crucibles were not available so another method of collecting the precipitate was scught and as any trans-ference is attended by risk of loss and is a source of doubt and anxiety it seemed desirable if possible to collect and weigh the precipitate in the vessel in which it is formed but at the same time to avoid any weighing of large glass vessels.After several attempts, a satisfactory method was evolved whereby the precipitate was collected in a detachable part of a suitably shaped precipitation flask shown in Figs. 2 and 3. A conical flask A of about 1700 C.C. capacity has two necks B and C fitted with hollow glass stoppers; the neck C has also an external taper both this and the stopper D being ground info the neck of the weighing bottle E. Four such flasks a.nd two spare weighing bottles with stoppers for use as tares were made of " Durosil " glass the joints and stoppers being very carefully ground and polished to an excellent fit * the various parts of each set * Our thanks are due to Mr. G. Ellison of this Department for his care arid skill in executing this difficult task A REDETERMINATION OF THE ATOMIC WEIGHT OF BIJOMIEU'E.145 mere etched with the same number. Before use the flasks and bottles were cleaned with chromic and nitric acids washed allowed t'o stand about three weeks containing a solution of arnmoniuni bromide and nitric acid allowed to soak for a further period in distilled water and well washed. In conducting a precipitation several buttons of pure silver, weighing about 4-5 g. were weighed and dissolved in pure nitric acid in a solution flask of the form previously described (Briscce, J. 1915 107 78) and the solution was boiled to expel nitrous fumes cooled and diluted with pure water to about 400 C.C. Mean-wli ile a measured volume of ammonium bromide solution COE-taining bromine abcut 5% in excessof that required by this silver, was acidified with pure nitric acid diluted to about 700 c.c.and filtered into the precipitation flask supported by a padded ring and resting on a pad of filter paper F in the position shown in Fig. 2. All succeeding operations up to and including the final weigliings were conducted in orange light in a darkened laboratory set aside for this work. A rapid rotary movement was given to the bromide solution the silver solution was poured into i t with all the usual pre-cautions to avoid loss and 6-43 rinsings of the solution flask followed, bringing the total bulk in the precipitation flask to about 1400 C.C. Then the flask was stoppered and vigorously shaken a t first every 30 mins.and later about twice a day. After 7 days the precipitate having become sufficiently dense and coherent for transference, the stopper C was removed and rinsed into the flask and the super-nat'ant liquor was syphoned off by means of the arrangement deseribcd later. By careful manipulation the liquor left ahove the precipitate was reduced to 2-3 C.C. Next the silver bromide was washed twice by decantation using in each case ahout 1000 c.c. of wash-water allowing the precipitate to settle 2-3 days and syphoning off as in the first case. Com-monly the precipitate tended to form a colloidal solution in the seccncl wash and a solution of about 0.5 g. of ammonium nitrate, prepared from pure ammonia and nitric acid was added to avoid this difficulty. Transference of the precipitate to the weighing bottle was e€fected by means of the second washing.The bottle having been carefully heated in dry air cooled and weighed against the tare was fitted t o the aask at C and by reversing the whole apparatus into the position shown in Fig. 3 and alternately giving a rotary agitation ajnd allowing the apparatus to stand the silver bromide '~~'tls washed down into E. Then the stoppcr B was removed the wash-water syphoned off the syphon tube and the wails of the flask were rinsed down with a fine jet of water and the wash was syphoned off 146 ROBINSON AND BRISCOE : Finally the syphon was partly withdrawn i t and the flask were again rinsed the bottle was detached and the neck C rinsed into it. The bottle containing the silver bromide with but 15-20 C.C.of water was then heated in an electrically heated sheet nickel air-bath, h t at 85-90" until all free water had evaporated and then,at about 250-300" for 12-14 hours. During the whole of this drying operation a current of air supplied by a water-blast and dried by passing through a glass train over sulphuric acid concentrated aqueous potassium hydroxide and solid potassium hydroxide successively was led into the bottle by a glass tube inserted through the lid of the oven. The tare was of course heated alongside the bottle and after cooling usually for 3 4 hours in the current of dry air both bottle and tare were stoppered transferred to the balance case and weighed. The heating and weighing were repeated once or twice to ensure that a constant weight had been attained.I n two of the preliminary experiments the silver bromide was afterwards fused in the bottle heated in a small vertical electric furnace of silica provided with a window to allow observation of the fusion. As this treatment had no appreciable effect upon the apparent weight it was omitted in the final analyses. To minimise the risk of appreciable solvent action on the bottles, care was taken that they were never in contact with washings for more than 12 hours. There remains the risk of mechanical loss of glass from the ground joint; although Richards has shown that with a well made and carefullyused joint this risk is small. For-tunately satisfactory evidence that both these effects were negligible in the present work is afforded by comparing the weights of three of the bottles (weighed against the same tare) before and after the first series of precipitations :-Bottle I.Bottle 11. Bottle 111. Apparent wt. before ......... 2.02985 2-22330 3.22 103 ) , after ......... 2.02987 2.22330 3.22104 Before the method above described was finally adopted attempts were made to find a substitute for the Gooch-Munroe crucible which may be recorded briefly here. An alundum crucible was digested with 1 2 nitric acid for 6 horns and with repeated changes of water for 24 hours washed thoroughly with water under suction dried at 130" for 10 hours and ignited to about 600" in a closed porcelain crucible. After cooling in a desiccator and standing on the balance pan for 30 minutes the crucible had a weight of 12.47830 g.Then a litre of a clear filtered solution of 20 Q. of ammonium bromide in 10% nitric acid was passed through the crucible and it was washed and dried as before when its weight was 12.47799 g. (loss 0.30 mg.). A repetition of the treatment reduced the weight t A REDETEEMINATIOS O F TI-IE BTOA'IIC WEIGHT OF BROMINE. 147 12.47'782 g. (loss 0-17 mg.) and a second repetition using 5% nitric acid caused a further change t o 12.47746 g. (loss 0.36 mg.). Thus, evidently there was a continuous and variable loss too great to permit tlie successful use of alundum crucibles. A Jena filter crucible having a mat of sintered glass similarly treated but dried for 14 hours a t 350" lost in weight 0.80 mg. and among several such crucibles i t was observed that the degree of fritting was not uniform, some mats being so friable that glass could be removed with the finger nail.Hence such crucibles too were dismissed its unsuitable. The device used in syphoning the liquors and washings may here be described. Prelimiiiary experiments showed that as the liquid level fell fine particles of precipitate were apt to be dislodged from the sides of the flask and carried over with the washings. It vas impracticable to reduce the syphon to a capillary fine enougli to stop these particles therefore although tlie syphon was never brought in contact with the mass of the precipitate some form of reverse filter was necessary. Disks of thin alundum or wads of spongy platinum fused into the ends of glass tubes proved much too slow but a porous glass filter of the form shown in Fig.4 proved satisfactory. It was made by taking powdered Dixosil glass passing a sieve of 100 meshes to the linear inch and retained on zt 200 mesh sieve cleaning the powder by boiling ixith hydrochloric acid and washing with water packing i t wet to form a 3 mm. layer in the end of a 12 mm. bore Durosil tube and then drjing at 100" and heating carefully in a luminous flarre until the glass sintered to a strong coherent yet porous mass adhering firmly to the tube. After experience had been gained it was possiblc thus to make filters which n-ould pass 1000-1500 C.C. of n-atcr per hour nith a pressure difference of 3 0 4 0 cm. of mercury and yet' retain the iincst partic-lcs of suspended matter. Each filter was fused to a long tube and then carefully cleaned ant1 soaked.I n use it was connected by a length of clean pure rubber pressure-tubing to a clean bottle in nhich the liquors were received in the manner shown in Fig. 5. In each analysis the fist mother-liquor was tested to ensure that an excess of bromide had been used as nephelometric tests gave in them no indication of silver they were rejected. Some small particles of silver bromide adhered to the inner surface of the precipitation flask or were retained by the glass filter therefore one filter was kept for each analysis and after the bottle and precipi-tate had been removed the flask was stoppered at B and re-inverted, and a small quantity of pure ammonia was passed in the reverse direction through the filter shaken round the walls of the flask and, with the rinsings of filter and flask made up to a definite bulk 148 ROBINSON AND BRISCOE : The silver content of this solution and of the main washings was then determined by nephelometric comparison of aliquot portions with standard silver solutions and the weight of silver 'lost thus ascertained was deducted from the weight of silver originally taken.Statement and Discussion of Results. Several preliminary analyses made on the middle fraction of ammonium bromide No. 57 whilst useful in establishing the methods are not comparable in value with those of the final series : hence the results are omitted here. The final series consisted of eight determinations four on the ammonium bromide from the head fraction No.71 and four on t,hat from the tail fraction No. 43. The essenthl data are given in Table I. TABLE I. Scries I . 1 71 A 3.69325 2 , , 5.06115 3 , , 3-93515 4 ? , 3.67332 Serie.9 I I . 5 43 A 5-14730 6 ,? 3.08140 7 , B 3.34631 S , , 3.43817 0.00058 3-69267 6.42874 0-00297 5.05818 8.80494 0.00374 3.93141 6.84337 0.00097 3.67235 6.39249 Mean of Series I. 0.00146 5-14584 8.95708 0.00111 3.08029 5-36261 0-00072 3.34559 5.82369 0.00251 3-43566 5.98102 Mean of Series 11. General mean 0,574400 79-933 0.574471 79.910 0.574484 79.906 0.574479 79.908 0.574459 79.914 0.574499 79.901 0.574401 79-933 0.574479 79.907 0.574427 79.916 0.574451 79.914 0.574455 79.914 +0.019 - 0.004 -0.00s -0.006 & 0.009 -0.013 + 0.01 7 - 0.009 + 0-002 fO.011 &0.010 In the first group of analyses the extreme variation in the ratio Ag AgBr is 0.000084 or 1 part in 6840 that in the atomic weight of bromine is 0.027 or 1 part in 2990.In the second group t.he cor-responding variations are 0-000096 or 1 part in 5984 and 0.032 or 1 part in 2493. The magnitude and sign of these variations indicate that the mean value of the atomic weight deduced from each series has an error probably less than 0.01 or in round numbers, about 1 part in 8000. The mean atomic weights from the two series differ by but 1 part in 40,000 as a change of lye in the isotope ratio would change the atomic weight by 0.02 and should have been appreciable in these determinations it appears that a fractionation of ammonium bromide involving 2700 crystallisations does not produce such a change.This result affording as it does more precise confirmation of th A REDETERMINATION OF THE ATOMIC WEIGHT OF BROMINE. 149 conclusion drawn by Richards and Hall is of some theoretical interest. It is in particular instructive to institute a comparison between t'he present results and those recorded for salts of kindred rare-earths where the solubility differences are of the same order. The solubilities of the hexahydrated nitrates of lanthanum and neodymium La(NB,),,6H20 and Nd(N0,),,6H20 are respect-ively 151.1 and 152.9 parts of anhydrous nitrate in 100 parts of water (James and Whittemore J . Amer. Chem. Xoc. 1912 34, 1168; James and Robinson ibid. 1913,35 754).As the molecular weights are 324.8 (La = 138-8) and 330-2 (Nd = 144.2) the mole-cular solubilities are 0,46520 and 0.46515. The coincidence of these numbers within one part in 9000 is obviously accidental but the molecular solubilities evidently differ by no more than the error of the solubility determinations say 2 or 3 parts in 1000 parts. The effect of fractional crystallisation in this case is shown in t'he investigations inter alios of Demargay (Cmpt. rend. 1896 122, 728; 1900 130 1021) and of Baxter and Chapin ( J . Amer. Chem. Soc. 1911 33 1). In the latter case for example about 1600 crystallisations as double ammonium nitrates and aboutl 1300 as simple nitrates produced a number of fractions of neodymium free from appreciable traces of other earths.The fact that 2000 crystallisations or fewer completely separate these earths whilst more than 2500 crystallisations do not appreci-ably change the proportions of the two kinds of molecule in ammon-ium bromide may be explained by and therefore affords chemical evidence supporting the theories of Bohr and of Bury (J. Amer. Chem. Xoc. 1921 43 1602) whereby the additional electrons in the atom of the heavier of a pair of rare-earth metals are supposed to be less intimately associated with the corresponding protons than are the additional electrons in the heavier of a pair of isotopes. Regarding the data as one series of determinations of the atomic weight of bromine it is to be noted that the extreme variation in the ratio Ag AgBr is from 0-574400 to 0.574498 or 1 part in 5745, whilst that in the atomic weight is from 79.901 to 79-933 or 1 part in 2500.The general mean values are given in Table I. A discussion of all the available data for the ratio of silver to silver bromide both direct and indirect has been given by Clarke (" A Recalculation of the Atomic Weights," 4th Edition Mem. Nnt. Acud. Sci. 1920 16 71) and need not be attempted here. The best direct determinations are those of Baxter (J. Amer. Chem. Soc., 1906 28 1322) whose general mean of 18 determinations of the ratio Ag AgBr is 0-57445 a figure substantially identical with that calculated by Clarke as the general mean of all determinations. Possible sources of error in the present determinations are 150 BRISCOE ROBINSON AND STEPHENSON : (1) errors in weighing ; (2) loss of silver in transference ; (3) impurity in the silver; (4) loss of silver bromide mechanically or in solution; (5) the presence of chlorine or iodine in the ammonium bromide.The error in weighing was undoubtedly too small to have any signifi-cant effect and the mode of transference of silver solution to the precipitation flask would appear to eliminate any risk of mechanical loss. The methods used for the preparation of pure silver have been shown to yield a metal containing less than 1 part of impurity in 300,000 parts of silver (Briscoe and Little J. 1914 105 1320). There can be little doubt that the chief error in the determination of weight lay in the nephelometric estimation of the silver lost (4) in washings and retained by the flask and filter a careful review of the analytical details leads to the conclusion that this error was not greater than 0.0001 g.but it was probably the one really significant error in the determinatioiis. With regard to the purity of the bromine it is certain that in such a prolonged fractionation the whole of any chlorine and iodine originally present would have become concentrated in opposite end fractions therefore the identity of the results obtained with these fractions shows that the chemical purification subsequent to fractionation had so far reduced the proportion of these halogens that their effect on the atomic weight of bromine was inappreciable. The final mean value of the ratio Ag AgBr now found differs from Baxter’s value by less than 1 part in 100,000 and the rounded mean value of the atomic weight Br = 79.92 with a probable error & 0.0031 confirms the accepted atomic weight.UNIVERSITY OF DURHAM ARMSTRONG COLLEGE, NEWCASTLE UPON TYNE. [Received November 3rd 1924. 138 ROBINSON AND BRISCOE: XXPI1.-A Redetermination of the Atomic Weight of The Inseparability of the Isotopes by Bromine. Fractional Crystallisation. By PERCY LUCOCK ROBINSON and HENRY VINCEPU'T AIRD ERISCOE. THE only precise experimental evidence in support of the generally accepted view that isotopes are inseparable by fractional crystallis-ation is the attempt by Richards and Hall ( J . Amer. Chem. Xoc., 1917 39 531) to separate thus the isotopic forms present in lead derived from Australian carnotite. Both they and Soddy (J., 1911 99 72; J .Amer. Chem. SOC. 1917 39 1614) refer to the earlier data which need not be quoted here. Richards and Hall obtained no evidence of separation by 904 effective crystallisations but their lead presented a relatively unfavourable case since it contained only about 20y0 of the heavier isotope and the nitrates of the isotopes differed in solubility by but o*7470 (Richards and Schumb J . Anzer. Chem. SOC. 1918, 40,1403). Since Aston's work has disclosed the existence of isotopes of many of the lighter elements it has become clear that cases may be found permitting a more precise test (or proof) of inseparability by fractional crystallisation. The isotopes of boron differ by 10% in atomic weight but the isotope ratio is unfavourable (approxim-ately 20% BlO and 80% Bl1) and an accurate determination of the atomic weight is both difficult and laborious.A prolonged frac-tionation of boric acid was effected but the determinations of atomic weight although they had other unexpected features o€ interest (see following paper) proved valueless for the purpose now under discussion. Bromine in the form of ammonium bromide seems to present the best test-case because (1) bromine has but two isotopes (Br79 and BrSl) present in nearly equal proportions and differing by 2.2y0 in atomic weight (2) hydrogen and nitrogen are " simple " elements and together form less than 20% of the salt (3) the atomic weight of bromine can be determined by direct and trustworthy methods and (4) ammonium bromide crystallises from water in a fashion convenient for a long series of fractionations.Therefore with the double object of making a re-determination of the atomic weight of bromine and of confirming in this more favour-able case the finding of Richards and Hall ammonium bromide was subjected to prolonged fractional crystallisation and the atomic weight of the bromine in the final head and tail fractions was deter-mined by measurement of the ratio Ag AgRr A EEDETERMIXATIOX O F THE ATORlIC WEIGHT O F SROXISE. 130 Pu?<Jicatio.rz of Reagmts. T17cter.-Laboratory distilled water already free from halogen, was redistilled from a little caustic soda and potassium perman-ganate in a 10-gallon copper still through a vertical spray-trap packed with glass beads and condensed in pure tin.The first and last portions of the distillate were rejected anif the main fraction was collected and stored (for but a short time) in 'O-litre stoppered resistance glass bottles which had been n-ell cleaned steamed out and kept for months full of distilled water. Many iwphclometric tests were made but on no occasion did the water give any indication of chloride. Ammo?zin.-Ammonia gas liberated by warming the purest commercial ammonia (d = O.SSU) was dissolved in pure water, in an apparatus constructed entirely of rcsistance glass with cz ground-in leading tube. SuZpJzur Dioxide.-The gas used both for the conversicii of bromine to hydrobromic acid and for the precipitation cf silver by Stas's method was aln-ays obtained from the middle fraction of a syphon of the liquid.n'ilric Acid.-Commercial nitric acid containing o d y small traces of halogen was thrice redistilled in an apparatus constrwted of '' Duro " resistance glass with ground joints large head and tail fractions being rejected. The final main fraction tested in the ncyhelometer was free from halogens. It was kept in a resistance glass bottle having c well-ground glass stopper protected by a glass cap. Formic Acid.-The purest obtainable reagent acid was twice redistilled from a fused silica ffask having a ground-in silica con-denser and adapter. Considerable head and tad fractions were rejected in each distillation and the main fraction n-as preserved i s a stoppered silica flask. il~ag?zesin.-~lagiiesiulrz nitrate was thrice precipitated from saturated aqueous solution by excess of nitric acid filtered on porcelain dissolvcd in water and precipitated with ammonia.The hydroxide was washed by decantation filtered dried a t lOO", ignited in air in a platinum dish and finally heated at about 1000" in a current of moist pure hydrogen. Magnesia boats moulded from the pure oxide nioisteried with dilute nitric acid were ignited and used on a silica plate. Hydrochloric Acid-Coinmercial reagent acid was freed from arsenic by treatment with a copper-tin couple and distillation from copper gauze according to the method of Thorne and Jefiers Zi?ic.-A sample of granulated electrolytic zinc kindly supplied ( d ? ~ d y s t 1906 31 101). F" 140 ROBINSON AND BRISCOE : by Messrs. Brunner Mond & Co. Ltd.proved t'o be free from arsenic and was used without further purification. Hydrogen.-This gas prepared from pure hydrochloric acid, containing a little platinic chloride and pure zinc in a Kipp's apparatus passed to a purifying train the several parts of which were sealed together in which it traversed successively a 12 inch column of concentrated potassium hydroxide solution two 12 inch columns packed with broken sticks of pot'assium hydroxide and two 10 inch columns packed with phosphorus pentoxide distributed on glass wool. XiZwer.-Two samples " A " and " B," were used prepared in the same way but at different times and from different samples of commercial silver nitrate. In each case a filtered aqueous solution of silver nitrate (200 g.) was precipitated in a volume of about 20 Iitres with ammonium bisulphite according to t'he method of Stas (Briscoe J.1915 107 69). The precipitated silver was washed six times with dilute ammonia allowed to stand over-night with ammonia washed six times with water and dried at 100". This silver in 100 g. portions was dissolved in pure nitric acid, precipitated hot with ammonium formate in a bulk of 2 litres, well washed with ammonia and water and dried first a t 100" and finally at 250" in an electrically heated covered beaker. Before weighing the silver was melted on a boat of pure magnesia in an atmosphere of hydrogen in an electrically heated silica tube furnace. The hydrogen issuing from the furnace during melting contained no impurities detectable by Marsh's test or by smell.At the end of each fusion the silver was collected into large buttons by shaking the furnace and these when cool were etched with 1 1 nitric acid well washed with water heated at 250" for 12 hours, and cooled in a desiccator over solid potassium hydroxide. The Determinatioia of Weight. All precise weighings were made on an Oertling balance con-structed specially for this work resembling in many respects the standard type known as " No. 7 S.W.," but with a beam of Firth's 36% nickel-steel having a low coefficient of expansion pan-supports of special design and a separate enclosure for the beam after the principle used by Manley (Phil. Trans. 1910 210 A 387). No proper situation giving constant temperature and freedom from vibration was available for the balance it was used on an ordinary stout wooden table in a room subject to vibration and to consider-able and rapid fluctuations of temperature.Even under these adverse conditions it gave results sufficiently precise and consistent for this work a remarkable performance which is undoubtedl A REDETERMINATION OF THE ATOMIC WEIGKT OF BRONINE. 141 attributable to the special features of construction indicated above. The sensitiveness of the balance increased very slightly with the load and during all the mork here described was very nearly 100 scale divisions per milligram. In weighing by the method of oscil-lations the zero could easily be determined to the nearest half-division and the apparent weight thus ascertained with an error not exceeding 0.00001 g.A set of gold-plated brass weights with platinum fractions by Oertling and a gold 5-mg. rider were used and were calibrated to ascertain the relative weights in air on three separate occasions before and during the weighings. The corrections applied to ascertain relative weights in a vacuum were calculated using the density of air at the temperature and pressure observed during the weighing and the densities 10.49 and 6.47 for silver and silver bromide respectively. As relative weights only were desired and the inequality of the arms of the balance was very small and constant all weighings were made directly. All were made in duplicate and many in triplicate in no case did the values for the vacuum weight thus obtained differ by more than 0.00003 g., thus it seems probable that the mean values arc in error by less than this amount.Silver was weighed directly on the balance pan silver bromide was weighed in a stoppered glass bottle against a tare of thc same glass and of closely similar form volume and weight which had been treated in all respects as the bottle. The balance case contained solid caustic potash and a piece of pitchblende. The Fractional Crystallisalioiz of mtio'i/iuitL Brmtiide. As starting material commercial ammonium bromide puriss., was used. Prolonged exposure to the laboratory atmosphere and contact with glass inevitably introduce impurities compared with which those originally present are insignificant ; hence no attempt was made to ascertain the nature and amount of the latter.About 2500 g. of ammonium bromide were dissolved in hot water in such proportion that about one-half crystallised out u n cooling. Each fraction was again fractionated in a similar manner until a series of 30 fractions had been built up. Thereafter the number of fractions was kept constant and fractionation was continued in the usual way by crystallising the whole series taking away the " head '' (most soluble) mother-liquor to form part of a new fraction, transferring each of the other mother-liquors to the crystals of the next higher fraction adding water to the crystals of the " tail " (least soluble) fraction and again crystallising the whole series 142 ROBINSON AND BRISCOE: A detailed scheme of such a fractionation is given by Richards and Hall (loc. cit.).I n all 80 crystallisations of the whole series were thus made * and the total number of crystallisations including those in-volved in the establishment of the series was approximately 2700. Throughout the later part of the fractionation each fraction contained about 80 g . of ammonium bromide of which one-half was transferred a t each crystallisation. The fractions were contained in 200 C.C. conical flasks closed against dust by loose hollow glass stoppers distilled water of good quality was used for all crystal-lisations. At the end of the fractionation the extreme end fract’ions, Nos. 42 and 72 were rejected and the ammonium bromide of Nos. 43,57 and 71 the ‘‘ tail,” middle and “ head ” fractions respectively, was taken for the atomic weight determinations.PuriJication of Bromine for Analysis. The volatility of bromine and its liberation from a bromide by oxidation afford an unexceptionable means of separating it sharply from all elements other than chlorine and iodine. It is therefore with these elements which would tend t o concentrate in the tail and head fractions respectively that the scheme of purification is concerned. At the same time as it was important that the determinations on head and tail fractions should be strictly com-parable both had to receive identical treatment. Bromine is usually purified from the other halogens by applying the facts that it liberates iodine from an iodide and is itself liberated from a bromide by chlorine. Whilst in ordinary analytical work the assumption that these reactions are complete and irreversible holds well enough they are probably not so in fact (see e.g.Schuyten, Chem. Ztg. 1908 32 619). The methods usually adopted for the rigorous purification of bromine for atomic weight work (see e.g., Scott J. 1900 97 614; Baxter J . Amer. Chem. Xoc. 1906 28, 1322) reject so great a proportion of the material in head and tail fractions (using “head ” and tail^' here in a chemical sense) as to ensure elimination of the other halogens but were for that reason inapplicable t o the small quantities (50-60 g. of bromine in each fraction) available for purification in this case. Therefore each fraction of ammonium bromide was evaporated t o dryness dried at 160° and weighed. A quantity of pure sodium carbonate about 5% in excess of the calculated amount dissolved in a small quantity of water was added to the bromide and the * The establishment of the series and the first 30 crystallisations thereof were carried out by Mr.T. Ranby whose valuable assistance in this tedious work w e desire t o acknowledge A REDETERMIXATION OF THE ATOMIO WEIGHT OF BROMINE. 143 whole was again evaporated to dryness and fused a t about 800" in platinum in an oxidising atmosphere. Thus any organic impurities were destroyed. A solution of about 1 g. of potassium dichromate and 30 C.C. of concentrated sulphuric acid in 170 C.C. cf water was boiled vigorously for 10 minutes to expel any possible trace of halogen cooled and used to dissolve the sodium bromide from the platinum dish. This solution was transferred to the flask A of the distillation apparatus shown in Fig.1 heated and kept boiling for about 10 minutes whilst the column B jacketed with cold water acted as a reflux condenser so that the free bromine might react with any trace of iodide present. Then the cooling was discontinued and the solution boiled until the bromine had distilled over into a flask containing pure water cooled externally by ice. Finally this flask was removed so that steam passed uncondensed right through the apparatus and carried away any residual traces of free halogen. The quantityof potassium dichromate used was calculated to displace about 2% of the total bromine and the procedure described evidently favoured the elimination therewith of any trace of iodine. After cooling the flask was detached from the apparatus and into it was introduced a further quantity of potassium dichromate ( 3 5 4 5 g.) calculated to displace about 95% of the total bromine originally present.Then A was refitted to the condenser heated very slowly a t fist using B as a reflux condenser and later distilling slowly over into a flask as already described. Thus a main fraction of bromine and bromine water was obtained under conditions favourable to the reaction of any trace of chlorine with the residual bromide in solution. As a check this solution was in each case treated with an excess of potassium dichromate and distilled a third time the liberation of 1-2 g. of bromine afforded satisfactory evidence that excess of bromide had been present during the distillation of the main fraction.Tests made with the largest (middle) fraction showed that digestion with sodium oxalate was unsatisfactory as a means of conversion to sodium bromide and that the quantities available were too small to permit a satisfactory distillation of aqueous hydrobromic acid. Therefore in working up the end fractions of ammonium bromide the main fraction of bromine from the first oxidation was converted directly to hydrobromic a,cid by the use of sulphur dioxide and the acid was subjected to a second fractional oxidation with potassium dichromate as already described. Thus it was possible to obtain in each final main fraction of the second series of oxidations about 90% of the bromine present in the original ammonium bromide. This bromine was collected in a 144 ROBINSON AND BRISCOE : excess of pure dilute aqueous ammonia and the resulting ammoniacal solution of ammonium bromide was transferred to a carefully cleaned, stoppered resistance glass bottle which was kept free from dust in a clean desiccator until required for analysis.Method of Anal@. To avoid the uncertainty attendant on drying and weighing ammonium bromide the ratio chosen for measurement was that of silver to silver bromide. A dilute aqueous solution of silver nitrate, made from an accurately known weight of silver was precipitated with a slight excess of ammonium bromide and the silver bromide was washed collected dried and weighed. FIG. 1. FIG. 2. FIG. 3. FIG. 4. FIG. 6. Platinum Gooch-Munroe crucibles were not available so another method of collecting the precipitate was scught and as any trans-ference is attended by risk of loss and is a source of doubt and anxiety it seemed desirable if possible to collect and weigh the precipitate in the vessel in which it is formed but at the same time to avoid any weighing of large glass vessels.After several attempts, a satisfactory method was evolved whereby the precipitate was collected in a detachable part of a suitably shaped precipitation flask shown in Figs. 2 and 3. A conical flask A of about 1700 C.C. capacity has two necks B and C fitted with hollow glass stoppers; the neck C has also an external taper both this and the stopper D being ground info the neck of the weighing bottle E. Four such flasks a.nd two spare weighing bottles with stoppers for use as tares were made of " Durosil " glass the joints and stoppers being very carefully ground and polished to an excellent fit * the various parts of each set * Our thanks are due to Mr.G. Ellison of this Department for his care arid skill in executing this difficult task A REDETERMINATION OF THE ATOMIC WEIGHT OF BIJOMIEU'E. 145 mere etched with the same number. Before use the flasks and bottles were cleaned with chromic and nitric acids washed allowed t'o stand about three weeks containing a solution of arnmoniuni bromide and nitric acid allowed to soak for a further period in distilled water and well washed. In conducting a precipitation several buttons of pure silver, weighing about 4-5 g. were weighed and dissolved in pure nitric acid in a solution flask of the form previously described (Briscce, J.1915 107 78) and the solution was boiled to expel nitrous fumes cooled and diluted with pure water to about 400 C.C. Mean-wli ile a measured volume of ammonium bromide solution COE-taining bromine abcut 5% in excessof that required by this silver, was acidified with pure nitric acid diluted to about 700 c.c. and filtered into the precipitation flask supported by a padded ring and resting on a pad of filter paper F in the position shown in Fig. 2. All succeeding operations up to and including the final weigliings were conducted in orange light in a darkened laboratory set aside for this work. A rapid rotary movement was given to the bromide solution the silver solution was poured into i t with all the usual pre-cautions to avoid loss and 6-43 rinsings of the solution flask followed, bringing the total bulk in the precipitation flask to about 1400 C.C.Then the flask was stoppered and vigorously shaken a t first every 30 mins. and later about twice a day. After 7 days the precipitate having become sufficiently dense and coherent for transference, the stopper C was removed and rinsed into the flask and the super-nat'ant liquor was syphoned off by means of the arrangement deseribcd later. By careful manipulation the liquor left ahove the precipitate was reduced to 2-3 C.C. Next the silver bromide was washed twice by decantation using in each case ahout 1000 c.c. of wash-water allowing the precipitate to settle 2-3 days and syphoning off as in the first case.Com-monly the precipitate tended to form a colloidal solution in the seccncl wash and a solution of about 0.5 g. of ammonium nitrate, prepared from pure ammonia and nitric acid was added to avoid this difficulty. Transference of the precipitate to the weighing bottle was e€fected by means of the second washing. The bottle having been carefully heated in dry air cooled and weighed against the tare was fitted t o the aask at C and by reversing the whole apparatus into the position shown in Fig. 3 and alternately giving a rotary agitation ajnd allowing the apparatus to stand the silver bromide '~~'tls washed down into E. Then the stoppcr B was removed the wash-water syphoned off the syphon tube and the wails of the flask were rinsed down with a fine jet of water and the wash was syphoned off 146 ROBINSON AND BRISCOE : Finally the syphon was partly withdrawn i t and the flask were again rinsed the bottle was detached and the neck C rinsed into it.The bottle containing the silver bromide with but 15-20 C.C. of water was then heated in an electrically heated sheet nickel air-bath, h t at 85-90" until all free water had evaporated and then,at about 250-300" for 12-14 hours. During the whole of this drying operation a current of air supplied by a water-blast and dried by passing through a glass train over sulphuric acid concentrated aqueous potassium hydroxide and solid potassium hydroxide successively was led into the bottle by a glass tube inserted through the lid of the oven. The tare was of course heated alongside the bottle and after cooling usually for 3 4 hours in the current of dry air both bottle and tare were stoppered transferred to the balance case and weighed.The heating and weighing were repeated once or twice to ensure that a constant weight had been attained. I n two of the preliminary experiments the silver bromide was afterwards fused in the bottle heated in a small vertical electric furnace of silica provided with a window to allow observation of the fusion. As this treatment had no appreciable effect upon the apparent weight it was omitted in the final analyses. To minimise the risk of appreciable solvent action on the bottles, care was taken that they were never in contact with washings for more than 12 hours. There remains the risk of mechanical loss of glass from the ground joint; although Richards has shown that with a well made and carefullyused joint this risk is small.For-tunately satisfactory evidence that both these effects were negligible in the present work is afforded by comparing the weights of three of the bottles (weighed against the same tare) before and after the first series of precipitations :-Bottle I. Bottle 11. Bottle 111. Apparent wt. before ......... 2.02985 2-22330 3.22 103 ) , after ......... 2.02987 2.22330 3.22104 Before the method above described was finally adopted attempts were made to find a substitute for the Gooch-Munroe crucible which may be recorded briefly here. An alundum crucible was digested with 1 2 nitric acid for 6 horns and with repeated changes of water for 24 hours washed thoroughly with water under suction dried at 130" for 10 hours and ignited to about 600" in a closed porcelain crucible.After cooling in a desiccator and standing on the balance pan for 30 minutes the crucible had a weight of 12.47830 g. Then a litre of a clear filtered solution of 20 Q. of ammonium bromide in 10% nitric acid was passed through the crucible and it was washed and dried as before when its weight was 12.47799 g. (loss 0.30 mg.). A repetition of the treatment reduced the weight t A REDETEEMINATIOS O F TI-IE BTOA'IIC WEIGHT OF BROMINE. 147 12.47'782 g. (loss 0-17 mg.) and a second repetition using 5% nitric acid caused a further change t o 12.47746 g. (loss 0.36 mg.). Thus, evidently there was a continuous and variable loss too great to permit tlie successful use of alundum crucibles.A Jena filter crucible having a mat of sintered glass similarly treated but dried for 14 hours a t 350" lost in weight 0.80 mg. and among several such crucibles i t was observed that the degree of fritting was not uniform, some mats being so friable that glass could be removed with the finger nail. Hence such crucibles too were dismissed its unsuitable. The device used in syphoning the liquors and washings may here be described. Prelimiiiary experiments showed that as the liquid level fell fine particles of precipitate were apt to be dislodged from the sides of the flask and carried over with the washings. It vas impracticable to reduce the syphon to a capillary fine enougli to stop these particles therefore although tlie syphon was never brought in contact with the mass of the precipitate some form of reverse filter was necessary.Disks of thin alundum or wads of spongy platinum fused into the ends of glass tubes proved much too slow but a porous glass filter of the form shown in Fig. 4 proved satisfactory. It was made by taking powdered Dixosil glass passing a sieve of 100 meshes to the linear inch and retained on zt 200 mesh sieve cleaning the powder by boiling ixith hydrochloric acid and washing with water packing i t wet to form a 3 mm. layer in the end of a 12 mm. bore Durosil tube and then drjing at 100" and heating carefully in a luminous flarre until the glass sintered to a strong coherent yet porous mass adhering firmly to the tube.After experience had been gained it was possiblc thus to make filters which n-ould pass 1000-1500 C.C. of n-atcr per hour nith a pressure difference of 3 0 4 0 cm. of mercury and yet' retain the iincst partic-lcs of suspended matter. Each filter was fused to a long tube and then carefully cleaned ant1 soaked. I n use it was connected by a length of clean pure rubber pressure-tubing to a clean bottle in nhich the liquors were received in the manner shown in Fig. 5. In each analysis the fist mother-liquor was tested to ensure that an excess of bromide had been used as nephelometric tests gave in them no indication of silver they were rejected. Some small particles of silver bromide adhered to the inner surface of the precipitation flask or were retained by the glass filter therefore one filter was kept for each analysis and after the bottle and precipi-tate had been removed the flask was stoppered at B and re-inverted, and a small quantity of pure ammonia was passed in the reverse direction through the filter shaken round the walls of the flask and, with the rinsings of filter and flask made up to a definite bulk 148 ROBINSON AND BRISCOE : The silver content of this solution and of the main washings was then determined by nephelometric comparison of aliquot portions with standard silver solutions and the weight of silver 'lost thus ascertained was deducted from the weight of silver originally taken.Statement and Discussion of Results. Several preliminary analyses made on the middle fraction of ammonium bromide No.57 whilst useful in establishing the methods are not comparable in value with those of the final series : hence the results are omitted here. The final series consisted of eight determinations four on the ammonium bromide from the head fraction No. 71 and four on t,hat from the tail fraction No. 43. The essenthl data are given in Table I. TABLE I. Scries I . 1 71 A 3.69325 2 , , 5.06115 3 , , 3-93515 4 ? , 3.67332 Serie.9 I I . 5 43 A 5-14730 6 ,? 3.08140 7 , B 3.34631 S , , 3.43817 0.00058 3-69267 6.42874 0-00297 5.05818 8.80494 0.00374 3.93141 6.84337 0.00097 3.67235 6.39249 Mean of Series I. 0.00146 5-14584 8.95708 0.00111 3.08029 5-36261 0-00072 3.34559 5.82369 0.00251 3-43566 5.98102 Mean of Series 11. General mean 0,574400 79-933 0.574471 79.910 0.574484 79.906 0.574479 79.908 0.574459 79.914 0.574499 79.901 0.574401 79-933 0.574479 79.907 0.574427 79.916 0.574451 79.914 0.574455 79.914 +0.019 - 0.004 -0.00s -0.006 & 0.009 -0.013 + 0.01 7 - 0.009 + 0-002 fO.011 &0.010 In the first group of analyses the extreme variation in the ratio Ag AgBr is 0.000084 or 1 part in 6840 that in the atomic weight of bromine is 0.027 or 1 part in 2990.In the second group t.he cor-responding variations are 0-000096 or 1 part in 5984 and 0.032 or 1 part in 2493. The magnitude and sign of these variations indicate that the mean value of the atomic weight deduced from each series has an error probably less than 0.01 or in round numbers, about 1 part in 8000.The mean atomic weights from the two series differ by but 1 part in 40,000 as a change of lye in the isotope ratio would change the atomic weight by 0.02 and should have been appreciable in these determinations it appears that a fractionation of ammonium bromide involving 2700 crystallisations does not produce such a change. This result affording as it does more precise confirmation of th A REDETERMINATION OF THE ATOMIC WEIGHT OF BROMINE. 149 conclusion drawn by Richards and Hall is of some theoretical interest. It is in particular instructive to institute a comparison between t'he present results and those recorded for salts of kindred rare-earths where the solubility differences are of the same order. The solubilities of the hexahydrated nitrates of lanthanum and neodymium La(NB,),,6H20 and Nd(N0,),,6H20 are respect-ively 151.1 and 152.9 parts of anhydrous nitrate in 100 parts of water (James and Whittemore J .Amer. Chem. Xoc. 1912 34, 1168; James and Robinson ibid. 1913,35 754). As the molecular weights are 324.8 (La = 138-8) and 330-2 (Nd = 144.2) the mole-cular solubilities are 0,46520 and 0.46515. The coincidence of these numbers within one part in 9000 is obviously accidental but the molecular solubilities evidently differ by no more than the error of the solubility determinations say 2 or 3 parts in 1000 parts. The effect of fractional crystallisation in this case is shown in t'he investigations inter alios of Demargay (Cmpt. rend. 1896 122, 728; 1900 130 1021) and of Baxter and Chapin ( J .Amer. Chem. Soc. 1911 33 1). In the latter case for example about 1600 crystallisations as double ammonium nitrates and aboutl 1300 as simple nitrates produced a number of fractions of neodymium free from appreciable traces of other earths. The fact that 2000 crystallisations or fewer completely separate these earths whilst more than 2500 crystallisations do not appreci-ably change the proportions of the two kinds of molecule in ammon-ium bromide may be explained by and therefore affords chemical evidence supporting the theories of Bohr and of Bury (J. Amer. Chem. Xoc. 1921 43 1602) whereby the additional electrons in the atom of the heavier of a pair of rare-earth metals are supposed to be less intimately associated with the corresponding protons than are the additional electrons in the heavier of a pair of isotopes.Regarding the data as one series of determinations of the atomic weight of bromine it is to be noted that the extreme variation in the ratio Ag AgBr is from 0-574400 to 0.574498 or 1 part in 5745, whilst that in the atomic weight is from 79.901 to 79-933 or 1 part in 2500. The general mean values are given in Table I. A discussion of all the available data for the ratio of silver to silver bromide both direct and indirect has been given by Clarke (" A Recalculation of the Atomic Weights," 4th Edition Mem. Nnt. Acud. Sci. 1920 16 71) and need not be attempted here. The best direct determinations are those of Baxter (J. Amer. Chem. Soc., 1906 28 1322) whose general mean of 18 determinations of the ratio Ag AgBr is 0-57445 a figure substantially identical with that calculated by Clarke as the general mean of all determinations.Possible sources of error in the present determinations are 150 BRISCOE ROBINSON AND STEPHENSON : (1) errors in weighing ; (2) loss of silver in transference ; (3) impurity in the silver; (4) loss of silver bromide mechanically or in solution; (5) the presence of chlorine or iodine in the ammonium bromide. The error in weighing was undoubtedly too small to have any signifi-cant effect and the mode of transference of silver solution to the precipitation flask would appear to eliminate any risk of mechanical loss. The methods used for the preparation of pure silver have been shown to yield a metal containing less than 1 part of impurity in 300,000 parts of silver (Briscoe and Little J. 1914 105 1320). There can be little doubt that the chief error in the determination of weight lay in the nephelometric estimation of the silver lost (4) in washings and retained by the flask and filter a careful review of the analytical details leads to the conclusion that this error was not greater than 0.0001 g. but it was probably the one really significant error in the determinatioiis. With regard to the purity of the bromine it is certain that in such a prolonged fractionation the whole of any chlorine and iodine originally present would have become concentrated in opposite end fractions therefore the identity of the results obtained with these fractions shows that the chemical purification subsequent to fractionation had so far reduced the proportion of these halogens that their effect on the atomic weight of bromine was inappreciable. The final mean value of the ratio Ag AgBr now found differs from Baxter’s value by less than 1 part in 100,000 and the rounded mean value of the atomic weight Br = 79.92 with a probable error & 0.0031 confirms the accepted atomic weight. UNIVERSITY OF DURHAM ARMSTRONG COLLEGE, NEWCASTLE UPON TYNE. [Received November 3rd 1924.