150 BRISCOE ROBINSON AND STEPHENSON : XX1V.-The Use of Fused Borax i n the Determination of the Atomic Weight of Boron. By HENRY VINCENT AIRD BRISCOE PERCY LUCOCK ROBINSON, and GEORGE EDWARD STEPITENSON. THOUUH a few determinations of the atomic weight of boron have been made by ascertaining the ratio of boron halides to silver, the majority have depended in one way or another upon weighing anhydrous borax in the form of borax-glass. A detailed discussion of the results may be found in Clarke’s “ A Recalculation of the Atomic Weights ” (Hem. Nat. Acad. Sci. 1920 16 205) and it is sufficient here to give a list of the determinations. Table I records all the previous data classified according to the nature of the method employed the wide divergence among the results render THE USE OF FUSED BORAX W THE DETERMINATION ETC.161 the “probable errors” obviously valueless as a measure of the weight to be given to each determination and they are therefore, omitted. TABLE I. Previous Determiirzations of the Atomic Weight of Boron. A. Determinations dependent upon weighing fused borax : Estimation of water in borax. 11.023 10-859 1869. DobrovoEsky. Estinmtion of water in bomx. 1st series. 2nd serk. 11.529 1892. Hoskyns- Abrahall. Decahydrate to borax glass. 10.703 1893. Ramsay and Aston. Decahydrate to borax glass. 10.946 10.955 11.059 1898. Armitage. Decahydrate to borax glass. 10,986 uric acid. 10.933 1918. Smith and van Haagen. Borax glass to sodium sulphate. 10.904 10.903 10.905 10.899 10-901 :z :::27} Borax glass to sodium chloride.Borax glass to silver chloride. Borax glass titrated with sulph-Borax glass to sodium fluoride. Borax glass to sodium carbonate. Borax glass to sodium nitrate. Borax glass to sodium chloride. Mean of all determinations. 10.964 Mean excluding the first four results.. 10.939 B. Determinations of the ratio of a boron halide to silver or silver halide : 1899. Gautier. BBr 3AgBr. 11.016 BC‘I 3AgCl. 10.947 1922. Hiinigschmid and BCl 3AgCI. 10.840 Birkenbach. 10.818 10,825 1892. Hoskyns-Abrahall. BBr 3Ag. 10-880 Mean excluding Gautier’s results. 10.827 ~ C. Determinations of ratios which do not involve fused borax or a boron 1893. Rimbach. Na2B40,10H20 hydrochloric acid solution. 11.006 1899. Gautier. B2S 3BaS0,.11.024 B,C 6C0,. 10.997 1923. Stock and Kuss. B,Hs 6H,. 10.806 halide : References Berzeliw Pogg. Ann. 1826 8 1 ; Laurent J . pr. Chem., 1849 47 415; Dobrovolsky Doctoral Disaertatwn Kiev 1869; Hoskyns-Abrahall J. 1892 61 650; Ramsay and Aston J. 1893 68 211; Armitage, P. 1898 14 22; Smith and van Haagen Carnegie Inst. Washington Pub-lication No. 267 1918; Gautier Ann. Chim. Phy8. 1899 [vii] 18 352; Hdnigschmid and Birkenbach Ber. 1923 56 [B] 1467; Rimbach Bcr., 1893 26 164; Stock and Kuss Ber. 1923 56 [B] 314; 8. anorg. Chem., 1923,128 49. Certain of these determinations (printed in italics) are unsatis-factory. In Group A those of Bemelius and Laurent have no pretensions to great accuracy Dobrovolsky’s two series are s 152 BRISCOE ROBINSON AND STEPHENSON : gravely discordant as to be valueless; and Abrahall’s results are definitely those of preliminary analyses thus these may be excluded making the mean result for the Group B = 10.939.I n Group B Gautier’s results show discordance so grave and so unaccountable in a halide ratio as to necessitate their immediate exclusion the remaining results give the mean value B = 10.827. I n Group C Gautier’s determinations involve eccentric ratios apparently unsuitable for atomic weight determination. The two remaining values may be retained but are not pertinent to the point now to be discussed. Reviewing all the results in Groups A and B not excluded for the reasons given it is immediately obvious that those dependent upon fused borax have a general tendency to be higher by about 0.1 than those deduced from halide ratios.As experience has shown that the halide ratios mually afford a very trustworthy means of determining atomic weights and as moreover determinations in this laboratory which will form the subject of another com-munication confirmed the general result thereby attained we were led to suspect that in the use of borax there lay some hitherto unsuspected source of systematic error. The present determinations were therefore undertaken with a two-fold object first to obtain independent confirmation of the values for the atomic weight of boron obtained by earlier workers by the analysis of borax; and secondly to attempt to ascertain whether the fractional crystallisation of boric acid from aqueous solution produces any change in the isotope ratio of boron.As, for reasons which mill appear in due course the results give no information on the second point and as the general nature of such an enquiry has been discussed elsewhere (Robinson and Briscoe, preceding paper) this secondary object need but be mentioned here in explanation of the nature of the boric acid used as starting material . Outline of the Method. Of the possible means of analysing borax its titration with aqueous hydrochloric acid was chosen because the experimental procedure is relatively simple and therefore little liable to error, and because borax can thus be referred directly to silver chloride, for which the antecedent atomic weights are well known. Rimbach (Zoc. cit.) in his work upon the titration of borax decahydrate has shown that the reaction proceeds qua’ntitatively according to the equation : Na,B,O + 2HC1+ 5H,O + 2NaC1+ 4H3B03.Preliminary tests described later showed that the end-point coul THE USE OF FUSED BORAX IN THE DETERMINATIOK ETC. 153 be observed with an accuracy sufficient for our purpose and the work of Smith and van Haagen afforded the strongest evidence that fused borax was a definite compound well suited to precise work. Borax prepared from boric acid and pure sodium carbonate, was purified by fractional crystallisation with centrifugal drainage of the crystals and the borax decahydra,te obtained was fused in platinum vessels in an electrically heated muffle furnace in a current of dry air free from carbon dioxide.The resultant borax glass was weighed dissolved in water and titrated with NI5-hydro-chloric acid solution delivered from a weight burette ; the titration being completed by the addition of A7/100-acid solution from an accurately calibrated burette using methyl-red as indicator. The acid was standardised by taking a weighed quantity and pre-cipitating the whole of the chlorine as silver chloride which was collected dried and weighed. Hence the ratio actually determined wits Na,B,O 2AgC1. Therefore the following method was adopted : The Determination of Weight. All weighings of borax and silver chloride were made on the special Oertling balance sensitive to 0-01 mg. or less already described (Robinson and Briscoe Zoc. cit.) with weights carefully calibrated for relative weight in air and all usual precautions.Fused borax was weighed directly on the balance pan silver chloride was weighed in the tared bottle used to collect the precipitate. The standard acid solution was weighed in a stoppered burette, using another similar burette as tare on a standard No. 7 S.W. Oert-ling balance ca,rrying a load of 300 g. in each pan and sensitive to 0.1 mg. A second set of calibrated weights was used in this case. As the standard solution only served as a means of referring borax to silver chloride the relation of the second set of weights to the first is immaterial. The relative vacuum weights of borax and silver are given to the nearest 0.01 mg. those of standard acid to the nearest 0-5 mg. The vacuum corrections applied were calculated using the densities : silver chloride D = 5.50 ; borax D = 2.357 (Smith and van Haagen, Preparation of Reagents.The water and nitric acid were prepared by the methods described in a recent communication (Robinson and Briscoe Zoc. cit.) and the silver used was a part of the stock prepared for the work there described. Hydrochloric Acid.-About 3 litres of pure reagent acid main-zoc. cit.) 154 BRISCOE ROBINSON AND STEPHENSON : tained at boiling point in a glass flask were treated for 30 minutes with a rapid current of chlorine gas from a cylinder of liquid chlorine and the excess of chlorine was then boiled off. Thus any traces of bromine and iodine were removed. The acid then stood over-night with copper foil previously cleaned with nitric acid and distilled water to remove arsenic.FIG. 1. FIG. 2. Next day the acid was transferred to a resistance glass still fitted with a ground-in condenser heated for 15 minutes to cause a vigorous evolution of hydrogen chloride gas and thus to eliminate any remaining volatile impurity and finally diluted to a density of 1-10 and twice distilled considerable head and tail fractions being rejected in each distillation. The main fraction from the second distillation was used forthwith for the preparation of the standard acid. The approximate strength of the acid having been determined by diluting a weighed portion and titrating it agains THE USE O F FUSED BORAX IN THE DETERMINATION ETC. 155 fused borax the requisite volume of the acid was transferred to the storage bot,tle and diluted to 4000 C.C.with pure water. In order to ensure as far as possible that the concentration of the standard acid should remain constant during the series of titrations the storage bottle was fitted with a well-ground glass stopper of the form shown in Pig. 1 where A is a stopcock and capillary tube for delivery of acid and B is a second tube passing to the bottom of the bottle and also fitted with a stopcock to admit air when the hottle being inverted acid was withdrawn. Before each withdrawal of acid the contents of the bottle mere mixed by vigorous shaking and in order to minimise the volume of acid withheld from this agitation the tube B was made of capillary bore. This stopper was wired in place and was not removed during the whole series of standardisations and analyses.Before use, the bottle had been wdl cleaned and rinsed allowed’ to stand for some weelis full of approximately S /5-hydrochloric acid and then again well washed with water. A quantity of approximately N/IBO-acid was prepared by diluting in a calibrated flask a weighed quantity of the standard AT/5-acid : thus it was possible by use of an appropriate factor to convert the volume of K/lOO-acid used in each titration to the corresponding weight of N/5-acid which was added to the weight determined in the weight burette. I n the standardisation of the AV/5-acid a weighed quantity was delivered into a slight excess of aqueous silver nitrate prepared from pure silver and nitric acid. and the precipitated silver chloride mas collected washed and weighed by the method already de-scribed (Robinson and Briscoe Zoc.cit.). In this case of course, the chloride found nephelometrically in the washings and ammoniacal rinsings was calculated to silver chloride and added to the observed weight. The essential data of the standardisations are given in Table 11. TABLE 11. I. 11. Vacuum %-eight of acid taken ............... 154.4826 g. 160.9230 g. Vacuum weight of silver chloride ............ 4.52544 g. 4.71377 g. Ratio silver chloride hydrochloric acid ... 0.0029294 0.0020292 Mean value of ratio .............................. 0.0029293 All flasks and burettes used here and in the titrations were carefully calibrated by weighing out with water in the usual way. E’ractional Crystallisat ion of Eoric Acid.About 14 kilos of commercial boric acid supplied by Messrs. Borax Consolidated Ltd but oE unknown origin were fractionall 156 BRISCOE ROBINSON AND STEPHENSON : crystallised (about 70 crystallisations) until two series A and B, each of 19 fractions were obtained. Each fraction contained about 350 g. of boric acid dissolved in 1500 C.C. of water. The fractions were contained in 2-litre round Jena flasks fitted with loose glass stoppers to exclude dust crystallisation was effected by slowly heating the whole set of flasks to boiling on a large hot plate and allowing them to cool over-night. The liquor was removed from each flask by pouring off through a filter of copper gauze which retained all h e crystals then the filter was reversed and liquor from the flask next in the series was poured in through the filter thus rinsing the fine crystals back into the flask from which %hey had come.In this way beginning with the head (most soluble) fraction all the liquors were moved up one place at each crystallisation and distilled water was added to the tail fraction. After several (usually three or four) crystallisafions the fail fraction disappeared and the head liquors were evaporated to form a new head fraction. Eoth series were systematically recrys-tallised after this fashion and by careful mana,gement the fractions were kept reasonably constant in size and number. After the 60th crystallisation of both series about 1150 crystal-lisations in all. having been made in each the extreme head and tail fractions were rejected and the heads A30 and B30 the tails A14 and B15 and the middle fractions A23 and E24 were taken for conversion into borax.Preparation and Purification of Borax. In order to obtain strictly comparable samples of borax the operations here described were carried out simultaneously on all six samples of boric acid. About 120 g. of boric acid were added gradually to a slight excess (about 0.5%) over the calculated equivalent of pure sodium carbonate dissolved in about 200 C.C. of hot water. The sodium carbonate originally very pure had been twice recrystallised from water and was spectroscopically free from potassium. The solution was boiled to remove carbon dioxide filtered whilst hot and allowed to cool slowly without movement until about one-tenth of the borax had crystallised out.Then the clear liquor was poured off quickly into another vessel and further cooled with agitation, yielding a main crop of fine crystals of borax decahydrate which were filtered with suction and immediately transferred to ghss tubes for centrifugal drainage. The drainage tubes had in the bottom a thick pad of dry cotton wool covered with a wad of small filter papers and a porcelain disk they were filled with the damp salt closed by corks covered with filter paper and centrifuged fo THE USE OB FUSED BOUX IN THE DETERMINATION ETC. 157 10 minutes a t a speed of 2500 revolutions per minute. Thus the mother-liquor was very effectively removed. The salt contained in a platinum dish was first dehydrated in an electrically-heated silica muffle kept a t 200-300' until intumes-cence ceased and then fused by raising the temperature to about 700-800".It thus formed a perfectly clear glass having only a faint blue tinge due to copper derived from the wire gauze filter mentioned above. Some difficulty was experienced at first in dis-solving the borax glass in water later it was found that if the dish containing the fused glass were floated on cold water the mass developed numerous cracks which allowed water to penetrate the glass and greatly hastened solution. To dissolve the glass, the dish containing it was wholly immersed in water in a covered Durosil beaker heated over a Rose burner solution was usually complete in about 2 hours. After fusion the borax was recrystallised four times from water in the manner already described with centrifugal drainage of the main crop a t each stage.Thus each sample was recrystallised five times in all about one-tenth of the material in the head frac-tion and the same amount in the tail fraction being rejected each time. The crystallisations were done in covered Erlenmeyer flasks of Durosil both these and the beakers having been thoroughly cleaned and boiled out with water for several days before use. Tests for the chief impurities carried out on the mother-liquors (head fraction) from each crystallisation in which the impurities present tended to concentrate gave a valuable indication of the purity of the main fraction without sacrifice of material therefrom. Copper was estimated colorimetrically in an ammoniacal solution containing 1 g.of borax decahydrate; the standard solution of copper sulphate contained 1 part of copper in 100,000 parts and in each case a " blank " test was made on the reagents alone. The head fractions from the earlier crystallisations contained traces of copper but those from the final crystallisafion of all six samples were absolutely free from copper. Phosphctte.-Mofher-liquor equivalent to 1 g. of borax deca-hydrate was treated hot with nitric acid ammonium nitrate and ammonium molybdate. No trace of yellow ammonium phospho-molybdate was observed at any stage of the purification. SuZphte.-The solution of borax was acidified with hydrochloric acid treated with barium chloride solution boiled and kept, covered over-night.The mother-liquor from the third recrystal-lisations showed traces of sulphate but tests 011 the crystals gave negative results. Two further recrystallisations were made on eac 158 BRISCOE ROBINSON AND STEPHENSON : sample and the mother-liquors from these gave negative results on the sulphate tests. Chloride.-Nitric acid and silver nitrate were added to a solution containing 1 g. of crystallised borax. The third recrystallisation showed a slight opalescence; later tests mere negative, The final main fractions of borax were partly dehydrated by standing for 3 weeks over solid potassium hydroxide in desiccators and were then stored in stoppered bottles capped to exclude dust. Fusion of the pure borax was carried out in a platinum dish in the electrically heated muffle in a current of air free from carbon dioxide and dried over solid caustic potash.The weight of borax taken was such as to yield from 3-10 g. of resultant " glass," as required and the dehydration and fusion were conducted in the manner described above. When a clear glass was obtained usually after about 1-14 hours' fusion a further period in no case less than 2 hours was allowed for complete dehydration. A shallow dish was used for fusion and thus a large surface of borax was exposed. Samples of the molten glass were taken in two ways : (i) by pouring small beads on to a clean cold platinum surface; (ii) by dipping the bottom of a clean platinum crucible into the melt. In the latter case the glass adhered whilst hot but on cooling cracked off in pieces weighing about 0.3 g.The samples were immediately transferred to and kept in it desiccator con-taining solid caustic potash. As tests showed that when a bead of the glass weighing about 2 g. mas kept on the balance pan for 3 days its weight did not change appreciably this method of storage was evidently satis-factory. In every case the glass was weighed within 2 days after its fusion. Method of Analysis. Preliminary tests with a number of indicators showed that under the conditions of the titration methyl-red gave much the sharpest colour change especially when the end-point chosen was slightly on the acid side. Under these conditions in N/lO-borax solution 2-3 drops of A7/100-acid or alkali gave a considerable and sharp colour change.During further tests which showed that the end-point in the borax titration was unaffected by moderate additions of boric acid and sodium chloride it was observed that in a solution made from boric acid and salt in the concentration which would be produced in the titration the end-point was quite indefinite. This curious fact had no direct significance in the analyses now reported but its further investigation at a future time is proposed. The titrations were made by two dist'inct methods described below THE USE OF FUSED BORSX I N THE DETERMINATION ETC. 159 (a) Using the weight burette. Weighed beads of borax glass were dissolved to a 2% solution in water by boiling in a covered flask. After cooling a quantity of NlSacid calculated to be 990/ of that required for complete neutralisation was added from the weight burette a fixed volume of a stock solution of methyl-red was pipetted in and the titratioii was completed by adding N/lOO-acid from a burette.The quantity of the latter required to give a match with a standard of definite pink shade could be ascertained with certainty within 0.1 C.C. Afterwards a measured excess of N/100-acid was added then a measured excess of an equivalent solution of borax sufficient to make the solution distinctly yellow and the end-point was again determined with X/lOO-acid. (b) B y direct zueighingr of the acid soluliorb. As several of the earlier determinations showed unaccountable discrepancies which might have been due to loss of borax (by spurting during solution) or of acid (by splashing when pouring in from the weight burette), a method of titration was devised to eliminate the possibility of such errors.The apparatus shown in Fig. 2 consisted of a 200 C.C. conical flask A and a reflux condenser 13 with a spray-trap C, connected to A by a carefully ground joint. TABLE 111. Data of Titratiom of Borax Glass. Sample number. A14 1 2 3 3a 4 5 6 A24 1 2 3 4 5 6 A30 1 2 3 4 5 6 7 8 9 10 Vacuum wt. borax glass. 1.99413 1.97982 1.93077 0.81470 1.01079 1-04G41 1-03879 1.00759 1.01246 2.14923 1.84391 1.90780 1.95540 1.21340 1.46470 1.41400 1.31097 0.67774 0.90300 1.01566 1 -122 62 1.22427 1.46529 (g.) of Vacuum mt. N / 5 -acid. 96.G881 05.9G36 93.5412 39-4923 48.9751 50.6956 50.3263 48.82 5 8 49.1306 104.1576 89.3821 92.4578 94.73 15 58.8361 70.981 2 68.5272 63.5162 32.8309 43.6951 49-1406 54.2960 59.2639 70.92U1 k.) of Calculated vacuum mt.silver chloride. 2.83229 2.81 107 2.7401 1 1.1 5655 1.43463 1.48303 1.47431 1.43026 1.43919 3.05110 2.G1828 3.70838 2.77498 1.72349 2.07926 2.00737 1.86059 0.96172 1.28005 1.43948 1.59060 1.73603 2.07747 (g.1 of Ratio : borax glass silver chloride' 0.704068 0.704294 0-704632 0-704239 0.704564 0- 70463 8 0-704640 0.704481 0.703494 0.704412 0.704245 0.704407 0-704686 0.70403 G 0.704434 0.704403 0.704600 0.704717 0.705439 0.706574 0-705829 0.705215 0.705324 Atomic weight .10.961 10.977 11.001 10.973 10.996 11.002 11.002 10.990 10.920 10-985 10.974 10.985 11-005 10.958 10.987 10.985 10.999 11-007 11.059 11.069 11-087 11.043 11.05 16Q BRISCOE ROBINSON AND STEPHENSON : This apparatus was made in duplicate of Durosil glass and was cleaned with all the precautions already described. In using it, the required amount of N/5-acid was weighed in A against another similar flask as tare the weighed borax was carefully slid down the dry side of the flask into the acid the condenser was fitted a little pure water was introduced into the spray trap and the flask was heated gently over a Rose burner until the borax had completely dissolved (about 1 hour).After cooling the contents of the bulbs and condenser were rinsed into the flask these were removed and the titration was completed with N/lOO-acid as described above. Working in this manner the total volume of the solution titrated was little greater than that of the NIS-acid used. Statement a.nd Disczmion of Reszdts. Table 111 gives the essential data of all the analyses made and the values of the atomic weight of boron calculated therefrom. The results are evidently subject to errors (1) in the weight of borax taken (2) in the volume of N/lOO-acid used (3) in the weight of N/5-acid used in the titration and (4) in the standardisation of the iV/6-acid the probable magnitude and effect of these errors are shown in Table IV. TABLE Iv. Corresponding variation in the atomic weight of boron.-J= 0.005 f 0.001 & 0.001 3 0-004 Total maximum error ........................... f0-011 Making all due allowance for these errors it would appear that the value obtained in each titration for the atomic weight of boron should not be in error by more than one or two units in the second decimal place and there remain among the data considerable differ-ences which must be otherwise explained. After much worry had been caused by unaccountable and erratic variations in the atomic weight it became apparent as data accumulated that the result was in many cases EL function of the duration of fusion of the borax. Fortunately precise records of the fusions were available and their correlation with the results is shown in Table V.It is clear that a high atomic weight iq always associated with a long period of fusion. The additional results on A30 Nos. 7 8 9 and 10 confirm this in a striking way. The most probable inference is clear a higher atomic weight corresponds with a lesser amount of acid used to neutralise a given weight of borax hence to a lesser content of soda (Na,O) in that Probable maximum value of error. (1) 0.1 C.C. of N/lOO-acid on 1 g. of borax-glass ...... (2) 0.02 Mg. on 1 g. of borax .............................. (3) 0-5 Mg. on 50 g. of N/5-acid ........................ (4) Difference between standardisations 1 in 15,000 THE USE OF FUSED BORAX IN THE DETERMXNATIOS ETC. 161 Sample N O . A14 1 2 3 3-4 4 5 6 A24 1 r) r* 3 4 5 6 A30 1 2 3 4 5 6 7 8 9 10 TABLE V.Notes on Treatment oJ- Samples. Atomic weight. 10.961 10.977 11401 1 10.973) 10.996 11.002 11.002 10.990 1 10.920) ;;::::} 10.9851 10.974 I PO-958 10.987 10.985 10.999 11.0073\ Condition of fuaion of Method of analysis. borax g l m . Molten 4-5 hours. Extra 5 g. of crystals added to residue (1). Fused 34 hours. Both samples taken Extra crystals added to reei-due (2) ; fused 3 hours. Further amount of crystals minations by weight SimuItaneous deter-burette method. 9 ) dehydrated. Fused 62 hours. Simultaneous deter- 30 G. of crystals dehydrated minations by weight to fusion; kept molten burette method. 1 2 hours. Residue further fused 2 hours.Residue from 3 and 4 further 9 9 39 } fused 24 hours. 20 G. of crystals dehydrated s a m p 1 in g . and fused 6 hours before Weight burette used. =Yusik par- Residue from (1-5) fused for { 4houm. j 2 0 G. of crystals dehydrated Residue from 7 and 8 fused for i a further 7 hours. 9 9 and fused 10 hours. 11.043 \ Done by weight 11.051 J burette method. borax therefore on fusion borax loses soda and the extent of this loss increases with the duration of fusion. Molten borax glass is known t o volatilise but previous workers have differed as to whether any decomposition takes place. Walde-bott ( J . Amer. Chem. SOC. 1894,16 410) as a result of experiments on the fusion of borax was of the opinion that sodium tetraborate volatilised unchanged; he analysed the fused residue by treatment with ammonium fluoride and found no change in composition.The application of this method to an atomic weight determination w&s rejected however by Smith and van Haagen (Zoc. cit.). On the other hand Leonard (Chem. News 1598 77 104) in criticism of the determinabions of the atomic weight of boron by Armitage (Zoc. cit.) stated definitely that the residue obtained after fusion of borax required less standard acid for neutralisation than the original sample this corresponds with a loss of soda. Smith and van Haagen assumed that acidic gases from burners had affected Leonard’s borax but no such explanation can be advanced in thk case since an electrically heated furnace was used to fuse the borax. Their own samples of borax glass were fused to constant weight POT2.CSXVII. 162 CHALLENGER JINKS AND HASLAM THE SULPHUR in a long-necked platiiium flask and they concluded that the material condensed in the cooler part of the apparatus was identical in composition with the fused residue and was pure sodium tetra-borate. But their borax had been fused previously in an open crucible one may infer for about the same time in the case of all samples and the sample for analysis was taken from the upper portion of the main mass of glass which by rotation had been spread in a thin layer over the upper sides of the crucible and cooled quickly. Therefore the volatilisation of this sample in the flask without material change is quite consistent with its having already changed in Composition during the open fusion in the way now observed.Whatever the explanation may be it seems c1ea.r that the borax ratios are subject to grave uncertainty and the authors are definitely of opinion that all values for the atomic weight of boron determined by methods involving the weighing of fused borax should be entirely rejected in favour of those deduced from analyses of boron halides. In summarising the present results it seems proper to discard A24 No. 2 for unaccountable discordance and A30 Nos. 6 7 8 9, and 10 because in these cases the fusion was deliberately pro-longed. The mean atomic weight deduced from the remaining determinations B = 10.99 is in general agreement with the results previously obtained with fused borax and is close to the results of Ramsay and Aston for the ratio Na,B40,/2AgCl but differs widely from the results of determinations of the halide ratios.For reasons which have been made clear the authors attach no weight to this result. One of the authors (G. E. S.) desires to acknowledge a grant from the Department of Scientific and Industrial Research enabling him to take part in this investigation. UNIXERSITY OF DURHAM ARMSTRONG COLLEGE, NEWCASTLE-UPON -TPNE. [Received N m e m b e r 20th 1924. 150 BRISCOE ROBINSON AND STEPHENSON : XX1V.-The Use of Fused Borax i n the Determination of the Atomic Weight of Boron. By HENRY VINCENT AIRD BRISCOE PERCY LUCOCK ROBINSON, and GEORGE EDWARD STEPITENSON. THOUUH a few determinations of the atomic weight of boron have been made by ascertaining the ratio of boron halides to silver, the majority have depended in one way or another upon weighing anhydrous borax in the form of borax-glass.A detailed discussion of the results may be found in Clarke’s “ A Recalculation of the Atomic Weights ” (Hem. Nat. Acad. Sci. 1920 16 205) and it is sufficient here to give a list of the determinations. Table I records all the previous data classified according to the nature of the method employed the wide divergence among the results render THE USE OF FUSED BORAX W THE DETERMINATION ETC. 161 the “probable errors” obviously valueless as a measure of the weight to be given to each determination and they are therefore, omitted. TABLE I. Previous Determiirzations of the Atomic Weight of Boron. A. Determinations dependent upon weighing fused borax : Estimation of water in borax.11.023 10-859 1869. DobrovoEsky. Estinmtion of water in bomx. 1st series. 2nd serk. 11.529 1892. Hoskyns- Abrahall. Decahydrate to borax glass. 10.703 1893. Ramsay and Aston. Decahydrate to borax glass. 10.946 10.955 11.059 1898. Armitage. Decahydrate to borax glass. 10,986 uric acid. 10.933 1918. Smith and van Haagen. Borax glass to sodium sulphate. 10.904 10.903 10.905 10.899 10-901 :z :::27} Borax glass to sodium chloride. Borax glass to silver chloride. Borax glass titrated with sulph-Borax glass to sodium fluoride. Borax glass to sodium carbonate. Borax glass to sodium nitrate. Borax glass to sodium chloride. Mean of all determinations. 10.964 Mean excluding the first four results..10.939 B. Determinations of the ratio of a boron halide to silver or silver halide : 1899. Gautier. BBr 3AgBr. 11.016 BC‘I 3AgCl. 10.947 1922. Hiinigschmid and BCl 3AgCI. 10.840 Birkenbach. 10.818 10,825 1892. Hoskyns-Abrahall. BBr 3Ag. 10-880 Mean excluding Gautier’s results. 10.827 ~ C. Determinations of ratios which do not involve fused borax or a boron 1893. Rimbach. Na2B40,10H20 hydrochloric acid solution. 11.006 1899. Gautier. B2S 3BaS0,. 11.024 B,C 6C0,. 10.997 1923. Stock and Kuss. B,Hs 6H,. 10.806 halide : References Berzeliw Pogg. Ann. 1826 8 1 ; Laurent J . pr. Chem., 1849 47 415; Dobrovolsky Doctoral Disaertatwn Kiev 1869; Hoskyns-Abrahall J. 1892 61 650; Ramsay and Aston J. 1893 68 211; Armitage, P. 1898 14 22; Smith and van Haagen Carnegie Inst.Washington Pub-lication No. 267 1918; Gautier Ann. Chim. Phy8. 1899 [vii] 18 352; Hdnigschmid and Birkenbach Ber. 1923 56 [B] 1467; Rimbach Bcr., 1893 26 164; Stock and Kuss Ber. 1923 56 [B] 314; 8. anorg. Chem., 1923,128 49. Certain of these determinations (printed in italics) are unsatis-factory. In Group A those of Bemelius and Laurent have no pretensions to great accuracy Dobrovolsky’s two series are s 152 BRISCOE ROBINSON AND STEPHENSON : gravely discordant as to be valueless; and Abrahall’s results are definitely those of preliminary analyses thus these may be excluded making the mean result for the Group B = 10.939. I n Group B Gautier’s results show discordance so grave and so unaccountable in a halide ratio as to necessitate their immediate exclusion the remaining results give the mean value B = 10.827.I n Group C Gautier’s determinations involve eccentric ratios apparently unsuitable for atomic weight determination. The two remaining values may be retained but are not pertinent to the point now to be discussed. Reviewing all the results in Groups A and B not excluded for the reasons given it is immediately obvious that those dependent upon fused borax have a general tendency to be higher by about 0.1 than those deduced from halide ratios. As experience has shown that the halide ratios mually afford a very trustworthy means of determining atomic weights and as moreover determinations in this laboratory which will form the subject of another com-munication confirmed the general result thereby attained we were led to suspect that in the use of borax there lay some hitherto unsuspected source of systematic error.The present determinations were therefore undertaken with a two-fold object first to obtain independent confirmation of the values for the atomic weight of boron obtained by earlier workers by the analysis of borax; and secondly to attempt to ascertain whether the fractional crystallisation of boric acid from aqueous solution produces any change in the isotope ratio of boron. As, for reasons which mill appear in due course the results give no information on the second point and as the general nature of such an enquiry has been discussed elsewhere (Robinson and Briscoe, preceding paper) this secondary object need but be mentioned here in explanation of the nature of the boric acid used as starting material .Outline of the Method. Of the possible means of analysing borax its titration with aqueous hydrochloric acid was chosen because the experimental procedure is relatively simple and therefore little liable to error, and because borax can thus be referred directly to silver chloride, for which the antecedent atomic weights are well known. Rimbach (Zoc. cit.) in his work upon the titration of borax decahydrate has shown that the reaction proceeds qua’ntitatively according to the equation : Na,B,O + 2HC1+ 5H,O + 2NaC1+ 4H3B03. Preliminary tests described later showed that the end-point coul THE USE OF FUSED BORAX IN THE DETERMINATIOK ETC. 153 be observed with an accuracy sufficient for our purpose and the work of Smith and van Haagen afforded the strongest evidence that fused borax was a definite compound well suited to precise work.Borax prepared from boric acid and pure sodium carbonate, was purified by fractional crystallisation with centrifugal drainage of the crystals and the borax decahydra,te obtained was fused in platinum vessels in an electrically heated muffle furnace in a current of dry air free from carbon dioxide. The resultant borax glass was weighed dissolved in water and titrated with NI5-hydro-chloric acid solution delivered from a weight burette ; the titration being completed by the addition of A7/100-acid solution from an accurately calibrated burette using methyl-red as indicator. The acid was standardised by taking a weighed quantity and pre-cipitating the whole of the chlorine as silver chloride which was collected dried and weighed.Hence the ratio actually determined wits Na,B,O 2AgC1. Therefore the following method was adopted : The Determination of Weight. All weighings of borax and silver chloride were made on the special Oertling balance sensitive to 0-01 mg. or less already described (Robinson and Briscoe Zoc. cit.) with weights carefully calibrated for relative weight in air and all usual precautions. Fused borax was weighed directly on the balance pan silver chloride was weighed in the tared bottle used to collect the precipitate. The standard acid solution was weighed in a stoppered burette, using another similar burette as tare on a standard No.7 S.W. Oert-ling balance ca,rrying a load of 300 g. in each pan and sensitive to 0.1 mg. A second set of calibrated weights was used in this case. As the standard solution only served as a means of referring borax to silver chloride the relation of the second set of weights to the first is immaterial. The relative vacuum weights of borax and silver are given to the nearest 0.01 mg. those of standard acid to the nearest 0-5 mg. The vacuum corrections applied were calculated using the densities : silver chloride D = 5.50 ; borax D = 2.357 (Smith and van Haagen, Preparation of Reagents. The water and nitric acid were prepared by the methods described in a recent communication (Robinson and Briscoe Zoc. cit.) and the silver used was a part of the stock prepared for the work there described.Hydrochloric Acid.-About 3 litres of pure reagent acid main-zoc. cit.) 154 BRISCOE ROBINSON AND STEPHENSON : tained at boiling point in a glass flask were treated for 30 minutes with a rapid current of chlorine gas from a cylinder of liquid chlorine and the excess of chlorine was then boiled off. Thus any traces of bromine and iodine were removed. The acid then stood over-night with copper foil previously cleaned with nitric acid and distilled water to remove arsenic. FIG. 1. FIG. 2. Next day the acid was transferred to a resistance glass still fitted with a ground-in condenser heated for 15 minutes to cause a vigorous evolution of hydrogen chloride gas and thus to eliminate any remaining volatile impurity and finally diluted to a density of 1-10 and twice distilled considerable head and tail fractions being rejected in each distillation.The main fraction from the second distillation was used forthwith for the preparation of the standard acid. The approximate strength of the acid having been determined by diluting a weighed portion and titrating it agains THE USE O F FUSED BORAX IN THE DETERMINATION ETC. 155 fused borax the requisite volume of the acid was transferred to the storage bot,tle and diluted to 4000 C.C. with pure water. In order to ensure as far as possible that the concentration of the standard acid should remain constant during the series of titrations the storage bottle was fitted with a well-ground glass stopper of the form shown in Pig.1 where A is a stopcock and capillary tube for delivery of acid and B is a second tube passing to the bottom of the bottle and also fitted with a stopcock to admit air when the hottle being inverted acid was withdrawn. Before each withdrawal of acid the contents of the bottle mere mixed by vigorous shaking and in order to minimise the volume of acid withheld from this agitation the tube B was made of capillary bore. This stopper was wired in place and was not removed during the whole series of standardisations and analyses. Before use, the bottle had been wdl cleaned and rinsed allowed’ to stand for some weelis full of approximately S /5-hydrochloric acid and then again well washed with water. A quantity of approximately N/IBO-acid was prepared by diluting in a calibrated flask a weighed quantity of the standard AT/5-acid : thus it was possible by use of an appropriate factor to convert the volume of K/lOO-acid used in each titration to the corresponding weight of N/5-acid which was added to the weight determined in the weight burette.I n the standardisation of the AV/5-acid a weighed quantity was delivered into a slight excess of aqueous silver nitrate prepared from pure silver and nitric acid. and the precipitated silver chloride mas collected washed and weighed by the method already de-scribed (Robinson and Briscoe Zoc. cit.). In this case of course, the chloride found nephelometrically in the washings and ammoniacal rinsings was calculated to silver chloride and added to the observed weight. The essential data of the standardisations are given in Table 11.TABLE 11. I. 11. Vacuum %-eight of acid taken ............... 154.4826 g. 160.9230 g. Vacuum weight of silver chloride ............ 4.52544 g. 4.71377 g. Ratio silver chloride hydrochloric acid ... 0.0029294 0.0020292 Mean value of ratio .............................. 0.0029293 All flasks and burettes used here and in the titrations were carefully calibrated by weighing out with water in the usual way. E’ractional Crystallisat ion of Eoric Acid. About 14 kilos of commercial boric acid supplied by Messrs. Borax Consolidated Ltd but oE unknown origin were fractionall 156 BRISCOE ROBINSON AND STEPHENSON : crystallised (about 70 crystallisations) until two series A and B, each of 19 fractions were obtained.Each fraction contained about 350 g. of boric acid dissolved in 1500 C.C. of water. The fractions were contained in 2-litre round Jena flasks fitted with loose glass stoppers to exclude dust crystallisation was effected by slowly heating the whole set of flasks to boiling on a large hot plate and allowing them to cool over-night. The liquor was removed from each flask by pouring off through a filter of copper gauze which retained all h e crystals then the filter was reversed and liquor from the flask next in the series was poured in through the filter thus rinsing the fine crystals back into the flask from which %hey had come. In this way beginning with the head (most soluble) fraction all the liquors were moved up one place at each crystallisation and distilled water was added to the tail fraction.After several (usually three or four) crystallisafions the fail fraction disappeared and the head liquors were evaporated to form a new head fraction. Eoth series were systematically recrys-tallised after this fashion and by careful mana,gement the fractions were kept reasonably constant in size and number. After the 60th crystallisation of both series about 1150 crystal-lisations in all. having been made in each the extreme head and tail fractions were rejected and the heads A30 and B30 the tails A14 and B15 and the middle fractions A23 and E24 were taken for conversion into borax. Preparation and Purification of Borax. In order to obtain strictly comparable samples of borax the operations here described were carried out simultaneously on all six samples of boric acid.About 120 g. of boric acid were added gradually to a slight excess (about 0.5%) over the calculated equivalent of pure sodium carbonate dissolved in about 200 C.C. of hot water. The sodium carbonate originally very pure had been twice recrystallised from water and was spectroscopically free from potassium. The solution was boiled to remove carbon dioxide filtered whilst hot and allowed to cool slowly without movement until about one-tenth of the borax had crystallised out. Then the clear liquor was poured off quickly into another vessel and further cooled with agitation, yielding a main crop of fine crystals of borax decahydrate which were filtered with suction and immediately transferred to ghss tubes for centrifugal drainage.The drainage tubes had in the bottom a thick pad of dry cotton wool covered with a wad of small filter papers and a porcelain disk they were filled with the damp salt closed by corks covered with filter paper and centrifuged fo THE USE OB FUSED BOUX IN THE DETERMINATION ETC. 157 10 minutes a t a speed of 2500 revolutions per minute. Thus the mother-liquor was very effectively removed. The salt contained in a platinum dish was first dehydrated in an electrically-heated silica muffle kept a t 200-300' until intumes-cence ceased and then fused by raising the temperature to about 700-800". It thus formed a perfectly clear glass having only a faint blue tinge due to copper derived from the wire gauze filter mentioned above. Some difficulty was experienced at first in dis-solving the borax glass in water later it was found that if the dish containing the fused glass were floated on cold water the mass developed numerous cracks which allowed water to penetrate the glass and greatly hastened solution.To dissolve the glass, the dish containing it was wholly immersed in water in a covered Durosil beaker heated over a Rose burner solution was usually complete in about 2 hours. After fusion the borax was recrystallised four times from water in the manner already described with centrifugal drainage of the main crop a t each stage. Thus each sample was recrystallised five times in all about one-tenth of the material in the head frac-tion and the same amount in the tail fraction being rejected each time.The crystallisations were done in covered Erlenmeyer flasks of Durosil both these and the beakers having been thoroughly cleaned and boiled out with water for several days before use. Tests for the chief impurities carried out on the mother-liquors (head fraction) from each crystallisation in which the impurities present tended to concentrate gave a valuable indication of the purity of the main fraction without sacrifice of material therefrom. Copper was estimated colorimetrically in an ammoniacal solution containing 1 g. of borax decahydrate; the standard solution of copper sulphate contained 1 part of copper in 100,000 parts and in each case a " blank " test was made on the reagents alone. The head fractions from the earlier crystallisations contained traces of copper but those from the final crystallisafion of all six samples were absolutely free from copper.Phosphctte.-Mofher-liquor equivalent to 1 g. of borax deca-hydrate was treated hot with nitric acid ammonium nitrate and ammonium molybdate. No trace of yellow ammonium phospho-molybdate was observed at any stage of the purification. SuZphte.-The solution of borax was acidified with hydrochloric acid treated with barium chloride solution boiled and kept, covered over-night. The mother-liquor from the third recrystal-lisations showed traces of sulphate but tests 011 the crystals gave negative results. Two further recrystallisations were made on eac 158 BRISCOE ROBINSON AND STEPHENSON : sample and the mother-liquors from these gave negative results on the sulphate tests.Chloride.-Nitric acid and silver nitrate were added to a solution containing 1 g. of crystallised borax. The third recrystallisation showed a slight opalescence; later tests mere negative, The final main fractions of borax were partly dehydrated by standing for 3 weeks over solid potassium hydroxide in desiccators and were then stored in stoppered bottles capped to exclude dust. Fusion of the pure borax was carried out in a platinum dish in the electrically heated muffle in a current of air free from carbon dioxide and dried over solid caustic potash. The weight of borax taken was such as to yield from 3-10 g. of resultant " glass," as required and the dehydration and fusion were conducted in the manner described above. When a clear glass was obtained usually after about 1-14 hours' fusion a further period in no case less than 2 hours was allowed for complete dehydration.A shallow dish was used for fusion and thus a large surface of borax was exposed. Samples of the molten glass were taken in two ways : (i) by pouring small beads on to a clean cold platinum surface; (ii) by dipping the bottom of a clean platinum crucible into the melt. In the latter case the glass adhered whilst hot but on cooling cracked off in pieces weighing about 0.3 g. The samples were immediately transferred to and kept in it desiccator con-taining solid caustic potash. As tests showed that when a bead of the glass weighing about 2 g. mas kept on the balance pan for 3 days its weight did not change appreciably this method of storage was evidently satis-factory.In every case the glass was weighed within 2 days after its fusion. Method of Analysis. Preliminary tests with a number of indicators showed that under the conditions of the titration methyl-red gave much the sharpest colour change especially when the end-point chosen was slightly on the acid side. Under these conditions in N/lO-borax solution 2-3 drops of A7/100-acid or alkali gave a considerable and sharp colour change. During further tests which showed that the end-point in the borax titration was unaffected by moderate additions of boric acid and sodium chloride it was observed that in a solution made from boric acid and salt in the concentration which would be produced in the titration the end-point was quite indefinite.This curious fact had no direct significance in the analyses now reported but its further investigation at a future time is proposed. The titrations were made by two dist'inct methods described below THE USE OF FUSED BORSX I N THE DETERMINATION ETC. 159 (a) Using the weight burette. Weighed beads of borax glass were dissolved to a 2% solution in water by boiling in a covered flask. After cooling a quantity of NlSacid calculated to be 990/ of that required for complete neutralisation was added from the weight burette a fixed volume of a stock solution of methyl-red was pipetted in and the titratioii was completed by adding N/lOO-acid from a burette. The quantity of the latter required to give a match with a standard of definite pink shade could be ascertained with certainty within 0.1 C.C.Afterwards a measured excess of N/100-acid was added then a measured excess of an equivalent solution of borax sufficient to make the solution distinctly yellow and the end-point was again determined with X/lOO-acid. (b) B y direct zueighingr of the acid soluliorb. As several of the earlier determinations showed unaccountable discrepancies which might have been due to loss of borax (by spurting during solution) or of acid (by splashing when pouring in from the weight burette), a method of titration was devised to eliminate the possibility of such errors. The apparatus shown in Fig. 2 consisted of a 200 C.C. conical flask A and a reflux condenser 13 with a spray-trap C, connected to A by a carefully ground joint.TABLE 111. Data of Titratiom of Borax Glass. Sample number. A14 1 2 3 3a 4 5 6 A24 1 2 3 4 5 6 A30 1 2 3 4 5 6 7 8 9 10 Vacuum wt. borax glass. 1.99413 1.97982 1.93077 0.81470 1.01079 1-04G41 1-03879 1.00759 1.01246 2.14923 1.84391 1.90780 1.95540 1.21340 1.46470 1.41400 1.31097 0.67774 0.90300 1.01566 1 -122 62 1.22427 1.46529 (g.) of Vacuum mt. N / 5 -acid. 96.G881 05.9G36 93.5412 39-4923 48.9751 50.6956 50.3263 48.82 5 8 49.1306 104.1576 89.3821 92.4578 94.73 15 58.8361 70.981 2 68.5272 63.5162 32.8309 43.6951 49-1406 54.2960 59.2639 70.92U1 k.) of Calculated vacuum mt. silver chloride. 2.83229 2.81 107 2.7401 1 1.1 5655 1.43463 1.48303 1.47431 1.43026 1.43919 3.05110 2.G1828 3.70838 2.77498 1.72349 2.07926 2.00737 1.86059 0.96172 1.28005 1.43948 1.59060 1.73603 2.07747 (g.1 of Ratio : borax glass silver chloride' 0.704068 0.704294 0-704632 0-704239 0.704564 0- 70463 8 0-704640 0.704481 0.703494 0.704412 0.704245 0.704407 0-704686 0.70403 G 0.704434 0.704403 0.704600 0.704717 0.705439 0.706574 0-705829 0.705215 0.705324 Atomic weight .10.961 10.977 11.001 10.973 10.996 11.002 11.002 10.990 10.920 10-985 10.974 10.985 11-005 10.958 10.987 10.985 10.999 11-007 11.059 11.069 11-087 11.043 11.05 16Q BRISCOE ROBINSON AND STEPHENSON : This apparatus was made in duplicate of Durosil glass and was cleaned with all the precautions already described.In using it, the required amount of N/5-acid was weighed in A against another similar flask as tare the weighed borax was carefully slid down the dry side of the flask into the acid the condenser was fitted a little pure water was introduced into the spray trap and the flask was heated gently over a Rose burner until the borax had completely dissolved (about 1 hour). After cooling the contents of the bulbs and condenser were rinsed into the flask these were removed and the titration was completed with N/lOO-acid as described above. Working in this manner the total volume of the solution titrated was little greater than that of the NIS-acid used. Statement a.nd Disczmion of Reszdts.Table 111 gives the essential data of all the analyses made and the values of the atomic weight of boron calculated therefrom. The results are evidently subject to errors (1) in the weight of borax taken (2) in the volume of N/lOO-acid used (3) in the weight of N/5-acid used in the titration and (4) in the standardisation of the iV/6-acid the probable magnitude and effect of these errors are shown in Table IV. TABLE Iv. Corresponding variation in the atomic weight of boron. -J= 0.005 f 0.001 & 0.001 3 0-004 Total maximum error ........................... f0-011 Making all due allowance for these errors it would appear that the value obtained in each titration for the atomic weight of boron should not be in error by more than one or two units in the second decimal place and there remain among the data considerable differ-ences which must be otherwise explained.After much worry had been caused by unaccountable and erratic variations in the atomic weight it became apparent as data accumulated that the result was in many cases EL function of the duration of fusion of the borax. Fortunately precise records of the fusions were available and their correlation with the results is shown in Table V. It is clear that a high atomic weight iq always associated with a long period of fusion. The additional results on A30 Nos. 7 8 9 and 10 confirm this in a striking way. The most probable inference is clear a higher atomic weight corresponds with a lesser amount of acid used to neutralise a given weight of borax hence to a lesser content of soda (Na,O) in that Probable maximum value of error.(1) 0.1 C.C. of N/lOO-acid on 1 g. of borax-glass ...... (2) 0.02 Mg. on 1 g. of borax .............................. (3) 0-5 Mg. on 50 g. of N/5-acid ........................ (4) Difference between standardisations 1 in 15,000 THE USE OF FUSED BORAX IN THE DETERMXNATIOS ETC. 161 Sample N O . A14 1 2 3 3-4 4 5 6 A24 1 r) r* 3 4 5 6 A30 1 2 3 4 5 6 7 8 9 10 TABLE V. Notes on Treatment oJ- Samples. Atomic weight. 10.961 10.977 11401 1 10.973) 10.996 11.002 11.002 10.990 1 10.920) ;;::::} 10.9851 10.974 I PO-958 10.987 10.985 10.999 11.0073\ Condition of fuaion of Method of analysis.borax g l m . Molten 4-5 hours. Extra 5 g. of crystals added to residue (1). Fused 34 hours. Both samples taken Extra crystals added to reei-due (2) ; fused 3 hours. Further amount of crystals minations by weight SimuItaneous deter-burette method. 9 ) dehydrated. Fused 62 hours. Simultaneous deter- 30 G. of crystals dehydrated minations by weight to fusion; kept molten burette method. 1 2 hours. Residue further fused 2 hours. Residue from 3 and 4 further 9 9 39 } fused 24 hours. 20 G. of crystals dehydrated s a m p 1 in g . and fused 6 hours before Weight burette used. =Yusik par- Residue from (1-5) fused for { 4houm. j 2 0 G. of crystals dehydrated Residue from 7 and 8 fused for i a further 7 hours.9 9 and fused 10 hours. 11.043 \ Done by weight 11.051 J burette method. borax therefore on fusion borax loses soda and the extent of this loss increases with the duration of fusion. Molten borax glass is known t o volatilise but previous workers have differed as to whether any decomposition takes place. Walde-bott ( J . Amer. Chem. SOC. 1894,16 410) as a result of experiments on the fusion of borax was of the opinion that sodium tetraborate volatilised unchanged; he analysed the fused residue by treatment with ammonium fluoride and found no change in composition. The application of this method to an atomic weight determination w&s rejected however by Smith and van Haagen (Zoc. cit.). On the other hand Leonard (Chem. News 1598 77 104) in criticism of the determinabions of the atomic weight of boron by Armitage (Zoc.cit.) stated definitely that the residue obtained after fusion of borax required less standard acid for neutralisation than the original sample this corresponds with a loss of soda. Smith and van Haagen assumed that acidic gases from burners had affected Leonard’s borax but no such explanation can be advanced in thk case since an electrically heated furnace was used to fuse the borax. Their own samples of borax glass were fused to constant weight POT2. CSXVII. 162 CHALLENGER JINKS AND HASLAM THE SULPHUR in a long-necked platiiium flask and they concluded that the material condensed in the cooler part of the apparatus was identical in composition with the fused residue and was pure sodium tetra-borate. But their borax had been fused previously in an open crucible one may infer for about the same time in the case of all samples and the sample for analysis was taken from the upper portion of the main mass of glass which by rotation had been spread in a thin layer over the upper sides of the crucible and cooled quickly. Therefore the volatilisation of this sample in the flask without material change is quite consistent with its having already changed in Composition during the open fusion in the way now observed. Whatever the explanation may be it seems c1ea.r that the borax ratios are subject to grave uncertainty and the authors are definitely of opinion that all values for the atomic weight of boron determined by methods involving the weighing of fused borax should be entirely rejected in favour of those deduced from analyses of boron halides. In summarising the present results it seems proper to discard A24 No. 2 for unaccountable discordance and A30 Nos. 6 7 8 9, and 10 because in these cases the fusion was deliberately pro-longed. The mean atomic weight deduced from the remaining determinations B = 10.99 is in general agreement with the results previously obtained with fused borax and is close to the results of Ramsay and Aston for the ratio Na,B40,/2AgCl but differs widely from the results of determinations of the halide ratios. For reasons which have been made clear the authors attach no weight to this result. One of the authors (G. E. S.) desires to acknowledge a grant from the Department of Scientific and Industrial Research enabling him to take part in this investigation. UNIXERSITY OF DURHAM ARMSTRONG COLLEGE, NEWCASTLE-UPON -TPNE. [Received N m e m b e r 20th 1924.