DEAN : TEE ATOMIC WEIGHT OF NITROGEN. 117 XII.-Tlze Atonaic Weiyht of -Nitrogen. By GEORGE DEAN, B.A. CONSIDERING the large number of very definite aud stable compounds in which nitrogen occurs as a constituent, it might have been ex- pected that the atomic weight of nitrogen would be known with the greatest exactness. The readiness and accuracy with which nitrates may be converted into chlorides and wice vem&, the determination of the percentage of silver in silver nitrate by synthetical and by analy- tical methods, the ratio between ammonium chloride or bromide and the silver required for complete precipitation, and many other exact processes seem fully to warrant this supposition, and t o enable us to deduce directly the ratio of the atomic weight of nitrogen to those of both hydrogen and oxygen.A glance a t the summary of results given below, however, will show that much work has still t o be done before we know the atomic weight of nitrogen as accurately as we have every reason t o believe it can be determined with the resources now at our command. The experiments described in the following paper were undertaken with the view of deducing the constant from a combination of elements which had not hitherto been used for the purpose. Before entering upon the account of the method adopted here and the results obtained, it may be of interest to briefly consider the most important work done by previous investigators. Both for variety of method and the number of experiments, the researches of Stas take the foremost place. Making oxygen the standard of comparison, and giving t o it a value of 16, the different numbers obtained by him for nitrogen, with the respective methods employed, are as follows : (1) Comparison of ammonium chloride and metallic silver (3) Conversion of silver into silver nitrate.(Two series of experiments, the atomic weight having, in each series, two values, deduced respectively from the weight of silver nitrate before and after fusion). First series, silver nitrate before Fusion.. ................ > Y Y 9 9 9 after ,, .................. Second ,, ,? before , , 9 9 after ,, P? 7, ,¶ bromide ,, 9 , .................. .................. >9 9 9 (4) Comparison of ammonium chloride and silver ni tmte,. ( 5 ) Conversion of potassium chloride into potassium nitrate (6) ¶? sodium ?) sodium ,) VOL, LXXVII.14.043 14.048 14,044 14.029 14.054 14.042 14.027 14.044 14.046 K118 DEAN : THE ATOMIC WEIGIHT OF NITROGIEN. (7) Comparison of silver nitrate and potassium chloride : First series. ..................................................... 14.083 Second ,, ...................................................... 14-1 05 Third ,, ...................................................... 14.043 The general mean of these values, assigning equal importance to each, is 14.051, the lowest value obtained being 14.027, the highest 14-105. Next in order of importance t o the work of Stas is probably that of Penny (Phil. Trans., 1839, 129, I, 32) ; his various methods and the results respectively deduced from them may be briefly given thus : (1) Conversion of metallic silver into nitrate ...............(2) 9 ) silver nitrate into silver chloride.. ....... (3) ,, p Aassium nitrate into potassium chloride (4) 9 9 ,, chloride ,, nitrate.. (5) (6) (7) (8) 9 , ,, chlornte 99 99 9 , sodium 99 sodium ,, 9 , 9 9 chloride 9 , 9 , 9 9 9 , nitrate ,, chloride 13.996 14.011 14.037 14,039 14-003 14.031 14-025 14.02 1 The mean of these is 14.020, the lowest value beicg 13.996, the highest 14.039. Marignac treated the question less exhaustively ; his methods with their corresponding results were : (1) Comparison of metallic silver and ammonium chloride 13.961 1 , 7, silver nitrate ...... 13-977 (3) ?, silver nitrate and potassium chloride.. . 14.150 (2) the mean being 14.029. Thus the values for the constant, determined by three of the masters in this branch of research, are practically 14.02, 14.03, and 14.05, the separate experiments yielding results varying from 13.961 to 14.150.Other workers have used methods more or less similar with the following results. Pelouze ( C m p t . rend., 1845, 20, 1047) determined the weight of pure silver which, when dissolved in nitric acid, was sufficient for the complete precipitation of a weighed amount of pure ammonium chloride. The atomic weight of nitrogen calculated from his numbers is 13.975. Hardin (J. Amer. Chem. Soc., 1896, 18, 995) electrolysed small weighed amounts of silver nitrate, and weighed the silver deposited. His data lead to the value 14.042. Turner (Phil. Trans., 1833, 123, 11, 537) converted a weighedDEAN : THE ATOMIC WEIUHT OF NITROGEN. 119 amount of silver nitrate into silver chloride, and determined the mass of chloride produced.Hibbs (J. Amer. Chem. SOC., 1896, 18, 1044) heated known masses of potassium nitrate in a stream of hydrogen chloride, and weighed the amount of chloride obtained. Moreover, the same process was applied to the sodium compound, the results being 14.032 and 14.026 respectively. Thomsen (Zeit. physikccl. Chem., 1894, 13, 398) determined the ratio of the weights of hydrogen chloride and ammonia which combine with each other. He passed pure, dry hydrogen chloride into a weighed apparatus containing distilled water, and weighed again. Then pure ammonia was led in until it was present in slight excess, and the in- crease in weight observed, the excess of ammonia being finally deter- mined by titration with standard acid, His mean result, 2.13934, leads to the value 14.021 for nitrogen.Excluding the value deduced by Thomsen-to which little weight can be assigned as his experiments also lead to the conclusion that the ratio of H : 0 is 1 : 16 instead of 1 : 15.88 or 1 : 15.89 as estab- lished by the laborious researches of Rayleigh, Leduc, Morley, Scott, Noyes, and others-the mean is 14.034, if equal importance is given to the result of each separate series of experiments. So far the results considered have been those which are based upon purely chemical methods. In the various determinations of the density of nitrogen, we have, on the other hand, a series of values obtained by physical means.The work of the earlier experimenters, Biot and Arago, Thomson, &c., was carried on without the refinements of accuracy brought to the aid of later research, and may be passed over here. More exact estimations have been made by Dumas and Boussingault, Regnault, von Jolly, Leduc, and Rayleigh, but chiefly with residual atmospheric nitrogen, and therefore still containing argon. In consequence of the admixture of this substance with the nitrogen, these results also are of little value for our present purpose. The most recent numbers obtained for p w e nitrogen, both by Lord Rayleigh and M. Leduc, however, are almosb identical. Taking oxygen as the unit, Leduc, from his own experiments, gives to nitrogen a density of 0.87508, and from Rayleigh’s data, 0937507 (Compt.qqend., 1898, 126, 415). Lord Rayleigh refers his results to the density of air as unit, and obtains for nitrogen and oxygen the densities 0.9673’7 and 1.10535 respectively; hence the relative densities are 14.003 : 16 (Proc. Roy. SOC., 1897, 62, 209). Now the great similarity in behavioiir of oxygen and nitrogen, with regard to changes in temperature and pressure, renders it almost impossible that any deviations from Avogadro’s lam mould be able t o reconcile the two values of 14,034 as found by chemical methods and His experiments give the value 14.013. K 2120 DEAN : THE ATOMIC WEIGHT OF NITROGEN. 14.003 by physical methods. This view is supported by D. Berthelot’s recalculation of the atomic weight of nitrogen from the densities of nitrogen and oxygen, on the assumption that Avogadro’s law is strictly accurate at low pressures.After applying a correction for the differences in compressibility of the two gases, tlhe ratio is only raised to 14.007 : 16 (Compt. rend., 1898, 126, 954). It was therefore thought to be of the greatest importance to redetermine this constant by some new method involving as few atomic weights as possible, and only those which are known with the highest degree of accuracy. itfethod Employed. Some years ago, when discussing the probable cause of the differ- ences between Stas’ numbers deduced from the weights of fused and unfused silver nitrate respectively, Professor Dewar suggested the use of silver cyanide in order to obtain an independent value. By deter- mining the amount of silver in a known weight of cyanide, the equi- valent of cyanogen could be estimated; by subtracting from this the atomic weight of carbon, that of nitrogen is obtained.Many preliminary experiments were made before the final method of treatment was decided upon. Of course, the simplest plan, which at once suggested itself, was to heat a weighed amount of cyanide, and weigh the silver left. Unfortunately, the formation of para- cyanogen and a carbide of silver in the mass of metal, and the appreciable volatility of silver when heated for a fair length of time in the air, led to an utter lack of agreement among the results obtained. Attempts were made, on the other hand, with varying degrees of success, to dissolve weighed amounts of the cyanide in nitric acid alone, in nitric acid with some other oxidising agent (for example, potassium permanganate), and in nitric acid under pressure, and to estimate the weight of silver in solution.Ultimately this difficulty was overcome, and the method resolved itself into the follow- ing steps : I. Preparation of pure silver sulphate ; 11. 99 ,, hydrocyanic acid ; 111. 9 , ,, silver cyanide, free from sulphuric acid ; IV. Y, nitric acid, free from haloid acids ; V. Drying of the cyanide until its weight remained constant ; VI. Conversion of the weighed cyanide into some soluble silver salt ; VII. Estimation of the amount of silver in solution. These objects were obtained in the following manner. First, ordinary silver nitrate was dissolved in water and precipitated by means ofDEAN : THE ATOMIC WEIGHT OF NITROGEN.121 redistilled sulphuric acid. The fine crystals of silver sulphate were drained, recrystallised twice from a large volume of distilled water, and finally made up into a dilute solution. Next, a weak solution of hydrocyanic acid mas prepared by the following method. Potassium ferrocyanida was recrystallised in a fine state of division, and distilled with dilute sulphuric acid, in an apparatus so arranged that any liquid thrown up in the act of ebullition was reflected back into the flask. The distillate was finally redistilled over a little magnesium carbonate, which was used to prevent ‘‘ bumping ” chiefly, but also to combine with any traces of sulphuric acid which might be present. The silver sulphate solution was placed in a stoppered bottle, dilute hydrocyanic acid added, and the whole well shaken, to render the pre- cipitate flocculent.More acid was added, and the process repeated until all the silver was precipitated as cyanide ; the clear liquid was then poured off, and more sulphate solution added ,and precipitated in a similar manner. When a suitable quantity of cyanide had been obtained, it was repeatedly washed with cold, and finally with hot distilled water, and allowed to stand for some weeks, the water being occasionally renewed. Any traces of sulphate of silver enclosed in the flocculi would thus have the opportunity of diffusing out. The cyanide was finally dried, as described later. Nitric Acid-The “pure” acid of commerce, sp. gr. 1-42, was twice redistilled, the first time with a few drops of silver nitrate solution, precautions being taken to avoid spirtings being carried over into the distillate.The acid finally collected, on being tested in the Stas chamber, was absolutely free from either hydrochloric acid or silver. Potassium Bromide, used in determining the amount of silver in solu- tion.-For this I am indebted to the kindness of Dr. Scott, Superin- tendent of the Davy-Faraday Research Laboratory. It is part of the sample used by Dewar and Scott in their determination of the atomic weight of manganese (Proc. Roy. Soc., 1883,35,44), and was prepared from potassium carbonate, obtained by decomposing carefully recrystal- lised potassium bitartrate, and pure hydrobromic acid, obtained from the distillation of potassium bromide and sulphuric acid somewhat diluted.Drying of the Silver Cyanide. The carefully washed precipitate was placed in a clean porcelain basin and as much as possible of the water poured away. After heating for about 12 hours in a steam oven, the basin was placed over concentrated sulphuric acid in a vacuum desiccator, which was then exhausted. The salt was left drying in this manner for a week, the surface of the acid being renewed by occasional agitation. A portion was then transferred to two platinum boats, which were enclosed in thin122 DEAN: THE ATOMIC WEIGET OF NITROGEN. glass tubes, sliding the one over the other, and fitting fairly tightIg, to prevent moisture being absorbed from the air. The whole-case, boats, and cyanide-was then weighed. The actual numbers obtained throughout in one determination are given below.July llth, 1898. Boats +tubes + cyanide. 67,947 grams. The boats were afterwards placed in a wide glass tube, drawn out at one end, and connected there with a series of U tubes, Bc., con- taining calcium chloride, phosphoric oxide, solid caustic potash, &c. Over the other end fitted a slightly wider tube, also drawn out at one end, and connected through a drying tube with a water pump. By means of this arrangement, a slow, steady current of pure, dry air (freed from traces of carbon dioxide and sulphuretted hydrogen by passage through a series of potash bulbs) could be drawn over the cyanide. The tube containing the boats was kept a t a fairly high temperature by means of an annular heater containing boiling xylene (b.p. 135'). c The boats were heated in this way in dry air, constantly renewed,, for 17 hours, allowed to cool all night in the heater, transferred next day to the tubes, and weighed. July 13th. Boats + tubes +cyanide. 67-9466 grams. The cyanide had lost 1.1 milligrams in 17 hours. The boats were again placed in the tube and heated for 18 hours, allowed to cool as before, and weighed. July 15th. Boats + tubes +cyanide. 67.94665 grams. After 18 hours heating, the weight was thus practically unaltered.DEAN: THE ATOMIC WEIGHT OF NITROGEN. 123 Solution of the Silver Cyanide. This was effected in a glass bulb of about 300 C.C. capacity, pro- vided with a long neck. The end of the neck was turned out slightly, and had a small lip in order to make it easier to pour accurately.A small glass condenser was ground to fit into the neck of the bulb, the upper end of the condenser being again fitted with a set of three small bulbs, also ground in. By taking these precautions, all loss of liquid by spirting was effectually prevented. The bulb was disconnected from the condenser, set up vertically over a sheet of paper, and a funnel placed in the neck. To transfer the cyanide, the boats mere carefully lifted over the funnel and gently tapped. any granules remaining on the sides of the funnel being finally washed in by means of pure nitric acid. The boats and tubes mere then weighed. Boats + tubes ........................ 55,8347 grams. Weight of cyanide taken ......... 12.11195 ,, In all 60 C.C. of nitric acid were added, the bulb then attached to the condenser, and the contents kept gently simmering on a sand-bath until the solid had completely dissolved.This took place in about 40 hours, thorough conversion into nitrate being marked by the ‘‘ bumping ” of the liquid. It was found that fuming nitric acid did not dissolve the silver cyanide nearly as readily as the 68 per cent, acid, on account of the insoluble nitrate being precipitated upon it, and protecting it against further action. Determination of the Xilvw. For this, the bulbs and condenser were carefully rinsed into a large stoppered bottle by means of distilled water and the contents of the large bulb added. The latter was repeatedly washed by boiling a little water in it and allowing the condensed water to run down the sides, the different washings being added to the main portion of the liquid.Next the amount of potassium bromide necessary for the theoretical weight of silver present was calculated, and weighed out into a small beaker. I n the particular experiment the weights were those given below : July 20th. 6eaker .............................. ,, +bromide ............... Weight of potassium bromide taken.. . under consideration, 11.8087 grams. 2205716 ,, 10.7629 ,,124 DEAN : THE ATOMIC WEiGHT OF NITROGEN. I n order to protect the silver bromide, when precipitated, from the action of the light, the bottle containing the silver solution was wrapped in red paper. The weighed amount of potassium bromide in the beaker was dissolved ir, a little distilled water, and the solution transferred with the utmost care to the silver solution.Finally, the bottle was vigorously shaken a t intervals in order to procure a perfectly clear liquid above the precipitated silver bromide, and allowed to stand for a day. The next step was the estimation of the amount of silver or potassium bromide present in excess. Standard solutions of these substances were employed (1 gram of the former containing 0.001 17 gram of silver, and 1 gram of the latter being equivalent to 0.00094 gram of silver) and small quantities added from weighed stoppered burettes until the end point was determined, that is, until further addition produced no turbidity. The weight of solution added furnished the weight of silver or bromide needed, and this, with the weight of bromide originally added, gave the means of determining the exact weight of silver in the weight of silver cyanide taken.The titration was performed in a dark room, a double box similar t o that used by Stas being employed to hold the vessel and lamp. Yellow light was passed through the upper portion of the clear liquid in the bottle, and a few drops of the standard solution of silver or potassium bromide, as the case might be, were added to determine which was in excess. Then five or six drops of solution were run in at a time until it was known that a slight excess had again been added. This excess was carefully titrated by means of the other solution, added a drop or so at a time. I n the experiment cited above, the weights of the solutions used before complete precipitation was ensured were : July 23rd.Potassium bromide 1.175 grams = 0.00110 grams silver. Silver . . . . . . , . , . . . . . 0.30 ,, = 0*00035 19 Cwrections. Having now obtained all the experimental data, the various cor- rections for buoyancy and for discrepancies between the actual masses of the weights used and their face values had to be applied. The former were obtained by means of the following densities : Silver cyanide, 3.94 ; potassium bromide, 2.69 ; brass, 8.4 ; air, 0.0012 ; silver, 10.6 ; the latter, by careful comparison of the set of weights among them- selves and reduction to expression in terms of one of them. (The weight thus adopted as unit was the third gram weight,, which is rarely used, and hence suffers little loss from abrasion, &c.) TheDEAN : THE ATOMIC WEIGHT OF NITROGEN.125 balance used was by Bunge of Hamburg, and the weights, a fine set of platinised brass weights, by Sartorius. Silver cyanide as weighed a t first ............... Correction for buoyancy of cyanide + 0.00368 12.11195 grams. I? ?? weights - 0.00175 ?, weights ............... - 0.001 71 Total correction ........................ + 0.00023 ,) Corrected weight of cyanide ..................... 12.1 12 13 ,, -- Potassium bromide as weighed at first.. ....... 10.7629 grams. Correction for buoyancy of bromide + 0-004 8 ?S 9 ) weights - 0.00153 ¶ I weights ............... - 0*00182 Total correction ........................ + 0.00145 Corrected weight of bromide ..................... 10.76435 ,, CcdcuZation of Equivulent of Cpnogen .Having obtained the exact weight of cyanide taken, the amount of silver contained in the cyanide is calculated from the weight of potassium bromide used; by simple proportion, the weight of silver cyanide which would contain the atomic weight of silver is estimated. This is its molecular weight. On subtracting the atomic weight of silver, the equivalent weight of cyanogen remains. The atomic weights used were those found by Stas, namely, Ag= 107.93, 0 = 1 6 . I n addition, it was necessary to know the weight of the potassium bromide employed which would completely precipitate 100 parts by weight of silver ; this was found to be 110.313 parts. The experiments con- ducted in order t o ascertain this fact will be referred to later.Weight of potassium bromide = 10.76435 grams. 100 x 10,76435 Its equivalent of silver = 110.313 = 9*75800 grams. Equivalent in silver of the potassium bromide solution needed for complete precipitation ... = 0.00075 .. Total silver present ............... = 9.75875 ), Weight of silver cyanide = 12.11213 grams. ?? I9 containing 107.93 grams of silver = 12*112*3 107*93 = 133.958 grams. 9.75875 Equivalent of cyanogen = 133.958 - 107.93 = 26.028.126 DEAN : THE ATOMIC WEIGHT OF NITROGEN, Series of Experiments. In the following list are briefly given the leading data and the results of the different determinations. In experiment 6, the method of procedure was varied somewhat, the cyanide being converted in this case into 8zcZpAate instead of nitrate of silver.For this purpose, sulphuric acid, the (' pure for analysis " of commerce, was redistilled in a vacuum! the first runnings being rejected. The cyanide was heated for some hours with the acid, diluted to nearly twice its volume with water, gas being gradually evolved and silver sulphate deposited. More of the concentrated acid was added and the heating repeated in order to bring the sulphate into solution, but although the crystalline salt seemed to disappear completely, there was a slight brownish, flocculent residue, possibly pqracyanogen, still undissolved. The solution was ultimately transferred t o a large volume of redietilled water, heated in a water- bath in order to have the sulphate in solution before adding the potassium bromide, and titrated in the usual way.The slight in- soluble residue still remained on heating the diluted liquid, The value deduced from this experiment is almost identical with that obtained in the one preceding it, which was performed by means of nitric acid. Expt. Wt. of AgCN. 6'2671 grams 17.60585 ,I 17'1049 ,, 179210 ,, 12.11215 ,, 14.6672 ,, 85'67820 grams Equiv. of Ag. 5.0490 grams 14.18496 ,, 13.7801 ,, 14'43881 ,, 11'81727 ,, 9'75875 ,, 69.02889 grams Mol. wt. of AgCN. 133.969 133'956 133.979 133.960 133'958 133,959 133.962 Equiv. of cyanogen. 26.039 26-026 26.049 26 *030 26.028 26'029 26'032 The mean value deduced from these experiments for the equivalent of cyanogen, calculating from the total weight of cyanide used and that of the silver found in it, is thus 26,032. Before we can use this result in determining the atomic weight of nitrogen we must know that of carbon.F. W. Clarke, in his BecaE cdation of the Atomic Veights, gives, as the mean of all the important numbers determined by different experimenters, the value 12-01 1. Since then, however, Scott has shown that certain serious errors have hitherto been neglected, which render some of the experiments con- sidered in that work useless for the time being. The best resultsDEAN : THE ATOMIC WEIGHT OF NITROGEN. 127 which are available, namely, those deduced from the combustion of a known amount of carbon and weight of the carbon dioxide formed, have been recalculated by him, and give a mean value of 12.001 practically (Trans., 1897, 71, 557). Subtracting this, then, from the cyanogen equivalent, we obtain for nitrogen an atomic weight of 1 4 '0 3 1.The accuracy of this number is, of course, strictly dependent upon the value for the atomic weight of carbon, and can only be relied upon to the same extent. The equivalent of cyanogen, however, is quite an independent value, cyanogen in the above experiments having been directly compared with silver, and, as far as method a t least is con- cerned, its equivalent is of the same order of accuracy as those of chlorine, bromine, &c., and mill be equally available for nitrogen or carbon as soon as the other constituent is evaluated. Remccrks on the Titration. On titrating the excess of silver or bromide with the standard solution, the turbid cloud produced in the illuminated layer of liquid became less pronounced as this excess diminished.Moreover,'after the excess had been totally precipitated and more of the solution added, a slight turbidity was produced on standing for a few minutes, which interfered with the accurate determination of the end point. With practice, however, it was possible to distinguish, to a drop or two, the point at which the excess had disappeared, as there was a difference in the appearance of the two turbidities. I n order, however, to check the first determination, the addition of solution was continued in some cases until it was in excess, and the other solution added until the end point was again obtained. In one experiment, the numbers given for the equivalent by successive end point determinations made in this way were (1) 26,0329.(2) 26.0344. The experience derived from many titrations has led to the same conclusion as that of Stas, namely, that titration of excess of silver by means of the bromide solution is more reliable than the converse process, and in most cases I have taken the result of the correspond- ing end point. From the numbers here quoted, however, an idea may be obtained of the magnitude of the maximum error likely to arise from this uncertainty. In some of the earlier experiments, the titration was proceeded with on the same day as the potassium bromide was added. On repeating the addition of the standard solutions on the following day in order toDEAN: THE ATOMIC WEIGHT OF NITROGEN, verify the end point, it was found necessary to add an extra amount of silver before precipitation was again complete.On following days, however, the liquid was quite free from either salt. It would appear that when the silver bromide is thrown down, the flocks enclose small quantities of potassium bromide which diffuse into the bulk of the liquid in the course of a day or so. I n the later researches, provision was made for this, and the titration of the excess postponed for a day at least. Remarks on the Weighings. It was found that the weighings could be made quite easily to a tenth of a milligram provided that the final weighing was not made until half an hour after the tubes and weights had been placed on the pans of the balance. Consecutive weighings, even after an interval of 12 hours, remained then absolutely constant. That the half-hour interval needed was not on account of a certain constant amount of hygroscopic moisture absorbed from the air by the material, is shown by the fact that the weighings taken immediately were always greater than those taken later.Estimation of the Silver Vulue of the Potassizcm Bromide used in the I’its.ations. In order to calculate the weight of silver present in solution from the weight of potassium bromide added, the amount of the latter equivalent t o 100 parts of silver given by Stas, namely, 110.346, was at first made use of. As the results so obtained uniformly gave values for cyanogen which seemed abnormally high, 1 suspected that the bromide used in my research might contain small quantities of the sodium salt. A given weight of the bromide would thus be equivalent t o a greater weight of silver than the pure salt, and the application of a corresponding correction would lead to lower molecular weight for silver cyanide and consequently a lower equivalent for cyanogen.The uncer- tainty arising from this cause was removed by obtaining directly the silver val e of the bromide used. Varying quantities of carefully pre- pared silver were dissolved in nitric acid, purified in the manner described above, and titrated with the bromide in question in the usual way. I n experiment A, the silver used had been obtained by the reduction of a solution of silver and copper nitrates by means of ammonium sulphite, Stas’ directions concerning the method being scrupulously followed. Before the final weighing, the silver was heated in a covered porcelain crucible over a Bunsen burner. In experiment B, a portion of the last fraction of silver prepared by the same method, but obtained on heating the solution, was taken,FORMATION OF a- AND ,8-ACROSE FROM GLYCOLLIC ALDEHYDE. 129 heated to redness for some hours in a current of dry hydrogen (evolved by dropping water upon sodium), and allowed to cool, the gas still passing. This precaution was taken in order that all possible occluded oxygen might be removed from the silver, I n experiment C, the same sample of pure silver was used as in experiment B, but instead of being heated in hydrogen was simply heated in the flame of an alcohol burner in a covered crucible. The bromide used was heated in a current of dry air in a tube provided with a ground cap before being weighed, to insure its thorough freedom from moisture, Wt. of KBr Equivalent of Wt. of KBr per taken. 4. 100 of Ag. A. 9040336 grams 8.52439 grams 110.311 B. 8.63900 ?, 7.83213 ?? 110.316 c. 9.84450 ,, 8.92422 ,, 110,312 Mean ...... 110,313 The author has much pleasure in expressing his indebtedness to Professor Dewar for much kindly criticism and advice, and for the great interest he has taken throughout this investigation.