EVANS THE DETERMINATION OF COBALT 363 The Determination of Cobalt. A new Method of Volumetric Determination and a new Method for its Determination in Steel* BY B. S. EVANS M.B.E. M.C. DSc. F.I.C. GRAVIMETRIC.-The literature of the analytical chemistry of cobalt is ex-tensive complicated and many-sided. Few elements have more or more multifarious methods of determination to their credit; on the other hand few have so many methods which are not quite good enough when one requires great accuracy. The electrolytic method beloved of text-books seems either not to remove all cobalt from solution or to give a contaminated deposit? The phosphate precipitate obtained by Schoeller and Powell's method2 leaves a certain amount of cobalt in solution for which a correction has to be made; precipitation as sulphide appears to be complete but the problem of washing the precipitate is acute; water carries it through the filter ammonium salt solutions dissolve appreciable amounts, alkali salt solutions are satisfactory but fatal to a gravimetric finish.The precipitation with a-nitroso-p-naphthoP5 appears to be complete and the pre-cipitate can be washed; also it has the merit of giving a separation from nickel; the difficulty here lies in the fact that it is (or was) apparently impossible to buy a reagent uncontaminated with iron. The finish of the determination is a further slight difficulty; doubts have been thrown (somewhat unnecessarily I consider) on the accuracy of the conversion to Co,O on ignition.814 On the other hand one may convert the cobalt into CoSO and weigh it as such but here again there is the doubt whether it is possible to eliminate all excess sulphuric acid without at the same time decomposing a trace of the sulphate495.In my view however with careful work both oxide and sulphate methods are capable of greater accuracy than has generally been assigned to them. Reduction to metal by ignition in a current of hydrogen is said to overcome the difficulty in both cases.39 A method has also been published for obtaining the cobalt complex with a-nitroso-#3-naphthol free from the precipitant and weighing it as such.34 Good results both gravimetric and volumetric appear to have been obtained by the 8-hydroxyquinoline So many metals are precipitated by this reagent however that a rigorous separation of the cobalt is first required.SEPARATION.-The separation of cobalt is in itself not an easy matter; it co-precipitates with stannic tin when the latter is thrown down by hydrogen sulphide in hydrochloric acid solution,g and with zinc when the precipitation is made in acetic acid ~olution.~ In view of these facts and of the work published showing that iron and zinc both co-precipitate with copper sulphide in quite strongly acid solution,1° there is at least a suspicion in lack of positive evidence to the con-trary that co-precipitation of cobalt with sulphides is a thing to be guarded against. Cobalt co-precipitates to a serious extent with metals of the third group thrown down with ammoniall. It of course follows nickel in most of its reactions. If cyanide is used at any stage for keeping cobalt from being co-precipitated in some reaction there is danger that some at least of the cobalt will be converted into * Communication from the Research Department Woolwich 364 EVANS THE DETERMINATION OF COBALT cobalticyanide in which form not only is the cobalt difficult to recover but the cobalticyanide itself gives precipitates with many metals in acid solution.One of the most difficult and also one of the most important metals to separate from cobalt is iron and here the valuable a-nitroso-/3-naphthol process as hitherto applied breaks down for the reagent precipitates the iron as well as the cobaltJ2 In consequence of this the usual procedure is to separate the cobalt from the iron group by precipitating the latter with zinc oxide assuming that the cobalt is left entirely in solution.This point has been investigated by Hoffman,ls who came to the conclusion that although the method gives approximate separation a certain amount of cobalt is always precipitated and he accordingly gave a method for eliminating this error but apart from the somewhat irksome restrictions in carrying out the precipitation the method demands re-precipitation which in steel analysis involves the solution and retreatment of the ferric hydroxide from 2 grams of steel. It is noteworthy that Hoffman followed up his investigation into the zinc oxide separation by a second paper a year later giving a method for determining cobalt in steel in which this separation is not used a t all the iron being removed by ether extractionJ4 A useful method for the separation of cobalt by means of precipitation with phenyl-thiohydrazoic acid eliminates many metals including the bulk of the iron, but not apparently nickel or the last few mg.of iron.4 The dinitroso-resorcinol precipitation requires the removal of iron and copperJ6 VOLUMETRIC.-Turning next to volumetric processes for the find determination of the separated cobalt we have the choice of a large number of methods. Of these, some depending on precipitation by 8-hydroxyquinoline etc. appear as mentioned above to be capable of giving good results. The same seems to be true of certain electrometric methods16; these latter however need of course special apparatus and technique. The attention of those in search of volumetric processes for cobalt has been largely centred on the fact that precipitated cobaltic hydroxide can be dissolved in a measured excess of a reducing agent and the excess of the latter titrated back.A number of methods of applying this reaction were examined by Willard and Hall who published many figures obtained by various modifications; many of these figures are excellent but on the whole the variations seem to fall well outside those obtained by the method about to be described. The process has some distinct drawbacks chiefly due to cobaltic hydroxide being a substance which does not dissolve readily in ordinary reducing agents. Willard and Hall's conclusions were that with potassium iodide the reaction was very slow with acid ferrous sulphate an empirical factor had to be used and with stannous or titanous chlorides air must be rigorously excluded.One of the most recent methods dependent on this principle is that of Sarver?' who reduces with ferrous sulphate in the alkaline liquid air of course being excluded. I have tried many experiments with a view to making the cobaltic hydroxide method readily workable but without obtaining consist en t 1 y accurate results. Other volumetric methods proposed are those based on titration of precipi-tated cobaltinitrite,l*J9Jo cobalt pyridine t h i o ~ y a n a t e 2 ~ ~ ~ ~ ~ ~ ~ or cobalt x a l a t e ~ ~ . All these processes like that using hydroxyquinoline involve filtration as a EVANS THE DETERMINATION OF COBALT 365 integral part of the volumetric determination and some need corrections for solubility.In another recent method cobalticyanide is titrated with silver nitrate, potassium chromate being used as an i n d i ~ a t o r . ~ ~ In face of all this wealth of processes it may seem waste of time putting forward yet another; there is however one possible method which has for reasons to be stated later been very much neglected of late years-the cyanide titration. THE CYANIDE TITRATION.-oVer and above the fact that most if not all, reliable methods of volumetric determination of cobalt involve filtI ation and its attendant evils the cyanide method has stood in need of investigation because it is perhaps the best method of determining nickel and hence the analyst has been faced with a troublesome separation of nickel from cobalt when as frequently happens the two occur together.One outstanding paper has been published of recent years-that of Glasstone and Speakman26; the authors investigated the cyanide titration electrometrically and confirmed the theoretical basis of Rupp and Pfenning’s method,27 which assumes a Co (CN) ratio of 1 5 at the end-point. Rupp and Pfenning’s method involves running the solution of nickel and cobalt into the cyanide-a procedure which though quite simple and straightforward when it is essential is one that involves an additional dilution to a known volume, with a certain amount of waste of a possibly limited sample and one instinctively avoids it where possible; in addition the end-point obtained is apparently difficult to observe and conditions as to alkalinity amount of ammonium salts etc.have to be somewhat carefully adjusted. The Rupp and Pfenning titration of course gives the sum of the nickel and cobalt and to obtain a measure of the individual com-ponents Glasstone and Speakman were driven back on the cobaltic hydroxide method which has been criticised above. A full discussion of the theoretical aspect of the question is given in the paper referred to and there is no point in repeating it here. The authors however dismiss their attempt to use the ordinary potassium cyanide-silver iodide titration for cobalt with the statement that they found it impossible to get an end-point. What .actually happens is that when one has titrated the liquid with potassium cyanide until it is clear after a moment’s stand it clouds again; the liquid can again be cleared with cyanide but it again clouds this time more slowly and so on, the recurrence of the clouding becoming more and more slow ; in these circumstances it is obviously impossible to get an end-point.Apparently the lower cyanide of cobalt first formed withdraws cyanide from the soluble silver cyanide complex, itself forming a higher cyanide silver iodide being precipitated. This process in itself slow naturally becomes slower as less and less cobalt remains to be converted. Attempts to speed up the conversion were not encouraging; heating accelerated it considerably but as cyanide is very apt to react with ammonia and the titration has to be conducted in its presence the liquid began to turn brown owing to the decomposition of the cyanide long before the cobalt was all converted.The first step necessary before one could add the excess of cyanide required for the conversion was to find some means of rendering the solution alkaline without using ammonia; sodium hydroxide was not satisfactory as the precipitate formed did not readily dissolve in the cyanide; sodium carbonate was better but by no means perfect; sodium bicarbonate better still. Eventually I found that an excess of bora 366 EVANS THE DETERMINATION OF COBALT gives a pink precipitate which can be approximately titrated away with cyanide without using silver at all. I t is necessary to run the borax soldtion in all at once; if it is added gradually a blue precipitate which is not readily soluble in cyanide, tends to be formed.A number of experiments were next tried to find out the best method of con-verting the cobalt to the higher cyanide; the requirements were these :-(a) com-plete conversion; ( b ) a stable end-product; (c) no loss of cyanide either by destruc-tion or volatilisation. Being at that time under the impression that cobalticyanide was the end-product to be aimed at I attempted the conversion by adding a variety of oxidising agents as well as the excess cyanide; the figures obtained fell roughly into two groups: (i) With mild oxidising agents e.g. air bubbling even at boiling temperature, the consumption of cyanide was considerably less than that required for the Co (CN) ratio. (ii) With more drastic treatment e.g. boiling with hydrogen peroxide the consumption of cyanide was much greater than that calculated and cyanide had obviously been destroyed.On examination I found that the figures obtained with the milder treatments tended to approximate to those required for a Co:(CN) ratio which agrees with Glasstone and Speakman’s finding; but the agreement was by no means perfect; even with the mildest treatment a decided excess of cyanide over that ratio had been used up and a considerable increase in the severity of treatment did not increase this excess by very much. This is illustrated by the following figures: Cobalt taken . . . . 0.0050 g. No. of ml. KCN solution required for Co (CN) . . . . 6.23 ml. ) ) > > J ) > > I 9 , Co:(CN) . . 7.46 ,, 9 , , t , , air-bubbled for 5 minutes cold 6-40 ,, J ) J J I ) > ) 2 ) heated to boiling removed from plate air-bubbled 5 minutes 6.45 ,, A search in the literature of the subject revealed the statement by Manchot and Herzog28 that hydrogen peroxide is set free during the atmospheric oxidation of potassium cobaltocyanide.This gave an explanation of the anomalous results, and I then found that if the potassium iodide which has in any case to be used as an indicator in the final titration is added before the initial titration with cyanide, results agreeing closely with the Co (CN) ratio are obtained. The following process was worked out: PROPOSED VOLUMETRIC METHOD.-SOZ&~O~ZS Required. Potassium cyanide (standard) 16.8 g. of KCN with approx. 10 g. of NaOH Silver nitrate (standard) 5.792 g. of pure AgNO per litre.Borax saturated aqueous solution. Potassium iodide 4 per cent. aqueous solution. Sodium carbonate 10 per cent. aqueous solution. Ammonium chloride 20 per cent. aqueous solution. Dilute ammonia (1 1). made up to 3500ml EVANS THE DETERMINATION OF COBALT 367 The faintly acid solution containing the cobalt to be titrated is placed in a flask and diluted to about 100ml. the flask is swirled round and 20ml. of the borax solution are tipped in from a measuring cylinder; this should result in the formation of a pale pink precipitate in the liquid; 10 ml. of the potassium iodide solution are next added and the mixture is titrated with the standard potassium cyanide solution until the precipitate has dissolved and the liquid is only slightly hazy.The volume of cyanide used V is read and an excess is added bringing the total to (1-3V + 5). The exact volume added does not matter so long as there is an excess which is provided for by the formula. As the titration has to be continued from the same burette after a few minutes and as cyanide burette taps are very apt to leak it is advisable to stand a small beaker under the burette while the process is being continued. To the liquid in the flask now containing an excess of cyanide are next added 10ml. of the sodium carbonate solution and a rapid stream of air is drawn or forced through it for six minutes. This can conveniently be done by means of a wash-bottle attachment with the short arm connected with a filter-pump. It is advisable to adhere fairly closely to the time of aeration stated as with less than five minutes the conversion seems to be incomplete whilst with more than seven or eight, results tend gradually to be slightly high.At the end of the aeration the pump is disconnected the attachment is rinsed in and a mixture of 10 ml. of the dilute ammonia and 25 ml. of the ammonium chloride solution is added. The beaker which has been standing under the cyanide burette is rinsed into the flask and the solution which should be ,quite bright is titrated with standard silver nitrate solution until it is permanently opalescent. Titration is now con-tinued from the cyanide burette drop by drop with vigorous shaking until the liquid again becomes bright. This is the end-point and the burette readings are noted. The burettes having been refilled an amount of the standard cyanide 0.5 ml.less than the previous burette reading is run into the solution. This is again titrated to opalescence with the silver solution and again brought back to brightness with the cyanide; the two solutions are then run in alternately drop by drop until with the original amount of cyanide added the end-point is reached. The new silver reading gives the amount of silver solution required to balance the amount of cyanide used the original one that required for the excess cyanide; the difference between the two is a measure of the amount of cobalt combined with the cyanide; this number multiplied by 0.803 gives the weight of cobalt in milligrams. Experimental results were as follows : TABLE I Cobalt taken g.0*05000 0.02992 0.02500 0*02000 0.01497 0~01000 0-00500 0*00200 Tit ra tion Ell. 80*00- 17*70=62.30 49.80- 12*55=37*25 44.10- 13*00=31* 10 38~00-13~00=25*00 28.45- 9*80=18*65 24055- 12.15=12*40 11-85- 5.75s 6.10 10.70- 8*20= 2.50 Cobalt found g-0.05001 0.02992 0.02498 0.02006 0-01498 0.00995 0*00490 0*0020 368 EVANS THE DETERMINATION OF COBALT Titrations carried out on known mixtures of nickel and cobalt showed that, as might be expected the figures for the two metals are merely additive; these results are given in Table 11. TABLE I1 Cobalt Nickel taken taken Titration g- g. ml. Titration calculated ml. 0.0300 0.0030 50.70 - 10*25=40.45 40.40 0.0240 0*0090 48.90- 10*00=38.90 38.90 0.0180 0.0150 48*90- 11*60=37.30 37.40 0.0120 0.0210 47.80 - 11*90=35.90 35.95 0.0060 0.0270 48.60 - 14-05 = 34.65 34.50 - 0.0330 46.00- 12.95=33*05 33-00 ordinary way 0.0330 36.80- 3*75=33*05 33.00 Direct titration in The last two determinations show that the titration of nickel gives the same result by the process as by the ordinary cyanometric method.The factor for nickel is 0.0010. Originally 10 ml. of 20 per cent. sodium hydroxide solution were used instead of the sodium carbonate added after the initial titration. This gave accurate figures with the higher amounts but the results tended to be low with the lower amounts. On the other hand when no alkali at all was used high results were obtained presumably owing to slight decomposition of the excess cyanide.TABLE I11 Sodium Sodium hydroxide carbonate soh tion solution Cobalt Cobalt taken (20 per cent.) (10 per cent.) recovered g. ml. ml. g-0-0250 10 - 0.0249 0.0250 - 10 0.0250 0~0100 10 - 0*0095 0*0100 10 - 0.0099 0~0100 10 - 0.0094 0.0100 - 10 0*0100 0*0100 no alkali 0-0104 TITRATION OF NICKEL IN PRESENCE OF CoBALT.-In view of the recently assumed industrial importance of electro-deposited nickel-cobalt alloys it seemed worth while to try and find a reliable method for the volumetric determination of one or other constituent independently which combined with the joint titration given above would give a complete analysis of the alloy. Not having had very happy experience of the cobaltic hydroxide titration (vide S Z ~ ~ ~ Y I X ) I turned my attention to the conversion of the cobalt into the very stable cobalticyanide form EVANS THE DETERMINATION OF COBALT 369 followed by recovery of the nickel from its more unstable complex.This principle has been used before by Feigl and Kapulit~as,~~ who however add formaldehyde to remove the excess cyanide thus rendering impossible a subsequent direct cyanide titration. The recovery of nickel from its cyanide complex is not quite so simple as at first sight appears. The complex is readily decomposed by acids, but the immediate result is the formation of an extremely insoluble white pre-cipitate-apparently a nickel nickelocyanide analogous to Prussian blue ; this precipitate resists all further attempts at decomposition. Prolonged boiling of the complex with a large excess of very faintly acid ammonium citrate effects de-composition and the nickel can then be titrated; results however though approximately correct seem to be invariably somewhat low; probably still a trace of the insoluble compound is formed.The following extremely simple procedure was found to give entirely satisfactory results : The solution is made slightly ammoniacal an excess of potassium cyanide (2-3 ml. of the saturated solution should be ample for an ordinary alloy) is added, followed by a few drops of hydrogen peroxide (20 vol.); the solution is then boiled for 5 minutes. The cobalt should now be entirely converted into the inert cobalticyanide while the nickel remains principally as nickelocyanide ; 20 ml. of ammonium chloride solution (20 per cent.) are added and the liquid is boiled for a further 15 minutes.The flask is now removed from the plate and 10ml. of dilute (1 1) ammonia followed by 10 ml. of hydrogen peroxide are added; the flask is replaced on the hot plate and boiled very gently for 10 minutes. This stage of the process is the only one which needs careful attention; it is necessary to dispel all the hydrogen peroxide but if the ammonia is boiled off too far nickel cobalticyanide comes down as a white precipitate. Nickel cobalticyanide is not readily re-dissolved in a small excess of ammonia; if it has been precipitated the titration can still be performed but it takes longer. The flask is now cooled, a few drops of citric acid solution (100 g. to 200 ml. of water) are added followed by 10 ml.of dilute (1 :1) ammonia and 10 ml. of potassium iodide solution (4 per cent.) and the solution is diluted to approximately 250-300 ml. with water. The titration is carried out as follows:-Standard potassium cyanide is run in until the blue colour of the solution is practically dispelled then 1 to 2 ml. of the standard silver solution are added from the burette. This probably produces a turbidity, and in that case more cyanide is run in until the turbidity is dispelled and the solution is bright. If nickel cobalticyanide has been precipitated it will still be visible at this stage and if so a few ml. excess of potassium cyanide solution must be added and the flask allowed to stand with occasional shaking until all solid particles are dissolved. In either case the turbidity is brought back by the addition of a little more silver nitrate and then just dispelled again by the addition of potassium cyanide added drop by drop.The silver nitrate value of the total potassium cyanide added is found in exactly the same way as in the cobalt titration by adding nearly the full amount to the titrated liquid; titrating this away with silver nitrate and then cautiously titrating alternately with the two solutions until the balance occurs when the full amount of potassium cyanide has been added. The difference in ml. between the two volumes of standard silver nitrate solution thus obtained is equal to the weight in milligrams of nickel present 370 EVANS THE DETERMINATION OF COBALT The following experimental results were obtained by this method : TABLE IV Taken 7 Nickel Titration found g .ml . g -P- Cobalt g. 0.0300 0.0030 35.60 -5.60=30.00 0.03000 0.0270 0*0060 36.00 - 9.00=27*00 0*02700 0.0240 0.0090 31.65 - 7.65=24*00 0.02400 0.0210 0.0120 24.40 - 3*50=20.90 0-02090 0.0 180 0.0 150 21.35 - 3.40 = 17.95 0.0 1795 0.0150 0.0180 17.70 -2*80= 14.90 0.01490 0-0120 0.0210 13.70 - 1.70 = 12-00 0~01200 0*0090 0.0240 12*50-3*45= 9.05 0.00905 0*0060 0.0270 8.25-2*10= 6.15 0.0061 5 0.0030 0*0300 4.50 - 1*50= 3.00 0.00300 The complete process therefore for the determination of nickel and cobalt in a mixture of these two metals gives us two titration figures expressed in ml. of the silver nitrate solution (i) combined nickel + cobalt titration = x ml. (ii) nickel titration = y ml.then: The method was tried on five electro-deposited alloys of nickel and cobalt supposedly free from other metals. Cobalt = ( x - y ) 0.803 mg. Nickel = y mg. Results were as follows: TABLE V Weight taken Combined f& each titration g . 0-0320 0.0320 0.0320 0.0320 0.0320 titration (ml. AgNO,) 34-05 34-45 35.45 36.15 38.75 Nickel titration (ml. AgNO,) 22.90 20.75 17.20 14.00 4.80 Cobalt found g. 0.00895 0-01 100 0.0 1466 0.01778 0.02725 Nickel found €5 0.02290 0.02075 0.01720 0.01400 0.00480 Composition per cent. CO 27.98 Ni 71.58 99.56 CO 34-38 Ni 64.85 99.23 CO 45.80 Ni 53-73 99.53 co 55-59 Ni 43.76 99.35 CO 85-12 Ni 14.98 100~11 -= -= EVANS THE DETERMINATION OF COBALT 371 It must be clearly borne in mind that before the cobalt titration can be carried out other metals especially iron copper and zinc must be eliminated.far the most widely used method for the separation of cobalt from most metals appears to be its precipitation by a-nitroso-IS-naphthol. It is especially un-fortunate that iron is also precipitated by this reagent in view of the fact that this metal almost always accompanies cobalt that many industrial alloys of cobalt also contain notable amounts of iron ( e g . cobalt steels) and that really reliable and simple methods of eliminating iron which do not involve filtration of large amounts of ferric hydroxide are conspicuously lacking (vide supra). The most usual way of removing the iron preliminary to determining cobalt in steel analysis is by zinc oxide separation.The objections to this method have been stated earlier (vide supra). Both the 8-hydroxyquinoline and the pyridine thiocyanate methods precipitate iron as well as cobalt. Many methods have been published which claim to effect the ~ e p a r a t i o n ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ but all or almost all of these methods, granted that they are all that is claimed for them suffer from the serious dis-advantage (from the point of view of steel analysis) that it is the iron which is precipitated whilst the cobalt remains in solution. The claim made for the process which follows is that the cobalt is precipitated substantially free from most metals including iron without any preliminary separation what ever.The usual solvent used for a-nitroso-IS-naphthol is acetic acid and this solution as has been said precipitates iron as well as cobalt. The reagent is, however soluble in a number of other liquids and in syrupy phosphoric acid it gives a solution which while still precipitating cobalt copper and ferrous iron,= no longer precipitates ferric iron. The ordinary medium in which precipitation is effected is dilute hydrochloric acid This may however be replaced by dilute nitric acid without ill effect so long as the nitric acid is free from oxides of nitrogen. This fact enables alloys to be readily brought into solution with the iron in the ferric condition without the introduction of possibly harmful reagents. One difficulty in the a-nitroso-p-naphthol precipitation has always been that the reagent itself is precipitated as well as the cobalt complex.This makes it difficult to know when an adequate excess of the precipitant has been added; the technique adopted in the process to be described overcomes to some extent this difficulty. A double precipitation has been adopted; this is principally due to the fact that the nitroso-naphthol precipitate is very bulky and difficult to wash and as a matter of fact has always been the correct procedure where large amounts of metal ( e g . nickel) in solution have to be eliminated. Copper still precipitates with the cobalt and as it reacts with cyanide must be removed; this is achieved by a hydrogen sulphide precipitation before the second nitroso-naphthol precipitation ; as pointed out earlier there is the possibility that cobalt may tend to co-precipitate with copper sulphide.The amount of copper in steels however is generally so small that the co-precipitation may be ignored; with an alloy containing copper it will probably be desirable to re-precipitate the sulphide. Attempts were originally made to determine the cobalt gravimetrically. These were unsuccessful as the results were invariably somewhat high; apparently it is difficult to remove all SEPARATION OF COBALT FROM IRON ETC. ; DETERMINATION I N STEEL.-B 372 EVANS THE DETERMINATION OF COBALT phosphoric acid from the precipitate but in view of the success of the volumetric process this question has lost its significance. The process finally worked out for the determination of cobalt in steel is as follows: DETERMINATION OF COBALT IN STEEL.-The sample (5g.for low amounts, less for higher) is dissolved in dilute nitric acid (spgr. 1.2) in the proportion of 75 ml. per 5 g.; the solution is boiled for 10 minutes and cooled and about 0.5 g. of urea is added. The reagent solution which should not be more than 2 or 3 days old is prepared by placing the wnitroso-j3-naphthol in a small beaker adding syrupy phosphoric acid in the proportion of 50 ml. to 3.5 g. of nitroso-naphthol, placing the mixture on the hot plate and stirring it with a thermometer until the temperature reaches 60" C. and finally cooling. The solution of the sample is diluted to about 100 ml. and the reagent solution is added 1 or 2 ml. at a time, with subsequent stirring a few seconds being allowed to elapse between additions during which stirring is continued.The following phenomena are observed The liquid darkens then a red turbidity which increases as the additions continue is produced and after two or three additions a stable froth is formed and carries up the red precipitate. At a certain point with continued addition and stirring the precipitate suddenly begins to darken the froth breaks or shows a tendency to break and a black resin-like scum floats to the surface. This marks the presence of an excess of reagent but it is advisable to add a final quantity of 3 to 4 ml. The liquid is next diluted with about its own volume of 20 per cent. acetic acid and stirred and about 3 ml. more of the reagent are run in (omission of this last precaution seems for some reason to lead to a tendency to slightly low results).The liquid is allowed to stand (cold) for about 25 minutes (not less than 20) with occasional stirring and the precipitate is filtered off on paper and well washed with cold dilute nitric acid (5 per cent. prepared by diluting 20 per cent. nitric acid which has been boiled for 15 minutes). . The precipitate so obtained is dark coloured probably nearly black and very bulky and may contain a good deal of water. The filter and its contents are transferred to a roomy platinum dish and placed on the plate to dry. (This is usually advisable as quite a considerable amount of liquid may separate on heating.) When sufficiently dry the dish is placed in the muffle and the filter is burnt off at not too high a temperature.If vanadium molybdenum or tungsten (vide inf~a) is present the residue is fused with fusion mixture taken up with hot water and filtered the precipitate is washed with hot water and the filter is replaced in the platinum dish and again burnt off; in the absence of these elements this step may be omitted. When completely burnt off the residue is fused with potassium bisulphate heating being continued until potassium sulphate begins to crystallise as a scum on top of the melt. If an appreciable amount of cobalt is present the melt will be dark blue when hot and pink when cold; in absence of copper it may be dissolved in about 100 ml. of 5 per cent. nitric acid the solution boiled and cooled and the second precipitation carried out forthwith but with a steel it is safer to proceed as follows: The melt is dissolved in water and the solution is transferred to a beaker, hydrogen sulphide is passed the liquid is heated to boiling and the precipitate is allowed to settle then filtered off and washed with hot water.The filtrate i EVANS THE DETERMINATION OF COBALT 373 boiled until hydrogen sulphide has been dispelled 25 ml. of the boiled-out 20 per cent. nitric acid are added and the liquid is boiled down to a volume of about 100 ml. After cooling the second precipitation is effected in exactly the same way as the first but no urea is added. The resulting precipitate should be crimson lake in colour. It is desirable to test the filtrates of both this and the first precipitation for excess of reagent; this may be done by removing about 50ml.of the filtrate (before washing) dividing it into equal parts and adding about 0.2 mg. of cobalt to one of them. After standing for some time the formation of the cobalt precipitate in one of them generally causes a distinct difference in appearance between the two parts. Too much stress must not be laid on this test; with the filtrate from the second pre-cipitation it is almost invariably and immediately successful but in the first precipitation the excess of the reagent seems itself to be mainly precipitated or destroyed. After washing the filter and precipitate are transferred this time to a porcelain crucible. Platinum must not be used as the subsequent solution of the cobaltic oxide in hydrochloric acid liberates chlorine; silica does not seem so satisfactory as porcelain.The filter will probably need drying again before ignition; it is then burnt off as before care being taken to see that the burning off is complete and that no carbon particles remain. After ignition the residue should be black; it may have white or blue patches presumably due to phosphoric acid which has escaped removal. When the residue is cool about 10 ml. of strong hydrochloric acid are poured on part of it being used to rinse down the sides. The crucible is then covered with a watch-glass and stood on a steam-bath. When the acid is hot a fairly vigorous reaction ensues which should result in the residue dissolving completely to a blue solution. When this is achieved the cover-glass is removed and rinsed in with hot water and the acid liquid is evaporated to dryness over steam.The sides of the crucible are rinsed down with about 15 drops of strong nitric acid, the solution is again evaporated to dryness and the residue of cobalt nitrate (which should be dark red) is taken up with 2 drops of nitric acid and about 10 ml. of hot water; this should give a pink solution which is probably slightly hazy. The solution is transferred to a beaker and cooled; it may contain a very small trace of iron which must be removed not because it is sufficient to influence the titration, but because it may cause a haze in the liquid and make the end-point difficult to observe. The removal is effected by the addition to ihe cold liquid of a suspension of barium carbonate in water added a few drops at a time until a trace of undissolved barium carbonate is floating in the solution as a semi-flocculent precipitate.* The precipitate along with any trace of iron that was present is filtered off and the precipitate is washed with cold water.The filtrate is now ready for the addition of borax and the remainder of the volumetric procedure is carried out as already described. Only one slight modification has to be introduced in this It is allowed to settle filtered off and washed as before. * Great care must be taken to avoid any excess of barium carbonate otherwise a little cobalt may be lost by adsorption (vide infra) 374 EVANS THE DETERMINATION OF COBALT procedure because of the soluble barium salts produced by the barium carbonate treatment.When sodium carbonate is added after the initial titration this of course precipitates the barium; therefore after aeration and the subsequent addition of ammonia and ammonium chloride the liquid is filtered and the pre-cipitate is washed with cold water. The final titration is then proceeded with as usual. The trace of iron requiring removal by barium carbonate is so small that no appreciable amount of cobalt is carried down with it. TABLE VI Trials of this method were carried through with the following results: Per cent. of cobalt Iron taken g. 5.00 5.00 5.00 5.00 5.00 5.00 5.00 Nickel Copper taken taken €!* g-0.050 0.050 0.050 0.050 0.050 0.050 0.050 0.010 Cobalt taken g. Blank 0*0020 0.0050 0~0100 0.0250 0.0500 0~0100 Titration ml.of AgNO soh. 3.40- 3.25= 0.15 7.65- 5.10= 2.55 16.35- 10*00= 6.35 24.90- 12.70=12*20 38.85- 7*55=31.30 77.80 - 15.90 = 61 -90 31-20 - 19.10 = 12.10 Cobalt found g. 0~00012" 0.00205 0.00510 0*00980 0.02513 0-04972 0.00972 f 1 found (corr. for added blank) - -0-0400 0-0386 0.1000 0.0996 0.2000 0.1960 0*5000 0.5002 1*0000 0.9920 0.2000 0.1920 * That this blank was really due to cobalt was shown by the colour of the hydrochloric acid extract of the residue. The method was tried without modification on steel "W" of Messrs. Ridsdale's The percentage composition of this steel as given in C 0.695; Si 0.187; S 0.075; P 0.028; Mn 0.101; Cr 3-01; V 0.791; As I was then unaware of the precipitation of vanadium etc.by the reagent, the first precipitate was fused direct with bisulphate the fusion with fusion mixture being omitted. The very considerable amount of tungsten (16.21 per cent.) was ignored throughout and was finally filtered off with the barium carbonate precipitate just before the titration was begun. British Chemical Standards. the certificate is: W 16.21; Co,'4.76; Ni 0.44; As 0.01; Cu 0.058; Mo 0.048. The results obtained were as follows: TABLE VII Weight taken Titration €5 ml. Cobalt Per Cent. 0.5000 47.20-18*70=28*50 4.58 0*5000 47.30- 19*00=28*30 4.55 0.5000 43.75- 15*30=28.45 4.57 0*5000 43.45 - 15*30=28*15 4.52 The cobalt figure 4.76 given in the certificate is the mean of a large number of determinations by 19 independent referees using a great variety of methods; the individual figures given vary from 4.52 to 4.99.As I made no attempt to work u EVANS THE DETERMINATION OF COBALT 375 the barium carbonate precipitate (vide infra) it is very probable that my figures are slightly low. Interference by Chromium Vanadium etc.-After completing the work so far described I came across the statement in Hillebrand and Lundell’s text-book:’ that bismuth silver chromium zirconium titanium vanadium tin and nitric acid* also “interfere” (with the gravimetric determination of cobalt by precipitation with a-nitroso-/3-naphthol). Time did not permit of an exhaustive examination of this question but in view of the use of chromium vanadium and titanium in d o y steels it was necessary to ascertain the effect of these metals at least on the process.As is proved by the figures given in this paper nitric acid is entirely harmless if certain simple precautions are taken; further experiments showed that : ( a ) Chromic salts do not seem to be precipitated by the reagent to any extent; a slight precipitation takes place but this is presumably due to adsorption, and possibly also to the fact that ( b ) Chromic acid appears to be precipitated. (c) Vanadic acid is precipitated (vanadyl salts were not tried). ( d ) Vanadic acid does not interfere with the cyanide titration. In the process as applied to steel the chromium is left as chromic salt; the vanadium on the other hand is largely converted into vanadic acid; consequently, in the first precipitate one has ( a ) All the cobalt; ( b ) all the copper; ( c ) at least a great part probably the whole of the vanadium molybdenum tungsten etc.; ( d ) a little chromium and iron. Of these the cobalt and copper have already been dealt with; the trace of chromium behaves in much the same way as a trace of iron and is eliminated therewith; such of the vanadium etc. as gets through into the titration liquid exerts no influence and in any case vanadium tungsten and molybdenum are removed by the fusion with fusion mixture. An initial experiment (made before I realised the extent to which vanadium is precipitated) was carried out with a solution containing 1.452 g. of chromium and 0.700 g. of vanadium. This was divided into two equal portions and 0.010 g.of cobalt was added to one of them. After adjustment of the nitric acid strength the two solutions were precipitated with the phosphoric acid-nitroso-naphthol reagent in the usual way; the precipitates obtained were burnt off in porcelain, and the residues were taken up with hydrochloric acid thus proceeding direct to the latter part of the process without a second precipitation. A great deal of vanadic acid did not go into solution when the crucible was rinsed out and was consequently filtered off with the barium carbonate precipitate. The titration behaved normally but the result was somewhat low-0.0090 g. of cobalt in place of the 0.01OOg. taken. The barium carbonate precipitate on being worked up (vide i n f y a ) furnished a further 0.0005 g.which had been entangled in the un-dissolved vanadic acid giving a total recovery of 0.0095 g.; the blank required 0-30 ml. which quantity was deducted from the titration figure. The lost 0.0005 g. * Lundell Hoffman and BrightS8 say “Iron copper molybdenum chromium zirconium, titanium vanadium tin and tungsten are also precipitated. Nitric acid interferes. 376 EVANS THE DETERMINATION OF COBALT was probably not precipitated owing to an insufficient quantity of reagent having been used for so large an amount of vanadium. The second experiment was carried out with smaller quantities of vanadium and chromium but with addition of titanium; each solution contained 0.025 g. V, 0-037 g. Ti and 0-145 g. Cr and 0.010 g. of cobalt was added to one of them. The process used was as for steel but the first precipitate after being burnt off was fused with fusion mixture.The melt was taken up with water and the precipitate was filtered off and washed with hot water. The precipitate was again burnt off and then fused with bisulphate the resulting melt being treated like the ordinary bisulphate melt of the first precipitate (reprecipitation etc.). Copper being absent hydrogen sulphide was not used. The titration of the cobalt sample required 11-65 ml. and that of the blank 0.15 ml.; the cobalt corresponding to the difference of these figures is 0.00924 g. There had therefore been a loss of 0-00076 g. , or 0.95 ml. The barium carbonate precipitate was then worked up (vide ircfra), and the titration resulting was 0.70 ml. bringing the total cobalt recovered up to 0.00980 g.It would seem therefore that the trace of chromium (or vanadium?) left after the second precipitation tends to drag down a fraction of a milligram of cobalt in the barium carbonate precipitate. It is clear that none of the three metals tested in any way prevents the pre-cipitation of the cobalt and the only form of interference shown was the removal of a trace in the barium carbonate precipitate. For ordinary purposes this trace can probably be neglected; for accurate work it can be recovered as follows:-The barium carbonate precipitate is burnt off in a platinum dish and fused with potassium bisulphate the melt is taken up with 5 per cent. nitric acid boiled and filtered the filtrate is then treated with the nitroso-naphthol reagent in the same way as the solution of the bisulphate melt of the first precipitate in the ordinary process and the determination is finished as usual.In the above trial about a milligram of blackish residue was obtained on burning off this precipitate and this dissolved in hydrochloric acid to give a blue solution; the hydrochloric acid solution of the “blank” residue was not blue and the 0.15 ml. titration figure was probably an experimental error. The vanadium extracted in the alkali fusion of the first precipitates was titrated; that from the cobalt sample contained 0.0135 g. and that from the blank 0-0158 g. out of a total in each case of 0.0250 g. The loss seems to have occurred during burning off. I hope to investigate the vanadium question later with a view to a possible method for determining vanadium in steel.Lack of time prevented further work being done on the separation of cobalt from chromium vanadium titanium, molybdenum etc. but as indicated above there seems no reason to fear that precipitation will be incomplete or that alkali fusion will fail to remove the vanadium molybdenum etc. It might be mentioned in conclusion that precipitation of cobalt in the usual way with an acetic solution of the nitroso-naphthol and with phosphoric acid added to the cobalt solution does not have the desired effect iron being precipitated to a considerable extent ; increase in the amount of phosphoric acid with the object of holding the iron up completely seems to prevent the complete precipitation of the cobalt.Several experiments carried out with the aim of preventing th EVANS THE DETERMINATION OF COBALT 377 reagent precipitating while allowing of the complete precipitation of the cobalt complex by means of the addition of excess of acetic acid acetone etc. appeared to indicate that under these conditions the solubility of the cobalt complex is by no means negligible. The cobalt solution used for these experiments was made up from a specially prepared sample of very pure electrolytic cobalt; the iron used was electrolytic iron of a high degree of purity; the nickel was a very pure sample of Mond nickel, which previous analysis had shown to contain 0.02 per cent. of cobalt. All reagents used were of analytical reagent quality except the a-nitroso-g-naphthol, which was old stock.SUMMARY.-(&) A new volumetric method of determining cobalt has been described. This method closely resembles the ordinary nickel titration. ( b ) A method has been described for titrating nickel in the presence of cobalt. (c) A new method of separation of cobalt from all metals interfering with (a) A preliminary survey of other metals precipitated by the reagent has the volumetric procedure has been detailed. been made and their non-interference with the method has been demonstrated. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. REFERENCES D. H. Brophy Ind. Eng. Chem. L4nal.Ed. 1931 3 363. W. R. Schoeller and A. R. Powell J . Iron and Steel Inst. 1918,97 444; ANALYST 1919,44, Recent Advances in Analytical Chemistry Vol. 11 pp. 266-7. H. H. Willard and D. Hall J . Amer. Chem. SOC. 1922 44 2219 2226 2237. A. A. Guntz and J. Barbier Chem. et Ind. 1929 21 711. R. Berg Z. anal. Chem. 1929 76 191. H. R. Fleck and A. M. Ward ANALYST 1933 58 388. M. Auger and L. Odinot Compt. rend. 1924 178 710. W. Funk 2. anal. Chem. 1907 46 93. W. Bottger and K. Druschke Annalen 1927 453 315. E. Swift and R. Barton J . Amer. Chem. Soc. 1932 54 2220. M. Ilinski and G. von Knorre Z. angew. Chem. 1893 264. J. I. Hoffman Bureau of Standards J . of Research 1931 7 883. W. R. Orndorff and M. L. Nichols J . Amer. Chewz. SOC. 1923 45 1439. E. Muller and H. Lauterbach Z. anal. Chem. 1923 62 23. L. A. Sarver Ind. Eng. Chem. Anal. Ed. 1933 5 275. G. Barbieri Atti. R. Accad. Lincei 1928 [v] 8 405. A. A. Wassilief Z. anal. Chem. 1929 78 439. H. Yagoda and H. M. Portridge J . Awzer. Chem. SOC. 1930 52 4887. G. Spacu and M. Kuras Bul. SOC. Stunte Cluj 1934,7 377; BY. Chem Abst. 1934 1323. P. Spacu Compt. rend. 1935 200 1595. J. T. Dobbins and J. P. Saunders Ind. Eng. Chem. Anal. Ed. 1934 6 459. J. Ledru and L. Hauss Bull. SOC. Chim. Belg. 1932 41 104. R. Uzel*and B. Jezek Coll. Czech. Comm. 1935 7 497; Br. Chew Abst. 1936 444. S. Glasstone and J. C. Speakman ANALYST 1930 55 93. E. Rupp and F. Pfenning Chew-Ztg. 1910 321, W. Manchot and J. Herzog Ber. 1900 33 1742. F. Feigl and H. J. Kapulitzas,,Z. anal Chem. 1930 82 417. F. Zeitsche and M. Nachmann Helv. Chim. Acta 1926 9 420 705 979. S. Krishna and H. Singh J . Amer. Chem. SOC. 1928,50 792. P. Ray and A. Chattopadhya 2. anorg. Chem. 1928 169 99. G. Font& and L. Thivolle Bull. SOC. Chim. 1924 35 641. C. Mayr and F. Feigl 2. anal. Ckem. 1932 90 15. M. Ilinski and G. v. Knorre Ber. 1885 18 699. L. Dufty J . Iron and Steel Inst. 1914 11 p. 52. Hillebrand and Lundell Applied Inorganic Analysis (1929) p. 326. Lundell Hoffman and Bright Chemical Analysis of Iron and Steel (1931) p. 337. 279. - Id. 1932 8 659. - Id. pp. 337-338