Solvent Extraction of Cobalt SX

Solvent Extraction of Cobalt SX

The United States produces essentially none of the cobalt and less than 10 pct of the primary nickel it consumes each year. However, it has been estimated that U.S. resources of cobalt and nickel are 842,000 and over 15 million short tons, respectively. Present yearly domestic consumption of cobalt is about 9,000 short tons, while that of nickel is about 200,000 short tons. Cobalt usually is recovered as a byproduct of either nickel or copper. A significant amount of the domestic nickel resources are nickel silicate and laterite deposits located in southern Oregon and northern California which contain from 0.5 to 1.2 pct Ni and 0.06 to 0.25 pct Co. Generally, commercial processes cannot efficiently and economically recover nickel and cobalt from low-grade domestic laterites.

The Bureau of Mines, U.S. Department of the Interior, as part of its goal to insure an adequate supply of minerals and metals to meet national economic and strategic needs, is developing a method to recover Ni, Co, and Cu metals from low-grade domestic laterites. Research concerning this method was begun in 1971 and has progressed through many stages of development. The present method, shown in figure 1, encompasses many separate systems and unit operations. This report describes the extraction of cobalt and the preparation of a suitable cobalt electrolyte using solvent extraction from liquor produced by this method.

The laterite ore is subjected to selective reduction at about 525° c with carbon monoxide. This reduces the nickel and iron to ferronickel, and the cobalt and copper to the metallic state. The reduced material is next leached with a 100 g/l ammonium hydroxide (NH4OH), 300 g/l (NH4)2SO4 solution in a controlled oxidation leach. This solubilizes the Ni, Co, and Cu as ammine complexes. In general, the nickel ammine formed is Ni (NH3)x(II) where x ranges from 2 to 6. A similar complex occurs for cobalt and copper except that cobalt is primarily cobalt (III).

The leach liquor then is stripped of 80 to 90 pct of its free ammonia. Stripped leach liquor has a pH of about 9.5. This improves the subsequent nickel extration. Nickel and copper are coextracted with General Mills LIX64N. They then are separated and concentrated by selective stripping. It is known that cobalt (III) will not be extracted by LIX64N during nickel and copper extraction, but remains in the raffinate.

The raffinate is recycled to the leach stages, thus cobalt will build in concentration unless removed. Eventually a point would be reached where the cobalt concentration would reduce the leaching power of the solution. In commercial processes (Nicaro, Cuba; Greenvale, Australia; and Marinduque’s Philippine operation) cobalt is removed by sulfide precipitation. This method often produces a precipitate with poor handling properties, introduces sulfide ions into the leach solution, and yields an impure sulfide precipitate that must be further refined.


Separating cobalt from sulfate solution by solvent extraction has been the subject of much work. These studies have dealt with solutions having pH ranges of either 5 to 6.5 or 11 to 12. The work at high alkalinity had shown that not greater than 40 g/l (NH4)2SO4 could be tolerated. Although a satisfactory solvent extraction method for cobalt that was amenable to this process was not known, this type of technology offers obvious advantages compared with sulfide precipitation. Some general processing advantages are: (1) leach reagent regeneration, (2) continuous processing, (3) usually low capital and operating costs, and (4) pure metal recovery when coupled with electrowinning. Thus, experimentation centered on applying solvent extraction techniques to the recovery of cobalt from the nickel-copper raffinate described above.

An experimental liquid ion exchange reagent developed by General Mills was evaluated for the recovery of cobalt from the ammoniacal nickel-copper raffinate stream. This proprietary compound is presently called XI-51. Although XI-51 is not an oxime, it operates on a similar hydrogen ion cycle, forming metal-organic chelates.

Experimental Techniques

The experimental methods used for the laboratory evaluation of XI-51 for the recovery of cobalt began with a series of tests to determine the maximum loading capacity of cobalt on the solvent. Equilibrium isotherms for extraction, washing, and stripping were then constructed. These are used to predict the number of stages necessary to obtain a desired efficiency or recovery. The kinetics of the various mass transfer operations and phase disengagement rates were also investigated. Similar experimental methods were applied in investigating the removal of zinc from the pregnant cobalt electrolyte.

The maximum loading capacity of the metal ions on the solvents were obtained by a “repeating contact” technique. Generally, equal volumes of fresh aqueous solution and solvent were contacted by vigorous shaking in a separatory funnel to equilibrium on a wrist-action shaker. Some loading tests were conducted at organic to aqueous volume ratios (O/A) of 0.5. Ten minutes of shaking was adequate to insure equilibrium with O/A ratios of near 1, with 15 min allowed for O/A ratios of 0.5. After phase separation, the aqueous phase was removed and replaced with a fresh aliquot of aqueous solution. The process was repeated until analyses of raffinate samples showed that extraction had ceased. In some situations, the monitoring of raffinate pH was accurate enough to determine when extraction had ceased. The analysis of the solvent for the appropriate metal after extraction had ceased yielded the value of the maximum loading capacity under the particular conditions used for that determination.

Equilibrium isotherms for extraction, washing, and stripping were determined by contacting the appropriate aqueous and solvent starting solutions to equilibrium at a range of O/A ratios (10:1, 5:1, 2:1, 1:1, 1:2, 1:5, and 1:10). At each ratio fresh portions of the feed solutions were used. The contacting was again achieved by vigorous shaking in a separatory funnel on a wrist-action shaker. Adequate time was allowed to insure equilibrium; 10 min for the intermediate ratios, and 30 min for the more extreme ratios. Most of the equilibrium isotherms for cobalt and zinc were made at room temperature (20° C); however, the cobalt extraction and stripping isotherms were obtained by heating the starting solutions to 40° C before contact. The temperature of the feed aqueous streams to the cobalt extraction and stripping stages under continuous flow conditions in a complete process is expected to average about 40° C.

For the extraction and stripping isotherms, both phases were analyzed for the appropriate metal after the phases were equilibrated. In the washing isotherm determination, the aqueous phase was analyzed for NH4OH, and the organic phase, after contact with the wash solution, was stripped with a 10-g/l H2SO4 solution (O/A = 1). This strip solution was subsequently analyzed for (NH4)2SO4. The resultant equilibrium values of ammonia were then calculated for the organic and aqueous phases. The validity of the equilibrium curves of all of the isotherms was checked by calculating material balances at random points.

Tests were made to define the kinetics of: (1) cobalt (II) extraction with XI-51, (2) NH3 removal from the loaded solvent by H2O-(NH4)2SO4 washing, and (3) the stripping of cobalt from the loaded solvent. “Kinetics” in this context refers to the rate of mass transfer under specified conditions and is expressed as the percentage approach to equilibrium in a given time.

The kinetic tests were begun by agitating a measured quantity of solvent in a 180-ml mixer compartment of a mixer-settler stage. The impeller speed (1,700 rev/min, 556 ft/min, or 169 m/min tip speed), solution compositions and quantities, and temperature (20° C) were all fixed. A measured quantity of the aqueous phase was then added rapidly with the simultaneous starting of a stop watch. This testing was run at an O/A ratio of 1. After the chosen time interval, the agitation was stopped, and as soon as sufficient clear phase was generated (about 25 ml), samples of the appropriate phase were withdrawn for analysis. The stop watch was stopped when sampling was begun, therefore, the reported times include the time required for some phase separation. The process was repeated, starting with fresh solutions, with agitation for different time intervals. The equilibrium value of the constituent of interest was established by agitation for 5 min.

A static method was used for preliminary evaluations of the effects of the following variables on phase separation: (1) solvent composition, (2) (NH4)2SO4 concentration in the wash solutions, (3) free NH3 concentration in the nickel-copper raffinate, (4) the extent of cobalt loading on the solvent, and (5) O/A ratios. In the static method, a measured amount of the appropriate fresh aqueous soltuion and solvent were contacted by vigorous shaking in a separatory funnel on a wrist-action shaker. Ten minutes of shaking was adequate to insure that equilibrium was reached and that a uniform dispersion was generated. After shaking was completed, the contents of the separatory funnel were immediately transferred to a graduated cylinder and a stop watch started. The mixture of the two phases was then allowed to separate quietly in the graduated cylinder. The distance of the top and bottom edges of the dispersion band were measured as they coalesced to one interface line. The dispersion band thickness was plotted as a function of time.

The phase disengagement characteristics of the extraction stage were studied further with a dynamic system. The test setup consisted of a mixer-settler stage equipped with a movable baffle in the settler. This stage had a 600-ml mixer compartment and a settler area that was variable between 2.5 and 10.6 sq in (16.1 and 68.4 sq cm). A turbine agitator having a diameter of 1.5 in (3.8 cm) was used in the mixer compartment. The depth of the settler was 14 in (35.6 cm). The organic and aqueous phases flowed from the settler to separate 2 liter reservoirs. Pumps continuously returned the two phases to the mixer compartment of the stage.

In the dynamic system, a continuous flow of both organic and aqueous phases to the mixer was fixed. The settler baffle was moved progressively in such manner as to decrease the settler area. The dispersion band thickness was measured at each setting and was plotted as a function of the specific flow rate. Generally, the tests were continued until the settler flooded (dispersion filled the settler).

Experimental Results

Solvent Composition

solvent-extraction-cobalt-effect-of-o-aIn preliminary tests of the XI-51 extractant, slow phase separation was frequently experienced and often stable emulsions were formed. In cobalt loading tests on a solvent composed of 10 vol-pct XI-51, 10 vol-pct isodecanol; and 80 vol-pct Kermac 470B, precipitation was experienced in the organic phase. Cobalt material balances indicated that the precipitate was at least partly a cobalt-organic compound. In subsequent testing with the 10-10-80 solvent, no precipitation was experienced as long as the solvent was not loaded to near its maximum capacity; however, poor phase separation still existed. Testing was then directed toward finding a solvent composition that would yield satisfactory phase separation characteristics.

solvent-extraction-cobalt-effect-of-isodecanol-concentrationThe effect of the O/A ratio on phase separation was investigated first. In this testing a synthetic liquor containing about 0.8 g/l Co, 25 g/l NH4OH, and 300 g/l (NH4)2SO4 was contacted with the 10-10-80 solvent on a wrist-action shaker. Three O/A ratios were tested: 0.5, 1.0, and 2.0. Temperature, agitation
intensity and time, solution compositions, and technique were fixed so that meaningful comparisons could be made. It was found that both speed and completeness of phase separation were favored by the higher O/A ratios, figure 2. The aqueous entrainment in the organic phase was determined by withdrawing organic samples after phase separation appeared complete and centrifuging them. The entrainment is reported in parts per million (ppm), and is based on the respective volumes. The organic entrainment in the aqueous phase was not measured. This appeared to be low, based upon the clarity of the aqueous phase.

The effect of isodecanol and XI-51 concentration in the solvent upon phase separation was determined in a similar manner. The optimum isodecanol concentration appears to be about 15 pct when using 10 vol-pct XI-51 solvent with an O/A ratio of 2 (fig. 3); 20 pct show no improvement in the rate of phase separation and slightly increased aqueous entrainment.

The rate of phase disengagement is apparently an inverse function of the concentration of XI-51 in the solvent, figure 4. Preliminary data indicated a high loading capacity of cobalt on XI-51 from synthetic solutions (a 3-vol- pct XI-51 solvent loaded up to about 1.9 g/l at pH 9.5). The low concentration of cobalt normally encountered in laterite produced liquors (0.10 to 0.30 g/l) allowed selection of the XI-51 concentration to be based upon optimum phase separation characteristics rather than extraction requirements. Therefore, subsequent testing was carried out with a solvent containing 3 vol- pct XI-51, 15 vol-pct isodecanol, and 82 vol-pct Kermac 470B.


solvent-extraction-cobalt-effect-of-concentrationThe raffinate from the nickel-copper solvent extraction system typically contained from 0.10 to 0.30 g/l cobalt (III); although one raffinate used contained about 0.50 g/l. The maximum loading capacity of the solvent: for cobalt (III) from typical raffinates was only about 0.06 g/l. Previous testing with fresh synthetic solutions (pH -9.5) showed that the solvent had a maximum loading capacity for cobalt (II) of up to about 1.9 g/l. At this level the solubility of the cobalt-organic complex was not exceeded in the 3-vol-pct XI-51 solvent so there was no precipitation in the organic phase. Methods were then investigated to reduce cobalt (III) to cobalt (II) before solvent extraction.

Treatment of nickel-copper raffinate by pumping through a column of cobalt metal shot reduced cobalt (III) to cobalt (II) according to the reaction,

2 Co (III) + Co → 3 Co (II)

In the cobalt reduction column, nickel-copper raffinate is pumped up through a bed of cobalt shot contained in a sealed column. The nickel-copper raffinate overflows at the top of the bed into a discharge tube. Air must be excluded from the reduction column or the cobalt shot would be leached. Also, some cobalt (II) would be oxidized to cobalt (III). Agitation can be supplied in the cobalt shot bed by introducing argon gas at the bottom of the column. The average residence time within the bed is determined by measuring both the void volume of the bed and the flow rate of the feed nickel-copper raffinate.

The effectiveness of the cobalt reduction column is illustrated in figure 5. In this testing the feed rate of the nickel-copper raffinate to the reduction column was varied, resulting in a range of residence times. Samples of the reduction column effluent were then contacted with fresh solvent at an O/A ratio of 1. Curve B represents the extraction performance relative to essentially plug flow conditions in the reduction column. In curve A, argon gas was continuously bubbled through a 1-in ID (25.4-mm) column at a moderate rate (about 0.2 cu ft/hr, 6 l/hr), supplying mild agitation and facilitating mass transfer. Clearly, high extraction of cobalt is possible when adequate contact is provided in the reduction column.

The cobalt shot used in the reduction column acquires a black coating after being exposed to typical nickel-copper raffinates for a long period of time. A reduction column that was used intermittently over a 9-month period and was kept full of nickel-copper raffinate during shutdown periods, showed no significant decrease in efficiency even though the cobalt shot acquired a very black coating. This coating is apparently noncrystalline, containing about 38 pct Co and 40 pct O. Much of the coating can be removed by wet tumbling of the shot in a mill. A dilute sulfuric acid solution (-11 g/l) is more effective than just water in the tumbling treatment; however, some cobalt shot is then leached.

The relative ease of cobalt (II) extraction is illustrated by the extraction isotherm, figure 6. In this testing the nickel-copper raffinate passed through the reduction column at a flow rate giving a residence time of 25 min. Agitation was not used in the reduction column during this testing. Argon gas was used in the separatory funnel during this testing to eliminate the oxidation of cobalt (II) to cobalt (III) during the shake-out contacts. The “S” shaped characteristic probably results from incomplete reduction of cobalt (III) to cobalt (II) in the reduction column. As previously noted, cobalt (III) is not strongly extracted by this solvent. One extraction stage (fig. 6) would produce a raffinate containing about 0.12 g/l Co, provided the loading of the solvent did not exceed about 85 pct; otherwise, two stages would be necessary. Increased agitation in the reduction column would reduce the cobalt concentration of the raffinate still further. As shown later in this report (table 7), treatment of nickel-copper raffinate in the agitated reduction column followed by contact in one extraction stage resulted in a raffinate containing only about 0.02 g/l Co.

solvent-extraction-cobalt-isothermThe loading of cobalt (II) on the solvent is strongly pH dependent, figure 7. Synthetic solutions containing about 0.5 g/l Co, 300 g/l (NH4)2SO4, and the pH being adjusted with either NH4OH or H2SO4, were used in this testing. Argon gas was used in the separatory funnel during this testing with aqueous solutions having a pH of 7.5 or greater. Cobalt (II) starts loading on the solvent at a PH above about 4.5, with the maximum loading taking place between about pH 8.0 and 9.5. Increased free NH3 concentration above pH 9.5 probably provides an increased driving force to form amine complexes, thus inhibiting cobalt (II) extraction. The maximum cobalt loading capacity on the solvent is about 1.4 g/l from laterite produced nickel-copper raffinate at pH 9.5. This determination was with a nickel-copper raffinate that had a residence time of 14 min in the reduction column prior to loading.

solvent-extraction-cobalt-loading-versus-phBesides cobalt, other metal ions can be extracted by the solvent. Of particular interest are nickel and zinc. Nickel is almost completely extracted by LIX64N prior to cobalt solvent extraction. Typical nickel-copper raffinate contains only about 0.01 g/l Ni and <0.001 g/l Cu. Zinc, on the other hand, is crowded (displaced) from LIX64N by the maximum loading of nickel and copper. Therefore, zinc will remain in the nickel-copper raffinate.

Methods for controlling impurities such as nickel and zinc in the cobalt electrolyte must be considered. Data from solvent-extraction-cobalt-zinc-extraction-isotherm continuous circuit tests indicate that some nickel is extracted from the nickel-copper raffinate. This is indicated by trace amounts of nickel being detected in the loaded solvent and in the pregnant cobalt electrolyte. If nickel contamination becomes a problem in the cobalt electrolyte, a small bleed stream should control it. Two methods were considered for controlling the impurity level of zinc in the cobalt electrolyte: (1) crowding of zinc from the XI-51 solvent by the maximum loading of cobalt, or (2) solvent extraction of zinc from the pregnant cobalt electrolyte with D2EHPA.

Research has centered on the latter approach. In this method, zinc is coextracted with cobalt by the XI-51 solvent. Zinc is then extracted from the resultant cobalt electrolyte with D2EHPA and ultimately recovered as crystallized zinc sulfate.

An extraction isotherm, figure 8, was developed for zinc extraction in the presence of cobalt (II). Nickel-copper raffinate used in this determination was first passed through the reduction column with a residence time of 8 min and operating under agitated conditions. The maximum loading capacity of zinc on the solvent appears to be about 0.1 g/l. The numbers on the curve correspond to the cobalt loading on the solvent. The dashed line represents a portion of the curve where the exact shape is unknown. The solvent will load both zinc and cobalt (II), but it is preferential to cobalt (II). If sufficient cobalt (II) is available in solution to load the solvent to near its maximum, then zinc will not be extracted to any appreciable extent. This is illustrated in figure 8 by the point corresponding to a cobalt loading of 1.3 g/l on the solvent, and also in table 1. If coextraction of zinc and cobalt is required, the solvent should not be fully loaded with cobalt (II).


The solvent also loads NH3; the 3-vol-pct XI-51 solvent has loaded up to 1.1 g/l NH3. Ammonia loading was quite variable, being affected by the free NH3 content of the liquor and by the cobalt loading on the solvent. Ammonia loading on the solvent increased with increasing NH3 concentration in the liquor. At least two mechanisms are apparently responsible in loading NH3 on the solvent: (1) NH3 dissolves in Kermac 470B, and (2) ammonium ions are extracted by XI-51. Data show that to some extent cobalt (II) loading will crowd NH3, Zn, and Ni from the solvent, table 1. In this test, nickel-copper raffinate was passed through the reduction column with a residence time of 21 min. Two samples of the column effluent were then contacted with solvents at O/A ratios of 0.25 and 1.0. This resulted in one solvent being essentially fully loaded with cobalt (II), while the other was only partially loaded. The loading of NH3 on the solvents was determined by stripping a portion of the loaded solvents with a 10 g/l H2SO4 solution (O/A of 1) and analyzing the strip solutions for (NH4)2SO4. The loading of NH3 was then calculated.

Water-Ammonium Sulfate Washing

Loading of NH3 on the solvent necessitates the washing of the solvent before stripping. Otherwise, excessive amounts of (NH4)2SO4 would be generated in the electrolyte causing possible precipitation of the double salt, CoSO4·(NH4)2SO4·6H2O. The NH3 could be removed with a dilute acid wash, but this would generate (NH4)2SO4 with a resultant NH3 loss.

Washing the loaded solvent with a H2O-(NH4)2SO4 solution effectively removed most of the NH3, figure 9. A 120 g/l (NH4)2SO4 optimized both NH3 removal and phase separation. Two countercurrent wash stages typically removed 70 to 80 pct of the total NH3 from the solvent. The washed loaded solvent typically contained between 0.15 and 0.20 g/l NH3. The stages plotted in figure 9 represent data obtained from a continuous circuit test (table 7). Ammonia could then be removed from the wash solution by heating and air sparging. This NH3 can then be combined with that from the liquor stripping operation and then recovered by absorption into the cobalt raffinate that is recycled to leaching.

The “J” shaped washing curve, figure 9, indicates that removal of NH3 and ammonium ion from the solvent is incomplete without a final neutralization wash. Ammonium ion loaded on XI-51 probably causes this characteristic. The H2O-(NH4)2SO4 wash system has several limitations. To recover a substantial amount of NH3 from the solvent without using an excessive number of stages requires: (1) the feed wash solution must have a very low NH3 concentration, and (2) the increase in NH3 concentration of the wash solution will be relatively small. The NH3-loaded wash solution is stripped of NH3 and recycled to the wash stages. Batch tests have shown that the NH3 concentration of the loaded wash solution, typically about 0.6 g/l, can be reduced by heating (50° C) and moderate aeration to less than 0.1 g/l in 6 hr.

pH-Controlled Wash

solvent-extraction-cobalt-washing-curveA pH-controlled wash stage removes the residual NH3 from the loaded solvent. Care must be exercised in this operation to control the pH so that adequate NH3 removal is achieved without appreciable stripping of cobalt. The pH-controlled wash stage of the continuous circuit was operated at successively lower pH levels. At each pH level, time was allowed for the system to equilibrate before samples were taken. A pH probe located in the settler compartment of the stage was connected to a pH-controller. The pH-controller operated a pump that injected dilute H2SO4 (0.1 M) into the mixer compartment where the loaded solvent and the recycled pH-controlled wash solution was being contacted.

Data indicate the optimum pH level to be between 5.5 and 6.0. In this range the NH3 loading on the loaded solvent was reduced to about 0.013 g/l while the cobalt stripped from the solvent averaged about 5 pct. A bleed stream from the pH-controlled wash stage can be used to control the (NH4)2SO4 level to about 300 g/l. This stream can then be used directly for leach solution makeup; thus the small amount of cobalt that is stripped in this stage will not represent a loss.


In selecting the strip solution concentrations of cobalt and H2SO4, three requirements were considered: (1) efficient cobalt stripping, (2) zinc extraction from the cobalt electrolyte, and (3) efficient cobalt electrowinning. A series of tests was run to determine the acid level necessary to obtain adequate stripping of cobalt, table 2. In these tests, portions of loaded solvent containing about 0.85 g/l Co were equilibrated with electrolyte samples containing about 75 g/l Co and variable amounts of H2SO4. An O/A ratio of 5.9 was used so that complete stripping of the solvent would result in a cobalt buildup of 5 g/l in the strip solution. The data indicate that adequate stripping of cobalt, <0.1 g/l Co on the stripped solvent, should be obtained at an equilibrium acid concentration of >1 g/l, in one stage.


The effect of the cobalt concentration in the strip solution upon the stripping of cobalt from the solvent was also investigated. Strip solutions containing about 12 g/l H2SO4 and cobalt ranging from zero to about 75 g/l were equilibrated with loaded solvent containing about 0.84 g/l Co at an O/A ratio of 5.9. The results from this testing, table 3, indicate that in the pH range of this test, the concentration of cobalt in the strip solution has a negligible effect upon the stripping of cobalt from the solvent.


Two cobalt stripping isotherms were constructed to show the possible variation caused by the use of different strength acid strip solutions, figure 10. The curves were generated by loading cobalt onto the XI-51 solvent (-1.4 g/l Co), washing the resulting loaded solvent three times (O/A = 1) with (NH4)2SO4 wash solutions to remove NH3, and then equilibrating portions of the solvent with strip solutions containing about 75 g/l Co and either 9.3 or 18.4 g/l H2SO4.

Assuming a buildup of 5 g/l Co (from about 75 g/l to 80 g/l)in the strip solution during actual circuit operation, the 9.3 g/l H2SO4 strip solution would be reduced to about 1.0 g/l in the pregnant strip solution (pregnant cobalt electrolyte). As will be shown later in this report, this is a desirable acid level from the standpoint of zinc extraction from the pregnant cobalt electrolyte.

solvent-extraction-cobalt-stripping-isothermIn some situations only insignificant amounts of zinc may be transferred to the cobalt electrolyte. This may be the case where: (1) the liquor being processed contains very little zinc, or (2) a zinc crowd stage is used in the cobalt solvent extraction circuit. In either case, it may be possible to eliminate the zinc extraction with D2EHPA operation from the cobalt electrolyte. This would allow a higher acid level to be used in the strip solution. As before, assuming a buildup of 5 g/l Co in the strip solution, the 18.4 g/l H2SO4 solution would be reduced to about 10 g/l in the pregnant cobalt electrolyte. Some expected advantages of a higher acid level are (1) more complete one-stage cobalt stripping, (2) lower voltage required in cobalt electrowinning, and (3)  possibly less cobalt oxide formation on the anodes during cobalt electrowinning.

The cobalt stripping isotherms, figure 10, are very similar. The curves indicate that cobalt is easily stripped from the solvent with either a 9.3 or 18.4 g/l strip solution. Only one stage should be necessary for good stripping of cobalt even though a strip solution containing a high cobalt concentration is used. The expected stripped solvent will contain about 0.03 g/l Co. The stage plotted on this figure represents data obtained from a continuous circuit test (table 7).

During extraction of cobalt (II) with the XI-51 solvent, zinc can be coextracted and thus transferred to the cobalt electrolyte. The zinc impurity must be held to a relatively low level in the cobalt electrolyte, possible upper limits being 5 to 100 ppm depending on other parameters. Zinc reportedly can cause cracking and peeling of the cobalt deposit. As mentioned previously, research was concentrated on controlling the zinc level by extraction of zinc with D2EHPA from the pregnant cobalt electrolyte. The location of the zinc recovery circuit relative to the cobalt solvent extraction system is shown in figure 11.

The maximum loading capacity of zinc on D2EHPA was then determined at various pH levels, table 4. Portions of solvent containing 20 vol-pct D2EHPA, 5 vol-pct tributyl phosphate (TBP), and 75 vol-pct Kermac 470B were repeatedly contacted with cobalt electrolytes to determine the maximum loading capacity.

The cobalt electrolytes contained about 80 g/l Co, 0.5 g/l Zn, and between 0.4 and 9.0 g/l H2SO4. The data indicate that zinc will load even from fairly acidic solutions on D2EHPA; also, some coextraction of cobalt is likely.


For preliminary parameter evaluation a zinc extraction isotherm was developed, figure 12. Cobalt stripping and zinc loading tests indicate that acceptable operation of these functions are expected when a pregnant cobalt electrolyte is produced with an acid level of between 1 and 3 g/l. Therefore, a pregnant cobalt electrolyte containing about 80 g/l Co, 0.6 g/l Zn, and 1.0 g/ H2SO4 with a pH of 2.1 was used along with the D2EHPA solvent for


this determination. The zinc extraction isotherm exhibits two important features: (1) the gradual
slope indicates that a number of stages will probably be required for an acceptable extraction (not yet established), and (2) the minimum zinc level in the raffinate from this electrolyte appears to be about 0.02 g/l (-20 ppm). The dashed line is an extension of the curve. The exact shape of the upper portion of the curve is unknown.

A method that is being evaluated for recovering zinc, consists of stripping it from the loaded D2EHPA solvent to produce a solution containing about 100 g/l Zn. The strip solution would then be partially evaporated to crystallize zinc sulfate. A test series was run to determine the H2SO4 concentration that would be necessary to efficiently strip zinc from the loaded D2EHPA solvent into the 100-g/l Zn solution, table 5. Portions of partially loaded D2EHPA solvent were equilibrated at an O/A ratio of 1 with zinc strip solutions containing from 5 to 100 g/l H2SO4.


The data in table 5 indicate that cobalt strips easily from the loaded D2EHPA solvent with a solution devoid of cobalt and containing about 5 g/l H2SO4. Zinc, on the other hand, loads from the strip solution onto the solvent until an acid level of about 40 g/l is reached. A strip solution containing 100 g/l Zn and 100 g/l H2SO4 would strip about 90 pct of the zinc from the solvent in one stage. It appears that the small amount of cobalt that is coextracted with zinc may be removed from the solvent either by crowding with zinc or by selective stripping with dilute H2SO4.

Cobalt electrowinning experiments indicate that good deposits can be produced under a variety of conditions. Some of the variables explored were: (1) temperature (50° to 85° C), (2) cobalt concentrations (55 to 80 g/l), and (3) H2SO4 concentration (9.75 to 30 g/l). The apparent flexibility of cobalt electrowinning allows selection of cobalt stripping conditions to be based primarily upon cobalt stripping and zinc extraction requirements.


A series of tests were made to define the kinetics of: (1) cobalt (II) extraction with XI-51 solvent, (2) NH3 removal from the loaded solvent by H2O-(NH4)2SO4 washing, and (3) the stripping of cobalt from the loaded solvent. These tests were conducted at 20° C in a 180-ml mixer compartment of a mixer-settler stage. The only variable investigated in this testing was the residence time in the mixer.

The extraction of cobalt (II) by the XI-51 solvent was very rapid. The minimum time tested was 45 sec. This residence time yielded a raffinate containing about 0.03 g/l Co and matched the equilibrium level that was established by agitation for 5 min. Nickel-copper raffinate used for this testing was treated in the reduction column, residence time of 12 min under agitated conditions, and contained about 0.38 g/l Co.

Ammonia removal from the loaded solvent was also accomplished rapidly. The minimum time used in this series was 1.25 min (15 sec of agitation followed by 1 min required to generate sufficient phases for sampling). The wash solution after contact contained 0.8 g/l NH4OH and matched the equilibrium value that was established by agitation for 5 min. The loaded solvent used in this series contained 1.3 g/l Co and 0.64 g/l NH3. The initial wash solution was NH3 free and contained 120 g/l (NH4)2SO4.

Stripping of cobalt from the loaded solvent proved to be slower than the other two mass transfer operations, table 6. The data indicate that essentially 100 pct of the equilibrium value was reached in about 2 min. A loaded solvent containing 1.3 g/l Co and a strip solution containing 75 g/l Co and 18.3 g/l H2SO4 were used in this series.


Phase Separation

Preliminary testing was done to define the phase separation characteristics of the XI-51 solvent. A plot of dispersion band thickness versus total flow rate of dispersion per unit settler area, figure 13, was developed to simulate the phase separation of the extraction stage. The dispersion band represents the area where incomplete phase separation exists. It is bound above and below by the separated phases. The dispersion band thickness when related to flow rate and settler area, can be used to size the settler. In this testing, solvent partially loaded with cobalt was contacted with typical raffinate from the continuous circuit in the previously described phase disengagement stage. Tests were run to evaluate the phase separation characteristics under various O/A ratios.

As found in the static phase separation tests, phase separation in the extraction stage was best when run at the higher O/A ratios. With the conditions given in figure 13, a dispersion band of about 4 in would be expected for a specific flow rate of 1.5 gal/min/sq ft (61 l/min/sq m) at O/A ratio of 1.2. It has been noticed that phase separation is also affected by the extent of cobalt loading on the solvent, NH3 concentration in the liquor, temperature of the solutions, and extractant batch. Therefore, a range of phase separation rates is possible depending upon the existing conditions.

Continuous Circuit Operation

solvent-extraction-cobalt-phase-disengagementAfter establishing the basic parameters for the cobalt solvent extraction system, the actual behavior of the system was determined by using a small mixer-settler continuous circuit, figure 11. The circuit consisted of a cobalt reduction column, one extraction stage, two countercurrent H2O-(NH4)2SO4 wash stages, one pH- controlled wash stage, and one strip stage. The zinc recovery system was not included in the continuous cobalt circuit in these tests.

The stages of the cobalt system had a mixer volume of 180 ml and a settler area of about 10 sq in (64 sq cm). Agitation and interstage pumping was accomplished by the impellers in the mixer compartments of the stages. Both the resultant mixer residence time and the settler area were oversized in all stages except the extraction stage; this was unavoidable because all of the mixer-settler units were the same size. The cobalt reduction column consisted of a glass tube having a 3-in ID (75 mm), packed with about 4.5 kg Co metal shot, and fitted with removable ends. The cobalt shot was random in both size and shape; ranging from 1/8 to ¾ in in diameter (3 to 19 mm). The measured void volume within the bed of cobalt shot was about 525 ml. Argon gas was continuously bubbled through the bed at about 2 cu ft/hour (57 l/hr).

Results from the continuous cobalt solvent extraction circuit are shown in table 7. Treatment of the nickel-copper raffinate in the agitated cobalt reduction column, 11 min residence time, followed by contact in one stage with the recycled stripped solvent resulted in about 94 pct Co extraction. The raffinate contained only 0.02 g/l Co, while the loaded solvent contained 0.85 g/l Co, and 0.77 g/l NH3.


A two-stage H2O-(NH4)2SO4 wash circuit using recycled wash solution recovered about 73 pct of the NH3 from the loaded solvent. Ammonia was stripped from the wash solution prior to recycling. After a final pH-controlled wash (pH-5.5), the NH3 concentration of the loaded solvent was reduced to about
0. 013 g/l. This resulted in about 98 pct overall NH3 removal from the loaded solvent. At pH 5.5 only about 3.5 pct of the cobalt was stripped in this wash stage.

The loaded solvent was then stripped in one stage with a solution containing 77.3 g/l Co and 18.1 g/l H2SO4. This testing represented the case when a relatively concentrated H2SO4 strip solution was used. This yielded a pregnant cobalt electrolyte (pregnant strip solution) containing about 81.3 g/l Co and a stripped solvent containing 0.05 g/l Co. The corresponding NH3 buildup in the pregnant cobalt electrolyte from residual ammonia ion loaded on the solvent was about 0.066 g/l. The balance of the (NH4)2SO4 generation in the strip solution was through aqueous entrainment (-260 ppm) in the organic phase from the pH-controlled wash stage.

The presence of zinc (-20 ppm) in the pregnant cobalt electrolyte indicated that only a very small amount of the zinc from the nickel-copper raffinate was extracted under these conditions. The cobalt raffinate is recycled to the leaching stages, therefore, the zinc level will build in the leach liquor. The extraction of zinc by the XI-51 solvent should become more pronounced as the concentration of zinc builds in the leaching circuit.


Solvent extraction using a 3-vol-pct XI-51 solvent was used to extract 94 pct of the cobalt in one stage from nickel-copper raffinate containing 0.24 g/l Co and 0.02 g/l Zn after treatment in a cobalt shot column. About 98 pct of the total NH3 loaded on the solvent was removed in two washing steps: (1) a two-stage H2O-(NH4)2SO4 wash, and (2) a pH-controlled wash stage. The H2O-(NH4)2SO4 wash recovered and recycled about 73 pct of the NH3 without neutralization. Cobalt electrolyte was prepared by stripping the loaded solvent in one stage with spent electrolyte containing about 75 g/l Co and 18 g/l H2SO4. Electrolytes containing less H2SO4 may also be used.

Preliminary data show that coextracted zinc can be removed from the pregnant cobalt electrolyte with a 20-vol-pct D2EHPA solvent. Cobalt stripping and zinc loading tests indicate that acceptable operation of these functions can be expected when a pregnant cobalt electrolyte of about 1 to 3 g/l H2SO4 is produced. Stripping zinc from the D2EHPA solvent into a solution containing 100 g/l Zn requires about 40 g/l or greater H2SO4.

The Bureau of Mines is developing a method to recover Ni, Co, and Cu from laterites containing less than 1.2 pct Ni and 0.25 pct Co. The method consists of the following basic unit operations: (1) reduction roasting, (2) leaching, (3) solvent extraction, and (4) electrowinning. The method reflects three Bureau of Mines objectives: (1) recovery of critical minerals that are domestically in short supply from low-grade domestic laterites, (2) lower processing energy requirements, and (3) solution recycling. This report deals with the extraction of cobalt and the preparation of a suitable cobalt electrolyte by solvent extraction from liquor produced by this method.

Nickel and copper are coextracted with LIX64N from an ammoniacal ammonium sulfate leach liquor containing about 1.00 g/l Ni, 0.30 g/l Co, 0.03 g/l Cu, and 0.02 g/l Zn. Cobalt (III) in the nickel-copper barren raffinate is reduced to cobalt (II) with cobalt metal. Reduction of cobalt (III) to cobalt (II) greatly aids subsequent extraction. Commercially available XI-51 extracts about 94 pct of the cobalt from the treated raffinate in one stage in a laboratory mixer-settler continuous circuit. Ammonia loaded on the solvent is removed in two washing steps. About 94 pct of the cobalt then is stripped from the XI-51 in one stage with spent cobalt electrolyte containing about 77 g/l Co and 18 g/l sulfuric acid (H2SO4). Electrolytes containing less H2SO4 also may be used. Preliminary data indicate that coextracted zinc may be removed from pregnant cobalt eletrolyte containing 3 g/l or less H2SO4 with di-(2 ethylhexyl) phosphoric acid (D2EHPA).