Table of Contents
Cobalt Flotation Studies were conducted on ores from the Viburnum trend in southeast Missouri. The Viburnum trend deposits occur along the flank of the St. Francois Mountains. Ores of the Viburnum trend are characteristically complex mixtures of the sulfide minerals galena, PbS; sphalerite, ZnS; chalcopyrite, CuFeS2; pyrite, FeS2; marcasite, FeS2; and the cobalt-nickel mineral siegenite, (Ni,Co)3S4, hosted in a gangue of dolomite, (Ca,MgXC03)2, with minor inclusions of quartz, SiO3, and calcite, CaCO3. Minor amounts of other sulfides are also found; these include: bravoite, (Fe,Ni)S2; millerite, NiS; bornite, CusFeS4; chalcocite, Cu2S; digenite, Cu2-xS; djurleite, Cu1.96S; polydymite, Ni3S4; vaesite, NiS2; gersdorffite, NiAsS; tennantite, (Cu, Fe)12As4S13; arsenopyrite, FeAsS; pyrrhotite, Fe1-xS; and covellite, CuS.
The major cobalt source in the ore is the mineral siegenite. Separation and concentration of siegenite by froth flotation is complicated by the mineralogy of the ore. The siegenite is often present as an intimate replacement intergrowth with chalcopyrite (fig. 1), sphalerite, galena (fig. 2), and the iron sulfides, pyrite and marcasite. Current grinding practice in the mills tends to leave the siegenite as binary locked particles in the various product streams (fig. 1). Therefore, further grinding to liberate the siegenite particles is required, but is limited by the lower size limitations of the froth flotation and by grinding energy considerations.
During beneficiation, the cobalt is distributed to the various mill product streams as shown in figure 3. The run-of-mine ore contains 0.015 wt pct Co. About 30 pct of this cobalt reports to the copper concentrate at an average grade of 0.50 wt pct Co. This represents an improvement ratio of 33. Because of the significant built-in improvement ratio, the copper concentrate was chosen for initial studies, and is the subject of this report.
Previous Bureau of Mines research to recover cobalt from the copper concentrates has included both hydrometallurgical and Cobalt Flotation studies. Hydrometallurgical schemes have utilized a ferric chloride solvent in an oxidative leach to treat the copper concentrate and leave the cobalt in a cobalt-enriched residue. Although successful, this scheme has the disadvantage of disrupting the production of a salable copper product and would involve a major capital investment in a leach and electrowinning plant to recover the copper and the lead from the leach liquor.
In laboratory flotation research, the first attempt was to float the cobalt-bearing siegenite from the chalcopyrite. This proved impossible except for some Halimond tube tests on simple two-mineral systems. Research, therefore, was directed at floating the chalcopyrite from the siegenite.
The authors would like to thank the management and staff of Cominco American, Inc., and Dresser Industries, Inc., Magmont operation, and especially the mill and maintenance personnel for their assistance in the set up and operation of research equipment at the mill.
Material Equipment and Procedures
The original experimental flotation test equipment used at the Cominco American Magmont concentrator consisted of grinding and flotation sections. The grinding section consisted of a 30- by 30-in ball mill inclosed circuit with a 3-in hydrocyclone. Feeding was accomplished by a periodic bleed from a continuously circulating stream, from an agitated holding tank. An air-actuated, timed valve accomplished the actual bleed. This feed system gave satisfactory control of parameters. Since the stream was continuously circulating, the feed density was relatively uniform and feed rate could be varied by adjusting the timed cycle on the valve. Two feed points were used. Initially, feed was introduced directly to the ball mill, but the feed point was later changed to the cyclone feed sump. The hydrocyclone was fed from a divided sump by a Vacseal pump. The pump served as a receptacle for both the ball mill discharge and the hydrocyclone overflow and, as mentioned, the feed. The Vacseal pump was charged from a bottom outlet from this cone-shaped sump and the rougher flotation circuit was fed from an overflow near the top of the sump. Provisions were made to feed directly into the No. 1 cell of the flotation section or into an attritioner and then into the No. 1 cell.
The flotation section included six Hazen-Quinn, subaeration cells in the series. These were No. 5 units with an effective volume of 0.25 ft³ per cell. Later, seven Fagegren type cells were used as roughers. These had an effective volume of 0.20 ft³ each.
By using three sets of launders on the Hazen-Quinn cells and by altering the flow patterns between the cells, it was possible to vary the flotation configurations.
Figure 4 shows the three of the variations used. In practice, the first pattern was used for roughing and the second for cleaning. The third was tried but was most difficult to control. When in the rougher configuration (pattern 1), concentrates were held in a tank awaiting a change to the second configuration for cleaning. Final tailings from each test were filtered in a pan filter and saved for subsequent testing. All pumping within the unit, except for the cyclone feed, was done by air-driven Warren Rupp Sandpiper pumps equipped with tranquilizer units to smooth the flow.
Description of Feed Material
Chalcopyrite concentrate was used as feed stock. A representative long-term chemical analysis of the material is shown in table 1.
Size analyses with assays for each size fraction and distribution of metal values are shown in table 2. It is important to realize there is a high degree of variability in the ore fed into the mill. This can vary day to day or even hour to hour.
The continuous flotation testing involved four phases: (1) initial short-duration tests to fix reagent dosages and determine grinding parameters, (2) extended roughing tests, (3) extended roughing tests with successive cleaning stages, and (4) continuous demonstration tests to obtain engineering data. The reagent suite and dosages that were used in prior, large-scale tests were used as a starting point for the continuous flotation tests. These reagents and their respective dosages were sodium aerofloat, 0.50 lb/st; Dowfroth 250 frother, 0.10 lb/st; and sodium cyanide, 0.10 lb/st. All but the frother were added to the ball mill; the frother was added down the cell bank as needed. The grind was targeted at a minimum of 50 wt pct passing 10µm. Prior tests indicated that the siegenite should be sufficiently liberated at this point to give satisfactory rougher recovery.
During the short-duration tests, it was discovered that some of the chalcopyrite was of such a refractory nature that additional collection would be needed. Both Minerac “B” and Z-200 proved effective in floating this more refractory fraction of chalcopyrite. The Z-200 was chosen because it was available on site.
The flotation flowsheet developed from these short-duration tests is compared in figure 5 to the classic rougher-scavenger-cleaner configuration. For lack of a better term, this flowsheet was dubbed “successive cleaning.” Table 3 shows the results of laboratory cleaning of rougher concentrates produced from a short-duration test. Two successive flotation tests were done in the laboratory at bench scale on a rougher concentrate, and the data from each tailing assay were combined into a hypothetical cobalt product. In each test, recoveries approached or exceeded 80 pct and the grades exceeded 3.5 wt pct Co on a starting grade of 0.5 wt pct Co. This flotation flowsheet appeared to allow feed grades lower than the previously determined benchmark of 0.8 wt pct Co.
Extended Rougher Tests
Extended roughing tests were run to optimize the initial flotation step in order to maximize cobalt grade. These were continuous tests, run around the clock, usually of 3 days to 1 week in duration. Tables 4 and 5 list some typical conditions and data from one of these series of extended tests. Results were evaluated in 8-h periods numbered from one to nine. The objective was to remove 30 to 50 wt pct of the cobalt at as high a grade as possible and leave the successive steps to make a high recovery. As can be seen from table 5, tests 2, 4, 5, and 9 met these standards.
One very interesting result from these tests was a correlation between the nonchalcopyrite iron (marcasite and pyrite) fraction of the feed and the lead grade of the cobalt product. Earlier work had indicated that there was a relationship between the amount of marcasite-pyrite present in the ore and the extent of galena oxidation. It was also noted that the marcasite was the particularly unstable component. This marcasite-pyrite content seems to fix the galena depression even stronger and makes any significant separation of galena from siegenite very difficult. This leads to an excess of lead in the cobalt product, which both dilutes the cobalt grade and makes further processing to recover the cobalt difficult. As can be seen from table 5, there is no direct correlation between the lead grade of the feed and the lead grade of the cobalt product.
Figure 6 plots lead grade in the cobalt product versus nonchalcopyrite iron in the feed for all continuous tests made with the test unit except for those few made on highly contaminated feedstock (greater than 10 pct Pb in the feed). From these data, two trend lines can be drawn, one with a steeper slope than the other. It is thought that the steeper trend represents a higher marcasite-to-pyrite ratio within the nonchalcopyrite iron fraction.
Extended Tests With Cleaning
To run the extended roughing tests with cleaning, some modifications to the continuous test unit were necessary.
A large agitation tank was used to hold rougher concentrate, until it was later refloated using a launder configuration that divided the six cells into two sets of three. The tailings from these two steps were added together with the original rougher tailings to form a cobalt product. Tables
6 and 7 list the parameters and summarize the results of a typical test.
Figure 7 plots cobalt improvement ratios versus mean particle size for the successive cleaning tests. The improvement ratios show an upward trend with diminishing particle size in the rougher stage but the line is virtually flat for the cleaning stages. This could indicate either a lack of liberation of the siegenite or entrainment of the siegenite in the froth.
Scanning electron microscope (SEM) studies tended to confirm the idea that the liberation of the siegenite from the chalcopyrite was incomplete. These data are summarized in table 8.
The first series of these studies was done in an attempt to determine the degree of liberation of the siegenite from the chalcopyrite. The samples of flotation test unit products
were searched for nickel by using X-ray mapping. When nickel was found the area was then remapped for cobalt and copper. The occurrence of cobalt and nickel without the copper would indicate a free siegenite particle. Indications were that the siegenite particles were not liberated. Some possible reasons given are that the lack of friability of the particles at this size range makes breakage difficult, or that the lack of distinct phase boundaries and grain structures between the siegenite and chalcopyrite also does not allow liberation of the siegenite.
The second series of studies involved the use of the SEM in conjunction with advanced image analysis and X-ray analysis. Particle distributions of test unit products were done by number of particles, by percent to total particles, and by weight percent of the sample. Again, the conclusion was that although some liberation of siegenite from chalcopyrite occurred, that this liberation involved relatively large composite particles, and that there remained a significant number and weight percent of composite particles in the less than 4-µm fraction.
Although liberation was not complete, the small size of the locked particles ruled out further grinding. Light microscopy studies done of polished sections prepared from the final cleaner concentrate revealed significant numbers of free siegenite particles in the minus 12-µm fraction. This indicated that the entrainment of the siegenite particles in the chalcopyrite froth was also contributing to the cobalt loss (fig. 8).
The copper concentrate is exposed to as many as seven separate reagents prior to cobalt separation. Several methods of reagent removal had been tried previously with little success. These included steaming, Na2S addition, and oxidation with H2O2 and ozone.
It was postulated that attritioning or scrubbing with a ceramic medium might produce a cleaner, more responsive mineral surface with which to work. With this in mind, a series of laboratory tests were initiated to evaluate the effect of attritioning. Bench testing was done with an opposed impeller scrubber with glass and ceramic media of sizes ranging from 500 to 3,000 µm for 2 h. Table 9 lists sizing data from cyclosizer tests of the attritioned products. For comparison, a typical float cell feed prepared by closed circuit grinding had the distribution in table 10.
Using the attrition mechanism for primary grinding was ruled out because of energy considerations but it was determined that a scrubber improved flotation results. In particular, the results of tests on attritioned concentrates using no reagents indicated that the mineral surfaces were being favorably altered. Also, these tests gave an indication that the attritioning might remove the starch from the surface of galena and make the galena more responsive. Table 11 shows the results of a test done on attrition-ground copper concentrate with no collector additions and with only a frother used for froth stability. Note in particular the lead assays. There is no upgrading into the tailings, which ran counter to the results on most previous tests.
Continuous Tests for Engineering Data
The final series of tests involved configuring the continuous test unit to obtain engineering data for the flowsheet and parameters developed from the prior tests. In order to be as realistic as possible, the feed for this series of tests was filtered copper concentrate taken from Magmont’s storage pad. It was felt that this offered the worst case for feed material and would best simulate the feed of any future circuit to recover cobalt. The concentrate was reslurried in an agitated tank at approximately 60-pct solids. A peristaltic pump was used to feed the slurry to the ball-mill sump at approximately 0.16 gal/min (1.4 lb/min). From the sump, the combined feed and ball mill overflow was pumped to the 3-in cyclone with the underflow going to the mill and the overflow back to the divided ball-mill sump. The overflow then fed by gravity to the attrition machine and from there to the float cells. This setup resulted in a heavy recycling of undersize through the cyclone, giving, in effect, a series of cyclones and resulting in a cyclone efficiency exceeding 97 pct.
By using an agitation holding tank for the concentrates, six stages of flotation were possible. Referring to figure 5 and the successive cleaning scheme, this flowsheet was simply run twice. Typical results from a test with six stages of flotation are given in table 12. Flotation was not carried out beyond six stages, but the grade in the final stage indicates that further stages could be used to increase recovery at little loss of grade. Cobalt grades stage by stage were, by percent: 1, 1.70; 2, 2.10; 3, 2.60; 4, 3.60; 5, 3.40; and 6, 3.00.
For the six-stage tests, the reagents and their dosages were: Sodium Aerofloat, 1.2 lb/st; Z-200, 0.25 lb/st; MIBC, 0.2 lb/st; and mercaptobenzothiozole, 0.5 lb/st. The Sodium Aerofloat was added to the attritioner, the Z-200 to the first stage, the MIBC to the first stage, and the mercaptobenzothiozole to stages 1, 2, 4, and 5.
Research using a continuous flotation test unit has resulted in a flowsheet that allows the recovery of 70 pct or more of the cobalt in Missouri lead ore copper concentrates, at grades of greater than 3 pct Co or 7 pct combined Ni-Co. This can be accomplished on feeds grading as low as 0.25 pct Co.
The flowsheet developed involves stagewise separations of the cobalt and copper-bearing materials and is more flexible than the classic rough, scavenge, and clean configuration.