The flotation separation has been carried out at laboratory scale in a discontinuous methyl polymethacrylate 40 mm ID column (Fig. 1a) with a length-over-diameter, L/D, ratio at 15. This ratio is significantly lower than in the continuous columns developed by Xu et al. [19] and used in more recent investigations [4, 18, 20], but the dimensions of the above column are in accordance to those recommended in [6, 21, 22]. The gas bubbles were generated by injection of a nitrogen flow through a sparger fixed at the bottom of the column (Fig. 1a). The sparger was a 40 mm disc of sintered stainless steel (poral, Class 5), whose pore dimensions are suitable for solid-liquid filtration of particles larger than 5.9 µm (> 98%). In a second configuration, a mechanical impeller with a rotation shaft coupled to the column bottom plug (Fig. 1b), has been mounted at the bottom of the above column: the impeller was a four tilted blade device with a diameter of 27 mm. Agitation has been tested in view to promoting the diffusion of collectors to particles and to producing smaller gas bubbles to enhance the probability of bubble-particle collision. Moreover, this more complex technique had been developed to avoid particle settling and to favor flow circulation in the column.

Fig. (1). (a) The flotation column used (40 mm internal diameter, 600 mm high); (b) Bottom plug of the column with agitation shaft.

Sodium hydroxide (in pellets) and sodium hydrogenphtalate were purchased from Sigma Aldrich (BioXtra grade) with a purity over 98 and 99.95% respectively. Cetyl trimethyl ammonium bromide (CTMAB) was also from Sigma Aldrich (> 98%). Sodium hydrogencarbonate (RP, France, Normapur) had a purity larger than 99.5%.

In the flotation experiments, chromite separation from olivine has been tested on three industrial olivines produced in Norway, differing from their particle size distribution, and kindly supplied by Sibelco (France): GL30 grade, GL10 and GL30 Milled (the last one being obtained from GL30 by milling). From the commercial chromite stock (Sibelco) having a very broad size distribution, three mineral fractions M4, M5 and T2 with different size distributions have been prepared by milling then by dry sieving. Serpentine was also provided by Sibelco.

2.4.1. Particle Size Distributions

As expected, the three olivine fractions supplied have very different particle size distributions (Fig. 2a). The distribution peak of GL30 olivine - considered here as the reference material – is centered around 120 µm to reduce the energy cost involved by further grinding in the carbonation process. A significant dispersion can be observed for GL30 Milled and GL10 olivine particles: for the first material the distribution exhibits a large peak near 80 µm followed by a broad plateau in the range 3-30 µm. For GL10 grade, the distribution presents also a broad plateau but in the range 5-50 µm. Serpentine has been calibrated from the supplied stock by dry sieving with an average particle size near 80 µm.

Fig. (2). Particles size distributions: (a) olivine and serpentine tested for chromite separation; (b) introduced chromite particles.

For development of the flotation method, M4 chromite grade was employed in a particle blend with regular GL30 olivine. As a matter of fact, this fraction has been selected because its size distribution centered at 70 µm (Fig. 2b), corresponds fairly well to the size of chromite particles contained in olivine, in the range 30-80 µm as shown by SEM observations (Fig. 3). M5 chromite is smaller than M4 grade with an average size around 35 µm, whereas coarser chromite T2 has an average size around 120 µm.

Fig. (3). SEM-EDX observations of mixed chromite-silicate particles in Sibelco chromite (left) and GL30 olivine after chromite concentration by magnetic technique (right). Top, SEM view of the particles; bottom, EDX cartography: red for Cr, blue for Mg and yellow for Si.

Preliminary tests have been carried out without stirring to determine the flow rate of injected gas. The solution was a dilute NaOH solution at pH 11, with 150 mg/L CTMAB [22]. Other collectors reported in the literature e.g. sodium oleate or dodecylsulfonate at pH 4 but the poor performances obtained led us to use exclusively CTMAB.

For experiments, 450 mL solution have been introduced first into the column; then the particles were introduced and the column wall was rinsed by pouring the remaining 50 mL. To avoid too rapid settling of the particles, gas had to be introduced at the pre-defined rate within a few seconds, initiating then the flotation run. With CTMAB, froth formation and possible flotation could not occur in a satisfactory manner for flow rates below 80 Ncm³/min, corresponding to superficial gas velocity, u_G, at 0.11 cm/s. In the range 100 – 250 Ncm³/min, in spite of different flow rates of particles collected in the froth, comparable extraction rate of chromite from the olivine stock was observed (data not shown). The reference flow rate of nitrogen injected was fixed at 150 Ncm³/min for the results presented below. For such conditions, the height of the froth was of 18 ± 2 cm. In most papers published in columns [18, 19, 21, 23], or in other flotation devices [17, 24, 25], the superficial gas velocity was in the range 1-2 cm/s. However, in other papers gas velocity in the order of 0.2 cm/s was also employed [4, 21, 22]. The gas holdup in the column, ε_G, has been estimated to approx. 2% from the variation of the upper liquid level in the column after sudden interruption of the gas feed: this value is slightly lower than values in the range 3-8% which can be deduced from formerly published investigations [5, 17-19, 25]. The actual average velocity of gas bubbles, u_b, defined as the ratio (u_G/ε_G) was near 10 cm/s, in accordance with visual observation of the gas bubbles in the Perspex column, in the range 6-10 cm/s. This real velocity corresponds to residence time of the gas in the liquid column in the range 4-7 s, in fair consistence with the residence times that can be estimated from the data reported in [18-24] and ranging from 5 to 20 s.

Prior to experiments in the column, zeta potential of the particles has been measured depending on the solution pH and collector concentration. First, the pH of a 150 mg/L (0.41 mM) CTMAB solution was adjusted before analyses, using 0.1 M solutions of sodium hydrogen phthalate and sodium hydrogen carbonate for pH 4 and 8, or aliquots of 1 M sodium hydroxide added to deionized water for pH 11. With a zeta potential of -30 mV in a 1 mM NaCl solution the surface of chromite particles with CTMAB can attain its point of zero charge (pzc) for pH near 4 (Fig. 4a). However, the surface charge of olivine particles was found comparable to that of chromite particles for pH 4 and 8 which is to affect the flotation selectivity. In contrast, different values of zeta potential of the two mineral surfaces could be obtained for pH 11.

Fig. (4). Zeta potentials of olivine and chromite particles: (a) Influence of the solution pH with CTMAB at150 mg/L; (b) Influence of the CTMAB concentration in solution at pH=11.

Conferring to this observation, the influence of CTMAB concentration is shown in Fig. (4b) for a pH fixed at 11. According to the data obtained, the optimal conditions for flotation could be found with a CTMAB concentration of 50 mg/L (0.14 mM), for which the zeta potential of chromite and olivine particles is respectively at 5.5 and 24.5 mV, whereas it attains respectively 42 and 25 mV with CTMAB concentration at 150 mg/L or higher (Fig. 4b).

Flotation tests have been made with 50 and 150 mg/L CTMAB at pH 11, with the flow conditions developed (section 3.1). With 50 mg/L CTMAB, the dispersion of gas bubbles was shown to be less efficient, and the froth quality was not sufficient to collect particles by overflow after 15 minutes, forcing the run to be stopped, with chromite recovery near 40% (15% of the solid recovered in the froth). Replicate runs confirmed the trend after comparable periods of time. The discrepancy between the flotation tests and zeta potential measurements, both with 50 mg/L, can be explained by the fact that CTMAB is acting as both frother and collector in the process. Contrary to what happens in the very diluted suspensions in zetameter cells, part of the collector added is adsorbed on the column wall and mainly on the gas bubbles (as frother): therefore, the theoretical concentration at 50 mg/L is far from sufficient for long period flotation, in addition conducted with appreciable fractions of suspended solids. In contrast, use of 150 mg/L CTMAB allowed to recover in the froth nearly 90% of the particles introduced as shown in Fig. (5a).

Fig. (5). Separation by flotation of the (GL30 Olivine - M4 chromite) system at 3.3% chromite added with 50 and 150 mg/L CTMAB at pH 11 without stirring; a) fraction of recovered particles in the froth vs. time; b) fraction of recovered chromite particles vs. time; c) Enrichment ratio in the floated solid versus the fraction of recovered chromite.

The amount of chromite extracted by flotation was also affected by the amount of introduced CTMAB (Fig. 5b). The enrichment ratio at time t was defined as the chromite concentration in the solid recovered from initial time to t (called chromite recovery) over the chromite content in the input solid blend. This ratio actually expresses the selectivity of the flotation process. The chromite recovery progresses along time but the enrichment ratio was observed to decrease with the fraction of recovered chromite in the discontinuous runs, without addition of fresh particle blend (Fig. 5c). With the higher CTMAB fraction, the first 5% solid recovered in the froth within the 5 first minutes, contained approx. 8 times more chromite that the initially introduced mineral blend, with a chromite recovery near 40% (Fig. 5c). With the lower CTMAB concentration, the enrichment ratio decreased from 6 to 2.3 after 15 minutes, for 40% chromite recovery. In the following sections, the flotation liquid was a NaOH solution at pH 11 with 150 mg/L CTMAB.

Test have been conducted with the agitation system (Fig. 1b) at 100 rpm and 200 rpm with the same gas flow rate to enhance the bubble-particle collision and the collector diffusion to the particle surface. It has been found that stirring in the column bottom led to faster recovery of solid particles in the froth than with the only injection of nitrogen, as shown in Fig. (6a). After 20 min around 75% particles could be collected with stirring at 100 and 200 rpm, whereas the corresponding particle recovery was less than 40% within this time lapse. Recovery of chromite particles was also observed to largely depend on agitation conditions: more intense agitation appears detrimental to chromite recovery (Fig. 6b): after 20 minutes, 80% of the chromite particles have been extracted from the initial blend at 100 rpm or without agitation, whereas the fraction was only near 60% at 200 rpm. All in all, the enrichment ratio was observed to be reduced by increased agitation rate. For tests without agitation, the enrichment ratio of chromite attained 8 for chromite recovery near 40%, with 5% collected particles, and 3.5 in the solids collected for 75% chromite recovery (Fig. 6c). This performance is far better than with agitation: the enrichment ratio at 40% chromite recovery attained only 2.2 and 1.3 at 100 and 200 rpm respectively, whereas it laid close to unity for chromite recovery larger than 60% at 200 rpm.

Fig. (6). Separation by flotation of the (GL30 Olivine - M4 chromite) system at 3.3% chromite added with 150 mg/L CTMAB at pH 11 depending on the stirring conditions; (a) fraction of recovered particles in the froth vs. time; b) fraction of recovered chromite particles vs. time; c) Enrichment ratio in the floated solid versus the fraction of recovered chromite.

Stirring initially thought to improve bubble-particle interactions, was actually shown here to induce the transport of particles to the column wall with the inhomogeneous dispersion of the gas bubbles which affects the flotation efficiency. In order to highlight this point, different conditions of agitation have been compared to observe flow phenomena in the flotation column without or with stirring from 100 to 400 rpm (Fig. 7). For possible visualization, mineral particles were not introduced in the photographs presented in Fig. (7). Without stirring, the gas bubbles appear evenly distributed in the column. Stirring the G/L medium was shown to be detrimental to quality and homogeneity of the gas dispersion in the solution, with the formation of a central vortex at high rotation speed. Mechanical agitation in the small column, with segregation of gas phase – in the central vortex - from the solid phase centrifuged to the column wall, actually impedes the flotation process.

This observation which may appear different from what is usually reported [5, 12, 13, 25], is due to the small diameter of the column and the short distance (near 12 mm) between the impeller and the sintered plate (Fig. 1b). In the present investigation, further experiments have been conducted without stirring.

Fig. (7). Visualization of the agitation effect on gas dispersion in the CTMAB solution at pH 11 (A: without agitation, B = 200 rpm, C = 400 rpm, D = 650 rpm). No particles added.

Because flotation is based on the upward transport of the particles by the action of bubbles, the particles size is to influence the separation efficiency because of gravity: as a matter of fact, both surface phenomena and aptitude to settling are to be involved in the overall flotation process. Tests have been conducted using the “regular combination” (GL30 olivine – M4 chromite), then the ‘fine blend’ formed with GL10 olivine and M5 chromite particles, and the ‘coarse blend’ prepared with GL30 olivine and T2 chromite particles Fig. (8). In all cases, chromite particles were added to olivine at 3 wt.% of the overall input fraction.

Flotation of the coarsest particles was found to be slightly more troublesome than with smaller particles (data not shown): as a matter of fact, flotation is known to be little suitable for particles larger than 100 µm or below [12, 13, 23]. Analysis of the floated particles revealed than the chromite recovery from the “fine” and the “coarse” blends is less effective than from the regular (GL30 olivine-M4 chromite) blend (Fig. 8a). As a consequence, the enrichment ratio is dramatically poor with the fine and the coarse particle blends (Fig. 8b) with ratio values below 2 even in the first minutes of the run, for which the selectivity is supposed to be maximal.

Attempts in interpretation are presented below. Flotation is actually to be hindered by occurrence of particle settling, whose significance is governed by adhesion of/to gas bubbles, drag forces and the intrinsic settling aptitude of these particles, as expressed in previous works [18, 23, 26].

In a first approach, the settling aptitude has been quantified here from the terminal falling velocity, assuming monodisperse, spherical particles and neglecting solid-solid particles in the media. Also neglected in the simple approach is the presence of the injected gas, whose motion is to affect the dynamics of the particles. For this purpose, the terminal velocity u_t has been calculated for the various particles using the correlation suggested by Turton and Clark [27]:

Fig. (8). Separation by flotation of various (olivine-chromite) systems at 3.3% chromite added with 150 mg/L CTMAB at pH 11. “Regular” corresponds to (GL30-M4) blend, “Fine” to (GL10-M5) system and “Coarse” to GL30-T2) combination; (a) fraction of recovered chromite particles vs. time; (b) Enrichment ratio in the floated solid versus the fraction of recovered particles.

(1)

with

(2)

where d is the average particle diameter, g is the acceleration due to gravity, ρ_L, the density of the liquid phase, ρ_S, the density of the suspended materials (approx. 3300 kg/m³ for olivine, 4100 kg/m³ for chromite), and μ the viscosity of the solution, taken at 10^-3 Pa.s. The terminal falling velocities of GL30 and M4 minerals were estimated respectively at 0.0172 and 0.0080 m/s (Table 3).

For such conditions, separation is favored by both the selectivity in particle adhesion to gas bubbles and the settling aptitude. For the two other mixtures of fine and coarse particles, the terminal falling velocity of chromite particles is higher than that of the considered olivine fractions: in such cases the gravity phenomenon opposes surface phenomena, leading to very poor flotation efficiency. For the fine particles, the enrichment ratio attained only 2 for a chromite recovery at 15%, far below that with the reference blend. The flotation of coarse particles (GL30 olivine -T2 chromite) appears really difficult because of the high value for u_t, larger than 0.02 m/s, somewhat larger than that of GL30 particles. For this last blend, it is likely that the chromite particles over 100 µm are too large for efficient flotation.

From the experience gained, further flotation tests have been conducted by treating 15g solids with higher contents of Sibelco chromite in the column to emulate a multistage flotation process with the same fraction of solids, to maintain constant the solid residence time in the column and the multiphase flow conditions. Regular (GL30 olivine -M4 chromite) combination has been considered here. The recovery of solids by the froth along the discontinuous run was little affected by the initial chromite content (data not shown), however the fraction of recovered chromite appeared very dependent on the input composition (Fig. 9a). With exception of data obtained with 3 and 5% chromite, the more concentrated the blend, the faster the recovery of chromite particles: in particular, for the 20% chromite blend, more than 93% of the introduced chromite could be extracted within 10 minutes, to be compared with the 63% chromite recovery obtained in this time lapse when using the 3.3% chromite enriched olivine. As expected, the enrichment ratio calculated is a decreasing function of the separation progress in the discontinuous tests. No real correlation could be found between the enrichment ratio and the fraction of recovered chromite (data not shown), but as shown in Fig. (9b), the experimental values for the enrichment factor E could be well correlated when plotted versus the particle recovery fraction R, and for the operating conditions considered, the following numerical law has been found:

Fig. (9). Separation by flotation of the (GL30 Olivine - M4 chromite) system at various amounts of chromite added (3.3-20%) with 150 mg/L CTMAB at pH 11; (a) fraction of recovered chromite particles vs. time; (b) Enrichment ratio in the floated solid versus the fraction of recovered solid.

(3)

where R is expressed in percentages, a=86.7%, b=6.3% and c=0.26. Coefficients a, b, and c of the above relation have a restricted physical meaning since they depend on the system used (column, electrolyte) and the flow conditions. Nevertheless, rel. (3) expresses that the enrichment ratio does not depend on the chromite content of the solid to be treated.

The operating protocol developed above has been applied to the treatment of the original minerals. Flotation tests have been conducted on the various olivines (GL30, GL30 Milled and GL10) to confirm the suitable particles size distribution, and with serpentine.

Best performance of chromite particles from the mineral has been obtained using olivine GL30: chromite recovery attained 90% after one hour (Fig. 10a) and the enrichment ratio peaked at 3 in the first minutes, however with a regular decreasing trend along the run (Fig. 10b). Flotation using G30 Milled olivine and serpentine allows enrichment ratio of 1.25 to 1.5 only in the first half of the flotation test. Chromite separation from GL10 olivine, which is the finest mineral was observed to be of nil efficiency. Because a significant fraction of particles is below 20 µm in olivine GL10 and GL30 Milled, mineral particles can be entrained with little effect of zeta potential in gas-solid interactions, resulting in a very poor selectivity: the reduction in gas flow rate should improve the separation efficiency, but other separation techniques would also be considered. For serpentine, which is a hydrated form of iron and magnesium silicate, the density of 2.5-3.0 g/cm³ - somewhat below than olivine - may also affect the flotation selectivity with the same consequence as for fine particles. Moreover, due to the presence of hydroxide groups on the silicon atoms, serpentine exhibits likely different affinity to the collector from that of olivine, resulting in less efficient adsorption on this mineral.

Fig. (10). Separation of chromite from olivine and serpentine fractions with 150 mg/L CTMAB at pH 11; (a) fraction of recovered chromite particles vs. time; (b) Enrichment ratio in the floated solid versus the fraction of recovered solid.

Considering the initial target, flotation of chromite particle as a pretreatment of the olivine feedstock is more selective with GL30 olivine. However, the modest efficiency of chromite separation is partly due to very poor chromite content (near 0.5 wt%) in the olivine studied. Moreover, chromite is also present in mixed chromite-olivine solid particles, as shown in Fig. (3), in addition to pure chromite particles: the behavior of these mixed particles in the flotation column is to be intermediate between those of pure chromite and pure olivine particles, rendering their separation from chromite-free olivine more difficult.