The flocculation efficiency of polydisperse polymer flocculants

Transcription

1 Int. J. Miner. Process. 73 (2004) The flocculation efficiency of polydisperse polymer flocculants Y.D. Yan*, S.M. Glover, G.J. Jameson, S. Biggs Centre for Multiphase Processes, The University of Newcastle, Callaghan NSW 2308, Australia Received 18 October 2002; received in revised form 22 October 2002; accepted 16 April 2003 Abstract The flocculation performance of three poly(acrylic acid) (PAA) samples (M w =9 10 4, and g/mol) has been investigated. Colloidal alumina particles were used as a model system and tests were performed at ph 5. Using a singlecomponent polyacid, it was found that the optimum dosage required to achieve supernatant clarity was similar between the g/mol PAA (23 ppm) and g/mol PAA (26 ppm), but increased dramatically with the g/mol PAA (83 ppm). For the two lower molecular weight samples, flocculation occurs through a charge neutralisation mechanism. In contrast, polymer bridging is inferred to be the dominant flocculation mechanism for the high molecular weight sample. The flocculation performance of a polymer mixture, produced by blending the high and low molecular weight polyacids to give an average molecular weight of g/mol, was also studied. Supernatant clarity from this system was found to be comparable to that from the single-component polyacid of the same (average) molecular weight. However, the optimum dosage required for the polymer mixture was about twice as much as that for the single-component reference polymer. The results suggest that for the polymer mixture no synergistic effects occur. Instead, analyses of aggregate sizes indicate an independent behaviour for the two polymers in the blend. We also examined re-suspension (under shear) and re-flocculation of the sediment formed in the initial flocculation experiments. For the three single-polymer systems, rapid re-flocculation after shear was seen for the two lower molecular weight samples suggesting a reversible aggregate breakage. For the high molecular weight sample, re-suspension resulted in the formation of a stable dispersion. This result was attributed to breakage of the high molecular weight polymer sample during re-suspension. In the case of the polymer mixture, rapid re-flocculation was again observed despite the presence of a large amount of the high molecular weight sample. This result may have important practical implications. D 2004 Elsevier B.V. All rights reserved. Keywords: flocculants; flocculation; sedimentation; settling 1. Introduction Since the introduction of synthetic water-soluble polymer flocculants in the early 1950s, their use has become a common practice across a wide range of industrial solid liquid separation operations such as * Corresponding author. Fax: address: cgyy@cc.newcastle.edu.au (Y.D. Yan). sedimentation, clarification, filtration, dewatering and flotation (Moudgil and Behl, 1995). Practical examples include thickening operations in mineral processing, industrial tailings dewatering, papermaking, and water and wastewater treatment. The use of polymers, especially those of very high molecular weights, has resulted in tremendous performance improvement for industrial separation processes. There are numerous polymer flocculants currently available on the market. However, they can conve /$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi: /s (03)

2 162 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) niently be grouped into two broad categories: nonionic polymers and polyelectrolytes (cationics and anionics). Various forces have been found to be responsible for the adsorption of polymers onto particle surfaces (Gregory, 1987). For nonionic polymers such as polyacrylamides, hydrogen bonding and/or hydrophobic forces are believed to be the main driving forces for adsorption. For polyelectrolytes interacting with oppositely charged solid surfaces, electrostatic attraction becomes the key factor for attachment. When nonionic polymers are used or when polyelectrolytes are added to particles of like sign of charge, particles are aggregated through the bridging flocculation mechanism, where segments of the same polymer molecule are attached to more than one particle. This type of flocculation mechanism has been found to be very efficient. The situation becomes much more complex when polyelectrolytes are used with oppositely charged particles. Three mechanisms have been suggested (see, e.g., Gregory, 1987). These are charge neutralisation, electrostaticpatch and bridging flocculation. It is generally believed that low molecular weight polymers tend to adsorb and neutralise the opposite charges on the particles. Consequently, flocculation is caused by the reduction in the electric double layer repulsion between particles. The electrostatic-patch model (Gregory, 1973) can be considered as another form of bridging flocculation and is believed to be operative for polymers of very high charge density interacting with oppositely charged particles of low charge density. The net residual charge of the polymer patch on one particle surface can attach to the bare part of an oppositely charged particle. Our understanding of the particle flocculation process involving adsorbing polymers has progressed considerably in recent years. Nevertheless, flocculation optimisation practices in industries are still reliant, to a very large extent, on trial and error. This reflects the highly complex nature of the flocculation process which can involve the following stages (Elimelech et al., 1995): (a) particle polymer mixing; (b) attachment of the polymer molecules onto the particle surface; (c) reconformation of the polymer molecules on the particle surface; (d) particle flocculation; and, (e) floc breakup due to shear mixing. These processes can take place concurrently and are often competing. Therefore, in order to achieve efficient flocculation it is crucial to gain a good fundamental understanding of the key factors which can significantly influence the time scale of a particular stage. Like shear mixing, polymer characteristics are amongst some of these factors. There exists a large body of literature on the effect of polymer characteristics on adsorption and particle flocculation. These include parameters such as polymer molecular weight or chain length (Ash and Clayfield, 1976; Gregory, 1976; Mabire et al., 1984; Wang and Audebert, 1987; Wong et al., 1988; Tanaka et al., 1990), conformation (Tjipangandjara and Somasundaran, 1991; Yu and Somasundaran, 1996a,b; Stoll and Buffle, 1998; Yates et al., 2001), charge density (Lindquist and Stratton, 1976; Mabire et al., 1984; Wong et al., 1988; Tanaka et al., 1990; Swerin et al., 1993) and functional groups (Somasundaran et al., 1996 and references therein). Interestingly, despite the fact that it is almost invariably true that all commercial polymer flocculants possess significant degrees of distribution in their molecular weight, there appears to be very limited literature on the effect of polymer molecular weight polydispersity on particle flocculation. One may expect that two polymers of the same type showing identical average molecular weight but rather different distributions may exhibit varying flocculation performance. Therefore, a better understanding of how polymer molecular weight distribution can affect its flocculation performance may lead to improved flocculant manufacturing processes and better choice of flocculants for the users. Ideally, one should compare the flocculation performance of a polydisperse polymer to that of a monodisperse polymer, with the latter possessing a molecular weight equal to the mean value of the former. Nevertheless, as a starting point, in this work we used three poly(acrylic acids), each possessing a degree of polydispersity in molecular weight. Examination of the polymer molecular weight distribution effect involved comparing the flocculation behaviour between the polyacid of intermediate molecular weight (the reference) and the mixture of the other two polyacids (the mixture). The mixing ratio for the latter was carefully chosen so that it yielded an average molecular weight equal to that of the reference. Poly(acrylic acid) or PAA samples were chosen

3 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) since they are the simplest common synthetic polycarboxylic acids and provide an ideal model for examining the flocculation performance of more complex polyelectrolytes. Our study represents the extreme case of polymer polydispersity and the results may have practical implications for applications such as flocculant blending. Flocculation tests on the performance of the PAA mixture as well as each of the three single-component polyacids were carried out using positively charged colloidal alumina particles. The size of the flocs formed and their settling behaviour were also examined using small-angle light scattering and turbidimetry techniques, respectively. 2. Materials and methods 2.1. Alumina particles The alumina particles used here were supplied by Sumitomo Chemical (Japan) in dry powder form (AKP-30, High Purity). The powder was first immersed in distilled water as a 40% w/w slurry and then ph adjusted to 5. The final concentration of this stock suspension was 25% w/w. After about 2 weeks the stock sample was sonicated using an ultrasonicator (Misonix, Ultrasonic Processor XL2020, equipped with a 1-cm diameter tip). This procedure was found to be effective in fully dispersing the alumina powder. Unless stated otherwise, the particle concentration used in all the flocculation tests reported in this paper was 2.5% w/w (ph 5). After preparation each sample was sonicated and then left standing for 3 h before use to avoid complications such as the possible formation of free radicals on the solid surfaces as a result of sonication. The dispersed alumina particles were characterised in terms of their size, shape and surface charging behaviour. Static light scattering (see below) indicated a fairly broad size distribution with a volume-averaged diameter of 0.39 Am. As was shown in our previous work (Glover et al., 2000), the particles were more polyhedral than spherical when viewed under TEM. Electrophoretic mobility measurements on these particles as a function of ph indicated an isoelectric point of ph 9.5. Below ph 9.5 the particles were positively charged and above this point they were negatively charged Polymer flocculants The (nominal) average molecular weights of the three poly(acrylic acid) or PAA samples used in this study were g/mol (Polysciences), g/ mol (Aldrich) and g/mol (Polysciences). They were used as received. They each contained a degree of distribution in molecular weight. The concentration of all the single-component PAA solutions used was 0.2% w/w, unless stated otherwise. They were also ph adjusted to 5. The PAA mixture with a calculated average molecular weight of g/mol was prepared by carefully mixing the and g/mol polyacids using a pre-estimated weight ratio. The overall polymer concentration of this mixture was also 0.2% w/w, amongst which the g/mol PAA amounted to 0.06% w/w and the g/mol PAA 0.14% w/w. The mixture was ph adjusted to 5. All ph adjustments in this work were done using 1 M HNO 3 or 1 M KOH. All water used was Milli-Q Plus water Flocculation tests The flocculation experiments were performed in a stirred beaker. For each test a 300-ml sample of the alumina suspension was placed into a 600 ml glass beaker (ID c 90 mm). A PVC baffle was inserted. Mixing was achieved using a Shelton overhead mechanical stirrer fitted with a standard six-blade Rushton impeller (blade length = 8.5 mm, circular pitch = 6.5 mm, shaft centre-to-blade distance = 9 mm). Unless stated otherwise, the mixing speed was kept at 977 rpm in all tests. Polymer was added to the slurry under continuous mixing in a single quick injection. The sample was then kept stirred at the same mixer speed for 2 min. After that, a portion (ca. 50 ml) of the flocculated sample was poured into a glass tube for both visual inspection and photography. It should be pointed out that one should not attempt to compare the sediment bed heights between the samples contained in the various tubes, since it was difficult to ensure that an equal amount of solids was transferred across each time.

4 164 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) Floc size and zeta potential measurements Floc sizes were measured on a small-angle static laser light scattering device (Mastersizer S, Malvern Instruments, UK). Dilution of a flocculated sample was necessary to keep the obscuration value to within a suitable range for sizing. This was done by carefully drawing and then releasing a small amount of the sample to an internally magnetically stirred cell which contained filtered water of the same ph. During measurement flocs were suspended by gentle stirring in the measuring cell to avoid settling. Electrophoretic mobility measurements were done using a ZetaPlus apparatus (Brookhaven Instruments, USA). Two types of flocculated samples were used for these measurements. The first one was taken from the supernatant containing tiny flocs, 5 min after the cessation of shear mixing. The second sample was again taken from the top part of a settling tube, but the sample had been subjected to sonication for 2 min (1 s on and 1 s off) and left standing for 3 h Floc settling behaviour The floc settling behaviour was measured here in terms of supernatant formation kinetics using a liquid dispersion optical characterisation instrument (Turbiscan MA2000, FormulAction, France). A flocculated sample to be analysed was carefully added into a cylindrical glass cell. The reading head scans across the height of the sample over time. It reports the percentage of the light either transmitted through or back scattered from the sample. The thickness evolution of the supernatant phase was obtained by analysing the data using a so-called absolute peak thickness method as described in the Turbiscan manual. 3. Results and discussion 3.1. Flocculation behaviour of the single-component polyacids alone Before exploring the flocculation characteristics of polydisperse polymer flocculants, it would be useful to gain a good understanding of the flocculation behaviour of the single-component polymers involved. Fig. 1a c shows the flocculation performance of each of the single-component poly(acrylic acids) used in this work. The molecular weights of these polyacids were , and g/mol, respectively. Observations common to each of these polymers are that there was an optimum polymer dosage at which the best supernatant clarity was obtained. Far below the optimum point the alumina suspension was underdosed, resulting in a cloudy supernatant. Far above the optimum point the suspension was overdosed and again the supernatant became cloudy. Under the experimental conditions used, the optimum dosage for the , and g/mol polyacids was found to be 23, 26 and 83 ppm, respectively. Note that for each PAA used there was a range of dosage over which reasonably good supernatant clarity was obtained. The optimum dosage for the two shorter chain polyacids could be even closer to each other than the values listed above. At ph 5 the polyacids should be negatively charged and the carboxylic functional groups on each PAA chain are expected to be about 50% ionised, whereas the alumina particles have, as mentioned before, positive surfaces. As a result, electrostatic attraction is expected to play a key role in the adsorption of these polyacids onto the alumina particle surfaces. As stated earlier, for oppositely charged polymers and particles, three possible mechanisms can be responsible for particle flocculation, i.e., charge neutralisation, electrostatic-patch and polymer bridging. We argue that flocculation of the alumina particles with the g/mol PAA (Fig. 1a) was most likely attributable to simple charge neutralisation. For the g/mol PAA (Fig. 1b) flocculation must be predominantly due to simple charge neutralisation, but at the same time a small proportion of the particles could be flocculated through true polymer bridging. Flocs with the g/mol PAA (Fig. 1c) were undoubtedly induced by true polymer bridging. These arguments were based upon the following observations: (i) the optimum dosage as identified in Fig. 1a c, (ii) the effect of shear mixing rate on flocculation, (iii) the effect of either prolonged shearing or sonication on flocculation, and (iv) floc size and size distribution at optimum polymer dosage. As can be seen from Fig. 1 and Table 1, the optimum dosage was nearly the same for the and g/mol polyacids, while that of the g/

5 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) Fig. 1. Flocculation of alumina particles with the single-component poly(acrylic acids) used: (a) g/mol, (b) g/mol and (c) g/mol. The polyacid dosage required to achieve the best supernatant clarity in (a) (c) was 23 ppm (or 3.5 ml), 26 ppm (or 4 ml) and 83 ppm (or 13 ml), respectively. Note that the dosage labels shown in the figure refer to millilitres of a 0.2% w/w polyacid solution used for each 300 ml of the 2.5% w/w alumina suspension. In all cases shear mixing was kept at 977 rpm for 2 min. mol PAA was significantly higher. Should simple charge neutralisation be the underlying flocculation mechanism for each of these polyacids, their optimum dosage should be close. The results lead us to postulate that with the two shorter chain polyacids, aggregation of the alumina particles may have followed a similar flocculation mechanism (e.g., charge neutralisation), whereas the g/mol PAA must have resulted in a very different flocculation mechanism (e.g., bridging flocculation). The zeta potentials

6 166 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) Table 1 Summary of the flocculation performance of poly(acrylic acids) of varying molecular weights used in this study PAA M w (g/mol) Optimum PAA dosage (ppm) Floc D[4,3] (Am) a Zeta potential (mv) b Zeta potential (mv) c 90, , , (mixture) 1,000, C alumina = 2.5% w/w. a The volume-weighted average of the equivalent spherical diameter. b Measured from the residual flocs in the supernatant after 5 min settling. c Measured from the particles or residual flocs, 3 h after the cessation of sonication. obtained from the residual flocs in the supernatant at the optimum polymer dosage (cf. Table 1) were indeed very small in the case of the two lower molecular weight polyacids. In the case of the g/mol PAA, such residual flocs showed a large negative value. In the latter case there were about three times as many charged carboxylic groups being added to the alumina suspension as in the case of the other two polyacids. Therefore, it was not surprising that the resultant flocs with the g/mol PAA showed a negative zeta potential. We also investigated the effect of shear mixing rate on flocculation. This was done by decreasing the mixer speed to almost one-third from 977 rpm as used in Fig. 1 to 332 rpm. The sample was also mixed for 2 min after polymer injection. All tests were carried out using the optimum polymer dosage as identified in Fig. 1. The results are shown in Fig. 2. Notice in Fig. 2 that at the shear mixing rate of 332 rpm the supernatant of the alumina suspension flocculated with 23 ppm of the g/mol PAA remained clear and that with 26 ppm of the g/ mol PAA became only slightly more cloudy. In contrast, the suspension flocculated with 83 ppm of the g/mol PAA became fairly cloudy. In this latter case another 20 ppm of this polyacid were required to restore the clarity. One can expect that a lowering in the shear mixing rate will result in a reduction in the interparticle collision rates. This will in turn increase the time available for a polyacid molecule to adopt a more equilibrium (or flatter) conformation on the alumina surface before encountering another alumina particle and hence causing flocculation. A flatter polymer conformation on the particle surface is expected to require a higher polymer dosage if flocculation is via the bridging mechanism, but should have no influence on flocculation if particles are destabilised by charge neutralisation. Hence the results in Fig. 2 seem to support our arguments for charge neutralisation with the g/mol PAA and bridging flocculation with the g/mol PAA. With the g/mol PAA the dominant mechanism must be charge neutralisation and polymer bridging may also have played a role. Further tests, by subjecting the optimally flocculated samples of Fig. 1 to either prolonged shear mixing or sonication, provided additional support to our arguments above. In the prolonged shear-mixing test, stirring was maintained for a total of 20 min after polymer addition. Interestingly, it was observed (Fig. 3a) that upon the cessation of shear the supernatant remained clear in the case of the and Fig. 2. Flocculation of the alumina suspensions at a lower shear mixing rate, with polyacid samples of varying molecular weights. From left to right: 23 ppm of g/mol PAA, 26 ppm of g/mol PAA, 46 ppm of g/mol PAA mixture and 83 ppm of g/mol PAA. Shear mixing was kept at 332 rpm for 2 min.

7 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) Fig. 3. Effect of shear mixing time and sonication on the supernatant clarity of the alumina suspensions flocculated with polyacid samples of varying molecular weights. From left to right: 23 ppm of g/mol PAA, 26 ppm of g/mol PAA, 46 ppm of g/mol PAA mixture and 83 ppm of g/mol PAA. (a) After shear mixing at 977 rpm for 20 min, and (b) after 2 min sonication. g/mol polyacids, whereas the supernatant with the g/mol PAA became rather cloudy. This floc breakup effect was magnified in the second test, where samples flocculated under the identical optimum conditions as in Fig. 1 were subjected to intermittent sonication (1 s on and 1 s off) for 2 min (Fig. 3b). During sonication it was observed that flocs formed using the g/mol PAA were completely broken up and they did not re-appear upon the cessation of sonication. In contrast, we noticed through visual observation that flocs formed with the and g/mol polyacids appeared to be resistant to the ultrasonic forces and they persisted throughout the sonication period. This must be due to the rapid floc re-formation and growth during the 1 s intermittent stop-period following each 1 s sonication. Indeed, when we subjected these two samples to 1 min continuous sonication, there were no flocs visually observable in each sample during sonication. However, flocs re-appeared right after the cessation of sonication. It was noticed (Fig. 3b) that powerful sonication resulted in a slight decrease in supernatant clarity in the case of the two shorter chain PAA samples. The prolonged shearing, and in particular the sonication test, provide strong corroborative evidence for the flocculation mechanisms postulated earlier for each of the polyacid samples used. Flocs which are initially formed via charge neutralisation and then broken up by shearing can easily re-form upon removing the shearing forces. However, polymer-bridged particles stay apart once broken up, since polymer tails and loops bridging across two or more particles are physically severed by the shearing forces. This chain breakage leaves residual polymer on both surfaces. Surface rearrangement of these polymeric components can lead to particle restabilisation against further aggregation by the presence of an electro-steric polymer layer on the surface. It should be pointed out that it is practically difficult, based solely on the above tests, to distinguish between simple charge neutralisation and the electrostatic-patch model. This is because the electrostaticpatch mechanism can also occur at apparent particle charge neutrality and flocs formed through this mechanism can also re-grow after being broken up by shear (Gregory, 1987). It can also be expected that this mechanism should be less influenced by the conformation of the polymer molecule on the particle surface and hence less dependent on the shear mixing rate. In view of the fact that with the two short-chain polyacids apparent charge neutrality was observed regardless of which of these two mechanisms was

8 168 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) operative, it would be reasonable to simply use charge neutralisation as a general description of the flocculation mechanism. More supporting evidence for the flocculation mechanism/s proposed above for each of the singlecomponent polyacids used in this work comes from direct measurement of floc size from the samples flocculated at optimum conditions in Fig. 1. Sampling was made just before the cessation of the 2 min mixing time. As can be seen from Fig. 4, for both the g/mol and g/mol PAA, only unimodal distributions in floc size were observed, whereas a second larger-size peak was present in the case of the g/mol PAA. Notice that the distribution curve of the smaller-size flocs in the g/mol PAA case peaked almost at the same position as that for the case of the g/mol PAA. This, together with the fairly small sizes of these flocs, indeed suggests that the dominant particle flocculation mechanism for these two polymers was charge neutralisation. The presence of the larger-size flocs with a relatively small volume percentage in the case of the g/mol PAA indicates that a small portion of the particles was flocculated via bridging flocculation. Finally, formation of huge flocs in the case of the g/mol PAA is in agreement with the polymer bridging mechanism. It should be remarked that the fact that the optimum dosage of the g/mol PAA was seen here Fig. 4. The size distribution of alumina flocs, formed under the optimum conditions as shown in Fig. 1, for each of the three singlecomponent polyacids used. to be considerably higher than those of the g/ mol and g/mol polyacids is contradictory to what one would normally expect from bridging flocculation. Interestingly, in a separate study we noticed that about 23 ppm of a g/mol anionic polymer (copolymer of acrylic acid and acrylamide) were required to flocculate the same alumina particles under similar conditions. Therefore, there appear to be some peculiar aspects about the conformation of high molecular weight poly(acrylic acids) on these alumina surfaces. It seems that in our case, the g/mol PAA (at low dosages) may not have the expected loops and tails conformation with segments extended far into the suspension which is needed for effective bridging. Instead, the polyacid molecules must have adopted a fairly flat conformation, possibly due to strong adsorption of these PAA molecules on the alumina surfaces. This is in line with the findings of a recent study by Vermohlen et al. (2000) on adsorption of polyelectrolytes onto oxides at ph = 5.2, where it was concluded that at low solution ionic strength PAA molecules (M w = g/mol) were adsorbed onto alumina surfaces in a flat conformation. Interestingly, in our case with the g/mol PAA, alumina particles obtained after ultrasonic breakup of the flocs were found to have a large negative zeta potential (see Table 1). The latter indicates that strong polyacid adsorption may have led to particle charge reversal. Thus requirement of high dosages of the g/mol PAA is likely attributable to the combined effects of chain conformation and particle charge reversal. It is worth noting that in a recent dual-polymer flocculation study involving alumina particles (Fan et al., 2000), there was indication that the flocculation efficiency of PAA improved as its molecular weight was increased from to g/mol. However, this study was carried out at a ph of 8, not far from the isoelectric point of their alumina particles. Thus the strong electrostatic attraction effect seen in our polymer/alumina system must be greatly reduced in Fan et al. s (2000) work, and hence the polymer conformation on the surface may not be as flat as in our case. We have also carried out a flocculation test similar to that in Fig. 1c by using a 10-fold dilution both for the alumina particles (i.e., 0.25% w/w) and the g/mol PAA solution (i.e., 0.02% w/w). We

9 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) did this to examine whether the unexpected large dosage required for this high molecular weight polyacid was due to insufficient particle polymer mixing. It was found that flocculation of the diluted system became, as expected, slightly less efficient than that in Fig. 1c, both in terms of optimum dosage required and supernatant clarity achievable. Hence, we are confident that mixing was not a significant issue in our systems Flocculation behaviour of the polyacid mixture The polyacid mixture having an estimated average molecular weight of g/mol was prepared by carefully mixing the appropriate proportions of the and g/mol polyacids. This mixture is thereafter referred to as g/mol PAA (mixture) or PAA mixture. The flocculation performance of this rather polydisperse (or bimodal) polyacid is shown in Fig. 5a. For ease of comparison with the PAA reference of equal average molecular weight, we show in Fig. 5b again the samples from the singlecomponent g/mol PAA (or PAA reference). To our knowledge, till now there have been few flocculation studies which have actually compared directly the flocculation performance of a polymer mixture with that of an appropriate reference polymer. As can be seen from Fig. 5, the PAA mixture (Fig. 5a) exhibited qualitatively similar flocculation phenomena to those of the single-component reference polymer (Fig. 5b). That is, for the PAA mixture there also exists a limited dosage range (or tolerable-dosage Fig. 5. Effect of polymer molecular weight polydispersity on alumina particle flocculation. (a) PAA mixture and (b) PAA reference. The estimated average molecular weight of the PAA mixture was g/mol. The optimum dosage was 46 ppm (or 7 ml) for the mixture and 26 ppm (or 4 ml) for the reference. Shear mixing was kept at 977 rpm for 2 min.

10 170 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) range) over which reasonably good supernatant clarity can be obtained, and the supernatant clarity at the respective optimum polymer dosage points is comparable. However, the optimum dosage for the PAA mixture (46 ppm) is significantly higher than that for the PAA reference (26 ppm); there is also not much overlap between the two tolerable-dosage ranges. Noticeably, the PAA mixture appears to have resulted in a markedly wider tolerable-dosage range. The fact that the optimum dosage of the PAA mixture was intermediate between those of the individual components and higher than that of the reference sample suggests that there was little synergistic flocculation effect between the low molecular weight ( g/mol) and high molecular weight ( g/mol) polyacids in the mixture. However, one should be careful not to generalise this finding. While this is true for the specific system under study, i.e., alumina particles and mixture of the chemically identical polyacid species, for other systems remarkable synergism has indeed been observed. Csempesz et al. (Csempesz and Rohrsetzer, 1984, 1988; Csempesz et al., 1987; Csempesz, 2000; Csempesz and Csaki, 2000) have studied the adsorption and flocculation behaviour of three neutral polymers (methylcellulose, polyvinyl alcohol and polyvinyl pyrrolidone), both individually and in 1:1 (w/w) binary mixtures, on aqueous dispersions of silver iodide sol, arsenic trisulphide sol and polystyrene latex. They found that with the silver iodide and arsenic trisulphide sols, binary mixtures involving polyvinyl pyrrolidone showed flocculation efficiency much higher than that of either component in the mixture alone. They attributed this synergistic effect to the formation of irregular polymer conformation at the particle/ water interface as a result of a considerable difference in the strength of adsorption of the competing polymers. Interestingly, no such synergistic effect from the binary mixtures of various combinations was observed with the polystyrene particles. As has been pointed out by Csempesz et al. (1987), to see synergism requires the weakly adsorbed polymer itself being a more effective flocculant than the preferentially adsorbed polymer. This was found to be the case with the silver iodide and arsenic trisulphide sols but not with the latex particles. For our PAA mixture it would be reasonable to assume that the g/mol PAA was more strongly adsorbed to the alumina surfaces than the g/mol PAA. This, together with our finding that the latter was also a more effective flocculant by itself, should have put our PAA mixture into the category of synergism. This plausible contradiction with experiment may, however, be explained by our observation that contrary to instincts, the two polyacids in the PAA mixture in fact adsorbed onto the alumina surfaces quite independently. This will become clear after we have discussed the floc size distribution data in the next section. Note that Fan et al. (2000) have examined the influence of molecular weight distribution of the primary polymer on alumina flocculation performance of the secondary polymer in a dual-polymer study. They mixed and g/mol PAA samples in a 1:1 (w/w) ratio at ph 8. While at very low dosage of the secondary polymer, the flocculation efficiency of this mixture was intermediate between that of the two individual components (i.e., no synergism in flocculation was observed), the polydisperse nature of this mixture did lead to a wider effective flocculation region when the secondary cationic copolymer was added. They speculated that this improved overall efficiency was possibly the result of a wider size distribution of the primary flocs. As it will be shown later in this paper, we indeed observed such broadening effects in floc size distribution in the case of our PAA mixture. Interestingly, in their work increasing the polydispersity of the secondary cationic copolymer showed no such effect. Here we found that, in comparison with the PAA reference, the PAA mixture was more strongly influenced by the mixing conditions, e.g., shear mixing rate (see Fig. 2), shearing time (Fig. 3a) and shearing intensity (Fig. 3b). This is to be expected since, in weight ratios, the PAA mixture was made up of 70% from the g/mol PAA and only 30% from the g/mol PAA. As we have discussed in the previous section, the flocculation mechanism of the alumina particles was charge neutralisation when the g/mol PAA was used alone, whereas the mechanism changed to polymer bridging with the g/mol PAA. We have already discussed why bridging-flocculated particles are more susceptible to the mixing conditions mentioned above. What is interesting is that with the presence of only 30% of the g/mol PAA, the alumina suspension flocculated using this PAA mix-

11 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) ture showed remarkable floc resistance against breakup when subjected to either prolonged shear mixing (cf. Fig. 3a) or more forceful sonication (cf. Fig. 3b). This apparent shear resistance must be due to the rapid re-formation through the charge neutralisation mechanism rendered by the g/mol PAA, upon the cessation of the shearing forces, of flocs in an otherwise (partially or fully) homogenised suspension. Similar to those suspensions flocculated using the and g/mol single-component PAA samples, it was also observed that, with the PAA mixture, flocs did not seem to disappear during the 2 min intermittent sonication test (1 s on and 1 s off). Again, this must be due to the rapid reformation of flocs during the 1 s off period, largely helped by charge neutralisation. Re-appearance of a reasonably clear supernatant (see Fig. 3b) from the PAA mixture after being subjected to intense sonication is fascinating. This may have important practical implications such as in papermaking, since one can try to blend a small portion of a low molecular weight polymer into a high molecular weight flocculant to create a flocculation process that will be much less susceptible to those adverse effects (e.g., floc breakup) caused by the shearing forces in the process. We also investigated the effect of the order of addition of the two polyacid species in the PAA mixture on alumina flocculation. Here we carried out the tests under the same experimental conditions as used in Fig. 5a. The tests done were: (a) 14 ppm of the g/mol PAA were added first to the suspension, followed 1 min later by 32 ppm of the g/mol PAA. The suspension was stirred for another 1 min. Thus the total mixing time was 2 min at 977 rpm; (b) Only the sequence of polymer addition as described in (a) was reversed; and (c) 23 ppm of the PAA mixture were added first, followed 1 min later by another equal amount of this polymer. Again, the suspension was stirred for another 1 min. Therefore, the total polymer dosage in all three tests was 46 ppm, which corresponds to the optimum dosage of the PAA mixture as identified from Fig. 5a. Fig. 6 shows the results from the above tests. It is surprising to find that the flocculation behaviour in these three tests was almost identical. We had expected to see a somewhat overdosed situation with test (a), since the short-chain polyacid was expected to Fig. 6. Effect of the order of PAA addition on alumina flocculation. From left to right: 14 ppm of g/mol PAA, followed by 32 ppm of g/mol PAA; 32 ppm of g/mol PAA, followed by 14 ppm of g/mol PAA; and 23 ppm of the PAA mixture, followed by 23 ppm of the same mixture. Shear mixing was kept at 977 rpm for a total of 2 min. attach onto the alumina surfaces first and thus might block some of the surface sites for the subsequent adsorption of the long-chain polyacid. Hence we had expected that the long-chain polymer would adopt a less flat conformation on the particle surface and result in more efficient bridging flocculation when compared with that in Fig. 5a, where a single dose (46 ppm) of the PAA mixture was performed. Site-blocking effects have been observed in a number of other systems (Swerin et al., 1997 and references therein). Based on the findings of Swerin et al. (1997) on adsorption onto and flocculation of microcrystalline cellulose particles by two cationic polymers (poly- DADMAC and polyacrylamide), it seems possible that both the concentration and molecular weight of the short-chain PAA used in our PAA mixture were too low to induce effective site blocking. For reasons that will be discussed in the next section, it suffices to say that the flocculation behaviour in test (b) above should be expected to be similar to that in Fig 5a. It is known in the literature that high molecular weight polymers tend to adsorb onto particle surfaces

12 172 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) preferentially as compared to the homologous low molecular weight ones (Fu and Santore, 1998). Furthermore, it has been observed that when added after the lower molecular weight polymers, the more affinitive high molecular weight polymers can cause desorption of the already adsorbed smaller species. In view of the rather short time period of sample mixing and the generally nonequilibrium nature of the flocculation process, we do not expect this type of preferential adsorption to play a significant role in test (a) above, which would otherwise effectively lead to the addition method of test (b). The lack of enhancement in the bridging flocculation efficiency of the g/mol PAA by first adding the short-chain polyacid is also likely to do with the high affinity of the g/mol PAA toward the alumina surfaces. The latter may mask any site blocking effect from the short-chain PAA Comparison of floc size and size distribution Flocculated alumina samples were taken for sizing using light scattering just prior to completion of the 2 min shear mixing period. The results from the samples flocculated under the optimum dosage conditions as identified in Fig. 5a and b are presented in Fig. 7. Notice that the size distribution curves from both the Fig. 7. The size distribution of alumina flocs, formed under the optimum conditions as shown in Fig. 5, from the PAA mixture and PAA reference. Fig. 8. Floc size distribution as a result of the order of PAA addition. The two samples measured here were the same ones as those reported in Fig. 6. PAA mixture and reference are bimodal. As discussed in Section 3.1, for the PAA reference, the peak at a larger size range (small volume percentage) may be attributed to a small contribution from bridging flocculation with this polymer. Considering the fact that the PAA mixture was synthesised using the and g/mol polyacids, the small- and large-size peaks in this mixture may be readily associated with respective contributions from the low and high constituent molecular weight polyacids. It is not surprising to see a more prominent larger-size peak in the case of the PAA mixture, as the g/mol PAA accounted for 70% w/w in the premixed polymer sample. The interesting part of the results from the PAA mixture is in the understanding of whether the two distinctive size peaks resulted from some complex interplay between charge neutralisation (due to the presence of the g/mol PAA) and bridging flocculation (due to the presence of the g/ mol PAA), or simply from the uncoordinated individual action of the two polymer species present. Our floc size data from the order of mixing tests gives strong support to the latter. As was done in Fig. 7, floc size was also measured on the two samples flocculated by changing the order of PAA addition (as reported earlier in Fig. 6). Fig. 8 shows that if the g/mol PAA is added first

13 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) and followed by the g/mol PAA, the resultant floc size distribution is unimodal. In contrast, when this order of addition is reversed (with other conditions being identical), the distribution curve becomes bimodal. Importantly, the abscissa positions of the two peaks in the latter case coincide closely to those in the PAA mixture (cf. Fig. 7), where the two polyacid species were dosed as premixed. The close resemblance of the two curves in Figs. 7 and 8 (solid circles) suggests that when the and g/mol polyacids were dosed as premixed, the long-chain polyacid must have caused alumina flocculation ahead of the shorter-chain polymer. One can imagine that there was not enough g/mol molecules in the mixture to completely flocculate the alumina suspension. Thus the remaining particles must have been destabilised through suppression of electrostatic repulsion due to adsorption of the g/mol molecules. In an aggregation process involving adsorbing polymers, there are two important time scales: the polymer adsorption time (t A ) and the particle flocculation time (t F ). Our sizing results discussed above indicate that, in the premixed polymer system, the sum of t A and t F for the bridging flocculation process due to the g/mol PAA must be smaller than that for the aggregation process due to particle charge reduction through adsorption of the g/mol PAA. This can be qualitatively explained if one treats, as a first approximation, both the polymer adsorption and subsequent particle flocculation processes as simple bimolecular reactions. Under orthokinetic conditions as used in this study, the rate of collision, J ij, between i (alumina particle) and j (polyacid) species is given by (Elimelech et al., 1995): J ij ¼ k ij n i n j ; ð1þ where k ij is the collision rate constant and can be calculated from (ignoring hydrodynamic effects): k ij ¼ 4G 3 ða i þ a j Þ 3 : ð2þ In the above equations, n i and n j are the respective number concentrations of particle and polymer in solution. Their respective radii are a i and a j. G is the mean shear rate of mixing, which can be calculated from the power input per unit mass of fluid. Note that, by setting j equal to i, Eqs. (1) and (2) can also be used to describe the collision process between the alumina particles. As has been pointed out by Gregory (1987), for both polymer bridging and charge neutralisation mechanisms, it is usually necessary for a substantial fraction of the added polymer to be adsorbed before the particles are adequately destabilised. It has been shown that the time required to adsorb a fraction f of the initial polymer concentration can be calculated from (Elimelech et al., 1995): lnð1 f Þ t A ¼ : ð3þ k ij N i In our work the shear rate ( G) and size (a i ) and initial concentration (N i ) of the alumina particles were fixed. Therefore, according to Eqs. (2) and (3), any difference in the adsorption time (t A ) between the two PAA species in the polymer mixture must come from their different sizes a j (and hence k ij )orf values. Since the radius of gyration of the g/mol PAA in solution is expected to be much larger than that of the g/mol PAA, k ij must be bigger in the case of g/mol PAA. Furthermore, it may be reasonable to speculate that bridging flocculation may start at an f value smaller than that required by charge neutralisation. In the former case, multiple bond formation (and hence increased floc strength) between two particles may occur through attachment of the loops and tail of a single polymer chain. In contrast, occurrence of particle destabilisation through surface charge reduction will necessarily require a large proportion of the added polymer to be adsorbed. Therefore, in our premixed polymer system, contributions from k ij (and probably also f) may lead to a shorter adsorption time for the g/mol PAA molecules as compared to their g/mol counterparts. Note that the expected effect of polymer size on adsorption time has been observed before (Gregory, 1988). As for the particle flocculation step, the characteristic time, t F, can be estimated using the following expression (Elimelech et al., 1995): t F ¼ 1 k a N i ; ð4þ

14 174 Y.D. Yan et al. / Int. J. Miner. Process. 73 (2004) where k a is the particle flocculation rate constant. Because of the increased collision radius and reduced hydrodynamic interactions (Gregory, 1982, 1988; Sonntag and Strenge, 1987) and hence a larger flocculation rate constant k a, between the alumina particles partially covered with the g/mol PAA, bridging flocculation due to the g/mol PAA in the mixture can be expected to be faster than aggregation via charge reduction due to the g/mol PAA. There is also the likelihood of simultaneous attachment of the different segments of the same polymer chain onto more than one particle in a bridging process, which can further increase the flocculation kinetics. We must say that the above discussions are a very simplistic attempt at explaining the data; what had actually taken place must be a lot more complex than this. Nevertheless, this simplified picture does seem able to offer an explanation to the floc size distribution data as observed in Figs. 7 and 8 (solid circles). Finally, we compare the settling properties of flocs formed with the PAA mixture and PAA reference. These properties were characterised here in terms of the formation kinetics of the supernatant phase. Fig. 9 shows the change in the surpernatant phase thickness as a function of floc settling time. It can be seen that the main difference in the settling behaviour of the flocs formed from the two polymer samples was in the first 10 min, during which, flocs from the PAA mixture settled much faster than those from the PAA reference. This is to be expected considering the much bigger (average) floc size in the former (cf. Table 1 and Fig. 7). For both polymer systems, slight changes in the supernatant thickness observed at the later stages were simply due to compaction of the sediment bed. 4. Conclusions We have investigated the flocculation efficiency for positively charged alumina particles of a very polydisperse flocculant and compared it with that of a single-component reference polymer. This polydisperse polymer was synthesised by carefully premixing two poly(acrylic acids) of very different molecular weights. It was estimated that the mixed sample had an average molecular weight equal to that of the reference. The nature of the bimodal distribution in the molecular weight of the PAA mixture thus represents a somewhat extreme case for flocculant polydispersity. Under the experimental conditions used we found no synergistic flocculation effects for the alumina particles from the two polyacid species in the PAA mixture. The flocculation performance of the PAA mixture was, to a large extent, a simple reflection of that of the constituent PAA species. Nevertheless, compared to the reference, the PAA mixture resulted in: (a) a wider range for effective flocculation across the optimum dosage point, and (b) bigger flocs and hence faster sample clarification upon standing. Most interestingly, our observation that presence of a relatively small amount of a short-chain polyacid in the mixture greatly improved floc shear resistance as compared to the case of a single-component high molecular weight polyacid used alone may have significant implications for industrial applications. Acknowledgements Fig. 9. Comparison of formation kinetics of the supernatant phases of the alumina particle flocs formed using the PAA mixture and PAA reference, respectively. The data reported here was based on a fixed level (2%) of light transmission. Samples were the same ones as used in Fig. 7. The authors wish to thank the support of the Centre for Multiphase Processes, a Special Research Centre of the Australian Research Council, at the University of Newcastle.