- Email: [email protected]
Surfactant effects on the mechanism of particle capture in high-gradient magnetic filtration
Separation and Purification Technology 51 (2006) 201–209
Surfactant effects on the mechanism of particle capture in high-gradient magnetic filtration Costas Tsouris a,b,∗ , Jeremy Noonan b , Tung-yu Ying b,1 , Sotira Yiacoumi b a
b
Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, TN 37831-6181, USA Georgia Institute of Technology, 200 Bobby Dodd Way, Atlanta, GA 30332-0355, USA
Received 30 December 2005; received in revised form 1 February 2006; accepted 1 February 2006
Abstract This study investigates the effect of surfactants and key design parameters on the removal efficiency (RE) of paramagnetic colloidal particles by high-gradient magnetic filtration, and attempts to predict the effect of these parameters by a trajectory model. Magnetic filtration offers an advantage over conventional filtration in that it can achieve a reversible and selective separation. An aqueous suspension of paramagnetic colloidal ferric oxide (Fe2 O3 ) particles was treated with sodium dodecyl sulfate (SDS) and delivered through a column containing a stainless steel wool filter matrix, which was mounted between the poles of an electromagnet. The RE of the surfactant-treated particles was measured by analyzing effluent samples for Fe2 O3 concentration. The effect of the applied magnetic induction, fluid velocity, and radius of the stainless steel wires on the RE was tested and compared for both surfactant-treated and untreated particles. These three factors had a marked effect on the RE of surfactant-treated particles. An increase in applied magnetic induction from 0.2 to 0.5 T increased the RE from 79.9 to 93.4%, a decrease in wire radius from 49 to 15 m increased the RE from 60.2 to 93.4%, and a decrease in fluid velocity from 0.5 to 0.1 cm/s increased the RE from 69.5 to 95.3%. In the absence of a magnetic field (0 T), the RE was 10.8%. The predictions of the trajectory model agreed closely with these results. The same factors had a negligible effect on the RE of untreated particles. Over the range of all three parameters, the RE varied from 90 to 99%, but these variations were not statistically significant. In the absence of applied magnetic induction, the RE was 90.1%. These results differed markedly from the trajectory model predictions and demonstrated that nonmagnetic filtration mechanisms are primarily responsible for the capture of particles without SDS. Regeneration experiments indicated that the particles were captured in the primary minimum of the potential energy. On the other hand, these results showed that the magnetic filtration mechanism is primarily responsible for the removal of particles treated with SDS and that these particles are captured in the secondary minimum of the potential energy. Therefore, surfactant adsorption onto colloidal particles can potentially preserve and enhance the advantages of magnetic filtration (e.g., reversibility and selectivity). © 2006 Elsevier B.V. All rights reserved. Keywords: High-gradient magnetic separation; Surfactant; Colloidal particles; Iron oxide; Magnetic filtration
1. Introduction High-gradient magnetic filtration (HGMF) relies on a magnetic force between particles and filter collectors as a mechanism for capturing magnetic colloidal particles. Used widely in the steel and mineral processing industries, HGMF also has many potential wastewater-treatment applications. Researchers have shown HGMF to be effective for the removal of phosphates and sludge from water [1,2], the recovery of radionuclides [3],
∗
Corresponding author. Tel.: +1 865 241 3246; fax: +1 865 241 4829. E-mail address: [email protected] (C. Tsouris). 1 Present address: Los Alamos National Laboratory, MS J580, ESA-AET, Los Alamos, NM 87545, USA. 1383-5866/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.seppur.2006.02.001
and the separation of heavy metals [4]. Coupled with novel techniques like magnetic seeding, in which magnetic particles flocculate with nonmagnetic particles, and the use of functionalized particles, in which magnetic particles are designed to have a specific affinity for dissolved contaminants, HGMF may also be used to remove synthetic organic compounds [5], biomolecules [6], ferrihydrate [7], and heavy metals [8,9]. Indeed, HGMF holds promise for the removal of any paramagnetic micron-size particles. The advantages of HGMF over conventional filtration are that HGMF can achieve a highly efficient separation of colloid-size particles that is potentially both reversible and selective. These advantages are inherent in the magnetic force mechanism, which captures the particles onto the filter collectors. Particle capture by this mechanism is reversible because the magnetic force can
202
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
be easily eliminated by simply shutting off the magnetic field. Particle capture by this mechanism is also selective because it captures particles based on their responsiveness to the magnetic force, which is measured by a physical property of particles called magnetic susceptibility. Particles with high magnetic susceptibility will be more responsive to the magnetic force than particles with low magnetic susceptibility. These advantages of reversible and selective particle capture in an HGMF process require minimal influence from nonmagnetic filtration mechanisms. Nonmagnetic filtration mechanisms do not distinguish between nonmagnetic and magnetic particles, and, unlike the magnetic force, the forces that attract particles to the filter collectors cannot be easily removed. Thus, particle capture by nonmagnetic filtration mechanisms is neither selective nor reversible. The efficient use of reversible particle capture also requires that particle capture onto the filter collector take place in the secondary minimum of the potential energy between a suspended particle and the collector surface or between a suspended particle and a particle that has already been captured by the collector. Colloidal forces, such as the electrostatic and van der Waal’s forces employed in the Derjaguin–Landau–Verwey–Overbeek (DLVO) theory, determine the profile of potential energy versus distance. The van der Waal’s potential energy is attractive and is given a negative sign, while the electrostatic potential energy is repulsive and is given a positive sign (Fig. 1a). Addition of these potentials leads to a total energy potential with one of the
following features: (1) a single minimum at the origin (Fig. 1b), which means that the particles attract each other until their surfaces are in contact; (2) an energy maximum (Fig. 1c), which means that an energy barrier repels the particles as they approach one another; or (3) a secondary minimum away from the origin (Fig. 1d), where the two particles may be at equilibrium in the suspension. The magnetic potential is attractive and of longer range than the electrostatic and van der Waal’s potentials. Thus, when added to the potential energy of a system with an energy barrier like the one of Fig. 1c, the magnetic potential causes a secondary minimum to appear (Fig. 1d) [10]. The secondary minimum is strong enough to hold the magnetic particles together and thereby separate them from a suspension of magnetic and nonmagnetic particles. Once the magnetic field is removed, there is no attractive force to hold the particles together, and the shear the particles experience during filter regeneration easily removes the particles from the filter fibers. Thus, particle separation, as well as filter regeneration, is more effective when a potential energy barrier exists between particles in the absence of the magnetic field. These advantages of magnetic filtration can be realized when the particle size is small in relation to the porosity of the filter. Larger particles are more susceptible to other mechanisms, including interception. Because colloidal-size particles are typically too small to be removed efficiently by nonmagnetic filtration mechanisms, common practice is to first
Fig. 1. Potential energy vs. distance between approaching surfaces: (a) electrostatic and van der Waal’s potential energy, (b) total potential energy with primary minimum, (c) total potential energy with energy barrier, (d) total potential energy with secondary minimum.
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
destabilize particles by coagulation in order to promote particle aggregation before filtering. These destabilized aggregates are more likely to be captured by short-range van der Waal’s forces in the irreversible primary minimum. Thus, higher removal efficiencies can be achieved for larger particles, but it is more difficult to recover the particles from the filter when they are attached in the primary minimum. The same theory applies to magnetic filtration. When the goal of a magnetic filtration process is not only to achieve an high removal efficiency (RE) but also to selectively separate and recover the particles for reuse, aggregation prior to filtration is undesirable. Instead of destabilizing particles to promote aggregation, it would instead be advantageous to stabilize particles to prevent aggregation. One way to stabilize colloidal particles is by surfactant adsorption onto the surface of particles. Surfactants have been shown to stabilize magnetic particles. Chin et al. [11,12] showed that when a magnetic field is applied to a suspension of magnetic particles stabilized by an anionic surfactant, the particles aggregate in the secondary minimum. Consequently, the aggregates break up when the magnetic field is removed. This finding suggests that a surfactant might be useful for stabilizing paramagnetic colloidal particles in order to achieve a high RE by capturing particles in the reversible secondary minimum. The purpose of this study was to test whether adding surfactant to a suspension of colloidal particles prior to filtration would both minimize the capture of particles by nonmagnetic filtration mechanisms and influence the magnetic filtration mechanism to capture particles in the secondary minimum of the potential energy. The effect of three key design parameters – applied magnetic induction, wire collector radius, and fluid velocity – on the RE of ferric oxide (Fe2 O3 ) particles by HGMF with and without surfactant was measured. The implication is that if surfactants can prevent the capture of particles by nonmagnetic filtration mechanisms and in the primary minimum of the potential energy while still achieving a high RE, then surfactants could be used together with HGMF to achieve a reversible, selective separation of paramagnetic colloidal particles.
2. Modeling of HGMF In the present study, a trajectory model was developed to investigate the effects of experimental parameters (e.g., applied magnetic induction and fluid velocity) on the RE of HGMF. Many researchers [13–17] have used trajectory analysis to describe the interaction between a particle and a magnetic collector. By solving a force balance equation including external forces, such as magnetic and drag forces, and interparticle forces, such as van der Waal’s and inertial forces, the trajectory of a paramagnetic particle as it approaches a matrix element can be obtained [16]. One important objective of the trajectory analysis is to determine the limiting trajectory, which is the exact path that divides the approaching particle trajectories into those leading to particle capture by the collector and those passing. The distance between the limiting trajectory and the axis parallel to the flow direction is defined as the critical radius (Rc ), a value that can
be used to determine the RE of magnetic filtration [13]: −4(1 − ε)LRc Cout = 1 − exp RE = 1 − Cin 3πa
203
(1)
where L is the length of a filter matrix that is packed randomly with ferromagnetic wire at a packing fraction of ε; Cout and Cin the effluent and influent concentrations, respectively; and a is the wire radius. In the present system, which is under a low Reynolds number flow, the inertial force is much smaller than the other forces and can therefore be neglected. By considering only the external forces acting on a particle, the force balance equation can be written as: Fg + Fm + Fd = 0
(2)
where Fg , Fm , and Fd are the gravitational, magnetic, and drag forces, respectively. By solving the above force balance equation numerically, the critical radius (Rc ) as well as the RE of the magnetic filter can be obtained. However, RE obtained from the trajectory model is valid only for a clean filter. Among the forces (external and interparticle) involved in the system, the magnetic and drag forces are the most significant ones and also compete with one another. The performance of magnetic separation is, therefore, examined in terms of magnetic velocity (Vm ) and superficial velocity (V0 ) [13]: separation performance ∝
Vm V0
(3)
where Vm =
2b2 µ0 χMH0 9ηa
(4)
In Eq. (4), b is the particle radius, µ0 (4π × 10−7 H m−1 ) the permeability of free space, χ the difference between the volume magnetic susceptibilities of the particle and background carrier (e.g., water), H0 the strength of the applied magnetic field, M the magnetization value of the matrix material, and η is the dynamic viscosity of the fluid. The size and magnetic susceptibility of the particles are usually specific to a process and thus are not considered as control parameters. The separation performance is, therefore, affected by the applied magnetic field, matrix material, and superficial velocity. In the present study, the effects of these parameters on the RE of the HGMF are examined experimentally and theoretically. Comparisons between the experimental data and model results based on these three variables are also presented and discussed in a later section. The parameters and their values used in the trajectory model are listed in Table 1. The output of the model is a dimensionless RE defined as [(mass)in − (mass)out ]/(mass)in . 3. Materials and methods Filtration experiments were conducted with Fe2 O3 particles (EM Science, Gibbstown, NJ) and 430 stainless steel wool pads (Aquafine Corporation). The experimental apparatus consisted
204
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
Table 1 Parameters used in the magnetic filtration trajectory model Inputs
Symbol
Level(s)
Saturation magnetization (T) Filter length (m) Filter porosity (dimensionless) Particle radius (m) Volume magnetic susceptibility (dimensionless) Particle density (kg/m3 ) Fluid density (kg/m3 ) Dynamic viscosity (kg/m s)
Ms L ε b χp
0.6 0.05 0.956 0.25 0.000480
ρp ρf η
5240 997 0.001
The output of the model is a dimensionless removal efficiency (RE).
of five main components (Fig. 2): (1) a water-jacketed feed tank, (2) a Branson 2210 ultrasonic cleaner (Branson Ultrasonic, Danbury, CT), (3) a peristaltic pump (Rainin Instrument Co. Inc., Oakland, CA), (4) a bipolar electromagnet (Applied Magnetics Laboratory, Baltimore, MD), and (5) a magnetic filter. Under continuous sonication (to prevent aggregation and maintain a uniform particle size), the feed was pumped to the magnetic filter, across which a magnetic field was applied. Effluent samples were collected at periodic intervals throughout the duration of the experiment. After particles from these samples were digested in trace-metal-grade hydrochloric acid (Fisher Scientific), the iron concentration was measured using atomic absorption spectroscopy. 3.1. Feed preparation The feed was composed of Fe2 O3 particles (500 ppm) suspended in deionized water. The ionic strength was adjusted to 0.001N by adding NaCl. To increase particle stability and help prevent aggregation via electrostatic repulsion, the pH was raised to 9.5–10.0 by adding NaOH (0.1N). The zeta potential of the particles was measured using a Laser Zee Meter (PenKem Inc.). At an ionic strength of 0.001N and a pH of 9.5–10.0, the zeta potential of the Fe2 O3 particles used in this work was in the range of −40 to −50 mV. At this range of zeta potential, the electrostatic repulsion should make the particles stable with respect to aggregation. However, results showed that aggregation occurred. In order to avoid aggregation, we then added surfactant to the suspension. Three types of surfactants were used:
anionic, cationic, and nonionic. The results were very similar for all three types of surfactants. Therefore, we proceeded using only anionic surfactant, which was readily available in our laboratory. Thus, in the surfactant-adsorption experiments of this work, 10 mM of sodium dodecyl sulfate (SDS) was added to the feed. Before each run, the feed was premixed via 30 min of sonication. To keep the temperature of the feed from rising, cool faucet water was circulated continuously through the water jacket. This method held the feed temperature between 26 and 30 ◦ C. 3.2. Filter preparation To achieve a packing density of 4.4%, 1.36 g of steel wool was randomly and uniformly packed into a glass cylinder (1cm i.d.) and compressed to a length of 5 cm. The same packing density was used for each experiment. The steel wool filter was mounted in an aluminum spacer (0.75-in. thickness), which was secured between the two poles of the electromagnet. To displace air from the column and remove residual oil from the surface of the wool, the filter was wetted with 20 mL of a surfactant solution (SDS, 10 mM), and then rinsed with 100 mL of water. 3.3. Magnetic field generation The magnetic field strength delivered by the bipolar electromagnet was controlled by a Precision Bipolar Magnet Controller (Applied Magnetics Laboratory Inc., Baltimore, MD). The applied magnetic field strength was set by adjusting the electric current on the controller. The level of current was determined by an empirical correlation between current, field strength, and spacer thickness, provided by the manufacturer. 3.4. Effluent concentration measurements Effluent samples were collected at periodic intervals, the length of which depended on the total run time. From each sample, a volume of 1 or 2 mL aliquot was added to 3 mL of HCl and left to sit overnight. The acid completely dissolved the solid particles, leaving free iron in the solution. The concentration of iron was measured with an AAanalyst 800 atomic absorption spectrometer (Perkin-Elmer, Norwalk, CT). From these iron concentration measurements, the Fe2 O3 concentration in the effluent was deduced. 3.5. Design parameters and levels
Fig. 2. Schematic of magnetic filtration experimental setup: (1) water-jacketed feed tank, (2) ultrasonic cleaner, (3) peristaltic pump, (4) electromagnet, (5) steel wool filter, (6) effluent tank, and (7) effluent samples.
The design parameters tested were applied magnetic induction, fluid velocity, and wire diameter. The effect of the magnetic field strength was tested at 0.2 and 0.5 T; the effect of fluid velocity was tested at 0.1, 0.3, and 0.5 cm/s; and the effect of wire diameter was tested for ultrafine-, medium-, and course-grade steel wool (average diameters were 37, 72, and 98 m, respectively). These parameters were studied with and without SDS in the feed.
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209 Table 2 Effect of applied magnetic induction and wire radius on the RE of Fe2 O3 particles (results are for non-SDS-treated particles) Design parameter
Level
RE, % (experimental)
Applied magnetic induction (B0 )
0T 0.2 T 0.5 T
90.1 ± 9.9 96.6 ± 11.0 98.1 ± 11.0
49 m 36 m 15 m
89.1 ± 9.8 91.9 ± 10.1 98.1 ± 11.0
Wire radius (a)
205
Figs. 3 and 4 compare the effect of these same parameters on the RE of Fe2 O3 particles with SDS (10 mM) to that of particles without SDS. At 0 T, adding SDS to the feed causes the RE to plummet from 90.1 to 10.8%. At 0.2 T, the RE increases sharply to 79.9% and reaches as high as 93.4% at 0.5 T. Increasing the wire diameter has the opposite effect. The RE is the highest at 15 m. Increasing the wire radius to 36 and 48 m causes the RE to decrease to 71.4 and 60.2%, respectively. These results demonstrate that for particles treated with SDS, the magnetic filtration mechanism dominates particle capture. 4.2. The effect of SDS on the reversibility of particle capture
Table 2 shows the effect of the applied magnetic induction and wire diameter on the RE of Fe2 O3 particles without SDS. Neither parameter had a statistically significant effect on the RE. A 90% RE was achieved at 0 T. These results indicate that particle capture is due primarily to nonmagnetic filtration mechanisms.
Fig. 5 compares the effect of fluid velocity on the RE of Fe2 O3 particles without SDS with that of particles with SDS. Although fluid velocity does not have a statistically significant effect on the RE of bare particles, however, it does have a marked effect on the RE of particles with SDS. The highest RE (95.3%) was achieved at 0.1 cm/s. Increasing the fluid velocity to 0.3 cm/s led to a decrease in RE to 78.4%. A further increase to 0.5 cm/s did not have a significant effect on the RE. These results indicate that particles captured with SDS adhere more weakly to the wire collectors. By stabilizing the particles and making them adhere weakly to surfaces, SDS adsorption lowers both the collector and the collision efficiency of the filter and thereby essentially eliminates the impact of the nonmagnetic mechanism on the overall RE of this process. SDS also makes the particles less adherent to surfaces. Two observations support this claim. First, when particles are treated with SDS, they do not leave behind an orange-red residue on the surface of the glass feed tank or the Teflon tubing. Second, it is much easier to regenerate filters that contain SDS-treated particles by simply turning off the magnetic field. Samples of the fluid from the filtration column were collected after filtration experiments were performed with and without SDS, using the same parameters, and the Fe2 O3 concentration was measured. With SDS, the mass of Fe2 O3 in the filter fluid was 514 mg, representing 89% of the total mass of Fe2 O3 captured by the filter. Without SDS, the mass of Fe2 O3 was 138 mg,
Fig. 4. Effect of wire radius on the RE of Fe2 O3 particles with SDS and those without SDS (fluid velocity = 0.3 cm/s; applied magnetic induction = 0.5 T). Solid circles indicate no SDS; solid squares indicate 10 mM SDS.
Fig. 5. Effect of fluid velocity on the RE of Fe2 O3 particle with SDS and those without SDS (applied magnetic induction = 0.2 T; wire radius = 15 m). Solid circles indicate no SDS; solid squares indicate 10 mM SDS.
Fig. 3. Effect of applied magnetic induction on the RE of Fe2 O3 particles with SDS and those without SDS (fluid velocity = 0.3 cm/s; wire radius = 15 m). Solid circles indicate no SDS; solid squares indicate 10 mM SDS.
4. Results 4.1. The effect of SDS on the capture mechanism
206
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
representing just 31% of the total mass captured. Samples were collected and measured again after each column was sonicated for 1 min. Of the mass remaining in the filter, the amount recovered after sonication was three times larger for the SDS-treated suspensions than for the untreated suspensions. Previous research on the effect of SDS adsorption on the agglomeration and breakup of magnetic particles sheds light on why surfactant treatment results in superior recovery of particles [12]. SDS adsorption onto superparamagnetic particles was shown to cause reversible secondary minimum aggregation in the presence of a magnetic field. This was demonstrated experimentally by the breakup of aggregates after the field was removed. Similarly, in this process, SDS adsorption creates a secondary minimum potential energy between a Fe2 O3 particle and the steel wool collector and between agglomerated particles on the steel wool surface. Consequently, the particles break up and are readily removed when the field is removed from the filter. 4.3. Filtration breakthrough experiments The effects of applied magnetic induction and SDS treatment on filter breakthrough were also tested in this study. The results are displayed in Figs. 6 and 7. The values for Cout * represent the effluent concentration normalized to the feed concentration. Fig. 6 shows the effects of pH and SDS treatment. It is clear that results from SDS-treated suspensions are very similar to those at pH 10.8, indicating that at these conditions, particle aggregation is minimized. On the other hand, at pH 9.4, the breakthrough curve is much different, because of aggregation occurring despite the relatively high zeta potential (∼−40 mV) at this pH. Fig. 7 shows that the applied magnetic induction not only improves the RE of the filter in the beginning stage of filtration but also increases its loading capacity. The latter parameter is germane to the overall filter performance because it determines how often the filter must be regenerated. At 0.2 T, the magnetic force is not sufficiently strong enough to overcome the interparticle repulsive forces created by electrostatic and steric interactions. Consequently, the rate of particle capture decreases
Fig. 6. Effect of SDS treatment on filter breakthrough (B0 = 0.2 T; V0 = 0.003 cm/s; a = 15 m). Cout * = effluent concentration normalized to feed concentration. Feed concentration = 500 ppm.
Fig. 7. Effect of applied magnetic induction on filter breakthrough (V0 = 0.003 cm/s; a = 15 m). Cout * = effluent concentration normalized to feed concentration. Feed concentration = 500 ppm.
sharply as particles build up on the surface of the collectors. However, at 0.8 T, the magnetic force remains strong enough to dominate these forces even as more particles are deposited on the surface. Consequently, the rate of particle capture remains steadier throughout the course of the experiment. 4.4. Comparison of trajectory model predictions with experimental results A comparison of the effects of applied magnetic field, radius of the ferromagnetic wire in the filter, and fluid velocity on the RE shows a good agreement between modeling and experimental results. Figs. 8–10 show that the model predicts well the experimental trends. Furthermore, the figures show that at each level of all three parameters, the RE predicted by the model lies within the experimental error of the corresponding experimental RE. That these experimental results closely parallel the trends predicted by the trajectory model further substantiates the claim that
Fig. 8. Comparison of trajectory model predictions () and experimental results () for the effect of applied magnetic induction on RE of Fe2 O3 particles with SDS (wire radius = 15 m; fluid velocity = 0.3 cm/s). Values for other parameters are provided in Table 1.
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
207
fluid velocity, V0 , is also observed. The model again predicts a remarkably similar response to changes in the wire radius and fluid velocity (Figs. 9 and 10). It is noteworthy that the trajectory model works well, even though it does not consider the repulsive forces between the particle and collector caused by electrostatic and steric interactions. Although these forces are related to particle removal by a nonmagnetic mechanism, they are overwhelmed by the magnetic force and thus present negligible opposition to capture by this mechanism. 5. Discussion
Fig. 9. Comparison of trajectory model predictions () and experimental results () for the effect of wire radius on RE of Fe2 O3 particles with SDS (magnetic induction = 0.5 T; fluid velocity = 0.3 cm/s). Values for other parameters are provided in Table 1.
the removal of SDS-treated particles is due primarily to the operation of the magnetic mechanism. If the experimental filtration mechanism corresponds to the filtration mechanism simulated by the model, then the sensitivity of the experimental process and the sensitivity of the model to changes in the design parameters should be similar. The impact of changes in the levels of the three design parameters on experimental and modeling results shows a remarkable consistency between model and experimental sensitivity to an increase in B0 from 0.2 to 0.5 T (Fig. 8). A somewhat greater sensitivity of the model over the range of 0–0.2 T can be explained by the approximately 10% removal of particles due to nonmagnetic filtration. Thus, in the experimental system, magnetic filtration is the dominant, but not the sole, mechanism for particle removal. A close similarity between the model and experimental sensitivity to the wire radius, a, and
Fig. 10. Comparison of trajectory model predictions () and experimental results () for the effect of fluid velocity on RE of Fe2 O3 particles with SDS (magnetic induction = 0.2 T; wire radius = 15 m). Values for other parameters are provided in Table 1.
These findings have important implications for the use of HGMS to remove paramagnetic particles from water or from a suspension of a mixture of particles. First, if the particles targeted for removal have properties similar to those of Fe2 O3 particles with respect to aggregation and attraction to surfaces, then HGMS may not be the best technique for separation. If 90% RE is achieved by a filter with such a high porosity (96.5%), then nonmagnetic filters with a lower porosity (e.g., granular filters) would perform even better. These filters would probably require less operating costs than an HGMS process. Second, HGMS could be useful for separating paramagnetic particles via magnetic filtration from diamagnetic particles (particles with very low and negative magnetic susceptibility). To achieve this separation, it would be necessary to minimize the impact of nonmagnetic filtration mechanisms to ensure that the magnetic filtration mechanism determines particle capture. This can be achieved by increasing particle stability. This study showed that SDS treatment stabilizes the particles and, overall, results in lower RE. Thus, although such a treatment provides more favorable comparisons with the trajectory model, adding SDS would offer an advantage in a separation process that aims solely to remove these particles from wastewater. However, if the goal of the process is not only to remove the particles from water but also to isolate them for the purpose of recycling and reuse, then SDS treatment coupled with HGMS is a promising strategy. If the wastewater contains diamagnetic particles, dispersing and stabilizing these particles with SDS would decrease their removal by ferromagnetic filter wires via nonmagnetic mechanisms. However, the paramagnetic particles, also stabilized, would be retained by magnetic filtration under a high applied magnetic field. Third, optimization of an HGMS process must account for these three design factors. To achieve the maximum RE, the applied magnetic induction and wire thickness must be varied to achieve the strongest magnetic force, and the flow velocity must be varied to achieve the weakest drag force. However, in an industrial process, other considerations must be taken into account. A filtration process must be capable of handling sizable throughput. To achieve a high throughput, it is necessary to increase the flow rate, the filter size, and possibly the fluid velocity. Operating cost is another important consideration. Generating a strong magnetic field requires electrical power. Achieving high throughput from a magnetic filter would require a stronger magnetic field and thus higher
208
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
operating costs. However, lowering the throughput would also raise costs. Therefore, these factors must be optimized. Fourth, SDS treatment could be useful for recovering particles from a magnetic filter. In industrial wastewater-treatment operations, the contaminants are often valuable materials that could be reused in the process. Therefore, it is important to be able to recover the contaminant from the filter. When filtration is due primarily to nonmagnetic mechanisms, the task of recovering particles from the filter is more difficult than when filtration is due primarily to a magnetic mechanism. This phenomenon occurs because it is more difficult to relax or reverse the attractive forces that are responsible for particle attachment. However, relaxing the magnetic force is a simple matter of turning off the electromagnet. If the magnetic force is primarily responsible for particle attachment, then the particles will more readily detach when the force is removed. 6. Conclusions A nonmagnetic filtration mechanism is primarily responsible for the removal of bare Fe2 O3 particles from water by a stainless steel filter, obscuring the effect on the filter RE of the parameters that determine the magnitude of the forces comprising the magnetic filtration mechanism. Conversely, the magnetic filtration mechanism is primarily responsible for the removal of SDS-treated Fe2 O3 particles. Adding SDS to the feed effectively disables the nonmagnetic mechanism by increasing particle stability via a steric repulsive force. By dampening the attractive forces behind the nonmagnetic mechanism, SDS treatment accentuates the effects of these same parameters on the RE, allowing for a meaningful comparison with the predictions of a trajectory model that simulates the magnetic removal mechanism. The reliability of a trajectory model depends on including the relevant forces in the right form. This model inadequately predicts the effect of the parameters on the removal of bare Fe2 O3 particles because it does not incorporate the appropriate nonmagnetic attractive forces. However, the model predictions agree nicely with the experimental data for the removal of SDS-treated particles because the only significant attractive force is the magnetic force, which the model simulates well. Better agreement with the predictions of a trajectory model does not justify the use of SDS in an HGMS process. If the performance of HGMS is measured by its RE, then SDS treatment is a disadvantage. However, SDS treatment can potentially improve other aspects of HGMS performance. SDS treatment allows for better recovery of particles from the filter when the magnetic field is disabled, making filter regeneration much easier. SDS treatment also allows for selective separation of nonmagnetic particles from magnetic particles. Stabilized nonmagnetic particles would escape capture in a magnetic filter, while stable magnetic particles would be retained. Thus, SDS treatment has the potential to preserve two chief advantages of HGMS: the reversibility and the selectivity of the magnetic force. Future studies should focus on testing these advantages of surfactant treatment. To study how well surfactants increase
the selectivity of an HGMS process, magnetic filtration experiments should be performed that measure the effect of surfactant treatment on the removal of Fe2 O3 particles and a nonmagnetic colloid, like kaolin, in the same suspension. Additional studies should examine the reversibility of the separation at higher magnetic field strengths and with different filter media. If the magnetic force is too high, the separation may not be as easily reversible. These studies could identify the optimal applied magnetic induction at which the attached particles remain in the domain of the secondary minimum potential energy, where they are easier to detach. Also of interest would be a more thorough investigation of the effect of different types of surfactants on particle removal. Using SDS in a large-scale wastewater-treatment process might be infeasible economically and undesirable environmentally. Therefore, less expensive, more environmentally friendly surfactants should be studied. Finally, in order to minimize the cost of the surfactant, the effect of surfactant concentration on reversibility and selectivity should be investigated to find an optimal concentration. This problem would also require a better understanding of the mechanism of surfactant adsorption. Acknowledgments Partial support for this work was provided by the Office of Basic Energy Sciences, Division of Chemical Sciences, U.S. Department of Energy, under contract DE-AC05-00OR22725 with UT-Battelle, LLC, and by Georgia Institute of Technology. The authors are also thankful to Dr. Marsha Savage for editing the manuscript. References [1] A.M.H. Shaikh, S.G. Dixit, Removal of phosphate from waters by precipitation and high gradient magnetic separation, Water Res. 26 (1992) 845–852. [2] E. Barrado, F. Prieto, J. Ribas, F.A. Lopez, Magnetic separation of ferrite sludge from a wastewater purification process, Water Air Soil Pollut. 115 (1999) 385–394. [3] A.S. Bahaj, I.W. Croudace, P.A.B. James, F.D. Moeschler, P.E. Warwick, Continuous radionuclide recovery from wastewater using magnetotactic bacteria, J. Magn. Magn. Mater. 184 (1998) 241–244. [4] P. Anand, J.E. Etzel, F.J. Friedlaender, Heavy metals removal by high gradient magnetic separation, IEEE Trans. Magn. MAG-21 (1985) 2062– 2064. [5] G.D. Moeser, K.A. Roach, W.H. Green, P.E. Laibinis, T.A. Hatton, Waterbased magnetic fluids as extractants for synthetic organic compounds, Ind. Eng. Chem. Res. 41 (2002) 739–749. [6] S. Bucak, D.A. Jones, P.E. Laibinis, T.A. Hatton, Protein separations using colloidal magnetic nanoparticles, Biotechnol. Prog. 19 (2003) 477– 484. [7] N. Karapinar, Magnetic separation of ferrihydrite from wastewater by magnetic seeding and high-gradient magnetic separation, Int. J. Miner. Process. 71 (2003) 45–54. [8] P. Phanapavudhikul, J.A. Waters, E.S.P. de Ortiz, Design and performance of magnetic composite particles for the separation of heavy metals from water, J. Environ. Sci. Health-Part A Toxic/Hazard. Subst. Environ. Eng. 38 (2003) 2277–2285. [9] M.D. Kaminski, L. Nunez, Extractant-coated magnetic particles for cobalt and nickel recovery from acidic solution, J. Magn. Magn. Mater. 194 (1999) 31–36.
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209 [10] C. Tsouris, T.C. Scott, Flocculation of paramagnetic particles in a magnetic field, J. Colloid Interf. Sci. 171 (1995) 319–330. [11] C.J. Chin, S. Yiacoumi, C. Tsouris, S. Relle, S.B. Grant, Secondary minimum aggregation of superparamagnetic colloidal particles, Langmuir 16 (2000) 3641–3650. [12] C.J. Chin, S. Yiacoumi, C. Tsouris, Agglomeration of magnetic particles and breakup of magnetic chains in magnetorheological fluids, Colloids Surf. A: Physicochem. Eng. Aspects 204 (2002) 63–72. [13] J.H.P. Watson, Magnetic filtration, J. Appl. Phys. 44 (1973) 4209.
209
[14] W.F. Lawson Jr., W.H. Simons, R.P. Treat, The dynamics of a particle attracted by a magnetized wire, J. Appl. Phys. 48 (1977) 3213. [15] H. Schewe, M. Takayasu, F.J. Friedlaender, Observation of particle trajectories in an HGMS single-wire system, IEEE Trans. Magn. MAG-16 (1980) 149. [16] T.-Y. Ying, S. Yiacoumi, C. Tsouris, High-gradient magnetically seeded filtration, Chem. Eng. Sci. 55 (2000) 1101–1113. [17] J. Svoboda, Magnetic Methods for the Treatment of Minerals, Elsevier, New York, 1987.
Surfactant effects on the mechanism of particle capture in high-gradient magnetic filtration Costas Tsouris a,b,∗ , Jeremy Noonan b , Tung-yu Ying b,1 , Sotira Yiacoumi b a
b
Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, TN 37831-6181, USA Georgia Institute of Technology, 200 Bobby Dodd Way, Atlanta, GA 30332-0355, USA
Received 30 December 2005; received in revised form 1 February 2006; accepted 1 February 2006
Abstract This study investigates the effect of surfactants and key design parameters on the removal efficiency (RE) of paramagnetic colloidal particles by high-gradient magnetic filtration, and attempts to predict the effect of these parameters by a trajectory model. Magnetic filtration offers an advantage over conventional filtration in that it can achieve a reversible and selective separation. An aqueous suspension of paramagnetic colloidal ferric oxide (Fe2 O3 ) particles was treated with sodium dodecyl sulfate (SDS) and delivered through a column containing a stainless steel wool filter matrix, which was mounted between the poles of an electromagnet. The RE of the surfactant-treated particles was measured by analyzing effluent samples for Fe2 O3 concentration. The effect of the applied magnetic induction, fluid velocity, and radius of the stainless steel wires on the RE was tested and compared for both surfactant-treated and untreated particles. These three factors had a marked effect on the RE of surfactant-treated particles. An increase in applied magnetic induction from 0.2 to 0.5 T increased the RE from 79.9 to 93.4%, a decrease in wire radius from 49 to 15 m increased the RE from 60.2 to 93.4%, and a decrease in fluid velocity from 0.5 to 0.1 cm/s increased the RE from 69.5 to 95.3%. In the absence of a magnetic field (0 T), the RE was 10.8%. The predictions of the trajectory model agreed closely with these results. The same factors had a negligible effect on the RE of untreated particles. Over the range of all three parameters, the RE varied from 90 to 99%, but these variations were not statistically significant. In the absence of applied magnetic induction, the RE was 90.1%. These results differed markedly from the trajectory model predictions and demonstrated that nonmagnetic filtration mechanisms are primarily responsible for the capture of particles without SDS. Regeneration experiments indicated that the particles were captured in the primary minimum of the potential energy. On the other hand, these results showed that the magnetic filtration mechanism is primarily responsible for the removal of particles treated with SDS and that these particles are captured in the secondary minimum of the potential energy. Therefore, surfactant adsorption onto colloidal particles can potentially preserve and enhance the advantages of magnetic filtration (e.g., reversibility and selectivity). © 2006 Elsevier B.V. All rights reserved. Keywords: High-gradient magnetic separation; Surfactant; Colloidal particles; Iron oxide; Magnetic filtration
1. Introduction High-gradient magnetic filtration (HGMF) relies on a magnetic force between particles and filter collectors as a mechanism for capturing magnetic colloidal particles. Used widely in the steel and mineral processing industries, HGMF also has many potential wastewater-treatment applications. Researchers have shown HGMF to be effective for the removal of phosphates and sludge from water [1,2], the recovery of radionuclides [3],
∗
Corresponding author. Tel.: +1 865 241 3246; fax: +1 865 241 4829. E-mail address: [email protected] (C. Tsouris). 1 Present address: Los Alamos National Laboratory, MS J580, ESA-AET, Los Alamos, NM 87545, USA. 1383-5866/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.seppur.2006.02.001
and the separation of heavy metals [4]. Coupled with novel techniques like magnetic seeding, in which magnetic particles flocculate with nonmagnetic particles, and the use of functionalized particles, in which magnetic particles are designed to have a specific affinity for dissolved contaminants, HGMF may also be used to remove synthetic organic compounds [5], biomolecules [6], ferrihydrate [7], and heavy metals [8,9]. Indeed, HGMF holds promise for the removal of any paramagnetic micron-size particles. The advantages of HGMF over conventional filtration are that HGMF can achieve a highly efficient separation of colloid-size particles that is potentially both reversible and selective. These advantages are inherent in the magnetic force mechanism, which captures the particles onto the filter collectors. Particle capture by this mechanism is reversible because the magnetic force can
202
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
be easily eliminated by simply shutting off the magnetic field. Particle capture by this mechanism is also selective because it captures particles based on their responsiveness to the magnetic force, which is measured by a physical property of particles called magnetic susceptibility. Particles with high magnetic susceptibility will be more responsive to the magnetic force than particles with low magnetic susceptibility. These advantages of reversible and selective particle capture in an HGMF process require minimal influence from nonmagnetic filtration mechanisms. Nonmagnetic filtration mechanisms do not distinguish between nonmagnetic and magnetic particles, and, unlike the magnetic force, the forces that attract particles to the filter collectors cannot be easily removed. Thus, particle capture by nonmagnetic filtration mechanisms is neither selective nor reversible. The efficient use of reversible particle capture also requires that particle capture onto the filter collector take place in the secondary minimum of the potential energy between a suspended particle and the collector surface or between a suspended particle and a particle that has already been captured by the collector. Colloidal forces, such as the electrostatic and van der Waal’s forces employed in the Derjaguin–Landau–Verwey–Overbeek (DLVO) theory, determine the profile of potential energy versus distance. The van der Waal’s potential energy is attractive and is given a negative sign, while the electrostatic potential energy is repulsive and is given a positive sign (Fig. 1a). Addition of these potentials leads to a total energy potential with one of the
following features: (1) a single minimum at the origin (Fig. 1b), which means that the particles attract each other until their surfaces are in contact; (2) an energy maximum (Fig. 1c), which means that an energy barrier repels the particles as they approach one another; or (3) a secondary minimum away from the origin (Fig. 1d), where the two particles may be at equilibrium in the suspension. The magnetic potential is attractive and of longer range than the electrostatic and van der Waal’s potentials. Thus, when added to the potential energy of a system with an energy barrier like the one of Fig. 1c, the magnetic potential causes a secondary minimum to appear (Fig. 1d) [10]. The secondary minimum is strong enough to hold the magnetic particles together and thereby separate them from a suspension of magnetic and nonmagnetic particles. Once the magnetic field is removed, there is no attractive force to hold the particles together, and the shear the particles experience during filter regeneration easily removes the particles from the filter fibers. Thus, particle separation, as well as filter regeneration, is more effective when a potential energy barrier exists between particles in the absence of the magnetic field. These advantages of magnetic filtration can be realized when the particle size is small in relation to the porosity of the filter. Larger particles are more susceptible to other mechanisms, including interception. Because colloidal-size particles are typically too small to be removed efficiently by nonmagnetic filtration mechanisms, common practice is to first
Fig. 1. Potential energy vs. distance between approaching surfaces: (a) electrostatic and van der Waal’s potential energy, (b) total potential energy with primary minimum, (c) total potential energy with energy barrier, (d) total potential energy with secondary minimum.
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
destabilize particles by coagulation in order to promote particle aggregation before filtering. These destabilized aggregates are more likely to be captured by short-range van der Waal’s forces in the irreversible primary minimum. Thus, higher removal efficiencies can be achieved for larger particles, but it is more difficult to recover the particles from the filter when they are attached in the primary minimum. The same theory applies to magnetic filtration. When the goal of a magnetic filtration process is not only to achieve an high removal efficiency (RE) but also to selectively separate and recover the particles for reuse, aggregation prior to filtration is undesirable. Instead of destabilizing particles to promote aggregation, it would instead be advantageous to stabilize particles to prevent aggregation. One way to stabilize colloidal particles is by surfactant adsorption onto the surface of particles. Surfactants have been shown to stabilize magnetic particles. Chin et al. [11,12] showed that when a magnetic field is applied to a suspension of magnetic particles stabilized by an anionic surfactant, the particles aggregate in the secondary minimum. Consequently, the aggregates break up when the magnetic field is removed. This finding suggests that a surfactant might be useful for stabilizing paramagnetic colloidal particles in order to achieve a high RE by capturing particles in the reversible secondary minimum. The purpose of this study was to test whether adding surfactant to a suspension of colloidal particles prior to filtration would both minimize the capture of particles by nonmagnetic filtration mechanisms and influence the magnetic filtration mechanism to capture particles in the secondary minimum of the potential energy. The effect of three key design parameters – applied magnetic induction, wire collector radius, and fluid velocity – on the RE of ferric oxide (Fe2 O3 ) particles by HGMF with and without surfactant was measured. The implication is that if surfactants can prevent the capture of particles by nonmagnetic filtration mechanisms and in the primary minimum of the potential energy while still achieving a high RE, then surfactants could be used together with HGMF to achieve a reversible, selective separation of paramagnetic colloidal particles.
2. Modeling of HGMF In the present study, a trajectory model was developed to investigate the effects of experimental parameters (e.g., applied magnetic induction and fluid velocity) on the RE of HGMF. Many researchers [13–17] have used trajectory analysis to describe the interaction between a particle and a magnetic collector. By solving a force balance equation including external forces, such as magnetic and drag forces, and interparticle forces, such as van der Waal’s and inertial forces, the trajectory of a paramagnetic particle as it approaches a matrix element can be obtained [16]. One important objective of the trajectory analysis is to determine the limiting trajectory, which is the exact path that divides the approaching particle trajectories into those leading to particle capture by the collector and those passing. The distance between the limiting trajectory and the axis parallel to the flow direction is defined as the critical radius (Rc ), a value that can
be used to determine the RE of magnetic filtration [13]: −4(1 − ε)LRc Cout = 1 − exp RE = 1 − Cin 3πa
203
(1)
where L is the length of a filter matrix that is packed randomly with ferromagnetic wire at a packing fraction of ε; Cout and Cin the effluent and influent concentrations, respectively; and a is the wire radius. In the present system, which is under a low Reynolds number flow, the inertial force is much smaller than the other forces and can therefore be neglected. By considering only the external forces acting on a particle, the force balance equation can be written as: Fg + Fm + Fd = 0
(2)
where Fg , Fm , and Fd are the gravitational, magnetic, and drag forces, respectively. By solving the above force balance equation numerically, the critical radius (Rc ) as well as the RE of the magnetic filter can be obtained. However, RE obtained from the trajectory model is valid only for a clean filter. Among the forces (external and interparticle) involved in the system, the magnetic and drag forces are the most significant ones and also compete with one another. The performance of magnetic separation is, therefore, examined in terms of magnetic velocity (Vm ) and superficial velocity (V0 ) [13]: separation performance ∝
Vm V0
(3)
where Vm =
2b2 µ0 χMH0 9ηa
(4)
In Eq. (4), b is the particle radius, µ0 (4π × 10−7 H m−1 ) the permeability of free space, χ the difference between the volume magnetic susceptibilities of the particle and background carrier (e.g., water), H0 the strength of the applied magnetic field, M the magnetization value of the matrix material, and η is the dynamic viscosity of the fluid. The size and magnetic susceptibility of the particles are usually specific to a process and thus are not considered as control parameters. The separation performance is, therefore, affected by the applied magnetic field, matrix material, and superficial velocity. In the present study, the effects of these parameters on the RE of the HGMF are examined experimentally and theoretically. Comparisons between the experimental data and model results based on these three variables are also presented and discussed in a later section. The parameters and their values used in the trajectory model are listed in Table 1. The output of the model is a dimensionless RE defined as [(mass)in − (mass)out ]/(mass)in . 3. Materials and methods Filtration experiments were conducted with Fe2 O3 particles (EM Science, Gibbstown, NJ) and 430 stainless steel wool pads (Aquafine Corporation). The experimental apparatus consisted
204
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
Table 1 Parameters used in the magnetic filtration trajectory model Inputs
Symbol
Level(s)
Saturation magnetization (T) Filter length (m) Filter porosity (dimensionless) Particle radius (m) Volume magnetic susceptibility (dimensionless) Particle density (kg/m3 ) Fluid density (kg/m3 ) Dynamic viscosity (kg/m s)
Ms L ε b χp
0.6 0.05 0.956 0.25 0.000480
ρp ρf η
5240 997 0.001
The output of the model is a dimensionless removal efficiency (RE).
of five main components (Fig. 2): (1) a water-jacketed feed tank, (2) a Branson 2210 ultrasonic cleaner (Branson Ultrasonic, Danbury, CT), (3) a peristaltic pump (Rainin Instrument Co. Inc., Oakland, CA), (4) a bipolar electromagnet (Applied Magnetics Laboratory, Baltimore, MD), and (5) a magnetic filter. Under continuous sonication (to prevent aggregation and maintain a uniform particle size), the feed was pumped to the magnetic filter, across which a magnetic field was applied. Effluent samples were collected at periodic intervals throughout the duration of the experiment. After particles from these samples were digested in trace-metal-grade hydrochloric acid (Fisher Scientific), the iron concentration was measured using atomic absorption spectroscopy. 3.1. Feed preparation The feed was composed of Fe2 O3 particles (500 ppm) suspended in deionized water. The ionic strength was adjusted to 0.001N by adding NaCl. To increase particle stability and help prevent aggregation via electrostatic repulsion, the pH was raised to 9.5–10.0 by adding NaOH (0.1N). The zeta potential of the particles was measured using a Laser Zee Meter (PenKem Inc.). At an ionic strength of 0.001N and a pH of 9.5–10.0, the zeta potential of the Fe2 O3 particles used in this work was in the range of −40 to −50 mV. At this range of zeta potential, the electrostatic repulsion should make the particles stable with respect to aggregation. However, results showed that aggregation occurred. In order to avoid aggregation, we then added surfactant to the suspension. Three types of surfactants were used:
anionic, cationic, and nonionic. The results were very similar for all three types of surfactants. Therefore, we proceeded using only anionic surfactant, which was readily available in our laboratory. Thus, in the surfactant-adsorption experiments of this work, 10 mM of sodium dodecyl sulfate (SDS) was added to the feed. Before each run, the feed was premixed via 30 min of sonication. To keep the temperature of the feed from rising, cool faucet water was circulated continuously through the water jacket. This method held the feed temperature between 26 and 30 ◦ C. 3.2. Filter preparation To achieve a packing density of 4.4%, 1.36 g of steel wool was randomly and uniformly packed into a glass cylinder (1cm i.d.) and compressed to a length of 5 cm. The same packing density was used for each experiment. The steel wool filter was mounted in an aluminum spacer (0.75-in. thickness), which was secured between the two poles of the electromagnet. To displace air from the column and remove residual oil from the surface of the wool, the filter was wetted with 20 mL of a surfactant solution (SDS, 10 mM), and then rinsed with 100 mL of water. 3.3. Magnetic field generation The magnetic field strength delivered by the bipolar electromagnet was controlled by a Precision Bipolar Magnet Controller (Applied Magnetics Laboratory Inc., Baltimore, MD). The applied magnetic field strength was set by adjusting the electric current on the controller. The level of current was determined by an empirical correlation between current, field strength, and spacer thickness, provided by the manufacturer. 3.4. Effluent concentration measurements Effluent samples were collected at periodic intervals, the length of which depended on the total run time. From each sample, a volume of 1 or 2 mL aliquot was added to 3 mL of HCl and left to sit overnight. The acid completely dissolved the solid particles, leaving free iron in the solution. The concentration of iron was measured with an AAanalyst 800 atomic absorption spectrometer (Perkin-Elmer, Norwalk, CT). From these iron concentration measurements, the Fe2 O3 concentration in the effluent was deduced. 3.5. Design parameters and levels
Fig. 2. Schematic of magnetic filtration experimental setup: (1) water-jacketed feed tank, (2) ultrasonic cleaner, (3) peristaltic pump, (4) electromagnet, (5) steel wool filter, (6) effluent tank, and (7) effluent samples.
The design parameters tested were applied magnetic induction, fluid velocity, and wire diameter. The effect of the magnetic field strength was tested at 0.2 and 0.5 T; the effect of fluid velocity was tested at 0.1, 0.3, and 0.5 cm/s; and the effect of wire diameter was tested for ultrafine-, medium-, and course-grade steel wool (average diameters were 37, 72, and 98 m, respectively). These parameters were studied with and without SDS in the feed.
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209 Table 2 Effect of applied magnetic induction and wire radius on the RE of Fe2 O3 particles (results are for non-SDS-treated particles) Design parameter
Level
RE, % (experimental)
Applied magnetic induction (B0 )
0T 0.2 T 0.5 T
90.1 ± 9.9 96.6 ± 11.0 98.1 ± 11.0
49 m 36 m 15 m
89.1 ± 9.8 91.9 ± 10.1 98.1 ± 11.0
Wire radius (a)
205
Figs. 3 and 4 compare the effect of these same parameters on the RE of Fe2 O3 particles with SDS (10 mM) to that of particles without SDS. At 0 T, adding SDS to the feed causes the RE to plummet from 90.1 to 10.8%. At 0.2 T, the RE increases sharply to 79.9% and reaches as high as 93.4% at 0.5 T. Increasing the wire diameter has the opposite effect. The RE is the highest at 15 m. Increasing the wire radius to 36 and 48 m causes the RE to decrease to 71.4 and 60.2%, respectively. These results demonstrate that for particles treated with SDS, the magnetic filtration mechanism dominates particle capture. 4.2. The effect of SDS on the reversibility of particle capture
Table 2 shows the effect of the applied magnetic induction and wire diameter on the RE of Fe2 O3 particles without SDS. Neither parameter had a statistically significant effect on the RE. A 90% RE was achieved at 0 T. These results indicate that particle capture is due primarily to nonmagnetic filtration mechanisms.
Fig. 5 compares the effect of fluid velocity on the RE of Fe2 O3 particles without SDS with that of particles with SDS. Although fluid velocity does not have a statistically significant effect on the RE of bare particles, however, it does have a marked effect on the RE of particles with SDS. The highest RE (95.3%) was achieved at 0.1 cm/s. Increasing the fluid velocity to 0.3 cm/s led to a decrease in RE to 78.4%. A further increase to 0.5 cm/s did not have a significant effect on the RE. These results indicate that particles captured with SDS adhere more weakly to the wire collectors. By stabilizing the particles and making them adhere weakly to surfaces, SDS adsorption lowers both the collector and the collision efficiency of the filter and thereby essentially eliminates the impact of the nonmagnetic mechanism on the overall RE of this process. SDS also makes the particles less adherent to surfaces. Two observations support this claim. First, when particles are treated with SDS, they do not leave behind an orange-red residue on the surface of the glass feed tank or the Teflon tubing. Second, it is much easier to regenerate filters that contain SDS-treated particles by simply turning off the magnetic field. Samples of the fluid from the filtration column were collected after filtration experiments were performed with and without SDS, using the same parameters, and the Fe2 O3 concentration was measured. With SDS, the mass of Fe2 O3 in the filter fluid was 514 mg, representing 89% of the total mass of Fe2 O3 captured by the filter. Without SDS, the mass of Fe2 O3 was 138 mg,
Fig. 4. Effect of wire radius on the RE of Fe2 O3 particles with SDS and those without SDS (fluid velocity = 0.3 cm/s; applied magnetic induction = 0.5 T). Solid circles indicate no SDS; solid squares indicate 10 mM SDS.
Fig. 5. Effect of fluid velocity on the RE of Fe2 O3 particle with SDS and those without SDS (applied magnetic induction = 0.2 T; wire radius = 15 m). Solid circles indicate no SDS; solid squares indicate 10 mM SDS.
Fig. 3. Effect of applied magnetic induction on the RE of Fe2 O3 particles with SDS and those without SDS (fluid velocity = 0.3 cm/s; wire radius = 15 m). Solid circles indicate no SDS; solid squares indicate 10 mM SDS.
4. Results 4.1. The effect of SDS on the capture mechanism
206
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
representing just 31% of the total mass captured. Samples were collected and measured again after each column was sonicated for 1 min. Of the mass remaining in the filter, the amount recovered after sonication was three times larger for the SDS-treated suspensions than for the untreated suspensions. Previous research on the effect of SDS adsorption on the agglomeration and breakup of magnetic particles sheds light on why surfactant treatment results in superior recovery of particles [12]. SDS adsorption onto superparamagnetic particles was shown to cause reversible secondary minimum aggregation in the presence of a magnetic field. This was demonstrated experimentally by the breakup of aggregates after the field was removed. Similarly, in this process, SDS adsorption creates a secondary minimum potential energy between a Fe2 O3 particle and the steel wool collector and between agglomerated particles on the steel wool surface. Consequently, the particles break up and are readily removed when the field is removed from the filter. 4.3. Filtration breakthrough experiments The effects of applied magnetic induction and SDS treatment on filter breakthrough were also tested in this study. The results are displayed in Figs. 6 and 7. The values for Cout * represent the effluent concentration normalized to the feed concentration. Fig. 6 shows the effects of pH and SDS treatment. It is clear that results from SDS-treated suspensions are very similar to those at pH 10.8, indicating that at these conditions, particle aggregation is minimized. On the other hand, at pH 9.4, the breakthrough curve is much different, because of aggregation occurring despite the relatively high zeta potential (∼−40 mV) at this pH. Fig. 7 shows that the applied magnetic induction not only improves the RE of the filter in the beginning stage of filtration but also increases its loading capacity. The latter parameter is germane to the overall filter performance because it determines how often the filter must be regenerated. At 0.2 T, the magnetic force is not sufficiently strong enough to overcome the interparticle repulsive forces created by electrostatic and steric interactions. Consequently, the rate of particle capture decreases
Fig. 6. Effect of SDS treatment on filter breakthrough (B0 = 0.2 T; V0 = 0.003 cm/s; a = 15 m). Cout * = effluent concentration normalized to feed concentration. Feed concentration = 500 ppm.
Fig. 7. Effect of applied magnetic induction on filter breakthrough (V0 = 0.003 cm/s; a = 15 m). Cout * = effluent concentration normalized to feed concentration. Feed concentration = 500 ppm.
sharply as particles build up on the surface of the collectors. However, at 0.8 T, the magnetic force remains strong enough to dominate these forces even as more particles are deposited on the surface. Consequently, the rate of particle capture remains steadier throughout the course of the experiment. 4.4. Comparison of trajectory model predictions with experimental results A comparison of the effects of applied magnetic field, radius of the ferromagnetic wire in the filter, and fluid velocity on the RE shows a good agreement between modeling and experimental results. Figs. 8–10 show that the model predicts well the experimental trends. Furthermore, the figures show that at each level of all three parameters, the RE predicted by the model lies within the experimental error of the corresponding experimental RE. That these experimental results closely parallel the trends predicted by the trajectory model further substantiates the claim that
Fig. 8. Comparison of trajectory model predictions () and experimental results () for the effect of applied magnetic induction on RE of Fe2 O3 particles with SDS (wire radius = 15 m; fluid velocity = 0.3 cm/s). Values for other parameters are provided in Table 1.
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
207
fluid velocity, V0 , is also observed. The model again predicts a remarkably similar response to changes in the wire radius and fluid velocity (Figs. 9 and 10). It is noteworthy that the trajectory model works well, even though it does not consider the repulsive forces between the particle and collector caused by electrostatic and steric interactions. Although these forces are related to particle removal by a nonmagnetic mechanism, they are overwhelmed by the magnetic force and thus present negligible opposition to capture by this mechanism. 5. Discussion
Fig. 9. Comparison of trajectory model predictions () and experimental results () for the effect of wire radius on RE of Fe2 O3 particles with SDS (magnetic induction = 0.5 T; fluid velocity = 0.3 cm/s). Values for other parameters are provided in Table 1.
the removal of SDS-treated particles is due primarily to the operation of the magnetic mechanism. If the experimental filtration mechanism corresponds to the filtration mechanism simulated by the model, then the sensitivity of the experimental process and the sensitivity of the model to changes in the design parameters should be similar. The impact of changes in the levels of the three design parameters on experimental and modeling results shows a remarkable consistency between model and experimental sensitivity to an increase in B0 from 0.2 to 0.5 T (Fig. 8). A somewhat greater sensitivity of the model over the range of 0–0.2 T can be explained by the approximately 10% removal of particles due to nonmagnetic filtration. Thus, in the experimental system, magnetic filtration is the dominant, but not the sole, mechanism for particle removal. A close similarity between the model and experimental sensitivity to the wire radius, a, and
Fig. 10. Comparison of trajectory model predictions () and experimental results () for the effect of fluid velocity on RE of Fe2 O3 particles with SDS (magnetic induction = 0.2 T; wire radius = 15 m). Values for other parameters are provided in Table 1.
These findings have important implications for the use of HGMS to remove paramagnetic particles from water or from a suspension of a mixture of particles. First, if the particles targeted for removal have properties similar to those of Fe2 O3 particles with respect to aggregation and attraction to surfaces, then HGMS may not be the best technique for separation. If 90% RE is achieved by a filter with such a high porosity (96.5%), then nonmagnetic filters with a lower porosity (e.g., granular filters) would perform even better. These filters would probably require less operating costs than an HGMS process. Second, HGMS could be useful for separating paramagnetic particles via magnetic filtration from diamagnetic particles (particles with very low and negative magnetic susceptibility). To achieve this separation, it would be necessary to minimize the impact of nonmagnetic filtration mechanisms to ensure that the magnetic filtration mechanism determines particle capture. This can be achieved by increasing particle stability. This study showed that SDS treatment stabilizes the particles and, overall, results in lower RE. Thus, although such a treatment provides more favorable comparisons with the trajectory model, adding SDS would offer an advantage in a separation process that aims solely to remove these particles from wastewater. However, if the goal of the process is not only to remove the particles from water but also to isolate them for the purpose of recycling and reuse, then SDS treatment coupled with HGMS is a promising strategy. If the wastewater contains diamagnetic particles, dispersing and stabilizing these particles with SDS would decrease their removal by ferromagnetic filter wires via nonmagnetic mechanisms. However, the paramagnetic particles, also stabilized, would be retained by magnetic filtration under a high applied magnetic field. Third, optimization of an HGMS process must account for these three design factors. To achieve the maximum RE, the applied magnetic induction and wire thickness must be varied to achieve the strongest magnetic force, and the flow velocity must be varied to achieve the weakest drag force. However, in an industrial process, other considerations must be taken into account. A filtration process must be capable of handling sizable throughput. To achieve a high throughput, it is necessary to increase the flow rate, the filter size, and possibly the fluid velocity. Operating cost is another important consideration. Generating a strong magnetic field requires electrical power. Achieving high throughput from a magnetic filter would require a stronger magnetic field and thus higher
208
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209
operating costs. However, lowering the throughput would also raise costs. Therefore, these factors must be optimized. Fourth, SDS treatment could be useful for recovering particles from a magnetic filter. In industrial wastewater-treatment operations, the contaminants are often valuable materials that could be reused in the process. Therefore, it is important to be able to recover the contaminant from the filter. When filtration is due primarily to nonmagnetic mechanisms, the task of recovering particles from the filter is more difficult than when filtration is due primarily to a magnetic mechanism. This phenomenon occurs because it is more difficult to relax or reverse the attractive forces that are responsible for particle attachment. However, relaxing the magnetic force is a simple matter of turning off the electromagnet. If the magnetic force is primarily responsible for particle attachment, then the particles will more readily detach when the force is removed. 6. Conclusions A nonmagnetic filtration mechanism is primarily responsible for the removal of bare Fe2 O3 particles from water by a stainless steel filter, obscuring the effect on the filter RE of the parameters that determine the magnitude of the forces comprising the magnetic filtration mechanism. Conversely, the magnetic filtration mechanism is primarily responsible for the removal of SDS-treated Fe2 O3 particles. Adding SDS to the feed effectively disables the nonmagnetic mechanism by increasing particle stability via a steric repulsive force. By dampening the attractive forces behind the nonmagnetic mechanism, SDS treatment accentuates the effects of these same parameters on the RE, allowing for a meaningful comparison with the predictions of a trajectory model that simulates the magnetic removal mechanism. The reliability of a trajectory model depends on including the relevant forces in the right form. This model inadequately predicts the effect of the parameters on the removal of bare Fe2 O3 particles because it does not incorporate the appropriate nonmagnetic attractive forces. However, the model predictions agree nicely with the experimental data for the removal of SDS-treated particles because the only significant attractive force is the magnetic force, which the model simulates well. Better agreement with the predictions of a trajectory model does not justify the use of SDS in an HGMS process. If the performance of HGMS is measured by its RE, then SDS treatment is a disadvantage. However, SDS treatment can potentially improve other aspects of HGMS performance. SDS treatment allows for better recovery of particles from the filter when the magnetic field is disabled, making filter regeneration much easier. SDS treatment also allows for selective separation of nonmagnetic particles from magnetic particles. Stabilized nonmagnetic particles would escape capture in a magnetic filter, while stable magnetic particles would be retained. Thus, SDS treatment has the potential to preserve two chief advantages of HGMS: the reversibility and the selectivity of the magnetic force. Future studies should focus on testing these advantages of surfactant treatment. To study how well surfactants increase
the selectivity of an HGMS process, magnetic filtration experiments should be performed that measure the effect of surfactant treatment on the removal of Fe2 O3 particles and a nonmagnetic colloid, like kaolin, in the same suspension. Additional studies should examine the reversibility of the separation at higher magnetic field strengths and with different filter media. If the magnetic force is too high, the separation may not be as easily reversible. These studies could identify the optimal applied magnetic induction at which the attached particles remain in the domain of the secondary minimum potential energy, where they are easier to detach. Also of interest would be a more thorough investigation of the effect of different types of surfactants on particle removal. Using SDS in a large-scale wastewater-treatment process might be infeasible economically and undesirable environmentally. Therefore, less expensive, more environmentally friendly surfactants should be studied. Finally, in order to minimize the cost of the surfactant, the effect of surfactant concentration on reversibility and selectivity should be investigated to find an optimal concentration. This problem would also require a better understanding of the mechanism of surfactant adsorption. Acknowledgments Partial support for this work was provided by the Office of Basic Energy Sciences, Division of Chemical Sciences, U.S. Department of Energy, under contract DE-AC05-00OR22725 with UT-Battelle, LLC, and by Georgia Institute of Technology. The authors are also thankful to Dr. Marsha Savage for editing the manuscript. References [1] A.M.H. Shaikh, S.G. Dixit, Removal of phosphate from waters by precipitation and high gradient magnetic separation, Water Res. 26 (1992) 845–852. [2] E. Barrado, F. Prieto, J. Ribas, F.A. Lopez, Magnetic separation of ferrite sludge from a wastewater purification process, Water Air Soil Pollut. 115 (1999) 385–394. [3] A.S. Bahaj, I.W. Croudace, P.A.B. James, F.D. Moeschler, P.E. Warwick, Continuous radionuclide recovery from wastewater using magnetotactic bacteria, J. Magn. Magn. Mater. 184 (1998) 241–244. [4] P. Anand, J.E. Etzel, F.J. Friedlaender, Heavy metals removal by high gradient magnetic separation, IEEE Trans. Magn. MAG-21 (1985) 2062– 2064. [5] G.D. Moeser, K.A. Roach, W.H. Green, P.E. Laibinis, T.A. Hatton, Waterbased magnetic fluids as extractants for synthetic organic compounds, Ind. Eng. Chem. Res. 41 (2002) 739–749. [6] S. Bucak, D.A. Jones, P.E. Laibinis, T.A. Hatton, Protein separations using colloidal magnetic nanoparticles, Biotechnol. Prog. 19 (2003) 477– 484. [7] N. Karapinar, Magnetic separation of ferrihydrite from wastewater by magnetic seeding and high-gradient magnetic separation, Int. J. Miner. Process. 71 (2003) 45–54. [8] P. Phanapavudhikul, J.A. Waters, E.S.P. de Ortiz, Design and performance of magnetic composite particles for the separation of heavy metals from water, J. Environ. Sci. Health-Part A Toxic/Hazard. Subst. Environ. Eng. 38 (2003) 2277–2285. [9] M.D. Kaminski, L. Nunez, Extractant-coated magnetic particles for cobalt and nickel recovery from acidic solution, J. Magn. Magn. Mater. 194 (1999) 31–36.
C. Tsouris et al. / Separation and Purification Technology 51 (2006) 201–209 [10] C. Tsouris, T.C. Scott, Flocculation of paramagnetic particles in a magnetic field, J. Colloid Interf. Sci. 171 (1995) 319–330. [11] C.J. Chin, S. Yiacoumi, C. Tsouris, S. Relle, S.B. Grant, Secondary minimum aggregation of superparamagnetic colloidal particles, Langmuir 16 (2000) 3641–3650. [12] C.J. Chin, S. Yiacoumi, C. Tsouris, Agglomeration of magnetic particles and breakup of magnetic chains in magnetorheological fluids, Colloids Surf. A: Physicochem. Eng. Aspects 204 (2002) 63–72. [13] J.H.P. Watson, Magnetic filtration, J. Appl. Phys. 44 (1973) 4209.
209
[14] W.F. Lawson Jr., W.H. Simons, R.P. Treat, The dynamics of a particle attracted by a magnetized wire, J. Appl. Phys. 48 (1977) 3213. [15] H. Schewe, M. Takayasu, F.J. Friedlaender, Observation of particle trajectories in an HGMS single-wire system, IEEE Trans. Magn. MAG-16 (1980) 149. [16] T.-Y. Ying, S. Yiacoumi, C. Tsouris, High-gradient magnetically seeded filtration, Chem. Eng. Sci. 55 (2000) 1101–1113. [17] J. Svoboda, Magnetic Methods for the Treatment of Minerals, Elsevier, New York, 1987.