3.1. Batch experiments
In order to simulate the behavior of immobilized ZVI particles in a fixed bed, 104 carbon steel spheres (~13.7 g), were placed in a polypropylene mesh sample bag, which was then immersed in 300 mL of 99.8 μg/L arsenic solution at pH 4, prepared from an As(III) standard (AAS TraceCERT™, 998 mg/L, Fluka) for a loading of 2.2 μg As/g Fe. The solution was stirred with a magnetic stir bar.
The effects of particle–particle abrasion were simulated in batch experiments conducted in ten 50 mL centrifuge sample tubes attached to a mechanical rotator, each containing 40 mL of buffered 99.8 μg/L arsenic solution and 14 carbon steel spheres (~1.82 g) for the same initial loading of 2.2 μg As/g Fe. The samples were rotated at 60 rpm, maintaining the ZVI particles in constant motion, creating abrasion among themselves and with the walls of the centrifuge tubes. One tube at a time was then taken off the rotator at selected times – the last at 24 h. If no visible corrosion products were visible in the tube, a 1 mL sample of the solution was taken and preserved. For the tubes where corrosion products were visible, the contents were filtered through a 60 mL fine Büchner filter funnel (pore size of 4.0–5.5 μm). If corrosion product material remained in the filtrate, 50 mg of CaCl2 flocculant was dissolved in 10 mL of filtrate. After the corrosion material settled out, a 1 mL sample of the supernatant solution was collected.
The results of these experiments are presented in . As shown, the agitated particle experiments resulted in the removal of about 45% of the initial arsenic from solution (arsenic concentration of 55.6 ± 4.0 μg/L) in 24 h, while the immobilized particle samples showed no signs of arsenic removal over the same time period. In addition, visual inspection of the sphere surfaces from the immobilized sample experiments showed no signs of reaction. Although regular sampling of the solutions in the immobilized samples was terminated after 24 h, solution/particle contact was allowed to continue for an additional 8 days. Corrosion of the spheres was finally visually observable at about 198.5 h, at which point the arsenic concentration had decreased to 36.1 ± 4.0 μg/L. For the agitated samples, reddish-brown corrosion products were visible in the solution and on the steel particle surfaces after approximately 1 h, coincident with the observed decrease in arsenic concentration.
Fig. 3 Removal of 99.8 μg/L As(III) in batch experiments with fixed (immobilized) and mixed (agitated) of carbon steel spheres in a pH 4 buffer electrolyte solution (6.3 mM NaOH and 2.4 M KHP) at ambient temperature. Fixed bed: V0 = 300 cm3, mspheres (more ...)
Although arsenic removal was observed for both the immobilized and agitated particle batch samples, the time scale was greater than that reported for other ZVI materials under similar experimental conditions [12
]. Biterna et al. [13
] observed nearly complete arsenic removal within 30 min for an initial concentration of 200 μg/L As(III) using 2 g/L of iron powder (325 mesh) at pH 4. At initial pH values of both 3 and 5, a loading of 150 mg/L ZVI powder removed 85% of the 500 μg/L As(III) in under 2 h, as reported by Katsoyiannis et al. [12
]. The difference in the time required for arsenic removal is attributed primarily to differences in surface area of the ZVI materials and the dependence of the removal mechanism on the iron corrosion rates Katsoyannis et al. [12
] reported a surface area of 0.1 m2
/g for their ZVI powder. In comparison, the carbon steel spheres have an estimated surface area of 2.4 × 10−3
/g. In addition, ZVI materials develop an oxide layer from exposure to air over time, and unless precautions are taken to prevent their oxidation, such as described by Kanel et al. [14
], then initial arsenic uptake can occur on pre-existing oxide layers. The ZVI materials used by both Biterna et al. [13
] and Katsoyiannis et al. [12
] were not pretreated prior to use. Although no pretreatment of the carbon steel spheres was performed in the current experiments, the large particle size allowed for visual inspection of the spheres. No corrosion was observable on the surfaces of the immobilized particles until about 8 days, which is also about the time of the onset of arsenic removal. So, in this case, under relatively anoxic conditions, it is likely that resistance to the transport of dissolved oxygen to the iron–water interface probably had an effect on the overall rate of corrosion and arsenic complexation.
The observed differences in performance are at least partially attributable to increased mass transfer rates of arsenic to available adsorption sites in the agitated system. Yu et al. [19
] found a strong correlation between arsenic removal and increasing Reynolds number, suggesting that the mass transfer rate is of significant importance for arsenic removal by ZVI. These authors also compared arsenite removal at varying stirrer speeds with an initial concentration of 2.88 mg/L As(III) and an iron loading of 50 g of ZVI (28–35 mesh) in 2 L of 0.1 M NaClO4 solution. At 333 rpm a 99% As(III) removal was observed after 1 h. In comparison, 5 days were required to attain 96.5% As(III) removal at 5 rpm. Yu et al. [19
] suggested that increased stirrer speed could shear iron oxide layers from the ZVI surface, which in turn exposes fresh iron material, thereby stimulating additional iron corrosion. In our batch experiments, in addition to enhanced mass transfer due to liquid shear, the continuous motion of the carbon steel spheres causes collisions between particles and with the centrifuge tube walls that continually removes corrosion products from the particle surfaces in a physical process that produces active high surface area colloidal material and exposes fresh ZVI material to continue/enhance the surface renewal/complexation process. This mechanism also occurs in the dense moving particle bed on the conical bottom of the spouted vessel.
3.2. Continuous flow experiments
Comparisons of the performance of continuous flow experiments in the spouted vessel and packed column systems, in terms of the generation of iron corrosion products, also support the preceding conclusions. In one set of experiments, the steel spheres in the spouted vessel exhibited a mass loss that was an order of magnitude greater than that in the comparable fixed bed experiment; 24.3 g in the former, and 2.5 g in the latter. Visual inspection of the carbon steel spheres from the spouted vessel experiments showed the expected signs of corrosion. Those from the comparable packed column contactor experiments also exhibited corrosion products on the surfaces that were much thicker than those observed in the spouted vessel experiments. Indeed, in the packed column experiments, corrosion products filled much of the interstices between the spheres, “cementing” them together and providing much greater resistance to flow. Thus, the packed column contactor promoted the formation of corrosion products on the carbon steel sphere surfaces, whereas the spouted vessel promoted the formation of colloidal corrosion products in the bulk liquid phase.
The occurrence of colloidal corrosion products was also cited in the arsenic removal experiments reported by Bang et al. [15
]. In the presence of dissolved oxygen, these authors observed that corrosion products readily formed as suspended particulates, as well as on the surfaces of ZVI filings. Furthermore, for a total As(V) removal of 72% after 90 min of mixing, only 44% was removed on the surfaces of the ZVI material, while the remainder was attributed to that on colloidal corrosion products. The role of colloidal corrosion products has also been discussed by Yu et al. [19
] and Triszcz et al. [16
In the current work, it was observed that exposed surfaces in the spouted vessel system were rapidly coated with colloidal corrosion products by operation with carbon steel spheres in the absence of arsenic. The carbon steel spheres and the solution were then removed, and the spouted vessel was filled with 24 L of fresh 100 μg/L As(V) solution. The arsenic concentration in solution was monitored over time by collecting and analyzing 1 mL samples of the bulk solution. After just 0.5 h of contact time with the pre-coated colloidal corrosion products, the arsenic concentration had decreased from 92.4 to 35.3 μg/L, and the presence of arsenic in solution was undetectable after 24.2 h. The contents of the spouted vessel were then drained, and the system was dried with ambient air without removing any of the pre-deposited colloidal corrosion products. Next, the spouted vessel was filled with 240 mL of a 1000 mg/L As(III) standard and filled with water to 24 L to yield an arsenic concentration of approximate 10 mg/L. The solution was then allowed to circulate for 20 days with samples taken periodically and analyzed by GFAAS. The results showed essentially no discernible arsenic removal by the colloidal corrosion products at this high arsenic concentration (10 mg/L). These results are interpreted to mean that colloidal corrosion products are quite effective in removing arsenic at low concentrations, but not at much higher concentrations. It is anticipated that under these conditions, all the available complexation sites on the pre-deposited colloidal corrosion material are rapidly saturated, and no additional arsenic can be removed without the generation of fresh colloidal corrosion material. Thus, the relative importance of suspended and deposited colloidal material vs. bulk surface iron corrosion products for the removal of arsenic from solution must be understood and carefully considered in order to understand performance differences between different arsenic removal systems for widely varying arsenic concentrations, and in the effective design of ZVI arsenic removal systems.
In another series of experiments, a filter was installed in the reservoir of the spouted vessel apparatus (see ). The results of arsenic removal at low concentrations (~100 μg/L) are presented in . As shown, the spouted vessel with internal filtering is a very effective technique for arsenic removal at low concentrations. The arsenic concentration of the solution attained a value of <2 μg/L in about an hour, which is significantly below the USEPA MCL of 10 μg/L [10
Comparison of arsenic removal from a solution of 100 μg/L As(V) with 500 mg/L NaCl in a spouted vessel with and without filtering of colloidal corrosion products. The vessel contained 24 L of solution at 45 °C, initially at pH 6.
3.3. Analysis of arsenic removal in the spouted vessel system
Assuming that the fundamental arsenic removal rate is defined by the encounter frequency of arsenic species, A
*, with “complexation sites in active colloidal iron corrosion products” (i.e.
, “active sites”), I
*, the rate of the removal process may be generally represented as:
is an effective second order rate constant that can incorporate mass transfer resistances that are often important in ZVI systems. Since [A
*] and [I
*] are local concentrations, Eq. (1)
can result in LHHW-type rate expressions when based on bulk phase concentrations that exhibit variable apparent reaction orders between zero and first for some systems [17
As established by the preceding experiments, the generation of “colloidal iron corrosion product material” (“colloidal material”) in the spouted vessel occurs primarily via
the self-polishing action of iron-containing particles in the moving bed that forms on the spouted vessel bottom, since that is where the particles reside for approximately 97% of their circuit time [18
]. If the generation rate of colloidal material is approximately constant, then the behavior of the active site concentration can be represented as:
* is the total generation rate of active sites in the moving bed (e.g.
, number of active sites/time), Veff
is the effective volume over which the colloidal material is active, and kd
] is the rate of iron complex deactivation which allows for depletion due to “age,” since it has been reported that fresh iron complexes are more active than more aged ones [17
]. For a constant generation rate of active sites that is more rapid than its sinks (i.e.
, complexation and aging), Eq. (2)
which assumes Veff
≈ constant, and [I
*] = 0 at t
In order to develop expressions describing the arsenic concentration in the bulk liquid phase in the spouted vessel system, this general formulation must be modified to reflect the characteristics of each portion of the active site population in the system. These are identified as colloidal material: (1) in the filter; (2) on accessible surfaces in the system; (3) in the circulating bulk liquid phase; and (4) on the primary ZVI particle surfaces where they are generated. The latter is assumed to be negligible in the current work since the surface area offered by the spouted ZVI particles is orders of magnitude less than that of the colloidal material. In addition, the “self-polishing” action of the particles in the moving bed is very effective in removing (i.e., “cleaning”) almost all the corrosion product material from the ZVI particle surfaces.
The colloidal material in the filter should be more effective for arsenic removal than the adsorbed colloidal material, and that in the circulating bulk liquid due to the high relative velocity between the “fixed” (captured) colloidal material and the bulk liquid solution. (The relative velocity should be close to zero on average in the bulk solution.) In addition, the filter is effective in removing almost all the colloidal material from the circulating bulk liquid solution. (The bulk liquid solution was always visually clear when operated with the filter.)
Insofar as the role of mass transfer resistance, Yu et al. [19
] found a strong correlation between arsenic removal and the liquid phase Reynolds number, which suggests that the effective mass transfer rate is of significant importance for arsenic removal by ZVI. These workers compared arsenite removal at varying stirrer speeds with an initial concentration of 2.88 mg/L As(III) and an iron loading of 50 g of ZVI (28–35 mesh) in 2 L of 0.1 M NaClO4
solution. At 333 rpm, 99% As(III) removal was observed after 1 h. In comparison, at 5 rpm, 5 days were required to attain 96.5% As(III) removal [19
]. It was suggested that increased stirrer speed could shear iron oxide layers from the ZVI surface, which, in turn, exposes additional surface area and stimulates further iron corrosion. In a similar fashion, in the current work, the continuous abrasion of the carbon steel spheres with one another and along the walls of the centrifuge tubes in the batch samples, and in the moving particulate bed on the bottom of the spouted vessel serve to continually remove corrosion products from the ZVI particles.
If it is assumed that arsenic removal in the hybrid spouted vessel/fixed bed filter system is dominated by the colloidal material accumulating in the filter, as our observations suggest, then substituting Eq. (3)
into Eq. (1)
is the effective filter volume, which is assumed to remain constant.1
Integration of this expression yields:
As shown in , the arsenic removal rate is much more rapid with the filter installed. Also, the resultant curve is not fit very well by a simple exponential, as would be the case for a first order process. Indeed, as shown, Eq. (5)
provides a much better fit of the data with the filter, with kk
= 2.6 × 10−3
. In this experiment, 1.81 g of colloidal material was collected by the filter in 2.5 h, or about 0.72 g/h, which provides a rough estimate of k
*. Approximating the effective filter volume as the geometric volume of the filter multiplied by an estimated void fraction of ε
= 0.4, Vf
= 156 cm3
, and k
= 34 ((g colloidal material/cm3
The other data set presented in , obtained in the absence of the filter, shows a considerably reduced arsenic removal rate. In this case, both the adsorbed colloidal material and that accumulating/circulating in the bulk liquid phase, may also be significant. The material in the bulk liquid phase also accumulates in time, such that the resultant formulation should be similar to Eq. (5)
Therefore, for the same constant rate of colloidal material generation, k
*, the timescale for arsenic removal is controlled by the ratio, k
in Eq. (6)
, or k
in Eq. (5)
Like the active colloidal material captured by the filter, the material adsorbed onto exposed vessel surfaces is captured and concentrated. However, unlike the material captured by the filter, its accumulation occurs relatively rapidly, due to the limited surface area per unit volume, and ceases when it attains its allowable thickness. This material will also experience deactivation due to loss of accessibility to active sites as a result of the diffusion barrier that develops in the multilayer deposit. Our observations support the rapid adsorption of this material onto system surfaces. Since we have not yet investigated the details of this mechanism, for the current purposes the behavior of this material is approximated by a simple model of a “step function” increase (i.e.
, “rapid” adsorption), followed by a first order deactivation process:
] is the available, active colloidal material per unit surface area on the system surfaces,
is the total amount of active sites in the adsorbed layers, and kd,s
is the first order deactivation rate constant of active sites due to decreasing accessibility and other factors. (Loss of active sites due to arsenic complexation is also a sink, but for low bulk phase arsenic concentrations, it is assumed to be negligible.) Substitution of this expression into Eq. (1)
″ is the apparent second order rate constant for complexation of arsenic species in the adsorbed material, and a
is the effective specific surface area of the spouted vessel system over which adsorption of active colloidal material occurs. The solution of this expression is:
Once this sink is exhausted, then the only remaining major sink is suspended, circulating colloidal material in the bulk liquid solution, which occurs at a much lower rate.
As shown in , the arsenic concentration-time history for operation without the filter is more complex than that with the filter. It is believed that this is due to contributions from more than one portion of the colloidal material population; i.e.
, it is a composite. In accordance with this hypothesis, the data without the filter were fit to a linear combination of Eqs. (6)
where the first term represents slower arsenic removal by the colloidal material in the bulk liquid phase, and the second (faster) term by colloidal material adsorbed onto the spouted vessel system surfaces. The resultant best fit parameter set for the data in for this model without the filter is: C1
= 0.5, C2
= 0.5, k
= 5 × 10−6
. These results indicate that about half of the total initial bulk arsenic will (eventually) be removed by the entrained, circulating colloidal material in the bulk liquid solution, and the other half by complexation with active colloidal material adsorbed onto the surfaces of the spouted vessel apparatus. Assuming k
* is similar for operation with/without the filter, (k
) ≈ 520, and if the volume ratio is roughly (Veff
) ≈ (Vb
) ≈ 24,000/156 = 154, then k
′ ≈ 520/154 = 3.4, which suggests that the apparent rate constant for arsenic complexation in the filter is approximately a factor of three or so greater than that in the bulk liquid solution, which seems reasonable based upon estimated mass transfer resistance differences alone. Also, k
′ ≈ 10 ((g colloidal material/cm3
As shown, the performance of the hybrid spouted vessel/fixed bed filter system for arsenic removal is quite effective. With the filter in place, there is no appreciable arsenic complexation in the bulk liquid phase since there is little or no colloidal material available in that phase. For the same reason, as well as its more limited capacity, the adsorbed surface layer is also an ineffective sink for arsenic under these conditions. Therefore, arsenic removal occurs almost totally within the filter.
In the absence of the filter, active colloidal material is available in the bulk liquid phase for arsenic removal, and for rapid formation of the adsorbed surface layer. In the current work, apparently both of these sinks are comparable in terms of the total amount of arsenic removal over the long run. It is also apparent, however, that the adsorbed colloidal material complexes arsenic at a much greater rate than that in the bulk liquid phase, primarily due to the much lower effective active site concentration in the latter sink. However, the adsorbed material is also rapidly saturated due to its more limited capacity.