Figure 1: Process flow diagram for the dead-end filtration test apparatus used in the calcium removal and membrane fouling experiments
Figure 2: Process flow diagram of the APS-MF integrated test system
Figure 3: Calcium removal as a function of calcium carbonate seed concentration and time (initial calcium concentration = 800 mg/L, pH = 10.5, T = 25°C). The test solution was the synthetic produced water
Results and discussion
Accelerated precipitation softening reaction kinetics and removal efficiency
The introduction of calcium carbonate seeds into the softening reactor resulted in increased precipitation kinetics (Figure 3) and greater calcium removal. The increase in precipitation kinetics was most notable for the seeded conditions at t ≤ 3 mins. After this point, there was a clear slowing in the kinetics for all seed concentrations. The rate of calcium precipitation did not change considerably from that in the absence of seeds until the seed concentration was ≥ 3 g/L. At this point the rate of calcium removal was 275 mg Ca2+/ min. This rate was approximately 15% greater than that measured in the absence of seeds (= 242 mg Ca2+/min). Increasing the seed concentration to 10 g/L resulted in a minor improvement, to 280 mg Ca2+/min, in the precipitation kinetics. Final calcium removal values reached 96% at a seed concentration of 5 g/L and 98% at 10 g/L. Previous works on precipitation softening have demonstrated that the practical minimum solubility limit for calcium at pH 10.5 is 10 mg/L as Ca2+ . Differences between theoretical and practical solubility limits for calcium can occur as a result of variations in kinetics, competing ion effects and impacts of solution ionic strength on water activity and calcium solubility . For this reason, higher or lower calcium removals may be achieved in complex mixtures like produced waters. Because the calcium concentrations at pH 10.5 and 11.6 were not statistically different within a 95% confidence level, all subsequent APS experiments were done at pH 10.5. In all cases, the reaction time required before the system reached a quasi-steadystate condition relative to calcium removal was 10 mins. Based on these results a seed concentration of 7 g/L and a reaction time of 10 mins were selected for subsequent testing
Calcium removal and fouling of MF membranes
Calcium removal by the PES MF membrane varied depending on the type of pretreatment applied to the produced water (Figure 4a). In the absence of precipitation (no treatment), no calcium was removed by the PES MF membrane, as would be expected for a pure filtration process. In contrast, when softening of any type was used upstream of the PES MF membrane calcium removal was ≥ 6%. Removal was slightly improved over the other conditions when APS was done upstream of the PES MF membrane (Figure 4a). The 3% increase in calcium removal, relative to that achieved in the absence of seeding, was attributed to the fact that the seeds are larger than the membrane pore size (= 0.45 µm) and thus, were more effectively retained by the membrane. From Figure 4b the calcium carbonate seeds grew in size within 5 mins of the initiation of the reaction. The primary peak size shifted from 2 µm to 51 µm, with no appreciable growth afterwards. This agrees with the calcium removal kinetics shown in Figure 3 where the majority of the calcium was removed within 3 mins. Based on this, the calcium was removed by the MF membrane through filtration of the calcium carbonate seeds.
Amongst all the treatment scenarios, the APS and MF treatment produced filtrate having the lowest turbidity Figure 4a. When APS was followed by conventional clarification the effluent turbidity was 26×that for the APS and PES MF treatment train. The presence and growth of a cake structure on the PES MF membrane was evidenced by the observed flux decline when APS was used (Figure 5). APS resulted in rapid flux decline with continued filtration, which contrasts to those conditions when no softening was done or when conventional softening was done. No observed flux loss was observed for the non-APS pretreatments (Figure 5). This suggests that any solids retained by the membrane did not appreciably affect the composite hydraulic resistance of the membrane. Further, calcium removal in these processes occurred prior to membrane filtration, if applied. Clearly calcium was removed during conventional softening; however, the type(s) of membrane fouling that occurred was distinct from that which occurred with the APS treatment. From the turbidity results (Figure 4b) more particulates were present in the PES MF filtrate when conventional softening was used relative to that for the APS pretreatment. This implies that more particulates were in fact retained by the MF membrane following APS. In fact, the mass of calcium carbonate retained on the PES MF membrane (= 75.3 g/m2 ) was an order of magnitude greater than that measured in the absence of softening (=1.0 g/m2 ), or following conventional softening (= 3.4 g/m2 ). The greater mass of solids retained following APS treatment was reflected in an exponential increase in cake resistance coefficient as filtration progressed, going from 3.60E1010 m-1 for the virgin PES membrane to 2.66E1010 m-1 after filtration of 500 mL of solution. Based on the particle size distribution in the feed (Figure 4a) and the resulting particle size to pore diameter ration, surface caking and complete pore blocking were the primary fouling mechanisms. These results are supported by SEM/EDS analysis of the PES MF membranes after each of the different treatment scenarios. From these analyses it was found that calcium carbonate cake formation was more prevalent after APS treatment, relative to the other conditions (Figures 6a-6c).
Figure 4a: Calcium concentration and turbidity of process flows subjected to varying treatment types (n = 3, initial calcium concentration = 800 mg/L, T = 25°C)
Figure 4b: Particle size distributions for the APS process at a seed concentration of 7 g/L at pH 10.5. The test solution was the synthetic produced water
Figure 5: Specific water flux as a function of filtrate volume for the PES MF membrane where the feed flow was subjected to different types of pretreatment (n = 3, PNET = 200 kPa, T = 25°C)
Integrated APS-MF system performance evaluation
Representative pictures of the feed water to the ceramic MF membrane and the MF filtrate are given in Figure 7a. The high turbidity of the feed water (>800 NTU) was due to the calcium carbonate seeds and the formation of precipitates during the APS reaction. No clarification was allowed after APS, as the APS reactor was continuously mixed, resulting in the highly turbid feed to the MF. From Figure 7b, the untreated water (prior to the initiation of the APS reaction) had a TSS concentration of zero. The TS were almost exclusively composed of TDS. Upon the initiation of the APS reaction the TSS concentration increased (= 9.0 ± 0.22 g/L) as a result of precipitate formation and seed addition. There was a 1.79 ± 1.0 g/L decrease in TDS for the APS effluent and MF filtrate due to removal of dissolved calcium and other co-precipitates like sulfate. Of note the turbidity of the MF filtrate was ≤ 0.18 NTU depending on the mode of operation and the sampling time during the test, representing an average turbidity removal of approximately 99.9%.
Figure 6a: Representative FESEM images of the PES-MF membrane surface with No softening
Figure 6b: Representative FESEM images of the PES-MF membrane surface with Conventional softening
Figure 6c: Representative FESEM images of the PES-MF membrane surface with APS softening
Figure 7a: Representative picture of the ceramic MF feed (Turbidity > 800 NTU) following APS treatment and the ceramic MF filtrate (turbidity = 0.15 NTU)
Figure 7b: Solids analysis of the untreated (raw), APS effluent and ceramic MF filtrate flows used in the integrated APS-MF test system (n = 3, T = 20° C). Samples for the MF filtrate were analyzed at the beginning of the filtration experiment
Compared to the dead-end filtration tests results for APS followed by MF, where 99% reduction in dissolved calcium was achieved (Figure 4a), the integrated APS-MF process achieved calcium removals between 96 to 98%. Here the filtrate calcium concentrations ranged between 16 to 35 mg/L as Ca2+. Turbidity was reduced from 4.8 NTU to ≤ 0.18 NTU for both operation configurations (filtrate return and filtrate withdraw). Filtrate turbidity remained constant throughout the durations of all tests. The slightly lower calcium removal between the dead-end and cross flow tests may be attributed to the material differences in the PES and TiO2 ceramic membranes, as well as to the relative lack of surface cake formation that was indirectly observed in the cross-flow system. The presence of calcium carbonate cake structures on membrane surfaces has been observed to result in greater calcium removal [15-17] due to improved precipitation conditions/kinetics within the cake structure. No significant decrease in calcium concentration, or turbidity, was observed overtime in the filtrate for either mode of operation in the cross flow tests. If a filter cake was forming or increasing overtime, filtration efficiency may improve and cause an improvement in filtrate quality (i.e., decreased calcium and turbidity) as has been observed in previous studies [15,16]. In addition, there was no appreciable change in membrane permeability over time. These observations were taken as indicators of a lack of appreciable cake formation on the membrane surface as was noted in the non-APS dead-end tests.
Marginal differences were observed between the two operational schemes in terms of the amount of calcium that was removed. Consistently, calcium removal was greater for the filtrate return mode (98% removal) when compared to the filtrate withdraw mode (96% removal). This is best attributed to the fact that the feed water chemistry and composition in the filtrate withdraw mode changed temporally (increased solids concentration) as a result of clean water being removed from this system. This had the effect of increasing the concentration of calcium carbonate and other particulate solids, in the feed water. It was postulated that this increase in particulate solids also resulted in an increase in the concentration of calcium, as small particulates, in the filtrate though no significant differences were observed in the turbidity values for the two configurations.
Membrane fouling in the integrated APS-MF process
Membrane fouling, measured in terms of TMP and filtrate quality, was assessed for both crossflow operational schemes. Because filtrate was not returned in the filtrate withdraw mode, the feed water quality changed over time. The TSS concentration increased time from 7.8 ± 1.5 g/L in the feed reservoir at the beginning of the tests to 59.1 ± 23.6 g/L at the end of the filtrate return tests. The increase in TSS concentration over time resulted in a corresponding increase in the solids loading rate on the MF membrane, going from roughly 1,560 to 11,820 g/m2hr. Conversely, in the filtrate return mode the TSS concentration in the feed reservoir and solids loading onto the MF membrane remained relatively constant through each test at 8.0 ± 1.1 g/L and 1,600 g/m2 hr, respectively
Representative water flux and TMP profiles for the cross flow MF process are reported in Figure 8 as a function of filtrate volume for both operational schemes. There was no change in TMP as filtration progressed for the filtrate mode of operation (constant feed water quality). This indicated that there was no appreciable accumulation of a cake structure on the membrane surface or pore blocking, which the two were fouling mechanisms expected for inorganic particle filtration. While a cake structure may have formed on the membrane surface, its hydraulic resistance was lower than the intrinsic hydraulic resistance of the virgin membrane. Therefore, the membrane may be said to have not fouled during filtration of ~200 L of the APS effluent (run time = 40 hrs). Notably no hydraulic backwashing, or pulsing, was used during this time and thus the only mechanism by which rejected solids were removed from the membrane surface was through the forces resulting from the cross flow shear. This result contrasts with what was observed in the dead end filtration experiments due to the cross flow shear that was present in the ceramic membrane tests. The shear forces continuously remove deposits from the membrane surface, where the interfacial conditions likely did not favor particle adhesion. This was due to the fact that both the membrane and calcium carbonate particles were hydrophilic and negatively charged at the pH condition used in the test. Another consideration was that the process was operated below the critical flux  for the process. The critical flux is defined by a force balance between permeate/filtrate drag forces and the tangential (shear) forces acting on a particle as it approaches the membrane surface. Regardless of its origin, these results demonstrated a resistance to performance degradation by the ceramic membranes in treating the APS slurry.
In filtrate withdraw mode (Figure 8b) a rapid increase in TMP was observed after approximately 1.7 L of water had been filtered (run time = 1 hr). This point corresponded to a feed TSS concentration of around 15 g/L. Note that the starting TSS concentration in the feed was 8 ± 1.5 mg/L. Like the process when operated in filtrate return mode, the process demonstrated a resistance to fouling until a relatively high solids loading was reached. Several possibilities exist that may explain the rapid loss of permeability that occurred once the TSS concentration reached 15 g/L. These include the onset of particle aggregation and an associated change in the aforementioned force balance due and changes in solution viscosity due to the high solids concentration. Of note, membrane permeability was fully restored when the feed water was changed to pure water having the same pH. This implied that particle caking onto the membrane process was determined by physical and not chemical, processes. Therefore, increasing the cross flow shear action at the membrane surface (fluid flowrate) may extend the solids concentration in the feed flow that may be achieved during treatment. Additionally, the use of backwashing and or back pulsing during treatment would be expected to mitigate membrane fouling.
Figure 8a: Specific water flux and TMP as a function of volume of water filtered for the filtrate return. No backwashing or back pulsing was used in either mode of operation
Figure 8b: Specific water flux and TMP as a function of volume of water filtered for the filtrate withdraw modes of operation. No backwashing or back pulsing was used in this operation
APS-MF treatment of representative produced water produces a superior product water quality in terms of turbidity and calcium concentration relative to that produced with conventional softening or APS/clarification. Filtration of APS effluent produced filtrate with a 3% lower dissolved calcium concentration and 8xlower turbidity than filtrate from water softened without seeds. An optimum calcium carbonate seed concentration of 7 g/L was identified for the produced water. The particulate seeds are well-rejected by polymeric and ceramic membranes. In a cross-flow configuration the particulates did not significantly foul the ceramic MF membrane until a solids concentration of 15 g/L was reached indicating that the APS-MF system is capable of treating high-solids content water. In addition, it could be operated without settling the water after the APS reaction, while maintaining a reasonable operating filtrate flux and TMP.