By Wendell James, Stephen Craik, Trevor Shu, Garry Solonynko and Tony Xie
The concept of using zeta potential (ZP) to guide coagulant dosing is not new, but techniques for measuring it are becoming more reliable and accessible to water treatment plant operators.
EPCOR Water Services is evaluating laboratory ZP measurements at its Edmonton water treatment plants, as a means of optimizing chemical dosing during two distinct modes of operation. The plants operate in conventional mode, with alum coagulation, settling clarifiers and dual media filters. They depend heavily on clarification during spring and summer when the North Saskatchewan River supply is subject to rapid changes in water quality. Maintaining the optimum alum dose is critical for maximizing turbidity removal through the clarifiers, especially during spring runoff and heavy rainfall events.
During fall and winter, the North Saskatchewan River water quality is generally more consistent, with turbidity less than 10 NTU and colour less than 8 TCU. This allows EPCOR to operate the water treatment plants in direct filtration (DF) mode. Lower flow makes the river more environmentally vulnerable during these seasons, but the natural suspended solids load is also lower. Substantially less coagulant is dosed during DF operation, to minimize discharge of coagulant residuals.
Filter polymer dosing must be carefully managed to ensure adequate particle removal, especially for parasitic protozoa such as Cryptosporidium oocysts. Even subtle increases in raw water colour can make DF operation difficult to maintain.
Setting ZP targets is proposed as an effective strategy for guiding coagulant dosing during conventional operation, and filter polymer dosing during DF operation.
Zeta potential measurements
The majority of suspended particles in surface water supplies have negative surface charges, which cause them to repel each other. Coagulants and/or cationic organic polymers are often added to neutralize charges and destabilize particles. Mixing energy is imparted to promote inter-particle collisions and formation of flocs, which can be removed by sedimentation and/or filtration.
It would be very difficult to directly measure surface charges on particles or the ionic clusters surrounding them. Indirect measurements can be used to infer charges and prescribe doses of treatment chemicals required to destabilize particles.
Electrophoresis, the movement induced when a charged particle is subjected to an electric field between two electrodes, can be used to estimate the ZP on particle surfaces, because particle velocity is proportional to the applied electric field strength. This principle is the basis of measurement for most commercially available ZP meters.
Instruments equipped with automated image-analysis software can measure ZP in samples taken from post-coagulant and pre-filter locations. These measurements can be used to establish ZP targets, which then allow chemical doses to be adjusted quickly and appropriately in response to changing influent water quality.
Bench-scale correlations between filter polymer dosing and ZP
A series of bench-scale experiments was conducted in 2014 to determine how ZP was affected by coagulant and cationic polymer dosing. Based on repeated analyses of raw water samples, the precision of ZP measurements was determined to be ± 0.4 mV. Adding either alum or cationic polymer caused an increase in ZP, but the dose response was much larger when using charged polymers. Figure 2 shows how ZP increased in response to doses of LT-7981 (an EPI DMA product) with and without a baseline alum dose. In each case, the response curve has a linear character.
The experiment conducted with and without alum on November 12, 2014 shows that the presence of alum increased the slope of the response curve. For the experiment on December 8, the sample was taken from the E. L. Smith water treatment plant upstream of the filters and immediately downstream of the hypochlorite injection point. The response curve was nearly parallel to the 3.4 mg/L alum curve obtained in November, but 3-4 mV lower. In general, samples collected downstream of hypochlorite injection had lower ZPs than samples collected upstream.
Zeta potential trends during conventional operation
To observe ZP trends during conventional operation, weekly samples were collected from the E. L. Smith plant after regular clarifier operation resumed in the spring of 2015. Samples representing raw water, post-alum injection, clarifier effluent and filter influent locations were analyzed. During the period between April 17 and July 22, raw water colour was generally decreasing, and the alum dose was reduced accordingly.
The box and whisker plot of Figure 3 shows upper and lower quartiles, median, minimum and maximum ZP values observed for each process location. The ZP of raw water samples was relatively consistent with a mean value of -12.7 mV. As expected, there was a large increase in ZP following alum addition. However, it was insufficient to approach electroneutrality (mean value -8.2 mV).
Several factors would have contributed to the large scatter in ZP observed in post-alum and post-clarifier samples. These include pH changes associated with alum addition, growth of particle flocs and the addition of anionic (primary) polymer. Due to sedimentation and sludge removal, post-clarifier ZP measurements would have applied to relatively few and smaller particles.
To explore reasons for the wide variability in post-clarifier ZP measurements, the results were divided into two groups to distinguish between two clarifier flow paths that receive alum doses separately. One flow path involved two clarifiers (C1 & C2) and the other a third clarifier (C3). Because the supervisory control system consistently showed similar alum doses applied to both flow paths, no differences in ZP measurements were expected.
However, as shown in Figure 4, C3 effluent ZP values were significantly lower than those measured in C1/C2 effluent. Summer C3 effluent turbidity readings had also been consistently higher than for C1/C2.
Later troubleshooting discovered a faulty valve had been allowing more alum to flow to C1/C2 than to C3. This explained the difference in ZP readings and confirmed the value of ZP as a monitoring tool.
Although filter performance remained acceptable during conventional operation, the mean ZP measured for filter influent particles was -11.7 mV, almost as low as values measured in raw water samples. The highest values reported for filter influent occurred during a period when the primary (anionic) polymer dose was lowest. ZP values were consistently much lower after the point of hypochlorite injection. These observations show the influence of chemical doses applied upstream of ZP monitoring locations.
Fall 2015: Transition to direct filtration
The original intent had been to start direct filtration operation at the E. L. Smith water treatment plant early in September 2015. As shown in Figure 5, raw water conditions were ideal during the first week when the turbidity and colour were both very low. However, several rainfall events created mild runoff conditions that increased the turbidity and kept the colour above 5 TCU until the second week of October. These conditions were generally unsuitable for DF operation.
However, a decision was made to convert part of the E. L. Smith plant to DF operation, starting on September 29, to test alternative strategies that might facilitate DF operation under such conditions.
Six filters being supplied by C3 were selected for DF operation because equipment had been installed to adjust the pH of coagulation on this flow path. Also, it had not yet become apparent that less alum than intended was being dosed upstream of C3. Conditions during the first half of October 2015 proved challenging, and the remainder of the E. L. Smith filters were not converted to DF operation until November 2.
During October, several alternative treatment strategies were applied, including lowering the pH of coagulation and adjusting the filter polymer dose. The six filters being operated in DF mode were monitored closely for parameters including influent ZP, ripening times and effluent particle counts. Effluent samples from individual filters were collected on three occasions and analyzed for Cryptosporidium oocysts.
Reducing the pH of coagulation from 8.0 to 7.5 decreased filter ripening times and increased the colour removed by C3. However, the pH adjustment had no measurable effect on filter influent ZP or reduction in particle counts achieved by the filters. The polymer dose was increased briefly on several occasions from a baseline dose of 0.5 mg/L. ZP measured in filter influent samples from two filters (Filters 1 and 5) didn’t always correspond in time to peak doses of polymer, but the general response is evident in Figure 6. When the polymer dose was returned to the baseline dose of 0.5 mg/L on October 13, filter influent ZP decreased to approximately -11 mV, similar to values measured in raw water samples.
A general downward trend in raw water colour was interrupted by a small resurgence between October 8 and 11, 2015. The ZP measured in the October 9 filter influent sample was quite low (-11.8 mV). The colour rise was also associated with an increase in filter effluent particle counts that lasted for approximately a week. These increased not only for the smallest (2-5 µm) category, but also for the 5-10 µm and 10-15 µm categories.
When filters shed more and larger particles, there is an increased likelihood of shedding Cryptosporidium oocysts. This concept proved true on October 19 and 26 when a few oocysts were detected in filter effluent samples. It should be noted that only about a third of the E. L. Smith production was from filters operating in DF mode at the time, and UV reactors downstream of the filters would have effectively inactivated any oocysts that weren’t physically removed.
Results of pilot trials
A pilot plant located at the E. L. Smith water treatment plant was used to mimic DF operation. A trial conducted on October 6, 2015 demonstrated the effect of filter polymer dose on ripening time. The experimental conditions and effluent turbidity trends are shown in Figure 7. Train 1 filters (1-1 and 1-2) received a polymer dose of 0.5 mg/L, similar to the dose being applied at the E. L. Smith plant. Train 2 filters (2-1 and 2-2) received a dose of 2.0 mg/L. Filter influent ZP readings were -10.8 mV and -4.8 mV for Train 1 and Train 2 filters, respectively.
Effluent turbidity settled much more rapidly for pilot filters receiving the higher polymer dose. Longer ripening times and lower ZP readings associated with Train 1 filters reflected the operation of E. L. Smith filters while they were receiving a polymer dose of 0.5 mg/L.
Pilot trials were also used to evaluate the effect of polymer dose on filter effluent particle counts. Figure 8 shows average log reductions in particle counts by size range, during four trials conducted between October 6 and 15. Clearly, the higher polymer dose (2.0 mg/L) provided a significant increase in ZP and resulted in more effective particle removal in each size range than the lower dose (0.5 mg/L).
Achieving a near-neutral charge on particles (as measured by zeta potential) is considered important for effective coagulation and to promote attachment of particles to filter media. Bench-scale experiments showed a linear relationship between coagulant and/or cationic filter polymer doses and ZP. ZP was measured in samples from various pre-filter locations during both conventional and DF plant operation to observe the impact of applying various process chemicals, including alum, hypochlorite and filter polymer.
Turbidity removal by clarifiers during conventional operation and particle removal by filters during DF operation were related to ZP and the associated doses of coagulant and filter polymer, respectively.
Efficiency of particle attachment is particularly important during DF operation when filters remove the bulk of the particle load. Pilot trials representing DF operation showed significant benefits of increased filter polymer dose, both in terms of decreasing filter ripening times and increasing log reductions of particles of each size category. The latter is particularly relevant for ensuring effective removal of Cryptosporidium oocysts during direct filtration operation.
Wendell James, Stephen Craik, Trevor Shu, Garry Solonynko and Tony Xie are with EPCOR Water Services.