Comparative evaluation of pilot-scale reactors based on pulsating floc blanket clarification and conventional clariflocculation technologies in simultaneous treatment of natural organic matter and turbidity Open Access

Water treatment technology may be seen as a means for efficiently implementing a given unit process(es), for instance, clariflocculation can be carried out in conventional upflow clariflocculators (CCs), floc blanket clarifiers (FBCs) and other systems. In the light of new findings/developments/imposed regulations, either processes need to be made more efficient and/or technologies may also evolve. Conventional clariflocculators have been facing issues related to space requirements, short life of mechanical moving parts, treatment performance under varied influent conditions and in meeting the regulatory requirements that are getting more stringent by the day (Dhabadgaonkar 2008; Edzwald 2020). A zone of densely concentrated suspension of flocculated particles in fluidized state, in other words, a floc blanket may provide some mechanistic advantages in water treatment. However, many aspects of floc blanket clarification, arguably one of the principal treatment processes are less understood (Hendricks 2010). Thus, comparative evaluation of pulsating floc blanket clarification and conventional upflow clariflocculation technologies on a pilot-scale experimental setup (Srivastava et al. 2021a, 2021b) can help in not only explicating the fundamental differences between the two processes but also demonstrate how the theoretical considerations translate to practical performance results.

Natural organic matter (NOM) released into surface water allochthonously or autochthonously poses a serious challenge before water treatment plants (WTPs). It produces aesthetic considerations such as unacceptable taste, color and odor, but more importantly, acts as a precursor to the formation of carcinogenic disinfection by-products. Enhanced coagulation is regarded as the best available technology (BAT) for NOM removal (Saxena et al. 2018, 2019; Lapointe et al. 2021) however high coagulant dosage administered in the process increases the risk of excessive residual aluminum in the finished water (Edzwald 2020; Lapointe et al. 2021). The quandary before WTPs is associated with serious health repercussions since potential neuro-toxic effects of aluminum (Krewski et al. 2007; Willhite et al. 2021) and carcinogenicity originating from NOM (Fang et al. 2010; Chaukura et al. 2020) in treated drinking water have been well documented. Thus, limits for residual aluminum concentration in treated drinking water under Indian regulations (BIS 2012) have been stated as 0.03–0.2 mg L⁻¹ Al and limits for dissolved aluminum concentrations is 0.05–0.2 mg L⁻¹ Al under the United States Environmental Protection Agency regulations (USEPA 1999a). Required percentage total organic carbon (TOC) removal ranges from 10 to 50% depending on the raw water TOC concentration and alkalinity (USEPA 1999b).

The problem is exacerbated since NOM has been reported to undergo complexation reactions with clayey turbidity as well as PACl, mostly resulting in the formation of soluble species (Saxena et al. 2019, 2021). In fact, several full-scale blanket clarifiers in Taiwan using PACl as coagulant had completely lost their blanket during low-turbidity period due to the presence of the humic acid (HA), a major constituent of NOM, in the raw water (Su et al. 2004). Distinct intrinsic mechanisms involved in a given reactor configuration influences inter-related parameters like particle aggregation and destabilization; surface adsorption and coverage; floc characteristics; microfloc formation and kinetics; and finally, the performance of subsequent unit operations. However, the physico-chemical significance of fundamentally different reaction environments provided by PFBC and CC with PACl for simultaneous influx of NOM and clayey turbidity has not been studied in much detail and hence is the focus of this paper. Comparative evaluation of the role that a floc blanket may have on properties of particles and their aggregation behaviour also lends fresh insight on operating WTPs based on either of the two technologies at peak efficiency in such inclement influent conditions.

For CC, average flow rate was maintained at 3.745 liters per minute (LPM), resulting in an upflow velocity of 0.463 mm s⁻¹ in the clarifier. Detention time (t_d) and velocity gradient (G) for the rapid mix unit were 30 and 600 s⁻¹ respectively, while t_d and G for flocculator of CC were 20 minutes and 40 s⁻¹ respectively.

In the given set of experiments, an actuator-solenoid valve system cycle time of 55 seconds was used with ‘on’ time of 10 s and ‘off’ time of 45 s. Continuous rise rate (upflow velocity) was 0.519 mm s⁻¹, and the upflow velocity reached a maximum 2.523 mm s⁻¹ for a period of 10 s during the admittance of pulse. The continuous flow together with the pulsed flow helped maintain a homogeneously suspended blanket for a continuous supply of treated water. The designed average flow rate of 3.745 LPM was maintained throughout the experimental phase in both the pilot plants.

TOC concentration in surface water generally ranges from 2 to 10 mg L⁻¹. Three years data of surface water characteristics admitted to WTPs in the state of Rajasthan, India show the average monthly peak input turbidity to be about 17 NTU (Saxena et al. 2020). To simulate the surface water conditions in synthetic raw water (SRW), turbidity was varied from 0 NTU to 20 NTU in increments of 5 NTU, while TOC was kept constant at 6 mg L⁻¹. PACl doses optimized for simultaneous NOM and turbidity removal (5, 6, 6, 6 and 8 mg L⁻¹ Al respectively) for each influent condition calculated from the jar tests results (Saxena et al. 2019) were administered continuously to both pilot plants.

An established stable floc blanket is a pre-requisite for satisfactorily operating most floc blanket clarifiers (FBCs), including the pulsating variant. Such a requirement is not posed by CCs. Hurst et al. (2017) had studied the floc blanket growth rate and steady-state suspended solids concentration, and had suggested a continuous dosing of 6.296 mg L⁻¹ Al for an influent kaolin turbidity of 500 NTU for optimal initial floc blanket formation. Thus, the PFBC was initially run as per the recommended settings until the blanket reached the height of 1.2 m before proceeding for experiments on assessment of performance under various influent conditions. The influent stream of SRW was treated in parallel on both pilot plants, simultaneously running under similar operational conditions (Figure 1). For continuous flow type pilot plants, steady state is reached after about 5 hours (⁠⁠), thus samples of clarified water were collected after the setup was run for 6 hours to allow proper stabilization.

Tap water supplied at Malaviya National Institute of Technology Jaipur having alkalinity in the range of 180–205 mg L⁻¹ (as CaCO₃) and negligible TOC (<0.069 mg L⁻¹) was used in preparation of SRW. Kaolin clay (Loba Chemie Pvt. Ltd, Mumbai, India) was used to synthetically impart turbidity (Hurst et al. 2017; Saxena et al. 2019) to the influent stream of raw water. NOM was simulated by the addition of HA (Sigma-Aldrich Corp., Missouri, USA) to influent water. For the preparation of stock solution, 1.5 g of powdered HA and 0.01 M NaOH (Rankem, Avantor Inc., Pennsylvania, USA) were added to 1 L of de-ionized (DI) water. After stirring the solution for 5 hours to ensure complete dissolution, the TOC of the stock solution was measured to be 736.9 mg L⁻¹. High basicity polyaluminum chloride (PACl) (Aditya Birla Chemicals, Grasim Industries Ltd, Nagda, Madhya Pradesh, India) was used as coagulant.

Measurement of turbidity was carried out in accordance with the 2130 B Nephelometric Method (APHA 2012, Part 2000 Physical and Aggregate Properties) on a portable digital nephelometer (TU-2016, Lutron Electronic Enterprise Co. Ltd, Taipei, Taiwan) and pH was measured on a portable pH meter (Hanna HI98130, Rhode Island, USA). The 2320 B standard titrimetric method (APHA 2012, Part 2000 Physical and Aggregate Properties) was followed for calculation of alkalinity. TOC was measured in accordance with the 5310 B high-temperature combustion method (APHA 2012, Part 5000 Aggregate Organic Constituents) on a total organic carbon analyzer, TOC-L (Shimadzu Corp., Kyoto, Japan) at combustion temperatures of 680 °C with catalytic oxidation.

The analysis for residual aluminum was carried out in accordance with the 3500-Al B Eriochrome Cyanine R Method (APHA 2012, Part 3000 Metals) on a spectrophotometer (UV 1800, Shimadzu Corp., Kyoto, Japan) at a wavelength of 535 nm using a 1 cm path length glass cuvette. Total and dissolved residual aluminum concentrations were measured in accordance with the methodology described by van Benschoten & Edzwald (1990). To measure total residual aluminum, solubilization was carried out with the help of trace metal grade concentrated HNO₃ (Fisher Scientific, Massachusetts, USA). Undigested clarified water samples were first passed through a 0.22 μm pore size polymeric membrane filter with the help of a vacuum pump. The filtrate so obtained was then tested for aluminum after acidification as per procedure described above to obtain the dissolved residual aluminum. Difference between total and dissolved residual aluminum gives the particulate fraction. All samples and reagents used in the process were prepared or stored in polypropylene ware instead of glassware to avoid any error in trace aluminum determination. Glass cuvettes and wares used in determination step were rinsed with 1 + 1 HNO₃ and deionized water.

A Zetasizer^® Nano ZS instrument (Malvern Instruments, Malvern, UK) utilizing the dynamic light scattering (DLS) technique was employed to measure zetapotential and the size of fine suspended colloidal particles in clarified water. The microscopic analysis according to a modified approach described by Srivastava et al. (2021a) was used to determine the size of flocs since they are larger and amenable to settle over time. A CX21i Microscope (Olympus Corp., Tokyo, Japan) and digital camera assembly, Magcam HD L (Magnus Opto Systems India Pvt. Ltd, New Delhi, India) using a 1/2.5′ CMOS image sensor was used to capture high resolution images (2,464 × 1,632 pixels) at ×10 optical magnification. Magvision^® software (Magnus Opto Systems India Pvt. Ltd, New Delhi, India) was used for layering, scaling and primary image analysis. Further image analysis, including particle size analysis (PSA) was done with the help of open source software ImageJ distributed by Fiji (Schindelin et al. 2015).

Sludge samples were collected from PFBC and CC pilot plants with the help of peristaltic pumps. The samples were first transferred to porcelain crucibles and then completely dried in a hot air oven (LabSmith, Inc., California, USA) at 150 °C temperature. Field emission-scanning electron microscopy (FE-SEM) was carried out on a Nova Nano FE-SEM 450 (FEI, Oregon, USA) at ×10,000 magnification to analyze the surface morphology and microstructure of sludge. Fourier transform infrared (FT-IR) analysis of sludge was carried out on IR Spectrum Two™ spectrophotometer (PerkinElmer, Inc., Massachusetts, USA) for identifying characteristic species.

Polynuclear aluminum species like Al₁₃ present in abundance in high basicity PACl were effective in capturing and neutralizing negatively charged colloidal particles which brought the ξ values within a range of −2.4 mV to −6.8 mV for PFBC and −3.2 mV to −11.3 mV for CC (Figure 2). Low standard deviation for PFBC (σ = 1.54) may be attributed to the buffering action of the blanket. It can also be seen that ξ_avg was lower in magnitude for PFBC (| − 4.7| mV) as compared to CC (| − 6.6| mV) which perhaps indicates better destabilization. However, on simultaneous observation of the combined variation trends in turbidity and ξ, it may inferred that particle destabilization was not the only factor responsible for such marked differences in the residual turbidity between PFBC and CC, and that perhaps other mechanisms, such as contact flocculation (Tambo & Hozumi 1979; Hauduc et al. 2019), charge neutralization (Lin & Ika 2019; Lapointe et al. 2021) and enmeshment (Chen et al. 2016) were dominant in the zone of densely concentrated suspension of pre-flocculated particles in fluidized state inside PFBC.

Firstly, to assess the size distribution of individual constituents comprising the SRW, PSA of stock HA and kaolin suspension were measured and the particle size distribution was found to be 267 ± 60 nm and 2,300 ± 540 nm respectively (Saxena et al. 2019). The respective stocks after requisite dilution were used in the preparation of SRW and subsequently underwent coagulation–flocculation unit process in the two types of reactors with coagulant PACl. Further PSA of fine particles which remain in suspension in the clarified water was carried out and it was found that Z_avg for CC (5,693 nm) was larger on average as compared to PFBC (2,341 nm). Equal average PdI value of 0.81 was recorded for samples from both PFBC and CC, and this relatively large value indicates the condition of hetero-dispersity. Saxena et al. (2021) noted that interactions between hetero-dispersed impurities improve coagulation and significantly impact floc properties.

The results depicted in Figure 3 observed in conjunction with zeta potential data points given in Figure 2 indicated that due to adequate colloidal destabilization, the microfloc formation proceeded well past the submicron range. However further growth/agglomeration of microflocs beyond a certain threshold is required to form a settleable floc, or else they may remain suspended. Smaller cut-off size along with lower residual turbidity for PFBC indicates that the fine particles in suspension may have aggregated to form larger flocs and settled down. Nevertheless, the fine colloidal particles in suspension were supramicron size (>1 μm) for both pilot plants and the subsequent filtration unit process has been reported to be more effective in their removal (Chowdhury et al. 1993). Owing to the size of these particles, a reduction in the particulate residual aluminum may also be expected after filtration for both PFBC and CC (Srivastava et al. 2021b).

The average total residual aluminum level for CC (141 ppb) was marginally lower than for PFBC (147 ppb). However, notwithstanding the rather similar total residual aluminum levels, the concentration of residual aluminum in dissolved form for CC (0.059 ppm) was markedly higher than the average dissolved residual aluminum for PFBC (0.031 ppm). It may be observed that the particulate fraction was found to comprise of total residual aluminum in clarified water for PFBC while the value for CC was ⁠.

Dissimilarity in the respective proportions can be explained on the basis of the compensation brought about by the presence of pre-formed floc blanket for PFBC. As the coagulated water sifts upwards through the fluidized bed of primarily kaolin flocs in the PFBC, perhaps precipitation of dissolved species occurs at the nucleation sites provided by the flocs and compensation by the clayey matter leads to residual aluminum fractionation results similar to those offered by Srivastava et al. (2021a). Conversely, the relative concentration of alumino-organic complexes ought to be higher in CC since the floc formation is effected by rapid mixed constituents (influent SRW and PACl) alone. Thus, it may be inferred that dissolved residual aluminum concentration was increased due to relatively higher NOM.

Furthermore, it can be observed from Figure 5 that the average TOC in clarified water from CC was 1.936 mg L⁻¹ whereas analysis of TOC for PFBC gave a slightly lower value of 1.776 mg L⁻¹ on average. This observation may be explained on the basis of interactions reported between HA and kaolin from the floc blanket if there is a low HA range of 4 and 8 mg L⁻¹ (Saxena et al. 2021). Although the alkalinity of SRW used was >120 ppm (as CaCO₃), both pilot plants were able to reduce TOC concentration by which was well above the maximum stipulated percentage reduction required (50%) by USEPA (1999b) for any alkalinity/source water TOC bracket. It may also be observed from Figure 5 that there is little mesial variation of the tight cluster of points with increase in influent turbidity as indicated by the average slope (m = 0.016). Thus, it may be inferred that PACl is first utilized to meet a significant aluminum demand raised by NOM before it can serve as an effective coagulant (Edzwald & van Benschoten 1990). Therefore, under low-turbidity and high organics conditions, optimization of coagulant doses in the enhanced coagulation range is a more important factor otherwise the HA would complex with clayey flocs and deplete the blanket.

Therefore, fractal dimensions so computed do not describe the property of an individual floc but represent an overall fitting parameter for entire population of flocs in a specific sample. For samples from CC (C₁₋₅) and PFBC (P₁₋₅), for different influent conditions the resulting fractal dimensions in two dimensions (D₂) and in three dimensions (D₃) with their standard error (SE), coefficient of determination (R²), population (N), mean characteristic length (l), mean projected surface area (A_s), mean circularity (C) and settling velocity (W_s) ratio are mentioned in Table 1.

Table 1

View Large

Fractal analysis data were highly correlated for all cases and presented low error values ⁠. The range of D₂ values as well as D₃ values for flocs formed with PACl is in good agreement with the respective reported values (Chakraborti et al. 2007; Lv et al. 2021). The average D₂ and D₃ values for CC were 1.93 and 2.86 respectively while the average D₂ and D₃ values for PFBC were calculated to be 1.90 and 2.79 respectively. It may be noted that the fractal dimensions for CC were higher than PFBC by a small margin. Generally, D₃ is < 2 for sweep flocculation mechanism, while the given D₃ values for both technologies indicate inter-particle bridging to have been the dominant mechanism for floc formation. From the D₃ and mean l values, it may also be inferred that reaction-limited aggregation (RLA) process was followed rather than diffusion-limited aggregation (DLA) process (Chakraborti et al. 2007). Therefore, more consolidated chain-like aggregates with deeply embedded particles were formed (Srivastava et al. 2021a).

It may be noted that while the correlation between W_s and l is direct, the three-dimensional fractal dimension (D₃) has a positive whereas the two-dimensional fractal dimension (D₂) has a negative exponential correlation with W_s. The color-coded scales shown in Table 1 allow ready comparison of the relative magnitudes of l, D₂ and D₃ of CC and PFBC flocs for mapping the combined effect on the settling velocity. On taking the weighted average on the basis of the population of flocs (N), the W_s for PFBC was found to be about 37% higher than the value for CC. Therefore, the results indicate that the overall size and structure of the flocs generated in a PFBC were such that they were more amenable to settling as compared to conventionally formed flocs.

Scanning electron microscopy showed that the sludge of PFBC (Figure 7(a)) was more compact than CC sludge (Figure 7(b)). Dissimilitude in the floc formation can be clearly observed at high magnification. This marked difference in the morphologies of the sludge from PFBC and CC may be attributed to the relatively higher concentration of clayey turbidity from kaolin due to the presence of the blanket in PFBC resulting in formation of surface features on the floc. For PFBC flocs, the main voluminous surface coverage is by kaolin, and Al₁₃ aggregates adsorbed on the surface of the kaolin particles extend to attach other particles in the interior of PACl-Al₁₃ agglomerate. Conversely, the relative concentration of NOM is higher in CC flocs and sludge. The results visually reveal the second level of the structure proposed by Verrelli et al. (2009) which had predicted the adherence of a thick ‘fluffy’ layer of material, perhaps NOM, to the surface of the otherwise hard and small particulates.

Comparative assessment of a PFBC and CC was carried out under a constant NOM influx for a range of influent turbidity conditions with PACl used as a coagulant. Mechanisms like contact flocculation and enmeshment were dominant in the conditions offered by the dense suspension of flocculated particles in fluidized state, which resulted in better colloidal destabilization and capture in a PFBC. The cyclical contraction and expansion also helped in significant compaction of sludge. Average total residual aluminum level was brought below the 200 ppb mark and an ∼ 68–70% reduction in TOC was achieved by both CC and PFBC. NOM was observed to readily interact with kaolin present in both influent as well as blanket and also raise a significant aluminum demand. The demand for aluminum was satisfied foremost by PACl before it could serve as an effective coagulant and their interaction resulted in the formation of predominantly soluble alumino-organic complexes. This led to a much higher average percentage of dissolved residual aluminum (⁠⁠) for CC. However, PFBC was less affected (average dissolved residual aluminum ⁠) since the blanket acted as a buffer and provided nucleation sites which aided precipitation. The flocs formed in a PFBC were on average larger by about 31% and their surface structure was observed to be markedly distinct from the flocs formed in CC. Therefore, the average settling velocity of PFBC flocs, computed on the basis of resulting fractal dimensions was ∼ 37% higher than for the CC flocs.

We acknowledge the support of Materials Research Centre (MRC) at Malaviya National Institute of Technology Jaipur, India in carrying out nanoparticle size analysis (nano-PSA), zeta potential analysis, FT-IR spectroscopy and SEM at their facilities.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.