Introduction

Coagulation, followed by sedimentation, is commonly used in drinking water treatment due to its removal of turbidity, viruses, heavy metal, algae, and precursor for disinfection by-products (Hu et al. 2006; Mertens et al. 2012; Shirasaki et al. 2014; Zheng et al. 2015). Floc properties, such as floc size and structure, have a profound impact on solid/liquid separation process, largely determining coagulation efficiency (Yu et al. 2009). Although coagulation process in water treatment works (WTWs) should avoid floc breakage as much as possible, breakage and re-growth of flocs occur in practice since regions of high shear force are prevalent (Rossini et al. 1999; McCurdy et al. 2004; Jarvis et al. 2005). Numerous studies, therefore, have been done to explore floc growth, breakage, and re-growth, and it has been verified that the flocs after breakage are limited re-growth except for the flocs formed under charge neutralization mechanism (Yu et al. 2010).

Factors, including mixing speed and duration, coagulant dosage and species, solution pH, and initial particle concentration, notably influence the floc characteristics during coagulation. Some studies with respect to the effect of mixing speed on floc properties were reported in recent years. It was shown that the increase of slow mixing intensity before breakage resulted in larger steady-state floc size before breakage and had less of an impact on floc size after breakage (Yu et al. 2011). Besides, higher rapid mixing speed during floc breakage period contributed to smaller floc size after breakage and re-growth, leading to significant decrease of turbidity removal efficiency (Yukselen and Gregory 2002; Zhao et al. 2011). However, previous work indicated that larger particles generally had better sedimentation property than smaller particles with similar density, implying the combined effect of floc size and structure on coagulation efficiency (Boller and Blaser 1998). Thus, in absence of floc structure, floc size cannot fully interpret the effect of slow mixing speed before breakage and rapid mixing speed during breakage on floc breakage and re-growth process. Thus, the investigation of floc structure in combination with floc size has significant indications for further understanding of underlying mechanism during floc breakage and re-growth process.

For floc breakage and re-structuring, few researchers paid attention to the influence of slow mixing speed after breakage on the properties of re-grown flocs. Li et al. (2007) stated that with the same slow mixing speed before and after breakage, an increase in slow mixing speed contributed to the re-formation of flocs with smaller size and more compact structure. This has obvious difference with the study of Yu et al. (2011) as mentioned above, where floc size after breakage and re-growth was nearly the same with different slow mixings before breakage. Thus, research on slow mixing speed after breakage is necessary and meaningful for floc breakage and re-growth process.

In this study, mixing speed before breakage, during breakage, and after breakage was separately varied to investigate the effect of mixing intensity on floc properties after breakage and re-growth. The mechanism involved was also discussed by characteristic analysis of floc size and structure. This topic would be beneficial to improve knowledge of coagulation process due to inevitable breakage and re-growth of flocs in WTWs.

Material and methods

Suspension and coagulant

Kaolin clay (Tianjin, China) was used as synthetic model suspension. Kaolin clay was prepared by dissolving 100.0 g of kaolin clay in 1 L deionized (DI) water with high-speed magnetic stirring for 24 h. After 1-h sedimentation, the top 800 mL was decanted as stock suspension of kaolin clay. The mean particle size was 4 μm, determined by a particle counter (2200PCX, HACH, USA).

For the flocculation tests, the stock solution was diluted in the tap water of Harbin, China, which has total hardness of 160 mg/L as CaCO3 and a pH of around 7.5 ± 0.3. A small amount humic acid was added in the stock kaolin suspension since humic acid could decrease the disturbance of divalent metal ions such as Ca2+ (Yukselen and Gregory 2004). The final suspension had a kaolin clay concentration of 95 mg/L and humic acid concentration of 2 mg/L. The turbidity determined by an online turbidimeter (MICROTOL, HF, USA) was around 90 ± 0.5 nephelometric turbidity unit (NTU).

Polyaluminum chloride (PACl) (Tianjin, China) with a basicity value of 75 % was used as coagulant. The PACl solution was prepared at a concentration of 1 % w/w with DI water. The PACl solution mentioned above was freshly prepared every day to prevent aging effects and used directly without prior dilution. All reagents used in this work were of analytical grade.

Apparatus

The flocculating reactor in this study was a modified version of jar test reactor, which was a rectangular stirred tank (5 mm thick plexiglas) with the length of bottom D = 150 mm and the height of liquid H = D = 150 mm, and filled with 3.4 L of testing water sample. To provide the stirring condition, a R1342-type impeller (IKA, Germany) with a diameter (d) = 50 mm was installed in the reactor and the center of the impeller was positioned at C = H/3 from the tank bottom. This mixing system was successfully used in our previous research (Nan et al. 2009; He and Nan 2012). The schematic diagram is shown in Fig. 1.

Fig. 1
figure 1

Schematic diagram of flocculation experiments with in situ recognition technique. 1 Automated stroboscopic lamp, 2 flocculation reactor, 3 impeller, 4 motor for digital agitator, 5 online turbidimeter, 6 peristaltic pump, 7 CCD camera, 8 personal computer

Full size image

In order to calculate floc size and structure, from the moment of PACl addition, a digital charged coupled device (CCD) camera (SVS-VISTEK GmbH, Germany) was used to capture digital images of particles and a process control and image processing (PCIP) software package (FMans 10, China) was used to determine floc geometrical parameters. The in situ recognition system, which comprises digital CCD camera and PCIP software package, has been widely used by our lab and other researchers since there is no sample handing to disturb floc during measurements (Chakraborti et al. 2003; He et al. 2012; Nan et al. 2016).

Procedure

For the flocculation tests, the floc growth, breakage, and re-growth were preformed with 3.4 L of suspension. PACl solution was added at optimal dosage in all cases. The procedure to ensure optimal dosage was same with our previous studies (He et al. 2012; Yao et al. 2015). In order to investigate the effect of different slow mixing speeds before breakage, the suspension was stirred at 400 rpm for 90 s after the addition of coagulant, followed by a slow stir phase at rates from 50 to 110 rpm for 20 min, respectively. After this, the suspension was mixed at 300 rpm for 1 min to break flocs and then reduced to 90 rpm for 10 min to allow floc re-growth. For various rapid mixing speeds to breakup flocs, coagulation experiments were carried out as follows: a rapid mix at 400 rpm for 90 s, followed by a slow stir at 90 rpm for 20 min. Then the mixing speed increased to 250, 300, 350, and 400, respectively. After breakage period, a slow mixing speed at 90 rpm for 10 min was carried out to re-grow flocs. Coagulant experiments involving different slow mixing speeds after breakage were started at 400 rpm for 90s, then 90 rpm for 20 min, 300 rpm for 1 min to break the flocs, and then at rates from 50 to 110 for 10 min to allow floc re-formation. For all tests, 20-min sedimentation after floc re-growth was performed to measure residual turbidity, which reflects the coagulation efficiency. All experiments in this study were repeated three times. To better understand, details of experimental setup are shown in Table 1. In addition, the relationship of G values and mixing speeds, which was derived from the study of Spicer and Pratsinis (1996), is shown in Table 2.

Table 1 Details of experimental setup
Full size table
Table 2 Hydraulic agitation values of each stage during coagulation
Full size table

Floc size and fractal dimension

The floc size and fractal dimension were calculated accurately using in situ recognition system. In order to obtain floc properties precisely at the corresponding moment, 12 images near this moment were used to calculate.

The size of flocs captured in the image is reflected by equivalent diameter which is closely related to floc projected area (A):

$$ d={\left(\frac{4A}{\pi}\right)}^{\frac{1}{2}} $$
(1)

where d is the floc size.

Fractal theory can well describe complicated floc aggregate structure as a new quantitative method since the flocs formed during coagulation have fractal characteristic (Jiang and Logan 1991; Zhu et al. 2009). In general, the floc structure can be characterized by two-dimensional fractal dimension. The two-dimensional fractal dimension, D 2, is the slope of a plot of projected area against the characteristic length of the aggregates on a log-log scale according to

$$ A\propto {l}^{D_2} $$
(2)

where l is the characteristic length of the aggregates (Logan and Kilps 1995; Chakraborti et al. 2000; Yu et al. 2010). In fact, the D 2 reflecting the projected floc morphology in the two-dimensional image can only take value between 1 and 2. A larger D 2 value indicates that the floc is more compact, while floc with highly branched and porous structure is formed when D 2 is lower. In particular, D 2 = 2 imply a spherical floc and D 2 = 1 signify a linear structure.

Results

Effect of mixing speed on coagulation efficiency

The changes of residual turbidity against rapid mixing speed during breakage and slow mixing speed before and after breakage are presented in Fig. 2. It was found that the residual turbidity elevated to 3.82 NTU as the rapid mixing speed during breakage increased from 250 to 400 rpm, implying that the re-grown flocs were more difficultly removed by sedimentation. In addition, the residual turbidity presented a slight rise with larger mixing speed before breakage adopted. This was similar with the finding of Yu et al. (2011), who studied kaolin flocs without humic acid. It indicated that lower mixing speed before breakage can gently improve coagulation efficiency after breakage and re-growth independent of humic acid addition in suspension. While the residual turbidity markedly decreased with higher mixing speed after breakage, which was notably different to the effect of mixing speed before and during breakage, it can be concluded that increasing mixing speed of each phase contributed to different variations of residual turbidity, perhaps as a result of the properties of re-grown flocs induced by different mechanisms.

Fig. 2
figure 2

The effect of various mixing speeds before breakage, during breakage, and after breakage on residual turbidity. For example, the slow mixing speed before breakage was performed at 50, 70, 90, and 110 rpm, respectively

Full size image

Effect of mixing speed on the evolution of floc size

To investigate the growth, breakage, and re-growth of flocs with various mixing speeds, the effects of rapid mixing speed during breakage and slow mixing speed before and after breakage on average size of flocs are shown in Fig. 3. As shown in Fig. 3a, for slow mixing speed before breakage, floc average size before and after breakage reached maximum at 50 rpm and decreased with mixing speed before breakage up to 110 rpm. The variation of floc average size before breakage was consistent with the study that larger slow mixing speed generated smaller floc size (Biggs and Lant 2000; Soos et al. 2008; Zhan et al. 2011). For average size of flocs after breakage, the result had a slight conflict with the study of Yu et al. (2011), who found that the flocculation index (FI) of flocs with different slow mixing after breakage was nearly the same. However, the difference of FI about 0.001 which was neglected in their study, indicated about 14 μm gaps for average size of flocs and the distinction between the maximum and the minimum of floc average size after breakage was only 19 μm in our research. Thus, there was no apparent discrepancy on slow mixing effect, although 2 mg/L of humic acid was added into the suspension in this paper. Figure 3b shows that floc average size after breakage markedly decreased with the increase of rapid mixing during breakage. This phenomenon was in accordance with other studies (Wang et al. 2009; Xu et al. 2010). The slow mixing speed after breakage, which was barely noticed, also influenced floc average size to some extent (Fig. 3c). It was found that higher slow mixing after breakage resulted in smaller average size of flocs after breakage and re-growth.

Fig. 3
figure 3

Temporal evolution of floc average size during floc growth, breakage, and re-growth process under different mixing speeds: a before breakage, b during breakage, and c after breakage

Full size image

In order to further quantify the floc re-growth process, the parameters including re-growth time, rate in the re-growth region, and floc size in the new steady-state region were used. The re-growth rate of flocs can be calculated as follows (Wang et al. 2009):

$$ Re\hbox{-} \mathrm{growth}\kern0.62em \mathrm{rate}=\frac{\varDelta \mathrm{size}}{\varDelta \mathrm{time}} $$
(3)

where the floc size was the mean value of floc average size during the whole steady-state region. The re-growth rate, re-growth time, and floc size for different mixing conditions are summarized in Table 3. It can be found that for different slow mixings before breakage, the floc re-grew to larger size with longer re-growth time and higher re-growth rate as a result of decreasing mixing speed. Under the condition of various rapid mixing during breakage, the floc size reached the largest value 115 μm at 250 rpm with fastest re-growth rate and shortest re-growth time. Although an increase in mixing speed dramatically prolonged the re-growth time to steady state, the floc size at steady state gradually reduced due to lower re-growth rate. While as the slow mixing after breakage increased, the re-grown flocs at steady state had smaller size after shorter re-growth time and the re-growth rate for different mixing speeds was nearly the same.

Table 3 The properties of re-grown flocs under various mixing speeds of each stage
Full size table

Effect of mixing speed on size distribution of re-grown flocs at new steady state

After floc breakage and re-growth, the average sizes of flocs are changed with varying mixing conditions. In order to further explore the influence of mixing condition on re-grown flocs at new steady state, the size distributions of flocs with different mixing conditions after re-growth are shown in Fig. 4. As expected, with higher slow mixing speed before breakage, the size distribution of re-grown flocs was narrower (Fig. 4a). When the mixing speed increased from 50 to 110 rpm, the frequency of flocs smaller than 100 μm elevated from 49 to 59 %. Figure 4b shows that increasing mixing speed during breakage facilitated the re-formation of flocs with smaller size. As the mixing speed during breakage exceeded 300 rpm, the flocs between 400 and 500 μm completely disappeared in floc size distribution at new steady state. Likewise, a similar variation could be seen in Fig. 4c, i.e., the floc size distribution was shifted toward smaller size. It indicated that the increase of mixing speed led to smaller size of re-grown flocs regardless of which stage the mixing speed belonged to.

Fig. 4
figure 4

Size distributions of re-grown flocs under different mixing speeds: a before breakage, b during breakage, and c after breakage

Full size image

Effect of mixing speed on re-grown floc structure at new steady state

The impact of mixing speeds before breakage, during breakage, and after breakage on re-grown floc structure at new steady state is presented in Fig. 5. It can be seen that the fractal dimension of re-grown flocs presented a similar response to increasing mixing speed in all cases. The distinction was that the change of mixing speeds in each phase had different influence extent on re-grown floc structure. As the slow mixing speed before breakage increased in the range of 50–110 rpm, the D 2 value of re-grown flocs increased from 1.79 to 1.83. While for the slow mixing speed after breakage, the D 2 value of re-grown flocs was in the range of 1.74–1.84. In addition, the D 2 value ranged from 1.76 to 1.87 when the rapid mixing speed during breakage phase increased from 250 to 400 rpm. Consequently, mixing speed also intensively affected re-grown floc structure besides re-grown floc size.

Fig. 5
figure 5

The effect of various mixing speeds before breakage, during breakage, and after breakage on re-grown floc fractal dimension. For example, the slow mixing speed before breakage was performed at 50, 70, 90, and 110 rpm to explore the re-grown floc fractal dimension, respectively

Full size image

Discussion

The results presented in this study clearly show that mixing speeds before breakage, during breakage, and after breakage influence the properties of re-grown flocs, resulting in different coagulation efficiency. Interpreting the effect of mixing conditions can give important indications for further understanding of floc breakage and re-growth process.

It can be found that the broken flocs could not re-grow to their previous sizes before breakage in all cases (Fig. 3). During the coagulation process, the broken flocs have complete recoverability under charge neutralization, whereas the broken flocs have poor re-growth ability under sweep or adsorption (Yu et al. 2010). For PACl, coagulation was dominated by charge neutralization at low dosage and by sweep at high dosage (Lin et al. 2014). It thus seems reasonable to infer that the dominated mechanism during coagulation was sweep. Under sweep mechanism, PACl could not completely neutralize the negative charge of particles (Xu et al. 2010). That is, the flocs before breakage were the same charged. It indicated that the broken flocs were not neutral. For broken flocs, surface charge would lead to mutual repulsion between each other (Lin et al. 2013). Therefore, the charge of broken flocs would hinder the re-growth of broken flocs, resulting in incomplete re-growth of broken flocs. In addition, the internal bonds of flocs were formed by physical interactions (for example, electrical effects) under charge neutralization, whereas chemical bonds also play an important role for floc formation under sweep mechanism. The breakage and re-growth of flocs would contribute to less reactive sites on broken flocs, resulting in fewer chemical bonds in broken flocs (Yukselen and Gregory 2002). This is another reason why the broken flocs could not fully re-grow.

With the increase of mixing speeds before breakage, the flocs had smaller average size before breakage (Fig. 3a). It was consistent with previous studies, which proposed that flocs with smaller size and more compact structure would be generated as slow mixing speed was elevated (Wang et al. 2011; Bubakova et al. 2013; Xu et al. 2014). Although the breakage condition was identical, larger flocs had different breakage extent with smaller flocs. One possible explanation is that smaller flocs were susceptible to be entrained within eddies, indicating slight influence of breakage, while larger flocs were greatly affected by breakage due to microscale eddies (Jarvis et al. 2005). This explains why with larger average size of flocs before breakage a smaller average size of flocs after breakage was observed. Since larger flocs broken into a greater number of smaller flocs increased the amount of newly exposed floc surface, smaller broken flocs could re-grow to larger size with larger re-growth rate (Table 3). In general, floc strength was considered to be positively related to floc structure since increasing number of internal bonds held aggregate together (Wang et al. 2011; He et al. 2012). Thus, although the slow mixing speed before breakage led to different floc structure, the same breakage condition indicated that the broken flocs after breakage had similar strength to withstand shear force, that is, the broken flocs had similar floc compaction after breakage. Under the same slow mixing speed after breakage, the floc structures at steady state were similar (Fig. 5).

As the rapid mixing speed during breakage was elevated, the average size of flocs tended to decrease. This phenomenon presumably occurs because of more severe breakage induced by larger shear force (Xu et al. 2014). Due to the same slow mixing speed and duration after breakage, the severe breakage flocs could only re-grow to smaller sizes at new steady state (Sun et al. 2011; Xu and Gao, 2012). For the same flocs before breakage, an increase in shear force during breakage indicated that the flocs should have larger strength to withstand shear and thus, the broken flocs were more compact after breakage. This can fully explain why severe breakage resulted in higher D 2 at new steady state.

With the same slow mixing before breakage and rapid mixing during breakage, the flocs after breakage had identical size and structure. Thus, the slow mixing speed after breakage, which dominated the re-structuring and re-arrangement of broken flocs during floc re-growth phase, intensively affected floc properties at new steady state. With lower stir after breakage, the function of re-structuring and re-arrangement was limited and thus generated larger flocs with looser structure. As slow mixing after breakage increased, the re-structuring and re-arrangement were enhanced, resulting in smaller flocs with more compact structure. Thus, the average size of re-grown flocs was in positive correlation with slow mixing after breakage, and the result for the fractal dimension was just the reverse. This proved that although the floc re-growth mechanism after breakage was a re-agglomeration of broken flocs and the floc formation before breakage was owing to sweep, the relation between the applied shear rate and the steady-state floc size was nearly the same.

Overall, the increase of shear rate contributed to the re-formation of flocs with smaller size and more compact structure irrespective of which phase the shear rate belong to. However, an interesting phenomenon was that the residual turbidity presented different changes with increasing shear rate in each phase. For example, the residual turbidity tended to decline with lower mixing speed before breakage, whereas the residual turbidity showed a rise as the mixing speed after breakage reduced. This can be explained based on the floc properties at new steady state. Boller and Blaser (1998) claimed that flocs with larger size and more compact structure had better settleability, leading to lower residual turbidity. Since the floc size decreased with increasing shear rate and the results for the fractal dimension was just the reverse, the smaller floc size and more compact floc structure had the opposite effect on settling rate. As applied shear rate before breakage and during breakage increased, the improvement of setting rate by higher floc density was lower than the inhibition of setting rate by smaller floc size. While the increase of floc compactness had greater influence on settling ability than the decrease of floc size as slow mixing after breakage was elevated, leading to lower residual turbidity.

Conclusions

The role of mixing speed before breakage, during breakage, and after breakage on floc properties and coagulation efficiency was investigated in a complex system of particles/colloid/soluble organic matter. In all cases, the increase of mixing speed facilitated the re-formation of flocs with smaller size and more compact structure. However, the mixing speed of each phase had different influence extents on floc size and structure at new steady state, which played an important role in solid/liquid separation efficiency. Thus, with increasing mixing speed before breakage, the residual turbidity tended to slightly rise. With increasing mixing speed during breakage, the residual turbidity was obviously elevated. However, it was notable that the residual turbidity presented a significant decrease as the mixing speed after breakage increased, which was markedly different from the effect of mixing speed before breakage and during breakage.

In view of the results in this study, it would improve the understanding of floc breakage and re-growth under various mixing conditions. According to the influence of different mixing conditions on floc properties and coagulation efficiency, some optimization of mixing conditions may further enhance coagulation efficiency since breakage and re-growth of flocs always exist in actual WTWs. This aspect should be explored in more detail in future actual WTWs.