Anda di halaman 1dari 11

Pergamon

Wat. Res. Vol. 28, No. 3, pp. 559-569, 1994 Copyright 1994Elsevier ScienceLtd Printed in Great Britain. All rights reserved 0043-1354/94$6.00+ 0.00

DYNAMICS OF COAGULATION OF KAOLIN PARTICLES WITH FERRIC CHLORIDE


HSIAO-WEICHING1, THEODORES. TANAKA and MENACHEMELIMELECHI*O 2 IDepartment of Civil and Environmental Engineering, University of California, Los Angeles, CA 90024-1593 and 2Metropolitan Water District of Southern California, 700 Moreno Avenue, La Verne, CA 91750, U.S.A.
(First received October 1992; accepted in revised form June 1993)

Abstract--An optical monitoring technique is used to investigate the dynamics of coagulation of kaolin suspensions with ferric chloride. Particular attention is given to the effects of coagulant dose, solution pH and mixing intensity on coagulation dynamics. Results show that this monitoring method provides valuable information on the dynamics of aggregates in coagulation with Fe(III) salts.
Key words--coagulation, ferric chloride, metal salt coagulant, clay particles, monitoring, coagulation dynamics, Fe(III) solution chemistry

INTRODUCTION Coagulation is an essential process in the overall solid-liquid separation scheme in water and advanced wastewater treatment. Hydrolyzing metal salts, such as aluminum and ferric coagulants, are widely used as primary coagulants to promote the formation of aggregates. For optimal performance of subsequent solid-liquid separation processes, it is imperative that aggregates of a certain size, strength and density be formed in the coagulation process (e.g. Ramaley et al., 1981; O'Melia, 1985). In coagulation with hydrolyzing metal salts, these properties are largely determined by coagulant dose, solution pH and mixing intensity (Amirtharajah and O'Melia, 1990; Dentel, 1991). The state of aggregation and dynamics of aggregates in a coagulating suspension can be measured by particle counters (e.g. Lawler et al., 1983; Francois, 1988; Amirtharajah and O'Melia, 1990). However, because of breakage of aggregates and formation of gelatinous hydroxide precipitates (Treweek and Morgan, 1977; Lawler et al., 1983), serious difficulties arise when particle counters are used to measure the size of aggregates formed in coagulation with hydrolyzing metal salts. In addition, particle counters are expensive and are not suitable for obtaining immediate and continuous information about the state of aggregation and dynamics of aggregates in a coagulating suspension, A simple but sensitive optical technique has recently been developed to monitor the state of aggregation of colloidal suspensions (Gregory, 1985). The technique is based on measurement of turbidity fluctuations in flowing suspensions. When a flowing suspension is illuminated by a narrow light beam, the *Author to whom all correspondence should be addressed,

transmitted light measured by a detector fluctuates about a mean level. It can be shown (Gregory, 1985) that the ratio of the root mean square of the fluctuating signal to the mean level is roughly proportional to the size of the aggregates flowing through the detector and to the square root of their concentration. This ratio, referred to as flocculation or coagulation index, can give immediate information on the state of aggregation of particle suspension over the entire period of coagulation. The technique described above has been successfully applied to study the dynamics and mechanisms of coagulation of dilute and concentrated suspensions with polymeric coagulants (e.g. Gregory, 1988; Gregory and Lee, 1990; Li and Gregory, 1991; Gregory and Li, 1991). In this paper, it is shown that this technique can be applied to study the dynamics of coagulation with metal salt coagulants (ferric chloride). It is further demonstrated that this method can give valuable insights into the mechanisms of particle aggregation with ferric chloride. BACKGROUND Chemistry o f Fe(llI)salts When Fe(III) salts are dissolved in water, the metal ion hydrates, coordinating six water molecules and forming an aquometal ion, Fe(H20)~ + . The aquometal ion can then hydrolyze and form monomeric and polymeric ferric species, the formation of which is highly pH dependent (Stumm and Morgan, 1981; O'Melia et al., 1989; Morel and Hering, 1993). It is possible to describe the formation of several hydrolysis species that are positively charged: FeOH 2+, Fe(OH)~', Fe2(OH)~ + and Fe3(OH)~+; neutral: Fe(OH); and negatively charged: Fe(OH)~-.

559

560

HSlAO WEI CHING et

al.

In general, the hydrolysis reactions of Fe(III) in aqueous solution can be written as ~3~ ~.) Fex(OH),. - + y H (1) At any pH, the maximum dissolved concentration of Fe(III) in equilibrium with the hydroxide solid is determined by the solubility of the solid phase, in this case amorphous ferric hydroxide, am-Fe(OH)3(s), and by the extent of formation of monomeric and polymeric hydrolysis species in solution. The solubility diagram of Fe(III), such as that shown in Morel and Hering (1993) or Johnson and Amirtharajah (1983), is very useful to describe the concentrations of the dissolved ferric species as a function of pH, at equilibrium with am-Fe(OH)3(s). It can be seen from the solubility diagram that amorphous ferric hydroxide is least soluble at a pH close to 8. Ferric adsorption onto particle surfaces can affect both the speciation of ferric in solution and the surface properties of the particles. Adsorption of ferric onto clay particles can occur by formation of surface complexes between ferric and surface hydroxyl groups of the mineral. For kaolinite, cation adsorption can involve both ion-exchange (particularly at low pH) and surface complex formation. This latter reaction is thought to occur predominantly at the basal gibbsite and edge surfaces of the clay mineral (Stumm, 1992). Reaction with monomeric ferric hydrolysis species may be written in a general form as = SOH + Fe(OH)~3-~)+ = -=SOFe(OH)~31n) + H20 (2) xFe3+yH20=

in sorption density of the precipitating cation, occurs below saturation in the bulk solution. Thus, amFe(OH)3(s ) may be formed at the surface of mineral particles and affect their surface properties, even when the solubility product for am-Fe(OH)3(s ) is not exceeded in the bulk solution. It should be emphasized that the chemistry of ferric adsorption and precipitation in natural waters is much more complicated because of the presence of humic substances. Interaction of humic substances with surfaces and dissolved ferric species can markedly influence coagulation processes. It is well known that actual ferric dosages in water treatment are controlled by the concentration of dissolved natural organic matter (Amirtharajah and O'Melia, 1990). These aspects, though beyond the scope of this paper, should be considered in actual treatment practices.

Principles of turbidity fluctuations


The principles of the optical technique used in this study and its application to coagulation processes have been described previously (e.g. Gregory, 1985; Gregory and Nelson, 1986). A short description of this technique is given below. When a flowing suspension is illuminated by a narrow light beam, the transmitted light intensity, monitored by a photo detector, fluctuates randomly about some mean value. The output from the photo detector consists of a steady (D.C.) signal and a fluctuating (A.C.) component. The D.C. value is a measure of the average transmitted light intensity and depends on the turbidity of the suspension. The fluctuating (A.C.) component is a result of random variation in the number of particles in the illuminated volume. It can be shown (Gregory, 1985) that the root mean square (RMS) value of the fluctuating (A.C.) signal is related to the average number concentration and the size of the suspended particles. In practice, it is convenient to divide the RMS value by the steady D.C. value to give a dimensionless term R = RMS/D.C. With the use of this ratio, one can avoid the effects of optical surface fouling and electronic drift. It can be shown (Gregory and Nelson, 1986) that, for a heterodisperse suspension, the ratio R can be expressed as

where = S indicates a surface species. Note that the surface-bound ferric can also undergo acid-base reactions. In the surface complexation model, the surface charge of the mineral is attributed to the presence of charged surface species ( = S O H f and = SO-, in the absence of any adsorbing species other than protons), Thus, formation of ferric surface complexes can alter surface charge (Stumm and Morgan, 1981; Stumm, 1992). Adsorption of polymeric ferric species, especially highly charged species, can dramatically affect surface charge (Tang and Stumm, 1987a, b; O'Melia et al., 1989). In practice, however, the adsorption of polymeric species may be indistinguishable from precipitation of am-Fe(OH)3(s) at the mineral surface. With increasing ferric adsorption and surface coverage, the acid-base properties of the surface will be increasingly characteristic of amFe(OH)3(s ). A continuum between surface adsorption (complexation) and precipitation is described by the surface precipitation model. This model allows for the formation of a surface phase with a composition that varies continuously between that of the original solid and that of a pure precipitate of the adsorbing cation (Farley et al., 1985). Formation of the precipitate at the mineral surface, evidenced by a gradual increase

( L ' ~ L2
R = \AJ

(Y.NiC 2)~/2

(3)

where L is the optical path length, A is the effective cross-sectional area of the light beam and Ni and C~ are the number concentration and scattering crosssection of particles of size i, respectively. This equation demonstrates that the fluctuating signal depends on the square root of particle concentration and on the first power of the scattering cross section. The latter is highly dependent on the size of the suspended particles.

Dynamics of coagulation of kaolin An analysis o f the term (Y,NiC 2)1/2 in equation (3) reveals that smaller particles have a negligible effect on R, and that, in a coagulating suspension, the larger aggregates have a dramatic influence on R (Gregory and Nelson, 1986). This means that as coagulation progresses, the value o f R increases. Although the ratio R does not provide quantitative information on aggregate size, the relative increase in the R value is a useful indicator of the degree of coagulation. For a given suspension, it can be assumed that larger R values imply larger aggregate size. This ratio will be referred to as the "coagulation index" in this study. = ~ ~ iili!i ili iiliii ii Coagulant

561

i i l~ - ~ m ~ ~ J, _ i -

Suspension

MATERIALS AND METHODS Synthetic kaolin suspension


A stock suspension of clay particles was prepared by dispersing purified kaolin in powder form (Fisher Scientific, Pittsburgh, Pa, U.S.A.) in deionized water to a concentration of 10 g/l. The test suspensions were prepared by diluting the stock suspension to a desired concentration. All test suspensions contained 1 mM of sodium bicarbonate to provide alkalinity. The size distribution of the kaolin suspension was measured by a particle counter (Multisizer II, Coulter Electronics, Hialeah, Fin, U.S.A.) with a 30-#m aperture tube (effective range 0.6-18 #m). The particles in the suspension had a mean volume diameter of 2.1 # m with a relatively narrow size distribution. The number and volume distributions of the clay particles, presented in the form discussed by Amirtharajah and O'Melia (1990), are shown in Fig. 1. Electrophoretic mobilities of the particles were measured by a Lazer Zee Model 501 apparatus (Pen Kern Inc., Bedford Hills, N.Y., U.S.A). The clay suspensions used in the electrophoretic mobility measurements were treated similarly to those used in the coagulation experiments, and the results were easily reproduced. It was found that the particles are negatively charged at all pH values examined (from 3.4 to 10.0).

Peristaltic Pump Fig. 2. Schematic description of the experimental setup for monitoring the dynamics of coagulation by the PDA 2000. coagulant solution of 0.3 M was prepared from the concentrated stock solution before every set of experiments. In all coagulation experiments, ferric chloride was added directly from the 0.3 M coagulant solution. Preliminary experiments showed that the dosage of coagulant needed to neutralize the negative charge of the particles is larger when more dilute ferric chloride stock solutions are used. The 0.3 M coagulant solution can be classified as type A based on the analysis of Tang and Stumm (1987a, b). They also reported that ferric coagulant solutions with concentrations higher than 0.1 M are more effective than those of lower concentration.

Experimental setup for monitoring coagulation dynamics


A light scattering instrument with a flow-through detector (PDA 2000, Rank Brothers Ltd, Cambridge, U.K.) was used to monitor the dynamics of coagulation. A schematic diagram of the experimental setup is shown in Fig. 2. The coagulation vessel was a l-liter beaker, mixed by a 762 x 254-mm rectangular flat blade located 45 mm above the base of the beaker. The blade was driven by an adjustable speed motor via a thin spindle, centrally located in the beaker. A 3-ram (i.d.) Tygon tube, located 30 mm below the liquid level, conveyed the suspension to the flow-through detector of the PDA 2000 monitor. The suspension was driven by a peristaltic pump (Masterflex, Cole Palmer Instrument Comp., Chicago, Ill., U.S.A.) located downstream of the monitor. The flow rate in the conveying tube was 22 cm3/min, and the time for the sample to pass from the coagulation beaker to the flow-through detector was approx. 10 s. The above conditions (i.e. the flow through the sampling tube and the location of the sampling tube) were strictly maintained throughout all the experiments, so that the coagulation index dynamics behaviors under various chemical and physical conditions could be compared to each other.

Preparation of ferric coagulant solution


A concentrated stock solution of ferric chloride was prepared by dissolving analytical reagent-grade FeCI 3 6H20 (Fisher Scientific, Pittsburgh, Pa, U.S.A.) in deionized water to a concentration of 2.25 M. A fresh 14 ,-, ,'~ u e 6
4

5 ,/ . ~ 4 ',-, ~ ,.~ ,-~ 3 ~=. ~ o 2 ~


z

12 10 8 ,

"~
2 0 . . . . . . ,a ' " ~ 0.01 0.1 1 dp (p.m) :. . . . . . 10

1
0 100

Coagulation experiments
A l-liter kaolin suspension, with 10 -3 M NaHCO 3 and a solid concentration of 50 mg/l, was used in the coagulation experiments. The experiments were conducted at room temperature (average of 20C). In these experiments, the target pH was adjusted by adding predetermined amounts of KOH or HCI to the suspension, followed by the addition of ferric chloride with a micropipette, directly from the 0.3 M ferric coagulant solution. Rapid mixing (for 2 min) by magnetic stirring at high speed was initiated after the addition of ferric chloride. At the end of the rapid mixing, a small sample was taken by a 25-ml pipette for electrophoretic mobility measurements. Following rapid mixing, the

Fig. 1. Number and volume distributions, dN/d(log do) (solid line) and d V/d(log de) (broken line) as a function of log d0, of the kaolin suspension used in this work (N is the number of particles per unit volume of suspension, V the volume of particles per unit volume of suspension and dp is the particle diameter),
WR 28/3--E

562

HS]AOWEI CHING et al. 40 - - o- - pH 6.0 ~ ~ "-" .~ ~" ~ 35 ---q3----pH 7.8 30 25 ,

suspension was mixed gently by a single flat blade at a desired rpm for a period of 15 min. The state of aggregation of the clay particles was monitored continuously by the PDA 2000, as described above. The suspension was allowed to settle for 15min after the slow mixing. Samples for residual turbidity measurements (Ratio/XR Turbidimeter, Hach Chemical Comp., Ames, Iowa, U.S.A.) were then taken by a 25-ml pipette, from about 25 mm below the water level.

RESULTS AND DISCUSSION


The results of our experiments are presented in terms of the effect of ferric chloride concentration and solution pH on (1) the electrophoretic mobility of particles, (2) the residual turbidity after coagulation and settling; and (3) the dynamics of the coagulation index during the coagulation process. Furthermore, the use of the optical technique to study the effect of mixing intensity on coagulation dynamics is demonstrated.

~
~ "4 ~,

2O
15 10

', ~

o, 5 0 0 ~ , , , 10 20 I , , ~, I , , v t . . . . 30 40 50 60

F e r r i c C h l o r i d e (ttM) Fig. 4. Residual turbidity (after settling) as a function of ferric chloride concentration for suspensions with different pH. The following chemical and physical conditions were used in the experiments: clay concentration = 50 rag/l, coagulation mixing intensity = 26 s-~, rapid-mixing time = 2 min, coagulation time = 15 min, settling time = 15 min and bicarbonate concentration = 1 mM. 1990; Stumm, 1992), and, as a result, charge reversal was observed at ferric chloride concentrations larger than about 10 # M . On the other hand, the charge of ferric hydroxide precipitates is close to zero at pH 7.8 (Dzombak and Morel, 1990; Stumm, 1992), and. therefore, the EM of the particles does not change significantly at high ferric chloride concentrations. At lower ferric chloride concentrations, the surface charge of the kaolin particles reflects the formation of ferric surface complexes and the acid-base properties of those complexes (Stumm, 1992). The residual turbidities, measured after coagulation and settling of the suspensions described above, are shown in Fig. 4. At these pH values (i.e. 6.0 and 7.8), the residual turbidity curves are comparable. The residual turbidity decreases continuously as the ferric chloride dose increases, and it reaches a constant value at ferric chloride concentrations larger than about 35/~ M. These results demonstrate that the clay particles aggregate even at low ferric dosages. where the particles are negatively charged. Coagulation in the experiments described in Fig. 4 is induced by formation of ferric hydroxide precipitates in solution and by adsorption of ferric precipitates and dissolved species onto the surface of the clay particles. One should note that the solution is supersaturated with respect to ferric hydroxide under all the chemical conditions used in the experiments described above (Stumm and Morgan, 1981). At pH 6 and low ferric chloride concentrations, adsorption of positively charged ferric hydroxide precipitates and dissolved species onto the clay particles reduces their negative charge and thus can promote aggregation at ferric dosages where the particles have no net

Effect of ferric dose on electrophoretic mobility and residual turbidity


The effect of ferric chloride concentration on the electrophoretic mobility (EM) of the kaolin particles at two pH values (6.0 and 7.8) is shown in Fig. 3. At pH 6.0, the EM of the kaolin particles increases significantly as ferric chloride is added, and charge reversal is observed at ferric chloride concentrations larger than 10 tiM. At pH 7.8, on the other hand, the particles remain negatively charged as ferric chloride is added, and no charge reversal is observed. At high ferric chloride concentrations, the particles are coated with ferric hydroxide precipitates, which then dominate the surface acid-base properties, surface charge and EM of the particles (James and Healy, 1972). At pH 6.0, the ferric hydroxide precipitates are positively charged (Dzombak and Morel, 0.5 0 "-" -0.5
- 1 "'

o - o . . . . . <>. . . . <>. . . . -o ~" 5 9 ,, 9 "~3

~-1 "-"

"l 0 .| .5 ~ ~ ~
2

/....o---.....___

-2.5

- 3 0

- - o - - - p H 7.8 ,,, I,,,, I,,,, I .... I .... I .... 10 20 30 40 50 60 F e r r i c C h l o r i d e (~tM)

- - - - p H 6.0

Fig. 3. Electrophoretic mobility (EM) of the kaolin particles as a function of added ferric chloride for two different pH values. Alkalinity was provided by adding 1 mM sodium bicarbonate,

Dynamics of coagulation of kaolin charge. This mechanism, however, cannot explain the coagulation of the clay suspension at pH 7.8, since the particles at this pH are highly charged at all ferric dosages (Fig. 3). As will be described later, formation of patches of positively charged ferric precipitates on the negatively charged surface of the clay particles may also induce coagulation. At both pH values, "sweep-floe" coagulation is the predominant aggregation mechanism at high ferric chloride concentrations. In this mechanism, the large ferric hydroxide precipitates enmesh suspended particles and effectively remove them by settling. The coagulation mechanisms in the experiments with the above suspensions are discussed in more detail in the following section. It is shown that measurements of coagulation index of the aggregating suspensions by the flowthrough monitor can provide more insight into the mechanisms of coagulation with ferric chloride.
Effect of ferric chloride concentration on coagulation index curves The change of the coagulation index with time for the coagulation experiments described above is shown in Fig. 5. As described before, the coagulation index, measured continuously by the optical monitor, is very sensitive to the state of aggregation of particulate suspensions. The data presented in Fig. 5 are for the period of slow mixing, that is, following the 2 min of rapid mixing,

563

The coagulation index for the experiments at pH 6.0 [Fig. 5(a)] increases gradually with time in the ferric chloride concentration range of 1.5-9.0 # M. At ferric dosages larger than 9.0 #M, the initial rates of the increase in the coagulation indices are greater, implying faster aggregation rates. Similar behavior is observed in the experiments at pH 7.8 [Fig. 5(h)], where the initial slope of the coagulation index curves increases with ferric coagulant dose. For both suspensions, the curves of the coagulation index at high ferric dosages pass through a maximum followed by a gradual decrease of the coagulation index. This phenomenon will be discussed later in this section. Furthermore, it should be noted that, in the absence of Fe(III), the coagulation index in all the experiments reported here remained unchanged, as shown in Fig. 5(b). Since the EM of the clay particles at pH 6.0 increases (becomes less negative) and passes through zero as ferric chloride is added, one might conclude that charge neutralization is the predominant coagulation mechanism in those experiments; however, the increase in the coagulation index at all ferric dosages (1.5#M and larger) for the experiments at pH 6 is much greater than that expected for coagulation by charge neutralization. The number concentration of particles larger than 0.6 #m in the clay suspension, as measured by the Coulter counter, is 4.02 x l0 s per cm 3. Calculations of coagulation rate based on the

7L
6

a. p H 6.0 ~

1.5 gM

5 ~
3

"

----o--3.0 ~M A 6.0~M -----o--9.0 ~ = 10.5glVl ~

O 2 I O 1 q o ~5 4 ~O ~2
o ~J

12.0 gM * 15.0gM ----a---18.0FM b. p H 7.8 = ~ 9 . ) ~. 0gM 6gM M 12

) ----o--- 15 gM

------cY-- gM 21
~

A *

30p_M 4Sp_M

O0

10

15

Time

(rain)

Fig. 5. Coagulation index as a function of time for various concentrations of ferric chloride for the coagulation experiments described in Fig. 4.

564

HSlAO WEI CHINGet al.

theory of orthokinetic coagulation (Ives, 1990) for this suspension, assuming favorable chemical conditions, shows that the residual (normalized) particle concentration after 15 min of coagulation at G = 26 s- ~ is 0.93. This means that, although the particles are destabilized, most of them will remain unaggregated because of the low interparticle collision rate. Coagulation experiments at high electrolyte (KCI and CaC12) concentration, under conditions where the particles are assumed to be destabilized, were conducted to test this hypothesis. It was found that the coagulation index in those experiments changed very little (by less than a factor of 2) after 15 min of coagulation, thus supporting the conclusion that the coagulation rates in the experiments shown in Fig. 5(a) are larger than those expected for charge neutralization, At pH 6.0 and low ferric dosages, ferric hydroxide precipitates and dissolved species are positively charged, whereas the kaolin particles are negatively charged (see Fig. 3). Under these conditions, adsorption of small ferric hydroxide precipitates or surface precipitation (Farley et al., 1985) might form small positively charged patches on the negatively charged surfaces of the clay particles. As a result, attractive forces could develop between a patch and an oppositely charged bare surface as particles collide during coagulation. This "electrostatic patch" model can explain the coagulation of negatively charged particles with cationic polymers observed at coagulant concentrations insufficient for charge neutralization, i.e. when the EM of the particles is substantially negative (Gregory, 1973, 1976). As discussed before, the coagulation rates at low ferric concentrations are higher than those for charge neutralization. It has been shown that coagulation rates of patched surfaces can be faster than those observed in a fully destabilized suspension, i.e. by charge neutralization (Gregory, 1976, 1988). As the ferric dose increases, the concentration of ferric hydroxide precipitates increases, resulting in higher coagulation rates and, hence, larger coagulation indices. At pH 7.8, the kaolin particles are negatively charged (Fig. 3), while the ferric hydroxide precipitates have a charge close to zero (Dzombak and Morel, 1990; Stumm, 1992). At low ferric dosages, the particles have a large negative charge; nevertheless, the aggregation rate is substantial. This, for instance, can be seen from the coagulation index curve for a ferric dose of 6 g M [Fig. 5(b)]. The aggregation at these conditions can be partly explained by the electrostatic patch model discussed above. In addition, collision between precipitates at pH 7.8 is favorable, resulting in larger precipitates and, therefore, higher aggregation rates. At high ferric dosages, the formation of ferric hydroxide precipitates dramatically increases the coagulation rate by (!) increasing the solids concentration and thus the rate of aggregation in the suspension and (2) enmeshing clay particles in larger

aggregates ("sweep-floe" coagulation). At high ferric chloride concentrations, the formation of hydroxide precipitates is very fast. This is evidenced by the high initial values of the coagulation index, immediately after the rapid mixing [Fig. 5(h) for ferric dosages of 30 and 45 #M]. An interesting feature in the coagulation index curves shown in Fig. 5 is the maximum observed at ferric chloride concentrations larger than about 10 #M. The coagulation index curves pass through a maximum followed by a continuous decrease. The maximum occurs at an earlier time as the ferric chloride concentration increases. This phenomenon apparently suggests that the aggregate size decreases after the time corresponding to the peak in the curve. A close inspection of this phenomenon, however, suggests that the decrease in the value of the coagulation index results from the settling of large aggregates (flocs) during the coagulation process. At low mixing intensities, it is difficult to keep the large floes suspended in the water, and a significant number of these flocs settle. Thus, the number of particles sampled by the Tygon tube leading to the photo dispersion analyzer decreases. Since the coagulation index is proportional to particle size and to the square root of particle concentration (Gregory, 1985), the coagulation index decreases as fewer particles enter the sampling tube. This hypothesis was further verified by lowering the sampling point in the coagulation vessel during aggregation at high dosages of ferric chloride. It was found that the coagulation index increased significantly soon after the sampling tube was lowered. It should be noted that higher concentrations of floes were visually observed below the sampling point during those coagulation experiments. In addition, as will be shown later in this paper, this maximum disappears at high mixing intensities. Effect of solution pH on electrophoretic mobility and residual turbidity The effect of solution pH on the electrophoretic mobility of the kaolin particles at different ferric chloride concentrations is shown in Fig. 6. With no ferric chloride in solution, the kaolin particles are negatively charged at all pH values. At pH values smaller than about 8, the EM of the particles increases (becomes less negative) as ferric chloride is added; the particles reverse their charge (from negatire to positive) at low pH values. This is attributed to adsorption of ferric species and ferric hydroxide precipitates onto the surface of the clay particles. Ferric hydroxide precipitates are positively charged at pH values smaller than about 8 (Dzombak and Morel, 1990). The isoelectric points (pH where the EM is zero), corresponding to ferric chloride dosages of 3.0, 10.5 and 36.0pM are 4.5, 6.0 and 6.5, respectively. Residual turbidity curves for the coagulation experiments under the solution chemistries described

Dynamics of coagulation of kaolin 3

565

charge (see Fig. 6), is smaller than that at higher pH values, where the particles are negatively charged. As 2 described earlier, coagulation rates higher than those --v--10.5 ~ for charge neutralization are possible when the elec1 --~--36.0~M trostatic patch model is invoked. The coagulation . ~ index at pH 8.7 remains unchanged during the entire coagulation period. At this pH, the particles and the 1 small ferric hydroxide precipitates are negatively ~..~A,.~ charged, so that precipitate-precipitate, particle,~ ~" particle and particle-precipitate interactions are -2 ~_ \ ~, . ~ 'oxh probably unfavorable. As a result, growth of aggre- 3 gates or solid precipitates is not feasible, resulting in negligible coagulation rates. The coagulation index curves at higher ferric -4 . noFeCl 3 dosages [10.5 and 36.0/IM in Fig. 8(b) and (c), - $ ",,,,I,.,, t,,,.I .... t .... t .... t .... t .... respectively] increase much faster that those at 3 4 5 6 7 8 9 10 1 3.0 #M. In addition, larger coagulation indices are pH reached, implying larger aggregates. At these ferric concentrations, formation of large hydroxide precipiFig. 6. Electrophoretic mobility (EM) of the kaolin suspen- tates is more favorable in the pH range of 6-8. These sion as a function of solution pH for different ferric chloride precipitates increase the solids concentration in susconcentrations. Alkalinity is provided by adding I mM of pension, and, as a result, the coagulation rate insodium bicarbonate. creases. At very large ferric dosages [Fig. 8(c)], the formation of large hydroxide precipitates is very fast, above are shown in Fig. 7. At low ferric dose (3/~M) and sweep-floe coagulation dominates. These precipiand low solution pH, coagulation is induced by tates are quite large after 2 rain of rapid mixing, as adsorption of ferric species and hydroxide precipi- evidenced by the high initial values of the coagulation rates which can form positively charged patches on index at this pH range. The maximum and subthe surface of the clay particles, and thus induce sequent decrease of the coagulation index at high aggregation (discussed in the previous section). In ferric dosages is attributed to the settling of large addition, formation of hydroxide precipitates can aggregates, as discussed previously in this paper. increase the solid concentration and, hence, the interparticle collision rate. At low pH values, it is not Effect of mixing intensity on residual turbidity and expected that significant amounts of large precipitates coagulation index curves would be formed at a ferric concentration of 3/z M. The effect of mixing intensity (expressed in terms of Therefore, because of low interparticle collision rates, the velocity gradient G) on the residual turbidity aggregates do not increase in size significantly and removal of turbidity by settling is low. At high ferric chloride concentrations, aggregates 4 0K o , ~ x I increase in size because of the higher coagulation rate, ~ 35 ~ ~ F and residual turbidities after settling can reach lower values. As described previously, the predominant "" 30 . ~ / mechanism of coagulation at high ferric concen..~ N,t~'~-~, /, trations is the formation of ferric hydroxide precipi25 A rates that increase the particle collision rate. In addition, at high ferric dosages, large precipitates 20 ~; O f I enmesh the smaller particles and remove them from '~ ~: I the suspension by settling. Efficient coagulation and ,~ 1 5 q'~ I , settling of aggregates is observed over broader pH "~ ' --<y-3.0 gM ~ , ~ . . . . . ~ ranges at higher total ferric concentrations, consistent ~ 10 ~ with the solubility diagram of Fe(III) (Stumm and - o -10.5 I~M ",, ,' Morgan, 1981). 5 - - ~-- 36.0 o_M ~ . . . . . ~-,~ ~" ~ "..'~ \ "" "~ "'. 4 , "'o,, ~ - - 0- - 3.0 p.M

Ji

"

,.,I.,.,I,,.,I,,,,I,,,.1,,.,

I,,.,l

Effect of solution p H on coagulation index curves

10

The change of the coagulation index at different pH values is shown in Fig. 8 for three ferric dosages. At low ferric dosages [3 ~tM in Fig. 8(a)], the coagulation index increases continuously at pH values smaller than 8. The slope of the coagulation index curve at pH 4.7, where the particles have no net

pH Fig. 7. Residual turbidity (after settling) of the kaolin suspensions as a function of solution pH for different ferric chloride concentrations. The physical conditions are similar to those described in Fig. 4.

566

HslAOWE! CHINGet al.

6
s I~

a ---O--pH 8.7 ---o-- pH 7.4 --O---pH 6.4 - - A - - pH 5.8 O pH4.7

;0D 2 U 1
{
0 I I

b 5
,~ 4 ~
3

---O--pH 8.2 --O--pH 7.7 ~D pH7.1 ~pH 6.4 pH 6.0 & pH 5.3 ---O--pH 4.0
!
; ' ' ' ' ' I ' ' ' ' ' ' ' ' '

!
I ' ' ' ' ' ' ' ' l

,~
I~ "~

3 2

---O--pH 9.0 ---CF-pH 8.1 +pH 7.6 ------A--- 6.4 pH --@--pH 5.5 --.o-- pH 4.0

5 T i m e (rain) Fig. 8. Coagulation index as a function of time for different solution pH at three ferric chloride concentrations: (a) 3.0 pM, (b) 10.5 pM, and (c) 36.0 #M. The coagulation indices were measured during the coagulation experiments reported in Fig: 7.

"1

10

curves is shown in Fig. 9. The mixing intensity in these experiments was controlled by adjusting the speed of the fiat-blade mixer (Fig. 2). The coagulation experiments reported in this figure were conducted at pH 6.0 with two different ferric chloride concentrations (10 and 36/~M). These concentrations of ferric chloride were selected so that the influence of mixing intensity could be investigated under different coagulation mechanisms. At the lower ferric concentration (I0/~ M), coagulation is induced by formation of ferric precipitates in solution (i.e. increasing particle collision rate) and by adsorption of the positively charged ferric precipitates and dissolved species on the clay particles. Particle-particle interactions under these conditions are favorable, since the particles have no net charge (see Fig. 3). On the other hand, at the higher ferric dose (36/~M), sweep-floc

coagulation is the predominant mechanism, as discussed earlier in this paper. At a ferric chloride concentration of 10/~ M, the residual turbidity decreases continuously as the mixing intensity increases from 5 up to 57 s t. At this range of mixing intensities, the coagulation rate increases with G because of the increase in the collision rate of the solid precipitates and the uncharged particles. As the coagulation rate increases, larger aggregates are formed, resulting in higher settling rates and consequently lower residual turbidities. At G values larger than 57 s- ~, strong shear forces develop, causing breakage of aggregates. As a result, the residual turbidity does not change as G increases from 57 up to 137 s-k. At higher ferric chloride dosages (36#M), large aggregates are formed by precipitation of ferric

Dynamics of coagulation of kaolin 30

567

increase in the particle collision rate. At G values larger than 57 s -1, the coagulation indices reach a 25 comparable value. In addition, at these mixing intensities, the coagulation indices reach their maximum value after only about 5 rain of coagulation. At larger 20 coagulation times, the coagulation indices do not change, thus indicating that the aggregates attain an 15 equilibrium (limiting) size. The aggregates cannot further increase in size due to breakage of aggregates caused by the high shear forces. At very low mixing 1 0 [] intensities (e.g. the curve for 5 s- 1), the coagulation ", index increases monotonically over the entire period 5 of coagulation. Because of the low particle collision O O b . . . . . . . v-. . . . . . . . ~ . . . . . . . . . . o rate at this mixing intensity, aggregates do not attain |1 i i i i i l | l | l l | l l | l l | l i iii i|| 0 their limiting size. 0 50 10 0 150 The coagulation index curves for high ferric dose G (a"1) (36/zM) are shown in Fig. 10(b). At this ferric dose, Fig. 9. Residual turbidity (after settling) as a function of large ferric precipitates are formed within a short coagulation mixing intensity (G) for different ferric dosages, period of time; sweep-floe coagulation is the predomAll experiments were conducted at pH 6.0 and I mM inant coagulation mechanism. At low mixing intensisodium bicarbonate, ties (G = 26 s-~), large floes settle and the coagulation index curve passes through a maximum hydroxide. The residual turbidity decreases as the (discussed earlier in this paper). At mixing intensities mixing intensity (G) increases up to a limiting value larger than 26 s-l, the coagulation index curves are between 26 and 57 s- ~. At greater mixing intensities, comparable. The coagulation indices attain a limiting a slight increase in the residual turbidity is observed, value within less than 3 min. At high mixing intensiThis may be attributed to the formation of small ties, the large aggregates do not settle, and the particles upon aggregate breakage; small particles coagulation indices remain unchanged after aggrecontribute more to the turbidity than large aggregates gates reach their limiting size. (van de Hulst, 1957; Kerker, 1969). Coagulation index curves for the experiments de- Analysis of the shear in the flow through detector scribed above are presented in Fig. 10. At a ferric Due to the nature of the signal analysis in the dose of l0/~M [Fig. 10(a)], the initial slope of the optical monitor used in this research, the frequency coagulation index curves increases as the mixing of the fluctuating signal should be about a few intensity increases, indicating higher aggregation hundred Hz (Gregory, 1985). With lower frequencies, rates. As discussed previously, this is caused by the some of the signal will be lost, and the RMS output will be lower than the true RMS value. To avoid this problem, the average velocity of the particles flowing 8 .a 7 through the detector should be a few cm/s. A flow , ~ ~ : ~ / ~ l rate of 22 ml/min was used in the experiments re6 C (s"1) ported in this paper. For a 3 mm (i.d.) tube, this flow .~ 5 ---o-aar rate corresponds to an average velocity of 5 cm/s. ~1 zl --C]---94 Similar velocity was used in all previous coagulation ---o-st studies employing this optical technique (e.g. Gret~ 3 ~ 2 6 2 --0-5 gory, 1988; Gregory and Lee, 1990; Gregory and Li, 1 1991). " ' " When a particle suspension passes through a "o narrow tube under laminar flow conditions, a shear 6 G(s-1) (velocity gradient) is developed. The average shear 5 ---o-la7 rate in the tube (G) is given by (Gregory, 1981) --o--10.51aM - - v- - 36.0 ItM 4 3 L
O

~ --~26 5 1'0'"" Time (rain) i5

8Q G - 31rR 03

(4)

where Q is the volumetric flow rate, and R0 is the inner radius of the tube. Equation (4) applies to laminar flow conditions. The 5 cm/s average velocity used in this investigation is within the laminar flow Fig. 10. Coagulation index as a function of time for different regime. mixing intensities (G): (a) ferric dose= 10/~M; (b) ferric dose=36/aM. All experiments were conducted at pH 6.0 For the flow conditions used in our work and 1 mM sodium bicarbonate. (Q = 22 ml/min, Ro = 1.5 ram), the average shear rate

t9 2

568

Hslno WEI CHING et al. REFERENCES Amirtharajah A. and O'Melia C. R. (1990) Coagulation processes: destabilization, mixing, and flocculation. In Water Quality and Treatment. McGraw-Hill, New York. Dentel S. K. (1991) Coagulation control in water treatment. Crit. Rev. Envir. Control 21, 41-135. Dzombak D. A. and Morel F. M. M. (1990) Surface Wiley-Interscience, New York. Farley K. J., Dzombak D. A. and Morel F. M. M. 0985) A surface precipitation model for the sorption of cations on metal oxides. J. Colloid Interface Sci. 106, 226-242. Franqois R. J. (1988) Growth kinetics of hydroxide flocs. J. Am. Wat. Wks Ass. 80, 92-96. Gregory J. (1973) Rate of flocculation of latex particles by cationic polymers. J. Colloid Interface Sci. 42, 448--456. Gregory J. (1976) The effect of cationic polymers on the colloidal stability of latex particles. J. Colloid Interface Sci. 55, 35-44. Gregory J. (1981) Flocculation in laminar tube flow. Chem. Engng Sci. 36, 1789-1794. Gregory J. (1985) Turbidity fluctuations in flowing suspensions. J. Colloid Interface Sci. 105, 357-371. Gregory J. (1988) Polymer adsorption and flocculation in sheared suspensions. Colloids Surfaces 31, 231--253. Gregory J. and Lee S. Y. (1990) The effect of charge density and molecular mass of cationic polymers on fiocculation kinetics in aqueous solution. J. Wat. SRT-Aqua 39, 265-274. Gregory J. and Nelson D. W. (1986) Monitoring of aggregates in flowing suspensions. Colloids Surfaces 18, 175-188. Gregory J. and Li G. (1991) Effects of dosing and mixing conditions on polymer flocculation of concentrated suspensions. Chem. Engng Commun. 10a, 3-21. van de Hulst H. C. 0957) Light Scattering by Small Particles. Dover, New York. Ives K. J. (1990) Coagulation and flocculation Part l I - Orthokinetic flocculation. In Solid-Liquid Separation (Edited by Svarovsky L.). Butterworth-Heinemann, Oxford. James R. O. and Healy T. W. (1972)Adsorption of hydrolyzable metal ions at the oxide-water interface. 11. Charge reversal of SiO2 and TiO2 colloids by adsorbed Co01), La(lll), and Th(IVI as model systems. J. Colloid Interface Sci. 40, 53~4. Johnson P. N. and Amirtharajah A. (1983) Ferric chloride and alum as single and dual coagulants. J. Am. Wat. Wks Ass. 75, 232-239. Kerker M. (1969) The Scattering of Light and Other Electromagnetic' Radiation. Academic Press, New York. Lawler D. F., Izuieta E. and Kao C.-P. (1983) Changes in particle size distributions in batch floeculation. J. Am. Wat. Wks Ass. 75, 604-612. Li G. and Gregory J. (1991) Flocculation and sedimentation of high-turbidity waters. Wat. Res. 25, 1137-1143. Morel F. M. M. and Hering J. G. (1993) Principles and Applications t~f Aquatic Chemistry. Wiley-lnterscience, New York. O'Melia C. R. (1985) Particle, pretreatment, and performance in water filtration. J. envir. Engng, Div. Am. Soc. Cir. Engrs I l l , 874-890. O'Melia C. R., Gray K. A. and Yao C. (1989) Polymeric Inorganic' Coagulants. American Water Works Association Research Foundation, Denver, Colo. Ramaley B. L., Lawler D. F., Wright W. C. and O'Melia C. R. (1981) Integral analysis of water plant performance. J. envir. Engng, Div. Am. Soc. Cir. Engrs 107, 547-562. Stumm W. (1992) Chemistry of the Solid-Water Interface:
Processes at the Mineral-Water and Particle-Water Interface in Natural Systems. Wiley, New York. Complexation Modeling: Hydrous Ferric Oxide.

calculated from equation (4) is about 90 s ~. Under this shear rate, it is likely that some aggregates would be disrupted on passage through the tube before reaching the optical flow-through detector. However, as was demonstrated in Fig. 10, the coagulation indices still increased when the mixing intensity was greater than 90s -I, thus indicating that aggregates were still growing in size, regardless of the high shear. Note that for the experiments at high mixing intensities (G > 90 s- l ), the average shear rate in the tube is smaller than that in the stirred vessel. These arguments may suggest that the extent of aggregate break-up in the tube is not very large. In addition, the average Gt value ("Camp number") in the tube during the experiment is small compared to that in the stirred vessel (the detention time in the tube is about 10s compared to 15min in the jar). Hence, the extent of coagulation in the tube is also negligible.

CONCLUSIONS The coagulation index measured by the flow through optical technique is very sensitive to the state of aggregation of particle suspensions coagulating with Fe(III) coagulants. Measurements of coagulation index dynamics complement the data obtained from residual turbidity and electrophoretic mobility measurements and thus can help in selecting optimal coagulant dose. Such measurements, together with electrophoretic mobilities, can provide additional insights into the mechanisms of coagulation with hydrolyzing metal salts. The shape of the coagulation index curves is determined in large part by the mechanisms of coagulation. In general, larger values of the coagulation index imply larger aggregates. Caution must be taken, however, in coagulation experiments at low mixing intensities when the aggregates are very large (as in the case of sweep-floc coagulation). Under these conditions, the coagulation index can decrease, due to settling of large aggregates. The results reported in this paper also illustrate that measurements of the coagulation index can be used to investigate the effect of mixing intensity on coagulation dynamics with hydrolyzing metal salts. Such measurements demonstrate that, at a given mixing intensity, an equilibrium aggregate size is reached. The size of the aggregate is determined by two competing processes: (1) growth of aggregates induced by particle collisions and (2) breakage of aggregates by shear forces.
Acknowledgements--This research was supported by the

Metropolitan Water District of Southern California (MWD). We thank Kevin Wattier and Mark Beuhler of MWD for their assistance, encouragement and ideas. The findings reported in this paper do not necessarily reflect the views of MWD, and no official endorsement should be inferred.

Dynamics of coagulation of kaolin Stumm W. and O'Mefia C. R. (1968) Stoichiometry of coagulation. J. Am. Wat. Wks Assc. 60, 514-539. Stumm W. and Morgan J. J. (1981) Aquatic Chemistry. Wiley, New York. Tang H.-X. and Stumm W. (1987a) The coagulating behaviors of F(III) polymeric spccics---I.Preformed polymers by base additions. Wat. Res. 21, 115-121. Tang H.-X. and Stumm W. (1987b) The coagulating

569

behavior of FeOlI) polymeric species--II. Preformed polymers in various concentrations. Wat. Res. 21, 123-128. Trcwcck G. P. nd Morgan J. J. (1977) Size distributions of flocculated particles: applications of electronic particle counters. Envir. Sci. Technol. 11, 707-714. van de Hulst H. C. (1957) Light Scattering by Small Particles. Dover, New York.

Anda mungkin juga menyukai