Chemical Engineering And Processing 47 (2008) 1509–1519

Coagulation of stable laterite suspension Zeta potential mV, turbidity reduction % are plotted as a function of the coagulant concentration expressed in mass ratio mg of AlCl3/g of later

Coagulation and flocculation of laterite suspensions with low levels of aluminium chloride and polyacrylamids

D´esir´e Dihang a , b , Pierre Aimar b , ∗ , Joseph Kayem a , Sylv`ere Ndi Koungou a , b

aTEFI Unit-ENSAI/IUT, University of Ngaoundere, P.O Box 455, Ngaoundere, Cameroon

bLGC-CNRS-UMR 5503 Universit´e P Sabatier Toulouse Cedex 9, France

Received 28 July 2006; received in revised form 4 July 2007; accepted 4 July 2007

Available online 18 July 2007

Flocculation either by non-ionic (PAM-N), cationic (PAM-C), or anionic (PAM-A) high molecular weight polyacrylamids promotes turbidity reduction of pre-coagulated laterite suspensions This turbidity reduction is independent on the amount of polymer added when the suspension is coagulated at the CCC In the other cases, turbidity reduction depends on polymer concentration For suspensions of high turbidity, flocculation does not improve significantly the efficiency as compared to coagulation At low concentration, PAM-N and PAM-A do not significantly modify the Zeta potential of the particles, enabling it to remain a relevant parameter to monitor destabilisation by combined coagulation and flocculation Laterite particles are very sensitive to the presence of PAM-C, which induces charge reversal even at very low concentration The critical concentration for flocculation is lower than 0.1 mg/L.

© 2007 Elsevier B.V All rights reserved.

Keywords: Laterite; Coagulation; Flocculation; Potable water

1 Introduction

Laterite is the major component found in raw water in most

tropical regions in Africa, and its removal represents the main

objective of the drinking water processes Laterite is a clay

mate-rial that confers to raw water a red colour and hazy aspect It is

also known to be the main vector of arsenic contamination in

ground water in many regions of the world Because of the lack

of knowledge on this clay and/or inappropriate process, it is

com-mon to find suspended particles in tap water, especially during

the rainy season Rather than the distribution system (Lehtola et

al [1] ), investigations reveal the inefficiency of the clarification

process, where particle removal is achieved by decantation and

∗Corresponding author at: LGC-CNRS-UMR 5503 Universit´e P Sabatier

Toulouse Cedex 9, France Tel.: +335 6 15 58 304; fax: +335 6 15 56 139

E-mail address:aimar@chimie.ups-tlse.fr(P Aimar)

filtration through sand beds Particle removal is promoted by coagulation with aluminium chloride and by flocculation with polymers In these tropical regions, where water plants are rather old, but where it is necessary to comply with turbidity standards for obvious health issues, it is important to use as little coagu- lant as possible, in order to fulfil both standards and economical requirements Moreover, the known implication of aluminium

in Alzheimer disease makes this issue of using low lant level a worldwide problem In drinkable water application, aluminium salt concentration depends on the physicochemical properties influencing salt precipitation So, the salt concentra- tion varies with the pH, ionic force, nature and concentration of ionic species, amount of organic matter, temperature, etc.

coagu-In practical, hydroxide precipitation occurs above 40 ␮M minium added but OMS recommended less than 200 ␮g/l of aluminium in treated “potable” water.

alu-However, using such low coagulant dosage makes it very difficult to locate the concentration giving maximum elimi-0255-2701/$ – see front matter © 2007 Elsevier B.V All rights reserved

doi:10.1016/j.cep.2007.07.002

Chemical Engineering and Processing 47 (2008) 1509–1519

1510 D Dihang et al / Chemical Engineering and Processing 47 (2008) 1509–1519

nation of the suspended particles, unless jar test experiments

are carried out almost continuously (Gregory and Duan [2] ,

Gregory [3] ) The literature abounds with studies on the

coagulation–flocculation of clay material, but most of them

focus either on concentrated suspensions as generally

encoun-tered in mining processes, or on waste water process with high

content in organic matter Contrarily, this work investigates

destabilisation by coagulation and flocculation of laterite

sus-pensions diluted to over 10-fold the usual cases studied in the

literature and where electrokinetics aspects of the suspended

particles are important It is aimed in filling the gap in the

litera-ture on the destabilization characteristics of laterite suspensions

with regards to the classical drinkable water process.

2 Materials and experimental methods

2.1 Materials

Raw laterite clay was obtained from the banks of the river

Bini in Cameroon The sample was washed with ultra pure water

and dried The coagulant, aluminium chloride, was prepared

daily as 1 g L−1 solution to avoid polymerisation in solution.

Graciously supplied by FLOERGER, the flocculants, a cationic

(FO8990 SEP), an anionic (AH912 SEP) and a non-ionic (FA920

SEP) high molecular weight polyacrylamids, were used for the

flocculation of the pre-coagulated laterite suspensions They

were prepared as 0.2 g L−1 solutions in distilled water and

kept for a week at room temperature Sodium hydroxide and

hydrochloric acid, used for pH adjustments, were prepared as

1% (w/w) solutions Potassium chloride was used to set the ionic

strength.

Laterite suspensions were prepared in two steps: first, a

dispersion–hydration of the dried powder in distilled water and

second a dilution of the obtained suspension to the required

tur-bidity The laterite powder (50 g) is added to water (5 L) under

vigorous agitation (750 rpm) and the pH raised to 10 Sodium

azide is added for preservation (0.02%, w/w) and mixing is

pro-longed for 12 h The suspension is then allowed to settle for 4 h

and the supernatant is withdrawn It is kept at room temperature

and can be used for a week Knowing that raw water turbidity

ranges from 22 NTU in the dry season to around 350 NTU

dur-ing the rainy season, we selected five turbidities (30, 90, 150,

180 and 300 NTU) for our study For this purpose, the

super-natant is diluted to the desired turbidity 10−3M in KCl was

added raising the conductivity to around 150 ␮S/cm and the pH

was adjusted to 7 All reagents used were of analytical grade.

Laterite suspensions exhibit negligible organic matter, around

1 mg/L of TOC measured on a TOC Analyzer model VCSN

from Shimadzu.

2.2 Experimental methods

2.2.1 Characterisation of laterite

The particle distribution measurements were carried out

using a Malvern Mastersizer 2000 analyser It also gives the

measure of specific surface area of the particles in square meter

per gram The characterisation of surface structure of the

lat-Table 1Mean hydrodynamic radius and the Zeta potential of the floculantsFloculant Mean hydrodynamic radius (nm) Zeta potential (mV)

Zeta potential indicating surface charge of the particles was measured with a Malvern Zetasizer 4 The measurement was carried out in various electrolytes, (KCl, NaCl and CaCl2) at various concentrations and pH.

Turbidimetry, light scattering using a Turbiscan-On-Line (Formulaction, France), particle counting (Malvern coulter counter) and dry mass determination by desiccation were used

to characterise the properties of laterite suspensions.

2.2.2 Characterisation of the floculants

We could not obtain the molecular weights from the ufacturer However, we measured the hydrodynamic radius so

man-as to have an idea of the molecular size, man-as shown in the table ( Table 1 ).

High charge density on PAM-C enable it to adopt a more extended configuration in suspension while PAM-A and PAM-

N, with their low charge density exhibit lower repulsion between charged polymer segment thus a less extended configuration.

2.2.3 Coagulation and flocculation study

Coagulation and flocculation experiments were carried out in

a classical jar test apparatus The flocculants were added after coagulation as it has proven to give better results than other combinations (Hogg [4] , Somasundaran [5] ) After 5 min of agitation at 230 rpm, the coagulant was added drop by drop and left to coagulate particles for 5 min The stirring speed was then reduced to 30 rpm for 15 min to allow flocs growth and the suspension was allowed to settle for 30 min For floc- culation of coagulated suspension, the flocculant was added under rapid mixing before the flocs are left to grow The super- natant was further taken out at a fixed 2 cm below the air–liquid interface for turbidity, Zeta potential, and TOC measurements The turbidity measurements were performed on a Hach 2000N turbidimeter and through light scattering measurements with

a Turbiscan-On-Line The Zeta potential was measured on a Malvern Zetasizer 4 All experiments were performed at room temperature The experimental procedure is summarised in

Fig 1 The performance of coagulation and flocculation processes

is assessed through the turbidity reduction (%) given by (Ozkan and Yekeler [6] ):

Turbidity reduction (%) = Ti− Tf

where Ti(NTU) is the initial turbidity of the suspension and Tf(NTU) is the supernatant turbidity Amounts of coagulant and

Fig 1 Protocol for coagulation (a) and coagulation/flocculation (b) of laterite suspension.

flocculant are expressed both in weight ratio (mass per mass of

laterite), associated to molar concentration for the coagulant and

mass per volume concentration for the flocculants.

3 Results and discussion

3.1 Properties of laterite

The investigation of the composition of the laterite by

whole-rock analysis and of the main components, using SEM/

microprobe, TEM and XRD, reveals in Fig 2 the presence of

platelets of clay material and clusters of iron oxides,

correspond-ing to gibbsite, goethite and hematite titanium oxide The clay

composition indicates the presence of high amount of kaolinite

and traces of smectite.

Youngue-Fouateu et al [7] ) reported similar results on

dif-ferent samples of laterite from Cameroon.

In Fig 3 , the Zeta potential of the laterite suspensions in

absence and presence of different salts is plotted as a function of

pH In the presence of CaCl2, the Zeta potential is constant and

independent of the pH Contrarily, in the absence, as well as in

the presence of added KCl and NaCl, laterite suspensions show a

pH dependence of the Zeta potential The isoelectric point (IEP)

is located around pH 3, the Zeta potential is positive below the

IEP and negative above Similar Zeta potential is obtained in the

absence and in the presence of 10−3M KCl At pH 7 where the

destabilisation tests are performed, the Zeta potential is −35 mV.

Silica and aluminium oxides have a more prominent influence on

the Zeta potential, as indicated by the significantly more negative

values measured under alkaline conditions in contrast to acidic

one.

This behaviour is in good agreement with the data on kaolinite

and oxides, but the Zeta potential of laterite in the absence of

salt is greater in magnitude than the values reported for kaolin

(Besra et al [8–10] , McFarlane et al [11] ) and lower than the

values for iron oxides (McGuire et al [12] ).

The particle size analysis ( Fig 4 ) shows a distribution with

an average diameter d50 of approximately 0.25 ␮m, which is

smaller than the value obtained by Besra et al [8] The specific

surface area calculated from this diameter equals 26.7 m2/g, and

it is very high compared to the result on kaolin (Besra et al.

[8] ) but closer to the one obtained on iron oxides (McGuire et

al [12] ) It is smaller than the BET area reported on laterite by

Blakey and James ( [13] ) These results suggest that our mode of

preparation promotes the maximum of individualization of the suspended particles and a complete elimination of the coarse phase.

Dispersion properties of laterite suspension examined through the measurement of the turbidity, particle number, light scattered and concentration are given in Fig 5 Graph (a) presents two regions of linear properties that separate at

160 NTU Above this turbidity, the slope is lower The same behaviour is obtained when plotting the turbidity versus the light scattered or the number of particles This suggests that for tur- bidity smaller than 160 NTU, the suspension behaves as dilute while for higher turbidities, it behaves as concentrated.

In concentrated suspensions, particles influence each other, and several authors suggest that for clay suspension, the posi- tive charges from the edge face orientate towards the negative charges from the lateral faces in an auto-flocculation reaction (Blakey and James [13] ) This observation is important because

it can have a critical impact on the destabilisation behaviour

of these types of suspension, both on the mechanism involved and the destabilisation results According to our results, this mechanism would be relevant for turbidities higher than 160 NTU.

3.2 Coagulation of stable laterite suspension

Zeta potential (mV), turbidity reduction (%) are plotted as a function of the coagulant concentration expressed in mass ratio (mg of AlCl3/g of laterite) and in mass per volume of suspension ( ␮M of AlCl3) for different initial turbidities ( Fig 6 ) As a con- sequence of particles removal, the turbidity reduction increases with coagulant concentration until it reaches a maximum value, then it decreases with the addition of more coagulant This is often called restabilisation of the system by addition of an excess

of coagulant For dilute suspensions, the mass per volume centration at the critical coagulation concentration CCC increase with the initial turbidity while for concentrated suspension, it decreases On the other hand, for both dilute and concentrated suspensions, the corresponding coagulant mass ratio diminishes

con-as the initial turbidity increcon-ases These results confirm the gestion of an autocoagulation of particles as the suspension turbidity increases As a consequence, the coagulant mass ratio

sug-at the CCC sug-at 30 NTU is three folds thsug-at sug-at 300 NTU For both cases, the residual turbidity remains high and not acceptable for drinking purposes.

1512 D Dihang et al / Chemical Engineering and Processing 47 (2008) 1509–1519

Fig 2 Scanning electron micrograph (SEM) (a), transmission electron micrograph (TEM) (b) and corresponding chemical composition (XRD) of global laterite(c); platelet material (d); iron oxides clusters (e)

The Zeta potential increases with the amount of coagulant

until charge reversal, and stabilization around 30 mV For dilute

suspensions, the point of charge reversal (PCR) corresponds to

restabilisation (30 and 90 NTU), while it is close to the optimum

turbidity reduction for higher turbidity (>150 NTU) suspensions

( Fig 6 ) Turbidity reduction increases with the initial turbidity

for dilute suspensions, but it remains constant, however higher,

for concentrated suspensions In all cases, the critical Zeta tial is constant and equal to ca −20 mV.

poten-In Fig 7 , the supernatant turbidity is given for various decantation times.

The results show that, as decantation time increases, the supernatant turbidity decreases and reaches the same maximum value for 30 and 300 NTU suspensions The curves also high-

Fig 3 Colloidal titration of a laterite suspension of initial turbidity 150 NTU.

lights that restabilisation of the suspension, due to excess of

coagulant disappears as decantation time increases This

sug-gests that what is called restabilisation would in fact be a kinetics

effect.

3.3 Discussion on coagulation

NMR spectra on the aluminium chloride solution used for

coagulation show that only the monomeric octahedral

hexahy-drate Al3+(H2O)6is added to the laterite suspension In all the

jar test experiments, this addition decreases the pH of the

sus-pension to values between 5.5 and 6.5 In this region of pH,

there is an intensive formation of in situ Al13 polymers that

significantly improves the efficiency of AlCl3 These polymers

coexist with lower amounts of monomeric and medium

poly-merised positively charged aluminium species (Chengzhi Hu et

al [14] ) Coagulation occurs by particle charge neutralisation

and depends closely on the quantity of added coagulant and

consequently on the Zeta potential of the suspended particles

(Lartiges et al [15] , Gregory [16] , Wang et al [17] , Duan and

Gregory [18] , Hu et al [14] At the CCC, we observed that the

Zeta potential is negative This can be due to soluble silica that

lowers the Zeta potential of laterite clay, as it is more negative

Fig 4 Particle size distribution of laterite suspension 90 NTU

Fig 5 Turbidity (NTU) vs concentration (mg/L) (a) and transmission (%) (b)

The maximum turbidity removal is obtained at the critical concentration of coagulation; it is around 60% for low turbidities and 90% for high turbidities When the settling time is increased, the turbidity reduction is constant and greater than 95% and the apparent restabilisation of the suspension by excess of coagulant disappears The point of charge reversal (PCR) corresponds to the restabilisation of the system for low turbidity suspensions, but it coincides with the optimum turbidity reduction for con- centrated suspensions and long settling times This PCR can therefore not be used as a process control parameter At CCC, the value of the turbidity reduction varies, but the Zeta potential

is constant (−20 mV) Therefore, Zeta potential proves to be a more relevant parameter to predict the coagulation of laterite sus- pensions than the turbidity of the supernatant or the PCR Then, using turbidity to control industrial plants can generate errors

as it is dependent on many parameters It can be emphasised

1514 D Dihang et al / Chemical Engineering and Processing 47 (2008) 1509–1519

Fig 6 Turbidity reduction after 30 min of decantation, TR30min(%) and Zeta potential vs coagulant mass ratio (mg/g) and coagulant concentration (␮M) for lateritesuspensions of various turbidities: 30 NTU (a), 90 NTU (b), 150 NTU(c), 180 NTU (d) and 300 NTU (e)

that the maximum concentration of aluminium used, 40 ␮M,

is five-fold smaller than the standards allowed in processes for

drinking water while the minimum value, 9 ␮M is quite

equiv-alent to the standards Definitely, coagulation does not promote

sufficient particle removal for water to become potable Table 2

summarises the coagulation characteristics of various laterite

suspensions.

3.4 Flocculation of coagulated laterite suspension

In Fig 8 , the turbidity reduction (%) for the flocculation

of pre-coagulated laterite suspensions of 30 NTU initial

tur-bidity is plotted as a function of the amount of non-ionic PAM-N (a), cationic PAM-C (b), and anionic PAM-A (c) poly- acrylamid polymers added At CCC, all polymers promote turbidity reduction, and there is no restabilisation of the sus- pension with excess of flocculant This particle removal is the maximum achievable, from around 50% for coagulation

to 90% for flocculation In addition, PAM-N also induces bidity reduction for salt concentration lower than CCC, and to

tur-a lower extent, slightly tur-above the CCC PAM-C extends the area of efficiency of PAM-N to coagulant concentration slightly greater than CCC Finally, PAM-A provides turbidity reduc- tion at almost all concentrations, except at very low aluminium

Fig 7 Effect of settling duration on turbidity reduction for 30 NTU (a) and 300 NTU (b) laterite suspensions (turbidity reduction, TRX, where x is the time ofdecantation).

dosage Contrarily to PAM-N, PAM-C and PAM-A induce a

dependence of the turbidity reduction to polymer

concentra-tion.

For high turbidity laterite suspensions ( Fig 9 ), PAM-N

promotes turbidity reduction for almost all coagulant

concen-trations, from around 80% to nearly 100%, but has no effect in

the absence of coagulant.

The critical flocculation concentration (CFC) at CCC

dimin-ishes of at least 30-fold with suspension turbidity, from 5 mg/g

at 30 NTU to 0.15 mg/g at 300 NTU.

PAM-C and PAM-A are more effective than PAM-N as they

promote turbidity removal on a wider range of coagulant

con-centration However, as PAM-C and PAM-A are not allowed

for drinking water processes, we present only flocculation with

PAM-N for laterite suspension of 300 NTU initial turbidity.

In Figs 10 and 11 , the Zeta potentials of the particles in the

supernatant of the previously flocculated systems are presented

as a function of the coagulant and flocculant concentration For

initial high turbidity (300 NTU), all the Zeta potential curves

obtained with PAM-N are similar and aligned on the curve

obtained for coagulation ( Fig 10 ).

For the lowest turbidity suspension (30 NTU), Fig 11 shows

that PAM-N and PAM-C tend to increase Zeta potential while

PAM-A decreases it At low flocculant concentration (1.5 mg/g

of laterite), the Zeta potential curves for coagulation and

floc-culation are similar for PAM-N and PAM-A Higher amounts

of PAM-N (≥5 mg/g) induce a rapid growth of Zeta potential at

low aluminium content, and just before the CCC the Zeta

poten-tial approaches a plateau value close to 5 mV As a consequence

of this plateau value, the flocculation curve intercepts with the coagulation curve at a point below the PCR At this particular point, the coagulated and flocculated particles have equal Zeta potential.

PAM-C inverted the Zeta potential of the particles, indicating that for the amount of polymer used, we are already overdosing

it This behaviour confirms the high charge density of the

PAM-C, as indicated by the supplier To compare the effect of the three polymers on laterite particles, the turbidity reduction and Zeta potential is plotted versus the flocculant dosage, for a laterite suspension at 30 NTU, coagulated at the CCC ( Fig 12 ) The results show that all polymers exhibit the same efficiency for particle removal at CCC This efficiency is independent of the particle Zeta potential.

3.5 Discussion on flocculation

PAM-C adsorbs on the laterite particles surface via gen bonding interactions between the silanol and aluminol OH groups at the particle surface and polymer’s primary amide functional groups The electrostatic attractions between the pos- itively charged polymer segments and the negatively charged laterite particles promote adsorption and result in raising the par- ticles Zeta potential The amount of polymer adsorbed increases with increasing polymer concentration, generating higher Zeta potential Similar results are reported for flocculation of kaolin with PAM-C (Nasser and James [20] ) High molecular weight combined to segment repulsion enable particle bridging by the adsorbed polymer.

hydro-Table 2

properties of coagulated laterite suspension

1516 D Dihang et al / Chemical Engineering and Processing 47 (2008) 1509–1519

Fig 8 Turbidity reduction (%) vs PAM-N (a), PAM-C (b) and PAM-A (c) concentration of coagulated laterite suspensions of 30 NTU initial turbidity

Fig 9 Turbidity reduction (%) vs PAM-N concentration of coagulated laterite

suspension of 300 NTU initial turbidity

Fig 10 Zeta potential (mV) (%) vs PAM-N concentration of coagulated lateritesuspensions of 300 NTU initial turbidity

Fig 11 Zeta potential (mV) (%) vs PAM-N (a), PAM-C (b) and PAM-A (c) concentration of coagulated laterite suspensions of 30 NTU initial turbidity.

Likewise, PAM-A adsorbed on laterite particles via hydrogen

bonding between the silanol and aluminol OH groups at the

par-ticle surface and polymer’s primary amide functional groups, but

the amount adsorbed is minimised by the electrostatic repulsion

between the particles negative charges and the negative polymer

segments The polymer adsorbed results in a shift in position

of the plane of shear, hence generating a small decrease in the magnitude of the Zeta potential (Nasser and James [20] , Mpofu

et al [21] ) PAM-A can also adsorb on the positively charged edges of the particles, creating extra negative charges that lower

Fig 12 Turbidity reduction (%) and Zeta potential of flocculated laterite suspensions containing AlCl3at the critical concentration for coagulation, 30 NTU initialturbidity

1518 D Dihang et al / Chemical Engineering and Processing 47 (2008) 1509–1519

the overall Zeta potential Moreover, through adsorption on

par-ticle negative surface via a polycation (Al3+, Ca2+, etc.), PAM-A

can similarly lower the particle surface charges The repulsive

forces between polymer segments allow the polymer molecules

to be extended and to produce loops and tails that promote

bridg-ing mechanisms and the formation of large open-structure flocs

(Gregory [3] , McGuire et al [12] ).

For PAM-N concentration less than 5 mg/g, adsorption is

probably patchy due to a contracted conformation and to the

fact that particle surface coverage is less than the optimum As a

consequence, there is a little change in the Zeta potential with the

amount of PAM-N added This suggests that the adsorbed

poly-mer layer thickness has a minor charge-shielding effect caused

by the shift in shear plane at which the Zeta potential is

mea-sured Brooks ( [22] ) noticed that the Zeta potential of particles

should increase after adsorption of a neutral polymer, as long as

the shear plane is not shifted too far from the negative particle

surface, due to the change in the ion distribution in the diffuse

double layer The adsorbed layer thickness of PAM-N on iron

oxides and kaolin plateaus at approximately 2.3 nm after the

addition of 3–5 mg/g of solid (Mpofu et al [21] , McGuire et al.

[12] ) Further addition of flocculant can lead to chemisorptions

on particle surface (Besra et al [10] , Besra et al [23] ), that can

promote aggregation without having any effect on Zeta potential.

The results indicate a greater adsorption of N than

PAM-A, probably due to adsorption on iron oxides (McGuire et al.

[12] ) Bridging mechanism is enhanced by the high suspension

concentration, which enables high molecular weight segments

of the PAM-N to bind towards many particles As a consequence

of this high molecular weight, the amount of PAM-N used for

flocculation diminishes as the suspension turbidity increases.

4 Conclusion

This work consists in destabilising laterite suspensions by

coagulation and flocculation, as usually encountered in the

classical potable water process The results pointed out the

auto-coagulation of laterite clay at turbidity greater than 160 NTU

and the influence on this phenomenon on destabilisation

pro-cess Another important result is the fact that coagulation can

be monitored through Zeta potential measurements, as turbidity

removal is always optimum at ca −20 mV, this value being

inde-pendent of the initial turbidity and other coagulation parameters.

For low polymer concentration, Zeta potential is also a relevant

parameter to study flocculation, as polymer adsorption does not

significantly modify the Zeta potential of the particles.

Coagulation happens by charge neutralisation, the positive

hydrolysis product of AlCl3reducing the global negative surface

charge of the particles Turbidity reduction increases with the

initial turbidity of the suspension.

Flocculation mechanism depends on the type and

concen-tration of flocculant All three polymers investigated here are

efficient in removing turbidity, and their efficiency is

indepen-dent on flocculant type at the CCC, PAM-N being the only one

agreed for drinking water processes At low flocculant

con-centration, the amount of flocculant adsorbed on the particles

surface is constant and independent on the degree of coagulation.

This amount is lower than the adsorption capacity of the particle surface making charge neutralisation limited by the amount and sign of charge carried by the polymer Therefore, the magnitude

of the particle Zeta potential is governed by coagulant dosage.

At higher concentration, PAM-C induces charge reversal due

to excess charge adsorb at the surface Contrarily, adding more PAM-A or PAM-N has little effect on the Zeta potential, as PAM-A exhibit little adsorption capacity on the negative surface and PAM-N can chemisorbs on the particle surface, resulting in negligible effect on the Zeta potential.

Acknowledgements

The authors wish to thank Formulaction (France) for providing access to the Turbiscan on-line equipment, SNF FLO- ERGER, (France), for free samples of polyacrylamides and the French Ministry of Cooperation and CNRS for financial support

to D.D and S.N.K.

References

[1] M.J Lehtola, T.K Nissinen, I.T Miettinen, P.J Martikainen, T Vartiainen,Removal of soft deposits from the distribution system improves the drinkingwater quality, Water Res 38 (2004) 601–610

[2] J Gregory, J Duan, Hydrolysing metal salts as coagulants, Pure Appl.Chem 73 (2001) 2017–2026

[3] J Gregory, Particles in Water, Properties and Processes, IWA Publishing,London, 2005

[4] R Hogg, Flocculation and dewatering, Int J Mineral Process 58 (2000)223–236

[5] P Somasundaran, Principles of flocculation, dispersion and selective culation, in: Proc Int Symp on Fine Particles Processing, Published by

floc-Am Inst Min., Metall., & Pet Engrs., Inc., 1980, pp 947–976.[6] A Ozkan, M Yekeler, Coagulation and flocculation characteristics ofcelestite with different inorganic salts and polymers, Chem Eng Process

43 (2004) 873–879

[7] R Yongue-Fouateu, R.T Ghogomu, J Penaye, G.E Ekodeck, H Stendal,

F Colin, Nickel and cobalt distribution in the laterites of the Lomie region,south-east Cameroon, J African Earth Sci 45 (2006) 33–47

[8] L Besra, D.K Sengupta, S.K Roy, Particle characteristics and their ence on dewatering of kaolin, calcite and quartz suspensions, Int J MineralProcess 59 (2000) 89–112

influ-[9] L Besra, D.K Sengupta, S.K Roy, P Ay, Polymer adsorption: its tion with flocculation and dewatering of kaolin suspension in the presenceand absence of surfactants, Int J Mineral Process 66 (2002) 183–202.[10] L Besra, D.K Sengupta, S.K Roy, Influence of unadsorbed and weaklyadsorbed flocculants on separation properties of kaolin suspensions, Int J.Mineral Process 78 (2006) 101–109

correla-[11] A McFarlane, K.E Bremmell, J Addai-Mensah, Improved dewateringbehavior of clay minerals dispersions via interfacial chemistry and particleinteractions optimization, J Colloid Interface Sci 293 (2006) 116–117.[12] M.J McGuire, J Addai-Mensah, K.E Bremmell, The effect of polymerstructure type, pH and shear on the interfacial chemistry, rheology anddewaterability of model iron oxide dispersions, Colloids Surf A: Physic-ochem Eng Aspects 275 (2006) 153–160

[13] B.C Blakey, D.F James, Characterizing the rheology of laterite slurries,Int J Mineral Process 70 (2003) 23–39

[14] C Hu, H Liu, L Qu, D Wang, J Ru, Coagulation behavior of aluminiumsalts in eutrophic water: significance of Al13species and pH control, Env-iron Sci Technol 40 (2006) 325–331

[15] B.S Lartiges, J.Y Bottero, L.S Derrendinger, B Humbert, P Tekely, H.Suty, Flocculation of colloidal silica with hydrolyzed aluminium: an Al-27solid state NMR investigation, Langmuir 13 (1997) 147–152

[16] J Bratby, Coagulation and Flocculation in Water and Wastewater

Treat-ment, Upland Press, Croydon, 1980

[17] D.S Wang, H.X Tang, J Gregory, Relative importance of charge

neu-tralization and precipitation on coagulation of kaolin with PACI: effect of

sulfate ion, Environ Sci Technol 36 (2002) 1815–1820

[18] J Duan, J Gregory, Coagulation by hydrolysing metal salts, Adv Colloid

Interface Sci 100–102 (2003) 475–502

[19] J Duan, J Gregory, Influence of soluble silica on coagulation by

alu-minium sulphate, Colloids Surf A: Physicochem Eng Aspects 107 (1996)

309–319

[20] M.S Nasser, A.E James, The effect of polyacrylamide charge

den-sity and molecular weight on the flocculation and sedimentation

behaviour of kaolinite suspensions, Sep Purif Technol 52 (2006) 241–252

[21] P Mpofu, J Addai-Mensah, J Ralston, Investigation of the effect

of polymer structure type on flocculation, rheology and dewateringbehaviour of kaolinite dispersions, Int J Mineral Process 71 (2003) 247–268

[22] D.E Brooks, Effect of neutral polymers on the electrokinetic tial of cells and other charged particles, J Colloid Interface Sci 43(1973)

poten-[23] L Besra, D.K Sengupta, S.K Roy, P Ay, Flocculation and dewatering ofkaolin suspensions in the presence of polyacrylamide and surfactants, Int

J Mineral Process 66 (2002) 203–232

The impact of polyacrylamide flocculant solution age

on flocculation performance

A.J Parker Cooperative Research Centre for Hydrometallurgy (CSIRO Minerals),

PO Box 90, Bentley, WA 6982, AustraliaReceived 20 July 2001; received in revised form 25 February 2002; accepted 9 April 2002

Abstract

Aqueous solutions of high molecular weight polyacrylamides used to flocculate mineral slurries undergo time-based changes in their properties Previous studies of the impact of ageing on flocculation performance have focused on time-scales of weeks or months, which has little relevance

to industrial practice In this study, ageing times from 1 h to 6 days were examined Flocculation was achieved continuously in a Couette mixing device (stationary outer cylinder, rotating inner cylinder) The extent of aggregation was assessed from batch settling tests and in situ size characterisation with

a focused beam reflectance measurement (FBRM) probe The polyacrylamide dosages required to achieve measurable flocculation decreased as the flocculant age was increased, with optimal performance attained at 72 h Flocculation using a 1-h-old flocculant solution consumed 75% more polymer than with an optimally prepared stock solution The relationship between hindered settling rate and FBRM chord length measurements was found to be independent of flocculant age, but was sensitive to shifts in aggregate density caused by variations in mixing intensity It is proposed that the early stages of flocculant dissolution involved the release of discrete polymer chains from highly agglomerated species, the former dominating flocculation activity Optimal ageing maximised the discrete polymer concentration available for flocculation, leading to a significant increase in the aggregate size distribution but did not appear to impact upon the aggregate packing structure (density).

D 2002 Elsevier Science B.V All rights reserved.

Keywords: flocculation; polyacrylamide; flocculant ageing; kaolin; aggregate size; settling rate

0301-7516/02/$ - see front matterD 2002 Elsevier Science B.V All rights reserved

PII: S 0 3 0 1 - 7 5 1 6 ( 0 2 ) 0 0 0 3 5 - 2

* Corresponding author Fax: +61-8-9334-8001

E-mail addresses: Andrew.Owen@csiro.au (A.T Owen), Phillip.Fawell@csiro.au (P.D Fawell),Jean.Swift@csiro.au (J.D Swift), John.Farrow@csiro.au (J.B Farrow)

1 Fax: + 61-8-9334-8001

www.elsevier.com/locate/ijminproInt J Miner Process 67 (2002) 123 – 144

1 Introduction

1.1 Flocculation in hydrometallurgy

The efficient solid – liquid separation of mineral suspensions is of critical importance to most hydrometallurgical processing operations A variety of reagents may be added to slurries to enhance sedimentation by inducing aggregation Salts of multivalent cations that effectively reduce the surface charge of the solids are termed coagulants while high molecular weight water-soluble polymers that are sufficiently large to bridge between particles are termed flocculants.

The most frequently used flocculants are polymers derived from the acrylamide monomer (Mortimer, 1991) The nonionic homopolymer (100% acrylamide) is an effective flocculant; however, its activity can be enhanced by copolymerisation with other monomers This can introduce functional groups that have a high affinity for a particular mineral phase (e.g hydroxamate for the iron minerals within bauxite residue), or simply provide charged groups that allow the polymer to take on an extended conformation in solution (e.g carboxylate or sulphonate).

The performance of a flocculant in any application (measured in terms of settling rate, clarity, sediment volume or flocculant consumption) is decided by the complex interplay between a number of factors, many of which have been reviewed in detail (La Mer and Healy, 1963; Mortimer, 1991; Farrow and Swift, 1996b; Hocking et al., 1999; Hogg, 2000) These factors may include:

 Slurry properties such as particle size, surface area, surface charge, solution composition, pH and ionic strength.

 Physical properties of the flocculant, such as molecular weight, charge density and functionality.

 Dynamic aspects of aggregate rupture and formation under applied shear.

The flocculation process is strongly influenced by the solution properties of the polymer molecules Flocculant adsorption and the bridging process depend not only upon the functional interactions with the surface, but also the solution dimensions of the polymer chain The latter is determined by both the molecular weight of the polymer and the conformation it takes in a particular electrolyte solution The application of excessive shear to long chain high molecular weight polymers during make-up or flocculation can lead to chain rupture, substantially reducing the capacity for bridging (Abdel-Alim and Hamielec, 1973; Nagashiro and Tsunoda, 1977; Nakano and Minoura, 1978; Henderson and Wheatley, 1987; Scott et al., 1996)

1.2 Flocculant solution ageing

The change in polymer properties as a function of aqueous solution age is also an important consideration Although such changes may be far less dramatic than those achieved as a consequence of shear degradation, they may still lead to significant reductions in flocculation performance and increased flocculant dosages.

The ageing of polyacrylamide solutions was first reported as a change in solution viscosity over a period of weeks or months Narkis and Rebhun (1966) explained this in terms of a disentanglement of polyacrylamide molecules that were agglomerated during the polymerisation process They stated that the viscosity decrease on ‘‘ageing’’ was not the result of molecular weight degradation, but rather a change in the polymer’s solution configuration.

Shyluk and Stow (1969) observed that there was a rapid and a slow stage in the decrease of viscosity of a polyacrylamide solution with time In proposing a mechanism involving both disentanglement and chemical degradation, they also reported for the first time that ageing reduced the ability of the polymer to flocculate a kaolin suspension Studies into the instability of polyacrylamide solutions have either detected no ageing

(Ma¨chtle, 1982; Henderson and Wheatley, 1987) or attributed change to microbial attack

(Chmelir et al., 1980) , radical attack from residual catalyst (Haas and MacDonald, 1972) , disentanglement (Gardner et al., 1978) or conformational changes (Klein and Westerkamp, 1981; Kulicke and Kniewske, 1981; Kulicke, 1986) The latter is generally accepted, with the polymer initially taking an extended conformation and water attacking intra-polymer hydrogen bonds, leading to a more stable and compact coil over time.

Almost all of the above studies have focused on long-term ageing over periods of up to

100 days Changes over such periods are of no practical interest in hydrometallurgical applications, where the residence time for a flocculant before dosing is rarely more than a day, and in some instances, may be less than an hour.

Gardner et al (1978) did include short ageing times when examining the behaviour of four different polyacrylamide solutions A rapid increase in reduced viscosity was initially observed, reaching a maximum between 5 and 24 h, followed by a much slower decrease This strongly suggests that there may be two distinct processes occurring, the first involving the dissolution of the powder polymer and disentanglement of the chains to give a homogeneous solution, the second involving more subtle conformational changes Clearly, this first stage is of much greater relevance industrially, with this short-term ageing required to ensure maximum activity Despite this, little has been published on the effect of this process on flocculation performance.

1.3 Flocculation performance testing

Hindered settling rates, measured from the rate of descent of a mudline, provide a qualitative indication of the aggregate dimensions, and serve as a guide to the throughput

of gravity thickeners They are normally obtained from cylinder tests, where dilute polymer is dosed into a standard cylinder of slurry and mixed, either by inversion or with plungers While such tests are simple and convenient, they suffer from ill-defined mixing The poor mixing can leave polymer-rich and polymer-poor regions, resulting in irreproducible flocculation (Farrow and Swift, 1996a) Batch flocculation measurements are also strongly dependent on variations in solids concentration and sampling procedures.

Farrow and Swift (1996a,b) developed a new tool to continuously characterise flocculation, which they termed the Shear Vessel An inner-rotating cylinder generates

an annular (Couette) shear zone for reproducible mixing Flocculant may be added at a number of points to change the mixing time, while varying the rotation rate changes the

A.T Owen et al / Int J Miner Process 67 (2002) 123–144 125

agitation intensity The Shear Vessel has been shown to provide highly reproducible results and allows measurement of both hindered settling rate and turbidity through isolation of the flocculated product in a settling tube It flocculates continuously under controlled agitation conditions, and is more appropriate for modelling the operation of a thickener feedwell than cylinder testing.

Hecker et al (1999) substantially increased the information obtained from the Shear Vessel by inserting a focused beam reflectance measurement (FBRM) probe in-line The principle of FBRM has been described in detail previously (Williams et al., 1992; Fawell

et al., 1997) The FBRM probe produces a rotating laser beam highly focused at a point near the probe window When the moving beam intersects the path of a particle or aggregate, some of the light is reflected back to a detector As the tangential velocity of the beam is known, the duration of the reflected light pulse is directly proportional to the intersected chord Thousands of reflected chords are measured each second, generating chord length distributions between 1 and 1000 Am Such in situ distributions not only relate to the size of the aggregates formed, but also indicate the efficiency of flocculation 1.4 Objective

This paper describes investigations into the effect of flocculant solution ageing on the resultant flocculation performance with a standard substrate Continuous flocculation was achieved in a Shear Vessel fitted with an FBRM probe, using hindered settling rates and chord length distributions as measures of performance Flocculant solution ages ranged from 1 h to several days, representing the time period of greatest relevance to most mineral processing operations.

2 Experimental

2.1 Materials

Farrow et al (2000) found that the particle size in kaolin slurries slowly decreased with continual stirring, due to rupturing of pre-existing kaolin aggregates Such changes in effective particle size with time substantially alter the flocculation characteristics of the slurry.

Kaolin (12.5 kg, density 2.68 g cm 3, RF grade Commercial Minerals, Perth, Australia) was soaked overnight in 60 l of deionised water The particle size for such slurries was highly irreproducible, with the d50(as measured by laser diffraction) varying from 7 to 15 Am To reduce the extent of aggregation, the concentrated slurry was recirculated through an in-line mixer (Sulzer SMV4DM25, Wohlen, Switzerland) by means of a self-priming centrifugal pump This was found to reduce the d50 to a fairly constant value of 6 Am following recirculation times of over 3 h For this study, standard recirculation times of 5 h were used The concentrated slurry was then pumped into a stirred 375-l baffled tank and diluted to 45 g l 1solids.

The kaolin size achieved by this method was still greater than that expected for fully dispersed primary particles (d f 2 Am) It was not practical to fully disperse the solids,

and indeed some degree of natural aggregation is observed in all kaolin-based mineral tailings While every effort was made to achieve a reproducible preparation procedure, the large volumes of slurry required meant that slight variations in size between batches were often unavoidable Such variations could result in small shifts in the required flocculant dosages It was therefore necessary to ensure that all direct comparisons of activity for aged flocculants were performed with the same slurry batch.

A commercial high molecular weight nonionic polyacrylamide, well known in mineral processing applications, was used in this study No evidence of acrylate moieties could be detected by FTIR,13C NMR or elemental analysis The weight-average molecular mass

Mw of this polymer as measured by multi-angle laser light scattering (MALLS) was

20  106 Polymer (1 g) was ‘‘wet’’ with absolute ethanol (2 g) and gently shaken by hand for 30 s and then let stand for 2 min Deionised water was then added to give a 0.50-wt.% stock solution This solution was shaken vigorously by hand for 2 – 3 min to ensure the powder was properly dispersed and then mixed gently on a tabletop shaker at 145 rpm for the entirety of the ageing experiments Working solutions (0.005 wt.%) were prepared at the pre-determined ageing times by diluting the stock solution in water The ageing times studied were 1, 3, 5, 24, 48, 72 and 144 h.

2.2 Viscosity measurements

Kinematic viscosities were determined using an Ubbelohde-type capillary viscometer

#0C-C96 with a calibration constant of 0.002882 cSt s 1at 40 jC (Canon Instrument, USA) 0.04 wt.% flocculant solutions were used, with all testing carried out at 40 jC Each diluted solution was passed through a 1.2-Am syringe filter before being transferred into the viscometer.

2.3 Focused beam reflectance measurement

The particle and aggregate dimensions of suspensions at different stages of flocculation were examined in situ by FBRM All such measurements were carried out with an M500 field unit fitted with a laboratory probe (LasentecR, Redmond, WA, USA) The probe had

a 12-mm-diameter flat window located at the base of the probe body (318 mm long, 25

mm diameter) The probe body joins to a larger casing that contains the electric motor for the scanning drive, which is connected to the controlling electronics in the field unit by a 10-m fibre optic cable In all experiments, the focal point of the laser was set to be at the surface of the probe window.

Using version 6.0 acquisition software, scans (10 s duration) were continuously measured over the full range of chord lengths (1 – 1000 Am), split into 90 channels in a logarithmic progression The unweighted chord length distribution is a number-sensitive distribution and provides the best indication of the presence of fines From this distribution, the total counts and mean chord length may be calculated Selected chord length ranges may also be monitored to provide an indication of trends in the fine and coarse regions Applying a length square-weighting (i.e volume weighting) to the chord length distribution accentuates the larger particles within the slurry The mean square-weighted chord length was also calculated.

A.T Owen et al / Int J Miner Process 67 (2002) 123–144 127

2.4 Shear Vessel flocculation assessment

2.4.1 Equipment

The Shear Vessel (Fig 1) consists of an external fixed cylinder (I.D 110 mm) and an internal rotating cylinder (O.D 100 mm) driven by a variable speed motor The annular gap width is 5 mm The inner cylinder is constructed from stainless steel, while for this work, an acrylic outer cylinder was used (for work at high temperatures, a stainless steel outer cylinder is used) The conical section at the base of the cylinders has a pitch of 45j, maintaining the gap of 5 mm.

Flocculant inlets (Ports 1 – 4) are positioned on one side of the outer cylinder, while the slurry inlet port is positioned on the opposite side Introducing the flocculant through different ports alters the ‘‘residence time’’ of the flocculant/feed contact, that is, the duration that the flocculant is mixed with the suspension (this correlates in a general sense with the residence time of the feed suspension within a feedwell) A column

is fitted to the base of the cylinders to allow the through-flow of the suspension, but

it is isolated (through the closure of two valves) for analysis of settling rates and turbidity.

Fig 1 The Shear Vessel system for continuous flocculation

2.4.2 Procedure

Slurry was pumped from the feed tank to the Shear Vessel using a peristaltic pump (‘‘Feed slurry pump’’, MasterflexR 7550-62 L/S: Tygon tubing size 16) at a constant delivery rate of 200 ml min 1 Dilute flocculant delivery was via a second pump (‘‘Flocculant pump’’, MasterflexR 7550-92 L/S: tubing size 14) through the flocculant ports at rates between 6.9 and 12 ml min 1 The dosages obtainable through such a delivery were from 38 to 70 g of flocculant per tonne (g t 1) of (dry) slurry The slurry passed through the vessel and the settling tube, with an underflow peristaltic pump located after the analysis column ensuring a controlled flow rate, set at 95% of the total incoming flows This allowed for f 5% of the flow to report to a levelling outlet, thus maintaining a constant volume of slurry within the vessel.

The agitation conditions experienced by the suspension in the annular gap are controlled by the rotation of the inner cylinder, the agitation intensity being directly proportional to the rotation speed The Shear Vessel was used at a range of rotation speeds from 75 to 400 rpm.

Hindered settling rates were determined by measuring the rate of fall of the mudline of

a sample isolated in the Shear Vessel’s analysis column The residual turbidity (in units of NTU) of the isolated sample after set time periods was measured with an Analite nephelometer The FBRM probe was inserted at an angle of 70j to the flow, allowing material to be suitably presented to the probe window but preventing any solid build-up The flocculation characteristics of the 45 g l 1kaolin suspension were assessed with the Shear Vessel over a range of rotation speeds (i.e the agitation intensity used to mix the flocculant into the feed suspension) In all cases, flocculant was added through Port 2 2.5 Aggregate density measurements

Aggregate density is one of the most important properties to be determined in the evaluation of flocculation performance (Ayyala et al., 1995) While direct measurement of density is not possible, it can be estimated from sedimentation and size data, using a modified version of the Stokes equation for ellipsoidal bodies (Happel and Brenner, 1973) The floc density analyser (FDA) is a device developed for determining microscopic properties of individual aggregates, such as size, shape, settling velocity and density, without removing them from their process liquor Details of the procedure have been described previously by Farrow and Warren (1993) A video camera, coupled to selected high magnification optical lenses, is used to record the sedimentation characteristics of individual aggregates as they settle under the influence of gravity within a thermostated analysis cell To ensure that the aggregates achieve free settling within the cell, dilution of the flocculated suspension by at least a factor of 10 is required The recorded images are then played back to allow the measurement of size and settling rate of the individual aggregates with custom-designed image analysis software The maximum horizontal and vertical dimensions were recorded, with the third dimension of the ellipsoidal aggregates assumed to be on average the same as the horizontal dimension The diameter of the sphere was then determined which had an equivalent Stokes settling velocity to the ellipsoid with these three dimensions A statistically representative number of aggregates (usually 200 – 300) are measured, covering a broad size range.

A.T Owen et al / Int J Miner Process 67 (2002) 123–144 129

Kaolin was flocculated in the Shear Vessel at 100 and 200 rpm with flocculant aged for

72 h at dosages that gave bulk settling rates of 1.7 m h 1 This low settling rate was selected to give a wide range of aggregate sizes while preventing too many large ( > 400 Am) aggregates forming (the fast settling rates of such large species cause them to impact upon smaller aggregates, interfering with free settling) The flocculated slurry was sampled for FDA measurements from below the analysis column and before the underflow pump

At this age, the amount of polymer dissolved is small and does not alter the viscosity to a great degree, while any undissolved polymer is filtered out to prevent blocking of the capillary.

At 3 h, there is a rapid increase in the viscosity followed by a sharp decrease to 5 h At 3

h ageing, the dissolved polymer is expected to be highly tangled producing ‘‘agglomerated polymer’’; as the agglomerates are disentangled, more discrete polymer chains are observed leading to a decrease in the viscosity A further slight decrease in viscosity is observed up to 7 h after which only minimal changes in viscosity occur up to the last measurement time of 144 h After 7 h, it is most likely that only subtle conformational changes of individual polymer chains occur which do not affect the overall solution viscosity but may very well have a large effect on flocculant activity.

Fig 2 Kinematic viscosity at 40 jC as a function of age for flocculant stock solutions (diluted to 0.04 wt.% and1.2 Am filtered before measurement)

3.2 Activity measurements

3.2.1 Effect of agitation intensity

In addition to the viscosity measurements shown in Fig 2 , the activity of each aged flocculant stock solution was assessed with a 45 g l 1 kaolin suspension in the Shear Vessel Each stock solution was diluted to an appropriate level before testing To examine the impact of agitation intensity, the applied dosages were chosen to ensure that measurable aggregation could be detected at settings below 300 rpm, without the onset

of over-flocculation at 100 rpm In most instances, the selected dosages produced similar settling rates at 200 rpm.

From Fig 3 , it can be seen that the applied agitation had a major effect on the extent

of aggregation achieved Under mild mixing (100 rpm), large, voluminous aggregates were readily formed and fast settling rates were measured As the agitation intensity increased, the measured settling rates decreased sharply, indicating a substantial reduction in the size of the aggregates achieved The maximum aggregate size that can survive under the applied shear conditions is expected to diminish at higher agitation intensities While it is possible that some degree of densification may also occur for aggregates formed under stronger mixing, this cannot be ascertained from the settling data alone.

The settling rate diminished rapidly up to an agitation intensity of 200 rpm At this point, the settling rate was low, typically in the range 2 – 3 m h 1, but still in excess of that for the unflocculated suspension ( f 0.3 m h 1) Any further increase in agitation intensity led to only a minor additional deterioration in flocculation performance.

Fig 3 Effect of Shear Vessel agitation intensity and flocculant stock solution age on hindered settling rates for theflocculation of kaolin (45 g l 1)

A.T Owen et al / Int J Miner Process 67 (2002) 123–144 131

low aggregate strength Farrow and Swift (1996b) found that the flocculation performance profile observed as a function of agitation intensity may vary for different flocculant pro- ducts, with the aggregates formed by some polymers able to tolerate more intense mixing.

A clearer picture of the impact of agitation intensity may be obtained from examination

of the FBRM chord length distributions from flocculation at both 100 and 200 rpm for the same dosages applied in Fig 3 Flocculation led to a significant drop in the total number of unweighted counts, with the formation of large aggregates indicated by a higher fraction of longer chords (Fig 4) While such aggregates contribute far fewer counts to the overall distribution than their unflocculated primary particles, they represent the dominant fraction

of the slurry volume Fig 4a shows that at 100 rpm, the flocculated distribution was typically broad and very low in counts Large aggregates were readily formed; however, the presence of bimodal character in the distributions, with a distinct fraction < 10 Am, suggests that the capture of fines during flocculation may not be fully efficient.

At 200 rpm, the total counts were higher, a consequence of smaller aggregates being formed (Fig 4b) However, the proportion of counts < 10 Am was less than that at 100 rpm, which may be indicative of more efficient fines capture under the stronger mixing, perhaps due to more efficient mixing of the flocculant with the slurry.

3.2.2 Effect of stock solution age

This general behaviour described in Section 3.2.1 was observed across the range of

h were identical to that obtained for 72 h and are not presented in Fig 4 for simplicity) At

144 h, the increase in counts relative to the 72-h-old stock solution was evidence of a reduction in flocculant activity.

Fig 4 Effect of flocculant stock solution age on FBRM unweighted chord length distributions for the flocculation

of kaolin (45 g l 1) at Shear Vessel agitation intensities of (a) 100 rpm and (b) 200 rpm (dosages as given inFig 3)

A.T Owen et al / Int J Miner Process 67 (2002) 123–144 133

Bimodal character was observed at 100 rpm for flocculant ages of 5 h and over, but not for the 1- and 3-h-old solutions (Fig 4a) It may be that at the faster settling rates measured for the former solutions, sedimentation was not fully hindered, that is, the settling network was less able to capture and ‘‘drag down’’ residual fine, unflocculated particles This was confirmed by examining the equivalent distributions acquired at lower flocculant dosages, for which bimodal character was far less evident (shown in Fig 6a for the 3-h-old stock).

3.2.3 Effect of dosage

A series of experiments was conducted to establish the full dosage response for each flocculant age at agitation intensities of 100 and 200 rpm In view of the large volume of kaolin slurry and time required for Shear Vessel testing, it was not possible to examine dosage effects concurrently with the agitation studies described in the previous section A separate set of aged flocculants was therefore examined, using another batch of kaolin slurry Slightly higher dosages were required for this slurry, reflecting a reduced degree of natural aggregation (Section 2.1), but otherwise, the general trends were unchanged.

Fig 5a confirms that fast settling rates may be achieved at 100 rpm for any flocculant age if a sufficiently high flocculant dosage is applied The slopes of the dosage-response curves were similar, but shifted to higher dosages for flocculant ages below 24 h For example, 75% and 25% more flocculant were required to achieve a settling rate of

15 m h 1for 1- and 3-h-old solutions relative to solutions aged for 24 h or more Even under the more intense mixing at 200 rpm, fast settling rates could be achieved at the expense of higher flocculant dosages (Fig 5b) In the case of the 3-h-old solution, the dosage required for a settling rate of 10 m h 1increased by 40%.

It can be seen from Fig 5 that the addition of flocculant has effectively no impact on the settling rate up to dosages in excess of 30 g t 1 Above this point, the settling rates typically rise sharply over a narrow range of dosages A better understanding of this behaviour is gained from examining the equivalent chord length distributions, shown in

Fig 6 for a 3-h stock solution at 100 rpm At a dosage of 36 g t 1, the measured settling rate was only 1 m h 1, but Fig 6a shows that substantial aggregation had actually taken place, with the total counts dropping from f 35 000 to 16 000 This suggests that the initial stages of flocculation may progress through a fines capture process, forming small, slow settling aggregates Once fines capture has achieved an equilibrium level, further addition of flocculant may lead to ‘‘cluster’’ flocculation of such aggregates into larger bodies (Farrow and Warren, 1993) As the external surface area of these aggregates is low relative to their volume, the formation of cluster aggregates with much higher settling rates may only require a small increase in flocculant dosage.

The application of a length square-weighting to the FBRM chord length distribution effectively represents a volume-based weighting, emphasising the contribution of the larger chords, which in this system are associated with aggregation Fig 6b shows the effect of flocculant dosage on the square-weighted chord length distribution, with higher dosages leading to increases in aggregate size While the aggregate size at a dosage of 36 g t 1 was significantly larger than that of the unflocculated slurry, the volumetric contribution of aggregates at dosages in excess of 50 g t 1 was clearly much greater.

Fig 5 Effect of flocculant dosage and stock solution age on hindered settling rate for kaolin (45 g l 1)flocculated at Shear Vessel agitation intensities of (a) 100 rpm and (b) 200 rpm.

A.T Owen et al / Int J Miner Process 67 (2002) 123–144 135

Fig 6 Effect of flocculant dosage (3-h-old stock solution) on (a) unweighted and (b) length square-weightedchord length distributions for the flocculation of kaolin (45 g l 1) at a Shear Vessel agitation intensity of 100 rpm.

3.2.4 Relationship between settling rate and FBRM chord length data

In the initial study of in situ FBRM detection of aggregates formed within the Shear Vessel, Hecker et al (1999) achieved reasonable correlations between hindered settling rate and the mean square-weighted chord length for the flocculation of kaolin with several different polymer products However, the correlation for each product was achieved over a range of dosages and Shear Vessel agitation intensities The aggregate structure (i.e density) may be expected to vary with conditions, particularly with agitation intensity, and therefore should influence the settling rate – size relationship.

Fig 7a shows the hindered settling rate plotted against the mean square-weighted chord length for the flocculant solutions of different ages over a range of dosages at a fixed Shear

Fig 7 Effect of flocculant stock solution age on hindered settling rate as a function of mean length weighted chord length for the flocculation of kaolin (45 g l 1) at Shear Vessel agitation intensities of (a) 100 rpmand (b) 200 rpm.

square-As expected, aggregate density was consistently higher at 200 rpm across the measured range of sizes For average 100-Am aggregates at 100 and 200 rpm, the densities were 1.028 and 1.040 g cm 3( F 0.005 g cm 3), respectively, corresponding

to volumetric liquor/solid ratios for these aggregates of 60:1 and 42:1, respectively The volumetric contribution of liquor may be inflated by the method of measurement, which assumes an overall ellipsoidal volume based on the maximum vertical and horizontal dimensions for each aggregate, that is, voids within the aggregate structure are therefore included It should also be remembered that in the undiluted slurry, there may be some degree of overlap between these flocculated ellipsoidal volumes However, these values clearly serve to confirm that the correlations between hindered settling rate and in situ chord length measurements have indeed provided an effective indication of changes in aggregate density.

3.3 The dissolution of powder flocculants

Recent studies have demonstrated that aqueous solutions of polyacrylamides, even after several days of ageing, are unlikely to consist of fully dispersed polymer chains.

Ying et al (1996) provided evidence of submicron chain entanglements, detecting solution species of four to six polymer chains by dynamic light scattering Submicron entanglements have also been suggested from light obscuration measurements (Hecker et al., 1998) Hecker et al (2000) , on the basis of flow field-flow fractionation with light

Fig 8 The data fromFig 7overlayed without identifying the different stock solution ages, thereby emphasisingthe effect of Shear Vessel agitation intensity

A.T Owen et al / Int J Miner Process 67 (2002) 123–144 139

scattering detection, proposed three solution states for polyacrylamide—(i) well-dispersed individual coils, (ii) submicron entangled chains and (iii) supramicron agglomerates containing up to thousands of chains The relative proportion of each phase may vary for different commercial products, and as a consequence of concentration, the presence of salts and solution age.

From the results presented above, a general mechanism can be proposed to explain the time-based changes in aqueous flocculant solution activity This may be described schematically by the four stages shown in Fig 10 Stage (a) represents the shortest ageing times, where the solid powder swells to large gel-lumps These gels are few in number compared to the dispersed chains already in solution and play no effective role during flocculation However, despite their low number, they still contain the majority of the polymer mass, which is therefore unavailable for flocculation From comparison with the required dosages for the well-aged stock solutions, approximately 40% of polymer within the 1-h-old stock may still be undispersed.

At Stage (b), the ageing process has progressed to the point that gel lumps are no longer visible, although the polymer chains may be far from fully dispersed The solution

Fig 9 Free settling rate and density versus size as measured by FDA for individual aggregates formed in theShear Vessel at 100 and 200 rpm (72-h-old flocculant stock solution, dosages 40 and 48 g t 1at 100 and 200rpm, respectively)

dimensions of species consisting of several entangled chains may be slightly larger than that for an individual chain, but this is unlikely to compensate for the reduced number of chains available for particle bridging A similar argument may be applied to any supramicron agglomerates Hecker et al (1999) suggested that under some conditions, agglomerates may form a bridge between existing flocculated aggregates to give large, fast settling clusters However, such clusters were very fragile and fines capture was poor due

to the reduced concentration of dispersed polymer.

Stage (c) represents the optimal ageing time for a flocculant solution—dispersion of the polymer chains is either complete or at the maximum level expected for the solvent and the required dosage for flocculation is at a minimum In the present study, this was achieved

Fig 10 Schematic representation of the proposed four main stages for the dissolution and ageing of flocculantsolutions

A.T Owen et al / Int J Miner Process 67 (2002) 123–144 141

after ageing for 72 h Extended ageing, as shown in (d), follows the behaviour expected from previous studies of flocculant ageing Once dispersion is complete, the dominant solution process is likely to be slow chain reconformation, with solvation of intra-polymer hydrogen bonds resulting in the initial extended conformation changing to a more stable, tighter coil (Kulicke and Kniewske, 1981)

It should be emphasised that in this study, the first step in the preparation of all flocculant stock solutions involved the ‘‘wetting’’ of polymer powder with ethanol This step prevents the powder particles sticking together after water addition, thereby ensuring

a more homogeneous dissolution process In extreme cases, poor wetting of flocculant powders can lead to the formation of gels that cannot be dispersed Many other procedures exist for achieving this initial wetting and dispersion of powder particles, often depending

on the scale of preparation For example, the use of a wetting agent may be avoided by carefully adding powder into the vortex formed in the aqueous phase by a suitable impellor rotating at high speed for a strictly limited period The choice of wetting procedure may influence the initial phase of dissolution, that is, Stage (a) in Fig 10 However, the subsequent stages are significantly slower and effectively independent of how Stage (a) is achieved The conclusions from this study should therefore be applicable across all standard wetting procedures.

For laboratory-based studies into the comparative flocculation behaviour of mineral slurries, it is clearly necessary to apply a set ageing period for flocculant stock solutions to ensure stable and reproducible behaviour Of potentially greater significance are the implications for industrial thickening operations We have observed at a number of mineral processing sites that the time-scale from powder flocculant make-up to addition can be less than 24 h, and in some instances, less than 1 h Invariably, the argument against longer ageing is the limited storage capacity available for the stock flocculant solution, although the increase in required dosages for short ageing in part makes this a self-fulfilling arrangement.

4 Conclusions

Continuous flocculation within the Shear Vessel with FBRM detection provides controlled mixing conditions and detailed characterisation of aggregate properties In particular, the relationship between the measured hindered settling rate and mean square- weighted chord length under controlled mixing may be used as a convenient indication of relative aggregate density, avoiding the complications associated with off-line measure- ments on diluted samples.

The dissolution of the powdered polyacrylamide flocculant was found to be a slow process The dosages required to achieve measurable flocculation decreased as the flocculant age was increased up to 72 h, reflecting the release of discrete polymer chains from highly agglomerated solution species Despite the much higher consump- tion of flocculant after short ageing times, under almost all conditions studied, the duration of ageing did not significantly impact upon the density of the aggregates formed, indicating that only the discrete polymer chains influenced the flocculation process.

This research has been supported under the Australian Government’s Cooperative Research Centre (CRC) Program, through the AJ Parker CRC for Hydrometallurgy This support is gratefully acknowledged.

Chmelir, M., Ku¨nschner, A., Barthell, E., 1980 Water soluble acrylamide polymers: 2 Ageing and viscous flow

of aqueous solutions of polyacrylamide and hydrolysed polyacrylamide Angew Makromol Chem 89, 145 –165

Ma¨chtle, W., 1982 Zur alterung von wa¨ßrigen polyacrylamid-lo¨sungen Makromol Chem 183, 2515 – 2525.Mortimer, D.A., 1991 Synthetic polyelectrolytes—a review Polym Int 25, 29 – 41.

Nagashiro, W., Tsunoda, T., 1977 Degradation of polyacrylamide in aqueous solutions by high-speed stirring J.Appl Polym Sci 21, 1149 – 1153

Nakano, A., Minoura, Y., 1978 Degradation of aqueous poly(acrylic acid) and its sodium salt solutions by speed stirring J Appl Polym Sci 22, 2207 – 2215

high-Narkis, N., Rebhun, M., 1966 Ageing effects in measurement of polyacrylamide solution viscosities Polymer(London) 7, 507 – 512

Richardson, J.F., Zaki, W.N., 1954 Sedimentation and fluidisation: Part 1 Trans Inst Chem Eng 32, 35 – 53.Scott, J.P., Fawell, P.D., Ralph, D.E., Farrow, J.B., 1996 The shear degradation of high-molecular-weightflocculant solutions J Appl Polym Sci 62, 2097 – 2106

Shyluk, W.P., Stow, F.S., 1969 Ageing and loss of flocculation activity of aqueous polyacrylamide solutions J.Appl Polym Sci 13, 1023 – 1036

Williams, R.A., Peng, S.J., Naylor, A., 1992 In situ measurement of particle aggregation and breakage kinetics in

a concentrated suspension Powder Technol 73, 75 – 83

Ying, Q., Wu, G., Chu, B., Farinato, R., Jackson, L., 1996 Laser light scattering of poly(acrylamide) in 1 M NaClaqueous solution Macromolecules 29, 4646 – 4654

Contents lists available atScienceDirect

Journal of Hazardous Materials

j o u r n a l h o m e p a g e :w w w e l s e v i e r c o m / l o c a t e / j h a z m a t

Polyaluminium silicate chloride—A systematic study for the preparation and application of an efficient coagulant for water or wastewater treatment

A.I Zouboulis∗, N.D Tzoupanos

Division of Chemical Technology, Department of Chemistry, Aristotle University of Thessaloniki, GR-54124 Thessaloniki, Greece

The coagulation performance of PASiC products were evaluated for the treatment of contaminated tapwater (in terms of turbidity and of NOM removal, as well as of residual Al concentrations and of zeta-potential measurements) Also, they were examined for the tertiary treatment of municipal wastewater(mainly for phosphates removal) Additionally, the new products were compared with the laboratory pre-

experiments were completed with the study of coagulation kinetics by using the Photometric DispersionAnalyzer (PDA), in order to compare the respective floc growth rates Overall, the obtained results suggestthat in order to produce a silica-based polyaluminium coagulant with improved coagulation properties,the basicity (OH/Al ratio) should be between 1.5 and 2.0, the silica content (Al/Si molar ratio) between 10and 15 and should be prepared preferable with the co-polymerization technique However, attention has

to be given in the specific application of these products, as in the case of tertiary wastewater treatment(phosphates removal) more efficient seem to be the silica-based coagulant with lower basicity (i.e OH/Al1–1.5)

© 2008 Elsevier B.V All rights reserved

1 Introduction

Several research efforts have been devoted to improve the

efficiency of coagulation–flocculation process, a basic and

essen-tial treatment technique both for water or wastewater treatment

facilities, especially for those handling hazardous or toxic

liq-uid wastes (e.g tanneries, metal plating, etc.) The tendency was

the production of coagulants with improved properties in

com-parison with the conventional ones, such as aluminium sulfate

(alum, Al2(SO4)3), or aluminium chloride (AlCl3) After studying

the aquatic chemistry and behaviour of simple Al salts, the way

for improvement seemed to be their (partial) polymerization,

∗ Corresponding author at: Division of Chemical Technology, Department of

Chemistry, Aristotle University of Thessaloniki, P.O Box 116, GR-54124 Thessaloniki,

Greece Tel.: +30 2310 997794.

E-mail address:zoubouli@chem.auth.gr (A.I Zouboulis).

before application The result of these efforts was the tion of a range of pre-polymerized aluminium solutions, referred

produc-as polyaluminium chloride (PACl), polyaluminium sulfate (PAS),

or polyaluminium chloro-sulfate (PACS), with variable degrees

of polymerization These products and especially the first one(PACl) are used extensively worldwide during the last two decades,with an ever increasing demand Their properties were intensivelyexamined and have proved to be more efficient in lower dosagesand in wider pH, temperature and colloids concentration ranges,than the conventional simpler products, leading to cost and opera-tive more effective treatment[1,2] Their superiority is related to thedifferent aluminium species distribution in the (pre-)polymerizedsolutions, as compared with the non-polymerized ones and partic-ularly, is connected with the specific polymer Al13(Keggin type)

[3,4]

In spite of their improved properties, the pre-polymerizedcoagulants performance still remains inferior, when compared

to the performance of organic polyelectrolytes, used, e.g in

0304-3894/$ – see front matter © 2008 Elsevier B.V All rights reserved.

water treatment [5] In recent years, the relevant research in

the coagulation–flocculation field was focused mainly in

under-standing the behaviour and aquatic chemistry of pre-polymerized

reagents, such as polyaluminium chloride, and to the further

improvement of their properties The main reason for the higher

efficiency of organic polymers is their higher molecular weight,

which implies better flocculation properties In order to improve

the aggregating power of PACl, some efforts have been made

dur-ing the past few years, regarddur-ing the incorporation of silica in its

structure Hasegawa et al [6]noticed that by introducing metal

ions into polymerized silicic acid solution, the molecular weight of

the product was increased and the respective stability and

coagula-tion performance were further improved In this case, the product

was rather an inorganic metal-polysilicate flocculant, where

sil-ica was the main component, than a pre-polymerized inorganic

coagulant More recently, research has focused in the

incorpora-tion of silica within the pre-polymerized metal soluincorpora-tions, aiming

to products with larger molecular weight by the application of two

techniques; i.e either by introducing polymerized silica in the

pre-polymerized metal solution, or by introducing pre-polymerized silica

in the metal solution, followed by polymerization The first method

is referred as composite polymerization, whereas the latter one as

co-polymerization

The polymerized silica can be obtained from a silicic acid

solu-tion, which has the tendency to polymerize by dehydration and to

form Si–O–Si anhydride bonds, according to the following general

scheme:

nSi(OH)4→ (OH)3Si–O–Si(OH)3(dimmer)→ oligomers

→ colloidal polymers → (SiO)n

The rate of silicic acid polymerization is strongly pH-depended,

being very fast in neutral and slightly alkaline solutions and

extremely slow at the low (acidic) pH values of 2–3[7] In order

to introduce silica in aluminium solution, the gelation must be

avoided and therefore, the maintenance of poly-silicates solution

respective pH should be applied around 2, as the method to handle

silicates for coagulants preparation Applying this concept,

cer-tain efforts have been made to synthesize and to understand the

behaviour and the coagulation efficiency of these coagulants The

coagulation performance particularly, is under extensive

investi-gation during the last few years[5,8–12]in order to evaluate the

advantages of new coagulants However, the coagulation

perfor-mance of silica-based coagulants was evaluated mainly for the

treatment of natural waters and their efficiency has not been

exam-ined thoroughly for the treatment of wastewater Moreover, the

evaluation was based mainly on the removal of turbidity and little

attention was given to residual Al concentration, or to the removal

of (natural) organic matter (NOM) The examination of as many

parameters as possible is desirable to get an intergraded approach

of coagulation performance, especially regarding the residual

alu-minium concentration and the removal of NOM[13]

This study aims to complete, intergrades and extent the

knowl-edge obtained by the aforementioned publications The efficiency

and behaviour of several derivatives of silica-based aluminium

coagulants was systematically examined, aiming to the production

of more generic conclusions Efforts have been made to define the

optimum hydroxyl to aluminium molar ratio and the aluminium

to silica molar ratio during their preparation, in order to

pro-duce improved coagulants, than the conventional (alum), or simple

pre-polymerized (PACl) ones For this purpose, the coagulation

effi-ciency of all prepared coagulants was tested in contaminated tap

water, simulating natural (surface) waters The coagulants with

the more efficient combination of OH/Al and Al/Si molar ratios

were further examined and their coagulation performance was

compared with the coagulation performance of alum, PACl-18 andlaboratory prepared PACl, regarding the impact of pH variation onthe treated sample Furthermore, selected coagulants were appliedfor the treatment of biologically pre-treated urban wastewater,aiming to the removal of phosphates, as relevant data for the respec-tive coagulants applied for wastewater treatment are rather rarelypublished in the literature Additionally, their behaviour and effi-ciency was compared with simple PACl, or alum coagulants Finally,the kinetics and dynamics of flocculation ability were studied forthe first time for silica-based coagulants by using the PhotometricDispersion Analyzer (PDA), a technique which allows the (relative)comparison of flocs’ growth for the tested coagulants during coag-ulation

2 Materials and methods

All used chemical reagents were analytically pure chemicals.Deionized water (with conductivity lower than 0.5␮S/cm) wasused to prepare all the solutions, except of the solutions used forthe synthesis of the coagulants In this case, deionized water madecarbonate free by boiling, was used

For comparison reasons, commercially available PACl-18 taining 17.15% Al2O3, with basicity 40% and density 1.365 g cm−3),PACl-14 (containing 14% Al2O3 with basicity 80% and density1.18 g cm−3), both obtained from Phosphate Fertilizers Industry(Greece), as well as aluminium sulfate (alum, Al2(SO4)3·18H2O,analytical reagent), were also examined

(con-2.1 Procedure for the preparation of coagulants 2.1.1 Preparation of polysilicic acid solution (pSi)

Water glass solution (containing 10% NaOH and 27% SiO2) wasdiluted to 0.5 M SiO2and placed in a plastic beaker Under magneticstirring, HCl (1N) was introduced drop wise until the pH reaches 4.The solution was aged for 90 min in pH 4 and then, the pH wasdecreased to 2, where the solution remained for 60 min before use(containing 0.37–0.38 M SiO2)

2.1.2 Synthesis of silica-based coagulants

The synthesis of coagulants took place at room temperature

by the application of two polymerization methods, accordingmainly to Gao et al.[9], although with certain modifications, i.e

by applying the co-polymerization or the composite tion techniques According to the first procedure, the appropriateamount of pSi was mixed with Al solution and in the mixture wasadded slowly (under magnetic stirring) the appropriate amount ofbase solution in order to achieve the desired OH/Al molar ratio.According to the second technique, the base solution was initiallyadded to the Al solution, creating an intermediate PACl solutionand then, the appropriate amount of pSi was introduced, in order

polymeriza-to achieve the desired Al/Si molar ratio The respective initial tions were 0.5 M AlCl3·6H2O, 0.5 M NaOH (as the added base) andthe aforementioned polysilicic acid solution The base additionrate (achieved by a peristaltic pump) was 0.1 ml/min and the stir-ring speed was 700–800 rpm The final volumes of the obtainedsolutions samples were about 40–65 ml and the final aluminiumconcentration was fixed for all coagulants at 0.13 M

solu-The total number of the newly prepared modified alternativecoagulants was 32, with the OH/Al molar ratios 1, 1.5, 2, or 2.5and the Al/Si molar ratios 5, 10, 15, or 20 Polyaluminium chlo-ride solutions were also prepared (PACllab) for comparison reasonsunder the same conditions, but without the addition of silicates.The coagulants prepared with the co-polymerization technique arereferred as PASiC, whereas the coagulants prepared with the com-posite polymerization technique are referred as PACSi According

to the basicity and to the Al/Si molar ratio, the coagulants are

referred as follows: PASiC with OH/Al = 2 and Al/Si = 10 as PASiC2/10,

while PACl with OH/Al = 2 as PACl2 Please note also that

basic-ity = [(OH/Al)]/3× 100[2] For PACl-18 the basicity is 40, therefore

the OH/Al ratio is 1.2, i.e relatively low, as compared to OH/Al ratio

2 for the case, e.g of PACl2

2.2 Coagulation performance

The zeta-potential was measured by using a Laser Zee Meter

501, the pH by using a Metrohm Herisau pH-Meter and the turbidity

measurements were performed by a HACH RATIO/XR Turbidimeter

The absorbance at 254 nm, as a convenient indicator of

natu-ral organic matter presence, was measured with a Schimadzu

UV/vis spectrophotometer, by using a 1 cm path length quartz

cuvette

2.2.1 Jar-tests using contaminated tap water (simulating surface

water)

For the determination of coagulation efficiency of the prepared

coagulants and their comparison with the performance of

conven-tional ones, a jar-test apparatus (Aqualytic) with six paddles was

used The treated sample (1 L) was made of tap water, clay (kaolin)

suspension (commercially available) and humic acids (Aldrich) The

initial concentration of clay suspended particles was 10 mg/L and

that of humic acids 5 mg/L.Table 1displays the specific properties of

the initial water sample As flocculant aid a common anionic

poly-acrylamide (Magnaflock LT25, Ciba SC LTD, commercially available)

was used in concentrations equal to the 1/10th of the respective

concentration of inorganic coagulants The jar-test experimental

conditions (based upon preliminary relevant experience) are

pre-sented inTable 2 Shortly (within 1 min) after the addition of the

coagulant (i.e during the initial rapid mixing stage) 30 ml of sample

was withdrawn for zeta-potential measurements The flocculant

aid was introduced just 15 s before the initialization of slow

mix-ing period At the end of these experiments about 50 ml of sample

was withdrawn 5 cm below the liquid surface for further analytical

determinations The concentrations of coagulants are expressed as

mg Al/L in the case of PACl and alum, and as mg (Al + Si)/L in the

case of PASiC and PACSi The experiments were repeated two to

three times and the average values are shown in the figures; usually,

the variance between the obtained separately values were within

3–5%

pH adjustment (when needed) was accomplished by the

addi-tion of the proper amount of 2N HCl or NaOH soluaddi-tions in the sample

under stirring

2.2.2 Jar-tests with biologically pre-treated wastewater

The prepared coagulants were also applied for the tertiary

treatment of biologically pre-treated municipal wastewater, to

evaluate their coagulation efficiency mainly by means of

phos-phates removal.Table 1displays the specific properties of the initial

waste water sample The coagulation experiments were conducted

in 500 ml samples and the applied jar-test experimental conditions

are shown inTable 2 The conditions were slightly different from the

respective in the aforementioned experiments (i.e those related to

contaminated tap water treatment), as resulted after preliminary

experiments

2.2.3 Residual aluminium concentration

The residual aluminium concentrations were determined with

the eriochrome cyanine R standard method [14] In dilute and

buffered to pH 6 solutions, aluminium complexes with eriochrome

cyanine R dye and results to a colored compound, which absorbs

light with maximum at 535 nm

2.2.4 Determination of phosphates concentration

The concentration of phosphates was determined with theascorbic acid standard method, according to APHA[14]

2.2.5 Study of coagulation kinetics

The extent of aggregation was also examined and the tion dynamics was accomplished by using a continuous flow opticalflocculation monitor (PDA 2000, Rank Brothers, UK).Fig 1displays

floccula-a schemfloccula-atic difloccula-agrfloccula-am of the experimentfloccula-al set-up, operfloccula-ated in recycling mode The test suspension of 1.5 L tap water, containing

non-5 mg/L of clay and non-5 mg/L of humic acids, was placed in a 2 L beakerand stirred with the paddle of a jar-test apparatus The suspen-sion flows through the measuring transparent plastic cell (having

3 mm diameter), where it was illuminated by a narrow light beam(of 850 nm wavelength), with the help of a peristaltic pump Thepump was placed after the PDA apparatus for preventing the even-tual floc breakage, eventually caused by the mechanical forces of thepump The applied flow rate was 30 ml/min in order to have lami-nar conditions in the transfer tube, hence avoiding flocs breakage.All experiments were conducted at room temperature, without theaddition of flocculant aid, to prevent the excessive growth of flocsand the blockage of connecting tubing (diameter 3 mm)

The PDA measures the average transmitted light intensity (dcvalue) and the root mean square (rms value) of the fluctuatingcomponent The ratio (rms/dc), or flocculation index (FI) provides

a sensitive measure for the aggregation of particles The FI value

is strongly correlated with the respective floc size and alwaysincreases as flocs grow larger, providing a useful (although rela-tive) indication of floc growth, eventually breakage and re-growth,which allows comparisons to be made between the differentcoagulants and under different shear conditions and coagulant con-centrations[15,16]

3 Results and discussion

3.1 Comparison of prepared coagulants

The main purpose of this study was to evaluate the optimumcharacteristics of silica-based coagulants, by means of differ-ent polymerization degrees (employing different OH/Al molarratios), silica content and preparation method, in order to achieveimproved products, when compared to the conventional coagu-lants, such as alum, or pre-polymerized (without the presence

of silica) coagulants, i.e polyaluminium chloride Considering theaforementioned properties of new coagulation reagents, it is worthnoting that the polymerization degree is mainly related with the %content of medium size Al polymers (mainly Al13), whereas theincrease of OH/Al molar ratio generally results to the increment ofthis content, as well as of the polymerization degree

Table 1

Initial sample properties (surface or waste water samples)

Type of sample to be treated Turbidity (NTU) Absorbance UV 254 nm pH Phosphates (mg/L)

The first step in this effort was the preliminary application of all

prepared coagulant samples, aiming to obtain an initial approach

about their coagulation behaviour and to define the relatively most

efficient between them For this purpose, coagulation experiments

were conducted with the synthetic clay-humics (model) sample

The initial concentration of coagulants in this case was 2 mg/L and

the measurements included the determination of residual

turbid-ity, the measurement of UV absorbance at 254 nm, as well as the

residual aluminium concentration

In the case of coagulants prepared with the co-polymerization

technique (denoted as PASiC,Fig 2), it seems that their behaviour

is quite different, in accordance with the obtained polymerization

degree and the silica content It is clear that the relatively worst

results are due to the higher silica content, i.e Al/Si 5 (for a given

OH/Al value), except of the residual aluminium parameter (but only

in the case of PASiC with OH/Al ratio 1.5) Regarding the residual

Al concentration, an important issue from the public health point

of few, which is also connected with the overall performance of

coagulation, it seems that the detrimental effect of higher silica

content is less significant (particularly for OH/Al ratio 1.5–2.5) The

aluminosilicates formation probably leads to molecules of larger

size, which can be removed easily through coagulation and

sedi-mentation Observing the impact of OH/Al ratio on the coagulation

performance, it can be seen that the least effective OH/Al ratio is

1 Based on the evaluation of best performance, e.g the lowest

residual turbidity, the UV absorbance at 254 nm and the residual

aluminium concentration, it seems that the more suitable OH/Al

molar ratio should be 2.0 Specifically, the overall lowest residual

turbidity, as well as lowest residual aluminium concentration, was

achieved by using the sample PASiC 2/10

The coagulants with the highest silica content are also the

less effective in the case of reagents prepared with the

compos-ite polymerization technique (PACSi,Fig 3), with some exceptions,

regarding the residual aluminium concentration Worth noting

is that with OH/Al ratio 2.5, all coagulants exhibit almost the

same performance, regarding the residual aluminium

concentra-tion It is obvious that the hydrolytic reactions taking place and the

respective products of them can differ accordingly to the applied

preparation technique In composite polymerization the higher

polymerization degree (achieved by OH/Al 2.5 ratio)[17], lead to

similar behaviour of coagulants, independently of silica content

The most efficient coagulant regarding turbidity removal seems to

be the PACSi with OH/Al ratio 1.5 and Al/Si ratio 10, whereas

regard-ing the residual Al concentration, seems to be the PACSi with OH/Al

ratio 2 and Al/Si ratio 15 Regarding the UV absorbance removal, the

PACSi with OH/Al 1.5 ratio and Al/Si ratio 15 was found to be the

most efficient reagent

It is obvious that the examined different preparation parameters

can influence significantly the properties of silica-based

coagu-lants Considering the polymerization degree, a medium or high

OH/Al ratio should be desirable, e.g 1.5–2.5 However, during theaforementioned experiments it was noticed that the coagulantswith OH/Al ratio 2.5 were quite unstable, as compared to the oth-ers Within 30 days after preparation, formation of precipitate wasobserved and sudden increase in the stored coagulants turbiditywas recorded The increase of turbidity was more intense in thecoagulants prepared with the co-polymerization technique andthose containing higher silica content Based on these observationsand on the coagulation performance, the more suitable OH/Al molarratio should be 1.5–2.0 The impacts of applied preparation tech-nique and of the Al/Si molar ratio on the coagulation performancewere further examined

3.2 Impact of Al/Si ratio and of preparation technique on coagulation performance

Jar-tests were conducted in order to evaluate the impact

of silica content and of specific preparation technique, i.e polymerization or composite polymerization, on the coagulationefficiency of obtained reagents The selected coagulants, based onthe previous results, were PASiC or PACSi with OH/Al 2.0 The exper-imental conditions were described in the previous Section2.2.1andthe applied coagulant doses were equal to 1, 2, or 3 mg (Al + Si)/L

co-InFigs 4 and 5the results of turbidity and UV254 nmabsorbancereduction, relatively to the Al/Si ratio and to the coagulant dosage,have been presented

In the case of PASiC (Fig 4a, turbidity removal) it is clear thatthe most efficient Al/Si molar ratio is 10, especially when applyinglower coagulant dosages At the concentration of 2 mg (Al + Si)/L,the performance of PASiC products with Al/Si ratio 15 converge

to that of PASiC with Al/Si ratio 10, while when applying evengreater dosages the differences between the efficiency of exam-ined coagulants become rather smaller, except of the coagulantwith the greatest silica content (Al/Si ratio 5) The latter sampleseems to be the least efficient for all the examined concentrations.Almost the same observations can be made, regarding the reduc-tion of UV254 nm absorbance (Fig 4b); PASiC with Al/Si ratio 10seems to be the most efficient and PASiC with Al/Si ratio 5 seems

to be the least efficient The difference in this case is that PASiCwith Al/Si ratio 20 exhibits equal or better performance than PASiCwith Al/Si ratio 15 For the PACSi products, Al/Si ratio seems tohave about the same impact on coagulation performance (Fig 5),besides the PACSi samples with Al/Si ratios 10 and 15, which presentalmost the same coagulation behaviour, regarding turbidity and UVabsorbance reduction Again, the least effective Al/Si molar ratio is5

Fig 6illustrates the turbidity and UV absorbance removal rate(%) with the use of silica-based coagulants It seems that the prepa-ration method does not present any major effect on the respectivecoagulation performance In the case of Al/Si ratios 5, 10 and 20,

Table 2

Coagulation experimental conditions (surface or waste water treatment)

Type of treatment Rapid mixing period Slow mixing period Sedimentation (min)

Duration (min) Mixing rate (rpm) Duration (min) Mixing rate (rpm)

Fig 2 Comparative coagulation experiments of several PASiC samples (prepared

with the co-polymerization technique) Experimental conditions: concentration of

coagulants 2 mg (Al + Si)/L, pH of sample to be treated 7.65; (a) residual turbidity

(initial turbidity 16 NTU), (b) UV absorbance at 254 nm (initial absorbance 0.125),

(c) residual aluminium concentration, with respect to OH/Al ratio.

PASiC seems to exhibit a slightly better performance than PACSi

and only in the case of Al/Si ratio 15 the PACSi seems to act

slightly better to PASiC Generally, the coagulants prepared with

co-polymerization, show better coagulation performance

Based upon these results, it is suggested that a silica-based

coagulant with improved properties should have medium to high

basicity (i.e OH/Al molar ratio 1.5–2.0), medium silica content (i.e

Al/Si molar ratio 10–15) and should be prepared preferable by the

co-polymerization technique In general, the obtained results are in

good agreement with the relevant observations of Gao et al.[5,9,10]

Fig 3 Comparative coagulation experiments of several PACSi samples (prepared

with the composite polymerization technique) Experimental conditions: tration of coagulants 2 mg (Al + Si)/L, pH of sample to be treated 7.65; (a) residual turbidity (initial turbidity 16 NTU), (b) UV absorbance at 254 nm (initial absorbance 0.125), (c) residual aluminium concentration, with respect to OH/Al ratio.

concen-However, Cheng et al.[11]concluded also that OH/Al ratio should bebetween 1.5 and 2, but regarding silica content, the more suitableAl/Si ratio should be 5

3.3 Impact of pH value on the coagulation performance

PASiC 2/10 and PACSi 2/10 samples were selected for the furtherexamination of pH influence on their performance Additionally,they were compared with the performance of laboratory preparedPACl2, which presents relatively higher degree of polymerization

[17], as well as with the commercially available product PACl-18(presenting lower degree of polymerization), and with alum (i.e therespective non-polymerized coagulant) The pH values were variedbetween 6 and 9, the dosage of examined coagulants was 2 mg/L

and the rest experimental conditions were remained the same, as

in the previous Section2.2.1.Fig 7demonstrates the coagulation

behaviour of these coagulants with respect to pH The removal of

turbidity and of UV absorbance at 254 nm, the residual aluminium

concentration and the zeta-potential values of the suspensions after

the addition of examined coagulants are also presented

It can be seen that the less effective coagulant, regarding pH

values greater than 6, is alum All pre-polymerized coagulants

behave better in a wider pH range, especially in the alkaline one,

which relies on the respective mechanisms of coagulation By the

dissolution of any aluminium salt, various hydrolysis reactions

immediately occur and several primary products can are formed

[3,4,18] In the case of non-polymerized, or with very low degree

of polymerization coagulants, the coagulation efficiency is based

mainly on the Al(OH)3precipitate formation, than on the charge

neutralization mechanism, which is favoured in pH values 6.5–7

[19] In the more alkaline region, Al(OH)4−formation begins and

with increasing pH it becomes dominant, resulting in the dramatic

drop of coagulation efficiency

This fact is also obvious, considering the residual Al

concen-tration (Fig 7c), where alum and PACl-18 exhibit the relatively

worst performance This is possibly due to the fact that in the

case of PACllaband silica-based coagulants, the higher degree of

pre-polymerization results in lower monomeric Al3+concentration

and in the presence of medium and larger size Al polymers[17],

Fig 4 Comparative coagulation experiments of PASiC samples with OH/Al ratio

2 and different Al/Si ratios Experimental conditions: concentration of coagulants

1–3 mg (Al + Si)/L, pH of sample to be treated 7.65; (a) residual turbidity (initial

turbidity 16 NTU), (b) UV absorbance at 254 nm (initial absorbance 0.125), with

Fig 5 Comparative coagulation experiments of PACSi samples with OH/Al ratio

2 and different Al/Si ratios Experimental conditions: concentration of coagulants 1–3 mg (Al + Si)/L, pH of sample to be treated 7.65; (a) residual turbidity (initial turbidity 16 NTU), (b) UV absorbance at 254 nm (initial absorbance 0.125), with respect to Al/Si ratio.

which are more resistant to further hydrolysis In this case it can

be suggested that the coagulation behaviour is at certain degree

“controlled” and it is not any longer depended upon the Al(OH)3precipitate formation The relevant mechanisms rely more uponthe charge neutralization mechanism, as well as to bridging andadsorption interactions between the colloids to be removed andthe Al polymers

However, at the pH value 6 the behaviour of coagulants is ent Regarding turbidity and UV absorbance removal, PACl-18 andalum seem to be more effective From the zeta-potential measure-ments (Fig 7d), it can be observed that at this pH value a chargereversal occurs in all cases, possibly leading to the re-stabilization

differ-of colloids The bigger the polymerization degree differ-of the lant, the greater the positive zeta-potential of resulting colloids is,although leading eventually to re-stabilization phenomena Never-theless, the residual aluminium concentration is again higher forthe case of alum, as shown inFig 7c Considering the fact that thehydrolysis of Al3+is retarded in pH 6 and that in this pH value theAl(OH)3specie exists mainly in dissolved form, the significance ofpre-polymerized aluminium species is obvious, and especially inthe case of silica-based coagulants, because they found to be themost effective

coagu-The pre-polymerized coagulants exhibit similar performance,regarding the removal of turbidity Slightly better seem to be thelaboratory PACl2 and the PASiC samples Regarding the reduction of

UV absorbance at 254 nm, the respective differences are easier to be

Fig 6 (a) Turbidity and (b) absorbance at 254 nm removal rate (%) for all prepared

coagulants with OH/Al 2 and Al/Si ratio 5–20.

distinguished, i.e PASiC and PACl present again better performance,

with little differences between them The relevant advantage in

the use of the silica-based coagulants can be concluded from the

lower residual aluminium concentration (Fig 7c) The differences

are quite large, when compared to the coagulants with none or

lower polymerization degree, such as alum or PACl-18 PASiC and

PACSi seem to create significantly less residual aluminium

concen-tration, especially at the higher (alkaline) pH region, whereas alum

and PACl-18 seem to be less appropriate for use, due to the

respec-tive higher residual aluminium concentration In the pH range 7–9,

i.e the usual pH range of natural waters, the residual aluminium

concentration remains under (or close) to 200␮g/L, which is the

EU maximum permissible concentration limit for drinking water

([20]) Considering that this concentration refers to total aluminium

and not only to the dissolved fraction, the superiority of silica-based

coagulants is obvious Moreover, PASiC and PACSi seem to behave

even better than the laboratory prepared PACl, although the later is

presenting higher polymerization degree, and between them PASiC

seems to be slightly better than the PACSi reagent It is noted that

the addition of silicates into PACl with certain OH/Al molar ratio can

result in the decrement of medium Al polymers content; therefore,

PACl has greater polymerization degree, than the PASiC or PACSi

samples, although with the same OH/Al molar ratio[17] The

forma-tion of aluminosilicate complexes seems to play a significant role,

probably because the presence of silica enhances the resistance of

aluminium species for further hydrolysis

Generally, the change of pH seems to have little effect on the

coagulation performance of PASiC and PACSi The optimum pH

range of alum is around the pH 7 and that of PACl-18 around the pH

7–8 Laboratory prepared PACl2 (with relatively higher tion degree) is effective in the pH range 6.5–8.5, whereas for silicabased coagulants and especially for PASiC the optimum pH rangecan be even broaden to pH values 6.5–9

polymeriza-InFig 7d the change of zeta-potential values with respect tothe pH values of test suspensions after the application of coagu-lants is presented In all examined coagulants a similar pattern wasobserved, i.e rather lower (more negative) values were observed

at the pH around 9, where the charge neutralization is less tive, whereas an increase of zeta-potential values was noticed(hence becoming less negative) towards the more acidic pH val-ues In all the examined cases at pH around 6 a charge reversal wasobserved Clearly, the greatest effect on surface charge exhibits thecoagulant with the highest polymerization degree, i.e PACl2[17],and it can be supposed that it presents also the greatest chargeneutralization capacity The least effective products were alumand PACl-18 PASiC and PACSi show moderate effect, with similarbehaviour The incorporation of silica reduces the charge neutral-ization capability of coagulants, a negative effect which seems to becompensated by the increase of colloids/particle size of these coag-ulants, as observed also in literatures[8,9] Additionally, observingthat the PACl, PASiC and PACSi samples at pH around 7 show zeta-potential measurements of created colloids/particles closer to zeroand hence, approaching the charge neutralization mechanism, thebetter performance of them in circum-neutral pH value is under-standable The better performance of silica-based coagulants withthe variation of pH values was also observed by other researchers

effec-[5,8,9,11], although it was not described with details

It should be mentioned that after the addition of aluminiumcoagulants in the conducted coagulation experiments, a decrease inthe pH values of the samples was observed (data not shown), espe-cially in the case of alum The acidic character of Al3+cation and itshydrolysis products (possessing higher positive charge) are respon-sible for this decrease The pre-polymerized coagulants containalready a percentage of these resistant to further hydrolysis speciesand show a lower (i.e less significant) effect to the pH value of sam-ples As the basicity, or the OH/Al molar ratio further increases,this impact is less noticeable In the case of silica-based coagu-lants, the incorporation of silica chains into the structure of PAClresults in a decrement of coagulants charge density, as shown bythe respective zeta-potential measurements Furthermore, the pos-sible conjunction of aluminium species with silica bridges enhancesthe resistance of them to hydrolysis As a consequence, the silica-based coagulants have shown a weaker influence on the pH values

of samples, than the simple pre-polymerized coagulants

3.4 Phosphates removal

Regarding the phosphates removal efficiency, all PASiC sampleswith molar ratio Al/Si 10 were tested and compared to the commer-cially available PACl-18, PACl-14 and alum coagulants The aqueoussample to be treated was biologically pre-treated urban wastew-ater, collected from the exit of a full-scale wastewater treatmentplant (Thessaloniki, N Greece) The main characteristics were: ini-tial concentration of inorganic phosphates 10.1 mg/L (measured as

ortho-phosphates), turbidity 13.5 NTU, UV absorbance (at 254 nm)

0.181 and pH 8

All tested coagulants were found to be efficient for the removal

of turbidity (i.e residual turbidity was below 0.15 NTU for allcases—data not shown) However, their efficiency was not equallygood, regarding the removal of natural organic matter, althoughtheir behaviour was similar, except that of PASiC1/10 sample, whichshowed better performance (data not shown).Fig 8a illustrates theresults of phosphates removal for initial coagulant concentrations10–50 mg/L It can be seen that the less effective coagulant was