Effect of coagulation conditions on the dewatering properties of sludges produced in drinking water treatment

https://doi.org/10.1016/j.colsurfa.2009.06.013Get rights and content

Abstract

The coagulation and flocculation processes in conventional drinking water treatment generate aggregates which settle to form a sludge waste. This sludge can be dewatered further by thickening, centrifugation and filtration operations in order to recover water and minimise the volume of the waste stream.

A range of water treatment sludges generated in the laboratory were characterised according to a phenomenological method that is valid from the dilute free-settling regime to the concentrated cake compression stages. These were compared with plant samples.

Experimental results show that raw water natural organic matter (NOM), coagulant dose and coagulation pH affected both the rate and potential extent of dewatering. Similar effects were observed for both aluminium sulfate and ferric chloride. These results suggest that increasing dose or pH leads to an increase in the proportion of rapidly precipitated material in the sludge or flocs, which form looser aggregates and hence exhibit inferior dewatering properties.

Introduction

The production of potable water is conventionally carried out by coagulation with a hydrolysing metal salt such as aluminium sulfate (‘alum’) or ferric chloride (‘ferric’). This process is effective at removing turbidity, colour, and micro-organisms, but also results in a waste by-product as the coagulant precipitates into particles that aggregate to form ‘flocs’. Settling of these flocs results in a sludge that can be thickened, centrifuged or filtered prior to ultimate disposal. These dewatering procedures reduce the volume of the waste stream, with both environmental and financial benefits.

Separation of the solid aggregates from the water is influenced by numerous factors, including structural configuration and density differences between the solid and liquid, depending upon the dewatering mechanism employed. However, from an operational perspective, it is useful to gain an understanding of how such practical parameters as coagulant dose and coagulation pH affect dewatering performance, or ‘dewaterability’.

A phenomenological theory developed by Landman, White and others provides a rigorous approach to modelling the dewatering behaviour of compressible materials, and has been adopted to model various dewatering unit operations [1], [2], [3]. In this approach the dewaterability is characterised in terms of an equilibrium term, the compressive yield stress, py, and a kinetic term, the hindered settling function, R, which both depend only on the volume fraction of solids, ϕ, for a given material. The compressive yield stress can be defined as the maximum stress that a given material is able to withstand without undergoing consolidation. Formally the hindered settling function is obtained from a force balance as a factor accounting for the enhancement of hydrodynamic retardation beyond that expected for a single particle in an infinite medium [1], [4]. R can also be expressed in terms of related hydrodynamic parameters [5], such as the Darcian permeability, KD,R=(1ϕ)2ϕηLKD,in which ηL is the fluid viscosity. Knowledge of these parameters allows prediction and optimisation of dewatering operations; conversely, monitoring dewatering performance allows the computation of these parameters. This method is valid for any suspension or particulate cake, and its application to water treatment plant (WTP) sludges was introduced by Aziz and co-workers [6], [7], [8].

This early work reported the solids diffusivity, D(ϕ), which is a summary parameter combining some information from both the py(ϕ) and R(ϕ) curves, useful for quick comparisons. However, D(ϕ) does not capture all of the information [9]: some unit operations are permeability limited, while others are limited more by the equilibrium dewaterability, and so the ability to separately assess the two constraints R(ϕ) and py(ϕ) is advantageous. Additionally the computation of D(ϕ) is less robust, normally depending upon numerical differentiation of either py or the permeate flux rate with respect to ϕ.

In order to obtain ϕ, the density of the solid phase, ρS, must generally be used. It is important to use an accurate value for ρS, to ensure the py(ϕ) and R(ϕ) curves are not inappropriately shifted. A recent extensive survey of the literature in this area [10] shows that a reliable single value for ρS is not available. Nominal values of 2000–3000 kg/m3 have been quoted for alum sludges and 1400–3400 kg/m3 for ferric sludges. Published measurements show almost as much variation, with estimates of 1900–2800 kg/m3 for alum sludges and 1900–2860 kg/m3 for ferric sludges. The wide ranges are indicative of substantial uncertainty, and could yield large errors if applied to the present materials, which is why we chose to make our own measurements.

On a fundamental level the sludge dewaterability is controlled by the chemical composition and physical configuration of the aggregates or flocs that make up the sludge. It is hypothesised that the coagulation conditions will determine these parameters and hence determine sludge dewaterability. Important variables in the coagulation process are the coagulant dose, coagulation pH and the raw water quality – for example, how much natural organic and inorganic material it carries. These variables will dictate the composition of the sludge, for example the proportion of natural organic matter (NOM) and the phase of coagulant precipitate formed. The proportional composition of the aggregates in the sludge was demonstrated to affect dewatering behaviour in an earlier study of synthetic sludge [11] and will be quantified in more detail here.

It is anticipated that the size and physical structure of the aggregates will also be affected by the coagulation conditions. A given mass of aggregated primary particles could take on a ‘streamlined’, compact form, which would provide little resistance to either settling or permeation (the same physical process, with different frames of reference) and allow efficient packing. Alternatively, a porous structure with the same mass would have a greater resistance, and would achieve a lower equilibrium solid volume fraction even when close packed, unless the structure were crushed.

Much published research in the area of drinking water coagulation and aggregate formation uses the fractal dimension to characterise aggregate structure [12] and it is often interpreted as a measure of aggregate compactness. For real aggregates there are numerous limitations to the application of fractal theory [10]. Nevertheless, estimates of the fractal dimension provide an indication of structure from which inferences may be drawn regarding behaviour in settling, up to the gel point, ϕg, at which concentration the particles form a space-filling network. Inferences about behaviour at higher concentrations, such as in filtration, are less justified due to the extreme structural changes, which progressively destroy the fractal particle arrangements.

A common application of fractal theory distinguishes the result of rapid (diffusion-limited) aggregation and slow (reaction-limited) aggregation [13], [14], [15]. The latter is known to generally favour the formation of more compact aggregate structures (of higher fractal dimension). Rapid reaction may be expected when the concentration of particles is high, which may be from high turbidity, but more likely from a large amount of precipitate generated in ‘sweep’ or enhanced coagulation from the dosed chemicals. Clearly, a larger amount of precipitate will be generated from a larger dose. However, we also expect that pH will influence the rate.

The overall solubility of the metal is pH-dependent. In theory a decrease in solubility will lead to an increase in the ‘driving force’ for precipitation. In practice, however, the coagulant doses tend to be at least an order of magnitude greater than the solubility, so that this effect is negligible. Thus the most important pH effect is purely based on the concentration of OH ions that drives the progressive hydrolysis of Al3+ and Fe3+. The rate of precipitation and coagulation of ferric salts has been related to the molar hydrolysis ratio, [OH]added/[Fe], and the equivalent for Al [16], [17], [18], [19], [20], [21]. It has been reported that there exists a threshold value of [OH]added/[Fe] in the range 2.7–2.8, above which rapid polymerisation occurs, yielding a poorly ordered precipitate [19], [20]. The hydrolysis ratio is less commonly used for Al, but an indicative threshold for the onset of rapid precipitation and coagulation of about 2.6–2.8 seems reasonable [16], [22], [23], similar to the value for Fe. The prevailing pH also affects the surface charge of existing species, such as NOM, thereby influencing the interaction of these species with the precipitate (or ions).

A pioneering systematic investigation by Knocke et al. [24] found that the dewaterability of both alum and ferric sludges improved as the coagulation pH decreased from ∼8 to ∼6 – across settling and filtration, in both rate and extent. Larger aggregates formed at high pH, but these exhibited poorer dewatering because of their lower density (which dominated any enhancement due to size) [24]. Tambo and Watanabe [25] showed that decreasing the pH from 8.0 to 6.5 led to a decrease in fractal dimension at low alum doses, but had little effect at higher alum doses. This is consistent with the solubility of aluminium, which is minimum at about pH 6 [26]; however the difference in solubilities would only be important for low total Al concentrations. Bottero et al. [17] reported increases in aggregate fractal dimension as the hydrolysis of aluminium (from aluminium chloride) was increased (i.e. increasing neutralisation) for a given pH, as well as for decreases in pH from 7.5 to 4.5 with a given hydrolysis ratio.

For turbid raw waters the sludge properties can also be affected by the proportion of precipitated coagulant to naturally present particles; arguably because the natural suspended solids are larger, and so denser, sludge produced from turbid water at low coagulant doses is widely reported to dewater faster and further [27], [28], [29], [30], [31]. (The ‘primary particles’ formed by precipitating coagulant may be as small as ∼2 nm [17], [18], [19], [32].) Tambo and Watanabe [25] measured a decrease in aggregate density as alum dose was increased in their coagulation of kaolinite clay suspensions. This is also consistent with the trend found by Dixon et al. [11] for py(ϕ) to increase with greater proportions of alum in the sludge. Some researchers have observed a maximum in aggregate strength at intermediate coagulant doses [33], [34], [35], [36], perhaps coinciding with a maximum in the bulk density of the sludge [37]. Strength was also observed to decrease with increasing pH [33].

Most of the literature comparing alum and iron sludges reports that the ferric sludges dewater further [29], [38], [39], [40] and faster [30], [41], [42] than alum sludges [43]. Yet the observations are often not quantified. Aggregates generated by coagulation of kaolinite clay from both alum and ferric “at the optimum coagulation conditions” did not show any significant differences in fractal dimension [25]. It seems that aggregates formed in turbid waters may have a structure similar to that formed by the precipitation of coagulant in pure water [17].

Natural fluctuations in raw water quality can cause large changes in the consistency of WTP sludge produced [40], [44], [45] through changes in the size, morphology, and strength of the underlying aggregate or floc structure [46]. However there is disagreement in the literature over whether NOM is beneficial to sludge quality [47], [48], detrimental [11], [39], [44], [49], [50], [51], [52], [53], or has no significant effect [54]. This discrepancy can be attributed to differences and limitations in the characterisation techniques employed, as well as differences in sludge concentrations and inherent dewatering properties.

Though not conclusive, the data of Tambo and Watanabe [25] suggest a decrease in aggregate density with increasing proportion of coloured organics relative to mineral (clay) content in the raw water. This is consistent with the trend found by Dixon et al. [11] for ‘model’ alum sludges, in which py(ϕ) increased in the presence of humic substances.

In contrast, a detailed investigation by Bottero et al. [17] indicated that aggregate density increased with increasing organic content, especially in the case of strongly chelating or bridging species, which was attributed to competition with OH ions for hydrolysis sites, which would otherwise form hydroxo bonds. Also, the presence of NOM has been reported to significantly increase the fractal dimension of aggregates formed from ferric chloride [47]. In conditions typical of conventional water treatment, the fractal dimension was estimated at 2.4 (pH 5.5) and 2.3 (pH 7.5) for a turbid, low TOC river water, and 2.9 (pH 5.5) and ∼2.1 (pH 7.5) for a non-turbid lake water of moderate TOC [47]. The decrease in fractal dimension at higher pH was attributed to adoption of less dense conformations by the NOM due to increased electrostatic repulsion arising from enhanced deprotonation [47].

This work seeks to better understand the relationship between clarification conditions and the dewaterability of the resulting sludge by focussing on exercising careful control over the sludge generation and handling procedures. Particular attention has been paid to obtain characterisation of both equilibrium and dynamic parameters across the full range of solid volume fraction of interest. The objective is the first quantitative characterisation of WTP sludge dewaterability across a broad range of systematically varied treatment conditions following from the analysis of experimental measurements according to a rigorous phenomenological theory.

Section snippets

Materials and methods

Sludge was prepared in the laboratory by addition of AR grade chemicals to 48 L batches of raw water obtained from a reservoir supplying Melbourne, Australia. Aluminium sulfate was added at 1 g(Al)/L, or 10 g(Al)/L for the highest doses. Ferric chloride was added at 2 g(Fe)/L, or 20 g(Fe)/L for the highest doses. Sodium hydroxide was added at 1.0 M, or 0.2 M for the lowest doses. These stock solutions were made up with deionised, RO water (Elix 3, Millipore), which was also used as a pure water blank.

Sludge composition and solid phase density

As noted, knowledge of the solid phase density is necessary to accurately compute the solid volume fraction, which is required to ensure a valid basis for comparison of dewatering properties. The sludge is made up of precipitated coagulant, in the form of a metal oxide or hydroxide, and material removed from the raw water.

Material removed from the raw water is mostly either suspended particles such as clay or sand, or dissolved natural organic matter – other matter such as protozoa or activated

Conclusions

The conditions under which coagulation is carried out have been found to influence the dewatering behaviour of the resulting precipitate across a range of solids concentrations from settling through to filtration. This dewatering behaviour has been fully characterised for each material in terms of the kinetic and equilibrium parameters R(ϕ) and py(ϕ).

For both ferric and alum sludges an improvement in dewatering was observed with a decrease in coagulant dose below a critical value of about 5 mg/L

Acknowledgements

The authors wish to acknowledge United Utilities (UK) and Yorkshire Water for project sponsorship, and the Australian Research Council and The University of Melbourne for provision of a postgraduate scholarship. Melbourne Water, the CRC for Water Quality and Treatment, United Utilities Australia, and United Water are acknowledged for assistance in obtaining samples. The assistance of Dr. Shane P. Usher and Dr. Ross G. de Kretser is also acknowledged.

References (73)

  • K.A. Landman et al.

    Adv. Colloid Interface Sci.

    (1994)
  • R. Bürger et al.

    Chem. Eng. J.

    (2000)
  • K.A. Northcott et al.

    Chem. Eng. Sci.

    (2005)
  • G.C. Bushell et al.

    Adv. Colloid Interface Sci.

    (2002)
  • R.J. Stol et al.

    J. Colloid Interface Sci.

    (1976)
  • N. Tambo et al.

    Water Res.

    (1979)
  • P.J. Murphy et al.

    J. Colloid Interface Sci.

    (1976)
  • D.H. Bache et al.

    Water Sci. Technol.

    (1997)
  • A. Vilgé-Ritter et al.

    Colloids Surf. A

    (1999)
  • L. Ho et al.

    Water Res.

    (2005)
  • P. Jarvis et al.

    Water Res.

    (2006)
  • J.H.A. van der Woude et al.

    Colloids Surf.

    (1984)
  • J.J. Ducoste

    Chem. Eng. Sci.

    (2002)
  • T. Thorsen

    J. Membr. Sci.

    (2004)
  • E.T. Gjessing et al.

    Environ. Int.

    (1999)
  • J. Duan et al.

    Adv. Colloid Interface Sci.

    (2003)
  • B.A. Bolto

    Prog. Polym. Sci.

    (1995)
  • K.A. Landman et al.

    AIChE J.

    (1997)
  • R.G. de Kretser et al.

    AIChE J.

    (2001)
  • R.G. de Kretser et al.

    Filtration

    (2007)
  • A.A.A. Aziz et al.

    Water Sci. Technol.

    (2000)
  • P.J. Harbour et al.

    Water Sci. Technol.

    (2001)
  • P.J. Harbour et al.

    J. Water Supply Res. Technol. AQUA

    (2004)
  • D.I. Verrelli, Drinking Water Treatment Sludge Production and Dewaterability. Ph.D. Thesis, The University of...
  • D.R. Dixon et al.

    J. Water Supply Res. Technol. AQUA

    (2004)
  • D.-H. Li et al.

    Environ. Sci. Technol.

    (1989)
  • D.A. Weitz et al.

    Phys. Rev. Lett.

    (1985)
  • D.W. Schaefer

    Science

    (1989)
  • P.M. Bertsch et al.

    Aqueous polynuclear aluminum species

  • J.Y. Bottero et al.

    J. Phys. IV – Colloque

    (1993)
  • J.Y. Bottero et al.

    Langmuir

    (1994)
  • D. Tchoubar et al.

    Langmuir

    (1991)
  • R.M. Cornell et al.

    The Iron Oxides: Structure, Properties, Reactions, Occurence and Uses

    (1996)
  • R.M. Cornell et al.

    The Iron Oxides: Structure, Properties, Reactions, Occurences and Uses

    (2003)
  • P.H. Hsu et al.

    Mineralogical Magazine

    (1964)
  • W.R. Knocke et al.

    J. AWWA

    (1987)
  • Cited by (48)

    • NOM removal by coagulation

      2022, Natural Organic Matter in Water: Characterization, Treatment Methods, and Climate Change Impact, Second Edition
    View all citing articles on Scopus
    View full text