w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Characterization of natural organic matter adsorption in granular activated carbon adsorbers

Silvana Velten a,*, Detlef R.U. Knappe b, Jacqueline Traber a, Hans-Peter Kaiser c, bastien Meylan a,e Urs von Gunten a,d, Markus Boller a, Se
a

Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Dubendorf, Switzerland Department of Civil, Construction, and Environmental Engineering, North Carolina State University, Raleigh, NC 27695-7908, USA c Stadt Zurich, Wasserversorgung, Hardhof 9, Postfach 1179, 8021 Zurich, Switzerland d Institute of Biogeochemistry and Pollutant Dynamics, ETH Zurich, 8092 Zurich, Switzerland e Cimo, Route de lIle-au-Bois, 1870 Monthey, Switzerland
b

article info
Article history: Received 12 August 2010 Received in revised form 31 March 2011 Accepted 28 April 2011 Available online 6 May 2011 Keywords: Adsorption Granular activated carbon Natural organic matter (NOM) NOM characterization Size exclusion chromatography

abstract
The removal of natural organic matter (NOM) from lake water was studied in two pilotscale adsorbers containing granular activated carbon (GAC) with different physical properties. To study the adsorption behavior of individual NOM fractions as a function of time and adsorber depth, NOM was fractionated by size exclusion chromatography (SEC) into biopolymers, humics, building blocks, and low molecular weight (LMW) organics, and NOM fractions were quantied by both ultraviolet and organic carbon detectors. High molecular weight biopolymers were not retained in the two adsorbers. In contrast, humic substances, building blocks and LMW organics were initially well and irreversibly removed, and their efuent concentrations increased gradually in the outlet of the adsorbers until a pseudosteady state concentration was reached. Poor removal of biopolymers was likely a result of their comparatively large size that prevented access to the internal pore structure of the GACs. In both GAC adsorbers, adsorbability of the remaining NOM fractions, compared on the basis of partition coefcients, increased with decreasing molecular size, suggesting that increasingly larger portions of the internal GAC surface area could be accessed as the size of NOM decreased. Overall DOC uptake at pseudo-steady state differed between the two tested GACs (18.9 and 28.6 g-C/kg GAC), and the percent difference in DOC uptake closely matched the percent difference in the volume of pores with widths in the 1e50 nm range that was measured for the two fresh GACs. Despite the differences in NOM uptake capacity, individual NOM fractions were removed in similar proportions by the two GACs. 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Natural organic matter (NOM) is an important water constituent that affects the performance of drinking water treatment processes and the quality of drinking water. For example, NOM exerts coagulant and oxidant/disinfectant demands,

contributes to membrane fouling, and serves as precursor material for disinfection byproducts (DBPs). Apart from the NOM concentration, which is typically measured by bulk parameters such as total organic carbon (TOC) or UV absorbance at 254 nm, the character of the NOM affects such factors as reactivity with disinfectants, membrane fouling, and biological

* Corresponding author. Tel.: 41 44 823 5067; fax: 41 44 823 5547. E-mail address: Silvana.Velten@eawag.ch (S. Velten). 0043-1354/$ e see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.047

3952

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9

stability of water in distribution systems. As a result, it is important for utilities to know both the concentration and the character of the NOM in their raw and nished waters as well as in the efuents of individual treatment stages. As utilities across the world need to increasingly rely on less pristine sources for the production of drinking water, the use of granular activated carbon (GAC) is becoming more widespread. Apart from trace contaminant removal, the need for enhanced NOM removal can be a driver for the installation of GAC adsorbers. Because aquatic NOM varies in character both spatially across different water sources and temporally within a water source (Owen et al., 1993), the rate and extent of NOM adsorption in GAC lters are difcult to predict (Matsui et al., 2002). Aspects of NOM character such as molecular weight distribution (MWD), degree of hydrophobicity, charge distribution, and ability to form hydrogen bonds with the GAC surface affect NOM adsorption. In addition, physical (surface area; size, shape and volume of pores) and chemical (charge, type, and number of surface groups; ash content) GAC characteristics inuence NOM adsorption (Summers and Roberts, 1988). Furthermore, solution pH, ionic strength, and hardness inuence the adsorption of NOM (Newcombe, 1999). Physical effects are largely governed by the MWD of NOM relative to the pore size distribution (PSD) of GAC (Newcombe et al., 1997; Karanl and Kilduff, 1999). NOM adsorption primarily takes place in mesopores (2e50 nm width) and large micropores (1e2 nm width) (Lee et al., 1981; Summers and Roberts, 1988; Newcombe et al., 1997; Pelekani and Snoeyink, 1999; Ebie et al., 2001; Li et al., 2003; Cheng et al., 2005) and is negligible on AC that primarily contains small micropores (<1 nm in width) (Pelekani and Snoeyink, 1999; Cheng et al., 2005). The high molecular weight (HMW) fraction (>w10,000 Da) of NOM does not readily adsorb because of size exclusion effects; however, this NOM fraction is generally well removed by coagulation and is frequently present only at low concentrations in GAC adsorber inuents (Vuorio et al., 1998; Matilainen et al., 2006). In contrast, intermediate molecular weight fractions (w500e4000 Da) are well removed by GAC. Although low molecular weight (LMW) NOM constituents have access to a large percentage of the GAC pore volume and thus could be well removed based on size considerations alone, LMW compounds may also be relatively hydrophilic and, hence, less adsorbable. Also, GAC adsorber efuents may contain LMW compounds that represent metabolic products of microorganisms living on GAC surfaces (Matilainen et al., 2006). Chemical interactions are inuenced by characteristics of the adsorbate, the adsorbent surface, and the water matrix. Because of the presence of carboxylic acid and phenolic groups in the NOM structure, NOM carries a negative charge at pH values typically encountered in drinking water treatment (Perdue and Lytle, 1983). Depending on the solution pH, the GAC base material and activation process, the net pore surface charge of GAC can be positive, negative, or neutral (Karanl and Kilduff, 1999). Consequently, electrostatic interactions can affect NOM adsorption. Overall, however, the adsorbent pore size distribution appears to be the principal factor controlling NOM uptake by GAC while electrostatic effects play a secondary role (Newcombe and Drikas, 1997; Newcombe et al., 2002).

To date, most information about NOM adsorption by activated carbon was collected in batch tests. Little information is available on the dynamic breakthrough behavior of individual NOM fractions along the depth of GAC adsorbers treating natural water. Prior studies employing SEC to study NOM removal in GAC adsorbers were limited to UV detectors (Vuorio et al., 1998; Nissinen et al., 2001; Matilainen et al., 2006; Li et al., 2007), and none of these studies developed breakthrough curves for individual NOM fractions. Therefore the objectives of this study were to (1) determine the breakthrough behavior of dissolved organic carbon (DOC) and individual NOM components such as biopolymers, humic substances, building blocks and LMW organics in pilot-scale GAC adsorbers and (2) compare the adsorbability of individual NOM fractions on two GACs with different pore size distributions and grain sizes.

2.
2.1.

Materials and methods

Experimental set up and water characteristics

Two pilot-scale GAC adsorbers, GAC 1 and GAC 2, were set up at the Zurich Water Works (WVZ-Lengg, Zurich, Switzerland) treating lake Zurich water. GAC 1 was operated with ozonated water from the intermediate ozonation step of the full-scale water treatment plant (treatment scheme: pre-ozonation, rapid sand ltration, intermediate ozonation, GAC ltration and slow sand ltration; Hammes et al., 2010). GAC 2 was part of a pilot plant (treatment scheme: ozonation, GAC ltration, ultraltration; Hammes et al., 2008). Both reactors were operated in downow mode. Operational and inuent water quality parameters for the two GAC adsorbers are given in Table 1. Given the short empty bed contact time (EBCT) of GAC 1 (1.65 min), only inuent and efuent samples were collected for NOM characterization. For GAC 2, aqueous samples were collected at the inuent, efuent, and at different bed depths

Table 1 e Operational and inuent water quality parameters for GAC 1 and GAC 2 adsorbers. Parameters
Carbon type GAC lter bed (m3) Filtration velocity (m/h) Backwash rate (m/h) Backwash time (min/d) Empty bed contact time (min) GAC depth (m) Column diameter (m) GAC particle diameter (mm) Bed porosity () Packed bed density (kg/m3) Inuent DOC (mg/L) Inuent UV254 (1/m) Inuent SUVA L/(mg*m) Inuent pH ()

GAC 1
Chemviron F400 0.015 8 22.9 10 1.65 0.22 0.3 0.125e0.71 0.47 425 0.96 (0.03) 1.07 (0.14) 1.12 (0.12) 8.1 (0.3)

GAC 2
Chemviron SGL 8 18 1.47 5.9 n/a n/a 15.76 1.55 1.1 1.002.38 0.39 460 1.1 (0.04) 1.84 (0.18) 1.58 (0.18) 7.79 (0.14)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9

3953

that corresponded to EBCTs of about 1, 4, 8, 12, and 16 min. GAC samples from GAC 1 were taken from the top of the lter. For GAC 2 sampling, a metal tube (inner and outer diameter 0.9 and 1.1 cm, respectively) was inserted 0.8 m into the GAC lter through sampling ports located on the side walls, and about 20 g GAC particles were collected from each sampling point. To not disturb the mass transfer zone in GAC 2, the adsorber was not backwashed during the course of this study.

Table 2 e Physical GAC characteristics of two GACs that were used in the GAC lters to assess NOM adsorption. BET surface area and pore volume distributions were obtained from N2 adsorption isotherm data. Parameters
Carbon Type BET Surface Area (m2/g) Primary Micropore Volumea 0e1 nm (cm3/g) Secondary Micropore Volumea 1e2 nm (cm3/g) DFT Mesopore Volumeb (cm3/g) BJH Mesopore Voumec (cm3/g)

GAC 1
Chemviron F400 1060 0.175 0.207 0.203 0.232

GAC 2
Chemviron SGL 8 18 790 0.165 0.127 0.124 0.146

2.2.

NOM characterization

Glass vials for DOC sampling were acid washed and baked at 450  C for 4 h. Upon in-line ltration (0.45 mm polyethersulfone lter) and acidication, NOM was characterized by size exclusion chromatography (SEC) and quantied by both ultraviolet (UV) and organic carbon (OC) detectors (SECOCD, DOC Labor Dr. Huber, Karlsruhe, Germany). Details about the SEC-OCD instrument, analytical method, and data interpretation can be found in Huber et al. (2011). The SEC column (Toyopearl TSK HW-50S, 250 20 mm) had a fractionation range of 100e10,000 Da. The UV detector (Knauer K200) provided an online UV signal and OC was detected by an infrared (IR) detector after oxidation of NOM to CO2 in a Graentzel Thin-Film Reactor. Chromatograms were interpreted with customized software (Chrom CALC, Huber et al., 2011). A phosphate buffer was used as the eluent (24 mM, pH 6.6) and the ow rate was set at 1 mL/min. Using an injection volume of 1 mL, the detection limit for NOM fractions eluting from the SEC column was 10 mg C/L.

a Micropore volume calculated by density functional theory (DFT) for pores with widths less than 2 nm. b Mesopore volume calculated by density functional theory (DFT) for pores with widths ranging from 2 to 36 nm (upper limit for DFT model). c Mesopore volume calculated by Barrett, Joyner, and Halenda (BJH) method for pores with widths ranging from 2 to 50 nm.

2.5.

Adsorbent characterization

2.3.

NOM uptake and partition coefcients

NOM uptake by GAC was calculated as follows: q

n Q X $ Dti $ 0:5$ cin;i1 cin;i 0:5$ cout;i1 cout;i mGAC i1

q ti cin,i1 cin,i cout,i1 cout,i Q mGAC

solid-phase NOM concentration (g/kg GAC) time interval between two sampling dates (d) inuent concentration at time i-1 (g/L) inuent concentration at time i (g/L) efuent concentration at time i-1 (g/L) efuent concentration at time i (g/L) ow rate (L/d) mass of GAC in column (kg)

Brunauer, Emmett and Teller (BET) surface areas, micropore and mesopore volumes, and pore size distributions (PSDs) were determined from N2 adsorption isotherm data collected at 77 K (Autosorb-1-MP, Quantachrome, Boynton Beach, FL, USA). Prior to analysis, adsorbent samples were outgassed for 24 h at 423 K. BET surface areas were determined from 18-point adsorption isotherms that were completed with a 0.1-g sample in the 0.01e0.3 relative pressure range. PSDs were determined by conducting N2 adsorption experiments in two separate phases: high and low pressure phase. Adsorption isotherms from both pressure phases were combined to produce one isotherm. Micropore volume and PSD were computed from N2 adsorption isotherm data using the Density Functional Theory (DFT) with the N2_carb1.gai kernel (PC software version 1.51, Quantachrome, Boynton Beach, FL, USA). In addition, the mesopore volume was computed using the Barrett, Joyner, and Halenda (BJH) method that captures the entire mesopore range (2e50 nm). Additional details of the experimental procedure are described in Knappe et al. (2007).

3.
3.1.

Results and discussion

GAC 1

The partition coefcient Kp of individual NOM components was calculated at pseudo-steady state conditions from the ratio of solid and aqueous-phase concentrations, q and c: KP q=c

2.4.

Activated carbons

Two bituminous-coal based GACs (Chemviron SGL 8 18 and Chemviron F400, Chemviron, Feluy, Belgium) were used. GAC characteristics are summarized in Table 2.

NOM adsorption by GAC 1 was studied for 112 days or 98,000 empty bed volumes (EBV). Fig. 1 shows DOC and UV absorbance data collected at the inlet and outlet of GAC 1. Following rapid initial breakthrough, efuent DOC and UV absorbance values reached a pseudo-steady state after about 40 days of operation or 36,000 EBV. At that time, the UV absorbance values measured in the adsorber inlet and outlet were almost identical whereas the DOC data differed by about 0.1 mg-C/L (Fig. 1). This difference may be explained by biodegradation of a DOC fraction with low UV absorbance. Bacteria attached to the GAC surface can metabolize the readily bioavailable

3954

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9

Fig. 1 e Inuent and efuent concentrations of DOC, UV absorbance, biopolymers, humics, building blocks and LMW organics measured for GAC 1.

fraction of DOC, which typically consists of LMW organic compounds with low UV absorbance (Herzberg et al., 2003; Huber et al., 2011). For GAC 1, 50% DOC breakthrough was reached after w2400 EBV. When GAC is used to remove NOM with the goal to lower DBP formation during chlorination, full-scale experience suggests that 50% total organic carbon (TOC) breakthrough is reached after w6400 and w7900 EBV for empty bed contact times (EBCT) of 10 and 20 min, respectively (Bond and DiGiano, 2004). For the purposes of this discussion, it was assumed that DOC was the principal contributor to TOC. The earlier occurrence of 50% DOC breakthrough obtained here resulted from the short EBCT of GAC 1 (1.65 min) relative to those used in practice. Regarding the onset of the pseudosteady state region, results obtained with GAC 1 were similar to those of a prior study (Speitel et al., 1989), during which a pseudo-steady state TOC concentration was reached

after 35,000e40,000 empty bed volumes when evaluating TOC removal from ozonated water in a GAC adsorber with an EBCT of 1.2 min. In Fig. 2, size exclusion chromatograms obtained with OC (panel A) and UV detection (panel B) are compared for inuent and efuent samples obtained from GAC 1 during the rst 50 days of operation. The size exclusion chromatogram of the inuent that was obtained with OC detection exhibits a peak for biopolymers (28 min retention time) followed by a large humic substances peak (42 min retention time) that constituted w50% of the total DOC. Building blocks (45 min retention time) are degradation products of humic substances. LMW organics (51 min retention time) are composed of LMW acids and LMW humics that elute simultaneously. LMW neutrals (>55 min retention time) are neutral or amphiphilic molecules that interact with the column and consequently have a longer retention time than the elution time of water (55 min).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9

3955

Fig. 2 e SEC chromatograms obtained with OC detection (A) and UV detection (B) for inuent and efuent samples of GAC 1.

Compared to the size exclusion chromatogram obtained with OC detection, that obtained with UV detection differs in four aspects: (1) the biopolymer peak is absent, (2) the relative signal strength for building blocks is lower, (3) the relative signal strength for LMW organics is higher, and (4) an inorganic colloids peak (e.g. clay particles or silicates) is present which is not seen in the OC-response. Fig. 2 qualitatively shows the removal of individual NOM fractions in a GAC adsorber. Biopolymers were not retained by the GAC lter as inuent and efuent concentrations were practically identical after one day of operation. Concentrations of humics, building blocks and LMW organics increased gradually in the efuent during the rst 50 days of operation. The three dominant NOM fractions (humics, building blocks, and LMW organics) were initially all well removed in the GAC adsorber, and this observation was consistent between chromatograms obtained with OC and UV detection. After 50 days of operation, the removal of humics became negligible while a small percentage of the two smaller NOM fractions (building blocks and LMW organics) continued to be removed across the GAC adsorber (Fig. 2). While biodegradation likely contributed to

the removal of LMW compounds, the appearance of soluble microbial products or extrapolymeric substances was not detectable by SEC-OCD analysis. During the entire study period, concentrations of all NOM fractions decreased along the length of the lter bed. This observation suggests that displacement of one adsorbed NOM fraction by another did not occur and that humics, building blocks, and LMW compounds adsorbed irreversibly. To permit a more quantitative assessment about the adsorbability of individual NOM fractions, breakthrough curves for biopolymers, humics, building blocks and LMW organics are depicted in Fig. 1. Panel C conrms that biopolymer removal was negligible by GAC 1 (Fig. 1). Biopolymers such as protein- and polysaccharide-like substances are large molecules with MW >10,000 Da. The number of adsorption sites for such large molecules on AC is limited because of size exclusion effects. Humics were initially well retained by GAC 1 but the AC rapidly lost its adsorption effectiveness (Fig. 1D). Fifty percent humics removal was reached after w4500 EBV and the inlet concentration was reached after w30,000 EBV. Compared to biopolymers, humics are smaller

Table 3 e Solid-phase concentrations and partition coefcients (Kp) of DOC and individual NOM fractions calculated for GAC 1 and GAC 2 at pseudo-steady state. Pseudo-steady state was reached after 36,000 EBV or 40 days in GAC 1 and at 120 cm after 20,000 EBV or 168 days in GAC 2. Fractions GAC 1 Solid-phase concentration (g-C/kg GAC)
DOC Biopolymers Humics Building Blocks LMW organics 28.6 1.1 11.3 7.1 6.8

GAC 2 Kp (L/g GAC)

34.4 14.7 24.2 49.7 53.5

Solid-phase concentration (g-C/kg GAC)

18.9 0.0 7.8 3.8 1.6

Kp (L/g GAC)
22.8 0.0 16.7 26.8 41.0

3956

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9

that the adsorbability of NOM fractions increased with decreasing size; i.e. biopolymers < humics < building blocks < LMW organics. The greater adsorbability of smaller NOM constituents is likely the result of their ability to access a larger percentage of the total AC pore volume.

3.2.

GAC 2

Fig. 3 e DOC breakthrough curves obtained with GAC 2 at different empty bed contact times and with GAC 1: efuent DOC concentration normalized with respect to inuent DOC concentration versus empty bed volumes treated.

molecules with an average molecular weight of w1000 Da and a diameter of <2 nm; they can therefore enter the mesopores and larger micropores of activated carbon (Huber and Frimmel, 1996; Wilkinson et al., 1999). Similarly the building blocks (w300e500 Da) were well retained and 50% building blocks removal was reached after w9000 EBV. Complete breakthrough of building blocks was not reached during the operation time of GAC 1 (Fig. 1E), and about 20% of the inuent building blocks were removed in the GAC lter at pseudo-steady state. LMW organics (<350 Da) followed a similar pattern as building blocks with 50% LMW organics removal after w17,000 EBV (Fig. 1F) and approximately 25% removal at pseudo-steady state. The decrease of building blocks and LMW organics is most probably due to biodegradation in the biolm that developed in GAC 1. A concurrent study focusing on the development of the microbiology on GAC 1 showed that the number of microorganisms reached a maximum after 33 days of operation (30,000 EBV) (Velten et al., 2007). Table 3 summarizes adsorption capacities and partition coefcients for DOC and individual NOM fractions after pseudosteady state was attained in GAC 1 and GAC 2. The solid-phase concentrations were determined by integrating the difference between the inuent and the efuent DOC concentration over time. Overall DOC uptake in GAC 1 was 28.6 g-C/kg GAC. Humics exhibited the highest solid-phase concentration on the GAC (11.3 g-C/kg GAC) followed by building blocks (7.1 g-C/kg GAC) and LMW organics (6.8 g-C/kg GAC) and nally biopolymers (1.1 g-C/kg GAC). The high solid-phase concentration of humics can in part be explained by its dominance in the inuent NOM composition (50%). To compare the adsorbability of DOC and individual NOM fractions, partition coefcients (Kp) were calculated from the pseudo-steady state solid-phase concentration and the corresponding aqueous-phase concentration in the GAC 1 efuent (Kp q/C ). The Kp values in Table 3 suggest

NOM removal by GAC 2 was studied as a function of bed depth (or EBCT) over a 4-month period. The evolution of DOC concentrations as a function of EBV and GAC bed depth is depicted in Fig. 3. Fifty percent DOC breakthrough in GAC 2 was reached immediately after start up of the reactor in the top 2 layers (10 and 40 cm) and was reached after 4300 EBV at 80 cm (EBCT 8.1 min); 5300 EBV at 120 cm (EBCT 12.2 min); and 6500 EBV at 155 cm (EBCT 15.8 min). EBV values to 50% DOC breakthrough that were obtained with GAC 2 are typical when compared to U.S. data that were gathered as part of the information collection rule (6449 3935 EBV at an EBCT of 10 min; Bond and DiGiano, 2004). Plotting normalized efuent DOC concentrations as a function of treated EBV, a comparison of the breakthrough curves at different lter depths (Fig. 4) shows improved DOC removal with increasing EBCT up to an EBCT of 12.2 min; an increase in EBCT from 12.2 to 15.8 min did not offer an advantage in terms of decreasing carbon usage rate, however. Breakthrough curves for biopolymers, humics, building blocks and LMW organics obtained with GAC 2 at different bed depths are shown in Fig. 4. In addition, results obtained with GAC 1 are included for comparison. As was the case with GAC 1, biopolymers were only slightly retained in GAC 2. Humics and building blocks were well retained, and concentrations decreased with increasing lter depth. At an EBCT of 15.8 min, 50% breakthrough of humics, building blocks, and LMW organics was reached after 5100 EBV (56 days), 6500 EBV (71 days), and 10,300 EBV (112 days), respectively. Adsorption capacities and partition coefcients of DOC and individual NOM fractions for GAC 2 are summarized in Table 3. The overall uptake of DOC in GAC 2 was 18.9 g-C/kg GAC. As was the case with GAC 1, humics showed the highest solid-phase concentration (7.8 g-C/kg GAC), followed by building blocks (3.8 g-C/kg GAC), LMW organics (1.6 g-C/kg GAC) and biopolymers (negligible adsorption). Partition coefcients for individual NOM fractions were calculated once the efuent DOC concentration reached a pseudo-steady state (w168 days at an EBCT of 12.21 min). The order of NOM fraction adsorbability matched that obtained with GAC 1 (i.e., biopolymers < humics < building blocks < LMW organics). These calculations were made by assuming that NOM fractions were only removed by adsorption even though biological degradation proceeded simultaneously in the GAC lter. Considering the population of bacteria in the GAC lter and a biological degradation capacity of 7.9 107 mg C/bacterium (calculated for this lter as part of another study, see Velten et al., 2007), true solid-phase concentrations could be 20e50% lower.

3.3.

Comparison of GACs

In this study two adsorbers containing GACs with different physicochemical properties were tested. The BET surface area

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9

3957

Fig. 4 e Breakthrough curves of biopolymers, humics, building blocks and LMW organics obtained with GAC 2 at different empty bed contact times and with GAC 1.

of the AC used in GAC 2 exhibited a BET surface area that was 75% of that used in GAC 1; furthermore, the AC used in GAC 2 had secondary micropore (1e2 nm pore width range) and mesopore (2e50 nm pore width range) volumes that were approximately 60% of those measured for the AC used in GAC 1 (Table 2). In contrast, primary micropore volumes (<1 nm width) were similar for the two ACs. Also, a smaller grain size was used in GAC 1 (US mesh 25 120) compared to GAC 2 (US mesh 8 18). Results of prior studies suggest that NOM adsorbs primarily in secondary micropores and mesopores. Therefore, DOC uptake was expected to be greater by GAC 1. The solid-phase concentrations and partition coefcients shown in Table 3 indeed conrm that GAC 1 was more effective for NOM removal than GAC 2. At pseudo-steady state, DOC uptake by GAC 2 was approximately 66% of that obtained by GAC 1, a result that is in reasonable agreement with the difference in mesopore and secondary micropore volumes between the two tested ACs. This result illustrates that extent of NOM uptake in GAC adsorbers is largely controlled by pores

with widths in the 1e50 nm range and conrms prior results obtained in batch tests. To assess whether GACs with different physical properties fractionate NOM differently, size exclusion chromatograms for the inuent and efuent of GAC 1 and GAC 2 were compared at points of operation, at which similar overall DOC removal was measured [for GAC 1 after 8 days or 7230 EBV and for GAC 2 after 42 days or 7434 EBV of operation (EBCT 8.14 min)]. As illustrated in Fig. 5, relative peak heights of the inuents to the two GAC adsorbers were similar despite the fact that there were small differences in the bulk inuent water quality parameters (Table 1). Furthermore, the data in Fig. 5 show that all NOM fractions adsorbed similarly in the two GAC adsorbers even though the physical characteristics of the two GACs differed (Table 2). Consequently, the efuent NOM composition at a given DOC removal percentage did not appear to be affected by differences in GAC pore structure. Finally, it should be noted that the smaller grain size of the AC used in GAC 1 aided the rate of NOM removal (Fig. 3).

3958

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9

Acknowledgments
We acknowledge nancial assistance from the Eawag Wave21 project, the Zurich Water Works (WVZ) and thank Heinz Bischofberger for technical and scientic assistance.

references

Fig. 5 e SEC-OCD chromatogram for inuent and efuent concentrations of GAC 1 and GAC 2 after treatment of 7230 and 7434 EBV, respectively.

As illustrated in Fig. 3, the initial performance of GAC 1 (EBCT 1.65 min) was similar to that of the GAC 2 layer with an EBCT of 8.14 min.

4.

Conclusions

To achieve a high NOM removal efciency with GAC adsorption, GAC with a high surface area and a large volume of pores with widths of 1e50 nm should be selected by utilities. The results of this study show that biopolymers do not adsorb on GAC while the adsorbability of the remaining NOM fractions increased with decreasing molecular size (i.e., humics < building blocks < LMW organics). All NOM fractions adsorbed similarly on the two GACs; i.e., the NOM composition in the efuents of GAC1 and GAC2 were similar even though the GACs exhibited different pore size distributions. Moreover, no displacement of individual NOM fractions occurred. The adsorption of individual NOM fractions can therefore be considered as irreversible. Biopolymers and humics play an important role in membrane fouling (Howe and Clark, 2002; Costa et al., 2006; Jermann et al., 2007). This study showed that biopolymers were not removed by GAC. Humics adsorbed on fresh GAC, but the adsorption capacity decreased rapidly. GAC ltration is therefore not expected to be an effective pretreatment technology for the removal of these membrane foulants. Regarding DBP precursors, prior studies showed that DBP yields (mmol DBP/mmol C) from higher MW humic and lower MW non-humic NOM fractions are similar (Hwang et al., 2001; Imai et al., 2003). The results of this research suggest that GAC adsorption is more effective for DBP precursor control in waters containing a larger percentage of LMW NOM.

Bond, R.G., DiGiano, F.A., 2004. Evaluating GAC performance using the ICR database. Journal American Water Works Association 96, 96e104. Cheng, W., Dastgheib, S.A., Karanl, T., 2005. Adsorption of dissolved natural organic matter by modied activated carbons. Water Research 39, 2281e2290. Costa, A.R., de Pinho, M.N., Elimelech, M., 2006. Mechanisms of colloidal natural organic matter fouling in ultraltration. Journal of Membrane Science 281, 716e725. Ebie, K., Li, F.S., Azuma, Y., Yuasa, A., Hagishita, T., 2001. Pore distribution effect of activated carbon in adsorbing organic micropollutants from natural water. Water Research 35, 167e179. Hammes, F., Berney, M., Wang, Y., Vital, M., Koster, O., Egli, T., 2008. Flow-cytometric total bacterial cell counts as a descriptive microbial parameter for drinking water treatment processes. Water Research 42, 269e277. Hammes, F., Berger, C., Koster, O., Egli, T., 2010. Assessing biological stability of drinking water without disinfectant residuals in a full-scale water supply system. Journal of Water Supply: Research and Technology e AQUA 59 (1). Herzberg, M., Dosoretz, C.G., Tarre, S., Green, M., 2003. Patchy biolm coverage can explain the potential advantage of BGAC reactors. Environmental Science and Technology 37, 4274e4280. Howe, K.J., Clark, M.M., 2002. Fouling of microltration and ultraltration membranes by natural waters. Environmental Science and Technology 36, 3571e3576. Huber, S.A., Frimmel, F.H., 1996. Size-exclusion chromatography with organic carbon detection (LC-OCD): a fast and reliable method for the characterization of hydrophilic organic matter in natural waters. Vom Wasser 86, 277e290. Huber, S.A., Balz, A., Albert, M., Pronk, W., 2011. Characterisation of aquatic humic and non-humic matter with size-exclusion chromatography e organic carbon detection e organic nitrogen detection (LC-OCD-OND). Water Research 45, 879e885. Hwang, C.J., Krasner, S.W., Sclimenti, M.J., Amy, G.L., Dickenson, E.,   Bruchet, A., Prompsy, C., Gisecle, F., Crouec, J.-P., Violleau, D., Leenheer, J., 2001. Polar NOM: Characterization, DBPs and Treatment. American Water Works Association Research Foundation, Denver, CO, USA. Imai, A., Matsushige, K., Nagai, T., 2003. Trihalomethane formation potential of dissolved organic matter in a shallow eutrophic lake. Water Research 37, 4284e4294. Jermann, D., Pronk, W., Meylan, S., Boller, M., 2007. Interplay of different NOM fouling mechanisms during ultraltration for drinking water production. Water Research 41, 1713e1722. Karanl, T., Kilduff, J.E., 1999. Role of granular activated carbon surface chemistry on the adsorption of organic compounds. 1. Priority pollutants. Environmental Science & Technology 33, 3217e3224. Knappe, D.R.U., Rossner, A., Snyder, S.A., Strickland, C., 2007. Alternative Adsorbents for the Removal of Polar Organic

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9

3959

Contaminants. American Water Works Association Research Foundation, Denver, CO, USA. Lee, M.C., Snoeyink, V.L., Crittenden, J.C., 1981. Activated carbon adsorption of humic substances. Journal American Water Works Association 73, 440e446. Li, Q., Snoeyink, V.L., Marinas, B.J., Campos, C., 2003. Pore blockage effect of OM on atrazine adsorption kinetics of PAC: the roles of PAC pore size distribution and NOM molecular weight. Water Research 37, 4863e4872. Li, F., Yuasa, A., Katamine, Y., Tanaka, H., 2007. Breakthrough of natural organic matter from xed bed adsorbers: investigations based on size-exclusion HPLC. Adsorption 13, 569e577. Matilainen, A., Vieno, M., Tuhkanen, T., 2006. Efciency of the activated carbon ltration in the natural organic matter removal. Environment International 32, 324e331. Matsui, Y., Knappe, D.R.U., Iwaki, K., Ohira, H., 2002. Pesticide adsorption by granular activated carbon adsorbers. 2. Effects of pesticide and natural organic matter characteristics on pesticide breakthrough curves. Environmental Science & Technology 36, 3432e3438. Newcombe, G., Drikas, M., Hayes, R., 1997. Inuence of characterised natural organic material on activated carbon adsorption: II. Effect on pore volume distribution and adsorption of 2-methylisoborneol. Water Research 31, 1065e1073. Newcombe, G., Drikas, M., 1997. Adsorption of NOM onto activated carbon: electrostatic and non-electrostatic effects. Carbon 35, 1239e1250. Newcombe, G., 1999. Charge vs. porosity e some inuences on the adsorption of natural organic matter (NOM) by activated carbon. Water Science and Technology 40, 191e198. Newcombe, G., Morrison, J., Hepplewhite, C., 2002. Simultaneous adsorption of MIB and NOM onto activated carbon. I.

characterisation of the system and NOM adsorption. Carbon 40, 2135e2146. Nissinen, T.K., Miettinen, I.T., Martikainen, P.J., Vartiainen, T., 2001. Molecular size distribution of natural organic matter in raw and drinking waters. Chemosphere 45, 865e873. Owen, D.M., Amy, G.L., Chowdhury, Z.K., 1993. Characterization of Natural Organic Matter and Its Relationship to Treatability. American Water Works Association Research Foundation, Denver, CO, USA. Pelekani, C., Snoeyink, V.L., 1999. Competitive adsorption in natural water: role of activated carbon pore size. Water Research 33, 1209e1219. Perdue, E.M., Lytle, C.R., 1983. Distribution model for binding of protons and metal-ions by humic substances. Environmental Science & Technology 17, 654e660. Speitel, G.E., Turakhia, M.H., Lu, C.J., 1989. Initiation of micropollutant biodegradation in virgin gac columns. Journal American Water Works Association 81, 168e176. Summers, R.S., Roberts, P.V., 1988. Activated carbon adsorption of humic substances .2. Size exclusion and electrostatic interactions. Journal of Colloid and Interface Science 122, 382e397. Velten, S., Hammes, F., Boller, M., Egli, T., 2007. Rapid and direct estimation of active biomass on granular activated carbon through adenosine tri-phosphate (ATP) determination. Water Research 41, 1973e1983. Vuorio, E., Vahala, R., Rintala, J., Laukkanen, R., 1998. The evaluation of drinking water treatment performed with HPSEC. Environment International 24, 617e623. Wilkinson, K.J., Balnois, E., Leppard, G.G., Bufe, J., 1999. Characteristic features of the major components of freshwater colloidal organic matter revealed by transmission electron and atomic force microscopy. Colloids and Surfaces A-Physicochemical and Engineering Aspects 155, 287e310.