Prediction of the adsorption capacities for four typical organic pollutants on activated carbons in natural waters

Water Research 111 (2017) 28e40

Contents lists available at ScienceDirect

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

Prediction of the adsorption capacities for four typical organic pollutants on activated carbons in natural waters Warisa Bunmahotama a, Wei-Nung Hung b, c, Tsair-Fuh Lin a, b, * a

Department of Environmental Engineering, National Cheng Kung University, Tainan, 70101, Taiwan Global Water Quality Research Center, National Cheng Kung University, Tainan, 70955, Taiwan c Green Energy and Environment Research Laboratories, Industrial Technology Research Institute, Hsinchu, 30011, Taiwan b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 4 September 2016 Received in revised form 18 December 2016 Accepted 19 December 2016 Available online 21 December 2016

A new model is developed to predict the competitive adsorption isotherms of atrazine, methyl tertiary butyl ether (MTBE), 2-methylisoborneol (2-MIB) and 2,4,6-trichlorophenol onto activated carbons (ACs) in natural water. Based on the Polanyi-Dubinin (PD) equation, with the limiting pore volume of adsorbent estimated from the pore size distribution data, and the Ideal adsorbed solution theory - equivalent background compound (IAST-EBC) model approximation, the model takes into account both the properties of ACs and the impact of natural organic matters in water. Only one set of isotherm in deionized water and one set in natural water are needed to obtain the parameters for the prediction of adsorption isotherms onto different ACs in natural water. The model was employed for the predictions of adsorption capacities for atrazine, MTBE, 2-MIB and 2,4,6-trichlorophenol onto 14 ACs in 22 synthetic and natural waters reported in 9 references, with errors between 14.9% and 44.5% SDEV only. The results suggest that in the proposed PD-IAST-EBC approach, prediction of adsorption capacity for organic compounds onto different ACs in the same natural water is feasible, if the ACs are thermally activated with known pore size information. The model may provide a simple approach for the prediction of adsorption of organic compounds in natural water, and thus greatly reduces the effort required for water utilities when change of AC is needed. © 2016 Elsevier Ltd. All rights reserved.

Keywords: Adsorption isotherm Polanyi-Dubinin model IAST model Micropore volume Atrazine 2-MIB

1. Introduction Powdered and granular activated carbons (ACs) are commonly used in drinking water treatment for the removal of organic micropollutants. It is well known that adsorption capacity of ACs is strongly dependent on the properties of ACs, targeted chemicals, and water matrix. The properties of ACs, including specific surface areas (Tan et al., 2008), pore volume and pore size distribution (PSD) (Pelekani and Snoeyink, 1999, 2000), and surface-chemical nature (Li et al., 2002a) have been observed to affect the adsorption of organic chemicals onto ACs. Chemical properties, including functional groups (Crittenden et al., 1999), hydrogen bonding capability (De Ridder et al., 2010), molecular sizes and/or molecular structures (Bunmahotama et al., 2015) are also the factors that

* Corresponding author. Department of Environmental Engineering, National Cheng Kung University, Tainan, 70101, Taiwan. E-mail address: tfl[email protected] (T.-F. Lin). http://dx.doi.org/10.1016/j.watres.2016.12.033 0043-1354/© 2016 Elsevier Ltd. All rights reserved.

affect adsorption. The properties of water matrix, such as solution pH, ionic strength, and temperature may also influence the adsorption (Al-Degs et al., 2008). However, for the adsorption of organic micro-pollutants in natural water systems, natural organic matter (NOM) has been considered to be the key factor to affect the adsorption (Hyung and Kim, 2008; Li et al., 2003b; Newcombe et al., 2002b). NOM is present in natural waters in a wide range of concentrations and is a collection of organic compounds of variable sizes, €€ molecular weights, functionalities and adsorbabilities (Sillanpa a, 2014). The existence of NOM can adversely impact the adsorption capacity and adsorption kinetics of micro-pollutants onto adsorbents (Smith and Weber Jr, 1989; Shih et al., 2003). Therefore, the adsorption capacity determined in deionized water system cannot represent that in natural water, since the impact of NOM needs to be considered. As the characteristics of NOM is source water dependent, the impact on the adsorption depends on the water €, 2014). Therefore, a model capable of quansource also (Sillanp€ aa titatively predicting the impact of NOM on adsorption capacity

W. Bunmahotama et al. / Water Research 111 (2017) 28e40

simply based on certain kind of NOM properties, such as molecular sizes and polarity (Hyung and Kim, 2008), would be very helpful for a better design of adsorption processes. Attempts have been made to model the impact of NOM on the adsorption of organic compounds in natural water (Hung, 2005). Among the models developed, a combination of equivalent background compound (EBC) with ideal adsorbed solution theory (IAST) has been widely used (Graham et al., 2000; Knappe et al., 1998; Najm et al., 1991; Newcombe et al., 2002b; Yu et al., 2008). In the IAST-EBC model, NOM is considered as a single solute (the EBC), while the targeted compound is considered as the other solute. The adsorption of the two solutes is then modeled together by the IAST. This IAST-EBC model has been employed successfully to simulate a variety of chemicals, including naproxen, carbamazepine, nonylphenol, atrazine, and 2-methylisoborneol (2-MIB) onto different ACs in natural waters collected in US, Australia, Canada and Taiwan (Hung and Lin, 2006a; b; Hung et al., 2005; Knappe et al., 1998; Newcombe et al., 2002b; Yu et al., 2008; Yu et al., 2016). IAST, developed by Myers and Prausnitz (1965) in 1965, is an extensively used (Walton and Sholl, 2015) thermodynamic basis for easily forecasting multi-component adsorption isotherms from only the pure-component adsorption isotherms at the same temperature. IAST rests upon the hypothesis that the adsorbed species form an ideal mixture, that is a sensible approximation in many systems (Babarao et al., 2007; Bae et al., 2008; Cessford et al., 2012; Hand et al., 1985; Keskin et al., 2008; Krishna et al., 2002; Myers and Prausnitz, 1965; Rother and Fieback, 2013; Yang and Zhong, 2006). Although the EBC-IAST model is able to describe the adsorption capacity for a variety of AC/chemical/source water combinations, the model does not take into account the properties of targeted chemicals and ACs and cannot predict the adsorption capacity for different chemicals onto different ACs. Recently, Bunmahotama et al. (2015) successfully developed a model to predict adsorption isotherms of 40 low-molecular- weight nonpolar organic compounds onto 14 ACs in deionized water system. The model, based on the Polanyi-Dubinin (PD) equation, the PSD data of ACs and the molecular structures of organic chemicals, is able to predict the adsorption capacities for organic chemicals onto ACs merely from the PSD data and the chemical properties. It would be very desirable if the model could be applied for natural water systems. In this study, a new approach, a combination of PD with the IAST-EBC model, is proposed to simulate and predict the adsorption capacity for organic compounds onto ACs in natural waters. The model is aimed to predict the adsorption capacity for the same natural water systems with capability of extrapolating to other chemicals and ACs. All the model parameters were either obtained from literatures or chemical database and no advanced experiments were required. The model was validated with extensive adsorption data for atrazine, methyl tertiary butyl ether (MTBE), 2MIB and 2,4,6-trichlorophenol onto 14 different ACs from 9 published reports. This novel approach may provide a simple way for the prediction of adsorption capacities for other ACs and chemicals, and therefore is very useful for water utilities when facing changes in targeted chemicals and ACs.

2. Methods 2.1. Modeling approach 2.1.1. PD model The PD equation (Dubinin, 1960) modified by Crittenden et al. (1999) is used in this study, as shown below,

h  W ¼ Wo exp 

ε n i 100N

29

(1)

where W is the volume of solute adsorbed (mL/g), Wo is the limiting volume of the adsorption space, (mL/g), n is the exponential constant (), N is a normalizing factor, 100 is a scaling factor, ε ¼ RT ln (Cs/C) in aqueous system (cal/mol), C is the aqueous phase concentration (mg/L), Cs is the aqueous solubility (mg/L), R is the gas constant (1.987 cal/mol/K), and T is the absolute temperature (K). To predict the adsorption capacity, W, three parameters, Wo, n, and N need to be determined in advance. Generally, Wo for a specific AC may be obtained from the regression of isotherm data, as shown by Crittenden et al. (1999). Although precise estimation of Wo for a specific GAC/chemical system may be obtained through this method, experiments for isotherm data are required. In this study, the method proposed by Urano et al. (1982), and later successfully used by Hung and Lin (2007) and Bunmahotama et al. (2015) is applied. In the method Wo can be estimated from PSD of ACs, as described in equation (2) Wo ¼ 0.055 mL/g þ V3.2

(2)

where V3.2 is the micropore volume for pores with diameter less than 3.2 nm for AC. For another parameter, n, that represents the property relevant to the surface heterogeneity, was determined by Bunmahotama et al. (2015) to be 1.1 and this value is used in this study. For the third parameter in the model, the normalizing factor N, its estimation is based on molecular connectivity indices (MCIs) as also described in Bunmahotama et al. (2015). 2.1.2. IAST-EBC model The IAST-EBC model, when combined with the Freundlich isotherm (q ¼ KC1/n) to describe the single-solute adsorption for both target organic compound and EBC, can be expressed as Najm et al. (1991):

C1;0  q1 CC 

  q1 n1 q1 þ n2 q2 n1 ¼0 q1 þ q2 n1 K1

(3)

C2;0  q2 CC 

  q2 n1 q1 þ n2 q2 n2 ¼0 q1 þ q2 n2 K2

(4)

in what subscript i (¼ 1 or 2) represents the target compound (i ¼ 1) and EBC (i ¼ 2), Ci,0 is the initial concentration of compound i (mmol/L), Cc is adsorbent dosage (g/L), qi is solid-phase concentration (mmol/g), and Ki and ni are single-solute Freundlich isotherm constants. As shown in equations (3) and (4), besides the adsorption parameters for the target compound (C1,0, K1 and 1/n1), three more EBC parameters (C2,0, K2 and 1/n2) are needed for the model prediction. To acquire the EBC parameters, a non-linear optimization algorithm that could simultaneously solve the IAST equations has been proposed (Najm et al., 1991). In the approach, three more equilibrium points at a specific initial concentration of target compound in natural water are required for defining the EBC parameters, along with single-solute parameters (Najm et al., 1991). In fact, model calibrations using two isotherms in natural water with different initial concentrations were usually used (Greene et al., 1994; Knappe et al., 1998). Najm et al. (1991) proposed that the EBC parameters (C2,0, K2 and 1/n2) are not unique and alike modeling results may be observed by using different sets of parameters. To evaluate the difference between model predictions and experimental data, Crittenden et al. (1999) proposed the percent sample deviation (SDEV), based on the relative error between

W. Bunmahotama et al. / Water Research 111 (2017) 28e40

325 325

% sample deviation ðSDEVÞ ¼

e

Ndata  1

 100

(5)

eb

2.2. Isotherm data and physical-chemical properties of adsorbates To evaluate the model predictions, the adsorption isotherm data in NOM water from 9 reports, Li et al. (2002b), Ding (2010), Hung and Lin (2006b), Newcombe et al. (2002b), Gillogly et al. (1999), Li et al. (2003a), Knappe et al. (1998), Najm et al. (1991) and Chen et al. (1997) were used. The data for these adsorption isotherms are listed in Table 1. The molecular weight (MW), water solubility and chemical density for the chemicals studied were gained from ChemSpider (ChemSpider) and Molinspiration (Molinspiration) online databases. 3. Results and discussion 44

WPH Hydrodarco B

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u Pq  q 2 p e uh i t q

where qp and qe are the solid-phase concentration determined from the prediction and the experimental data for various aqueous-phase concentrations, respectively, and Ndata is the number of data points.

eb

WPH SA UF

Hydrodarco B

WPH

22 eb eb

2-MIB

60 7.3e8 eb

2,4,6-trichlorophenol

experimental data and the correlation, to describe the degree of fitting between the model and experimental data. This was modified in this study and is expressed as

30 eb eb

Atrazine

43 eb eb

Atrazine

Chen et al. (1997) Knappe et al. (1998) Li et al. (2003a)

Najm et al. (1991)

F200 Lignite

30

eb

SA 30 P1100 PCO

eb eb

F400 F600

TAC WPH

60 eb eb

eb

WPH SA UF

Isotherms were given in the reports. Not specified in the reports.

eb eb Mesh size

b

SA UF SA Super W35 W20 WPH SA UF Carbon type

a

2-MIB 2-MIB

185 eb eb 53 7.7e8 eb

2-MIB MTBE

87 7.7e8 eb ea eb eb

Atrazine

68 eb eb

Atrazine

55 eb eb Data points pH Temp ( C)

Hung and Lin (2006b) Ding (2010) Li et al. (2002b)

References Compounds

Table 1 Summary of literatures and adsorption isotherms used in this study.

Newcombe et al. (2002b)

Gillogly et al. (1999)

3.1. Predictions of the adsorption capacity for atrazine Adsorption of atrazine onto ACs was first simulated and predicted for the adsorption capacity. Li et al. (2002b) studied adsorption of atrazine onto two ACs (SA UF and WPH) in deionized water and one natural water, and the data were used for the analysis. Fig. 1a shows the model fits and predictions of the adsorption capacities for atrazine onto SA UF. To predict adsorption isotherms in natural water, isotherm parameters for both targeted compounds and NOMs are needed. In this study, the adsorption data of atrazine in deionized water was first fitted with PD-DA isotherm equation (equation (1)) and further transformed to Freundlich isotherms of atrazine in deionized water. Then one set of the isotherm data in natural water was fitted with the IAST-EBC model (equations (3) and (4)) to obtain the isotherm parameters for NOMs. In the PD-DA equation, three parameters are included: the maximum adsorption capacity (Wo), the normalizing factor (N), and the exponent (n). In the fitting, Wo was estimated to be 0.455 mL/g based on the PSD data from Tang (2007) and the Urano approach (equation (2)), while the n value of 1.1 is used as suggested by Bunmahotama et al. (2015). Therefore, only N was used in fitting the isotherm equation to the data for atrazine in deionize water. As listed in Table 2 and Supplementary Table S1, the best fitted N was 20.45 for atrazine. The fitted PD-DA equation was then converted to Freundlich isotherm, as also shown in Table 2. The obtained Freundlich isotherm parameters for atrazine were used further for the IAST-EBC model. To acquire the EBC parameters for Central Illinois groundwater, one of the available isotherms of atrazine, at initial concentration of 40.6 mg/L, was first fitted, using the PD-IAST-EBC model. In fitting the PD-IAST-EBC model (equations (3) and (4)) to the isotherm data for natural water, the isotherm parameters obtained from atrazine in deionized water were used, and only the parameters relevant to NOM (C2,0, K2, and n2) were adjusted. The best fitted NOM parameters were 2000 mg/L, 45 ((mg/g)(mg/L)1/n) and 2.22 for C2,0, K2, n2, respectively. These parameters for the IAST-EBC model, are in the ranges reported in Hung and Lin

W. Bunmahotama et al. / Water Research 111 (2017) 28e40

31

Fig. 1. Comparison of model predictions and experimental data of the adsorption capacity for atrazine onto ACs in Central Illinois Water, where (a) is SA UF AC and (b) WPH AC. The data were from Li et al. (2002b), with Co, EBC ¼ 2 mg/L.

(2006b) C2,0, K2, n2 values are 11.7e43,100 mg/L, 0.515e103.231 ((mg/g)(mg/L)1/n) and 1.54e4.76, respectively and in Najm et al. (1991) these values are 785e5106 mg/L, 0.026e16.45 ((mg/g)(mg/ L)1/n) and 0.83e3.33, respectively. As shown in Fig. 1a, the models follow the adsorption data reasonably well for atrazine both in deionized water and in natural water with initial atrazine concentration of 40.6 mg/L. The best fitted isotherm parameters for

NOM were then applied to predict other isotherms for atrazine with different initial concentrations. Also, as illustrated in Fig. 1a, the model predicts the adsorption isotherms for two other atrazine concentrations well (with errors of 18.2%), showing that the model is able to capture the effect of initial concentrations on adsorption capacity. To extrapolate the developed PD-IAST-EBC model to another AC,

32

Table 2 Summary of the parameters used in the PD-IAST-EBC model. Ding (2010)

Hung and Lin (2006b)

WPH (0.383)a SA UF (0.455)a

SA UF (0.455)a SA Super (0.425)a W35 (0.355)a W20 (0.325)a

F400 (0.465)a F600 (0.449)a

TAC (0.382)a WPH (0.383)a

n N chemical Single-solute Freundlich isotherm parameters chemicals K1 ((mg/g)(mg/L)1/n) 1/n1 (e)

1.1 Atrazine (20.5) WPH (K1 ¼ 17.4) SA UF (K1 ¼ 20.9) 1/n1 ¼ 0.342

MTBE (16.5) F400 (K1 ¼ 0.194) F600 (K1 ¼ 0.187) 1/n1 ¼ 0.453

2-MIB (17.8) TAC (K1 ¼ 171) WPH (K1 ¼ 171) 1/n1 ¼ 0.424

N

Central Illinois groundwater (22.4)

FS surface water (10.4) STL groundwater (10.2)

FS surface water (327)

EBC

SA UF (K1 ¼ 20.3) SA Super (K1 ¼ 19.0) W35 (K1 ¼ 15.8) W20 (K1 ¼ 10.4) 1/n1 ¼ 0.355 Clinton water works (21.2)

Single-solute Freundlich isotherm parameters EBCs K2 ((mg/g)(mg/L)1/n) 1/n2 (e)

WPH (K2 ¼ 31.7) SA UF (K2 ¼ 45.0) 1/n2 ¼ 0.450

SA UF (K2 ¼ 55.0) SA Super (K2 ¼ 50.5) W35 (K2 ¼ 44.9) W20 (K2 ¼ 34.4) 1/n2 ¼ 0.400

F400FS (K2 ¼ 0.500) F600FS (K2 ¼ 0.447) 1/n2, FS ¼ 0.500 F400STL (K2 ¼ 0.500) F600STL (K2 ¼ 0.368) 1/n2, STL ¼ 0.500

TACFS (K2 ¼ 687) WPHFS (K2 ¼ 700) 1/n2, FS ¼ 0.025

%SDEV

30.5

44.5

18.7

14.9

a

Newcombe et al. (2002b)

Gillogly et al. (1999)

SA 30 (0.475)a P1100 (0.435)a PCO (0.365)a

WPH (0.383)a Hydrodarco B (0.305)a

SA 30 (K1 ¼ 191) P1100 (K1 ¼ 175) PCO (K1 ¼ 147) 1/n1 ¼ 0.424

WPH (K1 ¼ 171) Hydrodarco B (K1 ¼ 119) 1/n1 ¼ 0.424

Myponga reservoir water raw (42) dilute (341) effluent (294) F1 (251) F2 (185) F5 (137) SA 30F1 (K2 ¼ 1330) SA 30raw (K2 ¼ 1220) P1100raw (K2 ¼ 1200) P1100F1 (K2 ¼ 1300) PCOraw (K2 ¼ 935) PCOF1 (K2 ¼ 1020) 1/n2, raw ¼ 0.025 1/n2, F1 ¼ 0.045 SA 30dilute (K2 ¼ 696) SA 30F2 (K2 ¼ 1730) P1100dilute (K2 ¼ 650) P1100F2 (K2 ¼ 1800) PCOdilute (K2 ¼ 535) PCOF2 (K2 ¼ 1330) 1/n2, dilute ¼ 0.020 1/n2, F2 ¼ 0.075 SA 30effluent (K2 ¼ 639) SA 30F5 (K2 ¼ 1990) P1100effluent (K2 ¼ 600) P1100F5 (K2 ¼ 2200) PCOeffluent (K2 ¼ 491) PCOF5 (K2 ¼ 1530) 1/n2, effluent ¼ 0.022 1/n2, F5 ¼ 0.120 25.8

Kankakee river water (638) Lake Michigan water (288)

WPHKankakee (K2 ¼ 1310) HydrodarcoKankakee (K2 ¼ 1200) 1/n2, Kankakee ¼ 0.020 WPHMichigan (K2 ¼ 1070) HydrodarcoMichigan (K2 ¼ 1000) 1/n2, Michigan ¼ 0.045

23.1

Obtained from pore size distribution (PSD) data using equation (3), where the PSD for TAC was obtained from Hung (2005), for SA30, P1100 and PCO was from Newcombe et al. (2002a), and for Hydrodarco B was from Tennant (2004).

W. Bunmahotama et al. / Water Research 111 (2017) 28e40

Li et al. (2002b) Wo

W. Bunmahotama et al. / Water Research 111 (2017) 28e40

33

Fig. 2. Comparison of model predictions and experimental data of the adsorption capacity for atrazine onto ACs ion Clinton water works water, where atrazine concentration ¼ (a) 50 mg/L and (b) 10 mg/L. Data were from Ding (2010) (ref [D]), with Co, EBC ¼ 1.5 mg/L.

the properties of the AC need to be considered. According to Bunmahotama et al. (2015), Wo in PD-DA equation may change with the type of AC, causing the change of adsorption capacity. However, the parameters relevant to the chemical (here, atrazine) will remain the same. Therefore, to predict the adsorption of atrazine onto WPH in deionized water, the parameters used for SA UF is used except

that Wo is changed to that for WPH. A value of Wo ¼ 0.383 mL/g was estimated from the PSD reported in Hung (2005) using the Urano equation (Equation (2)). For NOM relevant parameters (C2,0, K2, and n2), the fitted Freundlich parameters from SA UF isotherms were first converted back to PD-DA equation, and N was obtain to be 22.37 for the NOM for Central Illinois groundwater as listed in

34

W. Bunmahotama et al. / Water Research 111 (2017) 28e40

Table 2. Similar to that for atrazine in deionized water, when extrapolating to WPH, Wo for WPH was used in the PD-DA equation. These two PD-DA equations, for atrazine in deionized water and NOM in deionized water, were then transformed to Freundlich isotherms again, and used for the PD-IAST-EBC model. Fig. 1b shows that the PD-IAST-EBC model predicts the experimental data well, with only 36.4% SDEV. Considering that a pure predictive model in such complex natural water system loaded with NOM, the predictions are reasonably well. Fig. 1a and b demonstrate that the model follows the experimental data reasonably well, with only 18.2 and 36.4% SDEV, respectively. Li et al. (2002b) developed correlations to simulate the adsorption data with the observed difference between model and data being 14.5% and 12.7% SDEV, respectively, for the same set of experimental data. Although the errors of our model are slightly higher than those reported in Li et al. (2002b), our model is satisfactory. This is the consequence of that our model is mainly prediction while the model of Li et al. (2002b) fits almost all data. The PD-IAST-EBC model was further applied to predict the adsorption capacity for atrazine in another natural water (Clinton Water Works) reported in Ding (2010). In the study, adsorption of atrazine onto four ACs (SA UF, SA Super, W35 and W20) in deionized water and one natural water was examined, and the data were used for the analysis. Fig. 2a shows the model fits and prediction of the adsorption capacities for atrazine onto the four ACs. We note that the procedures to obtain the parameters for the model fits and predictions in Fig. 2 are similar to those used for WPH in Fig. 1b. As used for the system of atrazine/WPH carbon in deionized water, the parameters for the PD-DA equation was used except that Wo was changed to those for the four ACs. The values for Wo were estimated to be 0.455, 0.425, 0.355 and 0.325 mL/g, for SA UF, SA Super, W35 and W20, respectively, based on the PSD data from Tang (2007) and the Urano approach (equation (2)). For the NOM parameters, again one set of data (SA UF carbon in natural water with 50 mg/L atrazine) was fitted to obtain C2,0, K2, and n2. The best fitted parameters were then used in the prediction of another isotherm for another atrazine concentration (10 mg/L) onto SA UF in the same natural water. For the estimation of the isotherm parameters for NOM onto three other ACs, the calculation was also based on those for SA UF, PSD data for three other ACs, and based on the Urano approach. The best fitted NOM parameters were 1500 mg/ L, 55 ((mg/g)(mg/L)1/n), and 2.5 for C2,0, K2, n2, respectively. The other model parameters are listed in Table 2. As shown in Fig. 2a, the model predicts the adsorption data reasonably well for atrazine onto the four ACs in deionized water (four isotherms) and in natural water (three isotherms), while only one data set (50 mg/L atrazine onto SA UF in natural water) being fitted. Fig. 2a demonstrates that the model predicts the adsorption isotherms for three other ACs well, with errors of 43.06%, showing that the model is able to capture the effect of initial concentrations on adsorption capacity. Fig. 2b further demonstrates that the model predicts well all the adsorption isotherms for another concentration of atrazine (10 mg/L) onto the four ACs. We note that the errors between model predictions and data are 43.1% and 46.7% for Fig. 2a and b, respectively, indicating that the model is able to take into account the effect of initial concentrations and types of ACs. We also note that the errors between the model predictions and data reduce to 24.7% and 29.3% for Fig. 2a and b, respectively, if the case of W20 AC is not considered. For the four tested ACs, the pore volume distribution of W20 is significantly different to that of the other three (Tang, 2007). As the current model assumes that the adsorption capacity is mainly governed by the available volume of the pores with size less than 3.2 nm, pore volume distribution is not accounted. From the observation of W20 adsorption, the pore volume (size) distribution may have some small effect on the

adsorption capacity. The effect of pore size distribution of AC on the adsorption capacity has been studied (Ebie et al., 2001; Lillodenas et al., 2005), the integration of the PSD into the current Ro model may improve model precision and thus further study is suggested. The PD-IAST-EBC model was further applied to predict the adsorption capacity for atrazine in synthetic-organic matter -laden water. In Ding (2010) and Li et al. (2003a), the authors studied the adsorption of atrazine onto two ACs (WPH and SA UF) using four dyes (Brilliant Yellow (BY), Congo Red (CR), Xylenol Orange (XO), and Evans Blue(EB)), and p-DCB as synthetic NOM. The data were used for the model fits of NOM parameters and predictions for other ACs. In addition, another case for adsorption atrazine onto Hydrodarco B in Missouri reported in Knappe et al. (1998) was also simulated for adsorption. Supplementary Fig. S1 shows the model fits and predicts the adsorption capacities for atrazine onto three ACs reasonably well, further suggesting that the model is satisfactory. 3.2. Predictions of the adsorption capacity for MTBE The PD-IAST-EBC model was further applied to predict the adsorption capacities for another chemical, MTBE, onto two ACs (F400 and F600, both from Calgon) in two natural waters (FS surface water and STL groundwater) reported in Hung and Lin (2006b). To obtain the model parameters, the isotherm data for MTBE onto one of the two ACs studied (F400) in deionized water and another isotherm data for one MTBE concentration onto F400 in natural waters was used. The procedures to obtain the parameters for the model fits and predictions for MTBE (Fig. 3a) are similar to those used for SA UF in Fig. 1a. The values for Wo were estimated to be 0.465 mL/g and 0.499 mL/g for F400 and F600 (Bunmahotama et al., 2015), respectively, and N was fitted to be 16.51 for MTBE. NOM parameters were 7000 mg/L, 0.5 ((mg/g)(mg/L)1/n), and 2 for C2,0, K2, n2, respectively, for FS water, and were 5500 mg/L, 0.5 ((mg/ g)(mg/L)1/n), and 2, respectively, for STL water. All the parameters are listed in Table 2. Figs. 3 and 4 show the model fits and predictions for the adsorption of MTBE onto F400 and F600 ACs in deionized water and in FS and STL natural waters. As shown in figures, based on the model fits of F400 in deionized water, the model is able to predict the adsorption capacities for MTBE onto another AC, F600, in deionized water. The figures also indicate that the model, where the parameters for NOM adsorption were obtained from fitting with only one MTBE concentration in natural water, is able to predict the adsorption for other MTBE concentrations. The errors between model predictions and experimental data are 14.1%, 22.2%, 16.3% and 23.1%, for different AC/natural water combinations: F400/FS (Fig. 3a), F600/FS (Fig. 3b), F400/STL (Fig. 4a), and F600/STL (Fig. 4b). This proves that the model is able to take into account the effect of initial concentrations and natural water sources. Hung and Lin (2006b) developed correlations to simulate the adsorption data with the observed difference between model and data being 14.2e19.6% SDEV, for the same set of experimental data. Since our model is mainly prediction, it has the advantage of reducing the experimental effort required for obtaining model parameters. 3.3. Predictions of the adsorption capacities for 2-MIB The PD-IAST-EBC model was further applied to predict the adsorption capacities for a taste and odor compound, 2-MIB, onto different ACs in NOM-laden water. Newcombe et al. (2002b) studied the adsorption of 2-MIB onto 6 ACs in deionized water and 4 ACs in the natural water, with NOM fractionated by different membranes of different pore sizes or ion exchange resin. As one of

W. Bunmahotama et al. / Water Research 111 (2017) 28e40

35

Fig. 3. Comparison of model predictions and experimental data of the adsorption capacity for MTBE onto ACs in FS Surface Water, where (a) is F400 AC, and (b) is F600. Data were from Hung and Lin (2006b) (ref [H]), with Co, EBC ¼ 7 mg/L.

the ACs is chemical activated, the 3 thermally activated ACs (SA 30 from Carbochem, P1100 and PCO from PICA) examined in both the deionized water and the NOM-laden water were chosen in this study. F-400, an AC only studied in deionized water, was used to fit for the PD equation parameter (N ¼ 17.80), and then transformed to Freundlich isotherm equation, as described in Section 3.1. Also

following the procedure shown in Sec. 3.1, the fitted N was then used to predict the adsorption isotherms for 2-MIB onto the three studied ACs by considering the limiting pore volumes (Wo) of the three ACs (0.475, 0.435 and 0.365 mL/g for SA 30, P1100 and PCO, respectively). Fig. 5 shows the model fits for 2-MIB adsorption onto F-400 and predictions for SA 30, P1100 and PCO, in deionized water.

36

W. Bunmahotama et al. / Water Research 111 (2017) 28e40

Fig. 4. Comparison of model predictions and experimental data of the adsorption capacity for MTBE in STL groundwater, where (a) is F400 AC and (b) is F600. Data were from Hung and Lin (2006b) (ref [H]) with Co, EBC ¼ 5.5 mg/L.

The reasonably good model fits and predictions suggest that the PD model approach is able to simulate the adsorption processes in deionized water. To obtain the model parameters for NOM adsorption, the isotherm data for 2-MIB onto one of the three ACs, P1100, in six

NOM waters fractionated from Myponga reservoir water: raw water (RW), 50% dilution with deionized water (50% RW), RW passing through ion exchange resin (F), F1 (filtrate through membrane with nominal molecular weight (MW) cut-off value ¼ 500), F2 (filtrate of MW ¼ 500e1000) and F5 (filtrate of 30,000) were used for model

Fig. 5. Comparison of model predictions and experimental data of the adsorption capacity for 2-MIB in natural water, where (a) is P1100 AC (b) is SA30 and (c) is PCO. Data were from Newcombe et al. (2002b) (ref [N]) with 2-MIB initial concentration Co ¼ 100 ng/L; NOM in Myponga Reservoir water raw Co, EBC ¼ 12.5 mg/L, in diluted Co, EBC ¼ 6.3 mg/L, in effluent Co, EBC ¼ 4 mg/L, in F1 Co, EBC ¼ 10 mg/L, in F2 Co, EBC ¼ 10 mg/L, and in F5 Co, EBC ¼ 10 mg/L.

38

W. Bunmahotama et al. / Water Research 111 (2017) 28e40

Fig. 6. Comparison of model predictions and experimental data of the adsorption capacity for 2-MIB onto three PACs in natural water. For the data from Gillogly et al. (1999) (ref [G]), Kankakee River Water, Co, EBC ¼ 10 mg/L, and Lake Michigan water, Co, EBC ¼ 10 mg/L. For the data from Hung and Lin (2006b) (ref [H]), FS Surface Water, Co, EBC ¼ 10 mg/L.

fits, also following the procedures described in Sec. 3.1. We note that in Newcombe et al. (2002b), all the experiments were conducted under one initial 2-MIB concentration, 100 ng/L. Fig. 5a shows that the models fit the experimental data reasonably well for all the 6 NOM water cases. These best fitted parameters, 12,500 mg/ L, 1200 ((mg/g)(mg/L)1/n), and 40 for C2,0, K2, n2, respectively, for raw water, 6300/650/50 for diluted water, 4000/600/45.5 for effluent water, 10,000/1300/22.2 for F1, 10,000/1800/13.3 for F2 and 10,000/2200/8.3 for F5, respectively, are listed in Table 2, and were used for the prediction for the other two ACs. Fig. 5b and c compare the model predictions and data for the adsorption of 2-MIB onto two other ACs, SA30 and PCO in NOM water. As shown in the figure, the model predicts the data for SA30 (Fig. 5b) reasonably well for all the studied water, suggesting the capability of extrapolation to other ACs. Although Fig. 5c demonstrates that the models captured the trends for 2-MIB adsorption onto PCO, the model overestimated the adsorption capacity for most waters. One of the probable reasons to cause this discrepancy is likely to be the PSD of the carbon. Similar to the case of adsorption of atrazine onto W20 (Sec. 3.1), the PSD of PCO is very different from the other two ACs studied (P1100 and SA30, with much less portion of pores less than 1.2 nm). Newcombe et al. (2002b) also employed the IAST-EBC model to simulate the adsorption of 2-MIB onto three ACs in all the studied water. Excellent fits between the model and data were obtained, with only 10.2e12.0% errors for the three AC cases. However, in that study, all the models were fitted with the experimental data, and different NOM concentrations (C2,0) were used for the same water in the model. Therefore, good simulation was expectable. In this current study, same NOM parameters were used for the same water, and pure predictions were conducted for two of the three studied ACs. Therefore, the slightly higher errors between the data and model, 14.4%, 15.7% and 37.7% for P1100, SA30, and PCO, respectively, are acceptable.

To further prove the applicability of this modeling approach, the data from two other studies were also tested. Gillogly et al. (1999) reported the adsorption of 2-MIB onto two ACs, Hydrodarco B and WPH, in Kankakee River and Lake Michigan water in US, while Hung and Lin (2006b) studied 2-MIB on 2 ACs, WPH and TAC, in Feng-Shen (FS) surface water in Taiwan. In modeling the adsorption of 2-MIB in the natural waters for these two studies, only the NOM parameters are needed, as the adsorption parameters for 2-MIB has been obtained from F-400 (N ¼ 17.80) as shown in Fig. 5a. The isotherms for one of the studied 2-MIB concentrations were used to fit the NOM parameters for each of the studied waters. The inner figure of Fig. 6 shows almost perfect fits between the model and experimental data obtained from the two studies, with error of 6.2%. The extracted NOM adsorption parameters for FS water/WPH, Kankakee River/Hydrodarco B and Lake Michigan/Hydrodarco B systems, the C2,0, K2 and n2, were 10,000/700/40, 10,000/1200/50 and 10,000/1000/22.2, respectively, and were used for the prediction. Fig. 6 demonstrates that the model predicts the adsorption of 2-MIB onto the three ACs in all the three tested waters excellently. For the same set data of FS water/WPH system, the errors between model and data were 11.8% for Hung and Lin (2006b) and it is 20.8% for this study. Although slightly higher error is found, this current study predicts the 2-MIB isotherm in DD water, while that isotherm was obtained from experiments by Hung and Lin (2006b). Therefore, the slightly higher error observed is reasonable. The small errors between model prediction and data, 25.4% on average (the main figure in Fig. 6), indicate that the model possesses good predictive capability for the studied systems. We also note that the fitted normalization factors of NOM (NEBC) for MTBE and 2-MIB in FS water are different, with values of 10.4 and 327.1, respectively. Although NEBC is expected to be the same theoretically, as they have the same NOM, difference is observed for 2-MIB and MTBE. It is interesting to see that for all the tested cases in this study if the concentration is in the level of mg/L, such as MTBE and atrazine

W. Bunmahotama et al. / Water Research 111 (2017) 28e40

39

(Table 2), then NEBC is in the range of 10.2e22.4. However, for 2MIB, a compound with concentration level of ng/L, the NEBC is much bigger, at around 136.8e638.2. Although the level of concentration is suspected to be an important factor to influence NEBC, more studies are needed to understand this alteration. In addition to the three studies tested, Chen et al. (1997) reported the adsorption of 2-MIB onto two other ACs, F200 and Lignite, with unknown PSD information, in the raw water of the Miramar Water Treatment Plant, San Diego, California. The PDIAST-EBC models also predict the experimental data reasonably well, with 35.5% and 16.0% errors between models and data for the four isotherms in DD water and natural water and for the two isotherms in natural water, respectively. The modeling procedures, parameters, and results are shown in Supplementary Table S1 and Fig. S2.

Although for almost all cases tested in this study, the model simulates and predicts the experimental data excellently, in a few AC cases with dissimilar PSD, large errors between model predictions and data were observed. One possible explanation is the level of activation during AC manufacturing. Different level of steam activation may lead to different pore size distribution development n et al., 2008), (Daud et al., 2000; Linares-Solano et al., 2000; Roma and this may affect the adsorption capacity. In this current study, the two ACs, P1100 and PCO, tested in the 2-MIB case (Sec. 3.3 and Fig. 5), were made from the same material with different levels of activation (Newcombe et al., 2002a). As shown in the figure, the model predicts the data for P1100 excellently, but does not for PCO, the lower activated one. In fact, the PSD of PCO is very different from that of P1100, and it may be one of the reasons to cause the discrepancy between model and data.

3.4. Applications and limitations of the model

4. Conclusions

The proposed PD-IAST-EBC model has been successfully employed to simulate and predict the adsorption capacity for three trace organic compounds, including atrazine, MTBE and 2-MIB onto 14 ACs in 19 synthetic and natural waters reported in 8 references. In addition, the approach has also been used to simulate the adsorption of 2,4,6-trichlorophenol onto WPH AC in 3 natural waters (see Supplementary Material, Supplementary Table S1, and Supplementary Fig. S1) (Najm et al., 1991). The good agreement between the model and data shows that the model is applicable to different chemical/AC/natural water combinations. Adsorption of organic compounds onto ACs in natural water has been studied for long time. However, predictions of the adsorption are always difficult, as NOM in natural water may affect the adsorption. Previous studies, such as Li et al. (2002b), Hung and Lin (2006b), Newcombe et al. (2002b), Li et al. (2003a), Knappe et al. (1998) and Najm et al. (1991), have developed excellent models to improve the simulation and/or prediction of the adsorption capacity for organic compounds onto ACs. However, their models do not have the capability to extrapolate to other ACs. When changing ACs, more adsorption experiments in both deionized water and natural water are needed before predicting the adsorption capacity. In the current PD-IAST-EBC approach, prediction of adsorption capacity for other ACs in the same natural water is possible, without any extra experiments needed, provided that the PSD information is available. This approach will greatly reduce the experimental effort for the water utilities when AC is to be changed. Although the proposed PD-IAST-EBC model has been successfully employed in prediction of the adsorption of many combinations of organic chemicals, ACs, and source waters, the model still has its limitations. NOM is the key component in natural water to affect the adsorption of organic compounds on to ACs. As it is source water dependent, in the current model, the isotherm parameters for NOM (C2,0, K2, n2) in a specific source water must be fitted with experimental data. Then the PD-IAST-EBC model can be used for prediction for other compounds/ACs in the same water. Attempts have been made to characterize the properties of NOM, based on approximate molecular sizes using size exclusion method and membrane separation (Cho et al., 2000; Her et al., 2002; Pelekani et al., 1999), fluorescence using liquid chromatography (Baghoth et al., 2011; Wu et al., 2003), and polarity using resin separation (Kitis et al., 2002), but none of the properties has been linked to AC adsorption. In our current model, the NOM parameters were obtained based on experimental fits, and no predictions were made from the physical/chemical properties of NOM. It would be useful if the adsorption isotherms of NOM can be linked to its properties so that prediction would be possible. Another limitation of the present model is the type of pore size distribution of ACs.

In this study a model, based on the PD equation, the PSD of AC, and EBC assumption for NOM, has been developed for estimating the adsorption capacities for organic compounds onto AC in the presence of NOM. In the model the adsorption parameters for the studied chemicals were first fitted with the data obtained in deionized water for a specific AC. After incorporating with the PSD of other ACs, the extracted adsorption parameters were successfully employed to predict the adsorption isotherms for other ACs in deionized water using the PD-DA equation. To predict the adsorption in natural water, a set of adsorption data in natural water were first fitted for the targeted chemical on a specific AC to obtain the EBC parameters. Then the EBC parameters obtained were combined with those parameters of the targeted chemical and of ACs to further predict the adsorption of the targeted chemicals in different concentrations and in other AC systems. This PD-IAST-EBC approach has been proven successful in estimating the adsorption of atrazine, MTBE, 2-MIB and 2,4,6-trichlorophenol onto 14 ACs in 22 synthetic and natural waters reported in 9 references. The proposed simple PD-IAST-EBC approach is able to predict the adsorption isotherms for the tested organic compounds onto different ACs in the same natural water, provided that the ACs are thermally activated and the PSDs are known in advance. To improve the model predictions in natural water, further studies are suggested to focus on NOM adsorption properties, on model extrapolation to very different concentration levels, and on the effect of pore size distribution on adsorption. Acknowledgements This work was supported in part by the Headquarters of University Advancement at the National Cheng Kung University sponsored by the Ministry of Education, in part by Industrial Technology Research Institute Project (FF55RN8000), and in part by Ministry of Science and Technology (MOST 104-2221-E-006 -021 -MY3), Taiwan. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.watres.2016.12.033. References Al-Degs, Y.S., El-Barghouthi, M.I., El-Sheikh, A.H., Walker, G.M., 2008. Effect of solution pH, ionic strength, and temperature on adsorption behavior of reactive dyes on activated carbon. Dyes Pigments 77 (1), 16e23. Babarao, R., Hu, Z., Jiang, J., Chempath, S., Sandler, S.I., 2007. Storage and separation of CO2 and CH4 in silicalite, C168 schwarzite, and IRMOF-1: a comparative

40

W. Bunmahotama et al. / Water Research 111 (2017) 28e40

study from Monte Carlo simulation. Langmuir 23 (2), 659e666. Bae, Y.-S., Mulfort, K.L., Frost, H., Ryan, P., Punnathanam, S., Broadbelt, L.J., Hupp, J.T., Snurr, R.Q., 2008. Separation of CO2 from CH4 using mixed-ligand metal organic frameworks. Langmuir 24 (16), 8592e8598. Baghoth, S., Sharma, S., Amy, G., 2011. Tracking natural organic matter (NOM) in a drinking water treatment plant using fluorescence excitationeemission matrices and PARAFAC. Water Res. 45 (2), 797e809. Bunmahotama, W., Hung, W.-N., Lin, T.-F., 2015. Predicting the adsorption of organic pollutants from water onto activated carbons based on the pore size distribution and molecular connectivity index. Water Res. 85, 521e531. Cessford, N.F., Seaton, N.A., Düren, T., 2012. Evaluation of ideal adsorbed solution theory as a tool for the design of metaleorganic framework materials. Ind. Eng. Chem. Res. 51 (13), 4911e4921. ChemSpider http://www.chemspider.com/. Chen, G., Dussert, B., Suffet, I., 1997. Evaluation of granular activated carbons for removal of methylisoborneol to below odor threshold concentration in drinking water. Water Res. 31 (5), 1155e1163. Cho, J., Amy, G., Pellegrino, J., 2000. Membrane filtration of natural organic matter: factors and mechanisms affecting rejection and flux decline with charged ultrafiltration (UF) membrane. J. Membr. Sci. 164 (1), 89e110. Crittenden, J.C., Sanongraj, S., Bulloch, J.L., Hand, D.W., Rogers, T.N., Speth, T.F., Ulmer, M., 1999. Correlation of aqueous-phase adsorption isotherms. Environ. Sci. Technol. 33 (17), 2926e2933. Daud, W.M.A.W., Ali, W.S.W., Sulaiman, M.Z., 2000. The effects of carbonization temperature on pore development in palm-shell-based activated carbon. Carbon 38 (14), 1925e1932. De Ridder, D.J., Villacorte, L., Verliefde, A.R., Verberk, J.Q., Heijman, S., Amy, G.L., Van Dijk, J.C., 2010. Modeling equilibrium adsorption of organic micropollutants onto activated carbon. Water Res. 44 (10), 3077e3086. Ding, L., 2010. Mechanisms of Competitive Adsorption between Trace Organic Contaminants and Natural Organic Matter on Activated Carbon. University of Illinois at Urbana-Champaign. Dubinin, M.M., 1960. The potential theory of adsorption of gases and vapors for adsorbents with energetically nonuniform surfaces. Chem. Rev. 60 (2), 235e241. Ebie, K., Li, F., Azuma, Y., Yuasa, A., Hagishita, T., 2001. Pore distribution effect of activated carbon in adsorbing organic micropollutants from natural water. Water Res. 35 (1), 167e179. Gillogly, T.E., Snoeyink, V.L., Newcombe, G., Elarde, J.R., 1999. A simplified method to determine the powdered activated carbon dose required to remove methylisoborneol. Water Sci. Technol. 40 (6), 59e64. Graham, M.R., Summers, R.S., Simpson, M.R., MacLeod, B.W., 2000. Modeling equilibrium adsorption of 2-methylisoborneol and geosmin in natural waters. Water Res. 34 (8), 2291e2300. Greene, B.E., Snoeyink, V.L., Pogge, F.W., 1994. Adsorption of Pesticides by Powdered Activated Carbon. American Water Works Research Foundation, Denver, CO, USA. Hand, D.W., Loper, S., Ari, M., Crittenden, J.C., 1985. Prediction of multicomponent adsorption equilibria using ideal adsorbed solution theory. Environ. Sci. Technol. 19 (11), 1037e1043. Her, N., Amy, G., Foss, D., Cho, J., Yoon, Y., Kosenka, P., 2002. Optimization of method for detecting and characterizing NOM by HPLC-size exclusion chromatography with UV and on-line DOC detection. Environ. Sci. Technol. 36 (5), 1069e1076. Hung, H.W., 2005. Remediation of MTBE-contaminated Groundwater Using Adsorbent-based Permeable Reactive Barriers. National Cheng Kung University, Tainan, Taiwan. PhD's thesis. Hung, H.W., Lin, T.F., 2006a. Adsorption of MTBE from contaminated water by carbonaceous resins and mordenite zeolite. J. Hazard. Mater. 135 (1e3), 210e217. Hung, H.W., Lin, T.F., 2006b. Predicting the adsorption capacity and isotherm curvature of organic compounds onto activated carbons in natural waters. Environ. Technol. 27 (3), 255e267. Hung, H.W., Lin, T.F., 2007. Prediction of the adsorption capacity for volatile organic compounds onto activated carbons by the dubinin - radushkevich - langmuir model. J. Air Waste Manag. Assoc. 57 (4), 497e506. Hung, H.W., Lin, T.F., Baus, C., Sacher, F., Brauch, H.J., 2005. Competitive and hindering effects of natural organic matter on the adsorption of MTBE onto activated carbons and zeolites. Environ. Technol. 26 (12), 1371e1382. Hyung, H., Kim, J.-H., 2008. Natural organic matter (NOM) adsorption to multiwalled carbon nanotubes: effect of NOM characteristics and water quality parameters. Environ. Sci. Technol. 42 (12), 4416e4421. Keskin, S., Liu, J., Johnson, J.K., Sholl, D.S., 2008. Testing the accuracy of correlations for multicomponent mass transport of adsorbed gases in metal organic frameworks: diffusion of H2/CH4 mixtures in CuBTC. Langmuir 24 (15), 8254e8261. Kitis, M., Karanfil, T., Wigton, A., Kilduff, J.E., 2002. Probing reactivity of dissolved organic matter for disinfection by-product formation using XAD-8 resin adsorption and ultrafiltration fractionation. Water Res. 36 (15), 3834e3848. Knappe, D.R.U., Matsui, Y., Snoeyink, V.L., Roche, P., Prados, M.J., Bourbigot, M.-M., 1998. Predicting the capacity of powdered activated carbon for trace organic compounds in natural waters. Environ. Sci. Technol. 32 (11), 1694e1698. Krishna, R., Calero, S., Smit, B., 2002. Investigation of entropy effects during sorption

of mixtures of alkanes in MFI zeolite. Chem. Eng. J. 88 (1), 81e94. Li, L., Quinlivan, P.A., Knappe, D.R.U., 2002a. Effects of activated carbon surface chemistry and pore structure on the adsorption of organic contaminants from aqueous solution. Carbon 40 (12), 2085e2100. ~ as, B.J., Snoeyink, V.L., Campos, C., 2003a. Three-component competitive Li, Q., Marin adsorption model for flow-through PAC systems. 1. Model development and verification with a PAC/membrane system. Environ. Sci. Technol. 37 (13), 2997e3004. ~ as, B.J., 2002b. Displacement effect of NOM Li, Q., Snoeyink, V.L., Campos, C., Marin on atrazine adsorption by PACs with different pore size distributions. Environ. Sci. Technol. 36 (7), 1510e1515. ~as, B.J., Campos, C., 2003b. Elucidating competitive Li, Q., Snoeyink, V.L., Maria adsorption mechanisms of atrazine and NOM using model compounds. Water Res. 37 (4), 773e784. denas, M.A., Cazorla-Amoro s, D., Linares-Solano, A., 2005. Behaviour of Lillo-Ro activated carbons with different pore size distributions and surface oxygen groups for benzene and toluene adsorption at low concentrations. Carbon 43 (8), 1758e1767. s, D., MartínLinares-Solano, A., Salinas-Martinez de Lecea, C., Cazorla-Amoro  n, I., 2000. Porosity development during CO2 and steam activation in a Gullo fluidized bed reactor. Energy & Fuels 14 (1), 142e149. Molinspiration http://www.molinspiration.com/. Myers, A., Prausnitz, J.M., 1965. Thermodynamics of mixed-gas adsorption. AIChE J. 11 (1), 121e127. Najm, I.N., Snoeyink, V.L., Richard, Y., 1991. Effect of initial concentration of a SOC in natural water on its adsorption by activated carbon. J. Am. Water Works Assoc. 57e63. Newcombe, G., Morrison, J., Hepplewhite, C., 2002a. Simultaneous adsorption of MIB and NOM onto activated carbon. I. Characterisation of the system and NOM adsorption. Carbon 40 (12), 2135e2146. Newcombe, G., Morrison, J., Hepplewhite, C., Knappe, D.R.U., 2002b. Simultaneous adsorption of MIB and NOM onto activated carbon: II. Competitive effects. Carbon 40 (12), 2147e2156. Pelekani, C., Newcombe, G., Snoeyink, V.L., Hepplewhite, C., Assemi, S., Beckett, R., 1999. Characterization of natural organic matter using high performance size exclusion chromatography. Environ. Sci. Technol. 33 (16), 2807e2813. Pelekani, C., Snoeyink, V.L., 1999. Competitive adsorption in natural water: role of activated carbon pore size. Water Res. 33 (5), 1209e1219. Pelekani, C., Snoeyink, V.L., 2000. Competitive adsorption between atrazine and methylene blue on activated carbon: the importance of pore size distribution. Carbon 38 (10), 1423e1436. Rom an, S., Gonz alez, J., Gonz alez-García, C., Zamora, F., 2008. Control of pore development during CO 2 and steam activation of olive stones. Fuel Process. Technol. 89 (8), 715e720. Rother, J., Fieback, T., 2013. Multicomponent adsorption measurements on activated carbon, zeolite molecular sieve and metaleorganic framework. Adsorption 19 (5), 1065e1074. Shih, T.C., Wangpaichitr, M., Suffet, M., 2003. Evaluation of granular activated carbon technology for the removal of methyl tertiary butyl ether (MTBE) from drinking water. Water Res. 37 (2), 375e385. €€ Sillanpa a, M., 2014. Natural Organic Matter in Water: Characterization and Treatment Methods. Elsevier Science. Smith, E.H., Weber Jr., W.J., 1989. Evaluation of mass transfer parameters for adsorption of organic compounds from complex organic matrices. ESandT Contents 23 (6), 713e722. Tan, I.A.W., Ahmad, A.L., Hameed, B.H., 2008. Adsorption of basic dye on highsurface-area activated carbon prepared from coconut husk: equilibrium, kinetic and thermodynamic studies. J. Hazard. Mater. 154 (1e3), 337e346. Tang, G.C.C., 2007. Roles of Background Compound Molecular Size and Adsorbent Pore Size Distribution in Competitive Adsorption on Activated Carbon. University of Illinois at Urbana-Champaign. Tennant, M.F., 2004. Activation and Use of Powdered Activated Carbon for Removing 2-methylisoborneol in Water Utilities. University of Florida. Urano, K., Omori, S., Yamamoto, E., 1982. Prediction method for adsorption capacities of commercial activated carbons in removal of organic vapors. Environ. Sci. Technol. 16 (1), 10e14. Walton, K.S., Sholl, D.S., 2015. Predicting multicomponent adsorption: 50 years of the ideal adsorbed solution theory. AIChE J. 61 (9), 2757e2762. Wu, F., Evans, R., Dillon, P., 2003. Separation and characterization of NOM by highperformance liquid chromatography and on-line three-dimensional excitation emission matrix fluorescence detection. Environ. Sci. Technol. 37 (16), 3687e3693. Yang, Q., Zhong, C., 2006. Molecular simulation of carbon dioxide/methane/ hydrogen mixture adsorption in metal-organic frameworks. J. Phys. Chem. B 110 (36), 17776e17783. Yu, Z., Peldszus, S., Huck, P.M., 2008. Adsorption characteristics of selected pharmaceuticals and an endocrine disrupting compounddnaproxen, carbamazepine and nonylphenoldon activated carbon. Water Res. 42 (12), 2873e2882. Yu, J.W., Yang, F.C., Hung, W.N., Liu, C.L., Yang, M., Lin, T.F., 2016. Prediction of powdered activated carbon doses for 2-MIB removal in drinking water treatment using a simplified HSDM approach. Chemosphere 156, 374e382.