Chemical Engineering Journal 229 (2013) 239–249

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Journal

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

pH dependence of chlordecone adsorption on activated carbonsand role of adsorbent physico-chemical properties

1385-8947/$ - see front matter � 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.cej.2013.03.036

⇑ Corresponding author. Tel.: +33 590 590 48 33 79.E-mail address: sgaspard@univ-ag.fr (S. Gaspard).

A. Durimel a, S. Altenor a, R. Miranda-Quintana b, P. Couespel Du Mesnil c, U. Jauregui-Haza b,R. Gadiou d, S. Gaspard a,⇑a Laboratoire COVACHIMM, EA 3592 Université des Antilles et de la Guyane, BP 250, 97157 Pointe à Pitre, Guadeloupe, French West Indies, Cedex, Franceb Facultad de Medio Ambiente, Instituto Superior de Tecnologías y Ciencias Aplicadas, Salvador Allende y Luaces, La Habana, Cubac Institut Pasteur de Guadeloupe, Morne Jolivière, 97183 Abymes, Guadeloupe, French West Indies, Franced Institut de Science des Matériaux de Mulhouse, UMR CNRS 7361, 15 rue Jean Starky BP 2488, 68057 Mulhouse, Cedex, France

h i g h l i g h t s

� Chlordecone adsorption isotherms onactivated carbons are well describedby the Fowler–Guggenheim/Jovanovic–Freundlich model.� Carboxylic groups at AC surface plays

a major role for CLD adsorption.� Higher amount of CLD is adsorbed at

pH = pHpzc, when the surface is notcharged.� Model structures of molecular

interactions of chlordecone with thesurface functional groups are shown.

g r a p h i c a l a b s t r a c t

a r t i c l e i n f o

Article history:Received 20 October 2012Received in revised form 7 March 2013Accepted 9 March 2013Available online 18 March 2013

Keywords:KeponeChlordeconeAdsorptionBagasseActivated carbonMolecular modeling

a b s t r a c t

From the 1960s to the 1990s, the large-scale production of banana in the French West Indies required anintensive use of chlorinated pesticides, such as chlordecone (Kepone), resulting in the diffuse contamina-tion of soil and surface waters in the banana-producing areas. For this reason, drinking water plants wereequipped with filters containing commercial activated carbons, being one of the challenges to find localadsorbents for the sustainable management of water treatment plants. In this paper, the adsorption ofchlordecone (CLD) on activated carbons (ACs) prepared from sugar cane bagasse is studied, aiming tounderstand the mechanism of CLD adsorption on the AC surface. First, textural, acido-basic and chemicalcharacteristics of the ACs were determined by thermal desorption, X-ray photoelectron and Boëhm stud-ies. Adsorption isotherms of CLD show that the adsorption capacity increases with the amount of carbonand acidic groups at the AC surface whereas basic groups, hydroxyl and ether groups are detrimental toadsorption. The adsorption capacity is maximized at a solution pH level equal to the pHpzc of the consid-ered AC. From temperature programmed desorption studies, it is proposed that chlordecone adsorptionmechanism onto ACs is mainly governed by interaction with carboxylic groups. These results were cor-related to molecular modeling studies of CLD interactions with surface functional groups of AC. The mod-els of preferential positions, corresponding to minimal value of association energy, of interactionsbetween CLD and AC functional groups were obtained.

� 2013 Elsevier B.V. All rights reserved.

240 A. Durimel et al. / Chemical Engineering Journal 229 (2013) 239–249

1. Introduction

Since the 1960s, the French departments of Guadeloupe andMartinique have based their economy on two major agriculturalproductions: banana and sugarcane crops. To prevent the propaga-tion of the banana weevil (Cosmopolite sordidus), which attacks theroots of the banana tree, chlorinated pesticides, such as chlorde-cone (CLD), hexachlorocyclohexane (HCH) and dieldrine wereextensively used until the beginning of the 1990s, resulting inthe contamination of soil and surface waters [1,2].

Chlordecone was used since 1972 and then definitely banned inearly 1993. In 2009, kepone was included in the Stockholm Con-vention on persistent organic pollutants, which bans its productionand use worldwide [3].

Chlordecone (C10H10O, is the common name of decachloropen-tacyclo[5.3.0.02.6.03.9.04.8]decan-5-one (CAS 143-50-0) (Fig. 1)has a high molecular weight (490.64 g/mol), a low solubility inwater (2.7 mg at 25 �C) and its vapor tension is less than2.25 � 10�7 mmHg at 25 �C [4]. Its partition coefficient Koc variesfrom 2000 to 2500 L/kg, depending on soil physico-chemical prop-erties [1]. Due to its strong persistence in natural environments, itshigh resistance to chemical reactions and microbiological degrada-tions, around 8–9% of the cultivation areas of Guadeloupe containCLD concentrations higher than 1 mg/kg in topsoil, and some bana-na fields exhibit CLD content higher than 9 mg/kg [1]. CLD may bebound to soils for several decades to half a millennium, dependingon soil type [1]. CLD has a strong affinity for lipids, accumulates inthe food chain [5,6] and is known for its endocrine-disrupting char-acter [7,8] and its carcinogenic potential [9–11]. In order to limitimpregnation of the population by CLD in Guadeloupe and Marti-nica, drinking water and production plants were equipped withactivated carbon filters.

Indeed, activated carbon is commonly used for treating water,especially to remove pesticides from contaminated water [12–16]. There is, however, no available research data on chlordeconeadsorption by activated carbons or by other any sorbent. In this re-search work, sugarcane bagasse activated carbons (SCB ACs) wereprepared and used to remove CLD from contaminated water. Prep-aration of activated carbons from sugar cane bagasse a by-productof sugar industry, has already been the subject of other studies[17–21]. Sugar cane bagasse allowed the production of ACs withinteresting textural and adsorption properties [17,20,21]. The useof this renewable resource provides a sustainable cleaning process,because a high value product is obtained from a low cost material,and, simultaneously, it brings, solutions to the problem of wastesand local water pollution. We prepared SCB ACs containing largemicropores and small mesopores with different pore size distribu-tion and chemical groups at their surface. Since the size of the mol-ecule is below 1 nm, these kinds of pores are expected to be themost efficient for CLD adsorption. Adsorption isotherm studies ofCLD onto SCB ACs were done. The influence of solution pH, acti-vated carbon surface properties and textural characteristics ofthe adsorption of CLD onto SCB ACs was also investigated. Theseresults were correlated to molecular modeling of CLD interactions

x

O Cl

Fig. 1. Structure of CLD molecule. Calculated critical dimensions: x = 6.52 Å,y = 5.89 Å and z = 5.30 Å.

with surface functional groups of AC. Overall, the results obtainedfrom this study allow us to bring some understanding to theadsorption mechanism of chlordecone onto ACs.

2. Experimental method

2.1. Chemicals

Chlordecone (97.5%) was provided by Cluzeau Info Laboratory.All safety and preventive measures were taken during all experi-ments. The operator wearied protective equipments, appropriatemask and gloves All CLD containing samples were prepared andhandled in a laboratory hood. CLD containing wastes were segre-gated and disposed in appropriate wastes containers. The structureof the molecule (Fig. 1) was studied with MOPAC software [22].

2.2. Activated carbons: preparation

The ACs were obtained from sugarcane bagasse collected inGuadeloupe, French West Indies. In this experiment, two conven-tional methods of AC preparation described in [23] were used.Sample prepared by steam activation was denominated BagH2O.Those prepared by phosphoric acid (H3PO4) activation withimpregnation ratio, XP, g H3PO4/g precursor: 0.5:1, 1:1 and 1:1.5,are called: BagP0.5, BagP1and BagP1.5, respectively. AC sampleswith particle size ranging between 0.4 and 1 mm were used for fur-ther experiments.

2.3. Activated carbon characterization

The textural characterization of the produced ACs was carriedout by N2 adsorption at 77 K using a Micromeritics model ASAP-2020 analyzer. The total surface area was determined by the BETmethod (SBET) while microporous surface (Smicro), external surface(Sext), total pore volume (VT) and micropore volume (Vmi) were ob-tained by the t-plot method. The mesopore volume (Vme) was com-puted by applying the BJH method on the desorption branch andthe mean pore diameter (Dp) was calculated by the equation4 � VT/SBET [22].

The percentage of the surface functional groups of the ACs wasestimated by X-ray Photoelectron Spectroscopy (XPS). XPS mea-surements were conducted on an Axis-Ultra DLD Model Kratos,equipped with a hemispherical electron analyzer and Al Ka(1253.6 eV) X-ray exciting source, as described in [23].

The total surface basicity and acidity of the samples and the pHat point of zero charge (pHpzc) were measured as described in [23].

2.4. Chlordecone quantification

Quantification of CLD was carried out using a liquid chromatog-rapher equipped with a mass spectrophotometer (AGILENT LC/MS1100 series system). Ionization of chlordecone was achieved byelectrospray (API-ES) in negative ion mode. The final parametersof the nebulizer chamber were defined as follow: drying gas flow:12 L/min; temperature of drying gas: 350 �C; pressure of atomizer:35 psi; capillary tension: 4000 V; and collision energy: 50 eV. Theproduct-ion mass spectra of CLD were m/z 507 and 509. The LC/MS analysis were carried out using a C8 column (2.1 � 150 mm,Eclipse X08-C8). The LC separation was performed at 80 �C usinga gradient composed of water and acetonitrile (ACN). The gradientchanged as follow: 0–6 min 55% of ACN, 6–7 min increasing to100%.

A. Durimel et al. / Chemical Engineering Journal 229 (2013) 239–249 241

2.5. Adsorption isotherms

In a preliminary experiment the equilibrium time of adsorptionwas determined (5 days). For adsorption equilibrium experiments,a fixed carbon mass (5 mg) was weighed into 200 mL conical flaskscontaining 100 mL of CLD solution at different initial concentra-tions ranging from 50 to 2800 lg/L of CLD. The solution was agi-tated at 200 rpm and at 25 �C until equilibrium was obtained.Each experiment was repeated at least two times under identicalconditions. The equilibrium concentration of CLD on solid phasewas calculated as qe = V(C0 � Ce)/M (mg/g). A pH adjustment wasdone using HCl or NaOH for equilibrium studies for BagH2O andBagP0.5 samples.

2.6. Temperature-programmed desorption (TPD) experiments

ACs were first dried overnight at 105 �C. A stock solution of CLDwas prepared at 1 g/L dissolving CLD powder in pure acetone.Impregnation of pollutant was carried out adding 30 mg of AC ina CLD solution at concentration of 2.5 mg/L, to ensure a maximaladsorption. Once equilibrium was reached, the contaminated ACswere filtered and then dried at room temperature in a desiccatorfor 48 h. Temperature-programmed desorption (TPD–MS) was car-ried out in a vacuum device equipped with a line-of-slight detec-tion quadrupole mass spectrometer [24,25]. TPD–MS tests wereconducted with about 10 mg of each AC, BagH2O, BagP0.5, BagP1and BagP1.5, before and after contamination by CLD. The acquisi-tion of mass was done in a mass range of 1–200 units. To allow aprecise quantification of desorbed gas, calibration of materialwas first made using four standard gases: H2 (m/z = 2), CO (m/z = 28), CO2 (m/z = 44) and H2O (m/z = 18). The total pressure ofthe gas release during the heat treatment was measured as a func-tion of the temperature, it was compared to the sum of the partialpressures of analyzed gases to allow a precise mass balance.

2.7. Adsorption modeling

The correlation of adsorption data using either a theoretical orempirical equation is essential for practical purposes. Seven differ-ent models were used to fit single component isotherms (Table 1).They can be classified into: (i) simple isotherm models for homoge-neous surfaces without lateral interactions, like the Langmuir [26]and Jovanovic models [27]; (ii) isotherm models for homogeneoussurfaces with lateral interactions, like the Fowler model [28]; (iii)isotherm equations for heterogeneous surfaces without lateralinteractions, like Freundlich and Jovanovic–Freundlich models[29,30], (iv) isotherm equations for heterogeneous surfaces withlateral interactions, like the Fowler–Guggenheim/Jovanovic–Freundlich model [31]; and (v) multilayer isotherm models for

Table 1Adsorption isotherm models: qe is the amount of adsorbate adsorbed at equilibriumper unit amount of adsorbent, qs is the monolayer capacity, ce is the concentration ofadsorbate in aqueous phase at equilibrium, v is the adsorbate–adsorbate interactionparameter, m is the heterogeneity parameter, K, K1 and K2 are affinity parameters.

Model Equation

Langmuir qe = qsKce/(1 + Kce)Jovanovic qe ¼ qsð1� e�ðKceÞÞFowler qe ¼ qsKce=ðe�vqe=qs þ KceÞFreundlich qe ¼ Kcm

e

Jovanovic–Freundlich qe ¼ qsð1� e�ðKceÞm ÞFowler–Guggenheim/Jovanovic–

Freundlichqe ¼ qsð1� e�ðKce evqe=qs Þm Þ

Multilayer Langmuir qe = (qs1K1ce/(1 + K1ce)) + (qs2K2(ce � a)/(1 + K2(ce � a)))

homogeneous surfaces without lateral interactions, like the multi-layer Langmuir model (MLM) [32].

Regressions of the experimental data to the adsorption iso-therm models were performed using a corrected Newton algo-rithm. The procedure calculates the values of the isothermparameters which minimize the residual sum of squares (RSSs):

RSS ¼Xn

i¼1

ðqexp;i � qt;iÞ2 ð1Þ

where qexp,i and qt,i are the experimental and calculated values foreach data point, respectively.

The selection of the most adequate model was performed usingFisher’s test. The model selected exhibited the highest value of theFisher parameter Fcalc [33]:

Fcalc ¼ðn� lÞ

Xn

i¼1

qexpe;i � qexp

e

� �2

ðn� 1ÞXn

i¼1

qexpe;i � qt

e;i

� �2ð2Þ

where qexpe is the mean value of the vector qexp

e and l is the number ofadjusted parameters of the model.

2.8. Molecular modeling of CLD interactions with surface functionalgroups of AC

In order to verify the compliance of the conditions that accord-ing to Giles et al. [34], favor the adsorption of the CLD, several the-oretical calculations with multiple minima hypersurface (MMH)[35,36] procedures were applied. This methodology allows explor-ing the whole conformational space in the interaction of CLD withthe functional surface groups of the AC (carbonyl, hydroxyl, andcarboxyl). This method combines quantum mechanical methodsfor the calculations of energy with statistical thermodynamics toobtain thermodynamic quantities related to the molecular associ-ation process. The AC is simulated by the simplest model: two aro-matic groups with different surface functional groups. Thechemical changes produced in the functional groups at differentpH were also taken into account. In this case, the MMH procedureswere used to study the following systems: CLD–CLD, CLD–car-bonyl, CLD–carboxyl, CLD–hydroxyl, H2O–carbonyl, H2O–carboxyl,H2O–hydroxyl, CLD–carbonylH+, CLD–carboxyl�, CLD–hydroxyl�,H2O–carbonyl+, H2O–carboxyl�, H2O–hydroxyl�. In all cases thefirst step was the complete optimization of the individual molecu-lar structures by semi-empirical AM1 Hamiltonian using MOPACsoftware package version 6.0 [22]. Then, using the programMMH48, it was proceeded to the randomly generation of 200 con-figurations of adsorbent–adsorbate clusters (note that in our casethe functional surface group of AC is modeled as adsorbent andthe adsorbate molecule is CLD or water). Each of the configurationsgenerated by the MMH48 was fully optimized using AM1 semi-empirical formalism also included in the package above men-tioned. The programs ENERGY and Q3 were used for the statisticalprocessing of the results of the optimization process. In the pro-gram Q3, for the calculation of the partition function and the calcu-lation of the thermodynamic quantities of molecular association,next parameters have to be defined: (a) energy (in eV) of the pre-vious optimized individual molecules, (b) temperature at whichthe calculations of thermodynamic magnitudes will be carriedout (298.15 K), (c) convergence value of energy (0.0001 eV), and(d) Tanimoto coefficient (0.85), used to evaluate the similarity be-tween two molecules or molecular configurations [36]. Finally,these results can be displayed using the graphical interface Gra-Multiple. This tool allows, by analyzing the populations of statesof the optimized configurations of the group studied, selecting

242 A. Durimel et al. / Chemical Engineering Journal 229 (2013) 239–249

those structures that have a greater contribution to the macro-scopic state of the studied system.

3. Results

3.1. Characterization of activated carbons

SCB ACs are essentially mesoporous as shown by the texturalparameters presented in Table 2. These parameters (specific sur-face area and pore volume) have been extensively described in aprevious paper [37]. Surface functional groups of ACs are veryimportant because they determine the surface properties of thecarbons. Total acidic and basic groups contents are presented inTable 3. The content of the basic groups was much lower in ACs ob-tained by chemical activation than for the sample obtained bysteam activation. Moreover, Table 3 shows that total basic groupsand pHpzc values decreased with increasing XP values, while thechemically prepared ACs have the lowest value of pHpzc, which isconsistent with the highest content of acidic groups. The highestcontent of basic groups is found in the sample prepared by steamactivation (BagH2O).

Surface chemistry of raw AC samples was characterized by TPDexperiments. Surface oxygen groups on carbon materials decom-pose upon heating producing CO and CO2 at different tempera-tures. TPD peaks can be tentatively assigned to the differentfunctional groups by comparison with the data listed in the litera-ture. It was established that CO2 desorption at low temperature isthe consequence of decomposition of acid groups such as carbox-ylic groups, anhydrides, lactones and peroxide [34,35]. CO obtainedat high temperature is the result of decomposition of neutral or ba-sic groups, such as phenol, ethers and carbonyls [38,39]. Recently, acomparison between TPD–MS and TPD–XPS analysis allowed amore precise correlation between desorbed species and surfacegroups [24]. CO2 desorption curves of all raw bagasse activated car-bon samples present two maxima at 230 �C and 800 �C (Fig. 2).Desorption obtained at 230 �C could correspond to the presenceof carboxylic groups, while that observed at 800 �C, could be char-acteristic of anhydride or lactones. TPD profile for BagH2O exhibitsa desorption peak at 580 �C, which may be attributed to carbonateson AC surface. To check this, elemental analysis of ACs was carriedout by EDX. The results, listed in Table 4, show for BagH2O a per-centage of calcium and magnesium of 3.3% and 1.4% respectively.Calcium present on AC surface could interact with CO2 from air,allowing the formation of carbonates. This hypothesis is reinforced

Table 2Textural characteristics of sugar cane bagasse activated carbons.

Sample SBET (m2/g) Smi (m2/g) Smi/SBET (%) Sext (m2/g) V

BagH2O 1242 1026 83 216 0BagP0.5 1269 1184 93 84 0BagP1 1502 911 60 591 0BagP1.5 1492 327 22 1165 0

Table 3Chemical characteristics of sugar cane bagasse activated carbons.

Sample Composition (%)

C O O/C

BagH2O 75.69 14.82 0.19BagP0.5 87.06 9.06 0.10BagP1.1 87.71 9.05 0.10BagP15 88.88 7.92 0.09

by the fact that only BagH2O has calcium on its surface and a max-imum for CO2 desorption is found at 580 �C. For samples activatedby H3PO4, the acid etching and the subsequent washing lead to aremoval of alcaline and alcaline earth species from the material.This is not the case for BagH2O for which no washing is needed.The formation of CO at around 800 �C for all ACs is due to the pres-ence of carbonyls, ethers or quinones. The surface chemistry ofBagH2O results from the mechanism of water activation, which isbased on a two step dissociation of the water molecule on the ac-tive sites of the carbon material. This lead mainly to the formationof CAH bonds and of semi-quinones [40] Therefore, BagH2O ismostly basic, not hydrophyllic material and it desorbs mostly car-bon oxides around 700 �C [24]. Fig. 2 shows also the evolution ofwater formation for all ACs. The curves obtained shows a waterdesorption peak between 200 and 400 �C, while for BagP0.5 andBagP1 an additional peak is observed at 600–800 �C. The low tem-perature peak is related to water desorption, the relatively hightemperature (above 200 �C) of this desorption shows that waterwas bonded to the carbon surface through relatively high energyinteractions, that is water in microporosity, or hydrogen bondswith hydrophillic oxygenated surface groups like carboxyls. Theformation of water at high temperatures is related to dehydrationreactions of neighbor carboxylic and phenolic groups which giveanhydrides, lactones and ethers [41,42]. These groups then decom-pose to a mixture of CO and CO2 above 600 �C [24]. Total amount ofCO and CO2 (CO2 + CO) and of H2O, released in TPD measurements,follows the order BagP1 > BagP0.5 > BagP1.5 > BagH2O (Table A1)as well water desorption occurs at the same desorption tempera-tures than CO2 and CO, which is in agreement with the hypothesisof dehydration of oxygenated groups. These results are confirmedby XPS studies showing that for the prepared AC samples, the C1s

spectrum can be deconvoluted into five components (Table A2),with chemical shifts corresponding to graphite (284.1–184.4 eV),hydroxyl groups, phenolic, alcohol or ether aromatic carbon(284.8–285.2 eV), carbonyl groups (285.5–286.1 eV), carboxyl andester groups (286.3–287.6 eV) and peaks corresponding to p–p�

transitions in the aromatic carbon (289.5–290.0 eV) [43,44]. TheO1s spectrum was fitted to the following three components: C@Ogroups (530–531.6 eV), CAOH or CAOAC groups (532.7–533.3 eV), with the last peak corresponding to chemisorbed oxygen(534.8–535.7 eV) [44]. According to the area-simulating curve, thepercentage of each functional group was calculated and listed inTable A2. The content of graphitic carbon is higher for BagH2Oand BagP1.5 and it increases with increasing XP value. On the otherhand, the hydroxyl and carboxyl groups are lower for the basic

mi (cm3/g) Vme (cm3/g) Vtot (cm3/g) Vmi/Vtot (%) Dp (nm)

.42 0.27 0.69 60 2.0

.52 0.27 0.79 65 1.9

.45 0.81 1.26 35 2.8

.14 1.49 1.63 8.5 4.2

Acidic groups (meq/g) Basic groups (meq/g) pHpzc

0.510 0.650 8.041.755 0.250 3.832.025 0.187 3.792.227 0.062 3.52

Fig. 2. TPD desorption profiles of bagasse activated carbons.

Table 4Results of elemental analysis (electron microscopy).

C% O% Ca% Mg% P% Si%

BagH2O 55.26 26.57 3.3 1.4 2.59 12.98BagP0.5 74.14 16.14 nd nd 9.72 ndBagP1 61.6 25.57 nd nd 12.84 ndBagP1.5 71.48 21.77 nd nd 7.61 nd

A. Durimel et al. / Chemical Engineering Journal 229 (2013) 239–249 243

BagH2O than for the other samples prepared by chemical activa-tion whereas, the presence of CAOH and CAOAC bonds is higherfor BagH2O.

3.2. Adsorption isotherm

Adsorption isotherm studies of CLD on the ACs were done at25 �C for all AC samples (Fig. 3). In order to understand the roleof ACs’ surface functional groups on CLD adsorption, samplesBagH2O and BagP0.5 with similar BET surface areas and mesoporevolumes (Table 2), but different chemical properties (Table 3) wereselected for additional studies at pH 2, 5, 7, 9 and 11. The averagerelative error of the measured concentrations in the liquid phasewas less than 5% in all cases. The adsorption isotherm curves ob-tained for adsorption of CLD on BagH2O are concave and reach a

plateau for pH values of 2, 5 and 11 (Fig. 4). These curves corre-spond to the L-type isotherm, according to Giles classification[34]. On the other hand, curve shape with a point of inflection, clas-sified as an S-type isotherm is obtained for adsorption of CLD onBagP0.5 (Fig. 5). It is necessary to comment that, at studiedadsorbant/solution ratio, in some cases it was not possible to reach

qe, µg/mg

Ce, µg/L

Fig. 3. Comparison between experimental adsorption isotherms of CLD (symbols)and calculated values (lines) for different sugarcane bagasse activated carbons at25 �C.

Table 5Model parameters and goodness of fit for adsorption isotherms of CLD onto BagH2O atdifferent initial values of pH. qs is the monolayer capacity,v is the adsorbate–adsorbate interaction parameter, m is the heterogeneity parameter, K, K1 and K2 areaffinity parameters, RSS is the residual sum of squares, Fcalc is the calculated Fisherparameter.

Model and parameters pH

2 5 7 9 11

1. Langmuirqs, lg/mgAC 21.2 44.9 51.2 123.1 28.1K, L/lg 0.11 0.0020 0.033 0.0018 0.0086RSS 4.0 10.7 100.1 25.2 53.5Fcalc 78.9 72.1 15.6 62.9 12.5

2. Jovanovicqs, lg/mgAC 20.5 33.2 43.2 74.5 24.8K, L/lg 0.062 0.0024 0.030 0.0028 0.0068RSS 9.0 6.7 175.1 22.8 33.3Fcalc 35.3 115.5 8.9 69.8 20.1

3. Fowlerqs, lg/mgAC 21.1 35.9 51.2 63.0 25.1K, L/lg 0.053 0.0016 0.033 0.0023 0.0019v 1.0 1.4 0.001 1.7 3.1RSS 4.3 3.3 100.1 5.3 15.4Fcalc 61.5 194.1 13.0 256.9 37.1

4. FreundlichK, lg1�m Lm mg�1

AC12.6 0.79 8.1 0.61 3.9

m 0.08 0.53 0.33 0.74 0.27RSS 4.8 40.4 40.7 43.0 101.9Fcalc 66.6 19.1 38.4 36.9 6.6

5. Jovanovic–Freundlichqs, lg/mgAC 21.7 31.9 59.6 54.7 24.3K, L/lg 0.083 0.0026 0.01 0.0047 0.0071m 0.33 1.12 0.51 1.30 1.74RSS 2.6 5.4 58.8 13.4 21.4Fcalc 100.2 119.8 22.1 101.6 26.7

6. Fowler–Guggenheim/Jovanovic–Freundlichqs, lg/mgAC 22.9 31.5 65.9 52.3 24.7K, L/lg 0.0001 0.0002 0.001 0.0001 0.0001v 10.1 4.1 3.1 5.7 6.8m 0.14 0.49 0.32 0.39 0.34RSS 1.2 2.7 21.4 3.6 4.3Fcalc 173.0 191.7 48.6 314.5 110.7

qe, µg/mg

Ce, µg/L

Fig. 4. Comparison between experimental adsorption isotherms of CLD (symbols)and calculated values (lines) for BagH2O activated carbon at 25 �C at different pH.

qe, µg/mg

Ce, µg/L

Fig. 5. Comparison between experimental adsorption isotherms of CLD (symbols)and calculated values (lines) for BagP0.5 activated carbon at 25 �C at different pH.

244 A. Durimel et al. / Chemical Engineering Journal 229 (2013) 239–249

the plateau for the adsorption isotherms due to low solubility ofCLD in water (2800 lg/L).

Tables 5 and 6 summarize the results of the nonlinear regres-sion analysis of the models evaluated in this work. In general,the agreement of the models with the experimental data is goodas Fcalc values show. In almost all cases the best fit was obtainedfor Fowler–Guggenheim/Jovanovic–Freundlich model with theaverage standard error estimation in the range of 1.5–15.1%. Intwo cases (BagH2O at pH = 5 and BagP0.5 at pH = 9) the best fitwas obtained for the Fowler model and for BagP0.5 AC at pH = 2the best fit was for the multilayer Lagmuir model. Figs. 3–5 showthe comparison between experimental adsorption data and valuescalculated using the best equation for each isotherm.

The high absolute values of v, in both Fowler and Fowler–Gug-genheim/Jovanovic–Freundlich equations, show the importance ofadsorbate–adsorbate interactions in the sorption process. Regard-ing the surface heterogeneity, it can be observed that the parame-ter m in Freundlich, Jovanovic–Freundlich and Fowler–Guggenheim/Jovanovic–Freundlich models is different to unityfor all supports. When the heterogeneity parameter is equal tounity, adsorption takes place on homogeneous surface. Then, as itcan be expected, the ACs studied here have strongly heterogeneoussurfaces. This fact is well known, due to both their pore size distri-bution and the presence of the different functional groups on theACs surface. It is not surprising that idealized models like Langmuirand Jovanovic do not fit well the data for all investigated ACs,although Langmuir equation is still often used to analyze adsorp-tion isotherms on such material. Moreover the presence of severalfunctional groups in the adsorbate molecule has been reported todiversify the interactions with activated carbon sites and thus toincrease energy dispersion [45].

As it was mentioned, the MLM gives a good agreement to theexperimental data for BagP0.5 AC at pH = 2 with a Fcalc value of47.1 (Table 6). The S isotherm shape may be due to strong compe-tition from the solvent molecules for the acidic surface sites and tomoderate intermolecular interaction between the CLD molecules,as it will be discussed later as a result of molecular modeling ofthe CLD–CLD and CLD-surface functional groups interactions. Con-sidering the adsorption capacity obtained from the Fowler–Gug-

Table 6Model parameters and goodness of fit for adsorption isotherms of CLD onto chemical activated BagP0.5, BagP1 and BagP1.5 carbons. qs is the monolayer capacity, v is theadsorbate–adsorbate interaction parameter, m is the heterogeneity parameter, K, K1 and K2 are affinity parameters, RSS is the residual sum of squares, Fcalc is the calculated Fisherparameter.

Model and parameters AC and initial pH value

BagP0.5 pH = 2 BagP0.5 pH = 5 BagP0.5 pH = 7 BagP0.5 pH = 9 BagP0.5 pH = 11 BagP1 pH = 5 BagP1.5 pH = 5

1. Langmuirqs, lg/mgAC 16.8 72.8 91.1 119.7 139.1 76.5 61.9K, L/lg 0.0035 0.0047 0.0004 0.001 0.0001 0.0013 0.024RSS 21.3 93.4 21.5 146.5 33.0 13.1 28.1Fcalc 10.8 12.1 38.0 11.0 13.7 84.3 59.9

2. Jovanovicqs, lg/mgAC 14.1 54.0 59.4 105.6 132.4 50.1 48.0K, L/lg 0.0028 0.0054 0.0006 0.001 0.0001 0.0018 0.025RSS 28.8 104.0 25.4 113.0 28.5 12.7 37.3Fcalc 8.0 10.8 32.2 14.2 15.9 87.1 45.2

3. Fowlerqs, lg/mgAC 15.7 57.9 41.8 51.1 28.2 57.9 60.2K, L/lg 0.0024 0.0050 0.0006 0.0008 0.0002 0.0015 0.0231v 1.0 0.8 1.5 2.9 2.8 0.8 0.2RSS 19.6 88.3 18.0 6.2 2.3 11.2 27.9Fcalc 10.3 10.7 39.7 222.6 167.7 82.3 50.2

4. FreundlichK, lg1�m Lm mg�1

AC0.64 1.71 5.24 0.0240 0.0021 0.48 4.66

m 0.44 0.57 0.38 1.22 1.25 0.66 0.49RSS 14.0 44.3 22.8 54.4 14.6 15.7 51.0Fcalc 16.5 25.5 35.9 29.6 30.9 70.8 33.0

5. Jovanovic–Freundlichqs, lg/mgAC 20.7 113.8 164.8 47.3 139.9 59.0 54.8K, L/lg 0.0010 0.0010 0.0001 0.0031 0.0001 0.0013 0.0181m 0.58 0.63 0.85 2.06 1.03 0.90 0.79RSS 16.0 53.5 23.1 34.9 25.1 12.1 25.6Fcalc 12.6 17.6 31.0 39.5 15.4 76.6 54.8

6. Fowler–Guggenheim/Jovanovic–Freundlichqs, lg/mgAC 17.9 46.2 34.1 43.2 23.5 44.3 48.6K, L/lg 0.0001 0.0000 0.0001 0.0003 0.0001 0.0001 0.0001v 4.3 13.5 4.1 4.0 3.4 4.9 8.7m 0.36 0.18 0.49 0.60 0.66 0.41 0.27RSS 4.1 6.6 9.7 5.7 1.7 0.5 12.4Fcalc 42.1 114.2 63.3 201.1 191.7 1637.8 90.7

7. Multilayer Langmuirqs1, lg/mgAC 11.5 27.1 50.7 40.0 281.8 – –K1, L/lg 0.0092 0.0434 0.0009 0.0023 0.0000 – –qs2, lg/mgAC 16.0 46.4 36.8 39.2 26.8 – –K2, L/lg 0.0411 0.0395 0.0067 0.1779 0.0057 – –a, lg/L 818.1 103.7 606.5 288.2 950.3 – –RSS 3.1 51.8 15.8 371.6 6.0 – –Fcalc 47.1 10.9 32.4 2.1 36.8 – –

A. Durimel et al. / Chemical Engineering Journal 229 (2013) 239–249 245

genheim/Jovanovic–Freundlich model, the adsorption capacity fol-lows the order: BagP1.5 > BagP0.5 > BagP1 > BagH2O.

Evolution of the adsorption capacity of CLD on BagH2O andBagP0.5 with pH was studied (Fig. 6). For BagH2O, adsorptioncapacity increases when the pH increases until pH = 7; it decreases

Fig. 6. Influence of pH solution on total CLD adsorption capacity (qmax) of BagH2Oand BagP0.5.

then. The pH at which the adsorption capacity reaches a maximumis close to the value of the pHpzc of BagH2O. Similarly, for BagP0.5,adsorption capacity increases from a pH value of 2 and thenreaches a maximum at a pH value of around 4, close to the pHpzc

value of BagP0.5, and then decreases. On the other hand, for thissample a second maximum is then reached around pH 9. Both re-sults indicate that CLD adsorption is favored when the AC surface isnot charged. The plot of the adsorption capacity of CLD as a func-tion of the BET surface area, the mesoporous and microporous vol-ume did not show any correlation between the porous distributionand the amount of CLD adsorbed. Fig. A1 shows clearly that highamount of C and acidic groups, and consequently low pHpzc favorCLD adsorption, whereas high contents of oxygen, basic groupsand the presence of ether and hydroxyl groups may be detrimentalto CLD adsorption.

3.3. TPD–MS study of interaction between CLD and ACs surface

In order to study the interactions between the AC surface andthe CLD molecule, TPD–MS profiles of raw ACs were compared tothose obtained for CLD contaminated ACs. The total pressure corre-

0

1 10 -5

2 10 -5

3 10 -5

4 10 -5

0 200 400 600 800

Température (°C)

Fig. 7. Difference between the measured and calculated pressures DP = Pmes � Pcalc

obtained during TPD–MS experiments for (s) BagH2O, (+) BagP0.5, (4) BagP1, (h)BagP1.5.

246 A. Durimel et al. / Chemical Engineering Journal 229 (2013) 239–249

sponds to the sum of all gases desorbed during TPD–MS curves ofCLD contaminated ACs and calculated pressure correspond only tothe sum of calibrated gases (see Section 2.6). The difference be-tween the measured and calculated pressures DP = Pmes � Pcalc cor-respond to the desorption of a mixture of hydrocarbons, that is tothe decomposition of the adsorbed CLD. For all samples, a peak ofDP is observed between 100 and 400 �C. It corresponds to thedecomposition of molecules which are strongly interacting withthe activated carbon surface. For the sample BagH2O (Fig. 7), onemore peak of DP is observed at a significantly lower temperature(100 �C), this could be related to the presence of molecules whichare more weakly bonded. However, the total pressure observed forBagP0.5 is lower than the others. This could be explained by thefact that CLD is more weakly bound to the AC surface, due to thepresence of water on the surface. This would be consistent withTPD–MS profiles which present two types of bounding for bothBagP0.5 and BagP1. For BagH2O, TPD–MS profiles of the raw andcontaminated samples, are different for both CO and CO2 desorp-tion. Fig. 8 shows three peaks of CO desorption at 110 �C, 295 �Cand 800–900 �C for CLD contaminated AC, while for raw AC, onlyone or two peaks are present at high temperature. Thus, CLD mol-ecules could be linked to the AC surface groups creating new chem-ical bonds that may cause desorption of CO at low temperature.The desorption of CO at high temperature is also strongly de-creased after CLD adsorption. This shows that CLD adsorption canproceed on oxygenated surface group like phenols or semi-qui-nones, resulting in a destabilization of these groups and a strongdecrease of CO desorption temperature. For raw AC only CO2

desorption is observed at 600 �C and 900 �C (Fig. 8), correspondingrespectively to the desorption of carbonate and lactones/anhy-dride. At 200–400 �C, observed CO2 desorption is the consequenceof carboxylic acid groups decomposition. In complement to this

Fig. 8. Comparison of TPD desorption profiles of raw BagH2O, C

analysis, TPD experiments were carried out for BagH2O impreg-nated with water (Fig. 8), which allows to show that the reactionof water molecules with lactones and/or anhydrides, lead to asmall increase of the amount of carboxylic acid groups. CLD mole-cules can then further interact with the carboxylic acids groups.

For contaminated BagP0.5 and BagP1 samples, TPD profilesshow a superposition of curves for CO and CO2 (Fig. 9). However,this superposition was not perfect, especially at 200–500 �C andaround 800 �C. From these observations, two hypotheses couldbe proposed: (i) CLD molecules are associated with oxygenatedgroups which inhibit desorption and allow formation of smallamount of carboxylic and anhydrides and (ii) the difference is

LD contaminated BagH2O and water impregnated BagH2O.

Fig. 9. TPD desorption profiles of CLD contaminated ACs BagP0.5, BagP1, BagP1.5.

A. Durimel et al. / Chemical Engineering Journal 229 (2013) 239–249 247

negligible; CLD molecules interact with other chemical groupssuch as the graphitic moiety of the carbon. Similar curves are ob-tained for BagP1, reinforcing our hypothesis. For BagP1.5 similarTPD profiles are found for the contaminated and raw sample. How-ever, at 800 �C, CO desorption is lower for the contaminated ACthan for the raw AC. Similarly, for CO2 curves, the peak observedat 250 �C is stronger for the contaminated AC, while the oppositeis obtained at 800 �C. This would indicate that CLD moleculesmay interact with lactones at the AC surface, allowing desorptionof carboxylic acid at 200 �C. Comparison of mass spectra of pureCLD, raw and contaminated ACs shows that mass spectrum ofthe pure CLD molecule is different to those obtained when CLD isreleased from the AC surface. This would imply that CLD is stronglylinked onto AC surface. Moreover, for the four samples, desorptionof HCl can be observed. CLD is the only source of chlorine in con-taminated material that can be attributed to its decomposition.

3.4. Molecular modeling of CLD interactions with surface functionalgroups of AC

Table A3 shows the results of the interactions of CLD with func-tional surface groups of ACs and the interactions CLD–CLD ob-tained by the MMH methodology, expressed through associationthermodynamic properties: energy of association (Eassoc); entropyof association (Sassoc) and Helmholtz free energy of association(Aassoc). These results demonstrate that the CLD-CLD intermolecu-lar interactions are moderate, which is a condition that, accordingto [33], can explain the shape of the obtained isotherms. In the caseof studies concerning the CLD and the surface functional groups ofACs, it was also found that the molecular interactions are weak,when the oxygenated functional groups are not charged, beinghigher for COOH, that confirms the observed behavior of adsorp-tion increase with increasing acidity groups in ACs (Fig. A1). When,due to the pH of the medium, the functional groups are modifiedacquiring a positive or negative charge, the interactions with theCLD increase appreciably. However, the interactions of CLD with

the water molecules of the solvent will be also increase with thevariation of pH, as it will be discussed below. Fig. A2 shows someof the preferential positions of molecular interactions for studiedsystems.

Moreover, Table A4 shows the results obtained by MMH proce-dures for the interaction of the aforementioned functional groupswith the aqueous medium expressed as one molecule of water. Itcan be seen, from the values of the energy of association, thatthe intermolecular interactions of water with surface functionalgroups are significantly higher if compared with CLD. This explainsthat the CLD is in competition with the molecules of the solvent foradsorption sites, being this in turn one of the conditions set byGiles et al. [34] to account for the form of the adsorption isotherms.As noted above for CLD, the interactions between water andcharged functional groups are greatly favored, when comparingthose of CLD with the charged functional groups. Taking advantageof the possibility of the MMH methodology to enable obtain modelstructures of molecular interactions, there are presented in Fig. A3the most important minimum energy structures for interactionsbetween water and different surface functional groups of ACs.

4. Discussion

Higher amounts of CLD were adsorbed on the acidic AC samples,indicating a higher affinity of CLD for acidic surface functionalgroups of the coals. From molecular modeling and TPD experi-ments, it is shown that carboxylic groups play a major role inCLD adsorption. We propose that CLD, may bound the activatedcarbons surface by hydrogen bounds between carboxyl groups atthe ACs surface and the chlorine atoms. Accordingly, the hydrogenbasicity of lindane, another chlorinated pesticide, was demon-strated by calculation of complexation constant, octanol–waterpartition coefficient and electrostatic potential measurements cal-culation, and as well FTIR measurements [46,47]. Isotherm exper-iments have shown as well that CLD adsorption was limited byformation of clusters between water molecules and oxygenated

248 A. Durimel et al. / Chemical Engineering Journal 229 (2013) 239–249

groups. This was confirmed by the S type of isotherm obtained foradsorption of CLD on BagP0.5, that may appear when the followingconditions are fulfilled: the solute molecule (a) is monofunctional,(b) has moderate intermolecular interactions, and (c) meets strongcompetition for substrate sites from molecules of the solvent orother adsorbed species. This may correspond to a competition be-tween the chlordecone molecules and the solvent molecules asdemonstrated by the molecular modeling studies, showing thatwater molecules has a higher affinity than CLD for the oxygenatedfunctional groups both, when they are in their neutral or chargedstate. It was proved that a large number of oxygenated groups de-creases the adsorption of hydrophobic molecules [48,49]. Themolecular modeling studies show that oxygenated functionalitiesinteract preferentially with water molecules in solution. Otheroxygenated functionalities, such as ethers or hydroxyls, may inter-act with water through hydrogen bounds [49,50], increasing thehydrophilic character of the AC and causing a decrease of CLD up-take. The adsorption mechanism of CLD should also be governed byhydrophobic interactions with the AC graphene layer. The study ofthe influence of pH on CLD adsorption on BagH2O and BagP0.5,which have similar textural characteristics but different surfaceproperties, confirmed this hypothesis. The adsorption capacity ofCLD on both carbons reaches a maximum where the solution pHis close to the pHpzc of the AC, where the charge surface is neutral.Thus, at this pH, chlordecone would mainly bind the AC surface byhydrophobic interactions. At pH values lower than pHpzc, very lowamounts of the molecule are adsorbed because (i) at pH < pHpzc,adsorption of protons would induce a steric hindrance and lowerthe affinity of the molecule for the surface and (ii) at pH > pHpzc

electrostatic repulsion between the negative surface of the ACand the molecule would occur. At a pH greater than 9, CLD formsCLD hydroxide (C10Cl10(OH)2) [51,52] by substitution of the car-bonyl function with two alcohol groups. The chlordecone hydrateformed is much more hydrophilic than CLD [1,51] and may havea higher affinity for BagP0.5 surface. At more basic pH levels, themolecule has a lower affinity for the carbon surface due to highamount of water on the surface. No clear influence of porous distri-bution on CLD adsorption was observed in this study, as the ACsprepared here contained significant amounts of both large microp-ores and mesopores, and their mean pore diameter was around2 nm. According to the distance between the oxygen atom andthe farthest chloride atom of CLD depicted in Fig. 1, and addingthe covalence radius of Cl and O, we obtain that the largest dimen-sion of the molecule is 0.81 nm, CLD could be able to enter intolarge micropores and mesopores of all ACs.

5. Conclusion

High amounts of carbon and acidic groups favor CLD adsorptionon studied SCB ACs. CLD adsorption is probably governed as wellby H bounding of chlorine groups with carboxylic groups and byhydrophobic interaction with the graphitic surface. This is rein-forced by results of TPD studies showing that carboxylic groupsat AC surface plays a major role for CLD adsorption and pH studiesshowing that higher amount of CLD is adsorbed at pH = pHpzc,when the surface is not charged. The results of the molecular mod-eling of CLD interactions with surface functional groups of AC arein agreement with experimental adsorption isotherms and the in-crease of adsorption capacity with acidity. In addition, importantmodels of preferential positions, corresponding to minimal valueof association energy, of interactions between CLD and AC func-tional groups were obtained. Future study to optimize CLD adsorp-tion will focus on the modification of the carbon surface bycreating carboxylic groups or by removal of hydrophilic specieswithout enlargement of the micropores.

Acknowledgments

Financial support for this work, provided by the French Over-seas Ministery (MOM), the Regional Council of Guadeloupe andthe Syndicat intercommunal d’alimentation en eau et d’Assainisse-ment de la Guadeloupe (SIAEAG), and the French Embassy in Cuba,are gratefully acknowledged.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.cej.2013.03.036.

References

[1] Y.M. Cabidoche, R. Achard, P. Cattan, C. Clermont-Dauphin, F. Massat, J.Sansoulet, Long-term pollution by chlordecone of tropical volcanic soils in theFrench West Indies: a simple leaching model accounts for current residue,Environ. Pollut. 157 (2009) 1697–1705.

[2] S. Coat, D. Monti, P. Legendre, C. Bouchon, F. Massat, G. Lepoint,Organochlorine pollution in tropical rivers (Guadeloupe): role of ecologicalfactors in food web bioaccumulation, Environ. Pollut. 159 (2011) 1692–1701.

[3] Stockholm Convention. Listing of POPs in the Stockholm Convention: Annex A(Elimination). <http://chm.pops.int/Convention/The%20POPs/tabid/673/language/fr-CH/Default.aspx>, 2011.

[4] INCHEM, Chlordecone, Environmental Health Criteria 43, InternationalProgramme on Chemical Safety, World Health Organization, Genève, 1984.

[5] K. Borga, G.W. Gabrielsen, J.U. Skaare, Biomagnification of organochlorinesalong a Barents sea food chain, Environ. Pollut. 113 (2001) 187–198.

[6] C. Dubuisson, F. Héraud, J.-C. Leblanc, S. Gallotti, C. Flamand, A. Blateau, P.Quenel, J.-L. Volatier, Impact of subsistence production on the managementoptions to reduce the food exposure of the Martinican population toChlordecone, Regul. Toxicol. Pharm. 49 (2007) 5–16.

[7] H. Fang, W. Tong, W.S. Branham, C.L. Moland, S.L. Dial, H. Hong, Study of 202natural, synthetic, and environmental chemicals for binding to the androgenreceptor, Chem. Res. Toxicol. 16 (2003) 1338–1358.

[8] H.T. Tapiero, B.G. Nguyen, New estrogens and environmental estrogens,Biomed. Pharmacother. 56 (2002) 36–42.

[9] M. Landau-Ossondo, N. Rabia, J. Jos-Pelage, L.M. Marquet, Y. Isidore, C. Saint-Aimé, M. Martin, P. Irigaray, D. Belpomme, Why pesticides could be a commoncause of prostate and breast cancers in the French Caribbean Island,Martinique. An overview on key mechanisms of pesticide-induced cancer,Biomed. Pharmacother. 63 (2009) 383–395.

[10] L. Multigner, J.R. Ndong, A. Giusti, M. Romana, H. Delacroix-Maillard, S.B.Cordier, B. Jégou, J.P. Thome, P. Blanchet, Chlordecone exposure and risk ofprostate cancer, J. Clin. Oncol. 28 (2010) 3457–3462.

[11] EPA, Toxicological Review of Chlordecone (Kepone). US-EPA Publications EPA/635/R-07/004F, 2009.

[12] A. Bembnowska, R. Pelech, E. Milchert, Adsorption from aqueous solutions ofchlorinated organic compounds onto activated carbons, J. Colloid Interface Sci.265 (2003) 276–282.

[13] V.O. Njoku, B.H. Hameed, Preparation and characterization of activated carbonfrom corncob by chemical activation with H3PO4 for 2,4-dichlorophenoxyacetic acid adsorption, Chem. Eng. J. 173 (2011) 391–399.

[14] B.H. Hameed, J.M. Salman, A.L. Ahmad, Adsorption isotherm and kineticmodeling of 2,4-D pesticide on activated carbon derived from date stones, J.Hazard. Mater. 163 (2009) 121–126.

[15] J.M. Salman, V.O. Njoku, B.H. Hameed, Bentazon and carbofuran adsorptiononto date seed activated carbon: kinetics and equilibrium, Chem. Eng. J. 173(2011) 361–368.

[16] C. Faur, A. Cougnaud, G. Dreyfus, P. Le Cloirec, Modelling the breakthrough ofactivated carbon filters by pesticides in surface waters with static andrecurrent neural network, Chem. Eng. J. 145 (2008) 7–15.

[17] D. Kalderis, D. Koutoulakis, P. Paraskeva, E. Diamadopoulos, E. Otal, J.O. Valle,C. Fernández-Pereira, Adsorption of polluting substances on activated carbonsprepared from rice husk and sugarcane bagasse, Chem. Eng. J. 144 (2008) 42–50.

[18] R.S. Juang, F.C. Wu, R.L. Tseng, Characterization and use of activated carbonsprepared from bagasses for liquid-phase adsorption, Colloids Surf., A 201(2002) 191–199.

[19] T.H. Liou, Development of mesoporous structure and high adsorption capacityof biomass-based activated carbon by phosphoric acid and zinc chlorideactivation, Chem. Eng. J. 158 (2010) 129–142.

[20] I.D. Mall, V.C. Srivastava, N.K. Agarwal, Adsorptive removal of Auramine-O:Kinetic and equilibrium study, J. Hazard. Mater. 143 (2007) 386–395.

[21] M. Valix, W.H. Cheung, K. Zhang, Role of heteroatoms in activated carbon forremoval of hexavalent chromium from wastewaters, J. Hazard. Mater. 135(2006) 395–405.

[22] J.J.P. Stewart, MOPAC Manual, 6th ed., Frank J. Seiler Research LaboratoryUnited State Air Force Academy, 1990.

A. Durimel et al. / Chemical Engineering Journal 229 (2013) 239–249 249

[23] S. Altenor, B. Carene, E. Emmanuel, J. Lambert, J.J. Ehrhardt, S. Gaspard,Adsorption studies of methylene blue and phenol onto vetiver roots activatedcarbon prepared by chemical activation, J. Hazard. Mater. 165 (2009) 1029–1039.

[24] R. Gadiou, E.A. Dos Santos, M. Vijayaraj, K. Anselme, J. Dentzer, G.A. Soares, C.Vix-Guterl, Temperature-programmed desorption as a tool for quantificationof protein adsorption capacity in micro- and nanoporous materials, ColloidsSurf., B 73 (2009) 168–174.

[25] P. Brender, R. Gadiou, J.-C. Rietsch, P. Fioux, J. Dentzer, A. Ponche, C. Vix-Guterl,Characterization of carbon surface chemistry by combined temperatureprogrammed desorption with in situ X-ray photoelectron spectrometry andtemperature programmed desorption with mass spectrometry analysis, Anal.Chem. 84 (2012) 2147–2153.

[26] I. Langmuir, The adsorption of gases on plane surfaces of glass, mica andplatinum, J. Am. Chem. Soc. 40 (1918) 1361–1403.

[27] D.S. Jovanovic, Physical adsorption of gases, Kolloidnii Zhurnal 235 (1969)1203–1208.

[28] R. Fowler, E.A. Guggenheim, Statistical Thermodynamics, Cambridge Univ.Press, Cambridge, 1965.

[29] H. Freundlich, Over the adsorption in solution, Z. Phys. Chem. 57 (1906) 384–470.

[30] I. Quiñones, G. Guiochon, Derivation and application of a Jovanovic–Freundlichisotherm model for single-component adsorption on heterogeneous surfaces,J. Colloid Interface Sci. 183 (1996) 57–67.

[31] I. Quiñones, G. Guiochon, Extension of a Jovanovic–Freundlich isotherm modelto multicomponent adsorption on heterogeneous surfaces, J. Chromatogr. A796 (1998) 15–40.

[32] P.G. Grant, S.L. Lemke, M.R. Dwyer, T.D. Phillips, Modified langmuirequation for S-shaped and multisite isotherm plots, Langmuir 14 (1998)4292–4299.

[33] S.L. Ajnazarova, V.V. Kafarov, Metodi optimisatsi eksperimenta v khimicheskoyteknologui, Vishaia Shkola, Moscow, 1985.

[34] C.H. Giles, T.H. MacEwan, S.N. Nakwa, D.J. Smith, Studies in adsorption. Part XI.A system of classification of solution adsorption isotherms, and its use indiagnosis of adsorption mechanisms and in measurement of specific surfaceareas of solids, J. Am. Chem. Soc. (1960) 3973–3993.

[35] L.A. Montero, A.M. Esteva, J. Molina, A. Zapardiel, L. Hernandez, H. Marquez,A.A. Acosta, Theoretical approach to analytical properties of 2,4-diamino-5-phenylthiazole in water solution. Tautomerism and dependence on pH, J. Am.Chem. Soc. 120 (1998) 12023–12033.

[36] L.A. Montero, J. Llano, J. Molina, J. Fabian, Multiple minima hypersurfaces ofwater clusters for calculations of association energy, Int. J. Quantum Chem. 79(2000) 8–16.

[37] N. Passé-Coutrin, S. Altenor, D. Cossement, C. Jean-Marius, S. Gaspard,Comparison of parameters calculated from the BET and Freundlichisotherms obtained by nitrogen adsorption on activated carbons: a new

method for calculating the specific surface area, Microporous MesoporousMater. 111 (1-3) (2008) 517–522.

[38] M. Molina-Sabio, M.T. Gonzalez, F. Rodriguez-Reinoso, A. Sepúlveda-Escribano,Effect of steam and carbon dioxide activation in the micropore sizedistribution of activated carbon, Carbon 34 (1996) 505–509.

[39] E. Castillejos-López, D.M. Nevskaia, V. Muñoz, A. Guerrero-Ruiz, On theinteractions of phenol, aniline and p-nitrophenol on activated carbon surfacesas detected by TPD, Carbon 46 (2008) 870–875.

[40] J.-C. Rietsch, P. Brender, J. Dentzer, R. Gadiou, L. Vidal, C. Vix-Guterl, Evidenceof water chemisorption during graphite friction under moist conditions,Carbon 55 (2013) 90–97.

[41] S. Haydar, C. Moreno-Castilla, M.A. Ferro-García, F. Carrasco-Marín, J. Rivera-Utrilla, A. Perrard, J.P. Joly, Regularities in the temperature-programmeddesorption spectra of CO2 and CO from activated carbons, Carbon 38 (2000)1297–1308.

[42] G.S. Szymanski, Z. Karpinski, S. Biniak, A. Swiatkowski, The effect of the gradualthermal decomposition of surface oxygen species on the chemical andcatalytic properties of oxidized activated carbon, Carbon 40 (2002) 2627–2639.

[43] W. Jung, K.H. Ahn, Y. Lee, K.P. Kim, J.S. Rhee, J.T. Park, K.J. Paeng, Adsorptioncharacteristics of phenol and chlorophenols on granular activated carbons(GAC), Microchem. J 70 (2001) 123–131.

[44] S. Figaro, S. Louisy-Louis, J. Lambert, J.-J. Ehrhardt, A. Ouensanga, S. Gaspard,Adsorption studies of recalcitrant compounds of molasses spentwash onactivated carbons, Water Res. 40 (2006) 3456–3466.

[45] A. Derylo-Marczewska, A. Swiatkowski, S. Biniak, M. Walczyk, Effect ofproperties of chemically modified activated carbon and aromatic adsorbatemolecule on adsorption from liquid phase, Colloids Surf. 327 (2008) 1–8.

[46] C. Ouvrard, M. Luçon, J. Graton, M. Berthelot, C. Laurence, Determination of thehydrogen-bond basicity of weak and multifunctional bases: the case of lindane(c-hexachlorocyclohexane), J. Phys. Org. Chem. 17 (2004) 56–64.

[47] M.H. Abraham, K. Enomoto, E.D. Clarke, G. Sexton, Hydrogen bond basicity ofthe chlorogroup; hexachlorocyclohexanes as strong hydrogen bond bases, J.Org. Chem. 67 (2002) 4782–4786.

[48] F. Ahnert, H.A. Harafat, N.G. Pinto, A Study of the influence of hydrophobicityof activated carbon on the adsorption equilibrium of aromatics in non-aqueous media, Adsorption 9 (2003) 311–319.

[49] C. Moreno-Castilla, Adsorption of organic molecules from aqueous solutionson carbon materials, Carbon 42 (2004) 83–94.

[50] A. Dabrowski, P. Podkoscielny, Z. Hubicki, M. Barczak, Adsorption of phenoliccompounds by activated carbon—a critical review, Chemosphere 58 (8) (2005)1049–1070.

[51] G.W. Dawson, W.C. Weimer, S.J. Shupe, Kepone – a case study of a persistentmaterial, Am. Inst. Chem. Eng. (AIChE) Symp. Ser. 75 (1979) 366–374.

[52] P.S. Guzelian, Comparative toxicology of chlordecone (Kepone) in humans andexperimental animals, Annu. Rev. Pharmacol. Toxicol. 22 (1982) 89–113.