Prediction of chlorophenols adsorption on activated carbons by representative pores method

The specification of a particular activated carbon adsorbents for removal of phenol and related derivatives, from dilute aqueous solutions, is still based on lengthy trial and error experimental tests. A predictive model of adsorption of these compounds would considerably reduce the carbon selection time and could also bring new information to support more efficient carbon synthesis. The use of molecular simulation and the methodology of representative pores proved to be adequate for quantitative prediction of phenol adsorption. Here the methodology is being extended to chlorophenols, an important class of phenol-derived pollutants. A set of ortho- and para-chlorophenol isotherms were simulated for different representative pores in order to predict carbon adsorption and determine the most significative pore size. At low concentrations (1 × 10−4 mol/L), the pores of 8.9 and 18.5 Å are the most effective. For concentrations above 3 × 10−4 mol/L, pores in the range of 27.9 Å must be present in the activated carbon. The simulation predicts a step for the 27.9 Å pore that can be correlated with experimental steps in literature. Finally, the adsorption isotherms of chlorophenols for other activated carbons were predicted with the help of the model.


Introduction
Chlorophenolic compounds are part of a group of pollutants often associated with unwanted scenarios of environmental contamination (Häggblom and Bossert 2003). The widespread use of chlorophenols in the production of fertilizers, biocides, medicines, and paints, in addition to the generation of industrial and landfill waste, transfers part of these contaminants to water, soil, and food chains. The majority of known environmental exposures are for surface water being classified as priority pollutants (ATSDR 2021).
Despite the usefulness and applications of these chemicals, the deleterious ecotoxicological effects to a variety of organisms are highlighted. Some of the outcomes in the health assessment parameters observed in laboratory animals exposed to chlorophenols after oral administration were effects on the liver, central nervous system, body weight, immune system, and reproductive function (ATSDR 2021). The need to reduce the entry of toxic compounds into the environment and to study methods for their removal from contaminated sites remains relevant (Olaniran and Igbinosa 2011). Adsorption is a method with many advantages over other methods due to its low investment and operating cost, ease of operation, simplicity and flexibility in design, and efficacy and versatility in a variety of adsorption processes (Hadi et al. 2021). Adsorption on activated carbon, a nanoporous material that has low production cost and abundance of raw materials (Gryglewicz et al. 2002;Marsh and Rodriguez-Reinoso 2006), is a first choice for the removal of organic pollutants in water (Wang et al. 2020).
The adsorption on activated carbons has been thoroughly studied by molecular simulation method (Vishnyakov et al. 1999;Ravikovitch et al. 2000;Blanco et al. 2010;Oliveira et al. 2011;Lucena et al. 2017;Soares Maia et al. 2018). The main reference kernels for a good characterization is obtained using N 2 , Ar, or CO 2 as probe molecules and they are all based in this technique (Frenkel et al. 1996). The molecular simulation of super diluted solutions of organic compounds presents huge obstacles because we do not have consistent vapor pressure data to apply the Monte Carlo method in the grand canonical ensemble, and the low concentration of the solute impairs the use of molecular dynamics.
In a previous study (Galdino et al. 2021), our group applied molecular simulation associated with the representative pore methodology (de Oliveira et al. 2021) to analyze the adsorption of dilute aqueous phenol solutions. That study was based in an experimental and theoretical work from Kowalczyk et al. (2018). They found that only phenol adsorbs inside the activated carbon micropores in a diluted solution at 298 K. According the authors, the activated carbon has selective molecular sieving effects with respect to water until phenol saturation inside the carbon pores. Other neutral phenol-derived molecules may also present molecular sieving effect; thus, the calculation of the adsorption isotherm in each pore size could be done by molecular simulation with the fugacity of the gas phase corresponding to the fugacity of phenol in equilibrium with the diluted solution (Chempath et al. 2004). The selective molecular sieving effect is based on hydrophobicity theories (Chandler 2005). Hydrophobicity is a characteristic of activated carbons. Kowalczyk et al. stated that the transfer of water molecules from aqueous solutions to hydrophobic micropores is thermodynamically unfavorable.
Our study aims to extend the technique to two neutral phenol derivatives: o-chlorophenol and p-chlorophenol. These phenolic species were chosen because, in addition to being an important class of pollutants, their experimental isotherms present an interesting adsorption delay (step), with o-chlorophenol (OCP) adsorbing more than p-chlorophenol (PCP) in a specific range (from 2 to 6 × 10 −4 mol/L) of the low concentration values in the aqueous solution (Gryglewicz et al. 2002). A similar behavior had been identified by Caturla et al. 1988 between the fully neutral and partially ionized 2,4-dinitrophenol (DNP) molecule. The authors suggest that the orientation of the molecules and the mixture of neutral and ionized species may be the cause of the step. The step was also observed for some carbons, for the o-chlorophenol molecule, in the Aktaş and Çeçen 2007 study.
In the literature (Ahmaruzzaman 2008;Garba et al. 2019), the step is not always present; therefore, it could be an experimental artifact. Another difficulty found in identifying more examples of steps is the wide dispersion in the methods (phenol concentration range, types of phenol molecules studied, carbon concentration in the medium, pH) and the lack of adequate characterization of the carbons used. Thus, our study has a double objective: (1) to verify the possibility of predicting the adsorbed amount using the methodology of representative pores and (2) to analyze the possibility of the existence of the step. Evidence of the non-existence of step by molecular simulation would reinforce the hypothesis of an experimental artifact.
Here, we present the predicted individual adsorption isotherms of o-and p-chlorophenol for selected representative pores (Lucena et al. 2013;Aguiar et al. 2016;Gonçalves et al. 2018), in the micropores and mesopores range sizes. We compared our simulation with experimental data, and identified the origin of the uptake difference between o-and p-chlorophenol. Finally, we estimated the adsorption capacity of the studied chlorophenols on a set of others activated carbons, carefully characterized for the pore size distribution.

Methods
Among the models proposed in the literature for activated carbon (Do and Do 2006;Lucena et al. 2008Lucena et al. , 2017Nguyen et al. 2008;Oliveira et al. 2011), we choose the most applied model. The molecular representation of activated carbon is then defined as slits of graphene layers walls. The simulation box is built with graphene walls of 40 × 40 Å (Fig. 1c). For the chlorophenols molecules, we implement the atom-atom model from Jorgensen et al. (1996) and Martins et al. (2015) ( Fig. 1a and b), for a better description of the interaction for aromatic ring, chlorine atom, and hydroxy group.
The Monte Carlo algorithm in grand canonical ensemble (GCMC) (Allen and Tildesley, 1987;Frenkel et al. 1996) was applied in the calculation of the adsorption isotherms of o-chlorophenol and p-chlorophenol in the selected pores. The molecules of both isomers were optimized before the simulations using DFT (LDA/PWC functional) as implemented in DMol3 code from BIOVIA materials studio (Fig. S1). The GCMC method is one of the most used in the adsorption studies and allows a direct calculation of phase equilibrium (Nicholson and Parsonage 1982). The interactions among the molecules were computed using the Lennard-Jones (LJ) potential: where ε gg and σ gg are the energetic and geometric parameters of LJ potential, respectively, and r is the distance between the particles. The electrostatic interactions were also computed. Table 1 shows all the parameters used in the simulation.
The kernel of chlorophenol isotherms were calculated at 298 K for three center-to-center pore size pore sizes (H cc = 8.9, 18.5, and 27.9 Å). The pore size distribution (PSD) of the carbons, except for the AGL3 carbon, was (1) U gg (r) = −4 gg gg r 6 − gg r 12 determined using an entire set of N 2 isotherms. After, representative pore volumes are assigned. This methodology of representative pores was used and validated in previous study reported in literature (Aguiar et al. 2016;Gonçalves et al. 2018). The integral equation of adsorption isotherm for PSD can be written as: where Q(P) is the total adsorbed amount per gram of adsorbent at pressure P (experimental isotherm), q (P, H) is a function that represents the adsorption isotherm for a material characterized by pores with size H, and f(H) is a PSD. The Monte Carlo steps used were 2·10 6 for equilibrium and 1·10 6 for production. The value of cutoff (r cutoff ) was 12.5 Å. For a cutoff above 12.5 Å, the energy values no longer change and there is an increase in the computational cost; thus, we decided to keep the value of 12.5 Å, also used by other authors in similar systems (Yang and Zhong 2006;Bae et al. 2008;Lucena et al. 2013). The PSD is obtained through deconvolution of Eq. 2 using the experimental isotherm and the kernel of simulated isotherms (Lucena et al. 2010).

Chlorophenol adsorption in individual pores
In Fig. 2 (see also Tables S1 and S2), we show the representative pore adsorption isotherms for both ortho-chlorophenol (OCP) and para-chlorophenol (PCP) that will be used in the simulated isotherm kernel. We observed that the pore filling occurs abruptly and at low concentrations for the pores of 8.9 and 18.5 Å. In the larger pore of 27.9 Å, the filling occurs more smoothly.
Adsorption discrepancies between the two species of chlorophenols already appear in the 18.5 Å pore with a small delay in pore filling for the PCP molecule. The difference in uptake is amplified considerably for the 27.9 Å pore. The difference in the filling of chlorophenols for the 27.9 Å pore may be the reason for the existence of the step in some experimental studies. To investigate why this discrepancy in the adsorption isotherms occurs for the 27.9 Å pore, we perform a calculation of the angular distribution of  the dipole moment of the chlorophenol molecules (Fig. S2) and the interaction energy distribution in the pore for similar loading (Fig. 3). The results show that OCP has a preference for angles close to 90° which means that the molecule is parallel to the carbon surface in the face-to-face conformation. This conformation has two consequences: it maximizes the fluid-fluid interaction allowing for faster filling from the pore to the same time that it optimizes the packaging of molecules. For the same loading, the PCP has a smaller number of molecules in the face-to-face position (less than the 90° angle); this distribution of configurations reduces both the molecule-molecule interaction and the packing. The interaction energy distribution of the OCP molecules in the 27.9 Å pore is higher than that for PCP molecules, also contributing to the pore filling discrepancy. These two factors result in an uptake decrease for PCP and the consequent delay of pore filling.

Predicting chlorophenol adsorption in carbons
As previously mentioned, the experimental isotherm of Gryglewicz et al. (2002) will be our reference isotherm. This isotherm was chosen as a reference because in addition to presenting the step, the activated carbon is well characterized in terms of pore size distribution. In this study, the authors evaluated the adsorption of chlorophenols on AGL3 activated carbon, highlighting the differences in the adsorption regime for OCP and PCP molecules. In Fig. 4, we present the pore size distribution (PSD) for the AGL3 carbon, obtained in the experimental study. This distribution was used to allow the evaluation of the respective volumes corresponding to each of the representative pores of 8.9, 18.5, and 27.9 Å. We emphasize that the representative pore methodology extracts the pore sizes distribution from a detailed N 2 isotherm at 77 K of the material. As we do not have a detailed N 2 isotherm for carbon AGL3, the volumes were determined only from the PSD presented by Gryglewicz et al. (2002). Knowing the volume corresponding to each pore, together with the pore isotherms presented in Fig. 2, and using Eq. 2, we can estimate the adsorption isotherms of chlorophenols for the entire activated carbon (Fig. 5).
In our previous study with phenol (Galdino et al. 2021), it was possible to correlate the pore pressure with the vapor pressure values estimated by Henry's law constant.
where C i,L and P i,G are the concentrations of the solute in the liquid and in the gas phases, respectively, and H pc is the Henry's law constant (HLC). Unfortunately, for chlorophenols, it was not possible to identify an estimate of H pc obtained by the liquid-liquid chromatographic method, with underlying phenomenology similar to the adsorption on activated carbons. In addition, the literature presents ranges of values with variations of the order of magnitude of 10 2 , as evidenced by Sander (2015). From the reconstituted isotherms, we estimated the H pc values to be 33 and 16 kPa L/mol for the OCP and PCP molecules, respectively. c) d) Figure 6 shows the experimental (Gryglewicz et al. 2002) and simulated isotherms with pressures converted to concentration from the estimated H pc values.

Fig. 4
Pore size distribution of AC AGL3 at 77 K (Gryglewicz et al. 2002) and the representative's pore volumes of 8.9 Å, 18.5 Å, and 27.9 Å. Red color-total volume of the 8.9 Å pore. Green colortotal volume of the 18.5 Å pore. Blue color-total volume of the 27.9 Å pore We obtained a good agreement between the experimental and simulated isotherms in the pressure above 2 × 10 −4 mmol/L. The maximum adsorption capacity obtained by simulation is compatible with that determined experimentally and others found in literature (Table S3). The smaller uptake for the PCP molecule was qualitatively reproduced. The disagreement between the predicted and the experimental adsorption for low concentrations (below 2 × 10 −4 mmol/L), which did not identify a delay in the adsorption of the phenols, is a consequence of both our model and the experimental procedure. Our model assumes that the pores are filled simultaneously which does not happen in real carbon; the connectivity of the pores is not taken into account. Our model also does not take into account the adsorption on the carbon surface that occurs in the first instants of contact of the solution. On the experimental side, measurements at low concentrations are always more imprecise particularly in the finite bath method with very diluted solutions. This value is obtained from the difference between the concentration of the initial solution and the concentration after adsorption, which is extremely sensitive to the analytical method used. The contact time for equilibrium to be reached is another factor that can impact the experimental measurement at low concentrations. For example, in Carluta et al.'s (1988) study that identified steps in the DNP molecule, there is a clear adsorption difference in the DNP molecule even for the lowest concentration values. To capture this difference, times of up to 100 h were used.
To demonstrate the versatility of the representative pore methodology, we estimate the adsorption isotherm for chlorophenol at 298 K on the commercial carbons WV1050, Maxsorb, NORIT-RB4 (Fig. 7) and the PC series of carbon (Fig. 7) that are produced from PET polymer and are neutral and without heteroatom species (Ania et al. 2007). The volumes of representative pores of each carbon are presented in Table 2.
Among the commercial carbons (Fig. 7), Maxsorb had a higher adsorption capacity, thanks to the volumes in the different regions of porosity, which are higher than the other carbons. Norit-RB4 carbon, despite having a greater volume of micropores than WV1050, adsorbs much less, highlighting the importance of pore volume in the range of 27.9 Å ( Table 2). The analysis of representative pores has the advantage of clarifying the contribution of each pore size to adsorption. This can be seen in the series of activated carbons shown in Fig. 8. The development of mesoporosity favors adsorption for both molecules and widens the adsorption difference between them. The four carbons have very similar pore volumes in the 8.9 Å range; however, the pore volumes of 18.5 Å and 27.9 Å for the samples PC35, PC58, and PC76 grow with the longer burn-off time (Table 2), therefore presenting a more developed mesopore volume. The maximum adsorbed amount of OCP grows from only 200 mg/g in sample PC12 to 1200 mg/g in sample PC76.
The pore size influence in the differences of amount adsorbed between chlorophenol molecules can be also examined in the carbon series. The carbons PC12 and PC35, with low mesoporosity, did not present maximum uptake differences for o-and p-chlorophenol. This interesting adsorptionstructure dependence can be used in separation process of the two species. These results represent an important step in obtaining predictive models for adsorption of phenol derivatives.

Conclusions
A predictive model of adsorption of phenol derivatives OCP and PCP was developed based on molecular simulation and the methodology of representative pores. The analysis allowed both the study of individual pore performance and the reproduction of isotherms on real carbons based on the N 2 PSD at 77 K. The prediction of the OCP and PCP isotherms quantitatively reproduced the maximum experimental adsorption capacity and described the difference in adsorption between the species. Experimental and model limitations resulted in discrepancies between experimental and simulated values at low concentrations. The simulation results suggested that the adsorption difference between the two chlorophenols is basically due to the configuration and interaction energy differences between the molecules, identified most intensely in the 27.9 Å pore. OCP molecules assume a face-to-face configuration more frequently than PCP molecules. Face-to-face configuration increases fluid-fluid interaction and optimizes molecular packing. Among the carbons tested, and considering the model limitations, those that were more favorable to the adsorption  of o-chlorophenol and p-chlorophenol were Maxsorb and PC76, as they have a more developed mesoporosity.
In addition to supporting carbon selection, we believe that methodology applied to phenols can be extended to other similar families of organic pollutants that are preferentially adsorbed by activated carbon.