Single and multi-component adsorption of psychiatric pharmaceuticals onto alternative and commercial carbons

This work describes the adsorptive removal of three widely consumed psychiatric pharmaceuticals (carbamazepine, paroxetine and oxazepam) from ultrapure water. Two different adsorbents were used: a commercial activated carbon and a non-activated waste-based carbon (PS800-150-HCl), produced by pyrolysis of primary paper mill sludge. These adsorbents were used in single, binary and ternary batch experiments in order to determine the adsorption kinetics and equilibrium isotherms of the considered pharmaceuticals. For the three drugs and both carbons, the equilibrium was quickly attained (with maximum equilibrium times of 15 and 120 min for the waste-based and the commercial carbons, respectively) even in binary and ternary systems. Single component equilibrium data were adequately described by the Langmuir model, with the commercial carbon registering higher maximum adsorption capacities (between 272 ± 10 and 493 ± 12 μmol g-1) than PS800-150-HCl (between 64 ± 2 and 74 ± 1 μmol g-1). Multi-component equilibrium data were also best fitted by the single component Langmuir isotherm, followed by the Langmuir competitive model. Overall, competitive effects did not largely affect the performance of both adsorbents. Binary and ternary systems maintained fast kinetics, the individual maximum adsorption capacities were not lower than half of the single component systems and both carbons presented improved total adsorption capacities for multi-component solutions.


Introduction
The global occurrence of pharmaceuticals in the environment is now a reasonably welldocumented reality. The number of studies reporting their presence in surface, ground and even drinking waters has grown exponentially in the last few years (Alygizakis et al., 2016;Calisto and Esteves, 2009;Daughton and Ternes, 1999;Li, 2014;Loos et al., 2010;Luo et al., 2014), constituting a first indication that the water resources quality is actually being affected. The environmental pressure exerted by these contaminants is expected to raise in the next decades, mainly due to the worldwide ageing of the population, high prevalence of chronic diseases and easy accessibility of medical care (OECD, 2015). Within this scenario, and in order to mitigate the environmental contamination by pharmaceuticals, the main sources of discharges of these pollutants into the environment should be effectively addressed. In fact, the European Union has already recognized the urgency of solving water pollution by pharmaceuticals by evaluating new ways of reducing their input into environmental matrices, taking into consideration public health and cost-effectiveness, according to the Directive 2013/39 EU (European Parliament, 2013). Waste Water Treatment Plants (WWTPs) were projected to reduce pollutants according with legislated parameters but not to eliminate pharmaceuticals. In fact, there is a large number of studies reporting that the treatments applied in conventional WWTPs are not adequate to remove these pollutants (Bahlmann et al., 2014;Calisto et al., 2011a;Margot et al., 2013). As a result, the discharge of contaminated effluents from these facilities (both urban and industrial) is the primary pathway of pharmaceuticals into the environment (Alygizakis et al., 2016;Cardoso et al., 2014;Jelic et al., 2011). Thus, developing effective methodologies to be applied as WWTPs tertiary treatment might be a feasible path for the minimization of the environmental impact of pharmaceuticals.
The study of the application of alternative adsorbents for the removal of pharmaceuticals from water is scarcely addressed in the literature (specially, in comparison with the huge number of studies concerning inorganic pollutants or other organic pollutants such as textile dyes or pesticides). Additionally, to our best knowledge, there are only a few studies on the adsorption of pharmaceuticals in multi-solute systems (Jung et al., 2015;Mansouri et al., 2015;Nielsen and Bandosz, 2016;Sotelo et al., 2014), all considering ultra-pure water as the background matrix. In fact, most of studies focus on the single adsorption of drugs so obviating multi-solute competitive effects occurring in real systems, where the performance of the adsorbent might be significantly modified due to possible competition of the different adsorbates for the vacant adsorption sites (Limousin et al., 2007). Accordingly, multicomponent adsorption data are essential for the design of treatment systems (Noroozi and Sorial, 2013) and, thus, the scarcity of such data was the main motivation for the work here presented. This manuscript describes the single, binary and ternary adsorptive removal of three widely consumed psychiatric pharmaceuticals (carbamazepine, oxazepam and paroxetine) from ultrapure water, by a commercial activated carbon and an alternative waste-based carbon produced by pyrolysis of primary paper mill sludge with the aim of evaluating and comparing the performance of the studied carbons under competitive effects .   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65 5

Adsorbents
Two powdered carbons were used as adsorbents for the removal of the pharmaceuticals from water: a commercially available activated carbon (PBFG4, provided by ChemViron Carbon) and an alternative waste-based non-activated carbon (PS800-150-HCl). PS800-150-HCl was produced by the pyrolysis of paper mill sludge, under nitrogen controlled atmosphere at 800 ºC, with a residence time of 150 min. Detailed information concerning the pyrolysis process and the selection of the production conditions can be found in a previous work (Calisto et al., 2014). After pyrolysis, the carbon was washed with 1.2 M HCl followed by distilled water until the washing solution reaching neutral pH. The washing procedure was aimed at removing ashes and other inorganic matter, improving the microporosity and the surface area of the material by unblocking obstructed pores. Finally, the carbon was oven dried at 105 ºC for 24h.

Characterization of the adsorbents
PBFG4 and PS800-150-HCl were characterized by means of elemental analyses, total and inorganic carbon analysis, point of zero charge (PZC) and N 2 adsorption isotherms for the determination of surface area and microporosity. Detailed information concerning these procedures is given in SM. where C 0 (µmol L -1 ) is the initial concentration of pharmaceutical, C is the pharmaceutical concentration after shaking (µmol L -1 ), V is the volume of the solution (L) and m is the mass (g)

Adsorption experiments -single, binary and ternary systems
of the corresponding adsorbent.

Adsorption kinetics
The time needed for each system to attain the adsorption equilibrium was determined by shaking single, binary and ternary pharmaceutical solutions with the corresponding adsorbent for different time intervals (between 5 and 240 min). The adsorbent concentration in kinetic experiments was 0.15 g L -1 for PS800-150-HCl and 0.025 g L -1 for PBFG4.
After shaking, the concentration of pharmaceuticals remaining in the aqueous phase for each time interval was determined and the amount of pharmaceutical adsorbed by mass unit of adsorbent at that time (q t , µmol g -1 ) was calculated by equation 1. The experimental data were 7 fitted to a pseudo-first and pseudo-second order kinetic models according to equations 2 (Lagergren, 1898) and 3 (Ho and McKay, 1999), respectively.

2) )
3) with t (min) representing the adsorbent/solution contact time, q t (µmol g -1 ) the amount of pharmaceutical adsorbed by mass unit of adsorbent at time t, q e the amount of pharmaceutical adsorbed when the equilibrium is attained (µmol g -1 ); and k 1 (min -1 ) and k 2 (g µmol -1 min -1 ) the pseudo-first and pseudo-second order rate constant, respectively. The mathematical modelling was performed using GraphPad Prism 5.
Additionally, fittings of equilibrium binary experimental data to multi-component isotherm models were determined. The following models were considered: competitive where all the parameters are defined as in equations 5 and 6, with the exception of b (L µmol -1 ) which corresponds to an affinity adsorption constant related with the non-competitive adsorption;

8)
where all the parameters are defined as in equations 5 and 6, with the exception of b i and b j (L µmol -1 ) which correspond to adsorption constants of the affinity of the specie i for adsorption sites occupied by j and vise-versa, respectively.
The modelling of the bi-component experimental data to equations 6 to 8 was performed using the curve fitting toolbox of Matlab (version R2014a).
Briefly, the electrophoretic separation was performed at 25ºC, in direct polarity mode at 25 kV, during 5 min runs. Ethylvaniline was used as internal standard, spiked to all samples and standard solutions at a final concentration of 3.34 mg L -1 . Also, sodium tetraborate, at a final concentration of 10 mM, was added to all the samples and standard solutions, resulting in better peak shape, better resolution and enhanced repeatability. Detection was monitored at 200 nm for CBZ and PAR and at 214 for OXZ. Separation buffer consisted of 15 mM of sodium tetraborate and 30 mM of sodium dodecyl sulfate and was renewed every 6 runs. Capillary was washed between each run with ultrapure water for 1 min and separation buffer for 1.5 min at 20 psi.
Additionally, capillary was washed with separation buffer, for 20 min, at the beginning of each working day in order to replenish the dynamic coating and with ultrapure water, for 10 min, at the end of the day. All the analyses were performed in triplicate. Calibration was performed by analyzing standard solutions with concentrations ranging from 1 to 20 µmol L -1 . Also, prior to the adsorption experiments, the adequacy of the optimized MEKC methodology to quantify the concentration of pharmaceuticals in the aqueous phase was tested by checking possible matrix effects. Table 1 summarizes the physico-chemical characterization of PBFG4, PS800-150-HCl and also primary sludge (the raw material used for the production of PS800-150-HCl). The commercial carbon PBFG4 has physical and chemical properties typical of an activated carbon, being mostly composed by carbon (81%, determined by elemental analysis) with a negligible percentage of inorganic carbon, and with a specific surface area of 848 m 2 g -1 . In turn, the produced carbon, PS800-150-HCl, has a lower carbon percentage (67%) and its surface area is 414 m 2 g -1 , which is approximately half of the one determined for PBFG4. However, one should note that PS800-150-HCl was not physically or chemically activated, having been produced by 10 the pyrolysis of paper mill sludge only followed by an acid washing step. On the other hand, as may be seen in Table 1, the PZC of the two carbons did not differ significantly.

Adsorption kinetics
The amount of each pharmaceutical adsorbed onto PS800-150-HCl or PBFG4 (q t , µmol g -1 ) versus contact time (t), for single and multicomponent solutions, is represented in Figure 1.
The fitting parameters of the pseudo-first and pseudo-second order kinetic models to the experimental data are summarized in Table 2. In general, both pseudo-first and pseudo-second order kinetic models satisfactorily describe the experimental data for all the studied adsorbentadsorbate(s) systems, with the large majority of the correlation coefficients above 0.97.
Considering that the fitting of the pseudo-second order model presents the highest correlation coefficient for the largest number of systems, this discussion will be based on the kinetic parameters of this model; in addition, for clarity purposes, this is the only fitting represented in Figure 1 for each data set. The experimental data depicted in Figure 1 clearly show that the adsorption equilibrium of the studied pharmaceuticals onto PS800-150-HCl is quickly attained, with equilibrium times varying approximately from 5 to 15 minutes of contact time. On the other hand, the adsorption equilibrium onto PBFG4 is attained after 60 to 120 min. With respect to the three drugs here considered (in single systems), as for the k 2 shown in Table 2, the adsorption rate onto PS800-150-HCl is PAR>OXZ≈CBZ and onto PBFG4 it is OXZ>PAR≈CBZ.
Overall, PS800-150-HCl showed larger k 2 than PBFG4. Also, except for PAR onto PS800-150-HCl, no lower k 2 values were determined for the multi-component than for the single systems.

Adsorption equilibrium
3  Table 3. The 3-parameter Langmuir-Freundlich equilibrium model was also tested; the results were not considered in this discussion as it was the one that worst fitted the experimental data. According to R 2 and S yx values in Table 3, the Langmuir model is the one that fits more adequately most of the adsorption isotherms onto PS800-150-HCl, while the Freundlich model is the one that best fits most of the adsorption isotherms onto PBFG4.
However, considering that Langmuir fittings to PBFG4 experimental data are still very satisfactory, the Langmuir parameters will be selected for further discussion in order to allow a direct comparison of both carbons. In what concerns single component systems, and for both adsorbents, the Langmuir maximum adsorption capacities followed the order CBZ>OXZ>PAR, with the most adsorbed pharmaceuticals corresponding to the compounds with lower water solubility. However PBFG4 was the carbon with the highest adsorption capacities it also showed the most accentuated differences between the maximum adsorption capacities of the three pharmaceuticals, with q m varying from 272 ± 10 µmol g -1 for PAR to 493 ± 12 µmol g -1 for CBZ. In opposition, PS800-150-HCl displayed very similar maximum adsorption capacities 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65   12 for the three pharmaceuticals, varying only from 64 ± 2 µmol g -1 for PAR to 74 ± 1 µmol g -1 for CBZ. Regarding the Langmuir affinity coefficient, values obtained for the single isotherms onto PBFG4 were all lower than those onto PS800-150-HCl, which point to a higher affinity to the first.
A comparison between single and multi-component systems reveals that, in the case of PS800-150-HCl, the adsorption of CBZ and OXZ registered a decrease in the amount adsorbed ( Figure 4). CBZ suffered the most accentuated difference from single to ternary systems, with a reduction in the amount adsorbed of approximately 50%. On the other hand, PAR was not negatively affected by the presence of the other two pharmaceuticals, with equivalent maximum adsorption capacities between single and binary systems and even slightly higher in ternary systems. The analysis of the Langmuir affinity constants shows that, besides the case of CBZ with no significant differences between systems, OXZ and PAR revealed a clear tendency to  13 decline of the maximum adsorption capacities of up to 30% in binary systems composed of diclofenac and caffeine, in comparison with the single adsorption of the same pharmaceuticals.
More accentuated reductions were observed by Nielsen and Bandosz (2016), who registered a fall of up to 77% in the adsorption of sulfamethoxazole, carbamazepine and trimethoprim in the ternary system (in relation to single adsorption) and by Jung et al. (2015), who observed a reduction of approximately 80% in binary systems and 95% in ternary systems when studying the competitive adsorption of naproxen, diclofenac and ibuprofen.
In summary, the experimental data of binary and ternary mixtures were adequately described by single component isotherm models, allowing for comparison of the adsorption parameters when in the presence of more than one adsorbate. The reduced individual maximum adsorption capacities obtained for the majority of the pharmaceuticals in the mixtures must be related with competitive effects.   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65 15 coefficients. These analyses are based on the assumption that the maximum adsorption capacity determined for single component data by the Langmuir isotherm is the same for multicomponent systems (as this models implies adsorption onto homogeneous surfaces with no interactions between adsorbed species), suggesting that the Langmuir affinity coefficient is the only parameter to be modified in the presence of mixtures (Janaki et al., 2012;Limousin et al., 2007). However, this approach might be questionable considering that the presence of other adsorbates often result in a significant variation in the adsorption parameters (as it is clearly demonstrated in the data presented in Table 4) and might be responsible to introduce some artificiality in the data analysis.

Modelling of multi-component isotherms to binary mixtures data
There are some works showing that single component isotherms models successfully
-For both carbons, single maximum adsorption capacities of each pharmaceutical decreased in the multi-component systems, suggesting the existence of competition between these pharmaceuticals; however, the decrease in the individual maximum adsorption capacities between single and multi-components systems was not higher than 50% in any case.
-Cumulative maximum adsorption capacities for binary and ternary systems are consistently higher than the single maximum adsorption capacities, indicating an actual increase in the total amount of pharmaceuticals adsorbed by the carbons.