Biosorption of pharmaceuticals from aqueous medium by Luffa Cylindrica Fibres: Application of the linear form of Redlich-Peterson isotherm equation

This study is focused on the removal of Dextropropoxyphene (DPP) and Paracetamol (PAR) from aqueous solutions by sorption on Luffa Cylindrica fibres, as a low-cost biosorbent, that was initially characterized using BET, FTIR spectroscopy, and SEM analysis. The sorption study has been realized by the batch method with the effect of the biosorbent amount, initial concentration, solution pH and batch temperature. The modelling of the sorption phenomenon was based on the mathematical approach of the modified Redlich-Peterson isotherm equation (RP), where the dimensionless form of this isotherm, corresponding to the optimal curve, allows the α parameter evaluation. The same value of α was obtained for both pharmaceutical compounds, while the 𝑏 𝑅𝑃 𝐶 𝑒𝛼 values where 2 and 10 for DPP and PAR, respectively. The linear regression of the Redlich-Peterson isotherm equation was also confirmed by analysis of variance (ANOVA). The obtained results show that p-value is less than 0.05, with correlation coefficients (R adj2 ) equal to 0.9515 and 0.9283 for DPP and PAR, respectively. The kinetics modelling shows that the sorption mechanism obeys the pseudo-second-order and intraparticular diffusion. The adsorption process is exothermic, spontaneous and the molecules of the two pharmaceutical products have a random behaviour on the Luffa Cylindrica active sites.


Introduction
In the last dictate, the removal of chemical, biochemical, and biological pollutants (heavy metals, dyes, and pharmaceutical products…) from natural aquatic mediums became a necessary operation to save the environment. The great dispersal of these hazards related 3 directly to the fast growth of the manufacturing activities in our cities [1,2] . Pharmaceutical pollution becomes an inevitable environmental problem of emerging concern; these kinds of compounds are persistent and they have incompletely degraded in sewage treatment stations [3][4][5] due to their resistance to biodegradation [6] and persistence in aquatic ecosystems. In addition, the presence of these products in aquatic media can produce serious damage to different microorganisms [7].
The consumption of pharmaceuticals is increasing in a dreadful way, which poses a great menace to life on earth [8] . Animals and humans in different levels of transformation emit these products [9] . Many of these pharmaceuticals degrade in nature and are transformed into byproducts resulting from direct/indirect photolysis or hydrolysis [10] . In fact, some of them can be removed by adsorption techniques, under random environmental conditions. In another way, the physicochemical parameters and the mechanisms that serve for the best natural degradation are not controlled. However, many compounds persist in aqueous systems around the world [11] . The presence of these chemicals in surface water can be explained by the incomplete removal of these hazards during treatment processes of wastewater [12] .
Unfortunately, there is a lack of research on the identification of by-products such as metabolites and products of transformation from pharmaceutical elimination and degradation that persist in all types of water [13,14] .
Since 2001, the environmental risk assessment has been done first by determining an environmental predicted concentration (PEC) at the surface water level [15] . Below 0.01 μg.L -1 , the considered molecule does not present any environmental risk under normal conditions of prescription and use. However, if the 'PEC' exceeds this limit value, the requisition of deeper data on the physicochemical, pharmacological, and toxicological properties with a study of degradability, persistence or bioaccumulation capacity of the molecule is envisaged. The limit values corresponding to predicted environmental concentration (PEC) are for the soil of 100 4 μg.kg -1 . This threshold is directly related to veterinary drugs. In the case of the marine environment, the convention 92 considers about twenty molecules for pharmaceutical use as the most dangerous for the marine environment, such as clotrimazole.
Dextropropoxyphene (DPP) and Paracetamol (PAR) are considered among the pharmaceuticals widely used. However, the occurrence of these drugs in water sources with high amount can be dangerous to health. Recent studies revealed that the (DPP) has been detected with important concentrations in surface waters. In the estuary water, Roberts and Thomas (2005) recovered the DPP at concentrations varied between 0.008 μg.L -1 and 0.033 μg.L -1 . The measured data in the sediments are almost non-existent, in particular, the studied molecule.
PAR is antipyretic and analgesic. It seems to inhibit cyclo-oxygenase (Cox1-Cox2) in the nervous system. It is frequently found in wastewater and its concentration at the inlet in different treatment plants is measured at 6 μg.L -1 [16] . The maximum concentration of PAR in surface waters is about 0.11 μg.L -1 [17] . Although it is a highly degradable compound, its highest concentration in seawater reached 250 μg.L -1 [18] . Protection of aquatic ecosystems and different water sources from the undesirable and toxic effects of pharmaceuticals and their transformation products has been the subject of several types of research works . The chemical [19][20][21][22][23] , biological [24] , electrochemical [25] , and bioelectrochemical [26] methods have proved their limitation towards the treatment of wastewater.
That is why the adsorption remains the most preferred technique compared to the other processes [19] . The principal advantages of an adsorption phenomenon are a short-term investment of both cost and land, and it remains a simple and very efficient process in the elimination of toxic and harmful substances [27,28] . The pharmaceuticals removal using the activated carbon as an adsorbent is a promising process. This material is characterized by high porosity, an important surface area, and it has a good capacity in the elimination of toxic 5 pollutant in a short time; it can be found as granular or powdery black material [29] , but it is too expensive [30] . The selection of highly efficient adsorbent is primordial for a successful adsorption process, essentially in the pharmaceutical elimination. However, the bio-sorption using natural adsorbents such as natural clays, marine algae, and plants fibres are among the most practical, green, and economical techniques used in pollutants removal. This study is undertaken to evaluate the use of Luffa Cylindrica (LC) for (DPP) and (PAR) sorptive removal from the aqueous medium at different temperature and pH values. The model analysis of the batch adsorption system is given by a graphical representation of the Luffa cylindrica residual amount in aqueous solution. This application is generally based on choosing an empirical relationship between experimental data and relevant dimensionless parameters using regression techniques [31] . Linear regression has been successfully employed to investigate the parameters of sorption isotherms [32] . Modelling of some adsorption system, by plotting solid-phase amount against the concentration of residual liquid using the two and three parameters of isotherm (Langmuir, Freundlich, and Redlich-Peterson) have shown some of the inaccuracies. The poor precision of these isotherms is due essentially to a poor linear fitting model. Feng-Chin-Wu has proposed a novel linear exponential form of Redlich-Peterson (RP) by estimating the three parameters isotherm using the characteristic line obtained from the dimensionless form of Redlich-Peterson equation [33] . To our knowledge, this study is first to its kind, which elucidates adsorptive properties of (LC) to removal pharmaceutical products from aqueous medium.
The aim of this work is the application of Luffa cylindrica (LC) in DPP and PAR removal from an aqueous medium in a batch system, firstly and the correlation of our experimental results using a mathematical approach. This approach is based on the linear form of Redlich-Peterson isotherm equation proposed by Feng, secondly. The exponential form has shown its superiority for fitting Redlich-Peterson isotherm to experimental data obtained from sorption 6 processes compared to the logarithmic form used currently by the selection of a suitable α value range, which was determined from the dimensionless form plot of Redlich-Peterson equation. Analysis of variance (ANOVA) was also employed to investigate the adequacy of the fitted model by determining the adjusted correlation coefficients (R 2 adj) and mean square of squares (SS). Statistical significance is judged by the Fisher value (F-value) and the probability value (p-value) which must be less than 0.05.

Reagents
The biosorbent Luffa Cylindrica (LC) was purchased from a local shop in Skikda, Algeria.
Only, the LC fibres were used.

Adsorbent pre-treatment and Characterization
100g of luffa fibres were soaked in 1L of hydrogen peroxide for a period of two days to remove adhering dirt. Thereafter, fibres were washed with distilled water and dried in an oven at 80 °C for 12 h. After, they were dipped in NaOH (0.1N) solution for about 1h to increase their hydrophilicity. Alkalized fibres were finally washed with distilled water and dried again at 80 °C for 12 h.
The Fourier transform infrared ray (FTIR) spectrum of the treated fibres was performed using 7 a Shimadzu Spectrometer (FTIR 8700, Japan) via KBr pellets method. The spectra were collected in the range of 400-4000 cm -1 with a spectral resolution of 4cm -1 . Their surface morphology was analysed using scanning electron microscopy (SEM) (JEOL, JSM-7600F, Japan). Their porosity was determined by nitrogen adsorption-desorption isotherms measured at 77 K using Cooltronic micro-meretics 2100E model surface area analyser. The N2 isotherms were used to calculate the specific surface area using the BET equation.

pH of zero point charge (pHPZC)
The point of zero charges (pHpzc) was carried out to determine the pH value for which the surface net charges of the Luffa cylindrica are zero. To quantify the pH at the point of zero charges, 0.1 g of LC has been added to 50 mL of distilled water, whose initial pH has been adjusted with NaOH or HCl in the range of 2 to 12. The containers are sealed and placed in a shaker for 24 h, after which the pH is measured. The PZC occurs when there is no change in the pH after contact with the carbon. The value of (pHPZC) is illustrated in Fig 1. The study of parameters affecting the biosorption process of DPP and PAR solutions onto LC allowed us to determine the optimal value of each parameter, which accords to the highest elimination rate. Thus, this study is done by changing the values of a studied parameter and fixing the other parameters.

Effect of different parameters of adsorption processes of DPP and PAR onto LC
The optimisation of the operating conditions has been studied by varying one parameter and The DPP and PAR removal efficiencies (R%) have been determined at various experimental conditions. The R% was calculated according to the following equation: where C0 and Ct are respectively the concentration (mg L -1 ) of adsorbat at initial and t timerespectively. Equation 2 was used to estimate the adsorption amount (qe) of DPP and PAR as shown in the following formula: In which Ce is the equilibrium concentration of pharmaceutical compounds in aqueous media (mg/L), W the weight of the biosorbent (g) and V the solution volume (L).

Adsorption isotherms
The Redlich-Peterson (RP) isotherm model is more accurate than the Freundlich and Langmuir equations in describing the sorption system [33][34] . The RP isotherm equation is generally used to explain the formation of monolayer and multi-sites interactions phenomena at the same time [35] . It also describes homogeneous adsorption mechanism systems [36] . In addition, it integrates the isotherms of Freundlich and Langmuir into a single equation [37] .
The non-linear Redlich-Peterson isotherm equation is presented as: where bRP and qmon are the parameters of Redlich-Peterson isotherm equation. The exponent, α, as it ranges from zero to 1, it has two limiting behaviors: Henry's law form for α equal to 0 and Langmuir forms for α equal to 1. At low concentrations, it is also similar to Henry isotherm and performs like Freundlich equation for high concentrations [38] . The linear, nonlogarithmic, form of equation (3) is given by Eq.4: However, the dimensionless of RP equation form proposed by Feng is expressed as follows: Here, Cref is the adsorption system highest equilibrium concentration and qref is the equilibrium adsorption quantity at Cref. α is a parameter attained by trial and error for = 2 and 10, respectively [33] . Its value must be less than 1. The plot of qe/qref against Ce/Cref of Eq. (5)  test, which allows the model to be evaluated [39] . If the p-value is inferior than 0.05, the studied model is significant. The low mean square sum of squares (SS) is also evident that the model is apparently the best.

Characterization of the biomaterial
The fibres of (LC) are composed of 60 % cellulose, hemicelluloses 30% and 10% lignin a fact that makes it flexible and durable biosorbent. Scanning electron microscopy (SEM) analysis of (LC) brought us the presence of macropores in the structure of (LC) fiber. using BET method, is 123 m 2 g -1 .

Fig 2. Scanning electron microscopy of Luffa Cylindrica (LC)
The point of zero charge (pzc) of LC and modified LC is about 5.59 and 8.60, respectively.
This former value may be attributed to the alkaline treatment. Consequently, the surface of modified LC is predominantly positively charged below 8.60 and negatively charged above this value. As a consequence, adsorption of compounds could depend from pH.

Fig 3. FTIR spectrum of alkalized Luffa Cylindrica (LC) biosorbent
The weak band at 1388.60 cm -1 is assigned to C-H, the shaper band at 1404.08 cm -1 corresponds to CH2 and CH3. The band located at 1515.93 cm -1 is due to benzene ring stretching (lignin) [30] . The peaks ranging between 1000 and 1300 cm -1 correspond to C-O (ketones), specially the clear peak at close to 1058.92 cm -1 represented C-O (cellulose) [40] and that's observed at 1176.50 cm -1 may correspond to C-O-C (cellulose and hemicellulose) [30] .
Finally, the carbohydrate (C-C) peaks are located between 813.90 and 941.20 cm -1 .

Effect of initial pH
It is well known that the sorptive uptake of the organic and inorganic pollutants can be significantly influenced by the initial pH of the aqueous medium. Indeed, the concentration of hydroxyl and hydrogen ions affects the adsorption process through the dissociation of functional groups on the adsorbent surface and/or the ionization degree of the adsorbed molecules [30] . The variations of DPP and PAR removal by LC at various pH values (2.0-12.0) were depicted in Fig 4a. It demonstrates that at all experimental pH values, except pH 10, the 13 adsorption process is more efficient with DPP than PAR. Moreover, the percentage of removal seems to exhibit dissimilar behaviour. Actually, the percentage removal of PAR slightly decreased from 81% to 66% with increasing pH from 2.0 to 10.0 and increased when the pH value was raised above 10.0 (72% at pH 12.0). Herein, it should be noted that the LC surface is positively or negatively charged when the solution pH is above or below the pHpzc and anions will be surrounded by water molecules and will not be significantly adsorbed.

Effect of LC masse
Regarding the used dose of LC, which range from 0.05 to 0.4 g, we note that 0.1 g is sufficient to eliminate 53.52% of DPP and 95.55% of PAR (Fig.4b). With this amount, the greater part of the molecules of each pharmaceutical can take place on sorbent sites.

Effect of pharmaceutical concentration
About the initial concentration for each pharmaceutical, the highest rate of DPP adsorption (54.29%) is attributed to 50 mg L -1 (Fig.4c). While, the optimal initial concentration of Paracetamol, appropriate for the highest elimination rate (R =70.70%), is equal to 20 mg.L -1 .

Effect Temperature
The media temperature effects on sorption onto (LC) is remarkable for the temperature 20°C and the adsorption efficiency (R %) is equal to 56.31% (Fig.4d), whereas for the optimal temperature for the highest elimination of PAR (R = 95.27 %) equals to 40 °C. 15 The obtained results from the application of the new mathematical approach of RP isotherm equations are depicted in Fig 5 and   The retained results are collected in Table 2, where bRP has become KL and qmon is taken qm.

Adsorption Isotherms
These lines proved to be linear over the majority of the experimental points of the concentration range studied for DPP and PAR, with an extremely high value of R 2 adj for Par (R 2 adj = 0.9364). Tables 1 and 2 show that the biosorption model constants of DPP onto luffa cylindrica fibres can be well defined by RP equation since a higher adjusted linear regression correlation coefficient (0.95) was obtained for this model. Thus, the same constant of PAR adsorption isotherm onto (LC) was found equal to 0.93 for Langmuir model (obtained using the linear equation RP when α equals 1) (see Table2). The monolayer sorption capacity for PAR is considerably highest (qm = 97.0873 mg g -1 ). It can be given information that luffa cylindrical fibres have homogeneous surface energy [41] . In the present research, the ANOVA has been introduced to guess the goodness of the studied model fit. If the model data forms are more significant, the sum of square (SS) or mean square (MS) will give small values, P-value must be smaller than 5% and F-value will be higher.
From Table S2, we can conclude that the regression explains well the studied phenomenon since the meaning of the model risk is less than 0.05 (p < 1.8411x10 -5 for PAR and p < 5.7114  Table 1 showed that only 4.85 % and 7.45% of the total variation of DPP and PAR respectively could not be explained by adapted model.

Adsorption kinetic modeling
The nature of the adsorption process depends essentially on the chemical and physical properties of the adsorbent surface and the nature of sorbates. The obtained kinetic outcomes of DPP and PAR adsorption were correlated using various conventional methods, namely, the intraparticle diffusion, pseudo-first and pseudo-second-order kinetic methods, in order to study the mechanism and rate of DPP and PAR biosorption process onto Luffa. Table 3 collects the principal experimental adsorption data obtained using DPP and PAR initial concentrations 40 and 60 mg.L -1 , respectively, and 0.1g of adsorbent. In 1898, Lagergren proposed the kinetic model of pseudo-first-order [42] using an empiric equation (Table 3), which plots the ln (qeqt) in terms of the contact time. Previous several studies indicated that this equation is a fruitful model, in which it can provide a good linear relationship between k1 and qe, which may be confirmed from its slope and intercept.
An analysis of the experimental qe values shows that the distance of intercept is aptness and failed in describing the experimental outcomes. In addition, the adsorption rate does not obey to this equation. On the other hand, the adsorption kinetics is explained using the pseudosecond-order kinetic model [43] . This last produces a straight line and high values of R 2 . The adsorption equilibrium capacity (qe) and k2, have been extracted respectively from the intercept and the slope curve and the plot of t/qt versus t. The theoretical capacity of adsorption (qe), the correlation coefficient values (R 2 ) and the experimental closeness values demonstrate that the second order model explain successfully our experimental data (Table 3). 18 The correlation coefficient (R 2 ) values reached 0.9999 and a strong similarity between the calculated values of adsorption equilibrium capacity (qe) and the experimental values of adsorption capacity was obtained under different experimental conditions. Moreover, the process has been also studied using the intraparticle diffusion model [44] , which is based on adsorbate mass transfer diffusion, where, the sorption rate is related to the square root (t 1/2 ).
The Kdiff and C values were determined respectively from the slope and the intercept of qt versus t 1/2 plot. The C values related to the boundary layer thickness and the rate constant intraparticle diffusion Kdiff are collected in Table 3. Since the curve of the intraparticle diffusion model does not pass through the origin, the diffusion kinetic model and the secondorder model proceed controls the process of sorption mechanism.

Thermodynamic study
The thermodynamic parameters of DPP and PAR sorption such as Gibbs energy (ΔG°), enthalpy (ΔH°) and entropy (ΔS°) have been determined by the following equations: Where, R is the gas universal constant, Kd the thermodynamic equilibrium constant and T the absolute temperature (Kelvin). The thermodynamic study (Table 4 and Fig 6) shows that all ΔG° have negative values, which indicate a spontaneous and feasible process. In addition, the ΔG° values decreased with increases in temperature. Thus, the ΔH° negative values indicate that an exothermic sorption process, which could be accredited to a physisorption process mechanism [45] ,

Conclusion
This study demonstrates the suitability and the applicability of the novel dimensionless and exponential linear forms of RP isotherm proposed by Feng, which allowed us to determine easily the parameters of the RP model for DPP and PAR. α value was equal to 0.8 for each studied pharmaceutical, corresponding to an optimum line which gives the other parameters, mainly bRP and qmon. The RP equation involving adsorption on homogeneous active sites is definitely the most suitable modelling tool to describe satisfactorily the DPP biosorption and the Langmuir model showed strong adequacy to describe the adsorption experimental data of PAR onto (LC), by providing the highest adjusted squared correlation coefficients R 2 and the lowest p-value. The kinetic and thermodynamic studies were investigated for the adsorption process. The kinetics curves are successfully characterized by the pseudo-second-order rate equation and the intra-particle diffusion was not the rate-limiting step. The thermodynamic parameters confirmed that the process is exothermic, spontaneous and that the molecules of studied drugs have random behaviour on the surface of (LC) used in this work.

Acknowledgments
The authors gratefully acknowledge Laboratory of Innovative Techniques for the Environment Preservation, University of Mentouri Brothers -Constantine 1 for providing instrumental facilities for sample characterizations.

Competing interests
The authors declare that they have no competing interests

Authors' contributions
The manuscript was mainly based on a draft written by AA, and written through contributions of all authors. All authors read and and approved the final manuscript. Laboratory of water treatment and recovery of industrial waste, Annaba University, Algeria.