A novel study on the preferential attachment of chromophore and auxochrome groups in azo dye adsorption on different greenly synthesized magnetite nanoparticles: investigation of the influence of the mediating plant extract's acidity

In this paper, the adsorption of Evans blue (EB) and methyl orange (MO) azo dyes on four greenly synthesized magnetite nanoparticles has been studied to investigate the effect of the mediating plant extract's acidity on magnetite surface reactivity in azo dye adsorption. Magnetite surface reactivity has been studied through the analysis of preferential attachment of dye chromophore and auxochrome groups on magnetite nanoparticles, and adsorption yields. According to the contents of chromophore and auxochrome groups in dye structures, the mediating plant extract's acidity effect on acid site types and densities was also deduced. Used plants for the green synthesis were: Artemisia herba-alba (L), Matricaria pubescens (L), Juniperus phoenicea (L), and Rosmarinus officinalis (L), and their extract pHs were respectively 5.25, 5.05, 4.63, and 3.69. The four greenly synthesized samples of magnetite were characterized by XRD, SEM, ATR-FTIR, and UV-Vis techniques. The novelty of this paper lies in highlighting the influence of the mediating plant extract's acidity on the greenly synthesized magnetite surface reactivity towards the preferential attachment of chromophore and auxochrome functional groups in azo dye adsorption, where obtained results show that the mediating plant extract's acidity has a clear effect on the preferential attachment of chromophore and auxochrome groups on magnetite surfaces as well as on azo dyes' adsorption yields and capacities. Indeed, the decrease in the plant extract's acidity leads to an increase in the attachment of chromophore groups and a decrease in the attachment of auxochrome groups. So, it leads to an increase in Lewis acid site density and a decrease in Brønsted acid site density of magnetite surfaces. Also, the decrease of the plant extract's acidity leads to an increase in the studied dye adsorption yields, and this is because the majority of functional groups of MO and EB dyes are chromophores that attach to Lewis acid sites. The difference found in adsorption yields of EB and MO on all four magnetite samples is due to the fact that the ratio of chromophore/auxochrome groups in EB is remarkably greater than that in MO. The linear and non-linear pseudo-first-order and pseudo-second-order kinetics of the adsorption as well as the intra-particle diffusion mechanism have been analyzed. Obtained results indicate that in all adsorption processes the adsorption kinetics followed a linear pseudo-first-order kinetic model, and film diffusion is the step that controlled adsorption mechanisms. The thermodynamic studies of EB and MO adsorption processes on the four magnetite surfaces have been analyzed in the temperature range of 303.15–318.15 K. Obtained results reveal the endothermic nature of the adsorption in all cases.


Introduction
Dyes are present in the effluent water of several industries, including textile, leather, paper, rubber, plastics, printing, cosmetics, pharmaceutical, and food industries. Dyes contribute to water toxicity and represent an increasing danger for the environment, humans, and animals. [1][2][3] They are generally resistant to light, oxidizing agents, and many chemicals and therefore difficult to degrade once released into aquatic systems. Thus, one of the major environmental problems related to the numerous industrial applications of dyes is their removal from effluents. 3 Dyes are composed of chromophores that are commonly electron withdrawing and auxochromes that are usually electron-releasing groups. 4 The most important chromophores, as dened in this way, are: N]N, C]O, -CH]N, NO 2 , NO, NOH, C]N, C^N, C]C, and C^C groups and ionizing auxochromes mainly include: SO 3 H, OH, COOH, NH 2 , NH 3 , NHCH 3 , and N(CH 3 ) 2 groups. 4,5 Adsorption is one of the most effective processes for the removal and recovery of colored materials and dyes from effluents. [6][7][8][9] Nanomaterials are widely used in the purication of aqueous media due to their advantages, such as high surface area and increased number of active sites. [10][11][12][13] They therefore allow a rapid thermodynamic equilibrium between adsorbent and adsorbate during the adsorption process and selective removal of pollutants. [7][8][9] Several factors inuence the adsorption process, mainly the solution chemistry, 7,14 the characteristics of the dye (adsorbate) 7,15 and the adsorbent surface properties. [16][17][18][19][20][21][22] Saha et al. 7 studied the preferential adsorption of seven different dyes on magnetite NPs. They reported that the magnetite surface preferred adsorbing dyes containing higher OH content. Xiao et al. 15 studied the preferential adsorption of different cationic and anionic dyes on iron nanoparticles. They reported that iron NPs preferred removing cationic dyes more than anionic dyes.
Other authors studied the effect of changing the adsorbent surface properties by binding ligands on the adsorbent surface. Khurshid et al. 17 found that the use of amine-functionalized cobalt-iron NP surface enhanced the removal of anionic azo dyes. Mahmoodi et al. 22 synthesized a titania/silica nano-hybrid (TSNH) and an amine-functionalized titania/silica nano-hybrid (AFTSNH) to use them in Reactive Red 198 and Acid Red 14 removal from wastewater. They found that the AFTSNH adsorbent showed high dye adsorption capacities compared to the TSNH adsorbent. The authors of ref. 23 prepared silica nanoparticles (SN) and amine-functionalized silica nanoparticles (AFSN) and then used them in Acid Red 14, Acid Black 1 and Acid Blue 25 removal. They reported that AFSN preferred adsorbing the studied dyes than SN. Meanwhile, the study of Madrakian et al. 24 reported that magnetite-modied activated carbon preferred adsorbing cationic dyes than anionic dyes. Moreover, a comparative study on adsorption of methylene blue on sericin-modied and unmodied magnetite NPs 25 reported that sericin-modied magnetite NPs were approximately 40% more effective than the unmodied magnetite NPs. Furthermore, in the study of the preferential adsorption of magnetite NP loaded g leaves (MNLFL) and magnetite NP loaded azolla (MNLA) to remove crystal violet and methylene blue, 26 the authors found that MNLFL preferred adsorbing crystal violet more than MNLA.
Surface acid site types and densities were also found to impact the adsorption process. Indeed, a study of the adsorption of organic contaminants on both natural and synthesized magnetite 20 found that the adsorption on natural magnetite was more efficient than that on the synthesized one, and this is because of its higher surface site density. Moreover, Gogoi et al. 27 studied the degradation of catechol using an Fe 3 O 4 -CeO 2 nanocomposite as a Fenton-like heterogeneous catalyst. They reported that the increase of Brønsted acid site density of this nanocomposite increased the degradation of catechol.
The impact of changing the mediating plants on the behavior of greenly synthesized metal oxide NPs in dye adsorption was studied in several works. The effect of three different tea extracts on the capacity of greenly synthesized iron oxide surfaces in the removal of methyl green dye from aqueous solution has been studied by Huang et al. 16 They reported that the plant extract had an effect on the adsorption yields of methyl green dye on the three iron oxide NPs, where yields varied from 81.2% to 75.6 to 67.1%. Duyen et al. 28 synthesized metal oxide NPs using the extracts of owers, bark, and leaf of Tecoma stans in order to use them in the removal of Congo red (CR) and crystal violet (CV) dyes. They reported that the adsorbent derived from ower extract gave better adsorption efficiency than those derived from other extracts. Islam et al. 29 synthesized magnetite NPs using six plant extracts in order to use them in the removal of methyl orange (MO) and crystal violet (CV) dyes. They reported that plant extract had an effect on magnetite NP surface reactivity in the adsorption, where magnetite NPs synthesized using tea extract showed the highest performance (MO 92.34%, CV 96.1%). Ahmouda et al. 30 used different greenly synthesized magnetite NPs in the removal of methyl green (MG) dye via the adsorption process. They reported that the mediating plant extract's acidity had an effect on the preferential adsorption of MG on the magnetite NP surface. Indeed, magnetite NPs synthesized using plant extract having the lowest acidity exhibited the highest acid site density (OH groups) and hence the highest performance in the removal of MG.
As it is known that the adsorption of dyes from wastewater is directly affected by the reactivity of the adsorbent surface towards the attachment of dyes' functional groups, looking for parameters that could control the greenly synthesized magnetite NP surface reactivity is always of importance. In this way, this paper looks for mediating plant extract parameters that could, in the case of green synthesis, impact the magnetite NP surface reactivity in azo dye adsorption.
Azo dyes are the largest and most versatile class of organic dyes. 31 These dyes contain one or more azo bonds (N]N). The complex aromatic structures of azo dyes make them more stable and more difficult to remove from the effluents discharged into water bodies. 31 The used dyes in this study are: methyl orange (MO) and Evans blue (EB) azo dyes, and their chemical structures are illustrated in Fig. 1. The EB molecule is composed of chromophore groups such as benzene, phenyl, phenyldiazonium, toluene, N]N, C]C, and C]N groups linked to the benzene ring, and of auxochrome groups such as sulphonic acid (SO À 3 ), phenol, and aniline groups. Meanwhile, the MO molecule is composed of chromophore groups such as benzene, phenyl, phenyldiazonium, N]N, C]C, and C]N, and of auxochrome groups such as sulphonic acid (SO À 3 ) and dimethylamine (N(CH 3 ) 2 ).
The attachment of these functional groups on the adsorbent surface is based on functional group properties (chromophore or auxochrome) towards the surface acid site type (Lewis or Brønsted). In the case of EB and MO dyes, 33 phenol, 34 and aniline 35 auxochrome groups are electron donating, and their attachment on the adsorbent surface is based on their ionic interaction with the developed positively charged Brønsted acid sites of the surface. Meanwhile, the chromophore groups phenyldiazonium, phenyl, 36 benzene, N]N, C]C, C]N, 32 and toluene 37 are electron withdrawing, and their attachment on the adsorbent surface is based on their interactions with the Lewis acid sites of the adsorbent surface.
In this paper, the effect of the mediating plant extract's acidity on the greenly synthesized magnetite surface reactivity in the adsorption of methyl orange (MO) and Evans blue (EB) azo dyes has been investigated through the analysis of MO and EB adsorption on four greenly synthesized magnetite samples, with the aim of studying the effect of the mediating plant extract's acidity on the preferential attachment of the studied dye chromophores and auxochromes (functional groups) on greenly synthesized magnetite NPs, and adsorption yields. For this purpose, aer the accomplishment of adsorption experiments, the free functional groups that are not attached on magnetite surfaces have been deeply analyzed in all dye residual solutions using ATR-FTIR spectroscopy, so as to perceive their preferential attachment on the four magnetite surfaces. Based on the analysis of preferential attachments of chromophore and auxochrome groups, it was possible to compare between Brønsted and Lewis acid site densities on each magnetite surface and their inuence on adsorption yields and capacities. To the best of our knowledge, there is no study in the literature that has dealt with the inuence of the mediating plant extract's acidity on magnetite surface reactivity towards the preferential attachment of dye functional groups, and thus on surface acid site type and density. The used plants are Artemisia herba-alba (L), Matricaria pubescens (L), Juniperus phoenicea (L), and Rosmarinus officinalis (L). Their extract pHs are respectively 5. 25

Materials and methods
This section focuses on listing needed materials and used apparatuses. It also provides methods utilized to perform adsorption experiments and characterization of iron oxide NPs. O 4 nanoparticle powders were greenly synthesised using iron salt (FeCl 3 $6H 2 O) (purchased from Biochem Chemopharma Co, Canada) as precursor and Artemisia herba-alba (L) (Asteraceae family), Matricaria pubescens (L) (Asteraceae family), Juniperus phoenicea (L) (Cupressaceae family), and Rosmarinus officinalis (L) (Lamiaceae family) plants as reducing agents. Magnetite samples were obtained aer 4 months of storage of the synthesized iron oxides under ambient conditions. Characterization of the freshly synthesized iron oxides samples is presented in Ahmouda et al. 38 2.1.2 Apparatuses. XPERT-PRO X-ray diffractometer (XRD) (Rigaku Miniex 600) with 30 keV and 30 mA as conditions of Xray generation and Ka radiation of copper (l ¼ 1.54056Å). Attenuated total reectance-Fourier transform infrared spectroscopy (ATR-FTIR): Shimadzu IR-Innity. Ultraviolet-visible spectroscopy (UV-Vis): Shimadzu UV-Vis spectrophotometer apparatus Model 1800 operating in the range of 200-900 nm. SEM-TESCAN VEGA3 Model XMU, LMH (Brno, Czech Republic) for speeding up the voltage of 15-20 kV in order to capture the scanning electron micrograph (SEM) of greenly synthesised iron oxide NPs.

Methods
In this section, used methods for solution preparation are described. Used protocols in adsorption and desorption experiments of dyes from magnetite NPs and characterization techniques are described as well.
2.2.1 Batch adsorption experiments of MO and EB dyes. First, the prepared standard aqueous solutions of EB and MO dyes were diluted several times as required. Then, 0.0015 g of JUN-Fe 3 O 4 , MAT-Fe 3 O 4 , ROS-Fe 3 O 4 and ARM-Fe 3 O 4 powders were added to a volume of 4 mL of dye aqueous solutions. The dye solution concentration was 0.0111 mg mL À1 . The ionic strength for all adsorption experiments was kept at 0.1 M by adding an appropriate amount of NaCl (0.023 g). A dilute solution of HCl was used to adjust the dye/Fe 3 O 4 solution pH to 4. This protocol is used to prepare, in total, 88 experiment sets (11 with each dye/magnetite sample). In addition, control experiment sets (without NPs) were also prepared.
All experiment sets are sonicated in an ultrasonic bath for 15 minutes and they were then stirred continually for 60 minutes until a steady state was reached. All adsorption experiments were carried out under ambient conditions in batch mode, and they were performed in triplicate for data consistency.
Kinetic experiments were performed by withdrawing samples of dye/Fe 3 O 4 solutions at a regular time interval to obtain, aer centrifugation, adequate aliquots for the purpose of quantifying residual dye concentrations and the adsorbed amounts. The concentrations of residual dye aqueous solutions were quantied using a UV-Vis spectrophotometer at absorbance maxima of EB (l max ¼ 602 nm) and MO (l max ¼ 463 nm). Furthermore, the adsorbed amounts of EB and MO molecules are calculated from the calibration curve for all adsorption experiments (Y ¼ 67.02X + 0.0442, R 2 ¼ 0.9987 and Y ¼ 31.39X + 0.0346, R 2 ¼ 0.9985, respectively). On another side, aer the adsorption was accomplished and steady state reached, the aliquots were centrifuged to separate liquid solutions and solid phases. The liquid solutions, which represent MO/magnetite and EB/magnetite residual solutions, were then analyzed using ATR-FTIR spectroscopy.
To calculate the adsorption capacity (q e in mg g À1 ) and the amount of MO and EB ions adsorbed per unit mass (q t in mg g À1 ) of JUN-Fe 3 O 4 , MAT-Fe 3 O 4 , ROS-Fe 3 O 4 and ARM-Fe 3 O 4 at equilibrium contact time, the following equations were used: Adsorption yield was calculated using the following equation: where C 0 , C e , C t , V and m are respectively: initial dye concentration in the liquid phase (mg mL À1 ), liquid phase dye concentration at the steady state (mg mL À1 ), liquid phase dye concentration at time t (mg mL À1 ), volume of dye solution used (mL), and the amount of adsorbent (g).

2.2.2
Linear and non-linear pseudo-rst-order and pseudosecond-order kinetics. The linear (eqn (4)) and non-linear (eqn (5)) pseudo-rst-order (PFO) or Lagergren, 39 and linear (eqn (6)) and non-linear (eqn (7)) pseudo-second-order (PSO) or Ho and McKay 40 kinetic models are selected to test the adsorption dynamics in this study due to their good applicability in most studies. [41][42][43] ln(q e À q t ) ¼ ln q e À K 1 t (4) where K 1 , q t and q e are respectively: the pseudo rst order kinetic constant (min À1 ), adsorbed dye quantity at instant t (mg g À1 ) and adsorbed dye quantity at thermodynamic equilibrium (mg g À1 ).
If the active surface of the adsorbent is regarded as invariable, the reaction could be treated as pseudo-rst-order kinetic. However, once the active sites have been saturated, the transfer at the adsorbate/adsorbent particle interface may be limited by mass transfer. 44 The pseudo second-order (PSO) model is proposed by Ho and McKay. 40 It is based on the adsorption capacity expressed as follows: where K 2 , q t and q e are the pseudo second order kinetic constant (mg g À1 min À1 ), adsorbed dye quantity at instant t (mg g À1 ) and adsorbed dye quantity at thermodynamic equilibrium (mg g À1 ), respectively. 2.2.3 Intra-particle diffusion kinetics. In order to gain insights into the adsorption mechanisms involved, a homogeneous particle diffusion model (HPDM) as shown in eqn (8), originally proposed by Boyd et al., 45 is used to describe the diffusive adsorption process. In this model, the rate-limiting step is usually described by either an intra-particle diffusion or a lm diffusion mechanism.
where F(t) is the fractional attainment at time t, i.e., F(t) ¼ q t /q e , D p (m 2 s À1 ) is the effective diffusion coefficient, r 0 is the radius of Fe 3 O 4 particles assumed to be spherical, and Z is an integer. For 0 < F(t) < 1, a simplied equation can be obtained for the adsorption on spherical particles: A further formula manipulation gives the following: where k p is the diffusion rate constant (1/s) and k p ¼ D p p 2 / r 2 0 . Eqn (10) is used for the calculation of effective intra-particle diffusivity (D p (m 2 s À1 )) from the experimental data. In the rst step, a graph of Àln(1 À , can then be obtained from D p ¼ k p r 2 0 /p 2 . Additionally, eqn (11) can be used when the rate of adsorption is controlled by liquid lm diffusion. 46 where D f is the lm diffusion coefficient (m 2 s À1 ) in the liquid phase, and C e (mol L À1 ) and C r (mol L À1 ) are respectively the equilibrium concentrations of dye in solution and in solid phases. d is the thickness of the liquid lm which was assumed to be 10 À5 m according to Yu and Luo. 47 A further formula deformation of eqn (11) gives the following equation: The linearity test of Boyd plots Àln(1 À F) and Àln(1 À F 2 ) versus time plots is employed to distinguish between the lm diffusion and particle diffusion-controlled adsorption mechanism. If the plot of Àln(1 À F) versus time is a straight line passing through the origin, then the adsorption rate is governed by the particle diffusion mechanism, otherwise if Àln(1 À F 2 ) versus time is a straight line passing through the origin then the adsorption is governed by lm diffusion.
2.2.4 Batch desorption experiments of MO and EB dyes. In order to regenerate used magnetite samples aer the end of dye adsorption, a dilute solution of NaOH was used to adjust the dye/Fe 3 O 4 solution pH in the range of 8-12. All experiment sets were sonicated in an ultrasonic bath for 15 minutes and they were then stirred continually for 60 minutes. All desorption experiments were carried out in ambient conditions in batch mode, and they were performed in triplicate for data consistency. This protocol is used to regenerate, in total, 8 magnetite samples. In addition, control experiment sets (without NPs) were also prepared. Aer centrifugation, adequate aliquots were taken for the purpose of quantifying desorbed dye concentrations and the desorbed amounts from magnetite surfaces. The concentrations of desorbed dye from magnetite surfaces were quantied using a UV-Vis spectrophotometer at absorbance maxima l max ¼ 602 nm for EB and l max ¼ 463 nm for MO. Furthermore, the desorbed amounts of EB and MO molecules are calculated from the calibration curve for all adsorption To calculate the desorption yields (R%) of MO and EB ions at contact time t ¼ 60 min from JUN-Fe 3 O 4 , MAT-Fe 3 O 4 , ROS-Fe 3 O 4 and ARM-Fe 3 O 4 surfaces, the following equation was used: where C 0 and C t are respectively: initial dye concentration adsorbed on the solid phase (magnetite) (mg mL À1 ) and liquid phase desorbed dye concentration at contact time t ¼ 60 min (mg mL À1 ).

2.2.5
Thermodynamic studies of EB and MO adsorption processes on magnetite surfaces. In order to study the thermodynamics of EB and MO adsorption processes, all sets of experiments are sonicated in an ultrasonic bath for 15 minutes. Then, they are stirred continually for 20 minutes at temperatures ranging from 303.15 to 318.15 K. The concentrations of residual EB and MO dyes in liquid phases are quantied using a UV-Vis spectrophotometer at absorbance maxima l max ¼ 602 nm for EB and l max ¼ 463 nm for MO. Furthermore, the adsorbed amounts of EB and MO molecules are calculated from the calibration curves for all adsorption experiments (Y The adsorption capacity, q eT (mg g À1 ), was calculated using the following equations: Adsorption yield was calculated using the following equation: where C 0 , C eT , V and m are respectively: initial dye concentration without any treatment (mg mL À1 ), residual dye concentration in  the liquid phase (mg mL À1 ) at a given temperature T, the volume of dye solution (mL), and the amount of magnetite NPs (g). The activation enthalpy (DH 0 ) of EB and MO adsorption processes on the magnetite NP surface was determined using the Arrhenius equation as follows: where R (1.987 cal mol À1 K À1 ) is the universal gas constant, T is the absolute solution temperature (K), and k d is the distribution coefficient which was calculated by: where C aeT (mg mL À1 ) and C eT (mg mL À1 ) are respectively the concentration of adsorbed dye on the solid (magnetite) and dye residual concentration in the liquid phase at a given temperature T.
The values of activation enthalpy DH 0 (kcal mol À1 ) and entropy DS 0 (cal mol À1 K À1 ) were respectively calculated from the slope and intercept of plots between ln k d and 1/T. DG 0 (kcal mol À1 ) was then calculated using the following equation: The free energy change indicates the degree of spontaneity of the adsorption process. The higher negative value reects more energetically favorable adsorption. The activation energy, DE a (kcal mol À1 ), of EB and MO adsorption processes on magnetite surfaces was determined using Arrhenius's equation: where k 2 is the distribution coefficient which can be calculated by: where q eT (mg g À1 ) and C eT (mg mL À1 ) are respectively the adsorption capacity of dye on the solid and dye residual concentration in the liquid phase at temperature T.

Results and discussion
3.1 X-ray analysis of magnetite nanoparticle samples X-ray patterns of all synthesised samples are presented in Fig. 2.
It is found that all synthesized samples have crystalline structures. The X-ray diffraction pattern in Fig. 2 where D, b, l, and q are the crystallite size, the full width at halfmaximum (FWHM) of the most intense diffraction peak, the Xray wavelength (1.54056Å), and Bragg angle, respectively.

ATR-FTIR spectroscopy analysis
FTIR spectra of the synthesized Fe 3 O 4 nanoparticle powders recorded between 4000 and 500 cm À1 are presented in Fig. 3. Fig. 3 shows that all IR spectra (A, B, C, and D) exhibit peaks in different ranges as is summarized in  Fig. 3 shows that the peaks of hydroxyl groups appear in remarkably different areas. Meanwhile, the hydroxyl group peak area appears to be the broadest on the ARM-

UV-vis spectroscopy analysis
The optical absorbance spectra of all the synthesized Fe 3 O 4 are measured in the wavelength range of 200-900 nm, to deduce their band gap energies. The band gap E g and the optical absorption coefficient (a) of a direct band gap semiconductor are related through the known following equation: 51 where a is the linear absorption coefficient of the material, hn is the photon energy, A is a proportionality constant, and the exponent n depends on the nature of electronic transition; it is equal to 1/2 for a direct allowed transition and 2 for an indirect allowed transition. E g of the direct transition of all samples were obtained from plotting (ahn) 2 as a function of ahn by the extrapolation of the linear portion of the curve (Fig. 4). However, E g of the indirect transition of all samples were obtained from plotting (ahn) 1/2 as a function of ahn by the extrapolation of the linear portion of the curve (Fig. 5 Fig. 6b. However, for MAT-Fe 3 O 4 a decrease in the dimensions of the mountains with more adherence to its structure is observed (Fig. 6c). Finally, the ARM-Fe 3 O 4 SEM image sometimes contains large-structured single bipyramid crystal as mentioned in Fig. 6d. The fast initial adsorption is mainly due to the rapid attachment of dye molecules to favorable active sites of magnetite surfaces through mass transfer. The subsequent adsorption is achieved through molecular diffusion of the dye molecules into pores of magnetite NPs. 54 3.5.2 Adsorption experiment kinetics analysis 3.5.2.1 Linear and non-linear pseudo-rst-order and pseudosecond-order kinetics. The results of linear and non-linear pseudo-rst-order (PFO) and pseudo-second-order (PSO) kinetics analysis of MO and EB adsorption processes on the four magnetite surfaces (Tables 5 and 6, and Fig. 8 and 9) indicate a good linearity and a good t of the experimental data to linear pseudo-rst-order. However, the comparison with the pseudosecond-order model indicates a poor linearity and a poor t to the experimental data of this model. Indeed, the equilibrium adsorption capacities q e,cal computed from linear pseudo-rstorder kinetics plots are in very close agreement with the Nanoscale Advances empirical q e,exp , unlike q e,cal calculated from linear pseudosecond-order plots which are far from empirical q e,exp . 3.5.2.2 Intra-particle diffusion mechanism analysis. The linearity tests of Boyd plots (Àln(1 À F) and Àln(1 À F 2 )) versus time are presented in Fig. 10. The kinetic data correlate well with the homogeneous particle diffusion model as conrmed by the high R 2 values. The results of linear regression analysis for eqn (10) and (12) are presented in Table 7. Film diffusion coefficients D f are found to be in the order of 10 À11 m 2 s À1 , while intra-particle diffusion coefficients D p are found to be in the order of 10 À20 m 2 s À1 . Michelson et al. 55 reported that the adsorption mechanism is controlled by lm diffusion and that lm diffusion is in control at D f ranging from 10 À10 to 10 À12 m 2 s À1 , while intra-particle diffusion is the rate-limiting step at D p in the range of 10 À15 to 10 À18 m 2 s À1 . Indeed, the obtained results indicate that lm diffusion is the step that controls the adsorption mechanism of MO and EB on magnetite surfaces, which is in agreement with the pseudo-rst-order kinetic model.
3.5.3 Activation thermodynamic parameters of MO and EB adsorption on magnetite surfaces. The calculated activation enthalpy DH 0 , entropy DS 0 , and free energy DG 0 are listed in Tables 8 and 9. DH 0 and DS 0 of EB/plant-Fe 3 O 4 systems were respectively calculated from the slopes and intercepts of the Arrhenius linear plots ln K D versus 1/T (Fig. 11b). Likewise, DH 0 and DS 0 of MO/plant-Fe 3 O 4 systems were respectively calculated from the slopes and intercepts of the Arrhenius linear plots ln K D versus 1/T (Fig. 12b)   As presented in Table 8, the found activation energies (E a ) for EB adsorption on ARM-Fe 3 O 4 , ROS-Fe 3 O 4 , MAT-Fe 3 O 4 , and JUN-Fe 3 O 4 surfaces are respectively: 2.79, 3.21, 5.59, and 6.29 kcal mol À1 . E a is calculated from the slopes of the Arrhenius linear plots ln K 2 versus 1/T (Fig. 11a). As presented in Table  9 Table 3 Average adsorption capacities q t of EB adsorption on magnetite samples. Standard deviation (STD) of three replicates is mentioned  are respectively: 3.27, 3.66, 6.28, and 8.45 kcal mol À1 . E a is calculated from the slopes of Arrhenius linear plots ln K 2 versus 1/T (Fig. 12a) Fig. 13a and b, and Table 10 present the comparison of EB and MO adsorption yields and capacities on the four Fe 3 O 4 surfaces at ambient temperature and in the temperature range of 303.15-318.15 K. The data show that EB and MO adsorption yields and capacities increase with the increase of temperature in all adsorption experiments, which conrms the endothermic nature of the adsorption processes as discussed in Section 3.5.3.
The tendency of adsorption capacities and yields on the four magnetite surfaces is the same in EB and MO adsorption processes at 298. 15 Table 4 Average adsorption capacities q t of MO adsorption on magnetite samples. Standard deviation (STD) of three replicates is mentioned  16.83 mg g À1 and 39.61%, 11.73 mg g À1 , respectively.When magnetite is immersed in an aqueous acidic solution, it develops its surface charge via the protonation and deprotonation of^FeOH sites on its surface according to the following equation: 59^F eOH + 2 4^FeOH 0 + H + sol (pk a1 ¼ 5.1) where^FeOH + 2 and^FeOH 0 are respectively the protonated positively charged surface group with two dissociable H + , and the neutral surface group with one dissociable H + . pK a1 ¼ 5.1 is the intrinsic acidity constant determined by Davis et al. 59 for magnetite.^FeOH + 2 known as the Brønsted acid site is an electron-pair acceptor and it is able to transfer H + from the solid to the adsorbed molecule, implying an H-bond with the surface. Meanwhile,^FeOH 0 known as the Lewis acid site is an electron-pair acceptor from the adsorbed molecule, implying a coordinated bond with the surface. 59 In order to study how the plant extract's acidity impacted preferential attachment of dyes' chromophore and auxochrome groups on synthesized magnetite surfaces, the free chromophore and auxochrome groups that were not attached to magnetite surfaces have been deeply analyzed in all dyes' residual solutions using FTIR spectroscopy, so as to perceive preferential attachment of chromophore and auxochrome groups on each surface and hence determine Brønsted and Lewis acid site densities. Based on this analysis, it was possible to infer Brønsted and Lewis acid site densities on each magnetite surface. For this purpose, aer the accomplishment of the adsorption in all experiments, the solid and liquid fractions were separated using a centrifuge. In the next sections, the analysis of residual dye chemistry changes of MO and EB will be studied in detail by comparison between FTIR spectra of MO and EB reference solution chemistry, and the chemistry of their residual solutions.
In the rest of this paper, MO residual solutions will be denoted as MO/ARM-Fe 3  3.6.1 Analysis of MO/plant-magnetite residual solution chemistry changes. The FTIR spectrum of the MO reference solution (Fig. 15A) 63 and C-H bending vibrations of the aromatic ring at 1047.03 cm À164 and at 1000-550 cm À1 . 65 In Fig. 15, spectra  The newly appeared peaks correspond to SO À 3 , ÀN þ H, asymmetric CH, phenyl, and the ring skeleton of benzene groups (RSB). More details about the identication of these peaks in MO/plantmagnetite residual solutions will be given in the next paragraphs, for the purpose of investigating the preferential attachment of chromophore and auxochrome groups on all four magnetite surfaces.
The decomposition of the dye is signicantly accelerated by the presence of acidic centers at the surface. 66 3.6.1.1 Analysis of preferential attachment of chromophore groups in the adsorption of MO. New peaks of the NH bond of phenyldiazonium groups shied from 3378.48 cm À1 to around 3308 cm À1 in all four residual solutions (Fig. 15), with however different peak areas. Furthermore, all spectra in Fig. 15 show that new peaks, with however different areas, appear at around 1633 cm À1 . These peaks correspond to the C]C bond of not attached phenyl groups. 67 Additionally, new peaks appear, with different areas, at around 646 cm À1 , which correspond to the CH of the ring skeleton of benzene. 21 The peaks of these groups in the FTIR spectrum of MO/ARM-Fe 3 O 4 are the narrowest ones, followed by those of MO/ROS-Fe 3 O 4 , then those of MO/MAT-Fe 3 O 4 , and nally, the broadest one is in the FTIR spectrum of The analysis of preferential attachment of chromophore groups shows, as summarized in Fig. 16, that: On JUN-Fe 3 O 4 , all MO chromophore groups were less attached compared to on the three other magnetite NPs.
On MAT-Fe 3 O 4 , N]N, C]C, and C]N groups were almost completely attached while other groups were more attached Fig. 10 Homogeneous particle diffusion plots of EB adsorption: (a) Àln(1 À F) and (b) Àln(1 À F 2 ). Homogeneous particle diffusion plots of MO adsorption: (c) Àln(1 À F) and (d) Àln(1 À F 2 ), on magnetite surfaces. Error bars represent the standard deviation of three replicates.   On ARM-Fe 3 O 4 , N]N, C]C, and C]N were completely attached, whereas phenyldiazonium, phenyl, and benzene groups were more attached on ARM-Fe 3 O 4 than on the other magnetite NPs.
As it is known that chromophore groups prefer to attach to Lewis acid sites, it is possible to infer that the density of Lewis acid sites of ARM-Fe 3 O 4 is the highest one, followed by that of ROS-Fe 3  Furthermore, FTIR spectra of all four residual solutions (Fig. 15) show that new peaks of asymmetric vibration of CH of  Fig. 17 shows a broad peak of the NH bond of the phenyldiazonium ring at 3375.89 cm À1 , 60 aromatic CH ring stretching at 2920.97 cm À1 , 60 C^C at 2300-2100 cm À1 , 61 aromatic C]C ring stretching at 1943.10 cm À1 , 1658.11 cm À1 , 62    groups. More details about the identication of these peaks in EB/plant-magnetite residual solutions will be given in the next paragraphs, for the purpose of investigating the preferential attachment of the EB chromophore and auxochrome groups on all four magnetite surfaces.
3.6.2.1 Analysis of preferential attachment of chromophore groups in EB adsorption. The spectra of all four EB/plantmagnetite residual solutions (Fig. 17) show that the NH bond peak of not attached phenyldiazonium groups shied from 3387.56 in the EB aqueous reference solution to around 3302 cm À1 , with however different peak bond areas. Additionally, new peaks, with different areas, appear at around 1634 cm À1 which correspond to the vibration of C]C of phenyl groups. 67 Furthermore, the peaks of the ring skeleton of benzene groups appear, with different areas, in all four FTIR spectra at around 642 cm À1 . 21 It is worth noting that peak areas of NH of phenyldiazonium, C]C of phenyl, and CH of benzene groups vary in the same manner, where the peak area in the FTIR spectrum of EB/ARM-Fe 3 O 4 is the narrowest one, followed by that of EB/ROS-Fe 3     The analysis of preferential attachment of chromophore groups on magnetite surfaces shows, as summarized in Fig. 18, that: On JUN-Fe 3 O 4 , all EB chromophore groups were less attached compared to on the three other magnetite NPs.
On On ARM-Fe 3 O 4 , N]N, C]C, C]N, toluene, and C^C were completely attached, while phenyldiazonium, phenyl, and benzene groups were more attached than on other magnetite surfaces.
As it is known that chromophore groups prefer to attach to Lewis acid sites, it is possible to infer that the density of Lewis acid sites of ARM-Fe 3 O 4 is the highest one, followed by that of ROS-Fe 3 69 and to the bending vibration of NH of aniline groups 70 appear at 1099.04 and 1204.58 cm À1 , respectively (Fig. 17). In the FTIR spectrum of EB/ ARM-Fe 3 O 4 , this peak is slightly more intense than in that of EB/ROS-Fe 3 O 4 . Meanwhile, no peaks corresponding to these groups appear in the FTIR spectra of EB/MAT-Fe 3 O 4 and EB/ JUN-Fe 3 O 4 (Fig. 17).   surface. These nding are in conformity with the results of several studies 16,28,29 witch reported that the variation of mediating plant extract has a clear effect on the greenly synthesised adsorbent surface reactivity in dye adsorption yield and capacity. As it is known that dye adsorption is directly affected by the reactivity of the adsorbent surface towards the attachment of dyes' functional groups, the focus here is on the effect of the mediating plant extract's acidity on the preferential attachment of EB and MO dye functional groups in their adsorption processes on the greenly synthesized magnetite samples. To the best of our knowledge, this is the rst report in which such a study has been conducted.     16.83 mg g À1 and 39.61%, 11.73 mg g À1 ), respectively. Accordingly, the adsorption yields and Lewis acid site densities varied in the same manner. Seeing that plant extracts used in the synthesis of ARM-Fe 3 O 4 , ROS-Fe 3 O 4 , MAT-Fe 3 O 4 and -Fe 3 O 4 NPs have respectively pH 5.25, 5.05, 4.63, and 3.69, one can conclude that plant extract pH has a clear effect on the preferential attachment of dye chromophore and auxochrome groups, magnetite nanoparticle acid sites, and adsorption yields. Indeed, the decrease of the mediating plant extract's acidity leads to the increase of Lewis acid site densities and the decrease of Brønsted acid site densities on magnetite NPs and hence an increase in the attachment of chromophore groups and a decrease in the attachment of auxochrome groups of dyes.
As most of the MO and EB functional groups are chromophores, the decrease of the mediating plant extract's acidity also leads to an increase in adsorption yields. The remarked difference in adsorption yields of EB and MO on all four magnetite NPs is due to the fact that the ratio of chromophore/auxochrome groups in EB is remarkably greater than that in MO.
Thus, the plant extract's acidity could provide a preconceived idea about the densities of Brønsted and Lewis acid sites of magnetite NPs to be greenly synthesized and therefore about azo dye adsorption yields. Dye adsorption yield can be predicted according to the content of chromophore and auxochrome groups in the azo dye structure.
3.7 Desorption efficiency of dyes from magnetite surfaces and their stability aer 3 cycles of reuse 3.7.1 Desorption of dyes from magnetite surfaces. Desorption yields of EB and MO from magnetite surfaces (adsorbed at pH ¼ 4) in the pH range of 8-12 are calculated using eqn (13) and they are shown in Table 12 and Fig. 20. It was observed that with the increase in solution pH desorption, EB and MO desorption efficiencies increase. At pH ¼ 8, the desorption efficiencies of EB on ARM-Fe 3 O 4 , ROS-Fe 3 O 4 , MAT-    ¼ 12), almost 100% desorption was achieved from the magnetite surfaces in each case. This is mainly due to the electrostatic repulsion between the negatively charged functional groups of EB and MO molecules and the negatively charged active sites of magnetite surfaces. 3.7.2 The stability of magnetite samples aer 3 cycles of reuse. The multi-cycle efficiencies of magnetite samples were also tested for each case to evaluate their stability during dye adsorption experiments. The relative adsorption efficiencies of magnetite samples for MO and EB dyes were signicant (97-95% and 94-91%) aer the second and third cycles of reuse, where the average adsorption efficiencies of the rst cycle were found to be 100% in all plant-Fe 3 O 4 samples (see Fig. 19). This indicates the stability of all plant-Fe 3 O 4 samples during EB and MO adsorption experiments.

Conclusion
The adsorption of methyl orange (MO) and Evans blue (EB) azo dyes on four greenly synthesized magnetite NPs has been studied. The pHs of plant extracts used for the green synthesis were 5.25, 5.05, 4.63, and 3.69. The aim of the study was the investigation of the plant extract's acidity effect on the magnetite surface reactivity through the analysis of the preferential attachment of the dyes' chromophore and auxochrome groups on magnetite nanoparticles, and adsorption yields, and therefore determination of the plant extract pH effect on acid site types and densities. To do so, the free chromophore and auxochrome groups that are not attached to magnetite surfaces have been deeply analyzed in all dye residual solutions using FTIR spectroscopy, so as to perceive preferential attachment of dye chromophore and auxochrome groups on the four magnetite surfaces.
Obtained results show that the mediating plant extract's acidity has a clear effect on preferential attachment of dye chromophore and auxochrome groups, magnetite nanoparticle acid sites, and adsorption yields. Indeed, the decrease of plant extract acidity leads to the increase of Lewis acid site densities and the decrease of Brønsted acid site densities on magnetite NPs and hence an increase in the attachment of chromophore dye groups and a decrease in the attachment of auxochrome dye groups. As most of the MO and EB functional groups are chromophores, the decrease of the mediating plant extract's acidity also leads to an increase in adsorption yields.
The linear and non-linear pseudo-rst-order and pseudosecond-order kinetics of the adsorption as well as the intraparticle diffusion mechanism have also been analyzed. Obtained results indicated that the adsorption kinetic followed a linear pseudo-rst-order kinetic model. Meanwhile, lm diffusion was found to be the step that controlled the adsorption mechanism of MO and EB adsorption processes. The thermodynamic studies of EB and MO adsorption processes have been analyzed in the temperature range of 303.15-318.15 K. They reveal the physical and endothermic nature of the adsorption in all cases.

Conflicts of interest
There are no conicts to declare.