Assessment of naproxen adsorption on bone char in aqueous solutions using batch and fixed-bed processes

Journal of Molecular Liquids 209 (2015) 187–195

Contents lists available at ScienceDirect

Journal of Molecular Liquids journal homepage: www.elsevier.com/locate/molliq

Assessment of naproxen adsorption on bone char in aqueous solutions using batch and fixed-bed processes Hilda E. Reynel-Avila a, Didilia I. Mendoza-Castillo a,b, Adrián Bonilla-Petriciolet a,⁎, Joaquín Silvestre-Albero c a b c

Tecnológico Nacional de México, Instituto Tecnológico de Aguascalientes, Aguascalientes 20256, Mexico CONACYT, Cátedras Jóvenes Investigadores, México D.F. 03940, Mexico Laboratorio de Materiales Avanzados, Departamento de Química Inorgánica, Universidad de Alicante, Apartado 99, E-3080 Alicante, Spain

a r t i c l e

i n f o

Article history: Received 15 March 2015 Accepted 5 May 2015 Available online xxxx Keywords: Adsorption Bone char Fixed-bed columns Naproxen Pharmaceuticals Water treatment

a b s t r a c t An integrated analysis of naproxen adsorption on bone char in batch and packed-bed column conditions has been performed. Kinetic, thermodynamic and breakthrough parameters have been calculated using adsorption models and artificial neural networks. Results show that naproxen removal using bone char in batch conditions is a feasible and effective process, which could involve electrostatic and non-electrostatic interactions depending mainly on pH conditions. However, the application of packed-bed column for naproxen adsorption on bone char is not effective for the treatment of diluted solutions due to the low degree of adsorbent utilization (below 4%) at tested operating conditions. The proposed mechanism for naproxen removal using bone char could include a complexation process via phosphate and naproxen, hydrogen bonding and the possibility of hydrophobic interactions via π–π electron. This study highlights the relevance of performing an integrated analysis of adsorbent effectiveness in batch and dynamic conditions to establish the best process configuration for the removal of emerging water pollutants such as pharmaceuticals. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Carbon-based adsorbents are versatile and the most popular materials for wastewater treatment and water purification due to their physical, chemical and structural properties [1–3]. These adsorbents can be obtained from different precursors and synthesis conditions and, consequently, they can be effectively applied for improving water quality via the removal of a great variety of pollutants, which include dyes, heavy metals, pesticides, arsenic, fluoride, and other inorganic and organic compounds [1,2,4–8]. Specifically, the use of carbon-based adsorbents for water purification is recognized as a prevailing method and it has become the focus of extensive research because of its simple design, easy operation, low cost and high efficiency [9,10]. It has been estimated that almost 80% of the carbon adsorbent production is used for aqueous adsorption processes [9]. Recently, these adsorbents have found applications for the removal of nontraditional water pollutants such as radioisotopes or pharmaceuticals [11,12]. To date, pharmaceuticals are considered as emerging water pollutants due to their widespread use worldwide [13,14]. These pollutants can enter into the sewer system via the human urine and fecal matter and, consequently, they can reach the wastewater treatment plants [12,14]. Recent studies have reported that conventional sewage ⁎ Corresponding author. E-mail address: [email protected] (A. Bonilla-Petriciolet).

http://dx.doi.org/10.1016/j.molliq.2015.05.013 0167-7322/© 2015 Elsevier B.V. All rights reserved.

treatment plants are not effective to remove/degrade these compounds [12,14]. Therefore, pharmaceuticals may be continuously introduced into water resources representing an environmental risk for both ecosystems and human beings due to its toxicological profile, even at trace concentrations [3,12,13,15,16]. The adsorption of pharmaceuticals has been studied mainly at batch operational conditions using activated carbons and other adsorbents such as soils, clays, nanocomposites, hydrous oxides, alumina and silica [3,12,14,17–19]. The adsorption capacity of traditional materials usually ranges from 0.1 to 10 mg/g, depending on the adsorbent, operating conditions and pharmaceutical compound used as solute. On the other hand, few authors have reported on the adsorption of pharmaceuticals at dynamic operational conditions using packed-bed columns [20–22]. It is convenient to highlight that, although the batch experiments provide valuable data about adsorbent effectiveness, including kinetic data and equilibrium uptakes, the evaluation of adsorbent performance at dynamic conditions is fundamental for the process scale up [23]. Dynamic adsorption experiments are necessary to obtain practical information about the adsorption capacity under flow conditions; they are needed to calculate design parameters for real-life applications and to identify the best operating conditions [23]. Note that the quantitative characterization of the adsorbent performance in packed-bed columns also involves the modeling and prediction of the breakthrough curves. Literature indicates that extensive studies in pilot plant scale can be avoided if the breakthrough curves for adsorption columns can be reliably modeled and predicted using laboratory scale measurements

188

H.E. Reynel-Avila et al. / Journal of Molecular Liquids 209 (2015) 187–195

[20,24]. Traditionally, the adsorbent performance is tested in batch or dynamic conditions, but not in both. However, the integrated analysis of both adsorption systems is fundamental for determining the best configuration and for performing a reliable design of treatment/purification processes used in the removal of priority water pollutants. With this in mind, the objective of the present study is to assess the performance of bone char as adsorbent for naproxen removal from aqueous solution at both batch and dynamic operating conditions. Naproxen is a member of the arylacetic acid group that exhibits antiinflammatory and analgesic effects in medical treatments [13,14]. However, this drug is considered as an emerging water pollutant with a high environmental risk [12]. The bone char is a low-cost and high performance carbon-based adsorbent that can be applied for the removal of both organic and inorganic compounds [8,25–27]. To the best of the author's knowledge, this adsorbent has not been used for naproxen adsorption from aqueous solutions. Therefore, batch and dynamic adsorption tests for naproxen on bone char have been studied at different operating conditions and data analysis has been performed to calculate kinetic, thermodynamic and design parameters using both adsorption models and artificial neural networks. The effectiveness of batch reactors and packed-bed columns for naproxen adsorption on bone char has been discussed including the advantages and limitations of each process configuration. 2. Methodology 2.1. Adsorbent description and its physicochemical characterization Naproxen adsorption experiments were performed using a commercial bone char supplied by Carbotecnia company (Mexico), which is produced from bovine bones. Bone char is considered as a mixed adsorbent constituted by carbon and calcium phosphate in the hydroxyapatite form [26]. In this study, the commercial bone char was washed with deionized water until obtaining a constant pH in washing solution. It was dried and sieved to obtain a mean particle diameter of 0.67 mm (i.e., 20–35 mesh fraction) and this raw adsorbent was employed for naproxen removal. Different characterization techniques were applied for the determination of physicochemical properties of the bone char samples. Specifically, textural parameters of the adsorbent (i.e., surface area, pore size distribution and pore volume) were calculated by N2 adsorption–desorption isotherm at 77 K using a home-made fully automated equipment designed and constructed by the Advanced Materials group (LMA, Universidad de Alicante), commercialized as N2 Gsorb-6 (www.g2mtech.com). FTIR analysis was performed to identify organic functional groups on adsorbent surface using a Bruker IFS 66/S spectrophotometer with 200 scans and a resolution of 4 cm− 1. The physical morphology of bone char was observed with a JEOL JSM-840 scanning electron microscope (SEM) where adsorbent particles were coated with gold. 2.2. Batch adsorption of naproxen and the calculation of kinetic and equilibrium parameters Naproxen adsorption kinetics and isotherms were obtained at different operating conditions using batch reactors with a mass–volume dosage of 10 mg/mL. These adsorption experiments were performed at pH 5–7 and 20–40 °C using aqueous solutions prepared with naproxen sodium and deionized water. Kinetic adsorption rates were calculated with naproxen initial concentrations of 50, 100 and 200 mg/L and samples were taken from t = 0.5 to 24 h; while the adsorption isotherms were obtained using initial concentrations from 20 to 250 mg/L where the equilibrium time was 24 h. The naproxen concentrations were determined using a Hach DR 5000 UV/Vis spectrophotometer at the corresponding characteristic wavelength of naproxen (i.e., 261 nm), and using a linear calibration curve. All the adsorption experiments were conducted in triplicate and the average values were

used for data analysis and modeling. The naproxen adsorption capacity of bone char (q, mg/g) was calculated using a mass balance q¼

C 0 −C t V m

ð1Þ

where C0 and Ct are the initial and final naproxen concentrations (mg/L) in adsorption experiments, V is the naproxen solution volume (L) and m is the adsorbent mass (g), respectively. The experimental adsorption data were analyzed using kinetic and equilibrium equations. Specifically, the pseudo-first and pseudosecond order rate constants [28,29] of naproxen adsorption on bone char were calculated and the intraparticle diffusion analysis [30] was performed; while the Freundlich, Langmuir and Sips models were used for isotherm analysis [31–33]. Kinetic and isotherm data modeling was performed via a non-linear regression where the coefficient of determination (R2) and the mean absolute percentage deviation (E) were considered to measure the goodness of data fitting. Thermodynamic parameters of naproxen adsorption process were calculated using the Gibbs free energy (ΔG°, kJ/mol) ΔG ¼ −RT ln K c

ð2Þ

where R is the universal constant of gases (kJ/mol K), T is the temperature of the adsorption process (K) and Kc is the equilibrium constant, which can be calculated as [15] Kc ¼

C Ae Ce

ð3Þ

where CAe and Ce are the equilibrium naproxen concentrations on the adsorbent and in the solution given in mg/L, respectively. Standard enthalpy (ΔH°) and entropy (ΔS°) for the naproxen adsorption were calculated using ln K c ¼

−ΔH  ΔS þ : RT R

ð4Þ

These thermodynamic parameters were determined from van't Hoff plot. 2.3. Fixed-bed adsorption of naproxen and the calculation of breakthrough design parameters Fixed-bed adsorption experiments were performed in polyethylene columns with an internal diameter of 1.8 cm and a length of 15 cm. The bed porosity of packed bed columns ranged from 25 to 30%. All dynamic adsorption experiments were performed at 30 °C and pH 7, which was the best pH for naproxen removal. Adsorption columns were operated at up-flow operation mode using a peristaltic pump and a feed flow rate of 3 mL/min. Naproxen breakthrough curves were obtained by collecting several effluent samples at different operating times. All fixed-bed columns were operated until the outlet column concentration Ct was 95% of the feed concentration C0. This operating point was established as the saturation/exhaustion condition (te) of the bone char. The experimental conditions tested in dynamic adsorption experiments were: the mass of bone char (i.e., 4.5–17 g) and the initial naproxen concentration in the feed (i.e., 1–10 mg/L). Design parameters of the packed-bed columns were determined from the experimental naproxen breakthrough curves and they were used to characterize the performance of bone char at different dynamic operating conditions. These parameters include the mass transfer zone (MTZ), the breakthrough point time (tb), the overall adsorption zone (Δt) and the retardation factor (rf). These parameters are given by MTZ ¼ L


 t e −t b te

ð5Þ

H.E. Reynel-Avila et al. / Journal of Molecular Liquids 209 (2015) 187–195

Δt ¼ t e −t b

rf ¼

ð6Þ

V 50% ALε

Fq ¼

qbed qe;max

ð8Þ

where qbed and qe,max are obtained at the same operating conditions (i.e., temperature and pH) and given in mg/g. For this calculation, qbed has been obtained via the integration of the experimental breakthrough data (i.e., Ct/C0 versus t) using Z

t t¼o

ðC 0 −C t Þ

Q dt W bed

ð9Þ

where the trapezoidal rule was employed for this calculation. The modeling of naproxen breakthrough curves was performed using the Thomas model and an artificial neural networks (ANNs) approach. Specifically, the Thomas's breakthrough model [34] is based on adsorption–desorption Langmuir kinetics and it assumes that axial and radial dispersions are negligible. This model is defined as Ct ¼ C0

1
  kTh  1 þ exp qbed;T W bed −C 0 V eff Q

ð10Þ

where Q is the feed flow rate in mL/min, Veff is the treated volume of naproxen solution in mL, Wbed is the amount of bone char packed in the column and given in g, qbed,T is the bed adsorption capacity estimated by the model in mg/g and kTh is the Thomas rate constant reported in mL/min mg, respectively. A non-linear regression was also used to estimate qbed,T and kTh using the breakthrough experimental data. On the other hand, a black-box ANNs model was used for representing the performance of naproxen adsorption in the bone char fixed-bed columns. This ANNs model consisted of input, hidden and output layers that are constituted by an interconnected group of artificial neurons [35]. These artificial neurons are interconnected to each other via connection weights where the net input Yij of the neuron j in the layer i is defined mathematically as Yij ¼

nX i −1

Naproxen initial concentration, mg/L Bone char weight, g Operation time, min

ð7Þ

where L is the bed height in cm, A is the cross sectional area of the column in cm2, V50% is the treated volume when the effluent concentration has reached Ct/C0 = 0.5 and ε is the void volume of the packed-bed column, respectively. Note that tb was established as the operating time when Ct/C0 = 0.05. Finally, the bed adsorption capacity (qbed) and the maximum adsorption capacity obtained from adsorption isotherms (qe,max) were used for the calculation of the adsorbent utilization in packed-bed columns (Fq). This parameter is given by

wi jk V i−1;k þ θi j

ð11Þ

k¼1

  Vij ¼ g Yi j

ð12Þ

where Vik is the neuron input, wijk is the connection weight and θij is the neuron bias. Eqs. (11) and (12) can be used for modeling nonlinear relationships between a set of input and output variables. Details of this ANNs approach for adsorption data modeling have been reported in previous studies [36,37]. In particular, the architecture used for the naproxen breakthrough ANNs model is reported in Fig. 1. Specifically, the input variables are the operating conditions of adsorption columns, i.e.,: naproxen feed concentration, bone char mass and operation time. The output values of ANNs model were the ratios Ct/C0 for naproxen adsorption.

189

Ct/Co Input neurons

Hidden Output neurons neurons

Fig. 1. Artificial neural network architecture used for modeling the naproxen adsorption on bone char at packed-bed columns.

The ANNs model included an input layer, one hidden layer with 10 neurons, and one output layer. This architecture offered the best performance without the data overfitting (i.e., overtraining). The classical backpropagation algorithm was employed for ANNs training and for determining the values of w and θ for each neuron given in Eq. (11). In summary, 142 experimental data were employed for training (70%), validation (15%) and testing (15%) of the ANNs model. Breakthrough data modeling was performed using the ANNs toolbox of MATLAB® with its default parameter values. 3. Results and discussion 3.1. Batch adsorption of naproxen and its kinetic and thermodynamic parameters Naproxen adsorption kinetics on bone char at different operating conditions (i.e., initial concentration, pH and temperature) are reported in Fig. 2. In general, the naproxen removal percentage is ~ 70% of the total uptake during the first 4 h and, at 12 h of the operating time, more than 92% of naproxen can be removed in the batch reactor. It is clear that naproxen adsorption was faster at the beginning of the batch experiments; its uptake increased with the operating time and reached the equilibrium at 24 h under the operating conditions tested. This adsorption behavior indicates that most of the naproxen adsorption may occur on the external surface of bone char where the binding sites could be more accessible to this organic molecule. As expected, the naproxen concentration in the solution favors the mass transfer driving force causing the increment in adsorption capacities, see Fig. 2a. Fig. 2b and c shows that naproxen uptake is significantly affected by solution pH and temperature. In particular, the solution pH affects the pollutant removal because the surface charge of the adsorbent may change [21]. Naproxen has a pKa value of 4.15 due to the carboxylic group present in its skeleton [14]. At tested operating conditions (pH N pKa), naproxen in aqueous solution is a negatively charged molecule, while bone char has a positive charge because the adsorption conditions are below its point of zero charge (i.e., pHz = 8.5). Therefore, a pH increment favors the naproxen removal (up to 25%) and, based on the organic nature of this molecule, it is expected that this adsorption process may involve both electrostatic and non-electrostatic interactions [14,20]. On the other hand, Fig. 2c shows that the temperature has also a significant impact on the naproxen removal. Note that the adsorption capacities increased in 32%–49% for tested temperature conditions. The increase in the mobility of naproxen molecules caused by temperature changes may enhance the interactions of adsorbent–adsorbate favoring the adsorption process [15]. Naproxen adsorption rates using bone char are reported in Table 1 including the results of kinetic data modeling. The values of k1 and k2 ranged from 0.43 to 1.07 h−1 and 0.13 to 1.04 g/mg h, respectively. In general, the pseudo-first order model showed the best correlation coefficients (R2) and a better performance than those obtained with the pseudo-second order model. The deviations between predicted and experimental kinetic data are given in Fig. 3a and the statistical analysis confirmed that the pseudo-first order model is the best for data correlation of the naproxen adsorption on bone char. Fig. 4 shows

190

H.E. Reynel-Avila et al. / Journal of Molecular Liquids 209 (2015) 187–195

3.5

a)

a)

3 2.5 2 1.5 1

0 0 3

5

10

15

20

qcalc mg/g

50 100 mg/L 200 Pseudo-first order model

C0

0.5

25

b)

b)

2.5

qt, mg/g

2 1.5 pH 1

5 6 7 Pseudo-first order model

0.5 0 0

5

10

15

20

25

qexp, mg/g

3.5

c)

3

Fig. 3. Experimental and calculated naproxen uptakes of bone char in batch reactors. a) Kinetic and b) Isotherm models.

2.5 2

the external surface of bone char. The intermediate stage can be attributed to the intraparticle diffusion; however, it is not the controlling step in naproxen adsorption according to the results of diffusion analysis. Finally, the third stage is attributed to the adsorption equilibrium [8,15]. Fig. 5 reports the naproxen adsorption isotherms for tested conditions of pH and temperature. These isotherms are L-type according to the Giles classification and correspond to a favorable adsorption process [38]. Fig. 5a confirmed that solution pH plays an important role in naproxen adsorption since it determines the predominant species in solution and the net charge on the adsorbent surface [20]. For pH changes from acidic to basic conditions, the naproxen adsorption capacities of bone char ranged from 2.5–3.2 mg/g and can increase up to 25%. Note that pH 7 is the best of the tested conditions for naproxen adsorption on bone char where the maximum experimental uptake is ≈3.2 mg/g. These results are consistent with the study of Nam et al. [39], who

1.5 T, °C 1

20 30 40 Pseudo-first order model

0.5 0 0

5

10

15

20

25

Time, h Fig. 2. Naproxen adsorption kinetics on bone char at batch reactors. Operating conditions: a) 30 °C and pH 7, b) 30 °C and 100 mg/L; and c) pH 7 and 100 mg/L.

the intraparticle diffusion analysis, which showed a multi-linearity behavior indicating that several stages could be involved in naproxen adsorption. The first stage corresponds to the naproxen diffusion to Table 1 Kinetic rates for naproxen adsorption on bone char from aqueous solutions at batch reactors.

Operating conditions of naproxen adsorption pH 7, 30 °C

pH 7, 30 °C

pH 7, 30 °C

pH 5, 30 °C

pH 6, 30 °C

pH 7, 20 °C

pH 7, 40 °C

Modela

C0, mg/L

50

100

200

100

100

100

100

Pseudo-first order   qt ¼ qe 1−e−k1 t

qe, mg/g k1, h−1 R2 E, % qe, mg/g k2, g/mg h R2 E, % qp, mg/ g h0.5 C, mg/g R2

1.120 1.073 0.94 5.52 1.240 1.037 0.913 6.44 0.0871 0.859 0.82

2.769 0.429 0.99 2.87 3.290 0.129 0.990 3.95 0.339 1.591 0.99

2.979 0.651 0.96 6.27 3.446 0.198 0.980 5.37 0.284 2.022 0.94

2.039 0.641 0.97 3.97 2.354 0.289 0.990 3.09 0.207 1.341 0.96

2.442 0.478 0.98 5.79 2.896 0.160 0.960 9.20 0.237 1.588 0.98

2.144 0.569 0.98 5.58 2.477 0.233 0.904 10.66 0.077 1.832 0.99

3.206 0.714 1.00 0.85 3.617 0.222 0.968 5.04 0.149 2.707 0.94

Pseudo-second order q2 k t

qt ¼ 1þke 22q

et

Intraparticle diffusion 1 q ¼ q t =2 þ C t

a

p

In all models, qt is the naproxen uptake (mg/g) at time t (h) and qe is the theoretical adsorption capacity (mg/g).

H.E. Reynel-Avila et al. / Journal of Molecular Liquids 209 (2015) 187–195

3.5

3.5 3

a)

a)

3

2.5

2.5

2

2

1.5

1.5

1

1

pH C0

0 0.5

3

1.5

2.5

50 100 mg/L 200

3.5

4.5

b)

5 6 7 Sips model

0.5

qe, mg/g

qt, mg/g

0.5

3.5

191

0 0 4

50

100

150

200

b)

3.5 2.5

3 2

2.5

1.5

2 T, °C

1 0.5 0 0.5

1.5

20 30 40 1.5

2.5

3.5

T, °C 20 30 40 Sips model

1 0.5 4.5

Time0.5, h

0 0

50

100

150

200

Ce, mg/L Fig. 4. Intraparticle diffusion analysis for the naproxen adsorption on bone char. Operating conditions: a) 30 °C and pH 7; and b) pH 7 and 100 mg/L.

have analyzed the removal of several pharmaceuticals including naproxen using activated carbon and the maximum adsorption capacity was obtained at pH 7. In particular, this result is relevant from a practical point of view based on the fact that water resources and wastewaters to be treated usually have a neutral pH. However, it is convenient to remark that the adsorption of organic molecules from dilute aqueous solutions on carbon-based materials is a complex process that may involve electrostatic and non-electrostatic interactions. The removal mechanism depends on the characteristics of the adsorbent, adsorbate and solution chemistry [20]. Naproxen is a pharmaceutical drug and consists of hydrophilic and hydrophobic moieties, see Fig. 6. The acidic and hydrophilic nature is due to the hydrogen of carboxylic group, which is ionisable at pH N pKa; while the hydrophobic behavior of this organic molecule is due to the aliphatic rings [17]. When solution pH increases, the naproxen molecule suffers a resonance structure change [17] and it gradually becomes more negatively charged [21]. These conditions favor its adsorption on bone char since it is expected that electrostatic interactions are present in this removal process (pKa of naproxen b pHsolution b pHz of bone char). Under these conditions, the bone char surface shows a positive charge thus promoting a hydrogen bonding. This hydrogen bonding is preferred at pH below pHz [20, 40]. The opposite charges of the naproxen–bone char system cause a coulombic attraction during adsorption process (i.e., an electrostatic interaction). Hydrophobic interactions could also be present since aromatic rings of naproxen may interact with the adsorbent surface. Similar findings have been reported for diclofenac removal on activated carbon [20]. Previous studies have established that the hydrophobic adsorbent–adsorbate interactions can be associated to driving forces caused by π–π electron dispersion interactions between aromatic rings of the adsorbate and the structure of carbon-based materials [12, 41]. Note that bone char also contains an amorphous carbon phase (~10%) that could be part of these interactions.

Fig. 5. Adsorption isotherms for the naproxen removal using bone char at batch reactors. Operating conditions: a) 30 °C and b) pH 7.

Fig. 5b shows the naproxen adsorption isotherms obtained at different temperatures and it is clear that an increase of temperature improves the naproxen uptake. Similar trends for the adsorption of naproxen and other pharmaceutical compounds using carbon-based adsorbents have been reported in previous studies [13,15,17,39]. In particular, Önal et al. [15] and Sarici-Ozdemir and Önal [13] have reported an increment in naproxen uptake by activated carbon with temperature. Similarly, Nam et al. [39] found that temperature influences the uptake of hydrophobic drugs (such as naproxen) on activated carbon. These authors stated that a low solution temperature can reduce the diffusion of molecules and inhibit their accessibility to the adsorbent structure. In addition, the increase in the temperature may affect the degree of hydration around the hydrophilic moiety and increase the solute hydrophobicity [42]. This phenomenon may favor the hydrophobic interactions between naproxen and bone char, thus enhancing the adsorption process. Thermodynamic parameters of naproxen adsorption on bone char are consistent with these findings, see Table 2. The negative ΔG° values indicate the feasibility and spontaneous nature of the naproxen adsorption process, while the positive ΔS° reflects the affinity of the bone char for naproxen molecule. ΔH° values confirmed the endothermic nature of the naproxen removal on bone char. Note that drug molecules have to lose part of their hydration sheath during adsorption and this process requires energy (i.e., it is an endothermic process) [12]. According to ΔH° values, it is expected that the naproxen adsorption on bone char could be caused by both physical and chemical processes. It is well known that the adsorption of aromatic compounds is partly physical and chemical [43]. Therefore, the naproxen uptake on bone char is a feasible thermodynamic process, which is more favorable at higher temperatures. Naproxen equilibrium data were fitted to Langmuir, Freundlich and Sips models and results of isotherm modeling are reported in Table 3 and Fig. 3b, respectively. The statistical analysis indicated that the Sips

192

H.E. Reynel-Avila et al. / Journal of Molecular Liquids 209 (2015) 187–195

a)

CH3

b)

CH3

O

O Hydrophilic group

H3C O

O

H3C O

O-H Hydrophobic groups

O

_

+

RO P OH O

Bone char surface CH3

c)

O O

H3C O

C O-H

O O -H

H3C O

O H+

CH3

d)

OH

+

Bone char surface

π-

H

π-

Bone char surface

Fig. 6. Description of a) naproxen molecule and b)–d) proposed adsorption mechanism using bone char.

model is the best isotherm equation, which showed high correlation coefficients R2: 0.97–0.99. This equation is an empirical model that includes the features of both Langmuir and Freundlich equations and can be used for reliable data modeling of naproxen adsorption on bone char in batch reactors. 3.2. Packed-bed adsorption of naproxen and its breakthrough parameters Experimental and modeled breakthrough curves for naproxen removal are reported in Fig. 7 for tested operating conditions. All the naproxen breakthrough curves show the common S-shape but they are unsymmetrical where their characteristics and shape depend on the bone char mass and feed concentration used in the dynamic adsorption experiments. Fig. 7a shows that the breakthrough time increases with the bone char mass packed in the column. In fact, breakthrough curves became steeper as the adsorbent mass decreased. This behavior is due to the increased residence time and the availability of more binding sites for naproxen removal with an increase in the mass of bone char. The effect of feed concentrations in naproxen dynamic adsorption is given in Fig. 7b–c. The breakthrough and exhaustion points occurred faster at high naproxen feed concentrations and the adsorption columns operating at the lowest feed concentration exhibited the higher operating times and treated solution volumes. As expected, the saturation rates of binding sites of the bone char surface depend on the feed concentration. It is interesting to remark that at low feed concentrations (i.e., 1 and 2 mg/L), the breakthrough curves show a region where the Ct/C0 profile is flat, which could be associated to the fact that the diffusion phenomena are the rate-limiting mass transport process in liquid-phase adsorption [36]. Table 4 contains the calculated breakthrough parameters for naproxen adsorption. These results indicate that an increase in the Table 2 Thermodynamic parameters for naproxen adsorption on bone char in batch reactors. T, °C

ΔG°, kJ/mol

ΔH°, kJ/mol

ΔS°, kJ/mol K

20 30 40

−0.33 −0.72 −1.22

75.11

0.23

bed depth (i.e., adsorbent mass) leads to an increment of the length of mass transfer zone (MTZ). However, slight increments of MTZ are observed when the feed concentration also increased. The retardation factors rf decreased with the feed naproxen concentration and, consequently, the molecules of this drug will move faster inside the adsorption columns with lower rf than those packed-bed columns with higher rf. The saturation rates, the breakthrough times and the overall adsorption zones depend on both the naproxen feed concentration and bone char mass used in packed-bed columns. In particular, the slow approach of Ct/C0 → 1.0 at the end of breakthrough curves impacts on the values of MTZ, Δt and the saturation times for naproxen removal, especially at lower feed concentrations. The most relevant results of dynamic experiments for naproxen removal are the bone char usage rates, which are necessary for determining the technical and economic feasibility of utilizing a dynamic adsorption process in water treatment. Specifically, the degree of adsorbent utilization in these columns ranged from 0.6 to 4% at tested operating conditions, see results reported in Table 4. These results indicated that ~95% of the theoretical bed adsorption capacity (i.e., the maximum adsorption capacity obtained at batch reactors) is not utilized during dynamic naproxen adsorption at tested operating conditions. In particular, the worst percentages of bed utilization were obtained with the lowest naproxen feed concentration (1 mg/L). Based on the fact that diffusion phenomena are affected by hydraulic conditions of adsorption columns and the equilibrium is usually not reached in these separation systems, it is expected that the adsorbent performance of fixed-bed columns is lower than those obtained with batch reactors and low values of Fp are obtained especially for low pollutant concentrations [24,36]. The best degree of bed utilization was reached using the highest feed concentration: 10 mg/L. Therefore, the adsorption columns packed with bone char appear to be not suitable for the treatment of diluted naproxen solutions while the batch reactor configuration is more adequate for this condition. Although, the naproxen uptakes of bone char obtained in batch reactors are competitive with respect to other adsorbents such as conventional activated carbons — 0.7 mg/g [16] or magnetic nanoparticles coated zeolites — 0.5 mg/g [21], the application of this adsorbent on packed-bed columns is not effective for naproxen removal. Alternatives to improve the bone char utilization in dynamic adsorption systems include the increment of bed length (i.e., adsorbent

H.E. Reynel-Avila et al. / Journal of Molecular Liquids 209 (2015) 187–195

193

Table 3 Results of adsorption isotherm modeling for naproxen removal using bone char in batch reactors. Operating conditions of naproxen adsorption Modela Langmuir

qm, mg/g B, L/mg R2 E, % kf, L/g n R2 E, % qs, mg/g as n R2 E, %

qm bC e qe ¼ 1þbC e

Freundlich 1

qe ¼ K f C e

=n

Sips q a Cn

s s e qe ¼ 1þa Cn s

a

e

pH 5, 30 °C

pH 6, 30 °C

pH 7, 30 °C

pH 7, 20 °C

pH 7, 40 °C

4.279 0.010 0.950 7.16 0.122 1.64 0.822 14.74 3.110 0.004 1.427 0.991 2.98

4.008 0.018 0.938 7.75 0.227 1.96 0.775 14.32 3.339 0.009 1.306 0.974 5.53

3.531 0.053 0.993 1.84 0.624 3.03 0.877 9.21 3.659 0.060 0.929 0.994 1.59

4.608 0.008 0.836 15.91 0.088 1.05 0.648 21.83 2.710 0.001 2.005 0.990 3.68

4.191 0.054 0.963 2.70 0.744 3.02 0.942 8.47 4.693 0.072 0.814 0.978 2.25

In all models, qe is the adsorption capacity at equilibrium (mg/g) and Ce is the equilibrium concentration (mg/L).

1 0.8

a)

0.6 W 9g

0.4

17 g 0.2

Thomas model

C0=2 mg/L

ANNs model 0 0

100

200

300

400

500

600

700

1

b)

0.8

Ct/C0

mass). In summary, these results highlight the importance of analyzing, in an integrated way, the performance of both batch reactor and packed-bed column configurations for selecting the best system and performing its proper design. Results of the Thomas model for breakthrough data modeling are reported in Table 5. Overall, this model showed an unsatisfactory performance for modeling some naproxen adsorption breakthrough curves and its correlation coefficients ranged from 0.36 to 0.96. These results also confirmed that the modeling of unsymmetrical breakthrough curves with Thomas equation can fail, which is consistent with previous studies [36]. In contrast, ANNs model offered the best performance for data correlation of all breakthrough curves, which have been modeled simultaneously with this black-box approach. Fig. 7 shows that this ANNs model is flexible and reliable for correlating and predicting the performance of packed-bed adsorption columns at different operating conditions. The correlation coefficient of ANNs model is better than

Table 4 Parameters of breakthrough curves for naproxen adsorption on bone char using packed bed columns.

0.6 C0 2 mg/L

0.4 0.2

W=9 g L/D=2

4 mg/L

Operating conditions

10 mg/L

C0, mg/L

W, g

qexp bed, mg/g

Δt, min

MTZ, cm

Fq

rf

Thomas model

1 2 2 4 4 10 10

17 9 17 4.5 9 4.5 9

0.019 0.043 0.052 0.105 0.048 0.113 0.087

218 203 189 299 163 78 239

5.41 3.16 5.67 1.99 3.36 1.94 3.59

0.006 0.013 0.016 0.033 0.015 0.035 0.027

78.59 53.99 63.68 28.95 8.76 11.02 8.47

ANNs model

0 0

50

100

150

200

250

300

1

Breakthrough parameters of naproxen adsorption

c)

0.8 0.6

Table 5 Results of data modeling of breakthrough curves of the naproxen adsorption on bone char using Thomas and ANNs models.

C0 0.4

1 mg/L

0.2

W=17 g L/D=4

0 0

100

200

300

400

500

2 mg/L

Operating conditions

Thomas model

C0, mg/L

W, g

k, mL/mg min

qbed, mg/g

R2

E, %

R2

E, %

ANNs model

1 2 2 4 4 10 10

17 9 17 4.5 9 4.5 9

36.454 140.727 25.464 23.001 110.557 86.899 35.636

0.022 0.025 0.037 0.057 0.024 0.046 0.021

0.93 0.83 0.96 0.88 0.36 0.95 0.72

27.55 23.77 11.19 11.44 14.75 3.34 7.69

0.98

9.97

600

700

Time, h Fig. 7. Breakthrough curves for naproxen adsorption on bone char at 30 °C and pH 7 using packed-bed columns.

Thomas model

ANNs model

194

H.E. Reynel-Avila et al. / Journal of Molecular Liquids 209 (2015) 187–195

1

ANNs model

0.8

Calculated Ct /C0

a)

Thomas model

0.6

0.4

0.2

0 0.6

0.8

1

Fig. 8. Performance of Thomas and ANNs model for breakthrough data correlation of the naproxen adsorption on bone char in packed-bed columns.

those obtained with the Thomas model and its deviations between predicted and experimental breakthrough curves are lower, see Table 5 and Fig. 8. Note that ANNs model is an alternative for adsorption data modeling when traditional models are inadequate and failed for data analysis. 3.3. Bone char characterization and its naproxen removal mechanism The characterization results of bone char used in naproxen removal experiments are reported in Fig. 9. In particular, Fig. 9a shows the SEM image of the adsorbent surface morphology, which is roughness. Nitrogen adsorption isotherm of Fig. 9b can be classified as type IV and it is a characteristic of micro-mesoporous solids according to the IUPAC classification. The hysteresis loop clearly shows the presence of mesoporosity [44,45]. Surface area estimated with the BET equation is 74 m2/g with a microporous volume of 0.026 cm3/g (28%) and a mesoporous volume of 0.067 cm3/g (72%). The pore size distribution was calculated using the QSFDT model. The bone char is essentially a mesoporous material with 2–50 nm pore size, see the insert in Fig. 9b. Fig. 9c shows all the FTIR characteristic absorption peaks of hydroxyapatite, which is the main component of bone char. These bands appear at 3430, 1630, 1458, 1040, 604 and 567 cm−1 [26,46,47]. The band at ~3430 cm−1 is characteristic of the stretching mode of OH groups of hydroxyapatite [26,47]. The main indication of hydroxyapatite is the strong broad band at ~1040 cm−1, which is associated to the asymmetric stretching vibration mode of PO4 group [26]. Also, the bands at 604, 567 and 465 cm−1 correspond to P–O mode vibration of the PO4 group [48,49]. The peak at 1385 is assigned to the vibration of the carbonate group [26,48]. The band at 1458 cm−1 is related to the vibration of the carboxylic group [50]. Remaining bands are associated to the vibrations of aromatic rings such as quinones and lactones in the carbon phase of bone char [50,51]. After naproxen adsorption, some changes were observed in the FTIR spectrum of bone char. The peaks of phosphate (567–604 and 1040 cm−1) changed and their intensity was lower than those obtained for raw bone char. This result could suggest some role of phosphate group in the binding of naproxen. On the other hand, the bands of oxygen groups decreased especially those at 1740 and 1630 cm−1 that represent aromatic groups [50,51]. Additionally, there are new bands at 702, 1220 and 1338 cm−1, which may support the interaction between naproxen and the adsorbent surface [52,53]. Note that the chemical structure of naproxen contains an aromatic ring besides carboxylic and ether groups [54]. In particular, the new peak at 1220 cm−1 corresponds to the aryl-O stretching vibration of the aromatic ring [54], which is consistent to the chemical structure of the drug molecule. In summary, these findings could confirm the presence of naproxen in bone char surface. The proposed removal mechanism and possible

b) dV, cc/A/g

0.4

Experimental Ct /C0

Adsorbed amount of N2, cm3/g

0.2

Pore Width, Å

Relative Pressure, P/P0

c)

Before naproxen adsorption

Transmittance, %

0

After naproxen adsorption

Wavenumber, cm-1 Fig. 9. Results of bone char characterization: a) SEM image; b) Nitrogen adsorption isotherm; c) FTIR analysis.

naproxen–bone char interactions are reported in Fig. 6 where: 1) Fig. 6b presents a complexation process [55] via phosphate– naproxen interactions, 2) Fig. 6c shows hydrogen bonds, which is a form of a dipole–dipole interaction [20,40] and 3) Fig. 6d describes hydrophobic interactions via π–π electron between aromatic rings of naproxen and the aromatic rings of the carbon phase of bone char [12,41,51]. 4. Conclusions In this study, the integrated analysis of the naproxen adsorption on bone char using batch reactors and packed-bed columns has been performed. Results indicated that the naproxen adsorption on bone char is significantly affected by solution pH and temperature. The maximum naproxen uptake was obtained at pH 7 and 40 °C in batch reactor and it corresponds to 3.2 mg/g. Thermodynamic calculations suggested that naproxen adsorption may be controlled by a combination of both physical and chemical mechanisms where both electrostatic and non-

H.E. Reynel-Avila et al. / Journal of Molecular Liquids 209 (2015) 187–195

electrostatic interactions are involved, depending mainly on solution pH. These interactions were confirmed via FTIR analysis. On the other hand, the naproxen adsorption breakthrough curves are unsymmetrical and the performance of bone char depends significantly on the dynamic operating conditions. The best filter usage in packed columns was ≈4% for the highest feed naproxen concentration. Consequently, bone char packed-bed columns are not effective for the treatment of diluted naproxen solutions and the batch adsorption reactors emerge as the better option. In summary, the naproxen adsorption on bone char using batch reactor appears to be a feasible process from both economic and technical points of view, especially for diluted naproxen solutions. This study shows the importance of evaluating an integrated analysis of adsorbent performance in batch and dynamic conditions for determining the best process configuration for the removal of emerging water pollutants such as pharmaceutical compounds. References [1] Y. Chen, Y. Zhu, Z. Wang, Y. Li, L. Wang, L. Ding, X. Gao, Y. Ma, Y. Guo, Adv. Colloid Interface Sci. 163 (2011) 39. [2] G. Mezohegyi, F.P. Van der Zee, J. Font, A. Fortuny, A. Fabregat, J. Environ. Manag. 102 (2012) 148. [3] S.P. Dubey, A.D. Dwivedi, C. Lee, Y. Kwon, M. Sillanpaa, L.Q. Ma, J. Ind. Eng. Chem. 29 (2014) 1126. [4] O. Aktas, F. Cecen, Int. Biodeterior. Biodegrad. 59 (2007) 257. [5] W.A.W. Daud, A.H. Houshamnd, J. Nat. Gas Chem. 19 (2010) 267. [6] J. Rivera-Utrilla, M. Sánchez-Polo, V. Gómez-Serrano, P.M. Alvarez, M.C.C. AlvimFerraz, J.M. Dias, J. Hazard. Mater. 187 (2011) 1. [7] A. Bhatnagar, W. Hogland, M. Marques, M. Sillanpaa, Chem. Eng. J. 219 (2013) 499. [8] D.I. Mendoza-Castillo, A. Bonilla-Petriciolet, J. Jáuregui-Rincón, Desalin. Water Treat. 54 (2015) 1651 (in press). [9] C. Moreno-Castilla, J. Rivera-Utrilla, MRS Bull. 26 (2001) 890. [10] K.Y. Foo, B.H. Hameed, J. Hazard. Mater. 170 (2009) 552. [11] H.A. Omar, H. Moloukhia, J. Hazard. Mater. 157 (2008) 242. [12] R. Baccar, M. Sarra, J. Bouzid, M. Feki, P. Blánquez, Chem. Eng. J. 211–212 (2012) 310. [13] C. Sarici-Ozdemir, Y. Önal, Chem. Eng. Process. 49 (2010) 1058. [14] H. Bagheri, A. Roostaie, M.Y. Baktash, Anal. Chim. Acta 816 (2014) 1. [15] Y. Önal, C. Akmil-Basar, C. Sarici-Ozdemir, J. Hazard. Mater. 148 (2007) 727. [16] Z. Yu, S. Peldszus, P.M. Huck, Water Res. 42 (2008) 2873. [17] A. Dwivedi, K. Gopal, J. Rajeev, Chem. Eng. J. 168 (2011) 1279. [18] J.C. Durán-Álvarez, B. Prado-Pano, B. Jiménez-Cisneros, Chemosphere 88 (2012) 84. [19] Y. Kim, J. Bae, J. Park, J. Suh, S. Lee, H. Park, H. Choi, Chem. Eng. J. 256 (2014) 475. [20] J. Sotelo, A. Rodríguez, S. Alvarez, J. García, Chem. Eng. Res. Des. 90 (2012) 967.

195

[21] T. Salem, X. Hu, Y. Qiang, Chemosphere 93 (2013) 2076. [22] S. Prabha, A. Dwivedi, C. Lee, Y. Kwon, M. Sillanpaa, L. Ma, J. Ind. Eng. Chem. 20 (2014) 1126. [23] K.S. Rao, S. Anand, P. Venkateswarlu, J. Ind. Eng. Chem. 17 (2011) 174. [24] I.A. Aguayo-Villarreal, A. Bonilla-Petriciolet, V. Hernández-Montoya, M.A. MontesMorán, H.E. Reynel-Avila, Chem. Eng. J. 167 (2011) 67. [25] C.W. Cheung, J.F. Porter, G. McKay, Water Res. 35 (2001) 605. [26] C.K. Rojas-Mayorga, A. Bonilla-Petriciolet, I.A. Aguayo-Villarreal, V. HernándezMontoya, M.R. Moreno-Virgen, R. Tovar-Gomez, M.A. Montes-Morán, J. Anal. Appl. Pyrolysis 104 (2013) 10. [27] A.H. Hyder, S. Begum, O. Egiebor, J. Environ. Chem. Eng. (2015) (in press). [28] S. Lagergren, K. Svenska, Vet. Akad. Handl. 24 (1898) 1. [29] Y.S. Ho, G. McKay, Process Biochem. 34 (1999) 451. [30] W.J. Weber, J.C. Morris, J. Sanit. Eng. Div. 89 (1963) 31. [31] H. Freundlich, Z. Phys. Chem. 57 (1906) 385. [32] I. Langmuir, J. Am. Chem. Soc. 40 (1981) 1361. [33] S. Sips, J. Chem. Phys. 16 (1948) 490. [34] H.C. Thomas, J. Am. Chem. Soc. 66 (1944) 1664. [35] I.A. Basheer, M. Hajmeer, J. Microbiol. Methods 43 (2000) 3. [36] R. Tovar-Gómez, M.R. Moreno-Virgen, J.A. Dena-Aguilar, V. Hernández-Montoya, A. Bonilla-Petriciolet, M. Montes-Morán, Chem. Eng. J. 228 (2013) 1098. [37] H.E. Reynel-Avila, A. Bonilla-Petriciolet, G. de la Rosa, Desalin. Water Treat. 1 (2014) 1. [38] C.H. Giles, D. Smith, I. Theoretical, J. Colloid Interface Sci. 47 (1974) 755. [39] S. Nam, D. Choi, S. Kim, N. Her, K. Zoh, J. Hazard. Mater. 270 (2014) 144. [40] N. Suriyanon, P. Punyapalakul, C. Ngamcharussrivichai, Chem. Eng. J. 214 (2013) 208. [41] O. Hamdaoui, E. Naffrechoux, J. Hazard. Mater. 147 (2007) 381. [42] J. Maszkowska, M. Wagil, K. Mioduszewska, J. Kumirska, P. Stepnowski, A. Bialk, Chemosphere 111 (2014) 568. [43] C. Moreno-Castilla, Carbon 42 (2004) 83. [44] E.K. Putra, R. Pranowo, J. Sunarso, N. Indraswati, S. Ismadji, Water Res. 43 (2009) 2419. [45] D. Saha, K. Warren, A. Naskar, Carbon 71 (2014) 47. [46] S. Lurtwitayapont, T. Srisatit, Environ. Asia 3 (2010) 32. [47] S. Mondal, B. Mondal, A. Dey, S. Mukhopadhyay, J. Miner. Mater. Charact. Eng. 11 (2012) 55. [48] F. Mohandes, M. Salavati-Niasari, M. Fathi, Z. Fereshteh, Mater. Sci. Eng. 45 (2014) 29. [49] K. Azzoui, E. Mejdoubi, A. Lamhamdi, S. Zaoui, M. Berrabah, A. Elidrissi, B. Hammouti, M. Fouda, S. Al-Deyab, Carbohydr. Polym. 111 (2015) 170. [50] C.K. Rojas-Mayorga, J. Silvestre-Albero, I.A. Aguayo-Villareal, D.I. Mendoza-Castillo, A. Bonilla-Petriciolet, Microporous Mesoporous Mater. 209 (2015) 38. [51] G. Kysas, N. Lazaridis, E. Deliyanni, Chem. Eng. J. 234 (2013) 491. [52] P. Parimoo, A. Bharathi, D. Mamatha, Asian J. Chem. 8 (1996) 247. [53] L. Nielsen, M. Biggs, W. Skinner, T. Bandosz, Carbon 80 (2014) 419. [54] A. Sharmin, P. Shovan, H. Ikramul, A. Navid, S. Syed, R. Selim, J. Pharm. Sci. 12 (2013) 11. [55] N.A. Medellín-Castillo, R. Leyva-Ramos, E. Padilla-Ortega, R. Ocampo-Perez, J. FloresCano, M. Berber-Mendoza, J. Ind. Eng. Chem. 20 (2014) 4014.