- Email: [email protected]
Filtration and removal performances of membrane adsorbers
Accepted Manuscript Title: Filtration and Removal Performances of Membrane Adsorbers Authors: Yilmaz Yurekli, Mehmet Yildirim, Levent Aydin, Melih Savran PII: DOI: Reference:
S0304-3894(17)30149-8 http://dx.doi.org/doi:10.1016/j.jhazmat.2017.02.061 HAZMAT 18416
To appear in:
Journal of Hazardous Materials
Received date: Revised date: Accepted date:
25-8-2016 20-12-2016 28-2-2017
Please cite this article as: Yilmaz Yurekli, Mehmet Yildirim, Levent Aydin, Melih Savran, Filtration and Removal Performances of Membrane Adsorbers, Journal of Hazardous Materials http://dx.doi.org/10.1016/j.jhazmat.2017.02.061 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Filtration and Removal Performances of Membrane Adsorbers
Yilmaz YUREKLI1*, Mehmet YILDIRIM2, Levent AYDIN3, Melih SAVRAN3
1
Department of Bioengineering, Celal Bayar University, Muradiye Kampusu, 45140, Yunusemre, Manisa, Turkey
2
Department of Metallurgical and Materials Engineering, Celal Bayar University, Muradiye Kampusu, 45140,
Yunusemre, Manisa, Turkey Mechanical Engineering Department, Izmir Katip Çelebi University, Balatçık Kampusu, Çiğli, Izmir, Turkey
3
*
Corresponding author: [email protected]
1
Highlights
NaX nanoparticles added PSf membrane was successfully synthesized and characterized by means of cross-flow filtration module. High recovery and reusability was attained for both of the lead and nickel cations. Nonlinear rational regression model perfectly described the whole filtration data. Lower initial metal concentrations, temperatures and pressures were selected to make the target response minimum.
Abstract Membrane adsorbers are promising candidates for the efficient and effective removal of heavy metals in waste water due to their unattainable adsorption and filtration capabilities. In the present study, zeolite nanoparticles incorporated polysulfone (PSf10) membrane was synthesized by means of non-solvent induced phase separation technique for the removal of lead and nickel ions in water. PSf10 showed a remarkable sorption capability and after repeated (adsorption/desorption)5 cycles in batch experiments, it preserved 77% and 92% of its initial sorption capacity for the lead and nickel, respectively. Addition of nanoparticles increased the pore radius of the native PSf from 10 to 19 nm, while bovine serum albumin rejection remained unchanged at 98%. Increments in the pore size and enhancement in hydrophilicity caused to increase hydraulic permeability of the native PSf from 23 to 57 L/m2.h.bar. Cross-flow filtration studies revealed that the filtrate concentrations were inversely affected by the initial metal concentration and transmembrane pressure due to reaction limited region. Nonlinear rational regression model perfectly described the filtration behavior of the PSf10 within the experimental range and suggested that lower initial metal concentration and pressure with a short filtration time should be selected for the target response to be minimum. Keywords: Cross-flow filtration, Heavy metals removal, Membrane adsorber, Rational regression
2
1. Introduction Strict regulations on the discharge levels of the industrial effluents for sustainability of clean water resources have pressurized the companies and researchers to develop more advanced functional materials that satisfy the criteria proposed by the environmental and health organizations. Nowadays, hybrid membranes prepared by impregnation of different types of inorganic nano-sized materials onto/into polymeric matrixes have been attracted attention due to their unattainable sorption capacities together with enhanced filtration performances [1-3]. Thereby, high energy cost of the membrane separation processes as in the case of nanofiltration or reverse osmosis can be reduced without altering the membrane selectivity by utilizing the hydrophilic nature of the adsorbents dispersed on/in the membrane. Besides elimination of difficulties in separation of the nano-adsorbents from treated water, hybrid membranes offer effective solution for removal of suspended charged solids and high molecular weight solutes (screening effect) presumably available in wastewater that may cause blocking the active sites of the adsorbents when they are used in direct contact mode in a simple batch adsorption process. In addition, the presence of charged nanoparticles in the membrane improves various properties like thermal [4], mechanical [5], diffusive [6] and electrostatic. Moreover, they introduce unique functionalities such as photocatalytic, antibacterial or adsorptive capabilities [7-11]. Thus, membrane adsorption is more eligible over conventional processes due to favorable hydrodynamic, high removal efficiency, acceptable reusability and small footprint. Polysulfone is one of the most popular polymer for membrane casting due to its well-known excellent mechanical, thermal and chemical stability. However, because of its hydrophobic nature, wide range of its application in membrane separation processes has been restricted. To increase its hydrophilicity, addition of inorganic fillers into the membrane has been accepted as an efficient alternative way [12]. Among the inorganic materials, zeolite is a good candidate for the polymer membranes because of its hydration effect and advanced ion exchange performance. Zeolites have well-defined porous structures and offer mobility of alkali and alkaline earth metals, in order to compensate net negative charge between Si4+ and Al3+ in the framework makes zeolites excellent adsorber for the removal of many target solutes [13-15]. Meanwhile, nanoparticles known as high efficient adsorbents are replaced with micrometer-sized counterparts due to advanced specific surface area and interfacial activity. The unique properties of the zeolite nano-particle such as high 3
ion exchange capacity and fast adsorption rate makes it excellent choice for the separation of heavy metals in wastewater treatment [16] and salts in desalination applications [17]. Response surface methodology (RSM), a flexible statistical tool can be applied to many systems to optimize the target response by suitably indexing the linear, quadratic and interaction effects of governing process parameters. It provides an empirical model to yield precisely configured experimental trials within the dictated ranges of variables thereby, reduces the time expended on the required analyses. For example, it has been used for optimization process variables of polymer enhanced ultrafiltration in order to maximize dye rejection efficiency [18]. In another study, the preparation conditions of the polymeric hydrogels have been optimized to gain the maximum adsorption capacity for Pb2+ and Cu+2 [19]. After the design of experiments for the chemical process, generally nonlinear regression models can be used to predict the phenomenological response of the process. This approach allows to write exact objective functions and constraints including design variables for mathematical optimization problems. Successful operation of membrane systems can only be achieved by using suitable membranes and determining their optimum operational conditions. In our previous study, the effects of membrane preparation conditions including NaX content and evaporation time during phase inversion process on the water hydraulic permeabilities and metal removal capacities of the membranes were elaborated [16]. Current study mainly focuses on the optimization of operating conditions and identification of transfer regimes during filtration of metal solutions. In this respect, the same membrane architecture prepared in the previous study was used to elucidate the effect of various process parameters in which case cross-flow mode of operation was selected since it provides more realistic data and simulates well the industrial processes. Before filtration, batch adsorption experiments which are important to explain the removal mechanisms of metal ions were carried out by using native and NaX added PSf membranes. Reusability of the membrane was also investigated by successive adsorption/desorption cycles. The performance of the hybrid membrane was then examined under variable transmembrane pressures, initial metal concentrations and temperatures. Metal concentrations in filtrate and retentate at each conditions were determined as a function of time. The raw data were then used in nonlinear multiple regression to build a rational model equation. The significance of each interaction parameters used in model equation were further analysed using analyses of variance (ANOVA) statistical tool. 4
2.
Materials and Methods
2.1. Materials Polysulfone beads (Mn=22,000 g/mol) and N-methyl-2-pyrrolidone (NMP) for the preparation of ultrafiltration membranes were supplied by Sigma-Aldrich. During production of zeolite nanoparticles, fumed silica having 0.007 m particle size and sodium aluminate from Sigma-Aldrich were used. Sodium hydroxide pellets was purchased from Merck and nickel (II) chloride (NiCl2) and lead (II) nitrate (Pb(NO3)2) solutions as atomic spectroscopic standards were supplied from Fluka. Bovine serum albumin (BSA) used for the rejection experiments was supplied by Aldrich. Molecular dimension of BSA is given by the supplier as 40 x 140 Å. Necessary dilutions were performed with Milli-Q water having resistivity higher than 18 M.cm. 2.2. Methods 2.2.1. Preparation of NaX Added PSf Membranes Asymmetric PSf membrane loaded with 10w% of NaX nanoparticles was prepared based on nonsolvent-induced phase separation (NIPS) technique. PSf10 membrane was obtained by directly immersing the polymer casting solution into coagulation bath at 4˚C and the details in the experimental procedure were given in our previous study [16]. 2.2.2. Determination of BSA Rejection Rejection experiment was performed by filtration of 1 mg/mL bovine serum albumin (BSA) solution prepared in 22 mM sodium phosphate buffer at pH 5 which is close to isoelectric point of BSA (IEPBSA 4.9). Dead-end stirred cell at 1 bar transmembrane pressure was used. Protein concentrations in the feed and filtrate were measured using UV/VIS spectrophotometer (Thermo evolution 201 model) at a fixed wavelength of 278 nm. Rejection values were then calculated based on the expression given below: % 𝑟𝑒𝑗𝑒𝑐𝑡𝑖𝑜𝑛 =
𝐶𝑓 −𝐶𝑝 𝐶𝑓
𝑥100
(1)
where, Cf and Cp denote the protein concentrations in the feed and permeate sides, respectively.
5
2.2.3. Batch Adsorption Ion exchange performance of the NaX added PSf membrane was initiated by introducing a small piece of composite membrane with 20 mL of 500 mg/L lead or nickel solutions. During the reaction a magnetic stirrer with a constant stirring rate of 300 rpm was used and the temperature was maintained constant at 25˚C. The change in metal concentrations in the solution was monitored by sampling at specified time intervals and they were analyzed using atomic absorption spectroscopy (Perkin Elmer AAnalyst 800 with air-acetylene oxidizing flame). The same procedure was followed with the native PSf membrane as a control. The amount of metal ions exchanged with cations in the zeolite frameworks at each time interval was calculated by the following equation: 𝑞𝑡 =
(𝐶0 −𝐶𝑡 )∙𝑉
(2)
𝑆
where, qt (mg/cm2), is the amount of metal ions adsorbed per unit surface area of the membrane S (cm2), C0 and Ct (mg/L), are the liquid phase metal concentrations at initial and at any time t, (min), respectively and V (L) is the permeate volume. According to our previous study, the amount of NaX available in a unit surface area of the PSf10 membrane was determined as 1.08 mg/cm2. Reusability of the hybrid membrane is an important factor which affects directly the cost of the process and decreases the environmental pollution in such a way that prolongs the shelf life of the membrane and hence, lowers the amount of waste materials. Reusability of the hybrid membrane was tested by series of adsorption and desorption experiments. Desorption process was carried out with the membrane which was previously saturated with one of the metal ions (lead or nickel). In order to release all the metal ions adsorbed by the membrane, an excess amount of NaCl (5 M) was used. Desorption was started by inserting the membrane into 50 mL of NaCl solution maintained at 25˚C. During 120 min of desorption period samples were withdrawn at the prescribed time intervals and analysed for their metal contents by using AAS. At the end of desorption process, the membrane was rinsed and then stored in water at 4˚C at least one day. Determination of removal rate of heavy metals through hybrid membrane provides a satisfactory design for a separation unit. In general, two or more mechanisms influence the rate of metal 6
adsorption by the hybrid membrane. The first one is the convective flow of the metal ions from the bulk solution to the nearby membrane interface. This is known as film type of mass transfer resistance and can be eliminated by applying adequate stirring rate. The second one is the diffusional flow of the metal ions through the membrane pores and through the pores of the NaX nanoparticles and the third one is the ion exchange reaction which is known to be occurred instantaneously. In order to identify the transport mechanism, one of the important dimensionless number called Thiele modulus, , can be used. From its definition, rate of reaction is compared with the mass transfer rate expressed in Equation 3 and 4 for the first order and second order of reactions, respectively. 𝜅1 = (
𝑘1,𝑖 𝐷𝑒𝑓𝑓,𝑖
𝜅2 = (
1/2
)
𝐿𝑚
𝑘2,𝑖 𝑆𝑎 𝜌𝑁𝑎𝑋 𝐶0 𝐷𝑒𝑓𝑓,𝑖
(3) 1/2
)
𝐿𝑚
(4)
where, k1,i (min-1), Deff,i (m2/min) and Lm (m) are defined as the first order reaction rate constant, effective diffusion coefficient of metal ions in the membrane and membrane thickness (137 m), respectively and k2,i (cm2/mg.min), Sa (cm2/g) and NaX (g/cm3) denote second order reaction rate constant, specific membrane surface area and density of NaX (0.5 g/cm3), respectively. A high value of implies that the system is mass transfer limited while in the case of small values of , the process is controlled by reaction. The mechanism of ion-exchange can be described by the reaction given below; 2+ + 2+ + 𝑀(𝑎𝑞) + 2𝑁𝑎(𝑧𝑒𝑜𝑙𝑖𝑡𝑒) ↔ 𝑀(𝑧𝑒𝑜𝑙𝑖𝑡𝑒) + 2𝑁𝑎(𝑎𝑞)
(5)
where, M2+ is a metal either in liquid phase or on the adsorbed phase. Lagergren pseudo-second order kinetic model is mostly suitable while describing this type of reaction. However, under experimental conditions one of the reactants can be limited, then the reaction follows pseudo-first order kinetics [20]. Therefore, the rate constants in Equation 3 and 4 were estimated from Lagergren pseudo first order (Equation 6) and second order (Equation 7) model equations, respectively. 𝑞𝑡 = 𝑞𝑒 (1 − 𝑒𝑥𝑝−𝑘1,𝑖𝑡 )
(6) 7
𝑞𝑡 =
𝑞𝑒2 𝑘2,𝑖 𝑡
(7)
1+𝑞𝑒 𝑘2,𝑖 𝑡
where, qe (mg/cm2) represent the adsorbed amount of metal ions in a unit surface area of the membrane at equilibrium. A nonlinear regression analysis by minimization of the sum of the squares of the residuals between experimental and theoretical kinetic data was performed by using the expression given below. (𝐸𝑟𝑟𝑜𝑟)𝑚𝑖𝑛 = ∑((𝑞𝑡 )𝑒𝑥𝑝𝑒𝑟𝑖𝑚𝑒𝑛𝑡𝑎𝑙 − (𝑞𝑡 )𝑡ℎ𝑒𝑜𝑟𝑒𝑡𝑖𝑐𝑎𝑙 )2
(8)
The effective solute diffusivities in the membrane are calculated by multiplying the free solution diffusion coefficients (DPb,=9.4510-10 m2/s and DNi,=6.7910-10 m2/s) with the partition coefficient,i, porosity, and diffusive hindrance factor, Ki,D. 𝐷𝑒𝑓𝑓,𝑖 = 𝜀 𝜙𝑖 𝐾𝑖,𝐷 𝐷𝑖,∞
(9)
The solute diffusive hindrance factor is a function of the ratio between the solute and the pore diameters (𝜆𝑖 = 𝑑𝑖,𝑠 /𝑑𝑝 ) [21]. If parabolic fully developed solute flow within the pore is assumed Ki,D can be defined as [22]: 𝐾𝑖,𝐷 = 1.0 − 2.30𝜆𝑖 + 1.154𝜆2𝑖 + 0.224𝜆3𝑖
(10)
The partition coefficient in Equation 9 is calculated by assuming spherical solutes in cylindrical pores and the expression is given below. 𝜙𝑖 = (1 − 𝜆𝑖 )2
(11)
2.2.4. Cross-Flow Filtration Filtration and metal removal performances of the PSf10 membrane was determined under dynamic conditions using cross-flow filtration system (Sepa CF cell, Sterlitech Corp., USA) with an holdup volume of 70 mL and an active surface area of 140 cm2. The experimental set-up is schematically illustrated in Figure 1. Flat sheet membrane with a feed spacer is placed inside the stainless steel filtration module and the cell body is installed into the cell holder. After pressurizing 8
the cell holder with a hydraulic pump up to 1000 psi, feed solution can be pumped from a temperature controlled feed tank (5 L) to the filtration module. The pressure of the feed solution is controlled with an increment of ±0.1 bar by adjusting the frequency of the feed pump. Volumetric flow rate of the retentate which is recycled into the feed tank is followed by a manual rotameter. Filtrate solution is collected in a beaker placed on a digital balance from which data acquisition is supplied to a computer at each 2 second. Figure 1 Throughout the filtration experiments, following procedure was applied. First, membrane was compacted twice with water at 3 bars for 20 min. If those two fluxes were close to each other, then, the pressure was lowered to 0.5 bar. Water filtration at constant pressure for 20 min was performed. By successive increasing the transmembrane pressure, water permeation at higher pressures were also measured. Filtration data was then used to calculate water fluxes, Jv (L/m2.h) by dividing the volumetric flowrate, Vp (L/h) to the effective membrane area S (m2). 𝐽𝑣 =
𝑉𝑝
(12)
𝑆
Hydraulic permeability, Lp (L/m2.h.bar), through the membrane was determined from the slope of the plot of the solution flux as a function of transmembrane pressure. 𝐿𝑝 =
𝐽𝑣
(13)
∆𝑃
Filtration of the metal solutions were carried out under different initial concentrations (50-200 ppm), temperatures (25-50˚C) and transmembrane pressures (0.5-2 bar) for 120 min. At certain times, samples were withdrawn from the filtrate and retentate. After necessary dilutions, sample concentrations were determined by AAS. 2.2.5. Regression Analysis Regression analysis is a statistical tool for estimating the relationships among the parameters which affect the physical phenomena. Several types of regression models are available in literature: Linear, Logistic, Nonlinear, and Stepwise. Although some of engineering processes may be described well by utilizing the linear models, there are many other inherently nonlinear processes 9
including more general class of functions. Examples of nonlinear functions are logarithmic, trigonometric, power, and rational functions [23]. Rational functions can be seen in many theoretical formulae in chemistry, chemical engineering, and biochemistry, and also have been shown to be a potential method for nonlinear empirical modeling [24]. Some of the advantages of the rational regression models are (i) having better interpolatory properties than polynomial models, (ii) having very good asymptotic properties, (iii) and they can be used to model complicated structure with a low degree in both the numerator and denominator. This in turn means that fewer coefficients will be required compared to the polynomial model. In contrast to conventional polynomials, rational regression models can be bounded. As a consequence, they can be used to describe some chemical effects, in cases where the response can be limited in a finite amount. In this study, a rational model, as shown below, was selected for the regression with the experimental data using Mathematica software:
a0 a1 t a2 t 2 a3 cin a4 cin 2 a5 p a6 p 2 a7 p t a8 cin p a9 cin t Cf,i = b0 b1 t b2 t 2 b3 cin b4 cin 2 b5 p b6 p 2 b7 p t b8 cin p b9 cin t
(14)
where, Cf,i is the metal concentrations in filtrate (mg/L), a0….a9, and b0…..b9 represent the regression coefficients, t, p, and cin are time, pressure and initial metal concentration, respectively. Present rational model was used for the prediction of nickel and lead concentrations in filtrate. The regression coefficients of the models were calculated by using Mathematica solver “𝐹𝑖𝑛𝑑𝐹𝑖𝑡”. 3.
Results and Discussion
3.1. Adsorption Results Adsorption capacity of the PSf10 membrane against metal ions was first determined in static condition. In order to estimate the exact contribution of NaX nanoparticles to metal adsorptions, native PSf membrane was also used in the adsorption process and their comparisons in terms of the amount of metal ions adsorbed in a unit surface area of a membrane with respect to time are illustrated in Figure 2. The metal ions retained by the native PSf membrane is almost negligible and it can be said that the removal of metal ions is mainly occurred by the NaX content in the membrane. This could be expected, since native PSf which is known as negatively charged polymer with a very low charge density provides insignificant ionic interactions with the metal cations in solution. From Figure 2, removal rate of metal ions is fast at the initial time (0-5 min) and gradually 10
decreased as time proceeds. This is due to the high number and availability of active sites, as well as due to the highest driving force for the mass transfer at the first stage of the adsorption. On the other hand, slower increase in the adsorbed amount of the cations at the later stage signify that the process became less efficient due to the gradual occupancy of the active sites and reduction in the metal cations in the liquid phase. Figure 2 The equilibrium adsorbed amount and the kinetic constants from the minimization of the error in Equation 8 were estimated as 0.48 mg/cm2 and 0.074 min-1 for lead and 0.18 mg/cm2 and 0.22 min1
for nickel, respectively. From theoretical calculation based on the molecular composition of the
NaX unit cell (Na104[Si119Al96O526]) obtained from EDX analysis [16], the maximum ion exchange capacity is expected to be 0.695 and 0.197 mg/cm2 for lead and nickel cations, respectively. It was thought that Pb2+ is not only adsorbed by ion exchange mechanism but also interacts with the lone pairs of electrons on oxygen located in the NaX framework [25]. The differences in the experimental and the theoretical metal capacities might be explained in two ways; first, driving force between metal ions in liquid and solid phases is reduced during proceeding of adsorption process leading to achieve equilibrium earlier than the complete saturation of NaX. Secondly, if the cations present in solutions as hydrated forms, then their sizes (~4Å) are comparable with the free dimensions of the NaX channels causing to exchange difficult. Reusability of the hybrid membranes were tested by successive adsorption/desorption cycles with the corresponding metal ions and NaCl solution was used as the desorbing agent. The metal ions weakly bonded on the surface of the membrane was removed by rinsing it with water at the end of each adsorption and desorption steps. The cycle which was repeated 5 times are represented in Figure 3 for both of the metal ions. In Figure 3, the percentage sorption values reflect the amount of desorbed or adsorbed metals relative to the amount that is adsorbed in the first cycle. From Figure 3, regeneration was successfully achieved with slightly diminished the hybrid membrane capacity for the metal adsorption after each cycle. At the end of the fifth cycle, the membrane retained 77 and 92% of its initial adsorption capacity for the lead and nickel ions, respectively. The accumulation due to the lower amount of desorption compared to the adsorbed amount may be the main reason for a decline in retention of metal ions upon recycling. The decrease in sorption capacity is due to the so-called hysteresis effect. A hysteresis arises when the amount of desorbed 11
is lower than the adsorbed [26]. In addition, the lower amount of metals released during desorption steps reflect that the zeolite nanoparticles are being more stable in the form of metal (PbX or NiX) than in the form sodium.
Figure 3 3.2.
Filtration Results
Physical properties of the native and NaX added PSf membranes are given in Table 1. An enhanced hydraulic permeability was obtained by the addition of NaX into PSf membrane while, BSA rejection values remained almost unchanged, even though the pore size of the hybrid membrane increased twice compared to the native PSf. The observed Lp increment is connected two factors: an increase in the pore size due to the instantaneous demixing occurred during phase separation of the membrane process and formation of hydrogen bonding between sulfonic groups of PSf and – OH groups of NaX that improve hydrophilicity and hence reduce the flow hindrance [12]. The effect of filler addition on the membrane structure has been characterized in detailed in our previous study [16]. It has been reported that the pore size, hydraulic permeability and metal removal capacity of the PSf membrane has been improved by the addition of NaX. Table 1 Sorption behavior of the PSf10 membrane was determined under different initial metal concentrations, operating pressures and temperatures by using cross-flow filtration module. During filtrations of metal solutions high linearities in permeate volume with respect to time were obtained which signified that membrane fouling did not occur. This could be expected since removal of heavy metals were mainly based on the ion-exchange mechanism. Effect of initial metal concentrations on the sorption kinetics of the membrane are illustrated in Figure 4 (a) through (d) as the metals adsorbed by the solid phase and remained in the liquid phase, respectively. In Figure 4 (a), there is an almost linear increase in the adsorbed amount of lead with respect to time was obtained. Increasing initial lead concentration enhanced the adsorption rate up to 100 ppm above which it was remained unaffected. On the other hand, lead concentration in the permeate increases more pronouncedly as the feed concentration is increased (Figure 4 (b)). Diffusion and convection 12
along with reaction (ion exchange) take place simultaneously during filtration of liquid stream through a membrane adsorber. In general, two limiting cases of reaction or diffusion determine the performance of the membrane adsorber. If the filtration process is reaction limited, the solute concentration at the membrane surface is negligible due to its desorption from the membrane to the permeate side. In the case of diffusion limited process, ultrafiltration is dominated by the difference in osmotic pressure [27]. The values of Thiele modulus and rate constants at each transmembrane pressure and initial metal concentration calculated by the Equations 3 to 8 are presented in Table 2. According to Table 2, all the estimated values are lower than 1 and the kinetic data were well described by the pseudo second order kinetic model based on the lower values of the errors. These two results reveal that the transport is controlled by chemical reaction and further increase in metal concentration does not alter the rate of reaction [28]. On the other hand, the values of Thiele modulus under static conditions have been found as 1.41 and 2.87 for the lead and nickel ions, respectively. The results reflect that the rate of reaction and diffusion are comparable. These findings allow us to make a practical comparison about the feasibility of the filtration process carried out under dynamic and static conditions. In static condition, diffusion and reaction take place simultaneously which result in fast equilibrium. On the other hand, at high concentration and pressures in dynamic filtration, increasing the volumetric flux shortens the residence time for the metal ions to contact NaX, consequently, the rate of metal permeation is increased. This result can be visualized clearly in Figure 4 (b) and (d). Before attaining saturation level on/in the solid phase, the metal concentrations in permeate are proceeding to increase. Figure 4 From Figure 4, the specific adsorbed amounts of the PSf10 membrane at the maximum initial metal concentrations during 120 min of filtration were calculated as 375 and 167 mg/g for the lead and nickel ions, respectively. The result reveals that the PSf10 membrane is the most promising adsorber among the other types of adsorbents reported in the literature [15, 29-30]. In addition, PSf10 hybrid membrane offers a simple way to purify water and eliminates the drawback to separate the nano-sized particles at the end of adsorption process. The presence of NaX on the other hand improves water permeability and throughput. As a conclusion, the present study is important to signify that the filtration and removal properties of a poor PSf membrane can be improved in a simple manner by incorporation of inorganic nanoparticles. 13
Table 2 Effect of temperature on the filtration performance of the PSf10 membrane were investigated in the range of 25-50˚C by keeping the initial metal concentration and operating pressure constant as 50 mg/L and 0.5 bar, respectively. Results are tabulated in Table 3, as percentage recoveries which were defined as the ratio of the adsorbed amount of metal ions to the initial amount. Increasing temperature did not demonstrate a systematic variation for both of the metal ions. Maximum recovery was obtained at maximum temperature for the lead, while nickel recovery was maximum at 35˚C. From Table 3, lower temperature, initial metal concentration and pressure adequately satisfy the maximum recoveries and make the process economically feasible. Table 3 3.3. Regression Results Analysis of variance (ANOVA) is important to signify the coefficients of variables and their interactions are significant or not. Results are summarized in Table 4 and 5 for the lead and nickel, respectively. Table 4 Table 5 F value and P value are the two key values used to determine which terms are statistically significant. The larger value of F (>1) and the lower value of P (<0.0001) indicates the corresponding term is more significant. According to Tables 4 and 5, all the linear terms “t, p, cin” are significant. Moreover, except the term “p.t” the other non-linear terms “cin.t, and cin.p” are also significant. Based on the ANOVA results, all independent variables with their interactions were included in Equation 5 and the coefficient of each terms were estimated. Table 6 shows the values of the coefficients of the model for the variation of nickel and lead concentrations in filtrate. Negative and positive signs of each coefficient reflect the effect of each term either in the purpose of descending or ascending of the response, respectively. The values of the coefficient of determination (R2) for both models were found to be higher than 0.99. It defines that observed outcomes can be replicated by the nickel and lead models very well based on the proportion of total 14
variation of outcomes explained by the models. However, it should be noted that the present models are only valid under the conditions 50
4. Conclusion Removal of heavy metals from wastewater include conventionally integration of process units with complementary treatment characteristics which make the overall process inefficient in terms of volume, energy and time. Membrane adsorber is a unique alternative for effectively and efficiently separate toxic substances from liquid phase. Adsorption and filtration are coupled in membrane adsorber provides high removal efficiency, favorable hydrodynamics, acceptable reusability and elimination of an extra step which can be required to separate adsorbent from treated water. However, the behavior of the membrane adsorber under different operating conditions needs to be clarified in order to minimize the filtrate concentration of target species. Increasing pressure will increase the volume filtrate and increase the transport of metal cations as well. Increasing pressure will reduce the contact time necessary for the reaction to take place especially in the reaction controlled regime. Similarly, initial metal concentration will cause to rise the metal concentration in permeate if rate of reaction is very small than the diffusional rate. Filtration time is another 15
parameter that should be controlled due to limited adsorption capacities of membrane adsorbers. In this study, these effects on the transport behavior of the membrane were determined by calculating Thiele modulus at each filtration condition. According to the values of the Thiele modulus, the transport of the lead and nickel cations were controlled by reaction. A nonlinear multiple rational regression has been introduced to analyze the whole filtration data of the lead and nickel. Both experimental and predicted results suggested that the PSf10 membrane should be used under low initial metal concentration and pressure in a short filtration time. As a conclusion, the present study is important to signify that the filtration and removal properties of a poor PSf membrane can be improved in a simple manner by incorporation of inorganic nanoparticles. Acknowledgement We would like to thank the Scientific and Technological Research Council of Turkey for the financial support through Grant 113Y159. Celal Bayar University, Food Engineering Department is gratefully acknowledged for providing AAS analysis.
References [1] R.J. Gohari, W.J. Lau, T. Matsuura, E. Halakoo, A.F. Ismail, Adsorptive removal of Pb(II) from aqueous solution by novel PES/HMO ultrafiltration mixed matrix membrane, Sep. Purif. Technol. 120 (2013) 59-68. [2] R. Mukherjee, S. De, Adsorptive removal of nitrate from aqueous solution by polyacrylonitrile–alumina nanoparticle mixed matrix hollow-fiber membrane, J. Membrane Sci. 466 (2014) 281-292. [3] E. Salehi, S.S. Madaeni, L. Rajabi, A.A. Derakhshan, S. Daraei, V. Vatanpour, Static and dynamic adsorption of copper ions on chitosan/polyvinyl alcohol thin adsorptive membranes: Combined effect of polyethylene glycol and aminated multi-walled carbon nanotubes, Chem. Eng. J. 215–216 (2013) 791–801. [4] J.W. Gilman, Flammability and thermal stability studies of polymer layered-silicate (clay) nanocomposites, Appl. Clay Sci., 15 (1999), pp. 31–49.
16
[5] A. Okada, A. Usuki, The chemistry of polymer–clay hybrids, Mater. Sci. Eng., C3 (3) (1995), pp. 109–115. [6] S.I. Yu, X. Zuo, R. Bao, X. Xu, J. Wang, J. Xu, Effect of SiO2 nanoparticle addition on the characteristics of a new organic–inorganic hybrid membrane, Polymer, 50 (2009), pp. 553– 559. [7] M. Fathizadeh, A. Aroujaliana, A. Raisi, Effect of added NaX nano-zeolite into polyamide as a top thin layer of membrane on water flux and salt rejection in a reverse osmosis process, J. Membrane Sci. 375 (2011) 88–95. [8] J. Huang, K. Zhang, K. Wang, Z. Xie, B. Ladewig, H. Wang, Fabrication of polyethersulfonemesoporous silica nanocomposite ultrafiltration membranes with antifouling properties, J. Membrane Sci. 423-424 (2012) 362-370. [9] S.J. Oh, N. Kim, Y.T. Lee, Preparation and characterization of PVDF/TiO2 organic-inorganic composite membranes for fouling resistance improvement, J. Membrane Sci. 345 (2009) 1320. [10] G. Arthanareeswaran, T.K. Sriyamuna Devi, M. Raajenthiren, Effect of silica particles on cellulose acetate blend ultrafiltration membranes: Part I, Sep. Purif. Technol. 64 (2008) 38-47. [11] R. Han, S. Zhang, C. Liu, Y. Wang, X. Jian, Effect of NaA zeolite particle addition on poly(phthalazinone ether sulfone ketone) composite ultrafilitration (UF) membrane performance, J. Membrane Sci. 345 (2009) 5-12. [12] F. Liu, B.R. Ma, D. Zhou, Y.H. Xiang, L.X. Xue, Breaking through tradeoff of Polysulfone ultrafiltration membranes by zeolite 4A, Micropor. Mesopor. Mat. 186 (2014) 113–120. [13] A. Charkhia, M. Kazemeinia, S.J. Ahmadib, H. Kazemian, Fabrication of granulated NaY zeolite nanoparticles using a new method and study the adsorption properties, Powder Technol. 231 (2012) 1–6. [14] Y. Wang, Z. Lei, B. Chen, Q. Guo, N. Liu, Adsorption of NO and N2O on Fe-BEA and HBEA zeolites, Appl. Surf. Sci. 256 (2010) 4042–4047. [15] S. Malamisa, E. Katsoua, A review on zinc and nickel adsorption on natural and modified zeolite, bentonite and vermiculite: Examination of process parameters, kinetics and isotherms, J. Hazard. Mater. 252– 253 (2013) 428– 461. [16] Y.Yurekli, Removal of heavy metals in wastewater by using zeolite nano-particles impregnated polysulfone membranes, J. Hazard. Mater. 309 (2016) 53–64. 17
[17] S.M. Hosseini, S. Rafiei, A.R. Hamidi, A.R. Moghadassi, S.S. Madaeni, Preparation and electrochemical
characterization
of
mixed
matrix
heterogeneous
cation
exchange membranes filled with zeolite nanoparticles: Ionic transport property in desalination, Desalination 351 (2014) 138-144. [18] J. Dasgupta, M. Singh, J. Sikder, V. Padarthi, S. Chakraborty, S. Curcio, Response surfaceoptimized removal of Reactive Red 120 dye from its aqueous solutions using polyethyleneimine enhanced ultrafiltration, Ecotox. Environ. Safe. 121 (2015) 271–278. [19] Y. Zheng, A. Wang, Removal of heavy metals using polyvinyl alcohol semi-IPN poly(acrylic acid)/tourmaline composite optimized with response surface methodology, Chem. Eng. J. 162 (2010) 186–193. [20] H. Bessbousse, T. Rhlalou, J.-F. Verchere, L. Lebrun, Removal of heavy metal ions from aqueous solutions by filtration with a novel complexing membrane containing poly(ethyleneimine) in a poly(vinyl alcohol) matrix, J. Membrane Sci. 307 (2008) 249–259. [21] W.M. Deen, Hindered transport of large molecules in liquid-filled pores, AIChE J. 33 (1987) 1409–1425. [22] W.R. Bowen, A.W. Mohammad, N. Hilal, Characterisation of nanofiltration membranes for predictive purposes—use of salts, uncharged solutes and atomic force microscopy J. Membrane Sci. 126 (1997) 91–105. [23] H. Pham, Springer Handbook of Engineering Statistics 2006, Springer-Verlag London [24] V.M. Taavitsainen, Rational function ridge regression in kinetic modeling: A case study, Chemometrics and Intelligent Laboratory Systems 120 (2013) 136–141. [25] X. Li, Z. Wang, Q. Li, J. Ma, M. Zhu, Preparation, characterization, and application of mesoporous silica-grafted graphene oxide for highly selective lead adsorption, Chem. Eng. J. 273 (2015) 630–637. [26] S.S. Madaeni, S. Zinadini, V. Vatanpour, Convective flow adsorption of nickel ions in PVDF membrane embedded with multi-walled carbon nanotubes and PAA coating, Sep. Purif. Technol. 80 (2011) 155–162. [27] S. Mondal, R. Mukherjee, S. Chatterjee, and S. De, Adsorption-Concentration Polarization Model for Ultrafiltration in Mixed Matrix Membrane, AIChE J. 60 (2014) 2354-2364.
18
[28] S. Habibi, A. Nematollahzadeh, S.A. Mousavi, Nano-scale modification of polysulfone membrane matrix and the surface for the separation of chromium ions from water Chem. Eng. J. 267 (2015) 306–316. [29] N. Abdullah, R.J. Gohari, N. Yusof, A.F. Ismail, J. Juhana, W.J. Lau, T. Matsuura, Polysulfone/hydrous ferric oxide ultrafiltration mixed matrix membrane: Preparation, characterization and its adsorptive removal of lead (II) from aqueous solution, Chem. Eng. J. 287 (2016) 28-37. [30] L. R. Rad, A. Momeni, B. F. Ghazani, M. Irani, M. Mahmoudi, B. Noghreh, Removal of Ni2+ and Cd2+ ions from aqueous solutions using electrospun PVA/zeolite nanofibrous adsorbent, Chem. Eng. J. 256 (2014) 119–127.
19
List of Figures
Figure 1: Experimental set-up of cross-flow filtration. Figure 2: Adsorption kinetics of the metal ions in the native (open symbols) and NaX added (solid symbols) PSf membranes. Lines denote the best fits from Lagergren pseudo first order kinetic model. (Cin = 500 mg/L). Figure 3: Adsorption-desorption cycles of the hybrid membrane against lead and nickel ions. Figure 4. Concentration of lead (a) at the solid phase, (b) in the liquid phase and nickel (c) at the solid phase and (d) in the liquid phase under operating pressure of 0.5 bar. Figure 5: Representative response surface plots for the effect of initial metal concentrations and filtration time on (a) Pb2+ and (b) Ni2+ concentrations in filtrate at 0.5 bar transmembrane pressure. (Red circles represent the experimental data).
20
Figure 1
0.6
Sorption capacity (mg/cm2)
lead-PSf10 0.5
nickel-PSf10 lead-PSf
0.4
nickel-PSf
0.3 0.2 0.1 0 0
10
20
30
40
Time (min)
Figure 2
21
50
60
70
100
lead adsorption
nickel adsorption
lead desorption
nickel desorption
% sorption
80 60 40 20 0 1
2
3
4
5
number of runs
Figure 3 160
(a)
50 ppm
Filtrate concentration (mg/L)
Adsorbed amount (mg/cm2)
0.5 100 ppm
0.4
200 ppm 0.3 0.2 0.1
50 ppm
140
(b)
100 ppm
120
200 ppm
100 80 60 40 20 0
0 0
50
100
0
150
50
Time (min) 0.2
150
160
(c)
50 ppm
Filtrate concentration (mg/L)
Adsorbed amount (mg/cm2)
100 Time (min)
100 ppm
0.15
200 ppm
0.1
0.05
0
50 ppm
140
(d)
100 ppm
120
200 ppm
100 80 60 40 20 0
0
50
100
150
0
50
100 Time (min)
Time (min)
Figure 4 22
150
(b)
(a)
Figure 5
23
Table 1: Properties of the membranes.
Membranes
Native PSf PSf10
Porosity* (%)
Pore radius* (nm)
BSA rejection (%)
Lp (L/m2.h.bar)
82
10
100
23.2
19
98
57.1
74 * from reference [16]
24
Table 2: Estimated values of Thiele modulus and the rate constants during filtration of lead and nickel ions.
C0 (mg/L)
P (bar)
50 50 50 100 100 100 200 200 200
0.5 1.0 2.0 0.5 1.0 2.0 0.5 1.0 2.0
50 50 50 100 100 100 200 200 200
0.5 1.0 2.0 0.5 1.0 2.0 0.5 1.0 2.0
k1 (1/min)
Error
1
k2 (cm2/mg.min)
Error
2
Pb2+
0.002 0.003 0.007 0.009 0.009 0.012 0.013 0.017 0.017
1.610-5 2.810-4 2.910-4 8.910-4 9.210-4 1.910-3 3.610-3 4.510-3 5.210-3
0.027 0.036 0.057 0.063 0.065 0.072 0.078 0.088 0.087
0.001 0.001 0.005 0.006 0.007 0.007 0.014 0.011 0.013
2.010-5 2.810-4 2.310-4 7.010-4 7.510-4 1.410-3 2.710-3 3.210-3 5.310-3
0.001 0.001 0.002 0.004 0.004 0.004 0.008 0.007 0.007
Ni2+
0.016 0.013 0.015 0.011 0.029 0.024 0.035 0.012 0.026
2.810-5 4.110-4 3.610-4 1.010-4 1.210-3 1.910-3 6.810-4 2.410-3 4.410-3
0.099 0.090 0.096 0.082 0.134 0.122 0.148 0.085 0.128
0.083 0.034 0.040 0.020 0.076 0.060 0.165 0.015 0.089
4.210-5 3.010-4 2.710-4 9.410-5 6.110-4 8.910-4 5.410-4 2.210-3 3.110-3
0.011 0.007 0.008 0.008 0.015 0.013 0.031 0.009 0.023
Metals
25
Table 3: Percentage recovery of the PSf10 membrane with respect to operating conditions (initial metal concentrations, operating pressures and temperatures).
Initial concentration (mg/L) 50 50 50 50 50 100 100 100 200 200 200 50 50 50 50 50 100 100 100 200 200 200
Pressure (bar) 0.5 1.0 2.0 0.5 0.5 0.5 1.0 2.0 0.5 1.0 2.0 0.5 1.0 2.0 0.5 0.5 0.5 1.0 2.0 0.5 1.0 2.0
Temperature (˚C) 25 25 25 35 50 25 25 25 25 25 25 25 25 25 35 50 25 25 25 25 25 25
26
Metals
Pb2+
Ni2+
% Recovery 86.1 89.8 55.0 79.0 90.2 50.9 50.2 38.8 33.9 32.2 21.5 40.6 29.8 33.7 46.8 31.2 39.4 27.2 24.2 25.5 14.9 12.9
Table 4: Analysis of variance (ANOVA) for the lead
t p cin pt cin t cin p Error Total R2=0.99
Degrees of freedom 24 2 2 48 48 4 95 223
Sum of squares
Mean Squares
98887.4 26066.7 452086.0 4593.4 32377.1 20314.8 9817.9 644143.0
4120.31 13033.30 226043.00 95.70 674.52 5078.70 103.35
27
F ratio
P value
39.87 126.11 2187.23 0.93 6.53 49.14
2.29 x 10-39 1.83 x 10-27 3.59 x 10-80 0.61 4.80 x 10-15 2.43x 10-22
Table 5: Analysis of variance (ANOVA) for the nickel
t p cin pt cin t cin p Error Total R2=0.99
Degrees of freedom 24 2 2 48 48 4 96 224
Sum of squares 73828.1 14443.1 420360.0 1456.7 43479.2 1937.1 2200.0 557705.0
Mean Squares 3076.17 7221.56 210180.00 30.35 905.82 484.28 22.92
28
F ratio 134.23 315.12 9171.43 1.32 39.53 21.13
P value 2.88 x 10-63 6.56x 10-43 2.48 x 10-110 0.12 5.14 x 10-46 1.61 x 10-12
Table 6: The regression coefficients of the rational model for nickel and lead
a0 a1 a2 a3 a4 a5 a6 a7 a8 a9 b0 b1 b2 b3 b4 b5 b6 b7 b8 b9
Cf,nickel
Cf,lead
-4420.09 66.75 1.81 37.44 -0.05 15388.89 -6720.15 -277.72 24.95 15.85 2383.30 18.75 -0.01 -1.43 -0.02 -2577.44 668.55 -1.42 2.76 0.01
10614.92 -8776.64 3.82 -2.65 -0.93 -28535.33 7000.03 2192.13 150.99 159.95 1294.90 155.29 -0.16 13.41 -0.02 2482.45 -1046.25 16.42 -6.59 -0.01
29
S0304-3894(17)30149-8 http://dx.doi.org/doi:10.1016/j.jhazmat.2017.02.061 HAZMAT 18416
To appear in:
Journal of Hazardous Materials
Received date: Revised date: Accepted date:
25-8-2016 20-12-2016 28-2-2017
Please cite this article as: Yilmaz Yurekli, Mehmet Yildirim, Levent Aydin, Melih Savran, Filtration and Removal Performances of Membrane Adsorbers, Journal of Hazardous Materials http://dx.doi.org/10.1016/j.jhazmat.2017.02.061 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Filtration and Removal Performances of Membrane Adsorbers
Yilmaz YUREKLI1*, Mehmet YILDIRIM2, Levent AYDIN3, Melih SAVRAN3
1
Department of Bioengineering, Celal Bayar University, Muradiye Kampusu, 45140, Yunusemre, Manisa, Turkey
2
Department of Metallurgical and Materials Engineering, Celal Bayar University, Muradiye Kampusu, 45140,
Yunusemre, Manisa, Turkey Mechanical Engineering Department, Izmir Katip Çelebi University, Balatçık Kampusu, Çiğli, Izmir, Turkey
3
*
Corresponding author: [email protected]
1
Highlights
NaX nanoparticles added PSf membrane was successfully synthesized and characterized by means of cross-flow filtration module. High recovery and reusability was attained for both of the lead and nickel cations. Nonlinear rational regression model perfectly described the whole filtration data. Lower initial metal concentrations, temperatures and pressures were selected to make the target response minimum.
Abstract Membrane adsorbers are promising candidates for the efficient and effective removal of heavy metals in waste water due to their unattainable adsorption and filtration capabilities. In the present study, zeolite nanoparticles incorporated polysulfone (PSf10) membrane was synthesized by means of non-solvent induced phase separation technique for the removal of lead and nickel ions in water. PSf10 showed a remarkable sorption capability and after repeated (adsorption/desorption)5 cycles in batch experiments, it preserved 77% and 92% of its initial sorption capacity for the lead and nickel, respectively. Addition of nanoparticles increased the pore radius of the native PSf from 10 to 19 nm, while bovine serum albumin rejection remained unchanged at 98%. Increments in the pore size and enhancement in hydrophilicity caused to increase hydraulic permeability of the native PSf from 23 to 57 L/m2.h.bar. Cross-flow filtration studies revealed that the filtrate concentrations were inversely affected by the initial metal concentration and transmembrane pressure due to reaction limited region. Nonlinear rational regression model perfectly described the filtration behavior of the PSf10 within the experimental range and suggested that lower initial metal concentration and pressure with a short filtration time should be selected for the target response to be minimum. Keywords: Cross-flow filtration, Heavy metals removal, Membrane adsorber, Rational regression
2
1. Introduction Strict regulations on the discharge levels of the industrial effluents for sustainability of clean water resources have pressurized the companies and researchers to develop more advanced functional materials that satisfy the criteria proposed by the environmental and health organizations. Nowadays, hybrid membranes prepared by impregnation of different types of inorganic nano-sized materials onto/into polymeric matrixes have been attracted attention due to their unattainable sorption capacities together with enhanced filtration performances [1-3]. Thereby, high energy cost of the membrane separation processes as in the case of nanofiltration or reverse osmosis can be reduced without altering the membrane selectivity by utilizing the hydrophilic nature of the adsorbents dispersed on/in the membrane. Besides elimination of difficulties in separation of the nano-adsorbents from treated water, hybrid membranes offer effective solution for removal of suspended charged solids and high molecular weight solutes (screening effect) presumably available in wastewater that may cause blocking the active sites of the adsorbents when they are used in direct contact mode in a simple batch adsorption process. In addition, the presence of charged nanoparticles in the membrane improves various properties like thermal [4], mechanical [5], diffusive [6] and electrostatic. Moreover, they introduce unique functionalities such as photocatalytic, antibacterial or adsorptive capabilities [7-11]. Thus, membrane adsorption is more eligible over conventional processes due to favorable hydrodynamic, high removal efficiency, acceptable reusability and small footprint. Polysulfone is one of the most popular polymer for membrane casting due to its well-known excellent mechanical, thermal and chemical stability. However, because of its hydrophobic nature, wide range of its application in membrane separation processes has been restricted. To increase its hydrophilicity, addition of inorganic fillers into the membrane has been accepted as an efficient alternative way [12]. Among the inorganic materials, zeolite is a good candidate for the polymer membranes because of its hydration effect and advanced ion exchange performance. Zeolites have well-defined porous structures and offer mobility of alkali and alkaline earth metals, in order to compensate net negative charge between Si4+ and Al3+ in the framework makes zeolites excellent adsorber for the removal of many target solutes [13-15]. Meanwhile, nanoparticles known as high efficient adsorbents are replaced with micrometer-sized counterparts due to advanced specific surface area and interfacial activity. The unique properties of the zeolite nano-particle such as high 3
ion exchange capacity and fast adsorption rate makes it excellent choice for the separation of heavy metals in wastewater treatment [16] and salts in desalination applications [17]. Response surface methodology (RSM), a flexible statistical tool can be applied to many systems to optimize the target response by suitably indexing the linear, quadratic and interaction effects of governing process parameters. It provides an empirical model to yield precisely configured experimental trials within the dictated ranges of variables thereby, reduces the time expended on the required analyses. For example, it has been used for optimization process variables of polymer enhanced ultrafiltration in order to maximize dye rejection efficiency [18]. In another study, the preparation conditions of the polymeric hydrogels have been optimized to gain the maximum adsorption capacity for Pb2+ and Cu+2 [19]. After the design of experiments for the chemical process, generally nonlinear regression models can be used to predict the phenomenological response of the process. This approach allows to write exact objective functions and constraints including design variables for mathematical optimization problems. Successful operation of membrane systems can only be achieved by using suitable membranes and determining their optimum operational conditions. In our previous study, the effects of membrane preparation conditions including NaX content and evaporation time during phase inversion process on the water hydraulic permeabilities and metal removal capacities of the membranes were elaborated [16]. Current study mainly focuses on the optimization of operating conditions and identification of transfer regimes during filtration of metal solutions. In this respect, the same membrane architecture prepared in the previous study was used to elucidate the effect of various process parameters in which case cross-flow mode of operation was selected since it provides more realistic data and simulates well the industrial processes. Before filtration, batch adsorption experiments which are important to explain the removal mechanisms of metal ions were carried out by using native and NaX added PSf membranes. Reusability of the membrane was also investigated by successive adsorption/desorption cycles. The performance of the hybrid membrane was then examined under variable transmembrane pressures, initial metal concentrations and temperatures. Metal concentrations in filtrate and retentate at each conditions were determined as a function of time. The raw data were then used in nonlinear multiple regression to build a rational model equation. The significance of each interaction parameters used in model equation were further analysed using analyses of variance (ANOVA) statistical tool. 4
2.
Materials and Methods
2.1. Materials Polysulfone beads (Mn=22,000 g/mol) and N-methyl-2-pyrrolidone (NMP) for the preparation of ultrafiltration membranes were supplied by Sigma-Aldrich. During production of zeolite nanoparticles, fumed silica having 0.007 m particle size and sodium aluminate from Sigma-Aldrich were used. Sodium hydroxide pellets was purchased from Merck and nickel (II) chloride (NiCl2) and lead (II) nitrate (Pb(NO3)2) solutions as atomic spectroscopic standards were supplied from Fluka. Bovine serum albumin (BSA) used for the rejection experiments was supplied by Aldrich. Molecular dimension of BSA is given by the supplier as 40 x 140 Å. Necessary dilutions were performed with Milli-Q water having resistivity higher than 18 M.cm. 2.2. Methods 2.2.1. Preparation of NaX Added PSf Membranes Asymmetric PSf membrane loaded with 10w% of NaX nanoparticles was prepared based on nonsolvent-induced phase separation (NIPS) technique. PSf10 membrane was obtained by directly immersing the polymer casting solution into coagulation bath at 4˚C and the details in the experimental procedure were given in our previous study [16]. 2.2.2. Determination of BSA Rejection Rejection experiment was performed by filtration of 1 mg/mL bovine serum albumin (BSA) solution prepared in 22 mM sodium phosphate buffer at pH 5 which is close to isoelectric point of BSA (IEPBSA 4.9). Dead-end stirred cell at 1 bar transmembrane pressure was used. Protein concentrations in the feed and filtrate were measured using UV/VIS spectrophotometer (Thermo evolution 201 model) at a fixed wavelength of 278 nm. Rejection values were then calculated based on the expression given below: % 𝑟𝑒𝑗𝑒𝑐𝑡𝑖𝑜𝑛 =
𝐶𝑓 −𝐶𝑝 𝐶𝑓
𝑥100
(1)
where, Cf and Cp denote the protein concentrations in the feed and permeate sides, respectively.
5
2.2.3. Batch Adsorption Ion exchange performance of the NaX added PSf membrane was initiated by introducing a small piece of composite membrane with 20 mL of 500 mg/L lead or nickel solutions. During the reaction a magnetic stirrer with a constant stirring rate of 300 rpm was used and the temperature was maintained constant at 25˚C. The change in metal concentrations in the solution was monitored by sampling at specified time intervals and they were analyzed using atomic absorption spectroscopy (Perkin Elmer AAnalyst 800 with air-acetylene oxidizing flame). The same procedure was followed with the native PSf membrane as a control. The amount of metal ions exchanged with cations in the zeolite frameworks at each time interval was calculated by the following equation: 𝑞𝑡 =
(𝐶0 −𝐶𝑡 )∙𝑉
(2)
𝑆
where, qt (mg/cm2), is the amount of metal ions adsorbed per unit surface area of the membrane S (cm2), C0 and Ct (mg/L), are the liquid phase metal concentrations at initial and at any time t, (min), respectively and V (L) is the permeate volume. According to our previous study, the amount of NaX available in a unit surface area of the PSf10 membrane was determined as 1.08 mg/cm2. Reusability of the hybrid membrane is an important factor which affects directly the cost of the process and decreases the environmental pollution in such a way that prolongs the shelf life of the membrane and hence, lowers the amount of waste materials. Reusability of the hybrid membrane was tested by series of adsorption and desorption experiments. Desorption process was carried out with the membrane which was previously saturated with one of the metal ions (lead or nickel). In order to release all the metal ions adsorbed by the membrane, an excess amount of NaCl (5 M) was used. Desorption was started by inserting the membrane into 50 mL of NaCl solution maintained at 25˚C. During 120 min of desorption period samples were withdrawn at the prescribed time intervals and analysed for their metal contents by using AAS. At the end of desorption process, the membrane was rinsed and then stored in water at 4˚C at least one day. Determination of removal rate of heavy metals through hybrid membrane provides a satisfactory design for a separation unit. In general, two or more mechanisms influence the rate of metal 6
adsorption by the hybrid membrane. The first one is the convective flow of the metal ions from the bulk solution to the nearby membrane interface. This is known as film type of mass transfer resistance and can be eliminated by applying adequate stirring rate. The second one is the diffusional flow of the metal ions through the membrane pores and through the pores of the NaX nanoparticles and the third one is the ion exchange reaction which is known to be occurred instantaneously. In order to identify the transport mechanism, one of the important dimensionless number called Thiele modulus, , can be used. From its definition, rate of reaction is compared with the mass transfer rate expressed in Equation 3 and 4 for the first order and second order of reactions, respectively. 𝜅1 = (
𝑘1,𝑖 𝐷𝑒𝑓𝑓,𝑖
𝜅2 = (
1/2
)
𝐿𝑚
𝑘2,𝑖 𝑆𝑎 𝜌𝑁𝑎𝑋 𝐶0 𝐷𝑒𝑓𝑓,𝑖
(3) 1/2
)
𝐿𝑚
(4)
where, k1,i (min-1), Deff,i (m2/min) and Lm (m) are defined as the first order reaction rate constant, effective diffusion coefficient of metal ions in the membrane and membrane thickness (137 m), respectively and k2,i (cm2/mg.min), Sa (cm2/g) and NaX (g/cm3) denote second order reaction rate constant, specific membrane surface area and density of NaX (0.5 g/cm3), respectively. A high value of implies that the system is mass transfer limited while in the case of small values of , the process is controlled by reaction. The mechanism of ion-exchange can be described by the reaction given below; 2+ + 2+ + 𝑀(𝑎𝑞) + 2𝑁𝑎(𝑧𝑒𝑜𝑙𝑖𝑡𝑒) ↔ 𝑀(𝑧𝑒𝑜𝑙𝑖𝑡𝑒) + 2𝑁𝑎(𝑎𝑞)
(5)
where, M2+ is a metal either in liquid phase or on the adsorbed phase. Lagergren pseudo-second order kinetic model is mostly suitable while describing this type of reaction. However, under experimental conditions one of the reactants can be limited, then the reaction follows pseudo-first order kinetics [20]. Therefore, the rate constants in Equation 3 and 4 were estimated from Lagergren pseudo first order (Equation 6) and second order (Equation 7) model equations, respectively. 𝑞𝑡 = 𝑞𝑒 (1 − 𝑒𝑥𝑝−𝑘1,𝑖𝑡 )
(6) 7
𝑞𝑡 =
𝑞𝑒2 𝑘2,𝑖 𝑡
(7)
1+𝑞𝑒 𝑘2,𝑖 𝑡
where, qe (mg/cm2) represent the adsorbed amount of metal ions in a unit surface area of the membrane at equilibrium. A nonlinear regression analysis by minimization of the sum of the squares of the residuals between experimental and theoretical kinetic data was performed by using the expression given below. (𝐸𝑟𝑟𝑜𝑟)𝑚𝑖𝑛 = ∑((𝑞𝑡 )𝑒𝑥𝑝𝑒𝑟𝑖𝑚𝑒𝑛𝑡𝑎𝑙 − (𝑞𝑡 )𝑡ℎ𝑒𝑜𝑟𝑒𝑡𝑖𝑐𝑎𝑙 )2
(8)
The effective solute diffusivities in the membrane are calculated by multiplying the free solution diffusion coefficients (DPb,=9.4510-10 m2/s and DNi,=6.7910-10 m2/s) with the partition coefficient,i, porosity, and diffusive hindrance factor, Ki,D. 𝐷𝑒𝑓𝑓,𝑖 = 𝜀 𝜙𝑖 𝐾𝑖,𝐷 𝐷𝑖,∞
(9)
The solute diffusive hindrance factor is a function of the ratio between the solute and the pore diameters (𝜆𝑖 = 𝑑𝑖,𝑠 /𝑑𝑝 ) [21]. If parabolic fully developed solute flow within the pore is assumed Ki,D can be defined as [22]: 𝐾𝑖,𝐷 = 1.0 − 2.30𝜆𝑖 + 1.154𝜆2𝑖 + 0.224𝜆3𝑖
(10)
The partition coefficient in Equation 9 is calculated by assuming spherical solutes in cylindrical pores and the expression is given below. 𝜙𝑖 = (1 − 𝜆𝑖 )2
(11)
2.2.4. Cross-Flow Filtration Filtration and metal removal performances of the PSf10 membrane was determined under dynamic conditions using cross-flow filtration system (Sepa CF cell, Sterlitech Corp., USA) with an holdup volume of 70 mL and an active surface area of 140 cm2. The experimental set-up is schematically illustrated in Figure 1. Flat sheet membrane with a feed spacer is placed inside the stainless steel filtration module and the cell body is installed into the cell holder. After pressurizing 8
the cell holder with a hydraulic pump up to 1000 psi, feed solution can be pumped from a temperature controlled feed tank (5 L) to the filtration module. The pressure of the feed solution is controlled with an increment of ±0.1 bar by adjusting the frequency of the feed pump. Volumetric flow rate of the retentate which is recycled into the feed tank is followed by a manual rotameter. Filtrate solution is collected in a beaker placed on a digital balance from which data acquisition is supplied to a computer at each 2 second. Figure 1 Throughout the filtration experiments, following procedure was applied. First, membrane was compacted twice with water at 3 bars for 20 min. If those two fluxes were close to each other, then, the pressure was lowered to 0.5 bar. Water filtration at constant pressure for 20 min was performed. By successive increasing the transmembrane pressure, water permeation at higher pressures were also measured. Filtration data was then used to calculate water fluxes, Jv (L/m2.h) by dividing the volumetric flowrate, Vp (L/h) to the effective membrane area S (m2). 𝐽𝑣 =
𝑉𝑝
(12)
𝑆
Hydraulic permeability, Lp (L/m2.h.bar), through the membrane was determined from the slope of the plot of the solution flux as a function of transmembrane pressure. 𝐿𝑝 =
𝐽𝑣
(13)
∆𝑃
Filtration of the metal solutions were carried out under different initial concentrations (50-200 ppm), temperatures (25-50˚C) and transmembrane pressures (0.5-2 bar) for 120 min. At certain times, samples were withdrawn from the filtrate and retentate. After necessary dilutions, sample concentrations were determined by AAS. 2.2.5. Regression Analysis Regression analysis is a statistical tool for estimating the relationships among the parameters which affect the physical phenomena. Several types of regression models are available in literature: Linear, Logistic, Nonlinear, and Stepwise. Although some of engineering processes may be described well by utilizing the linear models, there are many other inherently nonlinear processes 9
including more general class of functions. Examples of nonlinear functions are logarithmic, trigonometric, power, and rational functions [23]. Rational functions can be seen in many theoretical formulae in chemistry, chemical engineering, and biochemistry, and also have been shown to be a potential method for nonlinear empirical modeling [24]. Some of the advantages of the rational regression models are (i) having better interpolatory properties than polynomial models, (ii) having very good asymptotic properties, (iii) and they can be used to model complicated structure with a low degree in both the numerator and denominator. This in turn means that fewer coefficients will be required compared to the polynomial model. In contrast to conventional polynomials, rational regression models can be bounded. As a consequence, they can be used to describe some chemical effects, in cases where the response can be limited in a finite amount. In this study, a rational model, as shown below, was selected for the regression with the experimental data using Mathematica software:
a0 a1 t a2 t 2 a3 cin a4 cin 2 a5 p a6 p 2 a7 p t a8 cin p a9 cin t Cf,i = b0 b1 t b2 t 2 b3 cin b4 cin 2 b5 p b6 p 2 b7 p t b8 cin p b9 cin t
(14)
where, Cf,i is the metal concentrations in filtrate (mg/L), a0….a9, and b0…..b9 represent the regression coefficients, t, p, and cin are time, pressure and initial metal concentration, respectively. Present rational model was used for the prediction of nickel and lead concentrations in filtrate. The regression coefficients of the models were calculated by using Mathematica solver “𝐹𝑖𝑛𝑑𝐹𝑖𝑡”. 3.
Results and Discussion
3.1. Adsorption Results Adsorption capacity of the PSf10 membrane against metal ions was first determined in static condition. In order to estimate the exact contribution of NaX nanoparticles to metal adsorptions, native PSf membrane was also used in the adsorption process and their comparisons in terms of the amount of metal ions adsorbed in a unit surface area of a membrane with respect to time are illustrated in Figure 2. The metal ions retained by the native PSf membrane is almost negligible and it can be said that the removal of metal ions is mainly occurred by the NaX content in the membrane. This could be expected, since native PSf which is known as negatively charged polymer with a very low charge density provides insignificant ionic interactions with the metal cations in solution. From Figure 2, removal rate of metal ions is fast at the initial time (0-5 min) and gradually 10
decreased as time proceeds. This is due to the high number and availability of active sites, as well as due to the highest driving force for the mass transfer at the first stage of the adsorption. On the other hand, slower increase in the adsorbed amount of the cations at the later stage signify that the process became less efficient due to the gradual occupancy of the active sites and reduction in the metal cations in the liquid phase. Figure 2 The equilibrium adsorbed amount and the kinetic constants from the minimization of the error in Equation 8 were estimated as 0.48 mg/cm2 and 0.074 min-1 for lead and 0.18 mg/cm2 and 0.22 min1
for nickel, respectively. From theoretical calculation based on the molecular composition of the
NaX unit cell (Na104[Si119Al96O526]) obtained from EDX analysis [16], the maximum ion exchange capacity is expected to be 0.695 and 0.197 mg/cm2 for lead and nickel cations, respectively. It was thought that Pb2+ is not only adsorbed by ion exchange mechanism but also interacts with the lone pairs of electrons on oxygen located in the NaX framework [25]. The differences in the experimental and the theoretical metal capacities might be explained in two ways; first, driving force between metal ions in liquid and solid phases is reduced during proceeding of adsorption process leading to achieve equilibrium earlier than the complete saturation of NaX. Secondly, if the cations present in solutions as hydrated forms, then their sizes (~4Å) are comparable with the free dimensions of the NaX channels causing to exchange difficult. Reusability of the hybrid membranes were tested by successive adsorption/desorption cycles with the corresponding metal ions and NaCl solution was used as the desorbing agent. The metal ions weakly bonded on the surface of the membrane was removed by rinsing it with water at the end of each adsorption and desorption steps. The cycle which was repeated 5 times are represented in Figure 3 for both of the metal ions. In Figure 3, the percentage sorption values reflect the amount of desorbed or adsorbed metals relative to the amount that is adsorbed in the first cycle. From Figure 3, regeneration was successfully achieved with slightly diminished the hybrid membrane capacity for the metal adsorption after each cycle. At the end of the fifth cycle, the membrane retained 77 and 92% of its initial adsorption capacity for the lead and nickel ions, respectively. The accumulation due to the lower amount of desorption compared to the adsorbed amount may be the main reason for a decline in retention of metal ions upon recycling. The decrease in sorption capacity is due to the so-called hysteresis effect. A hysteresis arises when the amount of desorbed 11
is lower than the adsorbed [26]. In addition, the lower amount of metals released during desorption steps reflect that the zeolite nanoparticles are being more stable in the form of metal (PbX or NiX) than in the form sodium.
Figure 3 3.2.
Filtration Results
Physical properties of the native and NaX added PSf membranes are given in Table 1. An enhanced hydraulic permeability was obtained by the addition of NaX into PSf membrane while, BSA rejection values remained almost unchanged, even though the pore size of the hybrid membrane increased twice compared to the native PSf. The observed Lp increment is connected two factors: an increase in the pore size due to the instantaneous demixing occurred during phase separation of the membrane process and formation of hydrogen bonding between sulfonic groups of PSf and – OH groups of NaX that improve hydrophilicity and hence reduce the flow hindrance [12]. The effect of filler addition on the membrane structure has been characterized in detailed in our previous study [16]. It has been reported that the pore size, hydraulic permeability and metal removal capacity of the PSf membrane has been improved by the addition of NaX. Table 1 Sorption behavior of the PSf10 membrane was determined under different initial metal concentrations, operating pressures and temperatures by using cross-flow filtration module. During filtrations of metal solutions high linearities in permeate volume with respect to time were obtained which signified that membrane fouling did not occur. This could be expected since removal of heavy metals were mainly based on the ion-exchange mechanism. Effect of initial metal concentrations on the sorption kinetics of the membrane are illustrated in Figure 4 (a) through (d) as the metals adsorbed by the solid phase and remained in the liquid phase, respectively. In Figure 4 (a), there is an almost linear increase in the adsorbed amount of lead with respect to time was obtained. Increasing initial lead concentration enhanced the adsorption rate up to 100 ppm above which it was remained unaffected. On the other hand, lead concentration in the permeate increases more pronouncedly as the feed concentration is increased (Figure 4 (b)). Diffusion and convection 12
along with reaction (ion exchange) take place simultaneously during filtration of liquid stream through a membrane adsorber. In general, two limiting cases of reaction or diffusion determine the performance of the membrane adsorber. If the filtration process is reaction limited, the solute concentration at the membrane surface is negligible due to its desorption from the membrane to the permeate side. In the case of diffusion limited process, ultrafiltration is dominated by the difference in osmotic pressure [27]. The values of Thiele modulus and rate constants at each transmembrane pressure and initial metal concentration calculated by the Equations 3 to 8 are presented in Table 2. According to Table 2, all the estimated values are lower than 1 and the kinetic data were well described by the pseudo second order kinetic model based on the lower values of the errors. These two results reveal that the transport is controlled by chemical reaction and further increase in metal concentration does not alter the rate of reaction [28]. On the other hand, the values of Thiele modulus under static conditions have been found as 1.41 and 2.87 for the lead and nickel ions, respectively. The results reflect that the rate of reaction and diffusion are comparable. These findings allow us to make a practical comparison about the feasibility of the filtration process carried out under dynamic and static conditions. In static condition, diffusion and reaction take place simultaneously which result in fast equilibrium. On the other hand, at high concentration and pressures in dynamic filtration, increasing the volumetric flux shortens the residence time for the metal ions to contact NaX, consequently, the rate of metal permeation is increased. This result can be visualized clearly in Figure 4 (b) and (d). Before attaining saturation level on/in the solid phase, the metal concentrations in permeate are proceeding to increase. Figure 4 From Figure 4, the specific adsorbed amounts of the PSf10 membrane at the maximum initial metal concentrations during 120 min of filtration were calculated as 375 and 167 mg/g for the lead and nickel ions, respectively. The result reveals that the PSf10 membrane is the most promising adsorber among the other types of adsorbents reported in the literature [15, 29-30]. In addition, PSf10 hybrid membrane offers a simple way to purify water and eliminates the drawback to separate the nano-sized particles at the end of adsorption process. The presence of NaX on the other hand improves water permeability and throughput. As a conclusion, the present study is important to signify that the filtration and removal properties of a poor PSf membrane can be improved in a simple manner by incorporation of inorganic nanoparticles. 13
Table 2 Effect of temperature on the filtration performance of the PSf10 membrane were investigated in the range of 25-50˚C by keeping the initial metal concentration and operating pressure constant as 50 mg/L and 0.5 bar, respectively. Results are tabulated in Table 3, as percentage recoveries which were defined as the ratio of the adsorbed amount of metal ions to the initial amount. Increasing temperature did not demonstrate a systematic variation for both of the metal ions. Maximum recovery was obtained at maximum temperature for the lead, while nickel recovery was maximum at 35˚C. From Table 3, lower temperature, initial metal concentration and pressure adequately satisfy the maximum recoveries and make the process economically feasible. Table 3 3.3. Regression Results Analysis of variance (ANOVA) is important to signify the coefficients of variables and their interactions are significant or not. Results are summarized in Table 4 and 5 for the lead and nickel, respectively. Table 4 Table 5 F value and P value are the two key values used to determine which terms are statistically significant. The larger value of F (>1) and the lower value of P (<0.0001) indicates the corresponding term is more significant. According to Tables 4 and 5, all the linear terms “t, p, cin” are significant. Moreover, except the term “p.t” the other non-linear terms “cin.t, and cin.p” are also significant. Based on the ANOVA results, all independent variables with their interactions were included in Equation 5 and the coefficient of each terms were estimated. Table 6 shows the values of the coefficients of the model for the variation of nickel and lead concentrations in filtrate. Negative and positive signs of each coefficient reflect the effect of each term either in the purpose of descending or ascending of the response, respectively. The values of the coefficient of determination (R2) for both models were found to be higher than 0.99. It defines that observed outcomes can be replicated by the nickel and lead models very well based on the proportion of total 14
variation of outcomes explained by the models. However, it should be noted that the present models are only valid under the conditions 50
4. Conclusion Removal of heavy metals from wastewater include conventionally integration of process units with complementary treatment characteristics which make the overall process inefficient in terms of volume, energy and time. Membrane adsorber is a unique alternative for effectively and efficiently separate toxic substances from liquid phase. Adsorption and filtration are coupled in membrane adsorber provides high removal efficiency, favorable hydrodynamics, acceptable reusability and elimination of an extra step which can be required to separate adsorbent from treated water. However, the behavior of the membrane adsorber under different operating conditions needs to be clarified in order to minimize the filtrate concentration of target species. Increasing pressure will increase the volume filtrate and increase the transport of metal cations as well. Increasing pressure will reduce the contact time necessary for the reaction to take place especially in the reaction controlled regime. Similarly, initial metal concentration will cause to rise the metal concentration in permeate if rate of reaction is very small than the diffusional rate. Filtration time is another 15
parameter that should be controlled due to limited adsorption capacities of membrane adsorbers. In this study, these effects on the transport behavior of the membrane were determined by calculating Thiele modulus at each filtration condition. According to the values of the Thiele modulus, the transport of the lead and nickel cations were controlled by reaction. A nonlinear multiple rational regression has been introduced to analyze the whole filtration data of the lead and nickel. Both experimental and predicted results suggested that the PSf10 membrane should be used under low initial metal concentration and pressure in a short filtration time. As a conclusion, the present study is important to signify that the filtration and removal properties of a poor PSf membrane can be improved in a simple manner by incorporation of inorganic nanoparticles. Acknowledgement We would like to thank the Scientific and Technological Research Council of Turkey for the financial support through Grant 113Y159. Celal Bayar University, Food Engineering Department is gratefully acknowledged for providing AAS analysis.
References [1] R.J. Gohari, W.J. Lau, T. Matsuura, E. Halakoo, A.F. Ismail, Adsorptive removal of Pb(II) from aqueous solution by novel PES/HMO ultrafiltration mixed matrix membrane, Sep. Purif. Technol. 120 (2013) 59-68. [2] R. Mukherjee, S. De, Adsorptive removal of nitrate from aqueous solution by polyacrylonitrile–alumina nanoparticle mixed matrix hollow-fiber membrane, J. Membrane Sci. 466 (2014) 281-292. [3] E. Salehi, S.S. Madaeni, L. Rajabi, A.A. Derakhshan, S. Daraei, V. Vatanpour, Static and dynamic adsorption of copper ions on chitosan/polyvinyl alcohol thin adsorptive membranes: Combined effect of polyethylene glycol and aminated multi-walled carbon nanotubes, Chem. Eng. J. 215–216 (2013) 791–801. [4] J.W. Gilman, Flammability and thermal stability studies of polymer layered-silicate (clay) nanocomposites, Appl. Clay Sci., 15 (1999), pp. 31–49.
16
[5] A. Okada, A. Usuki, The chemistry of polymer–clay hybrids, Mater. Sci. Eng., C3 (3) (1995), pp. 109–115. [6] S.I. Yu, X. Zuo, R. Bao, X. Xu, J. Wang, J. Xu, Effect of SiO2 nanoparticle addition on the characteristics of a new organic–inorganic hybrid membrane, Polymer, 50 (2009), pp. 553– 559. [7] M. Fathizadeh, A. Aroujaliana, A. Raisi, Effect of added NaX nano-zeolite into polyamide as a top thin layer of membrane on water flux and salt rejection in a reverse osmosis process, J. Membrane Sci. 375 (2011) 88–95. [8] J. Huang, K. Zhang, K. Wang, Z. Xie, B. Ladewig, H. Wang, Fabrication of polyethersulfonemesoporous silica nanocomposite ultrafiltration membranes with antifouling properties, J. Membrane Sci. 423-424 (2012) 362-370. [9] S.J. Oh, N. Kim, Y.T. Lee, Preparation and characterization of PVDF/TiO2 organic-inorganic composite membranes for fouling resistance improvement, J. Membrane Sci. 345 (2009) 1320. [10] G. Arthanareeswaran, T.K. Sriyamuna Devi, M. Raajenthiren, Effect of silica particles on cellulose acetate blend ultrafiltration membranes: Part I, Sep. Purif. Technol. 64 (2008) 38-47. [11] R. Han, S. Zhang, C. Liu, Y. Wang, X. Jian, Effect of NaA zeolite particle addition on poly(phthalazinone ether sulfone ketone) composite ultrafilitration (UF) membrane performance, J. Membrane Sci. 345 (2009) 5-12. [12] F. Liu, B.R. Ma, D. Zhou, Y.H. Xiang, L.X. Xue, Breaking through tradeoff of Polysulfone ultrafiltration membranes by zeolite 4A, Micropor. Mesopor. Mat. 186 (2014) 113–120. [13] A. Charkhia, M. Kazemeinia, S.J. Ahmadib, H. Kazemian, Fabrication of granulated NaY zeolite nanoparticles using a new method and study the adsorption properties, Powder Technol. 231 (2012) 1–6. [14] Y. Wang, Z. Lei, B. Chen, Q. Guo, N. Liu, Adsorption of NO and N2O on Fe-BEA and HBEA zeolites, Appl. Surf. Sci. 256 (2010) 4042–4047. [15] S. Malamisa, E. Katsoua, A review on zinc and nickel adsorption on natural and modified zeolite, bentonite and vermiculite: Examination of process parameters, kinetics and isotherms, J. Hazard. Mater. 252– 253 (2013) 428– 461. [16] Y.Yurekli, Removal of heavy metals in wastewater by using zeolite nano-particles impregnated polysulfone membranes, J. Hazard. Mater. 309 (2016) 53–64. 17
[17] S.M. Hosseini, S. Rafiei, A.R. Hamidi, A.R. Moghadassi, S.S. Madaeni, Preparation and electrochemical
characterization
of
mixed
matrix
heterogeneous
cation
exchange membranes filled with zeolite nanoparticles: Ionic transport property in desalination, Desalination 351 (2014) 138-144. [18] J. Dasgupta, M. Singh, J. Sikder, V. Padarthi, S. Chakraborty, S. Curcio, Response surfaceoptimized removal of Reactive Red 120 dye from its aqueous solutions using polyethyleneimine enhanced ultrafiltration, Ecotox. Environ. Safe. 121 (2015) 271–278. [19] Y. Zheng, A. Wang, Removal of heavy metals using polyvinyl alcohol semi-IPN poly(acrylic acid)/tourmaline composite optimized with response surface methodology, Chem. Eng. J. 162 (2010) 186–193. [20] H. Bessbousse, T. Rhlalou, J.-F. Verchere, L. Lebrun, Removal of heavy metal ions from aqueous solutions by filtration with a novel complexing membrane containing poly(ethyleneimine) in a poly(vinyl alcohol) matrix, J. Membrane Sci. 307 (2008) 249–259. [21] W.M. Deen, Hindered transport of large molecules in liquid-filled pores, AIChE J. 33 (1987) 1409–1425. [22] W.R. Bowen, A.W. Mohammad, N. Hilal, Characterisation of nanofiltration membranes for predictive purposes—use of salts, uncharged solutes and atomic force microscopy J. Membrane Sci. 126 (1997) 91–105. [23] H. Pham, Springer Handbook of Engineering Statistics 2006, Springer-Verlag London [24] V.M. Taavitsainen, Rational function ridge regression in kinetic modeling: A case study, Chemometrics and Intelligent Laboratory Systems 120 (2013) 136–141. [25] X. Li, Z. Wang, Q. Li, J. Ma, M. Zhu, Preparation, characterization, and application of mesoporous silica-grafted graphene oxide for highly selective lead adsorption, Chem. Eng. J. 273 (2015) 630–637. [26] S.S. Madaeni, S. Zinadini, V. Vatanpour, Convective flow adsorption of nickel ions in PVDF membrane embedded with multi-walled carbon nanotubes and PAA coating, Sep. Purif. Technol. 80 (2011) 155–162. [27] S. Mondal, R. Mukherjee, S. Chatterjee, and S. De, Adsorption-Concentration Polarization Model for Ultrafiltration in Mixed Matrix Membrane, AIChE J. 60 (2014) 2354-2364.
18
[28] S. Habibi, A. Nematollahzadeh, S.A. Mousavi, Nano-scale modification of polysulfone membrane matrix and the surface for the separation of chromium ions from water Chem. Eng. J. 267 (2015) 306–316. [29] N. Abdullah, R.J. Gohari, N. Yusof, A.F. Ismail, J. Juhana, W.J. Lau, T. Matsuura, Polysulfone/hydrous ferric oxide ultrafiltration mixed matrix membrane: Preparation, characterization and its adsorptive removal of lead (II) from aqueous solution, Chem. Eng. J. 287 (2016) 28-37. [30] L. R. Rad, A. Momeni, B. F. Ghazani, M. Irani, M. Mahmoudi, B. Noghreh, Removal of Ni2+ and Cd2+ ions from aqueous solutions using electrospun PVA/zeolite nanofibrous adsorbent, Chem. Eng. J. 256 (2014) 119–127.
19
List of Figures
Figure 1: Experimental set-up of cross-flow filtration. Figure 2: Adsorption kinetics of the metal ions in the native (open symbols) and NaX added (solid symbols) PSf membranes. Lines denote the best fits from Lagergren pseudo first order kinetic model. (Cin = 500 mg/L). Figure 3: Adsorption-desorption cycles of the hybrid membrane against lead and nickel ions. Figure 4. Concentration of lead (a) at the solid phase, (b) in the liquid phase and nickel (c) at the solid phase and (d) in the liquid phase under operating pressure of 0.5 bar. Figure 5: Representative response surface plots for the effect of initial metal concentrations and filtration time on (a) Pb2+ and (b) Ni2+ concentrations in filtrate at 0.5 bar transmembrane pressure. (Red circles represent the experimental data).
20
Figure 1
0.6
Sorption capacity (mg/cm2)
lead-PSf10 0.5
nickel-PSf10 lead-PSf
0.4
nickel-PSf
0.3 0.2 0.1 0 0
10
20
30
40
Time (min)
Figure 2
21
50
60
70
100
lead adsorption
nickel adsorption
lead desorption
nickel desorption
% sorption
80 60 40 20 0 1
2
3
4
5
number of runs
Figure 3 160
(a)
50 ppm
Filtrate concentration (mg/L)
Adsorbed amount (mg/cm2)
0.5 100 ppm
0.4
200 ppm 0.3 0.2 0.1
50 ppm
140
(b)
100 ppm
120
200 ppm
100 80 60 40 20 0
0 0
50
100
0
150
50
Time (min) 0.2
150
160
(c)
50 ppm
Filtrate concentration (mg/L)
Adsorbed amount (mg/cm2)
100 Time (min)
100 ppm
0.15
200 ppm
0.1
0.05
0
50 ppm
140
(d)
100 ppm
120
200 ppm
100 80 60 40 20 0
0
50
100
150
0
50
100 Time (min)
Time (min)
Figure 4 22
150
(b)
(a)
Figure 5
23
Table 1: Properties of the membranes.
Membranes
Native PSf PSf10
Porosity* (%)
Pore radius* (nm)
BSA rejection (%)
Lp (L/m2.h.bar)
82
10
100
23.2
19
98
57.1
74 * from reference [16]
24
Table 2: Estimated values of Thiele modulus and the rate constants during filtration of lead and nickel ions.
C0 (mg/L)
P (bar)
50 50 50 100 100 100 200 200 200
0.5 1.0 2.0 0.5 1.0 2.0 0.5 1.0 2.0
50 50 50 100 100 100 200 200 200
0.5 1.0 2.0 0.5 1.0 2.0 0.5 1.0 2.0
k1 (1/min)
Error
1
k2 (cm2/mg.min)
Error
2
Pb2+
0.002 0.003 0.007 0.009 0.009 0.012 0.013 0.017 0.017
1.610-5 2.810-4 2.910-4 8.910-4 9.210-4 1.910-3 3.610-3 4.510-3 5.210-3
0.027 0.036 0.057 0.063 0.065 0.072 0.078 0.088 0.087
0.001 0.001 0.005 0.006 0.007 0.007 0.014 0.011 0.013
2.010-5 2.810-4 2.310-4 7.010-4 7.510-4 1.410-3 2.710-3 3.210-3 5.310-3
0.001 0.001 0.002 0.004 0.004 0.004 0.008 0.007 0.007
Ni2+
0.016 0.013 0.015 0.011 0.029 0.024 0.035 0.012 0.026
2.810-5 4.110-4 3.610-4 1.010-4 1.210-3 1.910-3 6.810-4 2.410-3 4.410-3
0.099 0.090 0.096 0.082 0.134 0.122 0.148 0.085 0.128
0.083 0.034 0.040 0.020 0.076 0.060 0.165 0.015 0.089
4.210-5 3.010-4 2.710-4 9.410-5 6.110-4 8.910-4 5.410-4 2.210-3 3.110-3
0.011 0.007 0.008 0.008 0.015 0.013 0.031 0.009 0.023
Metals
25
Table 3: Percentage recovery of the PSf10 membrane with respect to operating conditions (initial metal concentrations, operating pressures and temperatures).
Initial concentration (mg/L) 50 50 50 50 50 100 100 100 200 200 200 50 50 50 50 50 100 100 100 200 200 200
Pressure (bar) 0.5 1.0 2.0 0.5 0.5 0.5 1.0 2.0 0.5 1.0 2.0 0.5 1.0 2.0 0.5 0.5 0.5 1.0 2.0 0.5 1.0 2.0
Temperature (˚C) 25 25 25 35 50 25 25 25 25 25 25 25 25 25 35 50 25 25 25 25 25 25
26
Metals
Pb2+
Ni2+
% Recovery 86.1 89.8 55.0 79.0 90.2 50.9 50.2 38.8 33.9 32.2 21.5 40.6 29.8 33.7 46.8 31.2 39.4 27.2 24.2 25.5 14.9 12.9
Table 4: Analysis of variance (ANOVA) for the lead
t p cin pt cin t cin p Error Total R2=0.99
Degrees of freedom 24 2 2 48 48 4 95 223
Sum of squares
Mean Squares
98887.4 26066.7 452086.0 4593.4 32377.1 20314.8 9817.9 644143.0
4120.31 13033.30 226043.00 95.70 674.52 5078.70 103.35
27
F ratio
P value
39.87 126.11 2187.23 0.93 6.53 49.14
2.29 x 10-39 1.83 x 10-27 3.59 x 10-80 0.61 4.80 x 10-15 2.43x 10-22
Table 5: Analysis of variance (ANOVA) for the nickel
t p cin pt cin t cin p Error Total R2=0.99
Degrees of freedom 24 2 2 48 48 4 96 224
Sum of squares 73828.1 14443.1 420360.0 1456.7 43479.2 1937.1 2200.0 557705.0
Mean Squares 3076.17 7221.56 210180.00 30.35 905.82 484.28 22.92
28
F ratio 134.23 315.12 9171.43 1.32 39.53 21.13
P value 2.88 x 10-63 6.56x 10-43 2.48 x 10-110 0.12 5.14 x 10-46 1.61 x 10-12
Table 6: The regression coefficients of the rational model for nickel and lead
a0 a1 a2 a3 a4 a5 a6 a7 a8 a9 b0 b1 b2 b3 b4 b5 b6 b7 b8 b9
Cf,nickel
Cf,lead
-4420.09 66.75 1.81 37.44 -0.05 15388.89 -6720.15 -277.72 24.95 15.85 2383.30 18.75 -0.01 -1.43 -0.02 -2577.44 668.55 -1.42 2.76 0.01
10614.92 -8776.64 3.82 -2.65 -0.93 -28535.33 7000.03 2192.13 150.99 159.95 1294.90 155.29 -0.16 13.41 -0.02 2482.45 -1046.25 16.42 -6.59 -0.01
29