- Email: [email protected]
Nanofiltration applied in gold mining effluent treatment: Evaluation of chemical cleaning and membrane stability
Accepted Manuscript Nanofiltration applied in gold mining effluent treatment: evaluation of chemical cleaning and membrane stability L.H. Andrade, B.C. Ricci, L.B. Grossi, W.L. Pires, A.O. Aguiar, M.C.S. Amaral PII: DOI: Reference:
S1385-8947(17)30660-5 http://dx.doi.org/10.1016/j.cej.2017.04.116 CEJ 16870
To appear in:
Chemical Engineering Journal
Received Date: Revised Date: Accepted Date:
17 February 2017 24 April 2017 25 April 2017
Please cite this article as: L.H. Andrade, B.C. Ricci, L.B. Grossi, W.L. Pires, A.O. Aguiar, M.C.S. Amaral, Nanofiltration applied in gold mining effluent treatment: evaluation of chemical cleaning and membrane stability, Chemical Engineering Journal (2017), doi: http://dx.doi.org/10.1016/j.cej.2017.04.116
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.
Nanofiltration applied in gold mining effluent treatment: evaluation of chemical cleaning and membrane stability L. H. ANDRADE a*, B. C. RICCI a, L. B. GROSSI a, W. L. PIRES a, A. O. AGUIAR a, M. C. S. AMARAL a a
Department of Sanitary and Environmental Engineering, Federal University of Minas
Gerais. Address: Av. Antônio Carlos, nº 6627, Pampulha, Belo Horizonte, Minas Gerais, Brazil. * Corresponding author: Telephone: +55 (31) 34093669; Fax: +55 (31) 34091879; Email: [email protected] ABSTRACT The objectives of this study were (1) to determine the best conditions of chemical cleaning of a nanofiltration (NF) membrane (NF90) employed in the treatment of gold mining effluent, and (2) to investigate the effects of continuous exposure to mining effluent and an acid cleaning agent on NF characteristics. NF fouling was mainly due to inorganic species, and calcium sulfate (CaSO4.0.5H2O) was identified over membrane surface. Therefore, acid solutions were the most efficient in membrane cleaning, and among them hydrochloric acid (HCl) showed the best performance. Cleaning without recirculation (soak) was not effective and thus 90-minute recirculation was chosen as the best cleaning procedure. The NF membrane was exposed to the effluent and to the combined effluent and HCl solution for 285 days. Modifications in hydrophobicity, effective pore radius, fouling, and surface charge were observed. The retention of magnesium sulfate and glucose by the membrane exposed to the combined effluent and cleaning solution was statistically lower, what endorses the importance of cleaning conditions optimization. No indication of degradation of the polymeric material of the
1
membrane was observed. Thus, the results of this study are promising and demonstrate the robustness of NF technology for gold mining effluent treatment. KEYWORDS Gold mining effluent; Nanofiltration; Membrane lifetime; Membrane cleaning
1. INTRODUCTION Nanofiltration (NF) is a separation process that employs membranes with a typical pore size of 1 nm, which corresponds to a molecular weight cut-off (MWCO) of 300–500 Da. NF presents low rejection of monovalent ions, high rejection of multivalent ions, and higher flux compared to reverse osmosis (RO) membranes. These properties have allowed NF to be used in many niche applications, especially in the treatment of water and wastewater [1]. In this way, several authors have studied NF for the treatment of mining effluent and have obtained promising results regarding pollutant efficiency retention [2, 3]. In addition to the cited studies, NF has proved to be a favorable treatment for the effluent from the gold mining industry, specifically, from a sulfuric acid production plant and a calcined dam. As reported by Andrade et al. [4], the NF of the combined streams of these effluents allowed for generation of high quality treated effluent. Despite the excellent performance, membrane fouling management is essential to ensure consistent separation performance with minimal cleaning and membrane replacement. Fouling is a major limitation of NF and is related to operational conditions and to feed stream composition. Common fouling mechanisms include the following: pore plugging and external pore blocking, resulting from deposition of particles and colloids; precipitation of dissolved materials on membrane pores and surfaces; build-up of a 2
cake/gel-like layer on the upstream face of a membrane; absorption or adsorption; and molecular-membrane electrostatic interactions and acid–base interactions [5-8]. The main consequences of fouling include flux decline, permeate quality deterioration, and an energy consumption increase. Accordingly, fouling control is essential for increasing membrane operational lifetime and thus reducing process costs. Feed pre-treatment, membrane selection, module design, operation mode, and membrane cleaning are the main membrane fouling control strategies [9, 10]. Membrane cleaning methods are classified into physical, chemical, physico-chemical and biochemical [11]. Chemical cleaning is the most common method, especially in NF and RO systems. It is used to remove fouling by means of specific chemical agents, which typically include acids, bases, chelating agents, and/or surfactants [12] The optimal selection of the cleaning agent depends mainly on the membrane material and type of foulant. The ideal cleaning agent should remove the deposited material without damaging the membrane, but often prolonged exposure can result in morphological, structural, and superficial changes, reducing the ability of the membrane to reject solutes. This leads to a reduction of membrane module lifetime, which is related to the characteristics of the effluent, fouling rate, operating conditions, and hydraulic cleaning cycle [8]. In this context, the objectives of this study were (1) to determine the optimal conditions of chemical cleaning of an NF membrane (NF90) employed in the treatment of gold mining effluent, and (2) to investigate the effects of continuous exposure to mining effluent and a cleaning agent on NF membrane properties.
3
2. MATERIALS AND METHODS 2.1 Gold ore processing effluents Two effluents from a gold mining company in Brazil were studied, i.e., the effluent from the sulfuric acid production plant and the water from the calcined dam (Table 1). The industrial process description and the effluent generation points are as described in Andrade et al. [4]. Preliminary NF experiments have shown that membrane process performance was improved when applied to these two effluents in combination instead of to a single effluent. Thus, both effluents were mixed at a 1:1 ratio (corresponding to actual wastewater flow rates generated by the company, which are 140 m³·h-1 of effluent from the sulfuric acid production plant and 140 m³·h-1 of water from the calcined dam) before being transferred to the membrane treatment system. In this study, such a mixture will hereinafter be referred to as “effluent from gold mining”. Table 1 – Characteristics of gold ore processing effluent Parameters pH Conductivity (µS·cm-1) Dissolved organic carbon (mg·L-1) Dissolved inorganic carbono (mg·L-1) Total solids (mg·L-1) Suspended solids (mg·L-1) Sulfate (mg·L-1) Chloride (mg·L-1) Aluminum (mg·L-1) Arsenic (mg· L-1) Calcium (mg·L-1) Iron (mg·L-1) Magnesium (mg·L-1) Manganese (mg·L-1) Potassium (mg·L-1)
Effluent from the sulfuric acid production plant 1.82 9,420
Water from the calcined dam
Effluent from gold mining1
8.23 3,250
2.08 6,615
4.4
5.8
5.0
0.2
19.2
9.7
9,398 25 6,367 31 121 657 382 75 462 28 49
3,159 14 1,944 291 0.1 2 541 0.2 29 0.1 52
8,804 15 3,864 168 67 322 550 34 247 22 44 4
Sodium (mg·L-1) 45 168 77 -1 Zinc (mg·L ) 115 0.1 64 1 Mixture at a 1:1 ratio of effluent from the sulfuric acid production plant and of the water from the calcined dam 2.2 Experimental setup The treatment process proposed and evaluated in the bench-scale for the effluent consisted of a pH adjustment to 5.0 with a sodium hydroxide (NaOH) solution, ultrafiltration, and nanofiltration. Ultrafiltration (UF) was performed with a commercial submerged membrane module (ZeeWeed) with a filtration area of 0.047 m², PVDF-based polymer, and average pore diameter of 0.04 µm. A bench-scale unit was used; it comprised one pump equipped with a speed controller, one needle-type valve for pressure adjustment, and one manometer. NF treatment was performed with a Dow FilmTec NF90 membrane, which is a thin film composite membrane comprised of three layers: (1) a polyester support web, (2) a microporous polysulfone interlayer, and (3) an ultrathin aromatic polyamide active layer. The MWCO is 100 Da [13], the average membrane hydraulic resistance is 5.8 × 10 13 m-1, and the sodium chloride (NaCl) (2,000 mg·L-1) and MgSO4 (2,000 mg·L-1) rejection are 85-95% and 97%, respectively [14]. The maximum allowed NF operating pressure of the setup was 15 bar, which was provided by a rotary vane pump (Procon) with a maximum flow of 530 L·h-1. A needle valve was used to adjust the feed flow rate and the trans-membrane pressure. The stainless steel cell had a 9.8 cm diameter and an effective filtration area of 75 cm². The schematic view of the UF/NF setup is shown in Figure 1.
5
(a)
(b) Figure 1 – Schematic of (a) ultrafiltration and (b) nanofiltration bench-scale unit
2.3 Fouling evaluation 2.3.1 Experimental procedure To investigate NF fouling, a virgin membrane was inserted into the NF cell and compacted with distilled water at 10 bar. After flux stabilization, water permeability was measured by monitoring the stabilized water permeate flux at pressures of 10.0, 7.5, 5.0, and 2.5 bar using a measuring cylinder and a chronometer. Water temperature was also monitored and permeate flux was normalized to 25°C by means of a correction factor calculated as water viscosity at the temperature of permeation divided by water viscosity at 25°C [15]. 6
Water flux and permeability standard deviations were calculated according to the equation for determining the combined standard uncertainty (Equation 1).
= ∑ × ²
(Equation 1)
where is the combined standard deviation of the variable y, f is the function
y=f(x1, x2, ..., xN), and ² is the uncertainty related to the parameter xi. The water flux (J) and permeability (Lp) can be calculated using Equations 2 and 3.
= × × ² ×
(Equation 2)
!"º#
&
$% = '
(Equation 3)
where V is the volume of permeate (L) collected in a given time t (h), D is the diameter of the filtration cell (m), P is the pressure (bar), and () is the permeate viscosity at temperature T. The permeate viscosity was assumed to be equal to water viscosity and thus was calculated at a given temperature T (K) by Equation 4. () = 2.414 × 10/0 × 10 1.2⁄)/ 3
(Equation 4)
The combined uncertainty of J and Lp can thus be calculated by: = $% ×
5! !
$% = $% ×
+
5! !
5!
+
!
+
5! !
5!
+
!
+
5! !
5! ) 1.2!
+
)/ 3 7
5! ' '!
+
5! ) 1.2! )/ 3 7
(Equation 5)
(Equation 6)
Next, distilled water was substituted by 8.0 L of pre-treated effluent (after pH adjustment and ultrafiltration, as detailed in Andrade et al. [4]). NF was performed with a constant operational pressure of 10 bar and a feed flow rate of 144 L·h-1, which resulted in 1.9 m·s-1 cross-flow velocity and a Reynolds number of 847. The cross-flow velocity was chosen based on the literature (e.g. Sikder et al. [16] used cross-flow 7
velocities varying from 1.77 to 2.48 and Mukherjee et al.[17], Reynolds numbers from 800 to 1200). The system was operated with a continuous removal of permeate and recycling of concentrate back to the feed tank, reaching a permeate recovery rate (RR) of 71% after 2,100 minute-filtration. To evaluate salt precipitation on membrane surfaces, which results in scaling, the calcium sulfate supersaturation index at membrane surface (89: ) was determined for different RRs according to Ricci et al. [2]. Calcium sulfate was chosen as representative of salt precipitation as it has a low solubility and a high concentration in effluent from gold mining (Table 1) presenting, therefore, a greater precipitation potential. Hermia’s semi-empirical models modified for crossflow filtration [18] were used to identify the fouling mechanism involved during the NF process. Also, reversible and irreversible fouling resistances were also determined (Section 2.3.2). Feed and NF permeate samples were collected and analyzed according to the parameters conductivity (Hanna conductivity meter HI 9835) and ions sulfate, chloride, calcium, and magnesium (ion chromatograph Dionex ICS-1000 equipped with AS-22 and ICS 12a columns). The entire procedure was performed in duplicate and results presented in this paper correspond to averages. Following the procedure, two fragments of fouled membranes were obtained (from each duplicate). The first fragment was subjected to physical cleaning using distilled water recirculation, at 144 L·h-1 for 30 minutes. After physical cleaning, the water permeability was measured. This result was used to calculate the resistance to filtration (Section 2.3.2). The second membrane fragment was removed from the system. The precipitate formed on the membrane surface was gently removed with a spatula and analyzed by X-ray
8
diffraction (XRD) (Section 2.6.5). Samples of virgin and fouled membranes were also analyzed in a scanning electron microscope (SEM) coupled with an energy dispersive X-ray spectrometer (EDS) (Section 2.6.1).
2.3.2 Resistances in series calculations Resistances were calculated using the resistance in series concept. Total resistance (; ) comprises three components (Equation 7): intrinsic membrane resistance (;: );
reversible fouling resistance (; <=> , which comprised the concentration polarization resistance and the resistance from fouling removable by physical cleaning; and irreversible fouling resistance (; <<=> ), which cannot be removed by physical cleaning. ; = ;: + ; <=> + ; <<=>
(Equation 7)
Using the virgin membrane water permeability, the intrinsic membrane resistance to filtration (;: ) was calculated according to Equation 8: R@ =
ABCDECF ×G
(Equation 8)
where H>
L/MNO
G × PQRSTUQFV
(Equation 9)
where P is the applied transmembrane pressure (106 Pa), W is the reflection coefficient estimated by the averaged membrane rejection of the major constituents of effluent [19], and XY is the difference in osmotic pressure of the solution at membrane surface and 9
permeate streams. Thus, Z − WΔπ is the effective pressure, in Pa. In addition, Jeffluent is the permeate flux in m3·h-1·m-2 obtained at the end of the fouling experiment (after 2,100 minutes of filtration). The difference in osmotic pressure obtained with different permeate recovery rates were estimated by the Van’t Hoff equation, as described in Equation 10: Δπ = ;^ΣΔC
(Equation 10)
where R is the universal gas constant (L·Pa·K-1·mol-1), T is the permeation temperature (K), and ΣX` is the sum of the difference in molar concentrations (mol·L-1) of dissolved species present at membrane surface and the permeate. Only the most concentrated ions (sulfate, calcium, magnesium and arsenic) were taken into consideration to ΣX` determination. The resistance of the irreversible fouling can be calculated by the permeability with water after the physical cleaning process (H<<=> ) (Equation 11). R <<=> =
Aaabc ×G
− ;:
(Equation 11)
The reversible fouling resistance can be calculated by Equation 12. R <=> = R K − R @ − ; <<=>
(Equation 12)
2.4 Evaluation of NF membrane cleaning Cleaning tests were performed with different chemical agents and, after selection of the most appropriate, tests with different soak and recirculation times were carried out. The cleaning test consisted of six steps: (1) compaction of the virgin membrane at 10 bar 10
with distilled water; (2) measurement of distilled water permeability ($%J=d ; (3) performance of NF fouling procedure with effluent; (4) measurement of fouled membrane distilled water permeability ($% e5f=g ); (5) accomplishment of the cleaning procedure; and (6) measurement of the cleaned membrane distilled water permeability $%f=hJ=g . The fouling procedure consisted of feeding 3 L of pre-treated effluent into the NF unit, and performing NF at a constant pressure of 10 bar and a feed flow rate of 144 L·h-1 for 4 hours in concentration mode. For tests with different cleaning solutions, the cleaning procedure consisted of recirculating the solution in the NF unit for 90 minutes with a flow rate of 144 L·h-1 and a pressure of 1.5 bar. Different cleaning solutions were tested, specifically: 0.2% hydrochloric acid, 2.0% citric acid, 0.5% phosphoric acid, 0.2% nitric acid, 0.1% sulfuric acid, 0.4% sodium hydroxide, 0.03% dodecylbenzene sodium sulfate (DDBS), and 1.0% sodium ethylenediamine tetraacetic acid (EDTA) (similar to the solutions suggested by the supplier of the membranes [20]). A blank test was also conducted using only water. After that, the cleaning mode (only soaking or soaking associated with recirculation) and time were assessed for the best chemical cleaning agent. The procedure consisted of adding the cleaning solution in the NF unit, leaving it in static contact with the fouled membrane for different soak times (0 up to 60 minutes), then starting the pump and recirculating the solution for 15 up to 120 minutes. The cleaning efficiency in each test was calculated using Equation 13: `ijklmln joompjlp % =
rstubvwbx /rsyz{ubx rswb| /rsyz{ubx
(Equation 13)
11
2.5 Assessment of membrane stability to the effluent and the cleaning solution 2.5.1 NF membrane immersion procedure The effects of continuous exposure to effluent and chemical cleaning solutions on the characteristics of the NF membrane were evaluated by immersing four pristine membrane fragments in NF effluent concentrate (obtained using the conditions established in Andrade et al. [4]) for a period of 285 days. At four-week intervals, fragments were removed from the solution and rinsed with distilled water. Two of these fragments were immersed in a cleaning solution of 1.0% HCl overnight (16 hours), and the other two remained immersed in distilled water. The concentration of the cleaning solution and the cleaning time were higher than the ones used in Section 2.4 (given an exposition to cleaning agent approximately 50 times higher) so that the effect of the acid on the membrane could be more readily observed. Subsequently, filtration tests (Section 2.5.2) and analysis of the morphological and chemical characteristics (SEM, EDS, contact angle, atomic force microscopy, and attenuated total reflection Fourier transform infrared spectroscopy - Section 2.6) were carried out on two of the fragments, one that was subjected to contact with the effluent and cleaning solution (}~= f=hJ ) and one that was exposed only to effluent (}~= ). At the end of the exposure period, tests were also conducted to estimate the effective pore radius of the membrane (Section 2.5.3).
2.5.2 Filtration tests The filtration tests consisted of three steps: (1) hydraulic permeability determination; (2) evaluation of magnesium sulfate rejection; and (3) evaluation of glucose rejection. Magnesium sulfate retention tests were performed using a 2,000 ppm solution as the feed. Two liters of solution were fed into the NF unit and nanofiltrated at a pressure of 5 12
bar and temperature of 25±3°C, until a permeate recovery rate of 15% was achieved [21]. Feed and permeate were collected and analyzed for conductivity. A calibration curve was constructed and the magnesium sulfate removal efficiency was calculated. Glucose retention tests were performed in similar form, using a 500 ppm solution, pressure of 10 bar and final permeate recovery rate of 10% [22]. Feed and permeate carbohydrates were analyzed and glucose retention was calculated [23].
2.5.3 Estimation of the effective pore radius of the NF membrane Methanol rejections were determined as a function of permeate flux for pressures of 412 bar for virgin membrane and membranes after 285 days of exposure. A 2,000 ppm solution was used; since the concentration was not high, membrane swelling was not an issue. The rejection data were interpreted in terms of real rejection as shown in Equation 14. ;< = 1 −
(Equation 14)
Here, Rri and Cmi are the real rejection and concentration of solute i at the membrane surface, respectively. Results of methanol rejections as a function of pressure were analyzed using the Spiegler-Kedem and Steric Hindrance Pore (SHP) models [24]. The Spiegler-Kedem model equation is given by Equation 15 [25]. ;< =
σ/ /σ
ℎjj ~ = exp −
/σ '
(Equation 15)
where J is the permeate flux, P is the permeability of solute i, and σ is the reflection coefficient.
13
For neutral solutes where the convective flux is not influenced by electrostatic effects, separation occurs according to the size exclusion mechanism. Thus, only the ratio of the radius of the solute ( , considered equal to the Stokes radius, which corresponds to 0.135 nm for methanol [26]) to the pore radius (s ) of the membrane determines the reflection coefficient, as indicted in Equation 16, which is deduced from the SHP model [25]: < !
<
<
σ = 1 − 1 + < ! 1 − < 2 − 1 − <
(Equation 16)
As can be seen from Equation 16, the SHP model can be used to estimate the pore radius for a given membrane. For this purpose, the reflection coefficient of a given solute must first be determined by Equation 15 and then the pore radius must be calculated using Equation 16.
2.6 Evaluation of morphological and chemical characteristics of membranes 2.6.1 Scanning electron microscopy (SEM) and energy dispersive X-ray spectrometer (EDS) Membrane surface morphology was analyzed using a FEI Quanta 200 SEM with an additional EDS analysis. Prior to analysis, membrane samples were air-dried and coated with 5 nm carbon layer by a sputter coating machine (Leica EM SCD 500 with a pressure of 10 -2-10-3 mbar and a 2.5 A current).
2.6.2 Atomic force microscopy (AFM) Tapping mode AFM was performed with an Asylum MFP-3D-SA/AFM microscope (Asylum Research) over an area of 5 × 5 µm². A silicon probe (AC240TS-R3, Olympus) was used. At least three different areas of each membrane were visualized 14
and the mean value of the root-mean-squared roughness (RRMS) was determined using the AFM software.
2.6.3 Contact angle measurements Water contact angles were determined using a DIGIDROP-DI goniometer (GBX Instruments) equipped with a CCD camera and an automated liquid dispenser, using the standard sessile drop method. Six µL of deionized water droplets were placed at the membrane surface at room temperature and the contact angle was measured. Five droplets were applied to each membrane to establish an average.
2.6.4 Attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy analysis ATR-FTIR experiments were carried out using a Shimadzu FTIR IR Prestige-21 instrument equipped with an attenuated total reflectance (ATR) accessory. The spectrum was obtained in the range of 400–4000 cm−1 at a 4 cm−1 resolution.
2.6.5 X-ray diffraction (XRD) Precipitate was analyzed by a Philips diffractometer (Panalytical) with an X’Pert-APD system, PW 3710/31 controller, PW 1830/40 generator, and PW 3020/00 goniometer was used. The radiation was emitted from a copper tube (Z = 29) with a mean wavelength Kα of 1.54184Å and λα1 of 1.54056Å.
3. RESULTS AND DISCUSSION 3.1 Fouling evaluation 15
Figure 2 illustrates the permeate flux as a function of operation time for the NF of the gold mining effluent. It is evident that the flux decreased strongly (~80%) until approximately 750 minutes of NF, followed by a tendency toward stabilization. This flux decline is justified partially by the high ions rejection by the membrane (Table 2), which causes increase in the concentration of ions at membrane surface, rising osmotic pressure and reducing effective pressure (Figure 2).
Effective pressure
70
10 9 8 7 6 5 4 3 2 1 0
Fux (L/h.m²)
60 50 40 30 20 10 0 0
500
1000
1500
Effective pressure (bar)
Flux
2000
Time (min)
Figure 2 – Permeate flux (J) and effective pressure as a funtion of operating time for NF fouling evaluation during gold mining effluent filtration Table 2 – Characterization of gold mining effluent and NF permeate obtained during NF fouling evaluation Parameter NF feed1 NF permeate Removal efficiency -1 Conductivity (µS.cm ) 4,550 398 91% Sulfate (mg.L-1) 2,620 168 94% Chloride (mg.L-1) 196 15 92% -1 Calcium (mg.L ) 673 24 96% Magnesium (mg.L-1) 148 6 96% 1
Pre-treated effluent (after pH adjustment and ultrafiltration, as detailed in Andrade et al. [4]) In addition, flux decline was also caused by fouling which was mainly associated with inorganic matter, due to effluent characteristics. The effluent is saturated with calcium 16
sulfate, with a supersaturation index (SI) over membrane surface ranging from 1.68 to 2.94, for a permeate recovery rate between 0 and 70%, respectively, allowing the salt to precipitate. An X-ray diffraction from the precipitate deposited over membrane surface indicated the presence of bassanite (hydrated calcium sulfate CaSO4.0.5H2O) as the only crystalline phase (Figure 3).
Figure 3 – X- ray diffractogram from the precipitate formed over the membrane during the fouling test
Through SEM micrographs of the fouled surface in association with EDS analysis (Figure 4), it was observed that, besides crystalline calcium sulfate, the fouling composition included amorphous structures composed of magnesium, sodium, zinc, copper, aluminum, iron, arsenic, and silica.
17
(a)
(b) Figure 4 – SEM micrographs and EDS spectrum from (a) the virgin membrane and (b) after 2,100 minutes of filtration 18
According to the modified Hermia’s model, there are four main fouling mechanisms, i.e., complete blocking, intermediate blocking, standard blocking, and cake formation . Complete blocking occurs when the size of foulants is bigger than the membrane pore size, therefore, pore blocking takes place over the membrane surface and not inside the membrane pores. It is also assumed that a molecule never settles over another one that had been previously deposited. Intermediate blocking is valid when molecules can obstruct the pore entrance without blocking it completely. It also considers the possibility of some molecules to settle over others. Standard blocking assumes that particles approaching the membrane are adsorbed or deposited on the internal pore walls, thereby reducing the pore volume. This type of fouling is caused by molecules smaller than the membrane pores. At last, cake formation is characterized by foulants that do not enter the membrane pores. Instead, they deposit over membrane surface, which results in cake or gel layer formation over the membrane surface [18]. According to Kaya et al. [27], the construction of curves of ln (J-1), J-1/2, J-1 and J-2 (where J is the permeate flux) as a function of time and evaluation of the best model fitness (R²) allows the identification of which of the four fouling mechanisms (complete blocking, standard blocking, intermediate blocking, and cake formation, respectively) is the most important. Thus, the experimental data was tested against the four mechanisms (Figure 5).
19
(a)
(b)
(c) (d) Figure 5 – Comparison of experimental data with the Hermia’s model: (a) complete blocking, (b) standard blocking, (c) intermediate blocking, and (d) cake formation.
It was observed that the predominant mechanism was intermediate blocking. This means that the precipitated particles are capable of depositing on the membrane surface or even on top of each other, forming multilayers. This hypothesis is also corroborated by the SEM image (Figure 4b). Intermediate blocking occurs when the solute molecule is similar to the membrane pore size, thus some molecules can obstruct a membrane pore entrance without blocking the pore completely [18]. Therefore, the initial CaSO4 crystalline salts formed must had a nanometric size. The membrane, reversible fouling (physically removed by cleaning with water recirculation), and irreversible fouling (not removed by physical cleaning) resistances were calculated at the end of the fouling test (Table S1 – Supplementary Material). It was observed that the highest resistance to NF was due to the membrane itself (5.85 x 10 13 m-1, corresponding to 39% of total resistance), which is a typical result of using NF 20
membranes [17]. Part of the fouling resistance can be removed by physical methods (26%), and thus can be controlled by adjusting the feed flow conditions. However, a significant portion of the fouling is considered irreversible (35%), thus requiring chemical cleaning. The results show the importance of optimizing the chemical cleaning process to ensure the maintenance of membrane permeability and the sustainability of the process. 3.2 Cleaning membrane evaluation In order to remove fouling and restore membrane permeability to allow for an increase in the membrane module’s lifetime and the process sustainability, several chemical substances were evaluated as cleaning agents (Figure 6). The purpose of a cleaning agent is to reduce the foulant-foulant and foulant-membrane interaction in such a way that the fouling layer can be removed from the membrane’s surface through mass transfer [12]. Acid and chelating agents reduce the inorganic foulant-membrane interaction, while bases act mainly on the organic deposits and amino acids, by hydrolysis and solubilization [12, 28], and on silica [8].
Cleaning efficiency
100% 80% 60% 40% 20% 0% HCl 0.2%
Citric H3PO4 acid 2% 0.5%
HNO3 0.2%
H2SO4 0.1%
NaOH 0.4%
DDBS 0.03%
EDTA 1.0%
Water
Figure 6 – Cleaning efficiency of several chemical agents Due to the low organic compound concentration found in the effluent (Table 1), the NF membrane fouling was caused mainly by inorganic species, as observed in Figure 3 and 21
Figure 4. Thus, the NaOH solution presented the lowest cleaning efficiency among all chemical agents evaluated. EDTA is a strong metal-chelating agent. It can react with calcium ions in sulfate and carbonate of calcium to form soluble complexes. It can also react through ligandexchange with calcium ions in organic matter-calcium complexes [12]. However, in this study, EDTA did not demonstrate a good efficiency. Al-Amoudi et al. [29] also observed that permeability was not totally restored after cleaning by a mixed cleaning agent that contained EDTA. It was justified by a permanent alteration of membrane superficial charge [29], which may increase its interaction with foulant material. Among the acids, the worst performance was observed for sulfuric acid (H2SO4), which can be associated with the tendency of formation of sulfate salt precipitate when the solution comes in contact with the calcium presented on membrane surfaces. Phosphoric acid can also provide insoluble calcium phosphate formation [5]. Nitric acid did not present satisfactory results, which is consistent with previous results in other studies [28]. Despite some suggestions that it is the most efficient cleaning agent [12], citric acid had a less relevant effect in this study, probably because at this solution pH (~3.0) calcium sulfate solubility is only 7% higher than at the pH 5.0 of effluent. The best cleaning agent in this study was hydrochloric acid. The advantage of HCl over other acids such as HNO3 and H2SO4 is that HCl has no oxidation ability for degradation of organic matter. This degradation causes formation of secondary fouling. Moreover, the obtained salt from the HCl agent is more soluble compared to the salts from other acids [28]. Therefore, HCl was selected as the most suitable chemical for NF membrane cleaning procedures.
22
The Figure 7 shows the cleaning efficiencies obtained with 0.2% HCl solution over different cleaning times. The total cleaning time includes the soak time (ranging from 0 up to 60 minutes) plus the recirculation time. Bars refer to standard deviation.
Figure 7 – Cleaning efficiencies obtained with 0.2% HCl solution over different cleaning times. The total cleaning time includes soak and recirculation times.
According to Wei et al. [12] during cleaning process, the cleaning agent contacts membrane surface foulants via mass transfer. The agent reacts with the foulants especially those that closest to membrane surface, yielding loosened fouling layer. Then, these reaction products are removed from membrane surface via mass transfer. As hydrodynamic conditions interfere directly in the mass transfer coefficients, cleaning without fluid movement (soak) is not very effective. Thus, it can be seen that the longer the soak time, the greater the total cleaning time to achieve cleaning efficiency > 90%. The condition that required the lower total cleaning time to remove fouling was the one
23
with only recirculation cleaning (soak time = 0 min), in which it was possible to obtain 94% efficiency after 90 minutes.
3.3 Evaluation of membrane stability to the effluent and the cleaning solution Figure 8 shows the variation in water permeability of membranes exposed to the effluent (}~= ) and to the effluent and the cleaning solution (}~= f=hJ ) as a function of time.
Water permeability (L/h.m².bar)
NF eff
NF eff+clean
10 8 6 4 2 0 0
50
100
150
200
250
300
Exposure time (days)
Figure 8 – Water permeability variation as a funtion of time of exposure to the membrane in contact with effluent (}~= ) and in contact with effluent and cleaning solution (}~= f=hJ )
Retention of magnesium sulfate (ionic solute) and glucose (neutral solute) over the period were also monitored (Figure 9).
24
(a)
(b) Figure 9 – Retention of (a) magnesium sulfate and (b) glucose as a funtion of the exposure time for the membrane in contact with effluent (}~= ) and in contact with effluent and cleaning solution (}~= f=hJ ) In order to evaluate the changes observed in the permeability and the magnesium sulfate and glucose retentions, contact angle ( Table 3), SEM (Figure 10), and AFM (Figure 11) with root mean square (RMS) roughness calculations (Table 4) were carried out.
Table 3 – Contact angle measurments of the virgin membrane and the membrane exposed to the effluent and to the effluent and the cleaning solution }~= }~= f=hJ Virgin Sample NF90 105 days 285 days 105 days 285 days 25
Contact Angle (º)
(a)
Value
48.0 ± 0.7
47.0 ± 0.5
(b)
55.0 ± 0.8
49.5 ± 1.5
57.3 ± 1.0
(c)
Figure 10 – SEM images of the NF90 membranes: (a) virgin, (b) exposed to the effluent after 285 days, and (c) exposed to the effluent and the cleaning solution after 285 days
(a)
(b)
(c) 26
Figure 11 – AFM images of the NF90 membranes: (a) virgin, (b) exposed to the effluent after 285 days, and (c) exposed to the effluent and the cleaning solution after 285 days
Table 4 – Root mean square roughness (RMS) measurements of the NF90 membranes: virgin, exposed to the effluent, and exposed to the effluent and the cleaning solution for 285 days Sample
NF90 virgin
RMS (nm)
65.1 ± 3.6
}~=
52.5 ± 2.2
}~= f=hJ
64.3 ± 2.4
The membranes’ effective pore radiuses were also estimated through methanol real retention at several pressures and permeate fluxes. A good correlation between experimental data and theoretical model was obtained (Figure S1 – Supplementary Material). Once the σ and P parameters were determined by the nonlinear fit of the Spiegler-Kedem model, the pore radius (rp) was determined from (Table 5). As shown in Equation 15, the rejection of a given solute increases as the permeate flux increases, reaching a limit that corresponds to the reflection coefficient. Because the contribution of the diffusive flux of the solute can be neglected if the permeate flux becomes infinite, the reflection coefficient is a characteristic of the convective transport of the solute. A reflection coefficient equals to 1 indicates that the convective transport of the solute is completely prevented or that no convective transport occurs. The last scenario, in which there is no convective transport, represents the ideal RO membranes in which no pores are available for convective flux. A reflection coefficient smaller than 1 is obtained if the solute is sufficiently small to allow penetration into the pores of the membrane [25]. Thus, the reduction of the reflection coefficient indicates an increase in the solute convective transport and, therefore, an enlargement of membrane pores.
Table 5 – Methanol reflection coefficient values (σ), solute permeability (P), and membrane pore radius (rp) 27
Parameter
σ
P (L·h-1·m-2·bar-1)
rp (nm)
Virgin NF90 }~= }~= f=hJ
0.80
9.91
0.17
0.80 0.72
9.65 32.51
0.17 0.19
It is evident that contact with the effluent caused an increase in }~= permeability after approximately 105 days. This increase could be caused by (1) salt-induced membrane swelling, since the continuous contact of the membrane with the high salinity effluent contact increases the number of counter-ions inside pores, and, as consequence, the electrostatic repulsion [30]; and (2) slightly increase in the hydrophilicity, which can be observed through a small reduction of the contact angle to the membrane ( Table 3). After this period, the permeability reduced gradually, mainly due to fouling development, demonstrated by SEM (Figure 10) and AFM (Figure 11) images, and hydrophobicity augmentation ( Table 3). It is clear that the foulant accumulated in the ridge-and-valley structure of the membranes, reducing the overall surface roughness (Table 4). This result is similar to those of Nanda et al. [31], who studied NF fouling through calcium sulfate. As the charge density of the membrane at the feed pH is slightly negative [4], calcium sulfate was allowed to approach the membrane surface more closely due to the reduced Donnan’s effect [31], causing the roughness modification observed. On the other hand, the }~= f=hJ presented a lower water permeability variation (maximum variation of 20% over the study period). This may have been due to the antagonistic effects of salt-induced swelling, pore size, hydrophobicity, and fouling. While an increase in the pore radius (Table 5) tends to lead to an increase in permeability, the elevation of the contact angle (Table 3), which indicates an increase in the hydrophobicity, and fouling (Figure 10 and Figure 11), reduce it. Therefore, the 28
increase in the pore radius was compensated by the increase in hydrophobicity and fouling in such a way that permeability remained approximately constant. The SEM and AFM images (Figure 10 and Figure 11) show that the execution of a monthly cleaning procedure reduced the amount of foulants on the membrane. This can also be inferred through the minor reduction of roughness of }~= f=hJ in regard to the virgin membrane (Table 4). However, the effects of the cleaning procedure were not sufficient to completely avoid the emergence of fouling, which may be due to the cleaning procedure used in this part of the study (soaking the membrane in the cleaning solution overnight, without recirculation). This is consistent with results shown in Section 3.2. There are many mechanisms affecting solute rejection through NF membranes, including steric hindrance, Donnan’s effect, and dielectric effects. In this sense, factors such as the membrane pore size, surface charge, and pore charge directly influence solute retention. Regarding the magnesium sulfate retention, a small reduction over time was observed for both samples (Figure 9a). This can be related to the salt-induced membrane swelling, fouling occurrence (Figure 10) and/or changes at the membrane superficial charge. The increase in pore size due to salt-induced swelling reduces steric hindrance effect and allowed more solutes to move towards permeate. In addition, the formation of foulant deposits reduces solute diffusion from the vicinity of the membrane surface back to the bulk solution, resulting in the so-called cake-enhanced concentration polarization effect. This effect causes an increase in the solute concentration at the membrane surface and thus the solute transferred to the permeate increases. Furthermore, it has been proposed that the binding or adsorption of counter-ions with fixed charges would alter the charge property of membrane, especially by multivalent 29
cations (e.g.Ca2+, Mg2+) [32]. Therefore, the continued exposure of the membranes to the effluent may have reduced its superficial charge and the repulsion of the co-ions as SO42-. This superficial charge reduction can also contribute to roughness modification, as previously discussed. The statistical test of Mann-Whitney was performed using Matlab R2008a software (The MathWorks, USA) in order to compare the retention of MgSO4 for both membranes after 105 days of exposure. It was concluded that the membrane performances could be considered significantly different to a significance level of 5% (%>hf5= = 0.0079), and that the retention of }~= f=hJ was the lowest. Thus, the exposure to HCl intensified the neutralization of the superficial charge and the swelling processes. Due to protonation of the surface’s functional groups of the membrane while in contact with the cleaning solution and increase in charge density, the membrane became more neutral and the pores, larger. As a result of hysteresis effects, a considerable increase in permeability following acid cleaning could be observed, particularly if the membrane had a very thin active skin layer [33]. Glucose retention also diminished with time (Figure 9b) and, as previously discussed, membrane swelling and fouling can explain this phenomenon. It can be also verified that the decrease in rejection after 105 days of exposure is larger for }~= f=hJ
(significant difference with %>hf5= = 0.0022). It is explained by the enlargement of membrane pores caused by the periodic cleanings with HCl solution (Table 5). It is worth noting that, although there have been changes in the rejection of membranes over time, the reduction was not very intense. NF membranes still showed good performance even after 285 days of exposure to effluent and to cleaning solution. Final retentions of MgSO4 and glucose were 95-88% and 88-82%, respectively. This is partly due to the use of a good pre-treatment. The pre-treatment included a pH adjustment of 30
the effluent, which in its original pH is very acidic and may cause polymer degradation, and the UF removal of suspended and colloidal solids, which can result in fouling and physical damage of the membranes. The lifetime of the membranes was estimated. For this purpose, it was assumed that the concentration of calcium in the permeate should be kept below 100 mg·L-1 to allow its reuse by the mining company [4], which corresponds to 84.5% removal efficiency. It was also considered that the removal efficiencies of calcium sulfate and magnesium sulfate were similar. The magnesium sulfate rejection data as a function of time was adjusted to a second-order polynomial model (Figure 9a) and the time to reach the minimum rejection was calculated. For the membrane exposed only to the effluent, the estimated lifetime was 536 days, while for the membrane submitted periodic cleaning, it was reduced to 389 days. Although this result was obtained for an exposure to the HCl solution higher than the optimized cleaning condition, it emphasizes the influence of the cleaning on the lifetime of the membranes, and enlighten its importance for the sustainability of the process. The effect of continuous exposure to the effluent and the combination effluent and cleaning agent on the chemical composition of the NF90 membrane was evaluated through attenuated total reflectance Fourier transform infrared spectroscopy (ATRFTIR). The spectrums are presented in Figure 12.
31
Neff+clean
Neff
Virgin
0,30
0,25
1239
1505
Absorbance
1471 1342
0,20
0,15
1612 1577
1172
846
1370
0,10
0,05
0,00 1800
1600
1400
1200
1000
800
-1
Wave number (cm )
Figure 12 – Infrared spectrums in the region between 1800 to 800 cm-1 to NF 90 membrane exposed only to the effluent and to the combination of effluent and cleaning agent In the infrared spectroscopy for attenuated total reflectance, the selective and intermediate layers are analyzed, and all results consist in an overlapping of spectrums from each layer [34]. Thus, bands of the selective layer, whose base polymer is polyamide, and also from the support layer, whose polymer is polysulfone, were verified on the spectrums obtained for NF90. The main detected bands, specific for the mentioned polymers, can be found in Table 6. Table 6 – Main IR bands of polysulfone (PSF) and polyamide (PA) components of NF90 membrane in the region between 1800 and 800 cm-1 Assignment Reported peaks Observed Vibration Intensity -1 a (cm ) peaks (cm-1) 1609 1612 Aromatic amide (N-H Weak PA deformation vibration or C=C ring stretching vibration) ~1587, 1577, Aromatic in-plane ring bend Middle 1504,1488 1505, 1471 stretching vibration ~1245 1239 C-O-C asymmetric stretching Strong PSF vibration of the aryl-O-aryl group 1385-1365 1370 C-H symmetric deformation Weak vibration of -C(CH3)2 32
a
1350-1280
1342
1180-1145
1172
~830
846
Asymmetric SO2 stretching vibration Symmetric SO2 stretching vibration In-phase out-of-plane hydrogen deformation of parasubstituted phenyl group
Middle Strong Weak
Peak values of PA and PSU were reported by Tang et al. [34]
Appearance or disappearance of the characteristic bands was not verified during the evaluation period, as it can be seen through a comparison of the spectrums presented in Figure 12. Therefore, the results obtained suggest that the degradation of the polymer polyamide did not occur, or it happened on a very small scale, which corroborate the results obtained in this study.
4. CONCLUSIONS With this study it was concluded that the NF process is a suitable treatment for gold mining effluent, allowing for a high conductivity and ions removal efficiency. However, the effects of concentration polarization and Hermia’s intermediate blocking fouling caused a reduction of approximately 80% of the permeate flux when the permeate recovery rate was 70%. SEM and EDS analysis showed that fouling basically is made up of calcium, magnesium, sodium, zinc, copper, aluminum, iron, arsenic, and silica. Calcium sulfate precipitated as bassanite over membrane surface since SI was larger than 1 even for low recovery rate. HCl solution was the most efficient cleaning solution. It should be applied over 90minute recirculation, without soak. NF membrane was exposed to the effluent and to the effluent and HCl solution for 285 days. Alterations in hydrophobicity, increase in pore size due to salt-induced swelling,
33
surface charge changes, and occurrence of fouling led to a decrease in the retentions of ionic (magnesium sulfate) and neutral solutes (glucose) by the two membranes. The retention of the membrane exposed to the cleaning solution was statistically lower, reinforcing the importance of optimizing the cleaning condition to maximize cleaning efficiency with the least exposure. Positively, the infrared spectroscopy results showed no indication of degradation of the polymeric material of the membrane. Although there were changes in the rejection of the membranes over time, the reduction was not very intense, and NF membranes still showed good performance even after 285 days of exposure to the effluent and the cleaning solution. Thus, it can be said that the results of this study are promising and demonstrate the robustness of NF technology for gold mining effluent treatment.
5. ACKNOWLEDGEMENTS The authors thank the Coordination of Improvement of Higher Education Personnel (CAPES) and the Foundation for Research Support of Minas Gerais (FAPEMIG) for the scholarships and financial resources provided. We also thank the microscopy center of the Federal University of Minas Gerais for AFM and SEM analysis; the Department of Metallurgical Engineering of the Federal University of Minas Gerais for contact angle analysis; and the Department of Chemistry of the Federal Center of Technological Education of Minas Gerais for infrared analysis.
6. REFERENCES [1] A. Mohammad, Y. Teow, W. Ang, Y. Chung, D. Oatley-Radcliffe, N. Hilal, Nanofiltration membranes review: Recent advances and future prospects, Desalination, 356 (2015) 226-254. 34
[2] B.C. Ricci, C.D. Ferreira, A.O. Aguiar, M.C. Amaral, Integration of nanofiltration and reverse osmosis for metal separation and sulfuric acid recovery from gold mining effluent, Separation and Purification Technology, (2015). [3] C. Sierra, J.R.Á. Saiz, J.L.R. Gallego, Nanofiltration of Acid Mine Drainage in an Abandoned Mercury Mining Area, Water, Air, & Soil Pollution, 224 (2013) 1-12. [4] L.H. Andrade, A.O. Aguiar, W.L. Pires, G.A. Miranda, L.P.T. Teixeira, G.C.C. Almeida, M.C.S. Amaral, Nanofiltration and Reverse Osmosis applied to gold mining effluent treatment and reuse, in, Brazilian Journal of Chemical Engineering, 2016. [5] A. Al-Amoudi, R.W. Lovitt, Fouling strategies and the cleaning system of NF membranes and factors affecting cleaning efficiency, Journal of Membrane Science, 303 (2007) 4-28. [6] T. Chidambaram, Y. Oren, M. Noel, Fouling of nanofiltration membranes by dyes during brine recovery from textile dye bath wastewater, Chemical Engineering Journal, 262 (2015) 156-168. [7] S. Madaeni, E. Salehi, Adsorption of cations on nanofiltration membrane: Separation mechanism, isotherm confirmation and thermodynamic analysis, Chemical Engineering Journal, 150 (2009) 114-121. [8] E.-m. Gwon, M.-j. Yu, H.-k. Oh, Y.-h. Ylee, Fouling characteristics of NF and RO operated for removal of dissolved matter from groundwater, Water Research, 37 (2003) 2989-2997. [9] A.I. Schäfer, A.G. Fane, T.D. Waite, Nanofiltration: principles and applications, Elsevier, 2005. [10] K. Chon, J. Cho, Fouling behavior of dissolved organic matter in nanofiltration membranes from a pilot-scale drinking water treatment plant: An autopsy study, Chemical Engineering Journal, 295 (2016) 268-277. [11] Z. Wang, J. Ma, C.Y. Tang, K. Kimura, Q. Wang, X. Han, Membrane cleaning in membrane bioreactors: a review, Journal of Membrane Science, 468 (2014) 276-307. [12] X. Wei, Z. Wang, F. Fan, J. Wang, S. Wang, Advanced treatment of a complex pharmaceutical wastewater by nanofiltration: Membrane foulant identification and cleaning, Desalination, 251 (2010) 167-175. [13] S. Zulaikha, W. Lau, A. Ismail, J. Jaafar, Treatment of restaurant wastewater using ultrafiltration and nanofiltration membranes, Journal of Water Process Engineering, (2014). [14] DOWFilmtec™, Product Information: FILMTEC NF90-400 Nanofiltration Element, in: Form n. 609-00345-0312. [15] A. Drak, K. Glucina, M. Busch, D. Hasson, J.-M. Laîne, R. Semiat, Laboratory technique for predicting the scaling propensity of RO feed waters, Desalination, 132 (2000) 233-242. [16] J. Sikder, S. Chakraborty, P. Pal, E. Drioli, C. Bhattacharjee, Purification of lactic acid from microfiltrate fermentation broth by cross-flow nanofiltration, Biochemical Engineering Journal, 69 (2012) 130-137. [17] R. Mukherjee, M. Mondal, A. Sinha, S. Sarkar, S. De, Application of nanofiltration membrane for treatment of chloride rich steel plant effluent, Journal of Environmental Chemical Engineering, 4 (2016) 1-9. [18] M.C.V. Vela, S.Á. Blanco, J.L. García, E.B. Rodríguez, Analysis of membrane pore blocking models adapted to crossflow ultrafiltration in the ultrafiltration of PEG, Chemical Engineering Journal, 149 (2009) 232-241. [19] S. Mattaraj, C. Jarusutthirak, R. Jiraratananon, A combined osmotic pressure and cake filtration model for crossflow nanofiltration of natural organic matter, Journal of Membrane Science, 322 (2008) 475-483. 35
[20] DOWFilmtec™, Cleaning and Disinfection Procedures for FilmTec™ NF200 and NF270 Elements, in: Form n. 609-00388-1006. [21] DOWFilmtec™, Dow FilmTec NF90 Nanofiltration Elements for Commercial Systems, in: Form n. 609-00378-0811. [22] M. González, I. Saucedo, R. Navarro, P. Prádanos, L. Palacio, F. Martínez, A. Martín, A. Hernández, Effect of phosphoric and hydrofluoric acid on the structure and permeation of a nanofiltration membrane, Journal of Membrane Science, 281 (2006) 177-185. [23] M. Dubois, K.A. Gilles, J.K. Hamilton, P. Rebers, F. Smith, Colorimetric method for determination of sugars and related substances, Analytical chemistry, 28 (1956) 350356. [24] X.-L. Wang, T. Tsuru, M. Togoh, S.-i. Nakao, S. Kimura, Evaluation of pore structure and electrical properties of nanofiltration membranes, Journal of chemical engineering of Japan, 28 (1995) 186-192. [25] J. Schaep, B. Van der Bruggen, C. Vandecasteele, D. Wilms, Influence of ion size and charge in nanofiltration, Separation and Purification Technology, 14 (1998) 155162. [26] C. Bellona, M. Marts, J.E. Drewes, The effect of organic membrane fouling on the properties and rejection characteristics of nanofiltration membranes, Separation and Purification Technology, 74 (2010) 44-54. [27] Y. Kaya, Z. Gönder, I. Vergili, H. Barlas, The effect of transmembrane pressure and pH on treatment of paper machine process waters by using a two-step nanofiltration process: Flux decline analysis, Desalination, 250 (2010) 150-157. [28] S. Madaeni, S. Samieirad, Chemical cleaning of reverse osmosis membrane fouled by wastewater, Desalination, 257 (2010) 80-86. [29] A. Al-Amoudi, P. Williams, S. Mandale, R.W. Lovitt, Cleaning results of new and fouled nanofiltration membrane characterized by zeta potential and permeability, Separation and Purification Technology, 54 (2007) 234-240. [30] J. Luo, Y. Wan, Effects of pH and salt on nanofiltration—a critical review, Journal of Membrane Science, 438 (2013) 18-28. [31] D. Nanda, K.-L. Tung, Y.-L. Li, N.-J. Lin, C.-J. Chuang, Effect of pH on membrane morphology, fouling potential, and filtration performance of nanofiltration membrane for water softening, Journal of Membrane Science, 349 (2010) 411-420. [32] C. Mazzoni, L. Bruni, S. Bandini, Nanofiltration: role of the electrolyte and pH on Desal DK performances, Industrial & engineering chemistry research, 46 (2007) 22542262. [33] A. Simon, J.A. McDonald, S.J. Khan, W.E. Price, L.D. Nghiem, Effects of caustic cleaning on pore size of nanofiltration membranes and their rejection of trace organic chemicals, Journal of Membrane Science, 447 (2013) 153-162. [34] C.Y. Tang, Y.-N. Kwon, J.O. Leckie, Effect of membrane chemistry and coating layer on physiochemical properties of thin film composite polyamide RO and NF membranes: II. Membrane physiochemical properties and their dependence on polyamide and coating layers, Desalination, 242 (2009) 168-182.
36
7. HIGHLIGHTS •
The best cleaning procedure proved to be 90-minutes recirculation with HCl solution
•
Membrane exposed to combined effluent and cleaning solution had lower retention
•
In 285 days of exposure, membrane hydrophobicity and effective pore radius changed
•
No indication of degradation of the polymeric material of the membrane was observed
•
Results demonstrate the robustness of NF for gold mining effluent treatment
37
S1385-8947(17)30660-5 http://dx.doi.org/10.1016/j.cej.2017.04.116 CEJ 16870
To appear in:
Chemical Engineering Journal
Received Date: Revised Date: Accepted Date:
17 February 2017 24 April 2017 25 April 2017
Please cite this article as: L.H. Andrade, B.C. Ricci, L.B. Grossi, W.L. Pires, A.O. Aguiar, M.C.S. Amaral, Nanofiltration applied in gold mining effluent treatment: evaluation of chemical cleaning and membrane stability, Chemical Engineering Journal (2017), doi: http://dx.doi.org/10.1016/j.cej.2017.04.116
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.
Nanofiltration applied in gold mining effluent treatment: evaluation of chemical cleaning and membrane stability L. H. ANDRADE a*, B. C. RICCI a, L. B. GROSSI a, W. L. PIRES a, A. O. AGUIAR a, M. C. S. AMARAL a a
Department of Sanitary and Environmental Engineering, Federal University of Minas
Gerais. Address: Av. Antônio Carlos, nº 6627, Pampulha, Belo Horizonte, Minas Gerais, Brazil. * Corresponding author: Telephone: +55 (31) 34093669; Fax: +55 (31) 34091879; Email: [email protected] ABSTRACT The objectives of this study were (1) to determine the best conditions of chemical cleaning of a nanofiltration (NF) membrane (NF90) employed in the treatment of gold mining effluent, and (2) to investigate the effects of continuous exposure to mining effluent and an acid cleaning agent on NF characteristics. NF fouling was mainly due to inorganic species, and calcium sulfate (CaSO4.0.5H2O) was identified over membrane surface. Therefore, acid solutions were the most efficient in membrane cleaning, and among them hydrochloric acid (HCl) showed the best performance. Cleaning without recirculation (soak) was not effective and thus 90-minute recirculation was chosen as the best cleaning procedure. The NF membrane was exposed to the effluent and to the combined effluent and HCl solution for 285 days. Modifications in hydrophobicity, effective pore radius, fouling, and surface charge were observed. The retention of magnesium sulfate and glucose by the membrane exposed to the combined effluent and cleaning solution was statistically lower, what endorses the importance of cleaning conditions optimization. No indication of degradation of the polymeric material of the
1
membrane was observed. Thus, the results of this study are promising and demonstrate the robustness of NF technology for gold mining effluent treatment. KEYWORDS Gold mining effluent; Nanofiltration; Membrane lifetime; Membrane cleaning
1. INTRODUCTION Nanofiltration (NF) is a separation process that employs membranes with a typical pore size of 1 nm, which corresponds to a molecular weight cut-off (MWCO) of 300–500 Da. NF presents low rejection of monovalent ions, high rejection of multivalent ions, and higher flux compared to reverse osmosis (RO) membranes. These properties have allowed NF to be used in many niche applications, especially in the treatment of water and wastewater [1]. In this way, several authors have studied NF for the treatment of mining effluent and have obtained promising results regarding pollutant efficiency retention [2, 3]. In addition to the cited studies, NF has proved to be a favorable treatment for the effluent from the gold mining industry, specifically, from a sulfuric acid production plant and a calcined dam. As reported by Andrade et al. [4], the NF of the combined streams of these effluents allowed for generation of high quality treated effluent. Despite the excellent performance, membrane fouling management is essential to ensure consistent separation performance with minimal cleaning and membrane replacement. Fouling is a major limitation of NF and is related to operational conditions and to feed stream composition. Common fouling mechanisms include the following: pore plugging and external pore blocking, resulting from deposition of particles and colloids; precipitation of dissolved materials on membrane pores and surfaces; build-up of a 2
cake/gel-like layer on the upstream face of a membrane; absorption or adsorption; and molecular-membrane electrostatic interactions and acid–base interactions [5-8]. The main consequences of fouling include flux decline, permeate quality deterioration, and an energy consumption increase. Accordingly, fouling control is essential for increasing membrane operational lifetime and thus reducing process costs. Feed pre-treatment, membrane selection, module design, operation mode, and membrane cleaning are the main membrane fouling control strategies [9, 10]. Membrane cleaning methods are classified into physical, chemical, physico-chemical and biochemical [11]. Chemical cleaning is the most common method, especially in NF and RO systems. It is used to remove fouling by means of specific chemical agents, which typically include acids, bases, chelating agents, and/or surfactants [12] The optimal selection of the cleaning agent depends mainly on the membrane material and type of foulant. The ideal cleaning agent should remove the deposited material without damaging the membrane, but often prolonged exposure can result in morphological, structural, and superficial changes, reducing the ability of the membrane to reject solutes. This leads to a reduction of membrane module lifetime, which is related to the characteristics of the effluent, fouling rate, operating conditions, and hydraulic cleaning cycle [8]. In this context, the objectives of this study were (1) to determine the optimal conditions of chemical cleaning of an NF membrane (NF90) employed in the treatment of gold mining effluent, and (2) to investigate the effects of continuous exposure to mining effluent and a cleaning agent on NF membrane properties.
3
2. MATERIALS AND METHODS 2.1 Gold ore processing effluents Two effluents from a gold mining company in Brazil were studied, i.e., the effluent from the sulfuric acid production plant and the water from the calcined dam (Table 1). The industrial process description and the effluent generation points are as described in Andrade et al. [4]. Preliminary NF experiments have shown that membrane process performance was improved when applied to these two effluents in combination instead of to a single effluent. Thus, both effluents were mixed at a 1:1 ratio (corresponding to actual wastewater flow rates generated by the company, which are 140 m³·h-1 of effluent from the sulfuric acid production plant and 140 m³·h-1 of water from the calcined dam) before being transferred to the membrane treatment system. In this study, such a mixture will hereinafter be referred to as “effluent from gold mining”. Table 1 – Characteristics of gold ore processing effluent Parameters pH Conductivity (µS·cm-1) Dissolved organic carbon (mg·L-1) Dissolved inorganic carbono (mg·L-1) Total solids (mg·L-1) Suspended solids (mg·L-1) Sulfate (mg·L-1) Chloride (mg·L-1) Aluminum (mg·L-1) Arsenic (mg· L-1) Calcium (mg·L-1) Iron (mg·L-1) Magnesium (mg·L-1) Manganese (mg·L-1) Potassium (mg·L-1)
Effluent from the sulfuric acid production plant 1.82 9,420
Water from the calcined dam
Effluent from gold mining1
8.23 3,250
2.08 6,615
4.4
5.8
5.0
0.2
19.2
9.7
9,398 25 6,367 31 121 657 382 75 462 28 49
3,159 14 1,944 291 0.1 2 541 0.2 29 0.1 52
8,804 15 3,864 168 67 322 550 34 247 22 44 4
Sodium (mg·L-1) 45 168 77 -1 Zinc (mg·L ) 115 0.1 64 1 Mixture at a 1:1 ratio of effluent from the sulfuric acid production plant and of the water from the calcined dam 2.2 Experimental setup The treatment process proposed and evaluated in the bench-scale for the effluent consisted of a pH adjustment to 5.0 with a sodium hydroxide (NaOH) solution, ultrafiltration, and nanofiltration. Ultrafiltration (UF) was performed with a commercial submerged membrane module (ZeeWeed) with a filtration area of 0.047 m², PVDF-based polymer, and average pore diameter of 0.04 µm. A bench-scale unit was used; it comprised one pump equipped with a speed controller, one needle-type valve for pressure adjustment, and one manometer. NF treatment was performed with a Dow FilmTec NF90 membrane, which is a thin film composite membrane comprised of three layers: (1) a polyester support web, (2) a microporous polysulfone interlayer, and (3) an ultrathin aromatic polyamide active layer. The MWCO is 100 Da [13], the average membrane hydraulic resistance is 5.8 × 10 13 m-1, and the sodium chloride (NaCl) (2,000 mg·L-1) and MgSO4 (2,000 mg·L-1) rejection are 85-95% and 97%, respectively [14]. The maximum allowed NF operating pressure of the setup was 15 bar, which was provided by a rotary vane pump (Procon) with a maximum flow of 530 L·h-1. A needle valve was used to adjust the feed flow rate and the trans-membrane pressure. The stainless steel cell had a 9.8 cm diameter and an effective filtration area of 75 cm². The schematic view of the UF/NF setup is shown in Figure 1.
5
(a)
(b) Figure 1 – Schematic of (a) ultrafiltration and (b) nanofiltration bench-scale unit
2.3 Fouling evaluation 2.3.1 Experimental procedure To investigate NF fouling, a virgin membrane was inserted into the NF cell and compacted with distilled water at 10 bar. After flux stabilization, water permeability was measured by monitoring the stabilized water permeate flux at pressures of 10.0, 7.5, 5.0, and 2.5 bar using a measuring cylinder and a chronometer. Water temperature was also monitored and permeate flux was normalized to 25°C by means of a correction factor calculated as water viscosity at the temperature of permeation divided by water viscosity at 25°C [15]. 6
Water flux and permeability standard deviations were calculated according to the equation for determining the combined standard uncertainty (Equation 1).
= ∑ × ²
(Equation 1)
where is the combined standard deviation of the variable y, f is the function
y=f(x1, x2, ..., xN), and ² is the uncertainty related to the parameter xi. The water flux (J) and permeability (Lp) can be calculated using Equations 2 and 3.
= × × ² ×
(Equation 2)
!"º#
&
$% = '
(Equation 3)
where V is the volume of permeate (L) collected in a given time t (h), D is the diameter of the filtration cell (m), P is the pressure (bar), and () is the permeate viscosity at temperature T. The permeate viscosity was assumed to be equal to water viscosity and thus was calculated at a given temperature T (K) by Equation 4. () = 2.414 × 10/0 × 10 1.2⁄)/ 3
(Equation 4)
The combined uncertainty of J and Lp can thus be calculated by: = $% ×
5! !
$% = $% ×
+
5! !
5!
+
!
+
5! !
5!
+
!
+
5! !
5! ) 1.2!
+
)/ 3 7
5! ' '!
+
5! ) 1.2! )/ 3 7
(Equation 5)
(Equation 6)
Next, distilled water was substituted by 8.0 L of pre-treated effluent (after pH adjustment and ultrafiltration, as detailed in Andrade et al. [4]). NF was performed with a constant operational pressure of 10 bar and a feed flow rate of 144 L·h-1, which resulted in 1.9 m·s-1 cross-flow velocity and a Reynolds number of 847. The cross-flow velocity was chosen based on the literature (e.g. Sikder et al. [16] used cross-flow 7
velocities varying from 1.77 to 2.48 and Mukherjee et al.[17], Reynolds numbers from 800 to 1200). The system was operated with a continuous removal of permeate and recycling of concentrate back to the feed tank, reaching a permeate recovery rate (RR) of 71% after 2,100 minute-filtration. To evaluate salt precipitation on membrane surfaces, which results in scaling, the calcium sulfate supersaturation index at membrane surface (89: ) was determined for different RRs according to Ricci et al. [2]. Calcium sulfate was chosen as representative of salt precipitation as it has a low solubility and a high concentration in effluent from gold mining (Table 1) presenting, therefore, a greater precipitation potential. Hermia’s semi-empirical models modified for crossflow filtration [18] were used to identify the fouling mechanism involved during the NF process. Also, reversible and irreversible fouling resistances were also determined (Section 2.3.2). Feed and NF permeate samples were collected and analyzed according to the parameters conductivity (Hanna conductivity meter HI 9835) and ions sulfate, chloride, calcium, and magnesium (ion chromatograph Dionex ICS-1000 equipped with AS-22 and ICS 12a columns). The entire procedure was performed in duplicate and results presented in this paper correspond to averages. Following the procedure, two fragments of fouled membranes were obtained (from each duplicate). The first fragment was subjected to physical cleaning using distilled water recirculation, at 144 L·h-1 for 30 minutes. After physical cleaning, the water permeability was measured. This result was used to calculate the resistance to filtration (Section 2.3.2). The second membrane fragment was removed from the system. The precipitate formed on the membrane surface was gently removed with a spatula and analyzed by X-ray
8
diffraction (XRD) (Section 2.6.5). Samples of virgin and fouled membranes were also analyzed in a scanning electron microscope (SEM) coupled with an energy dispersive X-ray spectrometer (EDS) (Section 2.6.1).
2.3.2 Resistances in series calculations Resistances were calculated using the resistance in series concept. Total resistance (; ) comprises three components (Equation 7): intrinsic membrane resistance (;: );
reversible fouling resistance (; <=> , which comprised the concentration polarization resistance and the resistance from fouling removable by physical cleaning; and irreversible fouling resistance (; <<=> ), which cannot be removed by physical cleaning. ; = ;: + ; <=> + ; <<=>
(Equation 7)
Using the virgin membrane water permeability, the intrinsic membrane resistance to filtration (;: ) was calculated according to Equation 8: R@ =
ABCDECF ×G
(Equation 8)
where H>
L/MNO
G × PQRSTUQFV
(Equation 9)
where P is the applied transmembrane pressure (106 Pa), W is the reflection coefficient estimated by the averaged membrane rejection of the major constituents of effluent [19], and XY is the difference in osmotic pressure of the solution at membrane surface and 9
permeate streams. Thus, Z − WΔπ is the effective pressure, in Pa. In addition, Jeffluent is the permeate flux in m3·h-1·m-2 obtained at the end of the fouling experiment (after 2,100 minutes of filtration). The difference in osmotic pressure obtained with different permeate recovery rates were estimated by the Van’t Hoff equation, as described in Equation 10: Δπ = ;^ΣΔC
(Equation 10)
where R is the universal gas constant (L·Pa·K-1·mol-1), T is the permeation temperature (K), and ΣX` is the sum of the difference in molar concentrations (mol·L-1) of dissolved species present at membrane surface and the permeate. Only the most concentrated ions (sulfate, calcium, magnesium and arsenic) were taken into consideration to ΣX` determination. The resistance of the irreversible fouling can be calculated by the permeability with water after the physical cleaning process (H<<=> ) (Equation 11). R <<=> =
Aaabc ×G
− ;:
(Equation 11)
The reversible fouling resistance can be calculated by Equation 12. R <=> = R K − R @ − ; <<=>
(Equation 12)
2.4 Evaluation of NF membrane cleaning Cleaning tests were performed with different chemical agents and, after selection of the most appropriate, tests with different soak and recirculation times were carried out. The cleaning test consisted of six steps: (1) compaction of the virgin membrane at 10 bar 10
with distilled water; (2) measurement of distilled water permeability ($%J=d ; (3) performance of NF fouling procedure with effluent; (4) measurement of fouled membrane distilled water permeability ($% e5f=g ); (5) accomplishment of the cleaning procedure; and (6) measurement of the cleaned membrane distilled water permeability $%f=hJ=g . The fouling procedure consisted of feeding 3 L of pre-treated effluent into the NF unit, and performing NF at a constant pressure of 10 bar and a feed flow rate of 144 L·h-1 for 4 hours in concentration mode. For tests with different cleaning solutions, the cleaning procedure consisted of recirculating the solution in the NF unit for 90 minutes with a flow rate of 144 L·h-1 and a pressure of 1.5 bar. Different cleaning solutions were tested, specifically: 0.2% hydrochloric acid, 2.0% citric acid, 0.5% phosphoric acid, 0.2% nitric acid, 0.1% sulfuric acid, 0.4% sodium hydroxide, 0.03% dodecylbenzene sodium sulfate (DDBS), and 1.0% sodium ethylenediamine tetraacetic acid (EDTA) (similar to the solutions suggested by the supplier of the membranes [20]). A blank test was also conducted using only water. After that, the cleaning mode (only soaking or soaking associated with recirculation) and time were assessed for the best chemical cleaning agent. The procedure consisted of adding the cleaning solution in the NF unit, leaving it in static contact with the fouled membrane for different soak times (0 up to 60 minutes), then starting the pump and recirculating the solution for 15 up to 120 minutes. The cleaning efficiency in each test was calculated using Equation 13: `ijklmln joompjlp % =
rstubvwbx /rsyz{ubx rswb| /rsyz{ubx
(Equation 13)
11
2.5 Assessment of membrane stability to the effluent and the cleaning solution 2.5.1 NF membrane immersion procedure The effects of continuous exposure to effluent and chemical cleaning solutions on the characteristics of the NF membrane were evaluated by immersing four pristine membrane fragments in NF effluent concentrate (obtained using the conditions established in Andrade et al. [4]) for a period of 285 days. At four-week intervals, fragments were removed from the solution and rinsed with distilled water. Two of these fragments were immersed in a cleaning solution of 1.0% HCl overnight (16 hours), and the other two remained immersed in distilled water. The concentration of the cleaning solution and the cleaning time were higher than the ones used in Section 2.4 (given an exposition to cleaning agent approximately 50 times higher) so that the effect of the acid on the membrane could be more readily observed. Subsequently, filtration tests (Section 2.5.2) and analysis of the morphological and chemical characteristics (SEM, EDS, contact angle, atomic force microscopy, and attenuated total reflection Fourier transform infrared spectroscopy - Section 2.6) were carried out on two of the fragments, one that was subjected to contact with the effluent and cleaning solution (}~= f=hJ ) and one that was exposed only to effluent (}~= ). At the end of the exposure period, tests were also conducted to estimate the effective pore radius of the membrane (Section 2.5.3).
2.5.2 Filtration tests The filtration tests consisted of three steps: (1) hydraulic permeability determination; (2) evaluation of magnesium sulfate rejection; and (3) evaluation of glucose rejection. Magnesium sulfate retention tests were performed using a 2,000 ppm solution as the feed. Two liters of solution were fed into the NF unit and nanofiltrated at a pressure of 5 12
bar and temperature of 25±3°C, until a permeate recovery rate of 15% was achieved [21]. Feed and permeate were collected and analyzed for conductivity. A calibration curve was constructed and the magnesium sulfate removal efficiency was calculated. Glucose retention tests were performed in similar form, using a 500 ppm solution, pressure of 10 bar and final permeate recovery rate of 10% [22]. Feed and permeate carbohydrates were analyzed and glucose retention was calculated [23].
2.5.3 Estimation of the effective pore radius of the NF membrane Methanol rejections were determined as a function of permeate flux for pressures of 412 bar for virgin membrane and membranes after 285 days of exposure. A 2,000 ppm solution was used; since the concentration was not high, membrane swelling was not an issue. The rejection data were interpreted in terms of real rejection as shown in Equation 14. ;< = 1 −
(Equation 14)
Here, Rri and Cmi are the real rejection and concentration of solute i at the membrane surface, respectively. Results of methanol rejections as a function of pressure were analyzed using the Spiegler-Kedem and Steric Hindrance Pore (SHP) models [24]. The Spiegler-Kedem model equation is given by Equation 15 [25]. ;< =
σ/ /σ
ℎjj ~ = exp −
/σ '
(Equation 15)
where J is the permeate flux, P is the permeability of solute i, and σ is the reflection coefficient.
13
For neutral solutes where the convective flux is not influenced by electrostatic effects, separation occurs according to the size exclusion mechanism. Thus, only the ratio of the radius of the solute ( , considered equal to the Stokes radius, which corresponds to 0.135 nm for methanol [26]) to the pore radius (s ) of the membrane determines the reflection coefficient, as indicted in Equation 16, which is deduced from the SHP model [25]: < !
<
<
σ = 1 − 1 + < ! 1 − < 2 − 1 − <
(Equation 16)
As can be seen from Equation 16, the SHP model can be used to estimate the pore radius for a given membrane. For this purpose, the reflection coefficient of a given solute must first be determined by Equation 15 and then the pore radius must be calculated using Equation 16.
2.6 Evaluation of morphological and chemical characteristics of membranes 2.6.1 Scanning electron microscopy (SEM) and energy dispersive X-ray spectrometer (EDS) Membrane surface morphology was analyzed using a FEI Quanta 200 SEM with an additional EDS analysis. Prior to analysis, membrane samples were air-dried and coated with 5 nm carbon layer by a sputter coating machine (Leica EM SCD 500 with a pressure of 10 -2-10-3 mbar and a 2.5 A current).
2.6.2 Atomic force microscopy (AFM) Tapping mode AFM was performed with an Asylum MFP-3D-SA/AFM microscope (Asylum Research) over an area of 5 × 5 µm². A silicon probe (AC240TS-R3, Olympus) was used. At least three different areas of each membrane were visualized 14
and the mean value of the root-mean-squared roughness (RRMS) was determined using the AFM software.
2.6.3 Contact angle measurements Water contact angles were determined using a DIGIDROP-DI goniometer (GBX Instruments) equipped with a CCD camera and an automated liquid dispenser, using the standard sessile drop method. Six µL of deionized water droplets were placed at the membrane surface at room temperature and the contact angle was measured. Five droplets were applied to each membrane to establish an average.
2.6.4 Attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy analysis ATR-FTIR experiments were carried out using a Shimadzu FTIR IR Prestige-21 instrument equipped with an attenuated total reflectance (ATR) accessory. The spectrum was obtained in the range of 400–4000 cm−1 at a 4 cm−1 resolution.
2.6.5 X-ray diffraction (XRD) Precipitate was analyzed by a Philips diffractometer (Panalytical) with an X’Pert-APD system, PW 3710/31 controller, PW 1830/40 generator, and PW 3020/00 goniometer was used. The radiation was emitted from a copper tube (Z = 29) with a mean wavelength Kα of 1.54184Å and λα1 of 1.54056Å.
3. RESULTS AND DISCUSSION 3.1 Fouling evaluation 15
Figure 2 illustrates the permeate flux as a function of operation time for the NF of the gold mining effluent. It is evident that the flux decreased strongly (~80%) until approximately 750 minutes of NF, followed by a tendency toward stabilization. This flux decline is justified partially by the high ions rejection by the membrane (Table 2), which causes increase in the concentration of ions at membrane surface, rising osmotic pressure and reducing effective pressure (Figure 2).
Effective pressure
70
10 9 8 7 6 5 4 3 2 1 0
Fux (L/h.m²)
60 50 40 30 20 10 0 0
500
1000
1500
Effective pressure (bar)
Flux
2000
Time (min)
Figure 2 – Permeate flux (J) and effective pressure as a funtion of operating time for NF fouling evaluation during gold mining effluent filtration Table 2 – Characterization of gold mining effluent and NF permeate obtained during NF fouling evaluation Parameter NF feed1 NF permeate Removal efficiency -1 Conductivity (µS.cm ) 4,550 398 91% Sulfate (mg.L-1) 2,620 168 94% Chloride (mg.L-1) 196 15 92% -1 Calcium (mg.L ) 673 24 96% Magnesium (mg.L-1) 148 6 96% 1
Pre-treated effluent (after pH adjustment and ultrafiltration, as detailed in Andrade et al. [4]) In addition, flux decline was also caused by fouling which was mainly associated with inorganic matter, due to effluent characteristics. The effluent is saturated with calcium 16
sulfate, with a supersaturation index (SI) over membrane surface ranging from 1.68 to 2.94, for a permeate recovery rate between 0 and 70%, respectively, allowing the salt to precipitate. An X-ray diffraction from the precipitate deposited over membrane surface indicated the presence of bassanite (hydrated calcium sulfate CaSO4.0.5H2O) as the only crystalline phase (Figure 3).
Figure 3 – X- ray diffractogram from the precipitate formed over the membrane during the fouling test
Through SEM micrographs of the fouled surface in association with EDS analysis (Figure 4), it was observed that, besides crystalline calcium sulfate, the fouling composition included amorphous structures composed of magnesium, sodium, zinc, copper, aluminum, iron, arsenic, and silica.
17
(a)
(b) Figure 4 – SEM micrographs and EDS spectrum from (a) the virgin membrane and (b) after 2,100 minutes of filtration 18
According to the modified Hermia’s model, there are four main fouling mechanisms, i.e., complete blocking, intermediate blocking, standard blocking, and cake formation . Complete blocking occurs when the size of foulants is bigger than the membrane pore size, therefore, pore blocking takes place over the membrane surface and not inside the membrane pores. It is also assumed that a molecule never settles over another one that had been previously deposited. Intermediate blocking is valid when molecules can obstruct the pore entrance without blocking it completely. It also considers the possibility of some molecules to settle over others. Standard blocking assumes that particles approaching the membrane are adsorbed or deposited on the internal pore walls, thereby reducing the pore volume. This type of fouling is caused by molecules smaller than the membrane pores. At last, cake formation is characterized by foulants that do not enter the membrane pores. Instead, they deposit over membrane surface, which results in cake or gel layer formation over the membrane surface [18]. According to Kaya et al. [27], the construction of curves of ln (J-1), J-1/2, J-1 and J-2 (where J is the permeate flux) as a function of time and evaluation of the best model fitness (R²) allows the identification of which of the four fouling mechanisms (complete blocking, standard blocking, intermediate blocking, and cake formation, respectively) is the most important. Thus, the experimental data was tested against the four mechanisms (Figure 5).
19
(a)
(b)
(c) (d) Figure 5 – Comparison of experimental data with the Hermia’s model: (a) complete blocking, (b) standard blocking, (c) intermediate blocking, and (d) cake formation.
It was observed that the predominant mechanism was intermediate blocking. This means that the precipitated particles are capable of depositing on the membrane surface or even on top of each other, forming multilayers. This hypothesis is also corroborated by the SEM image (Figure 4b). Intermediate blocking occurs when the solute molecule is similar to the membrane pore size, thus some molecules can obstruct a membrane pore entrance without blocking the pore completely [18]. Therefore, the initial CaSO4 crystalline salts formed must had a nanometric size. The membrane, reversible fouling (physically removed by cleaning with water recirculation), and irreversible fouling (not removed by physical cleaning) resistances were calculated at the end of the fouling test (Table S1 – Supplementary Material). It was observed that the highest resistance to NF was due to the membrane itself (5.85 x 10 13 m-1, corresponding to 39% of total resistance), which is a typical result of using NF 20
membranes [17]. Part of the fouling resistance can be removed by physical methods (26%), and thus can be controlled by adjusting the feed flow conditions. However, a significant portion of the fouling is considered irreversible (35%), thus requiring chemical cleaning. The results show the importance of optimizing the chemical cleaning process to ensure the maintenance of membrane permeability and the sustainability of the process. 3.2 Cleaning membrane evaluation In order to remove fouling and restore membrane permeability to allow for an increase in the membrane module’s lifetime and the process sustainability, several chemical substances were evaluated as cleaning agents (Figure 6). The purpose of a cleaning agent is to reduce the foulant-foulant and foulant-membrane interaction in such a way that the fouling layer can be removed from the membrane’s surface through mass transfer [12]. Acid and chelating agents reduce the inorganic foulant-membrane interaction, while bases act mainly on the organic deposits and amino acids, by hydrolysis and solubilization [12, 28], and on silica [8].
Cleaning efficiency
100% 80% 60% 40% 20% 0% HCl 0.2%
Citric H3PO4 acid 2% 0.5%
HNO3 0.2%
H2SO4 0.1%
NaOH 0.4%
DDBS 0.03%
EDTA 1.0%
Water
Figure 6 – Cleaning efficiency of several chemical agents Due to the low organic compound concentration found in the effluent (Table 1), the NF membrane fouling was caused mainly by inorganic species, as observed in Figure 3 and 21
Figure 4. Thus, the NaOH solution presented the lowest cleaning efficiency among all chemical agents evaluated. EDTA is a strong metal-chelating agent. It can react with calcium ions in sulfate and carbonate of calcium to form soluble complexes. It can also react through ligandexchange with calcium ions in organic matter-calcium complexes [12]. However, in this study, EDTA did not demonstrate a good efficiency. Al-Amoudi et al. [29] also observed that permeability was not totally restored after cleaning by a mixed cleaning agent that contained EDTA. It was justified by a permanent alteration of membrane superficial charge [29], which may increase its interaction with foulant material. Among the acids, the worst performance was observed for sulfuric acid (H2SO4), which can be associated with the tendency of formation of sulfate salt precipitate when the solution comes in contact with the calcium presented on membrane surfaces. Phosphoric acid can also provide insoluble calcium phosphate formation [5]. Nitric acid did not present satisfactory results, which is consistent with previous results in other studies [28]. Despite some suggestions that it is the most efficient cleaning agent [12], citric acid had a less relevant effect in this study, probably because at this solution pH (~3.0) calcium sulfate solubility is only 7% higher than at the pH 5.0 of effluent. The best cleaning agent in this study was hydrochloric acid. The advantage of HCl over other acids such as HNO3 and H2SO4 is that HCl has no oxidation ability for degradation of organic matter. This degradation causes formation of secondary fouling. Moreover, the obtained salt from the HCl agent is more soluble compared to the salts from other acids [28]. Therefore, HCl was selected as the most suitable chemical for NF membrane cleaning procedures.
22
The Figure 7 shows the cleaning efficiencies obtained with 0.2% HCl solution over different cleaning times. The total cleaning time includes the soak time (ranging from 0 up to 60 minutes) plus the recirculation time. Bars refer to standard deviation.
Figure 7 – Cleaning efficiencies obtained with 0.2% HCl solution over different cleaning times. The total cleaning time includes soak and recirculation times.
According to Wei et al. [12] during cleaning process, the cleaning agent contacts membrane surface foulants via mass transfer. The agent reacts with the foulants especially those that closest to membrane surface, yielding loosened fouling layer. Then, these reaction products are removed from membrane surface via mass transfer. As hydrodynamic conditions interfere directly in the mass transfer coefficients, cleaning without fluid movement (soak) is not very effective. Thus, it can be seen that the longer the soak time, the greater the total cleaning time to achieve cleaning efficiency > 90%. The condition that required the lower total cleaning time to remove fouling was the one
23
with only recirculation cleaning (soak time = 0 min), in which it was possible to obtain 94% efficiency after 90 minutes.
3.3 Evaluation of membrane stability to the effluent and the cleaning solution Figure 8 shows the variation in water permeability of membranes exposed to the effluent (}~= ) and to the effluent and the cleaning solution (}~= f=hJ ) as a function of time.
Water permeability (L/h.m².bar)
NF eff
NF eff+clean
10 8 6 4 2 0 0
50
100
150
200
250
300
Exposure time (days)
Figure 8 – Water permeability variation as a funtion of time of exposure to the membrane in contact with effluent (}~= ) and in contact with effluent and cleaning solution (}~= f=hJ )
Retention of magnesium sulfate (ionic solute) and glucose (neutral solute) over the period were also monitored (Figure 9).
24
(a)
(b) Figure 9 – Retention of (a) magnesium sulfate and (b) glucose as a funtion of the exposure time for the membrane in contact with effluent (}~= ) and in contact with effluent and cleaning solution (}~= f=hJ ) In order to evaluate the changes observed in the permeability and the magnesium sulfate and glucose retentions, contact angle ( Table 3), SEM (Figure 10), and AFM (Figure 11) with root mean square (RMS) roughness calculations (Table 4) were carried out.
Table 3 – Contact angle measurments of the virgin membrane and the membrane exposed to the effluent and to the effluent and the cleaning solution }~= }~= f=hJ Virgin Sample NF90 105 days 285 days 105 days 285 days 25
Contact Angle (º)
(a)
Value
48.0 ± 0.7
47.0 ± 0.5
(b)
55.0 ± 0.8
49.5 ± 1.5
57.3 ± 1.0
(c)
Figure 10 – SEM images of the NF90 membranes: (a) virgin, (b) exposed to the effluent after 285 days, and (c) exposed to the effluent and the cleaning solution after 285 days
(a)
(b)
(c) 26
Figure 11 – AFM images of the NF90 membranes: (a) virgin, (b) exposed to the effluent after 285 days, and (c) exposed to the effluent and the cleaning solution after 285 days
Table 4 – Root mean square roughness (RMS) measurements of the NF90 membranes: virgin, exposed to the effluent, and exposed to the effluent and the cleaning solution for 285 days Sample
NF90 virgin
RMS (nm)
65.1 ± 3.6
}~=
52.5 ± 2.2
}~= f=hJ
64.3 ± 2.4
The membranes’ effective pore radiuses were also estimated through methanol real retention at several pressures and permeate fluxes. A good correlation between experimental data and theoretical model was obtained (Figure S1 – Supplementary Material). Once the σ and P parameters were determined by the nonlinear fit of the Spiegler-Kedem model, the pore radius (rp) was determined from (Table 5). As shown in Equation 15, the rejection of a given solute increases as the permeate flux increases, reaching a limit that corresponds to the reflection coefficient. Because the contribution of the diffusive flux of the solute can be neglected if the permeate flux becomes infinite, the reflection coefficient is a characteristic of the convective transport of the solute. A reflection coefficient equals to 1 indicates that the convective transport of the solute is completely prevented or that no convective transport occurs. The last scenario, in which there is no convective transport, represents the ideal RO membranes in which no pores are available for convective flux. A reflection coefficient smaller than 1 is obtained if the solute is sufficiently small to allow penetration into the pores of the membrane [25]. Thus, the reduction of the reflection coefficient indicates an increase in the solute convective transport and, therefore, an enlargement of membrane pores.
Table 5 – Methanol reflection coefficient values (σ), solute permeability (P), and membrane pore radius (rp) 27
Parameter
σ
P (L·h-1·m-2·bar-1)
rp (nm)
Virgin NF90 }~= }~= f=hJ
0.80
9.91
0.17
0.80 0.72
9.65 32.51
0.17 0.19
It is evident that contact with the effluent caused an increase in }~= permeability after approximately 105 days. This increase could be caused by (1) salt-induced membrane swelling, since the continuous contact of the membrane with the high salinity effluent contact increases the number of counter-ions inside pores, and, as consequence, the electrostatic repulsion [30]; and (2) slightly increase in the hydrophilicity, which can be observed through a small reduction of the contact angle to the membrane ( Table 3). After this period, the permeability reduced gradually, mainly due to fouling development, demonstrated by SEM (Figure 10) and AFM (Figure 11) images, and hydrophobicity augmentation ( Table 3). It is clear that the foulant accumulated in the ridge-and-valley structure of the membranes, reducing the overall surface roughness (Table 4). This result is similar to those of Nanda et al. [31], who studied NF fouling through calcium sulfate. As the charge density of the membrane at the feed pH is slightly negative [4], calcium sulfate was allowed to approach the membrane surface more closely due to the reduced Donnan’s effect [31], causing the roughness modification observed. On the other hand, the }~= f=hJ presented a lower water permeability variation (maximum variation of 20% over the study period). This may have been due to the antagonistic effects of salt-induced swelling, pore size, hydrophobicity, and fouling. While an increase in the pore radius (Table 5) tends to lead to an increase in permeability, the elevation of the contact angle (Table 3), which indicates an increase in the hydrophobicity, and fouling (Figure 10 and Figure 11), reduce it. Therefore, the 28
increase in the pore radius was compensated by the increase in hydrophobicity and fouling in such a way that permeability remained approximately constant. The SEM and AFM images (Figure 10 and Figure 11) show that the execution of a monthly cleaning procedure reduced the amount of foulants on the membrane. This can also be inferred through the minor reduction of roughness of }~= f=hJ in regard to the virgin membrane (Table 4). However, the effects of the cleaning procedure were not sufficient to completely avoid the emergence of fouling, which may be due to the cleaning procedure used in this part of the study (soaking the membrane in the cleaning solution overnight, without recirculation). This is consistent with results shown in Section 3.2. There are many mechanisms affecting solute rejection through NF membranes, including steric hindrance, Donnan’s effect, and dielectric effects. In this sense, factors such as the membrane pore size, surface charge, and pore charge directly influence solute retention. Regarding the magnesium sulfate retention, a small reduction over time was observed for both samples (Figure 9a). This can be related to the salt-induced membrane swelling, fouling occurrence (Figure 10) and/or changes at the membrane superficial charge. The increase in pore size due to salt-induced swelling reduces steric hindrance effect and allowed more solutes to move towards permeate. In addition, the formation of foulant deposits reduces solute diffusion from the vicinity of the membrane surface back to the bulk solution, resulting in the so-called cake-enhanced concentration polarization effect. This effect causes an increase in the solute concentration at the membrane surface and thus the solute transferred to the permeate increases. Furthermore, it has been proposed that the binding or adsorption of counter-ions with fixed charges would alter the charge property of membrane, especially by multivalent 29
cations (e.g.Ca2+, Mg2+) [32]. Therefore, the continued exposure of the membranes to the effluent may have reduced its superficial charge and the repulsion of the co-ions as SO42-. This superficial charge reduction can also contribute to roughness modification, as previously discussed. The statistical test of Mann-Whitney was performed using Matlab R2008a software (The MathWorks, USA) in order to compare the retention of MgSO4 for both membranes after 105 days of exposure. It was concluded that the membrane performances could be considered significantly different to a significance level of 5% (%>hf5= = 0.0079), and that the retention of }~= f=hJ was the lowest. Thus, the exposure to HCl intensified the neutralization of the superficial charge and the swelling processes. Due to protonation of the surface’s functional groups of the membrane while in contact with the cleaning solution and increase in charge density, the membrane became more neutral and the pores, larger. As a result of hysteresis effects, a considerable increase in permeability following acid cleaning could be observed, particularly if the membrane had a very thin active skin layer [33]. Glucose retention also diminished with time (Figure 9b) and, as previously discussed, membrane swelling and fouling can explain this phenomenon. It can be also verified that the decrease in rejection after 105 days of exposure is larger for }~= f=hJ
(significant difference with %>hf5= = 0.0022). It is explained by the enlargement of membrane pores caused by the periodic cleanings with HCl solution (Table 5). It is worth noting that, although there have been changes in the rejection of membranes over time, the reduction was not very intense. NF membranes still showed good performance even after 285 days of exposure to effluent and to cleaning solution. Final retentions of MgSO4 and glucose were 95-88% and 88-82%, respectively. This is partly due to the use of a good pre-treatment. The pre-treatment included a pH adjustment of 30
the effluent, which in its original pH is very acidic and may cause polymer degradation, and the UF removal of suspended and colloidal solids, which can result in fouling and physical damage of the membranes. The lifetime of the membranes was estimated. For this purpose, it was assumed that the concentration of calcium in the permeate should be kept below 100 mg·L-1 to allow its reuse by the mining company [4], which corresponds to 84.5% removal efficiency. It was also considered that the removal efficiencies of calcium sulfate and magnesium sulfate were similar. The magnesium sulfate rejection data as a function of time was adjusted to a second-order polynomial model (Figure 9a) and the time to reach the minimum rejection was calculated. For the membrane exposed only to the effluent, the estimated lifetime was 536 days, while for the membrane submitted periodic cleaning, it was reduced to 389 days. Although this result was obtained for an exposure to the HCl solution higher than the optimized cleaning condition, it emphasizes the influence of the cleaning on the lifetime of the membranes, and enlighten its importance for the sustainability of the process. The effect of continuous exposure to the effluent and the combination effluent and cleaning agent on the chemical composition of the NF90 membrane was evaluated through attenuated total reflectance Fourier transform infrared spectroscopy (ATRFTIR). The spectrums are presented in Figure 12.
31
Neff+clean
Neff
Virgin
0,30
0,25
1239
1505
Absorbance
1471 1342
0,20
0,15
1612 1577
1172
846
1370
0,10
0,05
0,00 1800
1600
1400
1200
1000
800
-1
Wave number (cm )
Figure 12 – Infrared spectrums in the region between 1800 to 800 cm-1 to NF 90 membrane exposed only to the effluent and to the combination of effluent and cleaning agent In the infrared spectroscopy for attenuated total reflectance, the selective and intermediate layers are analyzed, and all results consist in an overlapping of spectrums from each layer [34]. Thus, bands of the selective layer, whose base polymer is polyamide, and also from the support layer, whose polymer is polysulfone, were verified on the spectrums obtained for NF90. The main detected bands, specific for the mentioned polymers, can be found in Table 6. Table 6 – Main IR bands of polysulfone (PSF) and polyamide (PA) components of NF90 membrane in the region between 1800 and 800 cm-1 Assignment Reported peaks Observed Vibration Intensity -1 a (cm ) peaks (cm-1) 1609 1612 Aromatic amide (N-H Weak PA deformation vibration or C=C ring stretching vibration) ~1587, 1577, Aromatic in-plane ring bend Middle 1504,1488 1505, 1471 stretching vibration ~1245 1239 C-O-C asymmetric stretching Strong PSF vibration of the aryl-O-aryl group 1385-1365 1370 C-H symmetric deformation Weak vibration of -C(CH3)2 32
a
1350-1280
1342
1180-1145
1172
~830
846
Asymmetric SO2 stretching vibration Symmetric SO2 stretching vibration In-phase out-of-plane hydrogen deformation of parasubstituted phenyl group
Middle Strong Weak
Peak values of PA and PSU were reported by Tang et al. [34]
Appearance or disappearance of the characteristic bands was not verified during the evaluation period, as it can be seen through a comparison of the spectrums presented in Figure 12. Therefore, the results obtained suggest that the degradation of the polymer polyamide did not occur, or it happened on a very small scale, which corroborate the results obtained in this study.
4. CONCLUSIONS With this study it was concluded that the NF process is a suitable treatment for gold mining effluent, allowing for a high conductivity and ions removal efficiency. However, the effects of concentration polarization and Hermia’s intermediate blocking fouling caused a reduction of approximately 80% of the permeate flux when the permeate recovery rate was 70%. SEM and EDS analysis showed that fouling basically is made up of calcium, magnesium, sodium, zinc, copper, aluminum, iron, arsenic, and silica. Calcium sulfate precipitated as bassanite over membrane surface since SI was larger than 1 even for low recovery rate. HCl solution was the most efficient cleaning solution. It should be applied over 90minute recirculation, without soak. NF membrane was exposed to the effluent and to the effluent and HCl solution for 285 days. Alterations in hydrophobicity, increase in pore size due to salt-induced swelling,
33
surface charge changes, and occurrence of fouling led to a decrease in the retentions of ionic (magnesium sulfate) and neutral solutes (glucose) by the two membranes. The retention of the membrane exposed to the cleaning solution was statistically lower, reinforcing the importance of optimizing the cleaning condition to maximize cleaning efficiency with the least exposure. Positively, the infrared spectroscopy results showed no indication of degradation of the polymeric material of the membrane. Although there were changes in the rejection of the membranes over time, the reduction was not very intense, and NF membranes still showed good performance even after 285 days of exposure to the effluent and the cleaning solution. Thus, it can be said that the results of this study are promising and demonstrate the robustness of NF technology for gold mining effluent treatment.
5. ACKNOWLEDGEMENTS The authors thank the Coordination of Improvement of Higher Education Personnel (CAPES) and the Foundation for Research Support of Minas Gerais (FAPEMIG) for the scholarships and financial resources provided. We also thank the microscopy center of the Federal University of Minas Gerais for AFM and SEM analysis; the Department of Metallurgical Engineering of the Federal University of Minas Gerais for contact angle analysis; and the Department of Chemistry of the Federal Center of Technological Education of Minas Gerais for infrared analysis.
6. REFERENCES [1] A. Mohammad, Y. Teow, W. Ang, Y. Chung, D. Oatley-Radcliffe, N. Hilal, Nanofiltration membranes review: Recent advances and future prospects, Desalination, 356 (2015) 226-254. 34
[2] B.C. Ricci, C.D. Ferreira, A.O. Aguiar, M.C. Amaral, Integration of nanofiltration and reverse osmosis for metal separation and sulfuric acid recovery from gold mining effluent, Separation and Purification Technology, (2015). [3] C. Sierra, J.R.Á. Saiz, J.L.R. Gallego, Nanofiltration of Acid Mine Drainage in an Abandoned Mercury Mining Area, Water, Air, & Soil Pollution, 224 (2013) 1-12. [4] L.H. Andrade, A.O. Aguiar, W.L. Pires, G.A. Miranda, L.P.T. Teixeira, G.C.C. Almeida, M.C.S. Amaral, Nanofiltration and Reverse Osmosis applied to gold mining effluent treatment and reuse, in, Brazilian Journal of Chemical Engineering, 2016. [5] A. Al-Amoudi, R.W. Lovitt, Fouling strategies and the cleaning system of NF membranes and factors affecting cleaning efficiency, Journal of Membrane Science, 303 (2007) 4-28. [6] T. Chidambaram, Y. Oren, M. Noel, Fouling of nanofiltration membranes by dyes during brine recovery from textile dye bath wastewater, Chemical Engineering Journal, 262 (2015) 156-168. [7] S. Madaeni, E. Salehi, Adsorption of cations on nanofiltration membrane: Separation mechanism, isotherm confirmation and thermodynamic analysis, Chemical Engineering Journal, 150 (2009) 114-121. [8] E.-m. Gwon, M.-j. Yu, H.-k. Oh, Y.-h. Ylee, Fouling characteristics of NF and RO operated for removal of dissolved matter from groundwater, Water Research, 37 (2003) 2989-2997. [9] A.I. Schäfer, A.G. Fane, T.D. Waite, Nanofiltration: principles and applications, Elsevier, 2005. [10] K. Chon, J. Cho, Fouling behavior of dissolved organic matter in nanofiltration membranes from a pilot-scale drinking water treatment plant: An autopsy study, Chemical Engineering Journal, 295 (2016) 268-277. [11] Z. Wang, J. Ma, C.Y. Tang, K. Kimura, Q. Wang, X. Han, Membrane cleaning in membrane bioreactors: a review, Journal of Membrane Science, 468 (2014) 276-307. [12] X. Wei, Z. Wang, F. Fan, J. Wang, S. Wang, Advanced treatment of a complex pharmaceutical wastewater by nanofiltration: Membrane foulant identification and cleaning, Desalination, 251 (2010) 167-175. [13] S. Zulaikha, W. Lau, A. Ismail, J. Jaafar, Treatment of restaurant wastewater using ultrafiltration and nanofiltration membranes, Journal of Water Process Engineering, (2014). [14] DOWFilmtec™, Product Information: FILMTEC NF90-400 Nanofiltration Element, in: Form n. 609-00345-0312. [15] A. Drak, K. Glucina, M. Busch, D. Hasson, J.-M. Laîne, R. Semiat, Laboratory technique for predicting the scaling propensity of RO feed waters, Desalination, 132 (2000) 233-242. [16] J. Sikder, S. Chakraborty, P. Pal, E. Drioli, C. Bhattacharjee, Purification of lactic acid from microfiltrate fermentation broth by cross-flow nanofiltration, Biochemical Engineering Journal, 69 (2012) 130-137. [17] R. Mukherjee, M. Mondal, A. Sinha, S. Sarkar, S. De, Application of nanofiltration membrane for treatment of chloride rich steel plant effluent, Journal of Environmental Chemical Engineering, 4 (2016) 1-9. [18] M.C.V. Vela, S.Á. Blanco, J.L. García, E.B. Rodríguez, Analysis of membrane pore blocking models adapted to crossflow ultrafiltration in the ultrafiltration of PEG, Chemical Engineering Journal, 149 (2009) 232-241. [19] S. Mattaraj, C. Jarusutthirak, R. Jiraratananon, A combined osmotic pressure and cake filtration model for crossflow nanofiltration of natural organic matter, Journal of Membrane Science, 322 (2008) 475-483. 35
[20] DOWFilmtec™, Cleaning and Disinfection Procedures for FilmTec™ NF200 and NF270 Elements, in: Form n. 609-00388-1006. [21] DOWFilmtec™, Dow FilmTec NF90 Nanofiltration Elements for Commercial Systems, in: Form n. 609-00378-0811. [22] M. González, I. Saucedo, R. Navarro, P. Prádanos, L. Palacio, F. Martínez, A. Martín, A. Hernández, Effect of phosphoric and hydrofluoric acid on the structure and permeation of a nanofiltration membrane, Journal of Membrane Science, 281 (2006) 177-185. [23] M. Dubois, K.A. Gilles, J.K. Hamilton, P. Rebers, F. Smith, Colorimetric method for determination of sugars and related substances, Analytical chemistry, 28 (1956) 350356. [24] X.-L. Wang, T. Tsuru, M. Togoh, S.-i. Nakao, S. Kimura, Evaluation of pore structure and electrical properties of nanofiltration membranes, Journal of chemical engineering of Japan, 28 (1995) 186-192. [25] J. Schaep, B. Van der Bruggen, C. Vandecasteele, D. Wilms, Influence of ion size and charge in nanofiltration, Separation and Purification Technology, 14 (1998) 155162. [26] C. Bellona, M. Marts, J.E. Drewes, The effect of organic membrane fouling on the properties and rejection characteristics of nanofiltration membranes, Separation and Purification Technology, 74 (2010) 44-54. [27] Y. Kaya, Z. Gönder, I. Vergili, H. Barlas, The effect of transmembrane pressure and pH on treatment of paper machine process waters by using a two-step nanofiltration process: Flux decline analysis, Desalination, 250 (2010) 150-157. [28] S. Madaeni, S. Samieirad, Chemical cleaning of reverse osmosis membrane fouled by wastewater, Desalination, 257 (2010) 80-86. [29] A. Al-Amoudi, P. Williams, S. Mandale, R.W. Lovitt, Cleaning results of new and fouled nanofiltration membrane characterized by zeta potential and permeability, Separation and Purification Technology, 54 (2007) 234-240. [30] J. Luo, Y. Wan, Effects of pH and salt on nanofiltration—a critical review, Journal of Membrane Science, 438 (2013) 18-28. [31] D. Nanda, K.-L. Tung, Y.-L. Li, N.-J. Lin, C.-J. Chuang, Effect of pH on membrane morphology, fouling potential, and filtration performance of nanofiltration membrane for water softening, Journal of Membrane Science, 349 (2010) 411-420. [32] C. Mazzoni, L. Bruni, S. Bandini, Nanofiltration: role of the electrolyte and pH on Desal DK performances, Industrial & engineering chemistry research, 46 (2007) 22542262. [33] A. Simon, J.A. McDonald, S.J. Khan, W.E. Price, L.D. Nghiem, Effects of caustic cleaning on pore size of nanofiltration membranes and their rejection of trace organic chemicals, Journal of Membrane Science, 447 (2013) 153-162. [34] C.Y. Tang, Y.-N. Kwon, J.O. Leckie, Effect of membrane chemistry and coating layer on physiochemical properties of thin film composite polyamide RO and NF membranes: II. Membrane physiochemical properties and their dependence on polyamide and coating layers, Desalination, 242 (2009) 168-182.
36
7. HIGHLIGHTS •
The best cleaning procedure proved to be 90-minutes recirculation with HCl solution
•
Membrane exposed to combined effluent and cleaning solution had lower retention
•
In 285 days of exposure, membrane hydrophobicity and effective pore radius changed
•
No indication of degradation of the polymeric material of the membrane was observed
•
Results demonstrate the robustness of NF for gold mining effluent treatment
37