Optimization of a membrane cleaning strategy for advanced treatment of polymer flooding produced water by nanofiltration

Abstract

Recycling the polymer flooding produced water (PFPW) into polymer flooding oil extraction after advanced treatment by nanofiltration (NF) is a reasonable choice for the effective management of PFPW. However, membrane cleaning is indispensable because of the inevitable membrane fouling phenomenon. In this study, the NF process was illustrated to treat synthetic PFPW by taking anion polyacrylamide (APAM) and crude oil as target foulants, and during which membrane cleaning was investigated. Different cleaning agents were studied and used in combination. A single factor experiment and orthogonal experiment were conducted to optimize the cleaning strategy. Flux measurements, salt rejection experiments, AFM, a contact angle goniometer and ATR-FTIR were employed to evaluate the cleaning performance. The results indicate that the optimized alkaline formulated cleaning solution (pH = 11) consists of 0.05% EDTA, 0.2% sodium pyrophosphate and 0.2% SDS. The optimized cleaning strategy should include two steps: (i) the first step is to clean with the alkaline formulated cleaning agent for 0.5 h; (ii) the second step is to clean with a HCl solution (pH = 2) for 0.5 h. The optimized cleaning strategy can fully recover the membrane flux without damaging its other properties, including desalting ability, surface morphology, hydrophilicity and chemical bonds.

---

1. Introduction

As one of the enhanced oil recovery (EOR) technologies, polymer flooding involving anion polyacrylamide (APAM), on account of its easy operability, broad adaptability and relatively low capital cost, has become a significant method to ensure crude oil production in the Daqing oilfield in China.^1–4 Nevertheless, in the process of polymer flooding oil extraction, vast amounts of clean water is consumed for preparing the APAM flooding solution, followed by the generation of a great deal of polymer flooding produced water (PFPW). Direct discharging of the PFPW without proper treatment will not only bring much harm to the environment, but will also be a waste of water resources, particularly for water-deficient areas. Reclamation of the produced water from the petroleum industry for irrigation, livestock or wildlife watering and various industrial uses, has become the consensus of many researchers around the world.^5,6 Therefore, recycling of the polymer flooding produced water (PFPW) into polymer flooding oil extraction after proper treatment is a reasonable choice for the effective management of PFPW.

Recent advancements have demonstrated the reliability and potential for applying membrane technologies to produce water management, but several challenges persist, such as membrane fouling and the absence of efficient cleaning procedures.^6–13 In contrast with the conventionally produced water, PFPW contains not only crude oil, salts, etc., but also residual APAM, even after a series of pretreatments, such as floatation, coagulation, sedimentation, sand filtration and ultrafiltration (UF).^4,14 Cations, especially divalent cations, would cause the shielding of APAM’s negative charge, thereby leading to the linear polymer chain’s coiling up and then the decrease of solution viscosity, eventually making the recycled water not qualified for the preparation of the APAM flooding solution.¹⁵ In order to meet the water standards for polymer flooding and ensure stable oil production, nanofiltration (NF) has been used to treat the PFPW.¹⁴ However, salts, APAM and crude oil, which pass through the pretreatments, can cause membrane fouling of the NF membrane in the subsequent NF filtration step. According to Zhang et al.,¹⁴ membrane fouling was dominantly induced by APAM, as well as salts and crude oil. Membrane fouling results in productivity decline, increased energy consumption and treatment cost, as well as shortened membrane lifespan.

Although many strategies (such as pretreatment, membrane modification, optimization of operation conditions and so on) can be taken to mitigate membrane fouling, fouling cannot be completely avoided. Membrane cleaning is required when there is a serious drop in permeate flux, or when there is a need to increase the trans-membrane pressure significantly to maintain the desired water flux.¹⁶ Chemical cleaning is necessary because physical cleaning is generally inefficient for organic fouling and scaling. The commonly used chemical agents can be classified into six categories: acids, alkalis, surfactants, metal chelating agents, oxidizing agents and enzymes.¹⁷ Effective chemical cleaning can be divided into three mechanisms: (i) mass transfer of chemical agents from the bulk phase to the fouling layer, (ii) chemical reaction between cleaning agents and foulants in the fouling layer and (iii) mass transfer of foulants from the fouling layer to the bulk phase.¹⁶ After these processes, the membrane performance can be remedied.

In order to maintain and restore the membrane characteristics, the ideal cleaning strategy should not only be effective against target foulants, but also gentle to membrane materials.¹⁷ Up to now, most studies on membrane cleaning placed their focus mainly on the treatments of surface water, ground water, seawater, municipal wastewater and food industry wastewater.^18–21 The involved foulants in their research are mainly natural organic matters (NOMs), proteins, polysaccharides, fatty acids and inorganic precipitations. A systematic study of the NF membrane cleaning strategy involving PFPW, of which major foulants are APAM and crude oil, has not been found in the literature. As the optimal chemical cleaning strategy is a function of foulants and membrane material, it is necessary to carry on related research.

In this study, the NF process was illustrated to treat the synthetic PFPW by taking APAM and crude oil as target foulants, and during which membrane cleaning was investigated. Sodium hydroxide (NaOH), disodium ethylenediamine tetraacetate (EDTA), sodium dodecyl sulfate (SDS), sodium pyrophosphate (SPP) and hydrochloric acid (HCl) were studied and used in combination. A single factor experiment and orthogonal experiment were conducted to optimize the cleaning strategy by taking flux recovery as an assessment index. In addition, salt rejection experiments, AFM, a contact angle goniometer and ATR-FTIR were employed to evaluate whether the selected cleaning strategy deteriorates other membrane properties (desalting ability, surface morphology, hydrophilicity/hydrophobicity and chemical bonds). The present work can be a reference for membrane fouling control during the advanced treatment of PFPW by NF.

2. Materials and methods

2.1. Model foulants

The target organic foulants in this study were APAM and crude oil. APAM used in this work is a hydrophilic linear polymer with a molecule weight of 500 000 g mol⁻¹, and a purity higher than 99% (Daqing, China). The chemical structure of APAM is shown in Fig. 1.²² The crude oil was provided by the 5th Daqing oil production factory. Its moisture content is lower than 0.5% after electric dehydration. No further purification was conducted before use. The model PFPW was prepared by adding a certain amount of APAM or crude oil in the saline water (SW), which consists of ultrapure water, NaCl, K₂SO₄, CaCl₂, MgCl₂·6H₂O, and NaHCO₃. All the inorganic salts are of analytical grade from Tianjin Benchmark Chemical Reagent Co. Ltd (Tianjin, China). Water used in this research was ultrapure water (Milli-Q water, Millipore Milli-Q reference ultrapure water purification system, USA).The ion composition of SW is show in Table 1. According to ref. 14, the average concentrations of APAM and crude oil in real PFPW are 30 mg L⁻¹ and 1.5 mg L⁻¹, respectively. In order to accelerate the membrane fouling process, the concentration of APAM and crude oil in this study are fixed at 60 mg L⁻¹ and 3 mg L⁻¹, respectively. The specific methods of preparing different model solutions can be found in ref. 14. The model solutions used in this study are listed in Table 2.

Fig. 1 Molecular structure of APAM, SDS and the polyamide layer of the NF90 membrane.

Table 1 Ion composition of SW

Ion types	Na^+	K^+	Ca^2+	Mg^2+	Cl^−	HCO3^−	SO4^2−	Total salinity
Concentration (mg L^−1)	1442.63	44.83	7.21	75.76	1195.71	2178.57	55.17	5000

---

Table 2 Model solutions used in filtration experiments^a

Model solution types	Salts	APAM	Crude oil
a “✓” in the table denotes that the model solution contains this component. “—” in the table denotes that the model solution does not contain this component.
SW (saline water)	✓	—	—
SWCO (saline water containing 3 mg L^−1 crude oil)	✓	—	✓
SWCA (saline water containing 60 mg L^−1 APAM)	✓	✓	—
SWCAO (saline water containing 60 mg L^−1 APAM and 3 mg L^−1 crude oil)	✓	✓	✓

---

2.2. Characterization of the NF membrane

The thin-film composite NF90 (Dow FilmTec, Minneapolis, MN, USA) was used as the model NF membrane. NF90 is a fully aromatic polyamide membrane made from 1, 3-benzenediamine and trimesoyl chloride via the process of interfacial polymerization. The molecular structure of the polyamide layer of NF90 is illustrated in Fig. 1. Various properties of NF90 are summarized in Table 3.

Table 3 Summary of the characteristics of NF90

Parameters	Values
a Data from ref. 23 and 24.b According to the manufacturer.c Data from ref. 25.d Data from ref. 26.
MWCO^a (Da)	100–200
Rejection of NaCl^b (%)	85–95
Water permeability^b (L m^−2 h^−1 bar^−1)	5.2–8.1
Contact angle (°)	61.5
Max operating temperature^b (°C)	45
Max operating temperature when feed solution^b pH > 10 (°C)	35
Operating pH range^b	3–10
Cleaning pH range^b	1–12
Max operating pressure^b (bar)	41
Zeta potential^c (mv, pH = 6.3)	−31.1
Average pore diameter^d (nm)	0.64 ± 0.01
Max cleaning time when cleaning solution^b pH = 1 or 12 (min)	30

---

2.3. Chemical cleaning agents

As mentioned in the Introduction, the commonly used chemical agents can be classified into six categories: acids, alkalis, surfactants, metal chelating agents, oxidizing agents and enzymes.¹⁷ According to the proposed fouling mechanism involving treating PFPW,¹⁴ organic fouling induced by APAM and crude oil plays a dominant part. The interactions involving organic fouling include hydrogen bonds, complexing action, hydrophobic interaction and electrostatic interaction. Inorganic fouling induced by scaling also exists. As is well-known, the poor chlorine-resistance ability is a major drawback of the polyamide NF membrane,^27,28 so the oxidizing agents were not used here. The enzyme cleaning agent is always too expensive and needs strict conditions of use. Therefore, it is not chosen as well. Based on the fouling mechanism and cleaning mechanism of different chemical agents, the selected single chemical cleaning agents in this study include: NaOH (pH = 11.0) as an alkaline solution, HCl (pH = 2.0) as an acidic solution, certified grade disodium ethylenediaminetetraacetate (EDTA) as a metal chelating agent, certified grade sodium dodecyl sulfate (SDS) as an anionic surfactant, and sodium pyrophosphate (SPP) as a “generalist” which can act as a pH buffer agent, chelating agent and surfactant. As the chelating ability of EDTA will be maximized when the solution pH is up to 11.0,¹⁶ the EDTA solution pH was adjusted to 11.0 before use. The SDS molecule (as show in Fig. 1) has a hydrophilic head and a hydrophobic tail. The agents were purchased from Tianjin Kermel Chemical Reagent Co. Ltd (Tianjin, China) and used without further purification. The pH of various cleaning solutions was adjusted with 1.0 M NaOH as necessary.

2.4. Experimental set-up and operational procedure

Nanofiltration experiments were conducted by using the cross-flow NF set-up as illustrated in Fig. 2. The effective membrane surface area in the membrane cell is 0.0025 m². The piston pump together with the frequency converter and pressure sensor could provide relatively constant pressure under 1.5 MPa (pressure fluctuation ≤ 0.01 MPa). Electronic information of mass, pressure and temperature could be automatically collected by the computer. The concentrates were recirculated back into the feed tank. The re-circulation flow rate was controlled by a rotameter to keep a constant cross-flow velocity.

Fig. 2 The schematic diagram of the cross-flow NF set-up.

Operational procedure is illustrated in Fig. 3. The new membranes were soaked in ultrapure water for 72 h prior to use. Pre-pressuring with ultrapure water (temperature: 30 °C) at 1.0 MPa was conducted until the water flux became constant. And then the pump was stopped to decant the ultrapure water. The relevant feed solution (SWCA, SWCO or SWCAO) was added in the feed reservoir. In the fouling runs, operating pressure and re-circulation flow rate were fixed at 0.8 ± 0.01 MPa and 3.5 ± 0.05 mL s⁻¹, respectively. And the cross-flow velocity was 7.0 cm s⁻¹. The temperatures of various feed solutions were all fixed at 30 °C to mimic the real situation of the Daqing oilfield. Filtration experiments were carried out until the flux declined to 70% of the initial flux. The flux value in the first 5 min was regarded as the flux before fouling (J₀). The flux value in the last 5 min was regarded as the flux after fouling (J_f).

Fig. 3 Schematic diagram of the operational procedure.

In order to prevent the pipeline system of the NF set-up from being fouled by the cleaning agents, the membrane cleaning processes were conducted in a flask. According the manufacturer of NF90 (as shown in Table 3), the maximum cleaning time (when solution pH = 1 or 12) is 30 min and the max operating temperature (when feed solution pH > 10) is 35 °C. Therefore, the cleaning time and temperature were respectively specified as 30 min and 30 °C so as to prevent the membrane from being damaged. After the fouling run, the fouled NF membrane was taken out from the membrane cell and put into a flask containing 800 mL of cleaning solution with a temperature of 30 °C. The flask was placed in an air bath thermostat oscillator (30 °C, 120 rpm) for 0.5 h. Then the membrane was taken out and flushed with ultrapure water to remove the cleaning agent. Again, the cleaned NF membrane was installed into the membrane cell. The pump was started with a pressure of 0.8 MPa. The flux value in the first 5 min was regarded as the flux after cleaning (J_c). The cleaning efficiency of different cleaning strategies was evaluated by calculating and comparing flux recovery (FR). The FR was calculated according to eqn (1).

2.5. Salt rejection experiments

The membranes used for salt rejection experiments included: virgin membrane (VM), a new membrane that experienced the pre-pressure described in Section 2.5; fouled membrane (FM), the virgin membrane fouled by SWCAO, of which the water flux decreased to 70% of the initial flux, was regarded as the fouled membrane; cleaned membrane (CM), the FM cleaned by the optimized cleaning strategy obtained in Section 3.2.

VM, FM and CM were respectively employed to treat a 1000 mg L⁻¹ NaCl solution at 30 °C. The operating pressure and cross-flow velocity were fixed at 0.8 ± 0.01 MPa and 7.0 cm s⁻¹, respectively. The conductivities of the feed solution and permeate that accumulated in 30 min were measured and used for calculating conductivity rejection. The conductivity rejections of VM, FM and CM were regarded as their salt rejections.

2.6. Membrane characterization

All the membranes used for characterization (AFM, CA and ATR-FTIR) should be dried at 25 °C for 48 h prior to analysis. Atomic force microscope (AFM) (Bioscope, Veeco, USA) was used to analyse morphologies and roughness of different membranes. The scanning pattern of the probe was in tapping mode in the air. The images were flattened with order 2 after scanning, so as to remove curvature and slope from the images.²⁹ Then the RMS roughness (root-mean-squared roughness) was calculated according to eqn (2).where Z̄ is the average of the z values within a given area, Z_n is the current z value, and N is the number of data points within a given area.²⁹

Static contact angles (CAs) were measured for various membranes with a contact angle goniometer (SL200B3, Solon, China) by the sessile drop method (0.1 μL). Each CA value was the average of three different positions on the same membrane piece. Contact angles were calculated using the circle fitting method. ATR-FTIR (Spectrum One B, PerkinElmer, USA) was used to analyse functional groups of different membranes, and further speculate material composition. Samples were scanned in the range of 4000–650 cm⁻¹ with a resolution of 1 cm⁻¹.

3. Results and discussion

3.1. Cleaning efficiency of various chemical agents against an individual foulant

In order to understand the cleaning efficiency of different chemical agents, as well as the appropriate dosage, single factor experiments were conducted here by taking crude oil or APAM as the target foulant.

3.1.1 Cleaning efficiency of various chemical agents against crude oil. Here the fouling experiments were conducted by taking SWCO as the feed solution. Cleaning agents including a HCl solution (pH = 2), NaOH solution (pH = 11), SDS solutions with different concentrations, SPP solutions with different concentrations and EDTA solutions (pH = 11) with different concentrations were respectively studied. Cleaning with ultrapure water served as a baseline. Cleaning efficiencies of various cleaning agents against crude oil are shown in Fig. 4.

Fig. 4 Cleaning efficiencies of different cleaning agents against crude oil.

Fig. 4 clearly shows that cleaning with ultrapure water and HCl (pH = 2) is ineffective as their flux recoveries were just 21.15% and 26.01%, respectively. The flux recoveries of NaOH (pH = 11) and EDTA solution (pH = 11) with different concentrations were very close and around 35%. This points out that NaOH and EDTA cleaning performed at these conditions were not effective. Cleaning with SPP is relatively effective, and a higher concentration results in higher flux recovery. When the concentration of SPP is 0.2% (w/v), its flux recovery can reach 65.41%. Cleaning with SDS was obviously more effective than other cleaning agents. When the concentration of SDS reaches 0.1% (w/v), its flux recovery can reach 87.28%. However, further increasing the concentration cannot obviously promote the cleaning efficiency anymore.

A water molecule itself and HCl solution (pH = 2) cannot break up the relatively strong hydrophobic force between crude oil and the membrane surface. In addition, the possible scaling covered by the crude oil fouling layer can’t be dissolved and flushed away by HCl solution. Therefore, cleaning with ultrapure water and HCl (pH = 2) was ineffective. The cleaning mechanism of NaOH is hydrolysis and solubilisation of organic foulants, as well as generating an electrostatic repulsive force between the negatively charged membranes and foulants when the solution pH is elevated.²¹ However, there are some Ca²⁺, Mg²⁺ and much HCO₃⁻ in SWCO. The evaluated pH could lead to the transformation of HCO₃⁻ to CO₃²⁻, which aggravated the inorganic fouling (scaling) on the membrane surface. Hence the flux recovery of NaOH (pH = 11.0) was only around 35%. As the fouling mechanism of crude oil does not involve complexing action,¹⁴ the chelate cleaner (EDTA solutions (pH = 11) at different concentrations) merely acts as an alkaline cleaning agent (NaOH, pH = 11). SPP, as a “generalist”, can act as a pH buffer agent, chelating agent and surfactant, but its cleaning efficiency is always lower than SDS solutions, even at high concentrations. Because of the amphiphilic character of SDS, the hydrophilic tails of SDS molecules stretched into bulk solution, while the hydrophobic tails adsorbed onto crude oil molecules. So the amphiphilic SDS molecules can remove the crude oil from the membrane surface with the turbulent flow of the cleaning solution. In addition, some SDS molecules may be residual on the membrane surface by hydrophobic binding between its hydrophobic tail and membrane material. The hydrophilic tails stretching into feed solution could improve the membrane hydrophilicity. These two mechanisms can both increase water flux. Nevertheless, cleaning with SDS was ineffective to scaling. Therefore, we can see the flux recovery of SDS (0.1% (w/v)) is up to 87.28%, but still lower than 90%. The phenomenon that further increasing the concentration (0.2%, 0.3% and 0.4%, w/v) cannot obviously promote the cleaning efficiency any further indicates that a 0.1% (w/v) SDS solution is adequate for this fouling condition.

3.1.2 Cleaning efficiency of various chemical agents against APAM. Here the fouling experiments were conducted by taking SWCA as the feed solution. Cleaning agents including HCl solution (pH = 2), NaOH solution (pH = 11), SDS solutions with different concentrations, SPP solutions with different concentrations and EDTA solutions (pH = 11) with different concentrations were respectively studied. Cleaning with ultrapure water served as a baseline. Cleaning efficiencies of various chemical agents against APAM are shown in Fig. 5.

Fig. 5 Cleaning efficiencies of different cleaning agents against APAM.

Fig. 5 indicates that cleaning with NaOH (pH = 11.0) had the lowest flux recovery (57.91%), even under the flux recovery (63.78%) of ultrapure water cleaning. The flux recovery of acid cleaning with HCl solution (pH = 2) was 69.97%, higher than that of ultrapure water cleaning and alkaline cleaning. When the concentration of EDTA solutions (pH = 11) increases from 0.05% to 0.4%, the flux recovery just increases from 72.65% to 75.34%, implying that a higher concentration is unnecessary. As for SPP cleaning, when the concentration reaches 0.2%, the flux recovery can achieve 84.55%, and higher concentrations cannot visibly promote the cleaning efficiency. Just like the condition in Section 3.1.1, cleaning with SDS solution is the most effective, and the flux recovery is up to 87.55% when the solution concentration is 0.2%. However, further increasing the concentration cannot obviously facilitate the cleaning performance any further.

As APAM is a kind of hydrophilic polymer, ultrapure water can dissolve part of the fouling layer. Hence its flux recovery was up to 63.78%, much higher than that (21.25%) of ultrapure water cleaning when the membrane was fouled by crude oil. Although NaOH cleaning could break up the hydrogen binding, as well as enhance the repulsion between APAM and the membrane surface, it can improve the pH and aggravate the inorganic fouling (scaling, such as CaCO₃) on the membrane surface, consequently leading to a lower flux recovery than ultrapure water cleaning. Compared with ultrapure water cleaning, HCl solution (pH = 2) can dissolve the scaling in the fouling layer, thus getting a higher flux recovery (69.97%). When the APAM fouling layer was exposed to an EDTA solution, a ligand-exchange reaction between EDTA and APAM–calcium–APAM complexes/APAM–calcium–membrane complexes would set off. Consequently, the APAM gel layer was broken down. However, just like NaOH solution, the alkaline EDTA solution may also facilitate scaling. Thus its flux recovery (72.65–75.34%) was not too high. The cleaning mechanism of SDS against APAM is similar to crude oil fouling. But as the molecule weight of APAM is much higher than that of crude oil, it was more difficult to be dragged into the bulk solution by SDS molecules. SPP, as a “generalist”, can act as a chelating agent and surfactant. When it acted as a chelating agent, the chelating ability of SPP was lower than that of EDTA, and when it acted as a surfactant, its performance cannot match that of SDS. As a result, the flux recovery (84.55%) of SPP was a little higher than that of EDTA (75.66%), but lower than that of SDS (87.55%) when the solution concentration was 0.1%.

3.2. Optimizing of the formulated cleaning agent against combined fouling

According to the results obtained in Section 3.1, SDS had the highest cleaning efficiency against both crude oil and APAM fouling, but the flux recoveries were still under 90%. The alkaline EDTA solution (pH = 11.0) was relatively effective against APAM fouling. SPP can not only act as a surfactant and chelating agent, but also serve as a pH buffer agent. Therefore, an alkaline formulated cleaning agent consisting of SDS, EDTA and SPP is expected to play their respective functions and have a high cleaning efficiency against combined fouling induced by salts, APAM and crude oil. However, this formulated cleaning agent would be ineffective against inorganic fouling because all three chemical agents cannot break up scaling. On the contrary, the alkaline solution may improve pH and aggravate scaling. In general, acid cleaning is suitable for the removal of precipitated salts, such as CaCO₃.¹⁷ Moreover, previous research found that alkaline cleaning can result in the enlargement of effective membrane pores, while acid cleaning had the opposite effect.^16,30 If the effective membrane pores are enlarged, the solute rejection would be decreased. Therefore, alkaline formulated cleaning followed by acid cleaning is highly necessary. In this way, the inorganic precipitates can be removed. Meanwhile, the enlarged membrane pores resulting from alkaline cleaning can shrink back to its native state after acid cleaning.

An orthogonal experiment was conducted to optimize the mass fraction of SDS, EDTA and SPP in the alkaline formulated cleaning solution. SDS, EDTA and SPP were taken as the three factors. According to the results obtained from Section 3.1, when the solution concentration exceeds 0.2% (w/v), further increasing the concentration cannot obviously facilitate the cleaning performance any further. Therefore, the levels of each factor were set as 0.05% (w/v), 0.1% (w/v) and 0.2% (w/v), respectively. The pH of each alkaline formulated cleaning solution was adjusted to 11. All the alkaline formulated cleaning processes were followed by acid cleaning with HCl (pH = 2.0). Each cleaning process continued for 0.5 h as described in Section 2.5. The interactions between different factors were not considered. The evaluation index of each test was the flux recovery. The orthogonal experiment arrangement and results are show in Tables 4 and 5, respectively.

Table 4 The factors and levels of the orthogonal experiment

Level	Factor
	EDTA% (w/v)	SPP% (w/v)	SDS% (w/v)
1	0.05	0.05	0.05
2	0.1	0.1	0.1
3	0.2	0.2	0.2

---

Table 5 The results of the orthogonal experiment

No.	Test
	EDTA% (w/v)	SPP% (w/v)	SDS% (w/v)	Flux recovery (%)
1	0.05	0.05	0.05	45.59
2	0.05	0.1	0.1	90.21
3	0.05	0.2	0.2	102.35
4	0.1	0.05	0.1	60.38
5	0.1	0.1	0.2	65.54
6	0.1	0.2	0.05	92.02
7	0.2	0.05	0.2	96.64
8	0.2	0.1	0.05	60.75
9	0.2	0.2	0.1	90.83
K1	238.15	202.61	198.36	—
K2	217.94	216.5	241.42	—
K3	248.22	285.2	264.53	—
k1	79.38	67.54	66.12	—
k2	72.65	72.17	80.47	—
k3	82.74	96.07	88.18	—
R	10.09	27.53	22.06	—

---

Table 5 indicates that the factors in the order of significance are: SPP > SDS > EDTA. Group no. 3 and 7 have satisfying cleaning efficiencies as their flux recoveries are both above 95%, reaching 102.35% and 96.64%, respectively. However, the economic factor should also be considered. According to the Chinese market price, the prices of EDTA, SPP and SDS are RMB 15000 per t, RMB 7000 per t and RMB 5500 per t, respectively. The price of EDTA is much higher than that of SPP. Group no. 3 and 7 need the same amount of SDS. However, group no. 7 consumed more EDTA (0.2%, w/v) and less SPP (0.05%, w/v), while group no. 3 consumed less EDTA (0.05%, w/v) and more SPP (0.2%, w/v). That is to say, the formulated cleaning agent of group no. 3 is not only more effective (flux recovery = 102.35%), but also cheaper. Therefore, the optimized alkaline formulated cleaning solution consisted of 0.05% EDTA, 0.2% sodium pyrophosphate and 0.2% SDS. The optimized cleaning strategy should include two steps: (i) the first step is to clean with the optimized alkaline formulated cleaning agent for 0.5 h; (ii) the second step is to clean with an HCl solution (pH = 2.0) for 0.5 h. After this cleaning strategy, the flux recovery was up to 102.35%. The higher flux after cleaning than that of the new membrane may because of the increased hydrophilicity resulting from the residual cleaning agent on the membrane surface.

3.3. Flux attenuation characteristics of the whole operational procedure

The flux curve of the whole operational procedure including the pre-pressure process, fouling process and operation after cleaning is shown in Fig. 6. SWCAO was taken as the feed solution in the fouling process and operation after cleaning. The optimized cleaning strategy was adopted to clean the fouled membrane.

Fig. 6 Flux attenuation characteristics of the whole operational procedure.

As described in Fig. 6, the pre-pressure process (1.0 MPa, 30 °C) needs about 10.8 h to get a stable pure water flux, dropping from 99.12 L m⁻² h⁻¹ to 90.37 L m⁻²·h⁻¹. After substituting ultrapure water for SWCAO as the feed solution, the initial flux value sharply decreased to 26.88 L m⁻² h⁻¹.

The permeate flux is commonly described with the following resistance model:

where J is the flux, Δp is the trans-membrane pressure, σ is the reflection coefficient, Δπ is the osmotic pressure difference, μ is the solution viscosity and R is the total resistance.

In the initial stage of the fouling process, the trans-membrane pressure (Δp) decreased to 0.8 MPa from 1.0 MPa. Meanwhile, the solution viscosity (μ) and osmotic pressure difference (Δπ) increased because of the addition of the hydrophilic linear polymer (APAM) and various inorganic salts. Moreover, total resistance (R) increased as the hydrophilic APAM molecules could rapidly attach on the polyamide membrane surface by hydrogen bonding and complexing action.¹⁴ All the above factors would result in the sharp decrease of the initial flux value with SWCAO as the feed solution in the fouling process.

It took about 9.65 h to finish the fouling process and make J_f/J₀ reach 70%. After the membrane cleaning process, the operational procedure with SWCAO was re-conducted with the cleaned membrane. As illustrated in Fig. 6, the new membrane and the cleaned membrane have similar flux curves. The similar flux attenuation characteristics indicate that the optimized cleaning strategy can effectively remove the foulants and maintain the membrane stability at the same time.

3.4. Influence of the optimized cleaning strategy on NF membrane properties

Flux measurement is an indirect assessment of the fouling and cleaning process, but it cannot fully assess the cleaning efficiency, since destructive cleaning methods can also significantly improve the water flux at the cost of modifying membrane structure and physicochemical properties. Therefore, in this section, multiple evaluations were further conducted by employing salt rejection experiments, AFM, a contact angle goniometer and ATR-FTIR.

3.4.1 Influence on desalting ability. The salt rejections of the VM, FM and CM are shown in Fig. 7. As described in the figure, the fouled membrane had the highest desalting ability. And its salt rejection was up to 91.55%, obviously higher than that of the virgin membrane (86.21%). When the fouled membrane was cleaned with the optimized cleaning strategy, the salt rejection decreased from 91.55% to 87.62%, very close the salt rejection of the virgin membrane (86.21%).

Fig. 7 Salt rejections of various NF membranes.

As explained in ref. 14, the relative dense fouling layer can produce an additional barrier to ion penetration across the membrane. Thus the FM had a stronger desalting ability compared with the VM and CM. When the fouling layer was removed by chemical cleaning, the membrane was restored to its original desalting ability. The similar salt rejections of the VM and CM imply that the selected cleaning strategy could effectively remove the foulants and recover water flux without deteriorating desalting ability.

3.4.2 Influence on membrane surface morphology. AFM was employed to analyse the morphologies and roughness of the VM, FM and CM. As indicated in Fig. 8(a), there were obvious peaks and valleys distributed on the surface of the VM, the RMS roughness was 42.1 nm. Fig. 8(b) shows that fouling resulted in an obvious membrane morphology change, modifying the surface of the FM rather smooth. The RMS roughness was reduced to 18.0 nm. The reason for this phenomenon is that APAM and crude oil could attach on the membrane surface and cover the inherent peaks and valleys on the clean NF90. When the FM was cleaned by the selected cleaning strategy, peaks and valleys reappeared on the surface of the CM. Meanwhile, the RMS roughness increased to 39.4 nm, close to that of the VM. The AFM images of the VM, FM and CM imply that the selected cleaning strategy could effectively remove the fouling layer without destruction of the membrane surface structure.

Fig. 8 AFM images of various NF membranes: (a) virgin membrane, (b) fouled membrane and (c) cleaned membrane.

3.4.3 Influence on hydrophilicity/hydrophobicity. The contact angle indirectly reflects the hydrophilicity/hydrophobicity of the membrane surface. A contact angle goniometer was employed to analyse the contact angles of the VM, FM and CM. The results are shown in Fig. 9. Compared with the contact angle of the VM (61.61°), the contact angle of the FM was much higher, up to 78.57°. The increase of the contact angle was attributed to hydrophobic crude oil mixed in the fouling layer.¹⁴ When the fouled membrane was cleaned by the selected cleaning agent, the contact angle decreased to 59.62°, slightly lower than that of the VM (61.16°), indicating that the foulants on the membrane were effectively removed. The slight decrease of the contact angle after cleaning was probably because of the residual surfactant on the membrane surface.³¹

Fig. 9 Contact angles of various NF membranes.

3.4.4 Influence on ATR-FTIR spectra. ATR-FTIR was employed to infer the material compositions of different NF membranes from the viewpoint of chemical bonds. As shown in Fig. 10, curve (1), (2) and (4) are the infrared adsorption spectra of the VM, CM and FM (membrane fouled by SWCAO), respectively. In addition, curve (3) is the infrared adsorption spectrum of the membrane fouled by SWCO, which provides the typical IR of crude oil. The three peaks at 1487 cm⁻¹ (C=C aromatic ring stretching), 1238 cm⁻¹ (C–O–C asymmetric stretching vibration of the aryl–O–aryl group) and 1150 cm⁻¹ (C–SO₂–C symmetric stretching) are the typical IR absorption peaks of the NF90 membrane which consists of polyamide and polysulfone materials.^32,33 As the penetration depth of the IR spectrum in the low wave number region is deeper, these three peaks appear in all four adsorption curves.

Fig. 10 FTIR spectra of various NF membranes.

Curve (3) implies that crude oil fouling will result in the appearance of three peaks: the peaks at 2853 cm⁻¹, 2918 cm⁻¹ and 1739 cm⁻¹. The peaks at 2853 cm⁻¹ and 2918 cm⁻¹ are corresponding to the C–H stretch of an alkyl.³⁴ As there are alkyls in both crude oil and SDS, these two peaks can be regarded as a sign of crude oil or SDS.^34,35 The peak at 1739 cm⁻¹ is related to the specific absorbance peak of C=O bonds in aldehydes and ketones, which are the components of crude oil. Therefore, these three peaks in curve (3) and curve (4) can be used to confirm crude oil fouling. However, the peak intensity at 1739 cm⁻¹ in curve (4) is much weaker than that in curve (3). This is because APAM fouling is dominant when taking SWCAO as the feed solution.¹⁴

The peak at 3327 cm⁻¹ is the characteristic absorption peak of N–H in an amide bond. NF90 used in this study, a kind of polyamide composite membrane, and APAM both possess many N–H. Thus the peak at 3327 cm⁻¹ is present in all four curves, but is particularly strong in curve (4), because APAM is the main foulant in the fouling layer on the FM. The peak at 3100 cm⁻¹, which corresponds to C–H aromatic stretching, comes from the polyamide and polysulfone material of the NF90 membrane.³² It appears in curve (1), (2) and (3), but disappears in curve (4). This is because the IR spectrum in the high wave number region cannot penetrate the relative thick fouling layer on the FM, which can cover the membrane material and shield the IR of C–H aromatic stretching. In addition, a new peak at 3206 cm⁻¹ appears in curve (3), which is probably resulting from the hydrogen bonds between APAM molecules and the membrane surface.¹⁴

Compared with curve (1), there are two relatively strong peaks at 2853 cm⁻¹ and 2918 cm⁻¹ in curve (2). Although these two peaks can be regarded as a sign of crude oil or SDS, here it should be attributed to the residual SDS on the membrane. This is because crude oil can decrease membrane hydrophilicity while SDS can increase membrane hydrophilicity. The results in Section 3.4.3 have indicated that membrane hydrophilicity increased after cleaning. Therefore, the two peaks at 2853 cm⁻¹ and 2918 cm⁻¹ in curve (2) should be related to the residual SDS on the CM, rather than crude oil. Besides, a new peak at 1044 cm⁻¹, which is related to the stretching vibration of OSO₃⁻ in the SDS molecule, appears in curve (2). This peak further confirms the residue of SDS on the CM. Apart from this fine distinction, curve (2) is very similar to curve (1), implying that the selected cleaning strategy could effectively eliminate the fouling layer without damaging the membrane chemical bonds.

4. Conclusion

In this study, a membrane cleaning strategy for advanced treatment of PFPW by NF has been studied by taking APAM and crude oil as target foulants. NaOH, EDTA, SDS, SPP and HCl were investigated and used in combination. Flux measurements, salt rejection experiments, AFM, a contact angle goniometer and ATR-FTIR were employed to further evaluate the cleaning performance from other multiple perspectives. The obtained results are concluded as follows: (1) the optimized alkaline formulated cleaning solution (pH = 11.0) consists of 0.05% EDTA, 0.2% sodium pyrophosphate and 0.2% SDS; (2) the optimized cleaning strategy includes two steps: the first step is cleaning with the alkaline formulated cleaning agent for 0.5 h and the second step is cleaning with a HCl solution (pH = 2.0) for 0.5 h; (3) the optimized cleaning strategy can fully recover the membrane flux without damaging its other properties, including desalting ability, surface morphology, hydrophilicity and infrared spectroscopy.

Acknowledgements

The authors gratefully acknowledge the financial support provided by Natural Science Foundation of China (NSFC, Grant no. 51578390) and the State Key Laboratory of Urban Water Resource and Environment (Harbin institute of technology, no. 2016DX11).

Notes and references

1. V. Alvarado and E. Manrique, Energies, 2010, 3, 1529–1575.
2. A. Z. Abidin, T. Puspasari and W. A. Nugroho, Procedia Chem., 2012, 4, 11–16.
3. D. A. Z. Wever, F. Picchioni and A. A. Broekhuis, Prog. Polym. Sci., 2011, 36, 1558–1628.
4. H. Guo, L. Xiao, S. Yu, H. Yang, J. Hu, G. Liu and Y. Tang, Desalination, 2014, 346, 46–53.
5. A. Fakhru’l-Razi, A. Pendashteh, L. C. Abdullah, D. R. Biak, S. S. Madaeni and Z. Z. Abidin, J. Hazard. Mater., 2009, 170, 530–551.
6. S. Alzahrani and A. W. Mohammad, J. Water Process Eng., 2014, 4, 107–133.
7. S. Alzahrani, A. W. Mohammad, N. Hilal, P. Abdullah and O. Jaafar, Desalination, 2013, 315, 27–32.
8. S. Alzahrani, A. W. Mohammad, N. Hilal, P. Abdullah and O. Jaafar, Desalination, 2013, 315, 18–26.
9. R. M. Abousnina and D. N. Long, Desalin. Water Treat., 2014, 52, 570–579.
10. A. Reyhani, F. Rekabdar, M. Hemmati, A. A. Safekordi and M. Ahmadi, Desalin. Water Treat., 2013, 51, 7499–7508.
11. M. Cakmakci, N. Kayaalp and I. Koyuncu, Desalination, 2008, 222, 176–186.
12. S. Mondal and S. R. Wickramasinghe, J. Membr. Sci., 2008, 322, 162–170.
13. B. Zhang, W. Shi, S. Yu, Y. Zhu, R. Zhang and L. Li, RSC Adv., 2015, 5, 104960–104971.
14. R. Zhang, W. Shi, S. Yu, W. Wang, Z. Zhang, B. Zhang, L. Li and X. Bao, Desalination, 2015, 27–37.
15. M. R. Hashmet, M. Onur and I. M. Tan, J. Dispers. Sci. Technol., 2014, 35, 510–517.
16. W. S. Ang, S. Lee and M. Elimelech, J. Membr. Sci., 2006, 272, 198–210.
17. A. Al-Amoudi and R. W. Lovitt, J. Membr. Sci., 2007, 303, 4–28.
18. N. Porcelli and S. Judd, Sep. Purif. Technol., 2010, 71, 137–143.
19. Z. Wang, J. Ma, C. Y. Tang, K. Kimura, Q. Wang and X. Han, J. Membr. Sci., 2014, 468, 276–307.
20. W. S. Ang, N. Y. Yip, A. Tiraferri and M. Elimelech, J. Membr. Sci., 2011, 382, 100–106.
21. N. M. D’Souza and A. J. Mawson, Crit. Rev. Food Sci. Nutr., 2005, 45, 125–134.
22. X. Yi, W. Shi, S. L. Yu, N. Sun, L. M. Jin, S. Wang, B. Zhang, C. Ma and L. P. Sun, Desalination, 2012, 286, 254–262.
23. K. Boussu, Y. Zhang, J. Cocquyt, P. Van der Meeren, A. Volodin, C. Van Haesendonck, J. A. Martens and B. Van der Bruggen, J. Membr. Sci., 2006, 278, 418–427.
24. A. Simon, J. A. McDonald, S. J. Khan, W. E. Price and L. D. Nghiem, J. Membr. Sci., 2013, 447, 153–162.
25. X. Pei, C. Bellona and J. Drewes, J. Membr. Sci., 2010, 353, 111–121.
26. A. Simon, W. E. Price and L. D. Nghiem, J. Membr. Sci., 2013, 432, 73–82.
27. Y. N. Kwon and J. O. Leckie, J. Membr. Sci., 2006, 283, 21–26.
28. Y. N. Kwon and J. O. Leckie, J. Membr. Sci., 2006, 282, 456–464.
29. K. Boussu, B. Van der Bruggen, A. Volodin, J. Snauwaert, C. Van Haesendonck and C. Vandecasteele, J. Colloid Interface Sci., 2005, 286, 632–638.
30. R. Liikanen, J. Yli-Kuivila and R. Laukkanen, J. Membr. Sci., 2002, 195, 265–276.
31. A. Al-Amoudi, P. Williams, A. S. Al-Hobaib and R. W. Lovitt, Appl. Surf. Sci., 2008, 254, 3983–3992.
32. P. S. Singh, S. V. Joshi, J. J. Trivedi, C. V. Devmurari, A. P. Rao and P. K. Ghosh, J. Membr. Sci., 2006, 278, 19–25.
33. C. Y. Tang, Y. N. Kwon and J. O. Leckie, Desalination, 2009, 242, 149–167.
34. L. V. Meléndez, A. Lache, J. A. Orrego-Ruiz, Z. Pachón and E. Mejía-Ospino, J. Pet. Sci. Eng., 2012, 90–91, 56–60.
35. X. Gao and J. Chorover, J. Colloid Interface Sci., 2010, 348, 167–176.

---