Separation and Recycling of Concentrated Heavy Metal Wastewater by Tube Membrane Distillation Integrated with Crystallization

Tube membrane distillation (MD) integrated with a crystallization method is used in this study for the concurrent productions of pure water and salt crystals from concentrated single and mixed system solutions. The effects of concentrated Zn2+ and Ni2+ on performance in terms of membrane flux, permeate conductivity, crystal recovery rates, and crystal grades are investigated. Preferred crystallization and co-crystallization determinations were performed for mixed solutions. The results revealed that membrane fluxes remained at 2.61 kg·m−2·h−1 and showed a sharp decline until the saturation increased to 1.38. Water yield conductivity was below 10 μs·cm−1. High concentrated zinc and nickel did not have a particular effect on the rejection of the membrane process. For the mixed solutions, membrane flux showed a sharp decrease due to the high saturation, while the conductivity of permeate remained below 10 μs·cm−1 during the whole process. Co-crystallization has been proven to be a better method due to the existence of the SO42− common-ion effect. Membrane fouling studies have suggested that the membrane has excellent resistance to fouling from highly concentrated solutions. The MD integrated with crystallization proves to be a promising technology for treating highly concentrated heavy metal solutions.


Introduction
The increase of heavy metal wastewater has not only strained already limited freshwater but also caused the waste of heavy metal resources all over the world [1]. At the same time, electroplating, electrolysis, and other industrial processes are generating highly concentrated zinc and nickel wastewater every year [2]. These solutions are characterized by their high acidity, strong corrosion capacity, and elevated metal concentrations. Nowadays, the industry uses precipitation as a low-cost and straightforward method to treat this wastewater, but unfortunately, it produces a large amount of high-risk sludge in the meantime [3,4]. The traditional evaporation crystallization method takes lots of energy and needs reactors with excellent resistance to corrosion [5]. Besides, the most desirable goal of substantial metal wastewater treatment is to achieve the recovery of valuable

Materials
The reagents used in this study were ZnSO 4 ·7H 2 O and NiSO 4 ·6H 2 O. All are analytical grade and purchased from Tianjin Guangfu Fine Chemical Research Institute, Tianjin, China. As it is shown in Table 1 [23], these reagents have different solubilities at 10 • C (cooling crystallization temperature) and 65 • C (heat side temperature of MD). According to this condition, it is possible to cool crystallization to obtain solid salt before MD is blocked by membrane fouling. Table 1. Solubility and corresponding mass fraction of ZnSO 4 and NiSO 4 at different temperatures [22].

Material
Temperature/ • C Solubility Mass Fraction/% For the mixed system, if one reagent reaches its saturation point at 65 • C, while the other is not saturated at 10 • C, it would thus be possible to separate one from the other. Otherwise, co-crystallization will be a better option. Without considering the common-ion effect, three different ratios of ZnSO 4 and NiSO 4 were selected, and three different methods were chosen to treat the mixed solutions in this study accordingly, as listed in Table 2. Table 2. Three different methods to treat mixed solutions.

Ratios of Zinc Sulfate and Nickle Sulfate Method
3:1 Preferential crystallization of Zinc sulfate 0.5:1 Preferential crystallization of Nickle sulfate 1.5:1 Mixed co-crystallization PTFE tube membranes were kindly supplied by Huzhou Sano Environmental Technology, China. The membrane module is a self-made air gap type component.

Membrane Characterizations
The diameter and thickness were measured using a Vernier caliper. The measurement of the advancing contact angle was performed using a dynamic contact angle meter (DCAT21, Data Physics, Filderstadt, Germany). The membrane nominal and maximum pore sizes were conducted with a capillary flow parameter (Proulx 1000, Porometer NV, Nazareth, Belgium). The mechanical properties of the tube membrane were characterized in terms of strain and tensile stress at break using an electronic fabric strength tester (YG065CQS/pc, Changzhou No.1 Textile Equipment Co., Ltd., Changzhou, China). Table 3 shows the parameters of the membrane material and module.

MD Integrated with Crystallization Experiments
MD integrated with crystallization experiments were carried out by a laboratory-scale setup. The experimental apparatus diagram of air gap membrane distillation is shown in Figure 1. The membrane modules were self-made, and there were tube glasses, tube membranes, and tube condensers from the outside to the inside. During the MD process, hot feed flows between the membrane and the glass, and cold water flows inside the condense tube. There was not a certain width of the gap in this case. A 1000 mL beaker flask was used as a cooling mold. module driven by a magnetic force pump and recirculates on the module's hot side. The same procedure is followed for the chilled water system. The vapor passing through the membrane is condensed and recovered on the cold side. As the experiment goes on, when the feed solution is supersaturated, the concentrated solution is pumped out to the mold and cooled to crystallize. Finally, crystallization products are obtained by solid-liquid separation. The remaining crystalline solution returns to the MD unit and continues to concentrate. According to previous experiments, the optimized average operating conditions are as follows: temperature of the hot side, 65 °C; temperature of the cold side, 20 °C; velocity of the hot side, 0.16 m·s −1 ; velocity of the cold side, 0.26 m·s −1 ; temperature of the cooling crystallization, 10 °C; time of cooling crystallization, 8 h.

MD Integrated with Crystallization Characterizations
The obtained membrane flux is one of the most critical characterizations in the MD process [24,25]. It can be calculated using Equation (1), where V * ρ is permeated mass, A is the active membrane area, and t is the operating time. In this study, a 100 mL measuring cylinder was used to measure the volume of permeate. Rejection rate is also an essential characterization of operation stability in the MD process [26][27][28]. The rejection rate was indicated by permeate conductivity in this study, which was measured using a conductivity meter (S470 Seven Excellence, Mettler Toledo, Switzerland). The lower the permeate conductivity, the higher the reject rate. * / * The crystallization product is characterized by a recovery rate and grade, which can be calculated using Equations (2) and (3), respectively, where m is the weight of the target crystal product, M is the total weight of the target product, and p is the total weight of crystal product.
The crystals obtained from the single system were observed by stereomicroscope (XZT-CT, Shanghai Optics, Shanghai, China). Obtained mix crystals and membranes were determined using a Prepared feed solution is heated by a thermostatic hot bath, then flows through the membrane module driven by a magnetic force pump and recirculates on the module's hot side. The same procedure is followed for the chilled water system. The vapor passing through the membrane is condensed and recovered on the cold side. As the experiment goes on, when the feed solution is supersaturated, the concentrated solution is pumped out to the mold and cooled to crystallize. Finally, crystallization products are obtained by solid-liquid separation. The remaining crystalline solution returns to the MD unit and continues to concentrate.
According to previous experiments, the optimized average operating conditions are as follows: temperature of the hot side, 65 • C; temperature of the cold side, 20 • C; velocity of the hot side, 0.16 m·s −1 ; velocity of the cold side, 0.26 m·s −1 ; temperature of the cooling crystallization, 10 • C; time of cooling crystallization, 8 h.

MD Integrated with Crystallization Characterizations
The obtained membrane flux is one of the most critical characterizations in the MD process [24,25]. It can be calculated using Equation (1), where V * ρ is permeated mass, A is the active membrane area, and t is the operating time. In this study, a 100 mL measuring cylinder was used to measure the volume of permeate. Rejection rate is also an essential characterization of operation stability in the MD process [26][27][28]. The rejection rate was indicated by permeate conductivity in this study, which was measured using a conductivity meter (S470 Seven Excellence, Mettler Toledo, Switzerland). The lower the permeate conductivity, the higher the reject rate.
The crystallization product is characterized by a recovery rate and grade, which can be calculated using Equations (2) and (3), respectively, where m is the weight of the target crystal product, M is the total weight of the target product, and p is the total weight of crystal product.

of 13
The crystals obtained from the single system were observed by stereomicroscope (XZT-CT, Shanghai Optics, Shanghai, China). Obtained mix crystals and membranes were determined using a field emission scanning electron microscope (FE-SEM, SUPRA55, Carl Zeiss AG, Oberkochen, Germany) and an energy-dispersive X-ray spectroscopy (EDS, Oxford INCA x-act, High Wycombe, UK).

Single System Crystallization by MD Integrated with Crystallization
To investigate the feasibility of the MD integrated with crystallization process and to compare the effect of zinc ions with nickel ions, 20 wt % ZnSO 4 solution and 20 wt % NiSO 4 solution were used separately. The correspondences of saturation with mass score of ZnSO 4 and NiSO 4 at 65 • C are shown in Supplementary Tables S1 and S2. The variations in membrane flux and water yield conductivity as functions of saturation are represented in Figure 2. As can be seen, when the saturation was lower than 1 (42.64 wt % for ZnSO 4 and 36.64 wt % for NiSO 4 ), the fluxes of both solutions kept at 2.61 kg·m −2 ·h −1 . When the saturation was higher than 1 and lower than 1.38 (50.63 wt % for ZnSO 4 and 44.39 wt % for NiSO 4 ), the fluxes showed a slow decline. Finally, when the saturation was higher than 1.38, the fluxes decreased sharply. During the whole MD operation, the water yield conductivity of both solutions remained below 10 µs·cm −1 . Moreover, the rejection rates of heavy metal salt reached 99.99%.

Single System Crystallization by MD Integrated with Crystallization
To investigate the feasibility of the MD integrated with crystallization process and to compare the effect of zinc ions with nickel ions, 20 wt % ZnSO4 solution and 20 wt % NiSO4 solution were used separately. The correspondences of saturation with mass score of ZnSO4 and NiSO4 at 65 °C are shown in Supplementary Tables S1 and S2. The variations in membrane flux and water yield conductivity as functions of saturation are represented in Figure 2. As can be seen, when the saturation was lower than 1 (42.64 wt % for ZnSO4 and 36.64 wt % for NiSO4), the fluxes of both solutions kept at 2.61 kg·m −2 ·h −1 . When the saturation was higher than 1 and lower than 1.38 (50.63 wt % for ZnSO4 and 44.39 wt % for NiSO4), the fluxes showed a slow decline. Finally, when the saturation was higher than 1.38, the fluxes decreased sharply. During the whole MD operation, the water yield conductivity of both solutions remained below 10 μs·cm −1 . Moreover, the rejection rates of heavy metal salt reached 99.99%. Besides, the experiments involving ZnSO4 and NiSO4 solutions in air gap membrane distillation were consistent. Highly concentrated zinc and nickel ions did not have a different addition effect on the performance of MD, except that membrane fluxes were slightly lower than that of diluted solutions. Therefore, for the zinc and nickel mixed solution, it is possible to selectively separate the valuable metals according to their solubility differences at different temperatures.
After cooling crystallization at 10 °C, both the zinc and nickel sulfate crystalline products were obtained. According to crystallization chemistry, when the cooling temperature of the zinc sulfate solution is below 39 °C, and for the nickel sulfate solution, below 31.5 °C, the crystalline products will be ZnSO4·7H2O [29] and NiSO4·7H2O [30], respectively. In the view of the existence of carbonate and hydroxide ions (from the air and water) in the solution, the possible formation of carbonate and hydroxide precipitations is discussed in the supplemental information. The results indicate that it is impossible to form carbonate and hydroxide precipitations under the condition of the MD integrated with crystallization experiments. Obtained solid products were observed by stereomicroscope and compared with commercial products, respectively, as shown in Figure 3. The crystal products of zinc sulfate obtained from the MD integrated with crystallization process were colorless and needle-like, belonging to the orthorhombic system. The average size was 4 mm. Furthermore, its shape was more Besides, the experiments involving ZnSO 4 and NiSO 4 solutions in air gap membrane distillation were consistent. Highly concentrated zinc and nickel ions did not have a different addition effect on the performance of MD, except that membrane fluxes were slightly lower than that of diluted solutions. Therefore, for the zinc and nickel mixed solution, it is possible to selectively separate the valuable metals according to their solubility differences at different temperatures.
After cooling crystallization at 10 • C, both the zinc and nickel sulfate crystalline products were obtained. According to crystallization chemistry, when the cooling temperature of the zinc sulfate solution is below 39 • C, and for the nickel sulfate solution, below 31.5 • C, the crystalline products will be ZnSO 4 ·7H 2 O [29] and NiSO 4 ·7H 2 O [30], respectively. In the view of the existence of carbonate and hydroxide ions (from the air and water) in the solution, the possible formation of carbonate and hydroxide precipitations is discussed in the supplemental information. The results indicate that it is impossible to form carbonate and hydroxide precipitations under the condition of the MD integrated with crystallization experiments. Obtained solid products were observed by stereomicroscope and compared with commercial products, respectively, as shown in Figure 3. The crystal products of zinc sulfate obtained from the MD integrated with crystallization process were colorless and needle-like, belonging to the orthorhombic system. The average size was 4 mm. Furthermore, its shape was more regular than the observed commercial products. On the other hand, the obtained nickel sulfate crystals were green, transparent, and belonged to the orthorhombic system. Their average size was 2 mm, and shapes were quite different from commercial NiSO 4 ·6H 2 O, which was used to prepare the initial solution.
Membranes 2020, 10, x FOR PEER REVIEW 6 of 13 regular than the observed commercial products. On the other hand, the obtained nickel sulfate crystals were green, transparent, and belonged to the orthorhombic system. Their average size was 2 mm, and shapes were quite different from commercial NiSO4·6H2O, which was used to prepare the initial solution.

Preferential Crystallization of Zinc Sulfate by MD Integrated with Crystallization
In this experiment, 21.43 wt % ZnSO4 and 7.14 wt % NiSO4 mixed solution was used. MD integrated with crystallization experiments were performed to separate and obtain zinc sulfate crystal products selectively. As can be seen from Figure 4, membrane flux has an apparent decreasing trend with the increase of saturation. Additionally, when the saturation was higher than 1.15, membrane flux decreased rapidly. At the end of the process, when the saturation was 1.28, the flux was 0.38 kg·m −2 ·h −1 . The membrane flux drop was probably caused by the combined action of solution vapor pressure drop, concentration polarization, temperature polarization, and membrane fouling [31][32][33]. During the whole MD process, the conductivity of permeate remained below 10 μs·cm −1 . The rejection rate of heavy metal salt reached 99.99%. After cooling crystallization at 10 °C, a mixture of zinc sulfate and nickel sulfate crystalline products were obtained. Figure 5 shows that the recovery rate and grade of ZnSO4·7H2O crystals

Preferential Crystallization of Zinc Sulfate by MD Integrated with Crystallization
In this experiment, 21.43 wt % ZnSO 4 and 7.14 wt % NiSO 4 mixed solution was used. MD integrated with crystallization experiments were performed to separate and obtain zinc sulfate crystal products selectively. As can be seen from Figure 4, membrane flux has an apparent decreasing trend with the increase of saturation. Additionally, when the saturation was higher than 1.15, membrane flux decreased rapidly. At the end of the process, when the saturation was 1.28, the flux was 0.38 kg·m −2 ·h −1 . The membrane flux drop was probably caused by the combined action of solution vapor pressure drop, concentration polarization, temperature polarization, and membrane fouling [31][32][33]. During the whole MD process, the conductivity of permeate remained below 10 µs·cm −1 . The rejection rate of heavy metal salt reached 99.99%.
Membranes 2020, 10, x FOR PEER REVIEW 6 of 13 regular than the observed commercial products. On the other hand, the obtained nickel sulfate crystals were green, transparent, and belonged to the orthorhombic system. Their average size was 2 mm, and shapes were quite different from commercial NiSO4·6H2O, which was used to prepare the initial solution.

Preferential Crystallization of Zinc Sulfate by MD Integrated with Crystallization
In this experiment, 21.43 wt % ZnSO4 and 7.14 wt % NiSO4 mixed solution was used. MD integrated with crystallization experiments were performed to separate and obtain zinc sulfate crystal products selectively. As can be seen from Figure 4, membrane flux has an apparent decreasing trend with the increase of saturation. Additionally, when the saturation was higher than 1.15, membrane flux decreased rapidly. At the end of the process, when the saturation was 1.28, the flux was 0.38 kg·m −2 ·h −1 . The membrane flux drop was probably caused by the combined action of solution vapor pressure drop, concentration polarization, temperature polarization, and membrane fouling [31][32][33]. During the whole MD process, the conductivity of permeate remained below 10 μs·cm −1 . The rejection rate of heavy metal salt reached 99.99%. After cooling crystallization at 10 °C, a mixture of zinc sulfate and nickel sulfate crystalline products were obtained. Figure 5 shows that the recovery rate and grade of ZnSO4·7H2O crystals After cooling crystallization at 10 • C, a mixture of zinc sulfate and nickel sulfate crystalline products were obtained. Figure 5 shows that the recovery rate and grade of ZnSO 4 ·7H 2 O crystals varied at different saturation values during the MD integrated with crystallization process. As can be seen, with the increase of saturation, the recovery rate increased linearly. When the supersaturation was 1.18, the membrane flux was 1.5 kg·m −2 ·h −1 , the recovery rate of zinc sulfate reached 35%. Nevertheless, the grade of crystals was about 70% and changed little at different saturations.
Membranes 2020, 10, x FOR PEER REVIEW 7 of 13 varied at different saturation values during the MD integrated with crystallization process. As can be seen, with the increase of saturation, the recovery rate increased linearly. When the supersaturation was 1.18, the membrane flux was 1.5 kg·m −2 ·h −1 , the recovery rate of zinc sulfate reached 35%. Nevertheless, the grade of crystals was about 70% and changed little at different saturations. After drying, the obtained crystal products were observed under SEM and measured by EDS. The results are shown in Figure 6. The composition of the main elements of crystals was 29.52% O, 33.80% S, 29.44% Zn, and 7.23% Ni. Both the results obtained from ICP-OES and EDS analysis suggest that this process does not have a zinc sulfate preferential crystallization due to the common-ion effect of SO4 2− .

Preferential Crystallization of Nickel Sulfate by MD Integrated with Crystallization
The mass fractions of 10.34 wt % ZnSO4 and 20.69 wt % NiSO4 mixed solution were used to selectively separate and obtain nickel sulfate crystal products. As can be seen from Figures 7 and 8, the preferential separation of NiSO4 showed the same tendency as the preferred separation of ZnSO4. The conductivity of permeate remained below 10 μs·cm −1 during the MD process, but the membrane flux showed a decline when the saturation increased. The recovery rate rose linearly with the increase of saturation. When the saturation was 1.18, the membrane flux was 1.5 kg·m −2 ·h −1 , and the recovery After drying, the obtained crystal products were observed under SEM and measured by EDS. The results are shown in Figure 6. The composition of the main elements of crystals was 29.52% O, 33.80% S, 29.44% Zn, and 7.23% Ni. Both the results obtained from ICP-OES and EDS analysis suggest that this process does not have a zinc sulfate preferential crystallization due to the common-ion effect of SO 4 2− .
Membranes 2020, 10, x FOR PEER REVIEW 7 of 13 varied at different saturation values during the MD integrated with crystallization process. As can be seen, with the increase of saturation, the recovery rate increased linearly. When the supersaturation was 1.18, the membrane flux was 1.5 kg·m −2 ·h −1 , the recovery rate of zinc sulfate reached 35%. Nevertheless, the grade of crystals was about 70% and changed little at different saturations. After drying, the obtained crystal products were observed under SEM and measured by EDS. The results are shown in Figure 6. The composition of the main elements of crystals was 29.52% O, 33.80% S, 29.44% Zn, and 7.23% Ni. Both the results obtained from ICP-OES and EDS analysis suggest that this process does not have a zinc sulfate preferential crystallization due to the common-ion effect of SO4 2− .

Preferential Crystallization of Nickel Sulfate by MD Integrated with Crystallization
The mass fractions of 10.34 wt % ZnSO4 and 20.69 wt % NiSO4 mixed solution were used to selectively separate and obtain nickel sulfate crystal products. As can be seen from Figures 7 and 8, the preferential separation of NiSO4 showed the same tendency as the preferred separation of ZnSO4. The conductivity of permeate remained below 10 μs·cm −1 during the MD process, but the membrane flux showed a decline when the saturation increased. The recovery rate rose linearly with the increase of saturation. When the saturation was 1.18, the membrane flux was 1.5 kg·m −2 ·h −1 , and the recovery

Preferential Crystallization of Nickel Sulfate by MD Integrated with Crystallization
The mass fractions of 10.34 wt % ZnSO 4 and 20.69 wt % NiSO 4 mixed solution were used to selectively separate and obtain nickel sulfate crystal products. As can be seen from Figures 7 and 8, the preferential separation of NiSO 4 showed the same tendency as the preferred separation of ZnSO 4 . The conductivity of permeate remained below 10 µs·cm −1 during the MD process, but the membrane flux showed a decline when the saturation increased. The recovery rate rose linearly with the increase of saturation. When the saturation was 1.18, the membrane flux was 1.5 kg·m −2 ·h −1 , and the recovery rate was 35% by primary MD integrated with crystallization operating. However, the grade of crystals achieved 72% and changed little at different saturations.
Membranes 2020, 10, x FOR PEER REVIEW 8 of 13 rate was 35% by primary MD integrated with crystallization operating. However, the grade of crystals achieved 72% and changed little at different saturations.  The obtained crystal products were also observed under SEM and measured by EDS. The results are shown in Figure 9. The main element composition of crystals was 12.57% O, 26.07% S, 6.52% Zn, and 54.85% Ni. The results showed that preferential crystallization of NiSO4 by MD integrated with crystallization was feasible while there was still 7.23% Ni in the ZnSO4·7H2O and NiSO4·7H2O salt products.   The obtained crystal products were also observed under SEM and measured by EDS. The results are shown in Figure 9. The main element composition of crystals was 12.57% O, 26.07% S, 6.52% Zn, and 54.85% Ni. The results showed that preferential crystallization of NiSO4 by MD integrated with crystallization was feasible while there was still 7.23% Ni in the ZnSO4·7H2O and NiSO4·7H2O salt products. The obtained crystal products were also observed under SEM and measured by EDS. The results are shown in Figure 9. The main element composition of crystals was 12.57% O, 26.07% S, 6.52% Zn, and 54.85% Ni. The results showed that preferential crystallization of NiSO 4 by MD integrated with crystallization was feasible while there was still 7.23% Ni in the ZnSO 4 ·7H 2 O and NiSO 4 ·7H 2 O salt products. Membranes 2020, 10, x FOR PEER REVIEW 9 of 13 (a) (b)

Co-Crystallization of Zinc and Nickel Sulfate by MD Integrated with Crystallization
In these experiments, a ZnSO4 and NiSO4 mixed solution was used to investigate the cocrystallization. The mass fractions of ZnSO4 and NiSO4 were 20.00 wt % and 13.33 wt %, respectively. The variations in membrane flux and water yield conductivity as functions of filtration time are illustrated in Figure 10. It can be seen from the figure, as the experiment goes on, the membrane flux reduces continuously. Six hours later, the membrane flux decreased rapidly, and the conductivity of permeate increased. Furthermore, after 12 h of operation, the flux was 0.38 kg·m −2 ·h −1 . However, during the whole MD process, the conductivity of permeate remained below 20 μs·cm −1 . The rejection rate of heavy metal salt still reached 99.99%. After natural drying, the crystals were dissolved in distilled water and then analyzed utilizing an ICP-OES. Among the products, the mass fraction of ZnSO4·7H2O was 64.96%, and the mass fraction of NiSO4·7H2O was 35.04%.

Membrane Fouling
As mentioned before, during the MD integrated with crystallization process, a decrease in the membrane flux was observed. After washing the system with tap and MILIQ water, membrane flux returned to its initial value and permeated conductivity again remained below 10 μs·cm −1 . However,

Co-Crystallization of Zinc and Nickel Sulfate by MD Integrated with Crystallization
In these experiments, a ZnSO 4 and NiSO 4 mixed solution was used to investigate the co-crystallization. The mass fractions of ZnSO 4 and NiSO 4 were 20.00 wt % and 13.33 wt %, respectively. The variations in membrane flux and water yield conductivity as functions of filtration time are illustrated in Figure 10.

Co-Crystallization of Zinc and Nickel Sulfate by MD Integrated with Crystallization
In these experiments, a ZnSO4 and NiSO4 mixed solution was used to investigate the cocrystallization. The mass fractions of ZnSO4 and NiSO4 were 20.00 wt % and 13.33 wt %, respectively. The variations in membrane flux and water yield conductivity as functions of filtration time are illustrated in Figure 10. It can be seen from the figure, as the experiment goes on, the membrane flux reduces continuously. Six hours later, the membrane flux decreased rapidly, and the conductivity of permeate increased. Furthermore, after 12 h of operation, the flux was 0.38 kg·m −2 ·h −1 . However, during the whole MD process, the conductivity of permeate remained below 20 μs·cm −1 . The rejection rate of heavy metal salt still reached 99.99%. After natural drying, the crystals were dissolved in distilled water and then analyzed utilizing an ICP-OES. Among the products, the mass fraction of ZnSO4·7H2O was 64.96%, and the mass fraction of NiSO4·7H2O was 35.04%.

Membrane Fouling
As mentioned before, during the MD integrated with crystallization process, a decrease in the membrane flux was observed. After washing the system with tap and MILIQ water, membrane flux returned to its initial value and permeated conductivity again remained below 10 μs·cm −1 . However, It can be seen from the figure, as the experiment goes on, the membrane flux reduces continuously. Six hours later, the membrane flux decreased rapidly, and the conductivity of permeate increased. Furthermore, after 12 h of operation, the flux was 0.38 kg·m −2 ·h −1 . However, during the whole MD process, the conductivity of permeate remained below 20 µs·cm −1 . The rejection rate of heavy metal salt still reached 99.99%. After natural drying, the crystals were dissolved in distilled water and then analyzed utilizing an ICP-OES. Among the products, the mass fraction of ZnSO 4 ·7H 2 O was 64.96%, and the mass fraction of NiSO 4 ·7H 2 O was 35.04%.

Membrane Fouling
As mentioned before, during the MD integrated with crystallization process, a decrease in the membrane flux was observed. After washing the system with tap and MILIQ water, membrane flux returned to its initial value and permeated conductivity again remained below 10 µs·cm −1 . However, as an essential indicator of a comprehensive and in-depth evaluation of the membrane process [34][35][36], membrane surface morphology and surface deposits were investigated in this study. According to the differences in the treated liquids, tube membranes were labeled, as shown in Table 4. The working time of each membrane was calculated. The outer surface, inner surface, and cross-section of the membrane were observed by SEM, and the composition of the potential sediments was analyzed by EDS. In particular, to investigate how deeply the salts penetrate the pores of the membranes, EDS liner analysis was used instead of point analysis. In order to get neat cross-section samples, membranes were pre-treated with liquid nitrogen.As can be seen from Figure 11, the membranes used in this study have a density network structure. A small number of deposits can be seen on the outer surfaces of membrane 1 and membrane 3. This is mainly because of the highly concentrated solutions directly contacted with the outer surfaces. However, the network structure of the membrane surfaces can still be seen clearly. The deposition phenomenon on the membrane surfaces was not significant. Furthermore, there were no deposits found on the outer surfaces of membranes 0 and 1, nor the inner surfaces or cross-sections of all the membranes. The results of EDS liner analysis indicated that the deposits existed only on the membrane outer surfaces and did not penetrate deep into the membranes. Element and content analysis of the deposits on the outer membrane surface is shown in Table 5. The chemical formula of the PTFE membrane is (C 2 F 4 ) n . Apart from the C and F elements, the amounts of sediment elements such as Zn, Ni, S, and O were relatively small on the surface of the membrane. All this proved that the membrane has excellent resistance to the fouling of highly concentrated zinc and nickel solutions. as an essential indicator of a comprehensive and in-depth evaluation of the membrane process [34][35][36], membrane surface morphology and surface deposits were investigated in this study. According to the differences in the treated liquids, tube membranes were labeled, as shown in Table 4. The working time of each membrane was calculated. The outer surface, inner surface, and cross-section of the membrane were observed by SEM, and the composition of the potential sediments was analyzed by EDS. In particular, to investigate how deeply the salts penetrate the pores of the membranes, EDS liner analysis was used instead of point analysis. In order to get neat cross-section samples, membranes were pre-treated with liquid nitrogen.As can be seen from Figure 11, the membranes used in this study have a density network structure. A small number of deposits can be seen on the outer surfaces of membrane 1 and membrane 3. This is mainly because of the highly concentrated solutions directly contacted with the outer surfaces. However, the network structure of the membrane surfaces can still be seen clearly. The deposition phenomenon on the membrane surfaces was not significant. Furthermore, there were no deposits found on the outer surfaces of membranes 0 and 1, nor the inner surfaces or cross-sections of all the membranes. The results of EDS liner analysis indicated that the deposits existed only on the membrane outer surfaces and did not penetrate deep into the membranes. Element and content analysis of the deposits on the outer membrane surface is shown in Table 5. The chemical formula of the PTFE membrane is (C2F4)n. Apart from the C and F elements, the amounts of sediment elements such as Zn, Ni, S, and O were relatively small on the surface of the membrane. All this proved that the membrane has excellent resistance to the fouling of highly concentrated zinc and nickel solutions.   (1) zinc sulfate system; (2) nickel sulfate system; (3) mixed system.

Conclusions
The MD integrated with crystallization process has been confirmed to permit pure water and crystal products to be obtained from highly concentrated zinc and nickel mixed solutions. For the single element system, membrane fluxes remained at 2.61 kg·m −2 ·h −1 and showed a sharp decline until the saturation increased to 1.38. Water yield conductivity was maintained below 10 µs·cm −1 . Highly concentrated zinc and nickel ions did not have extra influence on the membrane. For the mixed solutions, membrane flux showed a sharp decrease as an increase in saturation was observed, while the conductivity of permeate remained under 10 µs·cm −1 during the whole process. In this study, co-crystallization proves to be a better method due to the existence of the SO 4 2− common-ion effect.
Furthermore, membrane fouling studies suggest that the membrane has excellent resistance to the fouling of highly concentrated solutions. The MD integrated with crystallization process has been demonstrated to be a promising technology in treating highly concentrated heavy metal solutions. However, nowadays, this technology is not used in the industry. Currently, one of the barriers to the implementation of MD integrated with crystallization is the price of the commercial membranes. It can be predicted that with the progress of membrane technology and materials science, MD integrated with crystallization technology will be widely used in the industry.