GDU membrane performance of humic acid as feed solution
To determine the performance of the GDU process, permeation flux of HA with different concentrations were evaluated. Membrane fouling was aggravated with the increased of HA concentration which led to lower flux and poorer membrane rejection. Fig. 2 shows that the average membrane rejection (triplicate results) declined from 94.96–90.39% as the initial concentration of HA (mg/L) increased from 10 mg/L to 50 mg/L, indicating that HA penetration was taken placed and its permeation was driven by the concentration gradient. It was found that membrane fouling via surface adsorption had caused the considerable HA permeation as indicated by the reduction of rejection capability.
Based on Fig. 3, the highest flux produced by HA solution was 2.07 kg/m2s with initial humic acid concentration of 10 mg/l. The system took about 400 s to totally wet the membrane pores. As mentioned earlier, flux reduction was due to the combined effects of fouling and depleting hydrostatic pressure. Flux reduction could be contributed by both reversible (concentration polarization) and irreversible (adsorption and deposition) fouling. Concentration polarization resulted in flux drop due to the reduction in the effective pressure (driving force). Concentration polarization become more crucial for dissolved solute separation as osmotic pressure is concentration dependence (Taheri et al. 2013). However, in this case, concentration polarization is not the controlling parameter for UF process. Instead, cake layer deposition is the dominating flux reduction factor. The membrane-filled module required 60 minutes to empty the whole solution and it is almost independent of the concentration of HA being used. This showed that concentration polarization was not the main cause of flux declination. However, it was also found out that the onset of fouling is proportional to the initial feed concentrations of HA. The higher the HA concentrations, the faster the onset of the fouling. Another interesting phenomenon observed from the graph was the initial declination slopes were almost similar which showed that the physical characteristic of the HA cake layer is quite similar under such a low-pressure operation.
The retentate solution concentrations are shown in Table 1. For the initial 10 mg/L of HA, the final retentate being concentrated to 69.74 mg/L which was about 7 times increased in feed concentration. Same goes to the initial feed concentrations of 20, 30 and 40 mg/L of humic acid. The concentration of retentate solution (50 mg/L) solution increased approximately 8 times to 378.90 mg/L compared to the initial feed concentration at the end of experiment. The HA was rejected via retention as well as surface adsorption. In that case, at higher HA feed concentration, the tendency for the solution to retain rather than adsorb on the membrane surface is more prevalent. It is because at higher initial feed concentration, the solubility of the HA was getting lower and it tends to appear as bigger aggregated molecule in the bulk solution (Rucka et al. 2019). In overall, the results showed that the HF membrane could retain the HA well even under gravitational driving force.
Table 1
Concentrated HA retained in the membrane module
Initial Feed Concentration (mg/L)
|
Retentate solution concentration (mg/L)
|
Concentrating Factor
|
10.0
|
69.74 ± 0.12
|
6.97
|
20.0
|
134.17 ± 0.46
|
6.71
|
30.0
|
197.97 ± 0.46
|
6.60
|
40.0
|
259.33 ± 0.39
|
6.67
|
50.0
|
378.90 ± 0.30
|
7.58
|
The size distribution of HA was measured using Zetasizer nano series instrument under room temperature. The mean apparent size distribution of aggregated HA was 2.443 µm. The aggregated HA cannot pass through the membrane pores with smaller pore size (0.03 micron) due to size exclusion. However, the dissolved humic substances have molecular weight around 1 kDa which could easily permeate through the membrane. Thus, dissolved HA could not be rejected effectively by the membrane (Jeong et al. 2014).
GDU membrane performance of Calcium Carbonate as feed solution
Figure 4 shows that the rejection of CaCO3 was higher than 97% with the initial feed concentration lower than 0.5 wt%. The rejection capability of the membrane reduced from 99.19 to 97.41% when the concentration of CaCO3 was increased from 0.1 to 0.5 wt %. Overall, its rejection capability towards calcium carbonate was higher than humic acid. The same separation principle applied whereby the higher the concentration gradient, the higher the mass transfer of the solute across the membrane. The higher the CaCO3 concentration, the tendency for it to scale the membrane become higher (Wang et al. 2013). The percentage of removal showed that the rejection of CaCO3 is significantly reduced if its concentration was increased.
Figure 5 shows that the onset of scaling by CaCO3 solution was also concentration dependent. The higher the CaCO3 concentration, the onset of scaling would be faster and subsequently the flux declined faster after the onset point. This mean that the physical characteristic of the cake layer is also concentration dependent. The higher the concentration of the CaCO3, the denser of the cake layer. The formation of cake layer reduced the membrane flux over the time. Figure 5 shows that the highest flux of 2.81 L/m2.hr was obtained when 0.1 wt% CaCO3 solution was filtered. Meanwhile, the lowest flux was recorded as 2.31 L/m2hr when 0.5 wt % CaCO3 was filtered. Nonetheless, the fouling of CaCO3 is not as serious as humic substance due to its different cake layer characteristic. CaCO3 crystallization could be initiated on the membrane surface or in the bulk solution followed by surface deposition. However, its deposition characteristic could be different for such a low-pressure environment. Based on the zeta sizer result, the mean size distribution of CaCO3 was 4.793 µm which is much larger compare to HA aggregate. In view of this, it was deduced that CaCO3 precipitation had led to a thicker but looser cake layer compared to humic acid. The concentrated calcium carbonate retained in the membrane module is shown in Table 2. The CaCO3 at 0.5 wt% showed a drastic increased of concentrating factor which indicated that the early onset of fouling prevents the removal of CaCO3.
Table 2
Concentrated CaCO3 retained in the membrane module
Initial Feed Concentration (wt%)
|
Retentate solution concentration (wt%)
|
Concentrating Factor
|
0.1
|
4.77 ± 0.08
|
47.67
|
0.2
|
10.85 ± 0.10
|
54.23
|
0.3
|
11.65 ± 0.10
|
38.83
|
0.4
|
17.78 ± 0.20
|
44.44
|
0.5
|
61.96 ± 0.40
|
123.92
|
Surface fouling analysis
Scanning electron microscope (SEM) is a common technique used for characterizing the morphology of membranes before and after fouling. Fig. 6 shows that after the humic substance filtration, some sparingly soluble salts were found deposited on the surface of the hollow fibre membrane. It was loosely packed and dispersive, and the colour of the membrane surface became brown (observed using naked eye). The SEM-EDX analysis of the surface showed that 7.73 wt% of Oxygen (O), 0.91 wt% of Silicates (Si), 1.01 wt% Magnesium (Mg) and 0.43 wt% Aluminium (Al) were the deposit materials associated with the humic acid structure and its associated components. Humic substance was more favourably adsorbed onto hydrophobic membranes due to the hydrophobic interaction. Similar to the CaCO3 precipitation, HA deposition can be initiated on the surface by the free humic substance or precipitated from the bulk solution (Mitrouli et al. 2016).
Figure 7 shows the SEM image of CaCO3 precipitation on the hollow fibre PVDF membrane surface after filtration. The precipitated CaCO3 led to the loose scaling layer due to the absence of high pressure. From the SEM-EDX analysis, the deposit on the membrane surface was mainly Calcium (Ca) and Silicates (Si) consists of 5.55 wt% and 2.27 wt% respectively. These irregular structures are the amorphous CaCO3 particles, while growing, tend to be transformed into the stable regular crystals over the time (Mitrouli et al. 2016; Mitrouli et al. 2016).
The rejection performance of the membranes was accordance to the particle size of the foulants. The membrane gave the highest rejection towards CaCO3 and followed by humic substance according to the size of the precipitate. Furthermore, it is important to note that the higher permeation velocity could lead to solute concentrating at membrane surface that favours nucleation and crystal growth (Mitrouli et al. 2016). Under such a low-pressure filtration, cake layer formation is favoured over the pore blocking which indicated that the fouling could be reversible (Xu et al. 2012).
Water Contact Angle Analysis
Surface hydrophilicity is one of the most important properties of membrane which could affect the flux and antifouling ability of the membrane (Kim et al. 2003; Lu et al. 2008; Oh et al. 2009). Contact angle value also indicates the membrane material affinity to water. As shown in Table 3, the contact angle value of the membrane surface increased significantly with different type of foulants. The increased contact angle value was due to the increased of surface roughness after surface precipitation. The contact angle of pristine PVDF membrane was 67.20˚ ± 1.09, after HA precipitation, the contact angle increased to 81.88˚ ± 0.74. Similarly, after ultrafiltration of CaCO3, the contact angle was increased to 87.47˚ ± 0.72. The heterogeneous nucleation and growth processes on the of calcite materials was the dominating structure during the incipient scaling process (Vela et al. 2008). increased of contact angle upon fouling is not favorable as it requires to overcome the extra capillary force of water permeating the pores.
Table 3
Contact angle value of DGU membrane towards different feeds solution
Hollow Fibre (HF) Membrane
|
Contact Angle (˚)
|
Pristine HF PVDF membrane
|
67.20 ± 1.09
|
HF membrane after filtration of Humic Acid
|
81.88 ± 0.74
|
HF membrane after filtration of Calcium Carbonate
|
87.47 ± 0.72
|
Cake layer resistance analysis
Table 4 displays the variation of total resistance under varying concentration of the feed solutions under gravitational driven membrane filtration. After running the filtration using two different feed solution, membrane resistance, Rm was estimated as 2.17x1010 m2/kg which is close to the value reported by Piry et al., with the membrane resistance, Rm ranging from 7.0x1010 to 8.9x1010. However, in their study, the flux resistance was barely affected by the deposit layer formation. It was found that the cake resistance (Rc) decreased under higher transmembrane pressure which means that some of the substances causing pressure-independent fouling are washed away (Piry et al. 2012).
Membrane fouling is started from the foulant deposition followed by cake layer build-up (Shao et al. 2017), the cake fouling resistance, Rc accounted for 76.4 to 99.1% of the total fouling resistance. The higher ratio of cake fouling resistance suggested that the characteristic of cake layer was playing the role in determining the flux of the GDU system. This finding is detrimental to the GDU system with poor hydrodynamic condition. According to Di Profio et al., they observed that cake layer forms on the membrane surface result in Rc value of (x 1012) for seawater filtration (Di Profio et al. 2011). and Listiarini et al. reported that the Rc value (x 1013) for humic acid, bromide and bromate filtration (Listiarini et al. 2010). In this work, the cake layer resistance had the value of at least one order of magnitude lower than their reported value which indicated that the cake layer resistance was much lower for a GDU process compared to the pressure driven process.
Table 4 HF membrane resistance value of GDU membrane system towards (a) HA and (b) CaCO3 filtration
(a)
Concentration
|
k'' (m3/s)
|
k' (kg/m2s)
|
Rm (m2/kg)
|
Rc (m2/kg)
|
10 mg/L
|
3.05x10−6
|
3.65x10−13
|
2.17x1010
|
4.53x1011
|
20 mg/L
|
2.11x10−6
|
2.53x10−13
|
2.17x1010
|
6.62x1011
|
30 mg/L
|
2.08x10−6
|
2.50x10−13
|
2.17x1010
|
6.71x1011
|
40 mg/L
|
2.06x10−6
|
2.47x10−13
|
2.17x1010
|
6.79x1011
|
50 mg/L
|
1.65x10−6
|
1.98x10−13
|
2.17x1010
|
8.52x1011
|
(b)
Concentration
|
k'' (m3/s)
|
k' (kg/m2s)
|
Rm (m2/kg)
|
Rc (m2/kg)
|
0.1 wt%
|
2.03x10−5
|
2.44x10−12
|
2.17x1010
|
4.92x1010
|
0.2 wt%
|
1.68x10−5
|
2.02x10−12
|
2.17x1010
|
6.40x1010
|
0.3 wt%
|
1.60x10−5
|
1.91x10−12
|
2.17x1010
|
6.89x1010
|
0.4 wt%
|
1.46x10−5
|
1.75x10−12
|
2.17x1010
|
7.72x1010
|
0.5 wt%
|
1.27x10−5
|
1.53x10−12
|
2.17x1010
|
9.15x1010
|
Table 4 shows the variation of the filter cake resistance, Rc for humic acid and calcium carbonate solution. In this study, the resistance of cake layer, Rc was found to be increased as the concentration of feed solution increased. Amongst the feed solutions, humic acid solution gave the highest filter cake resistance, Rc of 8.52x1011 m2/kg. This is in accordance to the previous SEM image that the cake layer is denser due to its smaller mean size distribution of HA (2.443 µm) compared to the calcium carbonate (4.793 µm). On the other hand, CaCO3 had the lowest Rc of 4.92x1010 m2/kg. This finding is in accordance to our previous flux declination result. As compared to humic substance, precipitated CaCO3 produced more porous layers due to its irregular particulate size.
Permeability analysis
Figure 8 shows the changes of permeability for the feed solution under different feed concentrations. It can be observed that the k’ value (m3/s) of the foulants were having an order of magnitude differences for different type of feed. The lowest k’ value (x 10−6) was recorded for the HA filtration and the CaCO3 filtration which appeared in particulate materials has the highest permeability (x 10−5). The inorganic substance, CaCO3 formed irregular particulate materials that enabling the maximum water permeation through the loose cake layer on the membrane surface. Thus, this layer in fact gave a much denser cake layer for the water permeation.
All the foulants showed that the water permeability decreased with the increased of feed concentrations (She et al. 2016; Shi et al. 2020). The turbidity showed that the permeability decreased almost linearly with the concentration. On the other hand, humic substance showed a non-linear decreasing trend of the permeability against feed concentration. The rapid decrease of permeability against the concentration of HA indicated that the fouling of HA might be contributed by both the cake layer as well as membrane adsorption by the free HA via hydrophobic interaction. The reduction of HA permeability in higher feed concentration can be attributed to aggregation and cake formation (Costa and de Pinho 2005; Sutzkover-Gutman et al. 2010).
Concentrating Effect of dead End GDU membrane system
Figure 9 is the simulated ideal concentration of solution (being concentrated) in the filtration module over the time. Based on Fig. 9, the concentration profile is linear and permeability (k”) dependant. The higher the k” value, the faster the solution being concentrated. The concentration of two feed solution including humic acid and calcium carbonate solution were estimated using Eq. 16 and compared to the actual concentration in the filtration module after the filtration time of 1 hour 40 minutes.
According the results shown in Table 1, for HA, the predicted concentration remains in membrane module by the developed model was in the range of 69.74 to 378.90 mg/L compared to the calculated concentration of HA in membrane module was about 10.89 to 521.53 mg/L. The calculated concentration CaCO3 was 4.77 to 61.96 wt % while the predicted CaCO3 concentration was around 1.13 to 78.76 wt %. From the obtained results, it was found that the ideal calculated concentrations for low concentrations feed was always lower than the experimentally determined concentration. On the other hand, at higher concentration, the calculated concentration was higher than the measured value. This indicated that for low concentration, the permeability constant is not dominated by the cake layer, so the actual permeability is higher than the predicted value. In contrary, at higher feed concentration, due to the rapid fouling, the concentration of the retentate is lower than the ideal conditions due to the adsorption of solutes on the membrane surface during fouling.
Real water filtration using the GDU membrane system
The purpose of carrying out the river water test was to assess the removal efficiencies of the impurities presence in the river water. For that purpose, 80 mL of the river water sample was filtered with no water circulation. The solution in the module was emptied naturally as the feed was done once. Based on Fig. 10, the maximum flux of 45.98 L/m2hr could be achieved around 500 s after which sharp flux drop was seen. The flux dropped approximately 52.47% from 45.98 to 21.86 L/m2hr within 500 s after the maximum flux. Then, the slope become gradually decreased to the lowest flux of 0.86 L/m2hr indicating the depleting hydrostatic pressure. Generally accepted membrane filtration theory assumes that the formation of fouling layer during ultrafiltration led to a continuous increase of hydraulic resistance and decreases of flux (Wang et al. 2019).
Table 5 shows the water quality analysis obtained from the ultrafiltration process. The results indicated that the GDU membrane system is able to polish the river water and produce permeate with better quality. The ultrafiltration process using GDU membrane system efficiently removed the contaminants or foulant of the feed river water including chloride, ammonia, nitrate, nitrite, sulphate, chemical oxygen demand (COD) and also turbidity, pH, conductivity and total dissolved solid (TDS).
During the experiment, partial reduction of chloride, ammonia, nitrate, nitrite and sulphate concentration were observed. Chloride was reduced from 4.73 ± 0.35 mg/L to 2.43 ± 0.25 mg/L which result in 48.63% removal while ammonia content dropped to 0.08 mg/L ± 0.01 from the initial concentration of 0.13± 0.02 mg/L. Ultrafiltration is known to be the membrane that has poor retention on the dissolved ionic salt due to its bigger pore size. Although at the beginning, the chemical oxygen demand (COD) was 22.33 ± 1.53 mg/L, settlement assisted GDU membrane system was able to reduce the COD to an undetectable level showing that the organic matter content can be effectively removed. COD is a crucial parameter in determining the water quality that represent organic loading in the water (Verma and Singh 2013).
Table 5
Water Quality Analysis for Kerian River Filtration Process
Parameter
|
Before Filtration
|
After Filtration
|
Chloride (mg/L)
|
4.73 ± 0.35
|
2.43 ± 0.25
|
Ammonia (mg/L)
|
0.13 ± 0.02
|
0.08 ± 0.01
|
Nitrate (mg/L)
|
Below 4 mg/L
|
Below 4 mg/L
|
Nitrite (mg/L)
|
0.04 ± 0.01
|
Below 0.01 mg/L
|
Sulphate (mg/L)
|
Below 5 mg/L
|
Below 5 mg/L
|
Hardness (mg/L)
|
16.33 ± 1.53
|
10.0 ± 1.0
|
COD (mg/L)
|
22.33 ± 1.53
|
Below 0 mg/L
|
Turbidity (NTU)
|
46.53 ± 0.15
|
0.25 ± 0.02
|
pH
|
7.45 ± 0.02
|
7.20 ± 0.03
|
Conductivity (µS/cm)
|
63.90 ± 0.02
|
58.97 ± 0.59
|
TDS (mg/L)
|
31.93 ± 0.306
|
29.50 ± 0.265
|
Unexpectedly, merely 60% of the hardness in the river can be removed which indicated that the hardness may be in a more dissolved form. On the other hand, the feed turbidity dropped significantly from an initial value of 46.53 to 0.25 NTU which showed that GDU membrane is very effective in suspended solid removal. For the other parameter such pH, conductivity and TDS, no significant changes and slightly fluctuation in the permeate quality were observed due to their smaller size. It can be concluded that the proposed system is effective in removing solid and bigger organic molecule which is a typical performance of ultrafiltration membrane (Wang et al. 2019).