MEMBRANE FOR WATER REUSE :

effect of pre-coagulation on fouling and selection

Y. Soffer*, R. Ben Aim** and A. Adin*

*Division of Environmental Sciences
The Hebrew University of Jerusalem
Jerusalem 9190
Israel

** Laboratory of Environmental Engineering
INSA/GPI
31077 Toulouse, France

ABSTRACT

Membrane filtration is adequate for producing disinfected clear water suitable for different kinds of applications. However, fouling of membranes is the main limitation. This study has been focused on the ability of flocculation to remove organic colloids from wastewater which play an important role in the fouling phenomena. First, Flocculation was optimized for high-efficiency removal of suspended solids and organics. Second, flocculation at selected ferric chloride dose–pH before membranes ranging from nanofiltration to 50KDa ultrafiltration was studied. Flocculant addition mostly resulted in large flocs which could be transported away from the membrane by shear forces or lateral migration and improved flux. Fouling increased when high molecular weight cut off membrane was used, the fouling mechanisms seems to be blocking, internal clogging, and cake formation with time. pH5.5 (charge neutralization zone) provided better removal and lower fouling intensity than pH7.8 (sweep coagulation zone). Ultrafiltration of 4KDa at acidic pH and 150 mg/L ferric chloride could reduce DOC by 70% and UV-254 nm by 60%. That quality better than nanofiltration at basic pH and same dose, but the fouling in ultrafiltration was lower. Thus coupling of flocculation with UF membrane might be the best compromise for producing in very compact units a very clear water for possible reuse in industrial areas.

INTRODUCTION

Filtration using membrane is sensitive to fouling phenomenon in which the membrane interacts with solutes in the feedstream, resulting in decrease in membrane flux. In-line flocculation membrane filtration is a method of reducing the degree of internal clogging. The flocculation is used to achieve two objectives: eliminating the penetration of colloidal particles or organic matter into membrane pores and modification of deposit characteristics (Ben Aim et al, 1988). Another possibility to reduce fouling can be in some cases selection of membrane with lower pore size range. But higher pressure has to be applied and lower flux will result in membranes with small pore diameter.

Lahusine-Turcaud et al. (1990a) found that iron coagulation of surface water slowed the decline in membrane flux most efficiently, when flocculation conditions produced particles with zeta potential close to zero. This is thought to be due to differences in particle size and cake porosity rather than improved removal of foulant. Lahusine-Turcaud et al. (1990b) investigated filtration (1KDa ) of tannic acid, humic acid and kaolin. They found that particles near 0.2 mm in diameter produce rapid fouling, while particles greater than 3 mm in size have a little effect on flux. Lahusine-Turcaud et al. (1992) found that fouling behavior of humic substances after coagulation with polyaluminum chloride depends on pH. For pH<7 the fouling lower than for pH>7. Under the last condition the fouling mechanism seems to be the filtration on a cake settled on the membrane surface. For pH<7, the mechanisms are mixture of cake and internal fouling.

This research examines the improvement of wastewater–primary effluent quality applying iron coagulation before membrane filtration (LMWCO-UF, low molecular weight cut off – ultrafiltration at cross flow, CF). The iron coagulation ability to improve dissolved organic matter removal was tested in parallel to fouling characteristics at different MWCO of membrane, dose-pH, with or without flocs separation prior filtration.

MATERIALS AND METHODS

Membrane filtration tests

A Stirrered cell has working volume of 500 ml. It could be fitted with 75 mm diameter membrane discs which manufactured by CELGARD (NADIR types – membrane characteristics are summarized in Table 1). The stirring rate was kept fixed at 500 rpm throughout the experiment. Nitrogen gas was used to pressurize the system and to maintain the transmembrane pressure, 3 atm for UF membrane and 6 atm for NF (nanofiltration). Samples for turbidity, UV-254 nm and DOC were collected. Flux reading were taken by weighing out samples on balance (Sartorius basic 3100, ± 0.01 g). All experiments runs ended after 345 ml of wastewater were filtered. The membrane influent were wastewater–primary effluent after iron coagulation (ferric chloride – FeCl_3× 6H₂O has a gram molecular weight of 270.3) at two different pH levels 7.8 (natural pH) and 5.5. The iron flocculation was performed using conventional jar test procedure (Phipps & Bird, Inc.– rapid mixing at 120 rpm for 60s, slow mixing at 30 rpm for the next 20 min, in some cases the suspension was allow to settle for 30 min). The effect of membrane influent was studied at three modes: A one, influent contains all the particles and flocs after flocculation without settling (ws), and the second, the influent was wastewater after flocculation and settling (as), the last without coagulant (wc). After filtration test, the membrane was soaked at ambient temperature for 12 hour in NaOH solution. This procedure was used to clean the membrane from the fouling materials. It is important to ascertain that all the discs groups have hydraulic permeability values within a narrow range. The variation of average pure water flux with different membrane discs was within 4% of the average value. All the fouling curves (J/Jo) reports with viscosity corrections to 20° and (J–The permeate flux at time t, and Jo– permeate flux of clean membrane)

TABLE 1.
MEMBRANE CHARACTERISTICS

Wastewater.

The raw wastewater used in this study was taken from INSA university and went through settling for 3 hours. The parameter range or average values of the primary effluent: turbidity – 42–70 NTU; UV-254 nm – 0.35–0.45 cm^-1; DOC – 100–130 mg/L; Zeta potential – -27 – -31 mV; TSS – 140 mg/L; COD (total) – 380 mg/L; COD (soluble) – 160 mg/L; pH – 7.8–7.9.

Analytical procedure.

The efficiency of the coagulant and the membrane filtratoin was determined by monitoring turbidiy, UV-254 nm absorbance and DOC (dissolved organic carbon, model-1555B-Ionics). Residual humic-like organic matter was estimated by measuring UV-absorbance at 254 nm (D-40, Backman). The samples were taken at the end of the filtration tests (average samples) or after settling and filtered through 0.45 m m. Zero absorbance was defined for distilled and dionized water. The spectrophotometer samples cells were made of quartz with a light path of 10 mm. Particle size distribution was measures using Granulometer Cilas Alcatel. Zeta potential was measured by using a malvern zeta master. Pore streaming potential of the membrane in presence of ferric chloride solution was also checked, the general mode of measurements is described elsewhere (Nystrom, 1989).

RESULTS AND DISCUSSION

Jar test results using ferric chloride.

Variation in quality parameters (C/Co of Turbidity, UV-254 nm, DOC) as a function of ferric chloride dosage at pH 7.9 (natural pH) and pH5.5 are shown in Fig. 1.

All the C/Co values are positive, except in narrow coagulant ranges for UV-254 nm absorbance (0-50 mg/L for pH7.9 and pH5.5 – insufficient coagulant dose and overdosing coagulant dosage which is close to 300 mg/L at the acidic pH). This «negative effect» may be affected by the sensitivity of the UV-254 nm absorbance measurement to small colloids or small Fe (OH)₃(s) which did not react with the organic fraction. For any practic use, reduction of the interference from turbidity can be achieve by subtracting 545 nm absorbance. The fact that maximum absorbance of FeCl₃ and some organic solution is closed to 254 nm (Friedman, 1952) has a significant effect at the case of residual matter (ineffective coagulation). In most of the coagulant range (for both pH levels) when the dosage is higher, the effluent quality is higher (direct relationship). At pH5.5, the optimal dose was 200 mg/L (for turbidity–150 mg/L), and at pH7.9 the dose is the maximum, 300 mg/L. At the minimum optimum dosage–150mg/L, the C/Co values are lower at pH5.5 than at pH7.9. In this dose, the improvement in the colloids removal (Turbidity) was higher than for the DOM (dissolved organic matter – which is represented by UV-254 nm and DOC). For each pH, the vertical distance between the turbidity and DOM, was higher than the distance between the DOC and UV-254 nm curves. It can be explained on the basis of the destabilization mechanism efficiency. Colloids and suspended particles can be remove by adsorption–charge neutralization or sweep coagulation. In the case of DOM it is different. Usually the efficiency of sweep coagulation (which plays an important role in particular at high doses).is lower than the other mechanism. The high existence of positive soluble iron species and positive iron hydroxide at acidic pH (lower than the iron IEP– iso electric point at pH closed to 8.5) increased the probability of adsorption and charge neutralization interactions with DOM (usually carry negative charge). Moreover, the humic substances become more hydrophobic and less negative with decreasing the pH. In acidic pH the complexation degree of humic material–iron is high. From the effluent pH curves it can be seen that the pH decreased after using acidic pH (in the initial stage ) is higher than the case of initial natural pH. Figure 2 show the PSD (particle size distribution) of the jar test effluent and the count mean diameter – d(50%) at the best pH (pH5.5). The contribution of small particles (reflect in minimum value of d(50%) increases in two cases. First, the case of raw water which contains high amount of colloids and particles. Acidification causes decrease in stabilityand as a result increases the natural coagulation rate of small colloids to particles in the range of particle analyzer (0.1m m<). The second case exist when the coagulation can produce flocs with high settling velocity (dose³ 100 mg/L).

Membrane filtration without flocculant.

Tests at different MWCO (4–50 KDa) were performed in order to find the MWCO influence on the quality parameters removal (Fig. 3a.) and on fouling (J/Jo– fraction of initial flux) intensity as a function of filtration time (Fig. 3b). For both pH levels, decreasing in MWCO value affect the improvement level of the quality parameters in linearly. Possitive deviation from this curve at pH5.5 (50KDa) was as a result of the test in this work point which was performed in dead end (DE) mode of operation in contrast to the other test (crossflow – CF, with aggitation).

In all the work points the removal efficiency was higher at pH5.5 than at natural pH. For both pH levels, the removal efficiency of DOC was higher than for UV-254 nm. The vertical distance between this two curves was higher at pH5.5 than at pH7.8 (at pH5.5 the DOC removal was almost double than that of UV-254 nm, and for UV-254, the two curves were close). The same variation was found at dose of 150 mg/L (jar test results, Fig. 1.) From here, it can be assumed that for pH7.8, the influence of physical separation is higher than the physical-chemical interaction (most of the organics have acidic IEP which affect size and charge). From fig. 3b it can be seen that the effect of the MWCO on fouling development is more pronounced than the pH. The fouling rate is more rapid with increasing the MWCO (at quasi steady state: J/Jo>0.65 for 4-10 KDa and J/Jo<0.35 for 30-50 KDa). Since the DOM removal percent of high MWCO is low it can be assumed that the origen of the high flux decline rate at the initial run is pore blocking (immediate process), in that case the effect of adsorption–polarization (fast process) is less significant (major mechanism for 4–10 KDa). The organic fraction cause with the time internal clogging and/or cake formation which affect the flux at moderate form (low process). For filtration test with high MWCO (50 KDa) the contribution of cake layer to membrane resistance is higher (32-50% at terms of fifference in J/Jo before and after surface wash) than the other. This can be attributed to activity of back transport mechanism. The lateral migration intensity of particles is low in high membrane pore size (Committee Report, 1992). In this situation, the minimum size of particles that will not deposit on the membrane is higher than in the case of small pore size membrane. For the last, the brownian motion value which affects also the back transport of small colloids is high. In most of the cases the fouling was higher at pH5.5 than at pH7.8 (2-3%), inversely related to the DOM removal. The difference can be also observed in small absolute turbidity levels (pH7.8: 0.11-0.15 NTU, pH5.5:0.01-0.1 NTU).

Membrane filtration of wastewater after iron coagulation.

Jar test effluents in the insufficient dose range (without settling-flocs or after settling) were filtered using coarse 50 KDa membrane, pH7.8 (Fig. 4a). The difference between the fouling curves increases with the addition of more coagulant and filtration after settling. But it can be noticed that the J/Jo values at the end of the runs were close (J/Jo of 0.03-0.07). The effect of more coagulant addition (75 mg/L–without settling) on removal efficiency was significant, also relativly to the jar test (32% and 23% for DOC and UV-254 nm correspondingly). The filtrated volume which is needed to get a stable curve is large in the cases of DOC (Fig. 4b) and UV-254 nm in contrast to the turbidity (data do not show), in particular when low doses of coagulant were used. In presence of coagulant particles and flocs in sticky layer (gel layer or secondary membrane which produce barrier energy), the selection activity rate is higher than without settling of the influent. Surface wash improved the flux also by 100% relative to without coagulant-test (the difference between after settling and without settling–tests wash was negligible). From here, when the test was performed without coagulant the effect of internal clogging on the flux was higher.

Membrane filtration using MWCO of 10 KDa.

Variation in DOM removal applying the optimum dosage from the jar test (150 mg/L) at pH5.5 and pH7.8 are shown in Fig. 5a. pH5.5 brought about better removal than pH7.8 (close to double for DOC), even when the influent at acidic pH was after settling and at pH7.8 contained flocs. For each pH, the operation mode had low effect on the quality. The turbidity values were also lower at pH5.5 (£ 0.09 NTU) than at pH7.8 (0.098–0.16 NTU). Addition of flocculant improved the flux declining (Fig. 7b). In presence of flocculant, the difference at the two pH levels was significant (maximum improvement of 19% at acidic condtion). pH5.5 gave low fouling rate compare to pH7.8, with or without settling of the influent. There are at least three reasons to this gap: the one, higher coagulation degree of organics matter applying acidic pH than for the basic pH (from fig. 1 at 150 mg/L the removal ratio: for DOC–56%/11%~5, for UV-254 nm–35%/15%~2.3) and enhancement of concentration polarization (e.g., the diffusion coefficient increases with the pH for humic substances). Second, even the coagulation is more effective in removing high molecular organic compounds (10⁵,<MW) acidic iron coagulation can remove massess bellow 750 Da most of them hydrophylic in nature, when for basic pH mass up to 10³ Da were detected in mass spectra (Dennett et al, 1996). The third, pH7.8 is closed to the minimum solublity point of iron and also to Fe(OH)3 (s) zero electric point. It is probably affect on deposit. Since DOC contain also components which do not absorb at UV (amino and aliphatic acids, sugars,), efficient coagulation (pH5.5) free to remove more hydrophylic fractions which have low probability to be removed by high UF (even at 10 KDa without coagulant). It can be assumed that at pH7.8 some organics removed contacting with the hydrophobic cake, and minimize the internal fouling. The residual fouling after cake surface wash was similar and low for all the experements (till 2%) including that of without coagulant. In the last case, it can be assumed that the high cake resistance can protect on the membrane pores from high internal clogging intensity. At both pHs, when the influent particle load is low the flux value is high.

Membrane filtration using MWCO of 4KDa.

In comparison to 10 KDa, the removal improvement applying 4 KDa (Fig. 6a) was limited for the DOC but on the other hand, for UV-254 nm the removal improvment reached even to 24% (pH7.8–without settling). pH5.5 continues to provide better removal at 4 KDa, but there is a difference in the price (fouling degree, Fig 6b). At 4 KDa, the fouling curves of filtration-without settling were lower even than the fouling curves of filtration-without coagulant. The curves of filtration test of influent-after settling at pH5.5 gave the minimum fouling as membrane of 10 KDa provided. The above mentioned difference can be explained by the possibility that some of the organic material (e.g., iron-organics complexes) that without coagulation could be transported away from the membrane surface by diffusion attach the membrabe as a result of insufficient diffusion coefficient. Even the fact that these fractions also exist in the membrane influent which is after settling , the effect of the non-diffusion fraction on the fouling structure can be more significant due to colloids–flocs interaction on LMWCO membrane pore entrances and colloid bridging (Visvanathan & Ben Aim, 1989). Another explanation can be the sensitivity of 4 KDa to pore blocking and adsorption relatively to 10 KDa (better DOM removal). The difference in the J/Jo values before and after surface wash can support this claim. The residual fouling was 1–5% for filtration influent after settling (basic and acidic pH correspondingly), while for the others that value was 10-12%.

Comparison between low MWCO UF membrane and NF applying ferric chloride coagulation–pH7.8.

There is a trend to improve DOM removal by using LMWCO membrane (Fig.7a). The improvement rate is higher with decrease MWCO from 10 KDa to 4KDa (UF range) than from 4KDa to nanofiltration (N-30-F). It is expressed in particular by UV-254 nm removal, in the category of filtration test-after settling. The data in table 2. illustrate the ratio between the total removal (include: jar test–settling–filtration) and the jar test removal alone. It can be noticed that the ratio for NF (pH7.8, 150 mg/L) is higher than the other at the same dose. Filtration test in 4 KDa, pH7.8 indicates also high ratio. The ratio at pH5.5 is relatively low resulting from the jar test effluent was better in this conditions than in the natural pH. This ratio can also explain why for UF (4KDa, 10 KDa ) the fouling intensity at acidic pH was minimal. From Fig. 7b, it can be seen that the intermediate membrane MWCO is less sensitive to fouling than the other, for both operating possibilities. The influence of the pretreatment mode on the NF fouling was negligible, and the fouling curves were almost the same. The residual fouling after surface wash, for NF was 28–32% (with or without floc correspondingly) whereas for the others this phenomenon was limited. From table 2 it can be also seen that when the membrane influent was after settling the effect of cake resistance on fouling increase with membrane pore size and inversely effect on wash fouling ratio (in both pH, in particular for the basic pH–Fe(OH)₃(S) precipitation on surface and less inside the pores as it can be expected in microfiltration, iron colloids becomes smaller with increase the pH). It can be assumed that the change in membrane material (table 1) also affect internal clogging by charge macromulaculates or organic-iron complexes. Pore streaming potential measurements show similar values in presence of 150 mg/L ferric chloride (at pH5.5 and 8: 10KDa–8 and 4 mV/bar, 4KDa–5 and 2.5 mV/bar), The latter deserves further investigation. The smaller the membrane pore, the higher the cumulative effect of adsorption with the time till pore blocking (e.g, NF).

TABLE 2.
COMPARISON BETWEEN JAR TEST AND FILTRATION EFFLUENTS

SUMMARY AND CONCLUSIONS

1. Pre-coagulation has significant effect on fouling and separation in particular at optimum dose, pH condition. Acidic coagulation at pH5.5 (adsorption and charge neutralization) can remove 56%-DOC and 35%-UV-254 nm absorbance (at pH7.8–sweep coagulation mechanism: 11% and 15% respectively).

2. In filtration without flocculant, the lower the UF–MWCO, the higher is the DOM removal in linear ratio at both pHs. Similar to jar test, DOC removal is higher than UV-254 nm. The removal efficiency is higher at pH5.5 than pH7.8. MWCO effect on fouling is more pronounced than the pH (change by 2-3%). The fouling increase with MWCO, high pore blocking at the beginning and cake formation and internal clogging with time (J/J_o>0.65 for 4-10 KDa and J/J_o<0.35 for 30-50 KDa, poor DOM removal).

3. In most cases of UF, the higher the coagulation degree before the membrane the lower the fouling. At natural pH (pH7.8), the DOM removal ratio (after filtration/after jar test) is more than double than pH5.5. This is why the fouling is higher at pH7.8 than at pH5.5 (by 14% for 10 KDa) for both operation modes (influent contain flocs or not). Larger flocs can be transported away by lateral migration and shear forces (less cake resistance and internal fouling effect). When the influent do not contain flocs, the fouling is less than the parallel mode, usually at acidic pH the difference is higher (it was noticed 19% improvement in permeate flux at 10 KDa relative to the case without applying pre-coagulation). But, there is no guarantee that applying coagulation always improve fouling. In some cases, the coagulation produce small flocs or iron-humic complexes with limited diffusion coefficient for back transport than their origin coefficient.

4. UF can not compensate on low coagulation degree. At this point, the removal efficiency for all the process is higher at acidic pH for both operation modes. The higher the DOM-flocs concentration, the higher DOM removal (an extreme case: at 10 KDa .the DOC removal after filtration at pH7.8 without flocs separation is 40%, whereas at acidic after flocs separation is 69%). For each pH, the operation mode has a significant effect on fouling degree than on DOM removal,

5. The smaller the membrane pore, the higher the cumulative effect of adsorption (28-32% for NF). Effect of cake resistance increase with membrane pore size. Filtration of influent with flocs brought up more internal fouling. At pH7.8 residual fouling after surface wash is lower than at pH5.5, in contrast to cake.

6. Turbidity can not be sensitive qualitative index for comparison (<0.112 NTU in all filtration tests).

7. DOM removal is improved using LMWCO membrane. The improvement rate is higher with the decreasing from 10 KDa to 4KDa than from 4KDa to nanofiltration, in particular for UV-254 nm. The influence of the pretreatment mode on the NF fouling was negligible, the fouling is almost the same.

8. On the basis of the above mentioned, there are two practical possibilities for membrane treatment for wastewater (put away the cost of activated sludge): (A.) Membrane filtration at pH5.5 can remove ~70%-DOC (150 mg/L ferric chloride, for 4-10KDa) and UV-254 nm at 60% (38%, in 10KDa). The filtrate quality is higher than NF at pH7.8 + 150 mg/L ferric chloride. The fouling intensity at pH5.5, influent–after settling is minimum in NF-10KDa range. The disadvantage of this treatment is the acidification. (B.) Membrane filtration at pH7.8 (150 mg/L ferric chloride)-On the base of filtrate quality, there is advantage to apply 4 KDa membrane, in conditions of influent without flocs separation. (DOC removal of 40% and UV-254 nm removal close to 57%, almost at the same level of NF). The disadvantage is high fouling rate. At this point, it can be concluded that the operation of membrane with acidification can provide high buffer capacity at non-stable influent conditions. The acidification price, in particular for large wastewater quantity is not negligible. This is a question of economic equilibrium, if the less fouling degree and high final effluent quality have the ability to compensate for acidification price.

 

ACKNOWLEDGMENT

This work was partially supported by a grant from the French and Israeli Ministries of Science and Technology (AFIRST). This paper is part of graduate work of Y. Soffer.

REFERENCES

Ben Aim, R., Mietton, P.M., Vigneswaran, S., Yamamoto, K., and Boonthanon, S. (1988). Application of Crossflow Microfiltration in Sewage Treatment. In Advances in water pollution control. ed. by T. Pandward, C. Polprasert and K. Yamamoto. Proceedings of the second IAWPRC Asian Conference of Water Pollution Control, Pergamon Press pp. 613–619.

Committee Report: Membrane Processes in Potable Water Treatment. (1992). Jour. AWWA, 84(1), 59-67

Dennett, K.E, Amirtharajah, A., Moran, T.F, Gould, J.P. (1996). Coagulation: Its Effect on Organic Matter. Jour. AWWA, 84?(4), 129-142

Friedman, H.L. (1952). The Visible and Ultraviolet Adsorption Spectrum of the Tetrachloroferrate (III) Ion in Various Media. Jour. of the American Chemical Society, 74(2), 3-10.

Lahoussine–Turcaud, V., Wiesner, M.R., Bottero, J.Y., Mallevialle, J. (1990a). Coagulation Pretreatment for Ultrafiltration of Surface Water. Jour. AWWA., 12:76 –81

Lahoussine–Turcaud, V., Wiesner, M.R., Bottero, J.Y., Mallevialle, J. (1992). Coagulation–Flocculation with Aluminium Salts: Influence on the Filtration Efficiency with Microporous Membranes. Wat. Res., 26:5:695–702

Nystrom, M., Lindstrom, M., Matthiasson, E. (1989). Streaming Potential as a Tool in the Characterization of Ultrafiltration Membrane. Colloids and Surfaces, 36:297–312

Vigneswaran, S. and Ben Aim, R. (1989). Studies On Colloidal Membrane Fouling Mechanisms in Crossflow Microfiltration. J. Membrane Sci 45, 3-15.

Page publiée par la Société Française de Filtration avec l'autorisation des auteurs. Tous droits réservés