Contents lists available at SciVerse ScienceDirect
Desalination
j ourna l homepage: www.e lsev ie r .com/ locate /desa l
Effect and mechanism of preoxidation using potassium permanganate in anultrafiltration membrane system
Tao Lin a,b, Liang Li b, Wei Chen a,b,⁎, Shaolin Pan b
a Ministry of Education Key Laboratory of Integrated Regulation and Resource Development on Shallow Lakes, Hohai University, Nanjing 210098, Chinab College of Environment, Hohai University, Nanjing 210098, China
⁎ Corresponding author at: Ministry of Education Keyulation and Resource Development on Shallow Lake210098, China. Tel.: +86 13913899869; fax: +86 0258
The use of potassium permanganate (KMnO4) for sand filter effluent pretreatment was investigated as well asits performance in combined ultrafiltration (UF) processes. The former process was implemented via directcontinuous dosing of KMnO4 into the influent of a UF system. Investigations into this option indicated thatthe optimal dose of KMnO4 was 0.3 mg/L and significant improvements to the efficiency of natural organicmatter (NOM) removal during the KMnO4/UF process were achieved. Due to membrane fouling, the use ofKMnO4 also resulted in lowered total membrane resistance in comparison with results obtainedwhen operatingwithout KMnO4. The KMnO4 oxidation of the sand filter effluent to UF also indicated some potential for mitigat-ing transmembrane pressures, due to the changed characteristics of organic pollutants. This observation wasconfirmed by Attenuated total reflection-Fourier transform infrared spectroscopy analysis (ATR-FTIR). Scanningelectronmicroscopy showed that a significant amount of fouling that had clogged themembranewas very looseand easy to be removed by hydraulic washing during the KMnO4/UF process. The fouling removal mechanismwas demonstrated by energy dispersive X-ray spectroscopy. It was noted that KMnO4 could oxidize metal ionsinto metal oxide particles, which were adsorbed on the membrane surface together with the pollutants. Theseformed loose fragments on the filtration cake that were easy to flush away.
Ultrafiltration (UF) processes are widely used in drinking watertreatment to produce better quality water and to meet more stringentregulations, particularly concerning the removal of pathogens andturbidity [1–3]. However, UF membrane fouling problems may im-pede their wide application [4]. Membrane fouling results in a reduc-tion in membrane permeability and hence a decline in membranepermeate flux or an increase in the applied pressure required,which leads to higher operating costs. Over time, fouling causes thedeterioration ofmembranematerials, resulting in compromised effluentwater quality and shorter membrane lifetimes. Membrane fouling canbe classified as organic fouling, biological fouling, inorganic fouling orscaling, and particulate fouling. In water treatment processes, mem-brane fouling, mainly due to natural organic matter (NOM) such ashumic substances, polysaccharides and proteins, is a major cause offlux decline [5,6]. NOM is a complexmatrix of organic chemicals derivedfrom a number of sources such as soil, living organisms and plant detri-tus. These affect the odor, color and taste of water, form complexes withheavymetals and pesticides, and reactwith chlorine to form chlorinated
Laboratory of Integrated Reg-s, Hohai University, Nanjing3787618.
rights reserved.
disinfection by-products (DBPs). Several studies on the membranetreatment of surface waters, including lake and river waters, havedemonstrated that the DBPs precursors of NOM are major foulants ofmembranes [7–9]. Membrane fouling is correlated with the ability of amembrane to reject NOM and the deposition of rejected organic con-stituents on its surface and/or in its pores. There are two types of foul-ing associated with membrane processes: surface fouling and internalpore fouling. Surface fouling can be reversed by cleaning, although theremoval of foulants may be difficult in certain situations. Internal porefouling might have reversible and irreversible components.
Membrane fouling is inevitable as long as foulants exist in themembrane feed water. The development of effective pretreatmentmethods to remove or transform NOM from membrane feed wateris therefore essential for improving the cost-effectiveness of mem-brane systems and for the broader application ofmembrane technologyin water treatment systems. Pretreatment of water has been used as ameans of reducing membrane fouling. Several studies have developedpotential pretreatment options for UF membrane fouling control[10,11] including preoxidation, which has been widely investigated.Two preoxidationmethods have been developed for advancedUF treat-ment processes, one of which is the preoxidation of raw water. Thereare numerous reports that oxidants used in preoxidation of rawwater, such as ozone, chlorine, or permanganate, can enhance the re-moval of pollutants by coagulation and filtration processes. These pre-oxidants serve as both algaecides and flocculent aids [12–14]. Very
380 T. Lin et al. / Desalination 286 (2012) 379–388
little information is however available on the oxidation pretreatment ofUF influents. Previous reports [15] have demonstrated that oxidationpretreatment of UF influent significantly mitigates organic fouling,which has been primarily attributed to changes in NOM molecularcharacteristics. The oxidants can fragment the macromolecules in theNOM into lower molecular weight organics, with some mineralizationor transformation of humic/hydrophobic NOM fractions and polysac-charides into less sorbable organic acids.
Potassium permanganate (KMnO4), an oxidant, is widely used inwater treatment processes. Compared with other oxidation processes,such as chlorination, KMnO4 preoxidation produces fewer by-productsand there is some evidence [16] that the preoxidation of raw waterusing KMnO4 can remove iron and/or manganese or algae. Few studieshave considered the preoxidation by KMnO4 of sand filter effluentsprior to entry into the UF system. It is also not clear whether preoxida-tion of the sand filter effluent by KMnO4 affects membrane fouling. Be-cause of the complex and unstable nature of organic materials presentin water, the fouling mechanism for low-pressure membrane systemsused in water treatment is poorly understood [17]. In this paper, we in-vestigated the effects and mechanisms of preoxidation using KMnO4 ina UF system. The objective of this research was to conduct preoxidationbefore membrane filtration experiments, to study permeate flux de-clines and solute rejection patterns in the UF process, and to evaluatethe potential for mitigating membrane fouling using KMnO4.
2. Materials and methods
2.1. Feed water
Water samples for this study were obtained from the sand filter ofthe Xiang Cheng Water Plant in Suzhou, which receives raw waterfrom Taihu Lake. The key water quality characteristics are given inTable 1.
2.2. Ultrafiltration membranes
Hollow fiber UF membranes made of modified polyvinylchloride(PVC) were used for experiments. A UF membrane filtration area of0.133 m2 was used in the bench-scale tests, while a filtration area of40 m2 was used for the pilot-scale tests. A UF membrane filtrationarea of 0.133 m2 in the bench-scale tests was used to optimize thedose of KMnO4 and to investigate the mechanism of membrane foul-ing, while a filtration area of 40 m2, the dimension of the actual mem-brane module, was used for the pilot-scale tests for comparing thewater treatment process with practical test situations. Although di-mensions of the membrane filtration area differed in different tests,the same key parameters were assigned, which guaranteed
accordance between the bench-scale tests and pilot-scale tests. De-tailed characteristics of the membrane are summarized in Table 2.
2.3. Membrane filtration experiments
2.3.1. Bench-scale testsBench-scale tests investigated the optimization of preoxidation by
KMnO4 in a UF membrane system; these results were then applied inthe pilot studies. Prior to the experiments, new membranes weresoaked in pure water overnight and then pretreated with purewater for about 8 h to achieve a more stable permeate flux. Duringthe bench-scale tests, several dosing levels were investigated. Anelectric stirrer was used to mix the water and feedwater pressure ofeach pretreatment condition was maintained at a consistent level.The permeate flux was monitored by measuring the flux in a volu-metric cylinder at specified time intervals. Prior to the commence-ment of continuous UF in each sample, the pure permeate flux (J0)was measured with deionized water. The permeate flux of samples(J1) was compared to the pure permeate flux (J0) to provide a com-parison between the different pretreatment conditions. The mem-brane resistance caused by the nature of membrane (Rm) wasobtained using the resistance-in-series model [18]. At the end of theexperiment, microscopic observations of surface changes to the drymembrane were performed using ATR-FTIR (Attenuated totalreflection-Fourier transform infrared spectroscopy), scanning elec-tron microscopy and energy dispersive X-ray spectroscopy.
2.3.2. Pilot testThe KMnO4/UF pilot system used for these experiments is shown
schematically in Fig. 1. KMnO4 was applied as a preoxidant beforethe mixing stage. The regulating tank ensured a retention time of10 min in the oxidation process and the dosages of KMnO4 in the ex-periments were based on the optimization results of bench-scaletests. Transmembrane pressure (TMP) reflects the resistance due tomembrane fouling, including the effects of pore narrowing and thepresence of a cake layer. It can be expected that TMP will increasein the course of filtration, due to fouling. After the 2 h pure waterfeed experiments, the effluent from the sand filter was pumped intothe system. The permeate flux during the experiments was fixed at75 L/m2h, filtration cycle was fixed at 30 min, and the following pro-cesses were applied at the end of each filtration cycle: first wash:6 m3/h, 15 s; backwash: 8 m3/h, 40 s; and wash: 6 m3/h, 15 s.
2.4. Analytical methods
2.4.1. TurbidityTurbidity was measured using a Turbidimeter (2100N, Hach,
USA).
2.4.2. Ultraviolet absorbanceLight absorbance at 254 nm,which is associatedwithNOM aromatic
groups, was used as a parameter for monitoring the concentration ofdissolved organic matter during water treatment, in a fast and easymanner [19]. Ultraviolet (UV) absorbance at a wavelength of 254 nm(UV254) was determined using a UV–visible spectrophotometer
Table 2Characteristics of ultrafiltration membrane.
Fig. 2. Effect on the permeate flux of washing with KMnO4 during continuous ultrafil-tration of various sand filter pretreatments.
381T. Lin et al. / Desalination 286 (2012) 379–388
(EV300, Thermo Fisher, USA). Samples were filtered through a 0.45-μmmembrane (Millipore, USA) to remove particles prior to UV absorbancemeasurements.
2.4.3. Total organic content/dissolved organic content measurementsThe total organic content/dissolved organic content (TOC/DOC)
concentration level was determined using a TOC analyzer (1030 W,OI, USA). All TOC/DOC samples were measured in duplicate. Thereported value was the average of the duplicate values, providedthe relative percent difference between duplicate samples and cali-bration check standards was ≤±10%. DOC was operationally definedas the organic carbon concentration of a sample filtered through a0.45-μm membrane filter.
2.4.4. Inorganic cationsIron, manganese, aluminum and calcium concentrations were ana-
lyzed using an inductively coupled plasma mass spectrometer (Optima2100 DV, PerkinElmer, USA).
2.4.5. Hydrophilic and hydrophobic componentsSurface water was fractionated to a hydrophobic (HPO) fraction,
adsorbed by Amberlit DAX-8 (Rohm and Haas Company), a transphi-lic (TPI) fraction, adsorbed by Amberlite XAD-4 and hydrophilic (HPI)components, was passed through both the resins without any adsorp-tion. DAX-8 and XAD-4 were washed with methanol and deionizedwater prior to use in fractionation. The surface water pH was adjustedto 2.0 prior to feeding onto the DAX-8 column (at a rate of 5 mL/min)[20] and the XAD-4 resin (at 15 mL/min). The DAX-8 is a non-functionalized resin that retains strongly hydrophobic organic matter(attributed to humic substances) while the XAD-4 retains weakly hy-drophobic material known as TPI [21].
2.4.6. Molecular weight distributionDetermination of apparent molecular weight distribution
(AMWD) was carried out using gel chromatography (Lc-10ADVP,SHIMADZU, Japan). The whole water and filtered fractions werestored at 4 °C until further analysis for UV254 and DOC.
2.4.7. Attenuated total reflection-Fourier transform infrared spectroscopystudies
The ATR-FTIR technique (NEXUS870, Thermo Nicolet C, USA) wasused to investigate functional groups and molecular structures on themembrane surface and deposited foulants. All samples, including thevirgin and fouled membranes, were gently washed with deionizedwater, and then dried overnight at room temperature.
2.4.8. MicroscopyThe membrane surface was visualized using a field emission scan-
ning electron microscope (Hitachi-3400N, Hitachi, Japan).
3. Results and discussion
3.1. Optimization of preoxidant dosage
Bench-scale tests were done to optimize preoxidant dosage. Com-parative experiments were performed for different pretreatment con-ditions, in which the KMnO4 doses varied from 0.1 to 0. 5 mg/L.
Fig. 2 shows the effect on the permeate flux of washing during thecontinuous UF of the sand filter effluent with different KMnO4 dosages.The use of KMnO4 pretreatment resulted in a relatively better permeateflux thanwas the casewhen using direct UF, which implies that it couldreduce the rate and extent of membrane fouling. This phenomenoncould be due to the KMnO4 pretreatment producing a cake layer thatshowed less resistance. The permeate flux of UF without pretreatmentdecreased sharply in 2 h (without washing) and then decreasedgradually to about 70% of the initial permeate flux, for which the J1/J0valuewas 99.3. The permeate flux declinedmore slowly as the pretreat-ment dosage of KMnO4 increased in the range of 0.1–0.5 mg/L.
Fig. 3 indicates the effect of intermittent washing during continuousUF on the permeate flux. This indicates that direct UF resulted in ahigher rate of membrane fouling while the permeate flux recoveredslowly under hydraulic cleaning. Pretreatment with KMnO4 alsoresulted in a more successful restoration of permeate flux than wasthe case for direct UF under hydraulic cleaning. It is noteworthy that,in a dosage range of 0.1–0.5 mg/L, a higher dosage KMnO4 pretreatmentresulted in a better reduction of membrane fouling, under the same
0 2 4 6 8 10 120.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
only UF 0.1mg/LKMnO4/UF
0.3mg/LKMnO4/UF
0.5mg/LKMnO4/UF
Perm
eate
flu
x(L
/m2 .m
in)
Time(h)
Fig. 3. Effect on the permeate flux of intermittentwashingwith KMnO4 during continuousultrafiltration of various sand filter pretreatments.
Table 3Various membrane resistances (in terms of four parameters) subjected to ultrafiltra-tion and ultrafiltration combined with KMnO4 pretreatment (KMnO4/UF).
UF: ultrafiltration.Rm = resistance due to the nature of membrane; Rp = resistance due mainly tosolutes adsorbed within the inner pore fiber; Rc = resistance caused by filtered cakeon the membrane surface; Rt = total resistance; KMnO4/UF = membranes subjectedto ultrafiltration combined with KMnO4 pretreatment.
a Resistance (×1012 m−1).b Proportion of total resistance.
5min30min
90min0
20
40
60
80
100
Rp
Rc
Rm
Per
sent
(%
)
Time
Fig. 4. The proportion of the various components of membrane resistance with ultrafil-tration alone.
382 T. Lin et al. / Desalination 286 (2012) 379–388
hydraulic cleaning regime. Applying the KMnO4 pretreatment processbefore UF membrane filtration was thus very effective in terms of re-ducing fouling. This suggests a strong adhesion force between themem-brane and organic compounds when they are in direct contact. Thiswould cause strongmembrane fouling that is difficult to remove via hy-draulic washing. KMnO4 oxidation could remove some organicsadsorbed to the inner pore fibers ormembrane surfaces andmanganesedioxide, produced by KMnO4 preoxidation, could absorb organics dur-ing filtration. The treatment methods described in this paper reducedthe risk of strong adhesion forces developing between membrane andorganics. Oxidation with KMnO4 as in the pretreatment methods de-scribed above resulted in the formation of a loose filtration cake, creat-ing a fouling layer that can be easily removed.
The results showed that the best dosage for maintaining a highpermeate flux was obtained at 0.5 mg/L KMnO4. When the dosagewas greater than 0.3 mg/L, however, the Mn and color of the mem-brane effluent sometimes exceeded water quality standards. Duringthe experiment, the concentration of organics in the sand filter effluentvaried between 1.984 mg/L and 2.609 mg/L. Themain goal of this inves-tigation was, however, to control membrane fouling rather than to de-crease the concentration of organic compounds in the product water.At KMnO4 dosages of than 0.3 mg/L, a better recovery of permeatefluxwas obtained and the concentration of Mn in the product water oc-casionally reached water quality standards. A dosage of KMnO4 of0.3 mg/L was found to be optimal in terms of both water quality andpermeate flux.
3.2. Effect of membrane resistance
The values of membrane resistances (in terms of three parame-ters: Rm, Rp and Rc, described below) and to fouling, with and with-out KMnO4 pretreatment, are shown in Table 3. Rm indicatesresistance caused by the nature of membrane: Rm values at 5 min,30 min, and 90 min with and without KMnO4 treatment were thesame; Rp, indicates resistance caused mainly by solutes adsorbedwithin the inner pore fibers: these were lower for the pretreatedmembranes; Rc, indicates resistance caused by the filtration cake onthe membrane surface. Because the presence of KMnO4 resulted inthe partial removal of NOM from the feedwater, the Rc with KMnO4
pretreatment continued to decrease, whereas that of the membranethat had not been pretreated only decreased until 5 min after com-mencement with the KMnO4 dosage. The Rc with KMnO4
pretreatment was larger than the one without KMnO4 after 90 min.This phenomenon indicated that the KMnO4 treatment could reduceirreversible fouling that formed mostly from pore blockage, as wellas reversible fouling that formed mostly from cake deposition. Rt in-dicates the total resistance. Overall, the Rt with KMnO4 pretreatmentwas much lower than that without KMnO4 pretreatment.
A comparison of the proportions of various resistances (Figs. 4 and 5)indicated that the Rp and Rc values at 90 min in KMnO4− pretreatedmembranes were much lower than those from membranes that hadnot received such pretreatment. This suggests that KMnO4 oxidizedand destroyed certain solutes, particularly organic matter, adsorbedonto the inner pore fiber and membrane surface. This phenomenon in-dicated that pretreatmentwith KMnO4 could remove inner pore foulingas well as membrane surface foulingmore effectively thanwas the casefor membranes that had not received such pretreatment.
3.3. Permeate water qualities
Table 4 provides details on the water quality after ultrafiltration inthe pilot test. KMnO4 pretreatment for UF was more effective in re-moving NOM than direct UF. The concentrations of organic matter,as measured by UV254, TOC and CODMn, were low and algae and bac-teria were not detectable in either sample. Compared to resultsobtained from the direct UF process, the average effluent concentra-tions as measured by UV254, TOC and CODMn were reduced by
Pers
ent (
%)
5min30min
90min0
20
40
60
80
100
Rp
Rc
Rm
Per
sent
(%
)
Time
Fig. 5. The proportion of the various components of membrane resistance with ultrafil-tration combined with KMnO4 pretreatment.
0 5 10 15 20 25
0.02
0.04
0.06
0.08
0.10
Only UF KMnO4/UF
Temperature
Time(day)
TM
P(M
Pa)
0
5
10
15
20
25
Tem
perature(°C)
Fig. 6. Transmembrane pressure variations of two runs.
383T. Lin et al. / Desalination 286 (2012) 379–388
14.29%, 4.83% and 3.75%, respectively, in the KMnO4/UF. The declinein UV values was much more substantial than that of TOC; this maybe due to oxidative degradation of aromatic groups in NOM (thathave strong UV absorbance at 254 nm) by KMnO4. It was also notedthat, after this treatment, the permeates of both samples met waterquality standards in terms of smell, taste and color.
A significant decline in turbidity was noted (Table 4), which canbe expected since UF can remove most colloidal particles in the feed-water and hardness was similar in the different samples; this was notimpacted by KMnO4. The average effluent concentration of magne-sium, aluminum and iron were, respectively, 0.021, 0.016 and0.056 mg/L when UF treatment alone was used, while they were, re-spectively, 0.043, 0.015 and 0.01 mg/L when the KMnO4/UF processwas employed. All of these values were below the water qualitythreshold limits. The concentration of iron in the permeates fromthe KMnO4/UF process was approximately 17.86% of that from theUF process. This can be attributed to the oxidation of ferrous to ferriciron by KMnO4 and the subsequent formation of iron hydroxide pre-cipitate that is rejected by UF. The concentration of magnesium in theKMnO4/UF permeates increased to some extent, but was lower thanthe water quality threshold limit. As a result, the quality of the UFpermeate could be guaranteed to meet the regulatory criteria.
3.4. Transmembrane pressure trends
Fig. 6 shows changes in the TMP during the UF operation, bothwith and without KMnO4 pretreatment. Several significant increases
Table 4Water quality parameters after ultrafiltration (mean values).
Component 0.0 mg/L 0.3 mg/L Water quality standards
Turbidity (NTU) 0.08–0.1 0.08–0.1 1Smell and taste N N NColor (Pt–Co) b5 b10 15UV254 (cm−1) 0.019–0.038 0.016–0.033 –
were noted in the TMP, which rapidly increased during the initialphase and during the final phase but increased at a slow rate in themid-period of filtration. During the initial phase there was a rapid in-crease in membrane pore pollution, which caused the TMP to increaserapidly. At this stage, pollutants that had deposited on the membranesurface slowly formed a gel layer that further increased TMP andcould not be removed by hydraulic cleaning. TMP for the formationof gel layer balanced with TMP reduction through hydraulic cleaningduring the mid-period of continuous UF. The TMP therefore slowlygrew during this short period. During the final phase, the membranefouling was gradually aggravated and could not be sufficiently miti-gated by hydraulic cleaning. The rapid increase of TMP in the thirdphase can be explained in terms of the resistance model, which pos-tulates that membrane fouling is caused by the formation of a densegel layer due to the continuous deposition of contaminants on themembrane surface. During this phase, the dynamic balance betweenaccumulation and release of contaminants in the filter cake is brokendown, resulting in a rapid increase in resistance to filtration. Thiscaused a sharp increase of TMP in the short-term operation.
Fig. 6 also shows that operation of the direct UF system withoutchemical cleaning, for about five days, resulted in an increase inTMP from 0.018 to 0.099 MPa. When the combined KMnO4/UF pro-cess was run for 24 days, however, the TMP increased from 0.019 to0.099 MPa. This confirmed that KMnO4 pretreatment was effectivein extending the operating period, and also indicated that permanga-nate pretreatment before continuous UF was more efficient than di-rect UF in terms of fouling mitigation. It was thus clear that KMnO4
addition enhanced the membrane flux substantially, since KMnO4
removes pollutants that would otherwise cause membrane fouling.The underlying reasons for this are discussed in the next section.
3.5. Effect of pretreatment by permanganate on hydrophilic andhydrophobic fractions
To elucidate the cause of membrane fouling during KMnO4/UF,batch tests on the hydrophilic and hydrophobic fractions of organicmatter were conducted on effluent samples from different processstages (Figs. 7 and 8).
As shown in Fig. 7, the effluent from the sand filter exhibited a rela-tively low hydrophobic fraction concentration and could therefore beclassified as a hydrophilic surface water type. This was consistent withits low specific ultraviolet absorbance (SUVA) value (1.29 L/mg.m)and the observation that the SUVA value of its surface water was nor-mally less than 3 lm−1 mg−1. Such waters are classified as low-humicwater [18]. Preoxidation by KMnO4 also played a role in the removal
Fig. 7. Effect of pretreatment conditions on removal of dissolved oxygen content.
384 T. Lin et al. / Desalination 286 (2012) 379–388
of DOC, mainly due to removal of strongly hydrophobic organic com-pounds (HPO). It was observed that major changes in the hydrophilicand hydrophobic fractions of organic matter had taken place beforeand after UF, regardless of whether the effluent from the sand filterwas oxidized by KMnO4. It was clear that UF resulted in a reduction inthe proportion of strongly/weakly hydrophobic organic compounds(HPO/TPI) and an increase in the proportion of hydrophilic organiccompounds (HPI). The removal rate of hydrophobic organic compoundswas therefore greater than that of hydrophilic organic compounds inthe UF effluent samples. The presence of hydrophilic UF membranesmeant that hydrophobic substances could not easily pass through theUF andwere thusmore likely to be retained by the UF, while the hydro-philic organic compounds could easily pass through the UF. The UF ef-fluents therefore included more hydrophilic organic compounds andfewer hydrophobic organic compounds. This suggested that the hydro-phobic organic compounds not only had themost significant foulingpo-tential, but were also associated with severe membrane fouling.Pretreatment by KMnO4 facilitated UF removal of hydrophobic organiccompounds, which greatly improved the permeate flux. There werehowever fewer hydrophobic organic compounds in the effluent of theKMnO4/UF process than in the direct UF process.
This study also showed that the reduction in UV254 was alwayshigher than DOC reduction for all the four fractions after the dosing
Fig. 8. Effect of pretreatment conditions on removal of compounds (measured as de-clines in ultraviolet absorbance at 254 nm).
of KMnO4. This could be due to the strong oxidizing action ofKMnO4, which resulted in breaking unsaturated chemical bonds of or-ganic compounds. It should also be noted that DOC is not a specificmeasurement of a certain compound but a total carbon concentrationcomprised of the three fractions and therefore it experienced a lowerpercentage change than the UV.
3.6. Effect of pretreatment by permanganate on molecular weightdistribution
The mechanisms relating to the effects of pretreatment by per-manganate on membrane fouling were assessed by measuring themolecular weight (MW) distribution of organic matter (Fig. 9). Themolecular weights of organic matter in water samples collectedfrom filtration experiments were generally less than 5 kDa. Chro-matographs of NOM in these water samples usually showed five dis-tinct peaks in terms of the UV response. Three high MW peaksappeared in a range of 1–5 kDa (mostly 1–3 kDa), and two low MWpeaks were in a range of several hundred Daltons or less. It was alsonoted that the AMWD of the dissolved organic substances in thesand filter effluent was concentrated in the 0–3 kDa range. After10 min contact with KMnO4, there was an increase in organic sub-stances with low MW (of less than 1 kDa), while hose in the1–3 kDa range declined. This change resulted from the oxidizationof particulate organic substances in water. When the high-MW or-ganic substances in water decreased, the low-MW organic substancesgenerally increased.
There were however some distinctive peaks present before andafter UF, regardless of whether the sand filter effluent had been sub-jected to KMnO4 treatment. Compared with the UF influent, the or-ganic matter concentration of the UF effluent samples in the1–3 kDa range decreased by a margin larger than that measured insamples containing compounds of MW less than 1 kDa. The small or-ganic matter (of less than 5 kDa), consisting mainly of humic sub-stances that could be adsorbed to the membrane, could noteffectively be removed. The UF preoxidation process however dis-played a greater capacity to remove organic matter that was concen-trated in the range of 0–5 kDa (mostly 1–3 kDa) than direct UF. Thiscould be interpreted as being due to oxidation by KMnO4, whichgreatly changed the properties of the organic matter and changedhigh MW organic compounds into small molecular organic com-pounds. The attachment of organics to the membrane, whichdepended on the absorption, was the main cause of fouling. Duringthe experiment, the MW of most NOM was less than 5000 Da, i.e.
0 1000 2000 3000 4000 5000molecular weight (Da)
Sand filter effluent Sand filter effluent/kMnO4
UF effluent UF effluent/kMnO4
Fig. 9. Changes in the molecular weight of organic compounds for different pretreat-ment conditions.
385T. Lin et al. / Desalination 286 (2012) 379–388
much lower than the molecular weight cut-off, of 50,000 Da, associatedwith UF. Oxidation by KMnO4 could fragment NOM macromoleculesinto lower-molecular-weight organics. These compounds would havea relatively higher penetration rate, while the NOM, which containedmacromolecules, was likely to be absorbed on the membrane, resultingin fouling. In addition, the formation of MnO2, due to KMnO4 oxidation,would result in the adsorption of a portion of the organic compoundsprior to the attachment of such substances tomembranes. TheMnO2 to-gether with organic compounds, therefore represents an importantfraction of the fouling cake, which was looser and had lower resistancethan compounds formed by organics that had not undergone preoxida-tion by KMnO4. The dosing of KMnO4 would therefore increase the po-tential for mitigating membrane fouling.
3.7. Attenuated total reflection-Fourier transform infrared spectroscopystudies
The FTIR spectrum for virgin and fouled polyvinylchloride hollowfiber membranes is presented in Fig. 10. This indicates that the absorp-tion peaks in the virginmembrane spectrawere either eliminated or se-verely attenuated due to coating byNOM foulants, particularly betweenwave numbers 675 and 1500 cm−1. All three spectra displayed peaks at1647 and 3372 cm−1, which were attributed to the membrane materi-al. The UF membrane that had been fouled with sand filter effluent hadstrong peaks at around 1735 cm−1, signifying the presence of the C_Obond of lipids and/or lipid-like material in the foulant layers. This peakwas not observed for the membrane fouled with preoxidized water, in-dicating that organicswith high fouling potential were transformed intothose with relatively low fouling potential during KMnO4 oxidation.Since the intensity of spectral bands of membranes exposed to preoxi-dized effluents were weaker than those for membranes exposed toraw water, it appears that NOM constituents in sand filter effluentwere transformed into less adsorbable substances during preoxidation.The peaks near 1242 and 1327 cm−1 were characteristic of the alkanes,in which the intensity of spectral bands was significantly higher for themembrane fouled with preoxidized water, in comparison with themembrane fouledwith sand filter effluent. In conclusion, the FTIR analy-sis of the filtration cake, with and without KMnO4 pretreatment, indi-cated that the primary difference was that the C\H bonds in thearomatic rings had disappeared; more C\H bonds appeared in the al-kanes. It was apparent that the KMnO4 had reacted with aromaticrings, converting them to linear chain alkanes. Because chromophoremolecules are primarilymade up of aromatic rings, the KMnO4 changedthe chemical structure and the characteristics of the filtration cake thatoriginated from the ultrafiltration of sand filter effluent.
500 1000 1500 2000 2500 3000 3500 4000
0
20
40
60
80
100
Ref
lect
ance
%
Wavenumbers(cm-1)
virgin membranefouled with sand filter effluentfouled with KMnO4 treated water
Fig. 10. The FTIR spectra of membranes at various phases of membrane operation.
3.8. Microscopy results
SEM micrographs of the new membrane (Fig. 11a and b) clearlyshowed the pore surface of UF, while the micrographs of fouled mem-branes (Fig. 11c, d and e, f) exhibited different surface morphologies,with the film pores being clogged by pollutants. Images of the UFmembrane surfaces that had been subjected to KMnO4 pretreatmentand those that had not been subjected to such pretreatment are dis-played in Fig. 10. A comparison of SEM micrographs of membranesthat had been moderately fouled with various feedwaters after 12 hof continuous UF, indicated that the surface of the KMnO4/UF pre-treated membrane had a sparse, loosely-bound foulant layer on thefiltration cake (Fig. 11c and d). The membrane that had not been sub-jected to KMnO4 pretreatment (Fig. 11e and f) appeared to be rela-tively dense and compact. This indicates that the process ofpreoxidation by KMnO4 caused the characteristics of the filtrationcake to change, due to regional adsorption of macromolecules andthe uneven nature of NOM deposition. The uneven structure of thefoulant deposition was consistent with the large degree of surfaceroughness seen in the SEM images (Fig. 11c and d); the loose sparsestructure was easily disturbed by hydraulic washing. The KMnO4
treatment destroyed the organic matter (or scaling), making it easierfor the water to filter into the cake. This indicates that the resistancewas reduced by KMnO4 pretreatment resulting in an increase in thepermeate flux. After 12 h of filtration the membrane that had beenfouled with sand filter effluent exhibited decreased surface roughnessdue to NOM adsorption onto depression areas of the accumulatedfoulant layer (Fig. 11e and f). After the formation of a densely com-pacted gel layer, further accumulation of NOM did not significantlyinfluence surface morphology but merely increased the layer thick-ness until reaching a steady-state was reached between adsorptionand desorption.
SEMmicrographs display the intuitive structure of foulant layer onfouling membranes. It can be seen that there is a more loosely-boundfoulant layer on fouling membrane with KMnO4 preoxidation thanone without preoxidation. It may be attributed to the changing ofstructural composition of foulant layer. Therefore, EDS analysis is neces-sary for the in-depth investigation on foulant layer. A comparison of theEDS analysis of the filtration cake (Fig. 12 and Table 5) after hydraulicwashing, shows that the weight proportions of each element that hadnot undergone KMnO4 pretreatment were C (39.27%), O (26.01%), S(0.81%), Al (3.00%), Cl (27.01%), Ca (0.28%), Fe (0.94%), Si (0.76%), Cu(0.98%), and Zn (0.92%). On the other hand, the proportions of thosethat had undertone KMnO4 pretreatment were C (26.14%), O(28.87%), S (1.13%), Al (3.69%), Cl (22.97%), Mn (5.27%), N (9.65%), Cu(1.03%), and Zn (1.24%). It is apparent that the KMnO4 pretreatmentlessened the CaCO3(s) and CaSO4(s) precipitation that, in the irrevers-ible fouling, formed from pore blockage and, in the reversible fouling,formed from the cake deposition, as indicated by the disappearance ofCa during the EDS analysis. CaCO3(s) and CaSO4(s) adsorbed on themembrane surface readily, resulting in irreversible fouling that wouldbe difficult to be remove by hydraulic washing. MnO2, as the intermedi-ate products, were formed during the oxidation by KMnO4. Particles ofMnO2 formed were very small and unregulated. The particle size ofMnO2 varied from a minimum of 20 nm to a maximum of 100 nm[22]. Therefore, the hydrated manganese dioxide usually existed as acolloid or a suspended solid in aqueous medium. The intermediateMnO2 had a large surface area with strong adsorptive capacity, whichadsorbed some CaCO3(s) and CaSO4(s) that were then retained on themembrane surface to form a loose filter cake. This was later partly re-duced by intermittent hydraulic cleaning. Mn adsorbed on foulingmembrane surface was found due to dosing of KMnO4. When dosingof KMnO4, the chemical reaction generated intermediate products(MnO2), which might adsorb organic or inorganic matter and then beretained on membrane surface to form filter cake. Although the filtercake was washed intermittently by water, there was still part of the
a b
c d
e f
Fig. 11. Scanning electron micrographs of the ultrafiltration hollow fiber membranes tested: (a) new membrane (magnified 20,000); (b) new membrane (magnified 5000);(c) used membrane, with KMnO4 treatment (magnified 20,000); (d) used membrane, with KMnO4 treatment (magnified 5000); (e) used membrane, without KMnO4 treatment(magnified 20,000); and (f) used membrane, without KMnO4 treatment (magnified 5000).
386 T. Lin et al. / Desalination 286 (2012) 379–388
manganese remained on the membrane. The manganese contentremained on cake layer would be gradually increasing with filtration.When filtration was over, manganese is found on the foulant cake.
In this study, Cl was also detected, which was derived from the in-trinsic membrane, while the element Mn may have been derivedfrom the KMnO4 dosing process. It was also noted that Fe disappearedduring the EDS analysis. This could be attributed to the Fe being oxi-dized to metal oxide particles by KMnO4, which was adsorbed duringUF and discharged together with the backwash water. The accumula-tion of the metal oxide particles on the membrane surface may nothave caused severe fouling but have resulted in improved permeability.This is possible because oxide particles can adsorb NOM, which is alsoknown as a foulant. Similar findings by other researchers showed thatmetal oxide adsorption contributed to the reduction in fouling [23,24].Filtration cake, caused by organics adhering to the membrane, had aclose structure and low permeability; membrane fouling caused bymetal ions was mostly irreversible, which could lead to a decrease inpermeate flux. Hydrous manganese dioxide that was converted by
KMnO4, and metal oxide particles that were oxidized by metal ions,could absorb organics andweremore readily attached to themembranesurface than to the membrane pores. This could be due to its volumebeing larger than the membrane pores. During filtration this would re-sult in a reduced level of membrane pore blockage, reducing the proba-bility of direct contact between organics and the membrane. Such aphenomenon would, to some extent, also reduce direct physical orchemical action between the membrane and organics. Membrane foul-ing, caused by adsorption between metal oxide particles and organics,contributed to partial filtration caking, resulting in a loose structureand a higher permeability in comparison to membranes that had notundergone KMnO4 preoxidation.
4. Summary and conclusions
A water treatment system was investigated by pretreating sandfilter effluent with KMnO4 prior to UF. It was found that using theKMnO4/UF process improved the effluent water quality and reduced
Fig. 12. Energy dispersive X-ray spectroscopy analysis of the ultrafiltration hollow fibermembranes tested: (a) new membrane; (b) used membrane with KMnO4; (c) usedmembrane without KMnO4.
387T. Lin et al. / Desalination 286 (2012) 379–388
the rate of UF membrane fouling. This approach thus reduced energyconsumption and improved the water production rate. The optimaldose of KMnO4 was 0.3 mg/L, which is the safety limit for permeate
Table 5Energy dispersive X-ray spectroscopy values obtained with various ultrafiltrationmembranes.
water. The total resistance due to membrane fouling was significantlylower in pretreated membranes than in those that had not been trea-ted with KMnO4. Measurements of permeate flux and TMP of the UFmembrane provided further evidence that pretreatment by KMnO4
improved the UF membrane system. The mechanisms by whichmembrane fouling mitigation was achieved by KMnO4 are describedbelow.
a A decline in the concentration of NOM (a principal cause of mem-brane fouling).
b Alteration to the characteristics of NOM, as well as to hydrophilicand hydrophobic, or molecular weight, distributions; this resultedin the removal of major NOM components responsible for mem-brane fouling.
c Oxidation of metal ions to metal oxide particles that were thenadsorbed on the membrane surface, which reduced the probabili-ty of direct contact between organics and the membrane. Themetal oxide particles contributed to partial filtration caking,resulting in a loose structure and a higher permeability.
Acknowledgments
This work was supported by the Major Science and TechnologyProgram for Water Pollution Control and Treatment (2011ZX07410-001) and the Fundamental Research Funds for the Central Universi-ties (Project 2009B17114).
References
[1] K. Glucina, A. Alvarez, J.M. Laîné, Assessment of an integrated membrane systemfor surface water treatment, Desalination 132 (1–3) (2000) 73–82.
[2] J.A.M.H. Hofman, M.M. Beumer, E.T. Baars, J.P. van der Hoek, H.M.M. Koppers, En-hanced surface water treatment by ultrafiltration, Desalination 119 (1–3) (1998)113–125.
[3] P. Lipp, G. Baldauf, R. Schick, K. Elsenhans, H.-H. Stabel, Integration of ultrafiltra-tion to conventional drinking water treatment for a better particle removal — ef-ficiency and costs, Desalination 119 (1–3) (1998) 133–142.
[4] A. Maartens, P. Swart, E.P. Jacobs, Feed-water pretreatment: methods to reducemembrane fouling by natural organic matter, J. Membr. Sci. 163 (1999) 51–62.
[5] C.F. Lin, T.Y. Lin, O.J. Hao, Effects of humic substance characteristics on UF perfor-mance, Water Res. 34 (4) (2000) 1097–1106.
[6] T.F. Speth, A.M. Gusses, R.S. Summers, Evaluation of nanofiltration pretreatmentsfor flux loss control, Desalination 130 (1) (2000) 31–44.
[7] M.R. Teixeira, M.J. Rosa, pH adjustment for seasonal control of UF fouling by nat-ural waters, Desalination 151 (2) (2003) 165–175.
[8] A.I. Schäfer, A.G. Fane, T.D. Waite, Fouling effects on rejection in the membranefiltration of natural waters, Desalination 131 (1–3) (2000) 215–224.
[9] H. Lee, G. Amy, J. Cho, Y. Yoon, S.H. Moon, I.S. Kim, Cleaning strategies for flux re-covery of an ultrafiltration membrane fouled by natural organic matter, WaterRes. 35 (14) (2001) 3301–3308.
[10] S. Lee, K. Lee, W.M. Wan, Y. Choi, Comparison of membrane permeability and afouling mechanism by pre-ozonation followed by membrane filtration and resid-ual ozone in membrane cells, Desalination 178 (2005) 287–294.
[11] J.J. Qin, M.H. Oo, K.A. Kekre, F. Knops, P. Miller, Reservoir water treatment usinghybrid coagulation–ultrafiltration, Desalination 193 (2006) 344–349.
[12] J.J. Chen, H.H. Yeh, The mechanisms of potassium permanganate on algae removal,Water Res. 39 (2005) 4420–4428.
[13] J.D. Plummer, J.K. Edzwald, Effect of ozone on algae as precursors for trihalomethaneand haloacetic acid, Environ. Sci. Technol. 35 (2001) 3661–3668.
[14] J.D. Plummer, J.K. Edzwald, Effects of chlorine and ozone on algal cell propertiesand removal of algae by coagulation, J. Water SRT-Aqua 21 (2002) 307–318.
[15] Liang Heng, Yang Yanling, Gong Weijia, Li Xing, Li Guibai, Effect of pretreatmentby permanganate/chlorine on algae fouling control for ultrafiltration (UF) mem-brane system, Desalination 222 (2008) 74–80.
[16] D.M. Owen, D.L. Amy, Z.K. Chowdhury, NOM characterization and treatability, JAm Water Works Assoc 87 (1) (1995) 46–63.
[17] H. Haiou, N. Lee, T. Young, et al., Natural organic matter fouling of low-pressure,hollow-fiber membranes: effects of NOM source and hydrodynamic conditions,Water Res. 41 (17) (2007) 3823–3832.
[18] J.O. Kim, E.B. Shin, W. Bae, S.K. Kim, R.H. Kim, Effect of intermittent back ozona-tion for membrane fouling reduction in microfiltration using a metal membrane,Desalination 143 (2002) 269–278.
[19] G.V. Korshin, C.-W. Li, M.M. Benjamin, Monitoring the properties of natural or-ganic matter through UV spectroscopy: a consistent theory, Water Res. 31 (7)(1997) 1787–1795.
[20] C. Lin, T. Lin, O.J. Hao, Effects of humic substance characteristics on UF perfor-mance, Water Res. 34 (4) (1999) 1097–1106.
388 T. Lin et al. / Desalination 286 (2012) 379–388
[21] J.P. Croue, Isolation, fractionation, characterization and reactive properties of naturalorganic matter, Proc. AWWA 18th Federal Convention Adelaide, Australia, 1999.
[22] Lizhu Zhang, Jun Ma, Yu. Min, The microtopography of manganese dioxideformed in situ and its adsorptive properties for organic micropollutants, SolidState Sciences. 10 (2) (2008) 148–153.
[23] Y.J. Chang, K.H. Choo, M.M. Benjamin, S. Reiber, Combined adsorption/UF processincreases TOC removal, J. Am. Water Works Assoc. 90 (5) (1998) 90.
[24] K.W. Lee, K.H. Choo, S.J. Choi, K. Yamamoto, Development of an integrated ironoxide adsorption/membrane separation system for water treatment, Water Sci.Technol. Water (Suppl. 2) (2002) 293.
