Factors affecting membrane coalescence of stableoil-in-water emulsions
A. Honga, A.G. Fanea,∗, R. Burfordb
a UNESCO Centre for Membrane Science and Technology, School of Chemical Engineering and Industrial Chemistry,The University of New South Wales, Sydney, NSW 2052, Australia
b Centre for Applied Polymer Science, School of Chemical Engineering and Industrial Chemistry,The University of New South Wales, Sydney, NSW 2052, Australia
Abstract
This paper presents results on membrane coalescence of stable oil-in-water emulsions. The method involves a two-stageprocess in which the membranes are used to enlarge oil droplets such that a traditional gravity settler can be used for oilphase separation. Several factors that affect the performance of the membrane coalescer have been studied. In particular, theeffect of membrane pore size and the imposed in-pore shear rates on coalescence of stable micron and submicron oil dropletspresent in oil-in-water emulsions were investigated. Hydrophobic membranes made of Teflon (PTFE) with nominal averagepore sizes of 0.22, 0.45, 1.2 and 5.0�m were tested to assess their ability to coalesce oil emulsions with an averaged50
Keywords: Water treatment; Coalescence; Oil emulsion
1. Introduction
Emulsions are homogeneous mixtures that consistof a dispersed phase distributed uniformly in a finelydivided state in a continuous phase. In the case ofoil-in-water emulsions, finely divided oil droplets areuniformly dispersed in water. The presence of suchemulsions in industry fall into two general categories:
existing emulsions, which must be treated or elimi-nated; and emulsions to be made.
This paper focuses on treatment of oil-in-wateremulsions, where large volumes of wastewater con-taining finely dispersed amounts of oil are producedby industries such as the food industry, pharmaceuti-cals, cosmetics, steel works, metal finishing, etc.
Currently, the most popular method commerciallyavailable for the treatment of oil emulsions involveschemical demulsification followed by air flotation.This system requires the use of expensive specialitychemicals and it generates sludge that has to be dis-posed of[1,2]. Several physical methods are also used
20 A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39
for the treatment of oily water in a more conventionalway. Techniques vary from the use of centrifuges,packed beds[3], to electrolytic cells[4,5]. How-ever, the most conventional and simple process is theuse of gravity flotation, but applicability diminisheswhen oil droplets fall below the micron range. Theproblem becomes more pronounced when surfactantsare present which coat the surfaces of oil droplets,increasing the repulsive forces between them. Suchsurfactants are deliberately added to emulsions in themetal working industry to enhance their stability.
Stoke’s law illustrates the importance of droplet di-ameter on phase separation of a mixture. According toStoke’s law, the gravitational separation rate dependson the rising velocity of the dropletsν, which is pro-portional to the square of the droplet diameterd:
ν = d2(ρw − ρo)g
18µ(1)
whereg is the acceleration due to gravity,µ the wa-ter viscosity and (ρw − ρo) the density difference be-tween the water and oil phase, respectively. Therefore,it is advantageous to increase the droplet size of sta-ble emulsions such that the use of gravity separationcan be practical. For example, a droplet size of 1�mand density of 877 kg m−3 has a rise velocity of about6 × 10−8 m s−1, whereas a droplet size of 10�m hasa rise velocity of 6× 10−6 m s−1. In other words, itwill take the 1�m droplet about 93 h to rise 2 cm asopposed to only 0.9 h required for a 10�m droplet torise the same distance.
Several studies have reported on the use of filtrationmembranes for the treatment of oil emulsions, how-ever, most studies have focused on the use of ultrafil-tration [6,7] and microfiltration membranes[8–10] inrejecting oil droplets. Ideally, in such a mode of op-eration, oil droplets are completely retained and thecontinuous phase is permeated. However, because theoil droplets are deformable, depending on the appliedpressure, they can be squeezed through the pores andcontaminate the permeate. Farnand et al.[11] and Lippet al.[12] have both reported such permeate contami-nation. Nazzal and Wiesner[13] also investigated theeffects of transmembrane pressure and membrane poresize. They found that if transmembrane pressures werebelow a critical pressure, emulsion rejection couldbe maximised. Conversely, if the membranes wereused as coalescers, the applied pressure has to be
above the critical pressure, enabling the oil dropletsto wet the membranes and initiate coalescence. Animportant difference between membrane coalescersand rejecting membranes is the potential flux, whichcould differ by an order of magnitude in favour ofcoalescers.
The current project aims to study the use of micro-filtration membranes in coalescing oil droplets suchthat gravity separators could then be used for emul-sion separation. In such a mode of operation, bothdispersed and continuous phase permeate through themembrane. Daiminger et al.[14] and Hlavacek[15]and more recently Kawakatsu et al.[16], reported onthe use of hydrophobic membranes in the treatmentof oil emulsions, but the mechanisms of emulsioncoalescence within the membrane pore are still notfully examined. This study attempts to gain a betterunderstanding of emulsion coalescence within themicrofiltration membrane. In particular, the effects ofmembrane pore size and shear rates within the porehave been investigated.
2. Experimental
2.1. Membranes and materials
Four hydrophobic microfiltration membranes madeof Teflon (Sartorius AG; 11807-142G, 11806-142G,11803-142G, 11842-142G) with nominal pore sizesof 0.22, 0.45, 1.2 and 5.0�m were used (Table 1). Ineach experiment, unless stated otherwise, the mem-brane was wetted with 100 ml of acetone for 1 h andwas then allowed to dry at room temperature for 30 s,to remove surface acetone. Also, in each experiment,unless stated otherwise, the membrane was operatedunder ‘reversed’ orientation (seeSection 3.1).
The acetone used was supplied by Q-Store, Aus-tralia. Oil emulsions were prepared from commercialrolling oil supplied by Worth Recycling Pty Ltd. Theexact composition of the oil is proprietary, however,it contains a blend of polyol ester and triglyceride(mainly derived from vegetable oil and animal fats);<60%, dimer and fatty acids; ethoxlylated and phos-phated non-ionic surfactants, phenolic antioxidant,tolutriazole, and narrow cut paraffinic mineral oil.Milli-Q water with a quality greater than 18 M cm−1
was used for all experiments.
A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39 21
a Calculated usingρTeflon = 2.28 g cm−3 (CRC Handbook of Chemistry and Physics, 81st ed.).b Measured with a micrometer and averaged from six random membrane samples.c Measured at 100 kPa and after membranes were wetted in ethanol for 30 min.
Pore size distributions of membranes before exper-iments were measured by a Coulter Porometer (Coul-ter Electronics Ltd.) which is an automated liquiddisplacement instrument. Before each measurement,membrane samples were soaked in a wetting fluid(Coulter Porofil) with surface tension of 16 mN m−1
for 5 min.
2.2. Apparatus
Experiments were carried out in a laboratory scaledead-end batch cell of 110 ml capacity with a mem-brane area of 15.2 cm2. The cell consisted of a cylindri-cal vessel containing the feed emulsion surmounting asupport on which the membrane was placed. (Fig. 1).
Nitrogen gas was used to pressurise a feed reser-voir, which contained the feed emulsion and wasconnected to the dead-end cell, to start emulsion flow.During experiments where constant flux was required,
Fig. 1. Experimental apparatus.
a pump (Easy-Load Masterflex, Model 7518-00) wasconnected to the permeate line to control the fluxthrough the membrane. After treating 400 ml of feedemulsion, the experiment was stopped and the col-lected permeate was poured into a 500 ml separatingfunnel (flotation vessel). Following 1 h of separationtime (flotation time), a 10 ml sample of the bottomlayer of the separating funnel was collected for oilconcentration analysis.
2.3. Preparation and characterisation of stableoil-in-water emulsion
2.3.1. Emulsion preparationFeed emulsions of different concentrations were
prepared by mixing rolling oil with Milli-Q water.Different mixing equipment was used to identify thebest method to produce a stable emulsion with rel-atively small and narrow droplet size distribution.
22 A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39
Table 2Summary of the mixer types used in the preparation of rolling oil-in-water emulsions
Table 2summarises the mixers and mixing times usedto prepare feed emulsions with varying oil concentra-tions ranging from 0.5 to 5.0 wt.%.
The high shear 1.25 l capacity blender (KambrookKB815C) produced smaller droplets than the Heidolphmechanical stirrer. After 5 min, average droplet sized50 for emulsions prepared with the high shear blenderwere 1.5�m as opposed to 6.0�m obtained from themechanical stirrer.
2.3.2. Emulsion ξ potentialsZeta potentials were measured with a ZetaPlus in-
strument (Brookhaven Instrument, USA) to determinethe surface properties of the emulsion. The emulsionswere diluted up to 1000 times to be within the measur-able range. Theξ potential of the prepared emulsionwas about−46.4 ± 1.3 mV monitored over a periodof 5 days, which indicates strong repulsive forces andsuggests that the oil emulsions were stable[17].
Fig. 2. Interfacial surface tension measurements at different emulsion concentrations;T = 18.5◦C.
2.3.3. Emulsion interfacial surface tensionInterfacial surface tension was measured at different
emulsion concentrations using a tensiometer (Analite2140). Measurements were performed at 18.5◦C andplotted inFig. 2.
2.3.4. Emulsion droplet size distributionsThe droplet size distribution of the feed emulsion
was determined by means of a laser light scatter-ing apparatus (Malvern Mastersizer E) with 45 and100 mm lenses. Feed samples were diluted up to1000 times to be within the measurable range; andpreliminary experiments detected no alteration in thesize distribution after dilution. Droplet distributionsobtained with the 100 mm lens demonstrated a limi-tation towards detection of smaller droplets. This wasexpected since the quoted size range for the 100 mmlens is between 0.5 and 180�m whereas the range for45 mm lens is 0.1–80�m. During emulsion droplet
A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39 23
Fig. 3. Typical prepared feed emulsion droplet size distribution. Rolling oil-in-water 0.5, 1.0 and 5.0 wt.%. Measured with 45 mm lens.
size measurements, it was crucial that the correct focallenses were selected prior to each measurement. Thus,to accurately measure untreated feed distributions, the45 mm focal lens was required and to accurately mea-sure the permeate samples, the 100 mm focal lens wasmore appropriate. The 45 mm lens tends to underesti-mate droplets bigger than 80�m whereas the 100 mmlens tends to underestimate droplets below 0.5�m.The problem arises when feed and permeate dropletsize distributions have to be compared but differentlenses are indicated. Therefore when feed and perme-ate droplet size distributions have to be compared, the100 mm lenses was chosen for consistency purposes.
Fig. 3 shows typical droplet size distributions ofthe feed emulsions (measured with 45 mm lenses) at0.5, 1.0 and 5.0 wt.% oil prepared with the high shearblender. The average droplet sized50 (volume based)was consistently 1.5 ± 0.5�m over the concentrationrange. Since the bulk emulsion droplet sizes were be-low 2.0�m and were also highly charged the emul-sions were considered very stable (seeSection 2.3.6).
The droplet coalescence on passage through themembrane, meant that a direct and accurate measure-ment of the permeate droplet size distribution was notpossible due to instantaneous free oil formation thatcaused sampling difficulties. Nevertheless, permeatedroplet size distributions were measured by dividing
it into four separate fractions (bottom, middle, top andfree oil), after it was left to stand in the flotation vesselfor 1 h. Only the droplet size distributions of the bot-tom, middle and top fractions were measured.Fig. 4illustrates the sampling involved.
2.3.5. Oil content analysis and membranecoalescence performance
Methods for oil content analysis of emulsions variessignificantly and depends greatly on the type and theamount of oil present in the permeate.
Fig. 4. Schematic diagram of bottom, middle, and top fractionstaken for sampling.
24 A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39
Several methods have been used by other resear-chers: atomic absorption spectrophotometer (AAS)[10,18]; total organic carbon[7,13,19,20]; ultra violetspectrophotometer (UV)[5]; turbidity meter[21,22];gas chromatography (GC)[14]; freeze drying[16];and standard methods involving solvent extraction[15]. Each of these methods has its advantages anddisadvantages, but a detailed discussion is beyond thescope of this paper.
The quantification of oil content of the treated emul-sion in our studies was difficult due to the formation offree oil on the samples. Therefore, three different char-acterisation methods were compared; TOC, turbidity,and gravimetric analysis, such that the best techniquesuitable for our studies could be identified.
Gravimetric analysis was performed by collecting aknown amount of the sample into a pre-weighed cleansample tube. The emulsion was heated in a 85◦C wa-ter bath until all water evaporated. Acetone was con-stantly added to increase the solubility of surfactantsinto water to allow faster demulsification. After all thewater and acetone was evaporated, the weight of oilcontent was measured and separation efficiency cal-culated (Eq. (2)). TOC was measured by a SK12 totalorganic carbon analyser (Skalar, The Netherlands) andturbidity was measured with a Hanna (HI 93793-11)turbidity meter. TOC and turbidity samples both re-
Fig. 5. Comparison of measured oil content from gravimetric analysis with oil content from prepared standards.
quired dilution up to 1000 times to be within the mea-surable range.
The membrane coalescence performance was mea-sured as the separation efficiencyη, obtained by mea-suring the oil concentration in the lower layer after 1 hseparation of membrane coalesced fluid in a flotationvessel. The separation efficiency is calculated by
η =
1 −
oil concentration in lowerlayer of flotation vessel
oil concentration in feed
× 100 (2)
Samples of feed and permeate were collected for anal-ysis. After 1 h of settling, 10 ml of the lower phase wascollected for oil content analysis. Ten millilitres feedsamples were collected at the start of the experimentand analyzed as is. We appreciate that a process plantbased on membrane coalescence would probably havea continuous gravity flotation vessel rather than thebatch unit we have used. However, the batch flotationhas been applied here since we are attempting to com-pare the effects of membrane type and operation onthe coalescence per se. Direct measurement of dropletsize distributions from membrane permeate sampleswere very unreliable due to ‘creaming’. Furthermore,it should be noted that batch tests are commonly usedin the design of continuous flow chambers (based
A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39 25
Fig. 6. Separation efficiencies obtained from different analytical methods and compared with gravimetric analysis.
on surface overflow rate, vessel depth and retentiontime).
The measurement of oil content was best withgravimetric analysis.Fig. 5 compares the mea-sured oil content from gravimetric analysis with
Fig. 7. Separation efficiency of feed emulsion without membrane coalescence compared with separation efficiency obtained with 0.45�mmembrane operated at 150 kPa.
prepared emulsion standards of known oil concen-tration. Oil content measurements to an accuracy>99% were obtained, and therefore this methodwas used to measure the oil concentration in ourexperiments.
26 A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39
Fig. 6compares results obtained from the other tec-hniques applied. TOC provided inconsistent data andthe results were very scattered when plotted againstgravimetric analysis data and no particular trend wasobtained. The turbidity meter also gave scattered databut it showed a better linear relationship (R2 = 0.88)with the data obtained from gravimetric analysis. Insummary, both turbidity and TOC gave inconsistentresults, although turbidity could be used qualitativelyfor a reasonable estimate of general trends.
2.3.6. Emulsion stabilityAlthough the prepared feed emulsions were stable
and their droplet size distributions did not change overseveral days, it was noted that there was a tendencyfor the emulsions to cream due to buoyancy effects.Calibration tests indicated that our emulsions requiredseveral hours before they started to cream.
Fig. 7shows the separation efficiency, based on theoil concentration at the bottom of the flotation vessel,over time with and without membrane coalescence.Within 1 h the separation efficiency without membranecoalescence was only 5% as opposed to 76% when a0.45�m membrane was used at 150 kPa. The separa-tion efficiency obtained from the 0.45�m membranedid not change with creaming time. Higher separa-tion efficiencies of about 40% were obtained withoutmembrane coalescence after 24 h of creaming time.However, flotation times of 24 h are not really prac-tical, and the ‘creamed’ layer would still be mainlydispersed droplets rather than coalesced oil.
Fig. 8. FESEM photograph at 3500 magnification of new 1.20�m PTFE membrane from A: reversed orientation and B: normal orientation.
3. Results and discussion
3.1. Effect of membrane orientation
PTFE 0.45 and 1.2�m membranes were used un-der normal and reversed orientation. The membraneis operated under normal orientation if the topside ofthe membrane (as indicated by the supplier) is usedfor the feed side stream. Reversed orientation mode iswhen the bottom side of the membrane is used for thefeed side stream.
Fig. 8 shows photographs of a new 1.2�m mem-brane at normal and reversed orientation taken with afield emission scanning electron microscope at 3500magnification. It can be seen that the pores on the bot-tom side of the membrane were more stretched thanon the topside and the porosity on the bottom sideappears to be much higher than the topside. This isalso evident in the porosimeter results (Fig. 9), whichprovide different size distributions depending on theorientation of the membrane; larger pores are detectedfor the reverse orientation. The fact that the porosime-ter provides different pore sizes depending on orien-tation may be due to the effect of the different pore‘structure angles’ presented by the two orientations[23].
Operating the membrane under reversed orientationresulted in higher permeate fluxes and better separa-tion efficiencies.Fig. 10shows that with the 0.45�mmembrane at 120 kPa, permeate flux dropped from350 to about 200 l m−2 h−1 and separation efficiency
A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39 27
Fig. 9. Pore size distributions for PTFE 0.45�m, and PTFE 1.20�m membranes obtained under normal and reversed orientation.
was 72% when operated under normal orientation.However, when operated under reversed orientationand at the same pressure, permeate flux reached430 l m−2 h−1 and stabilised at about 400 l m−2 h−1
with separation efficiency of 75%.
Fig. 10. Permeate flux for emulsion treated with 0.45�m membrane at 120 kPa under normal and reversed orientation. No pre-treatment.
Fig. 11 shows that with the 1.2�m membrane at100 kPa, permeate flux dropped from 750 to about100 l m−2 h−1 and separation efficiency was 58%when operated in normal orientation. However, whenoperated in reversed orientation and at the same
28 A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39
Fig. 11. Permeate flux for emulsion treated with 1.20�m membrane at 100 kPa under normal and reversed orientation. No pre-treatment.
pressure, the initial permeate flux was 1700 l m−2 h−1
stabilising at about 1000 l m−2 h−1 with separationefficiency of 66%.
The difference in porosity between the top and bot-tom part of the membrane may be due to differentialstretching occurred during the membrane manufactur-ing process. When the membrane is used in reversedorientation, the pore contours through the membraneprovide a conical funnel flow path that improves coa-lescence and hence separation efficiency.
Figs. 10 and 11also shows some interesting resultson the flux history during membrane coalescence op-eration. During operation under reversed orientation(for both membranes), there is a rise in permeate fluxfrom the gradual opening of pores until a maximumflux value is reached (or all pores are open), and after-wards a slow decline in flux is observed. However, asimilar initial flux rise is not observed for membraneswhen operated during ‘normal’ orientation. During op-eration in the normal orientation, the maximum initialflux value and the subsequent flux decline is reachedfaster than for reversed orientation membranes.
It is important to note that concentration factors of1.00 were obtained throughout our experiments in-dicating that neither oil rejection nor concentrationpolarisation occurred, and in theory no flux declineshould have been observed. However, it is possible
that a thin film of oil may have formed due to the in-teractions between the oil droplets and the membranesurface during the passage of oil through the mem-brane. Droplet coalescence may occur at the surfaceof the membrane as well as inside the membrane pores(see later for discussion ofFig. 14). The coalescenceof droplets at the membrane surface would result inthe localised formation of a thin oil layer that wouldreduce the total number of available pores for per-meate flow, resulting in lower permeate flux. Whensteady state is reached, the thin oil layer slowly per-meates through the pores via surface flow due to thehydrophobic nature of both oil droplets and membranepores. During reversed membrane operation, wherepore openings are larger than during normal operation,oil droplets are able to reach into the membrane poreswithout substantial film formation. In this condition,there is less reduction in the total number of availablepores exposed in the membrane surface, which resultsin higher permeate fluxes and steady state is reachedwithout significant flux decline.
3.2. Effect of pore size and appliedoperating pressure
Fig. 12 shows the effect of membrane pore sizeand applied pressure on coalescence performance.
A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39 29
Fig. 12. Effect of membrane pore size and applied pressure on coalescence performance.
A general trend observed is that as applied pressureincreased, coalescence performance improved gradu-ally for the 1.2 and 5.0�m membranes. An increasein pressure drop across the membrane would increaseflux and in-pore shear. This would enhance wettingand coalescence of the droplets since there are greaterapplied forces pushing droplets together or against
Fig. 13. Volume frequency droplet size distribution of top fraction permeate after treatment with 1.20�m membrane. Measurement with100 mm lens.
the pore surface at higher pressures. The separationefficiency reaches a maximum point (as indicatedin Fig. 12) and then remains constant over a certainrange of pressure, which we call the ‘critical range’.The gradual improvement in coalescence at low pres-sures was not observed with either the 0.22�m or the0.45�m membrane. The critical range has already
30 A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39
been reached at 100 kPa for these membranes andfor the 0.22�m membrane the separation efficien-cies started to drop gradually as applied pressurewas increased. Further investigation of the separationefficiency for the 0.22 and 0.45�m membranes atpressures below 150 and 100 kPa, respectively wasnot possible due to the local capillary pressures in-side the membrane pores acting against feed flow andpreventing permeation.
The effect of transmembrane pressure on coa-lescence phenomena for the 1.2�m membrane isillustrated inFig. 13 which shows volume frequencydroplet size distributions from a top fraction sample(Fig. 4) of permeate at different transmembrane pres-sure drops. After 1 h flotation time, a bimodal dropletsize distribution was obtained. As the pressure wasincreased from 150 to 500 kPa, the volume frequencyof droplets from the lower end of the distribution de-creased and the volume frequency of droplets in thehigher end of the distribution gradually increased.
Examination of the membrane surface after use alsoprovides insights into the coalescence process.Fig. 14shows scanning electron micrographs of new and used1.2�m membranes. Figures A, B, C, and D correspond
Fig. 14. FESEM photographs of PTFE 1.20�m membranes at3500 magnification: (A) new membrane; (B) used at 60 kPa; (C)used at 80 kPa; (D) used at 100 kPa.
to a new membrane, used at 60 kPa, used at 80 kPa,and used at 100 kPa, respectively The photographs in-dicate that the extent of the oil film on the membraneincreased as transmembrane pressure drops were in-creased.
The increased oil coverage with applied pressure isthought to be linked to the increased flux of oil dropletsto the membrane surface. The increased flux of oil intothe membrane as pressure is raised, would increase lo-cal in-pore shear forces overcoming the mutual repul-sion of the charged oil droplets favouring coalescence(seeSection 3.3). Furthermore, at increased appliedpressures, preferential phase flow within membranepores may occur. For the smaller pores in the distri-bution, partial reduction of oil flow could take placeallowing preferential waterflow. The oil phase wouldpreferably flow into the bigger pores, which wouldtherefore tend to increase the bulk-pore oil concentra-tion. The increased local concentration would increasethe probability of droplet collision and coalescence sothat oil droplets grow in size until they are releasedout of the porous matrix.
As the pressure is increased further the critical rangeis reached and a different phenomenon occurs. Themembranes used have a distribution of pores (Fig. 9),and it is postulated that over the critical range the im-proved coalescence obtained from the higher pressuresis counter balanced by the rate of oil droplet break-upat the higher shear conditions encountered in the largerpores for the same membrane (seeSection 3.3). Thus,the rate of oil droplets coalescence would be balancedby the rate of oil droplet break-up resulting in con-stant separation efficiency as the pressure is increased(Fig. 12).
A corollary of this is that at even higher operatingpressures, the rate of oil droplet breakage could ex-ceed the rate of oil droplet coalescence resulting in adeteriorating membrane coalescence performance andconsequently lower separation efficiencies. There isevidence for this inFig. 12which shows a decrease inseparation efficiency as applied pressure drop was in-creased during treatment with the 0.22�m membranesbased on replicate experiments. Kawakatsu et al.[16]reported a similar trend for membrane coalescence ofsunflower emulsion. Using a 5.0�m pore size PTFEmembrane they observed similar coalescence perfor-mance at 25 and 50 kPa but less at 100 kPa. No suchtrend was observed with a 10�m membrane. Although
A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39 31
the results of Kawakatsu et al.[16] show similar trendsto this work it is likely that our emulsions were smallerand more stable (commercial rolling oil versus sun-flower oil/surfactant mixture); this may explain ourobservations requiring the 0.22�m pore size mem-brane. The high flowrates obtained at increased pres-sures would result in increased shear rates in the poreswhich could break and re-dispersed some of the oildroplets, reducing the performance of membrane coa-lescence. The effect of in-pore shear rates is discussedin more detail inSection 3.3.
Fig. 12also shows a significant increase in coales-cence performance as membrane nominal pore sizeis reduced. Depending on the operating conditions,the 0.22�m membranes showed separation efficien-cies of up to 86% whereas the 0.45�m membranesshowed separation efficiencies up to 80%, the 1.2�mshowed separation efficiencies of up to 63% and the5.0�m showed up to 30% separation. A similar poresize trend was observed by Kawakatsu et al.[16]who found an improvement in separation efficiencyfor sunflower emulsions as the PTFE membrane poresize decreased from 10.0, 5.0, to 1.0�m, respectively.In contrast, Daiminger et al.[14] reported that withPTFE membranes of 0.22, 1.0 and 5.0�m, mem-brane coalescence performance deteriorated as thepore size decreased. A probable explanation for thisdifference is the initial average feed droplet size ofabout 5.0�m used by Daiminger et al.[14]. Such afeed emulsion with a relatively large and wide dropletsize distribution would have experienced significantdroplet breakage and re-dispersion of the oily emul-sion when pushed through the smaller 0.22 or 1.0�mpores.
The results demonstrate that pore size significantlyaffect coalescence performance. It suggests that theremust be an optimal relationship between feed dropletsize distribution and pore size distribution for coales-cence to take place. Although our data shows that coa-lescence improves as pore size decreases, it is intuitivethat a limiting pore size exists for coalescence to takeplace. If membrane pore size is too big compared tothe feed droplet size, no coalescence will take place.However, if pore sizes are too small, droplet break-age and re-dispersion may be favoured and eventuallythe membrane will stop functioning as a coalescer andwill start rejecting oil droplets instead of coalescingthem (e.g. Lipp et al.[12]).
3.3. Effect of shear rates
In Section 3.2, where the effect of membrane poresize was investigated, it was postulated that shearrates inside the membrane pore play a crucial rolein membrane coalescence. At low shear rates, verylittle or no coalescence would take place. In contrast,high shear rates would eventually break up the oildroplets. Between these extremes there must be an op-timum shear rate such that membrane coalescence isfavoured.
An analogous situation occurs for the flocculationof particles in stirred tanks. Orthokinetic flocculationtheory [24] states that flocculation of particles is de-pendant on collision efficiencyϕ, applied shear rateγ,particle diameterD and the total number of particlesN present in the system, i.e.
�N
�t= −2ϕγD3N2
3(3)
According to Eq. (3) the rate of coalescence isincreased as shear rateγ increases. However, inpractice there is an upper limit onγ to avoid flocbreakage. To estimate the in-pore shear rates in ourexperiments, it is assumed that our feed emulsionbehaves like a Newtonian fluid, that the flow insidethe membrane pore is laminar and the feed emul-sion behaves as an homogeneous fluid (1 wt.% feedemulsion). The in-pore shear rateγ is then calculatedby
γ = 8u
dp(4)
whereu is local velocity inside the membrane poresanddp the membrane pore size. The local velocity in-side the membrane pore can be determined accordingto the Hagen–Poiseuille model:
u = d2p
32µ
(dP
dx
)(5)
where (dP/dx) is the pressure gradient across the mem-brane andµ the viscosity of the solvent. Accordingly,permeate fluxJ can also be described given the mem-brane porosity,ε:
J = εd2p
32µ
(dP
dx
)(6)
32 A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39
Fig. 15. Effect of in-pore shear rates on membrane coalescence.
CombiningEqs. (4) and (5), for a given pressure drop,the in-pore shear rate can be expressed as
γ = dp
4µ
(dP
dx
)(7)
Rearranging the above equations, the average in-poreshear rate can be described as
γ = 8
(J
ε
)1
dp(8)
Eq. (7) shows that for a given pressure drop theaverage in-pore shear rate increases with pore size,whereas according toEq. (8)for a given imposed fluxthe shear rate increases with decrease in pore size.Fig. 15 presents separation efficiency versus in-poreshear rates for experiments at constant flux. The shearrates are calculated fromEq. (8) using the imposedfluxes and membrane properties given inTable 2.The separation efficiency gradually increased for allmembranes used until it reached a constant valuefrom which it did not increase further regardless ofthe amount of shear imposed. The results agree withthe trends discussed earlier on the effect of increasedoperating pressure on membrane coalescence.
Fig. 16 shows volume-based droplet size distribu-tions of the bottom fraction of the flotation vesselafter being treated with a 5.0�m membrane. After
the emulsion was treated by the membrane, there wasa small shift in droplet size towards the lower end ofthe distribution. These distributions, however, did notappear to change as the pressures (or imposed in-poreshear rates) were increased from 150 to 300 kPa.Fig. 17 shows the same distributions presented inFig. 16 but converted into number frequency. Thefigure shows a slight increase in the number of oildroplets in the lower end of the distribution, suggest-ing that some droplet breakage may have occurredduring the process. It is important to note that thenumber frequency and the actual number of dropletspresent in the system would show similar trends be-cause bottom sample concentrations are similar ateach pressure above 100 kPa (see separation efficien-cies inFig. 12).
During emulsion treatment with 1.2�m mem-branes, different trends in the droplet size distributionswere obtained.Fig. 18 shows volume-based dropletsize distributions of permeate bottom fractions aftertreatment with the 1.2�m membrane. When operatedat 150 and 400 kPa, similar droplet size distributionswere obtained and both indicated a shift in dropletsize towards the lower end of the distribution similarto results obtained with 5.0�m membranes. However,at a higher pressure of 500 kPa, a bimodal distributionwas obtained, and larger oil droplet diameters were
A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39 33
Fig. 16. Volume frequency droplet size distribution of bottom fraction permeate after treatment with 5.0�m membrane. Measurement with100 mm lens.
detected. When the same distribution is presentedas number frequency (Fig. 19), operating pressuresfrom 150 to 500 kPa all produced distributions wherethe number of droplets in the smaller end increased,but operation at 500 kPa showed the highest increase
Fig. 17. Number frequency droplet size distribution of bottom fraction permeate after treatment with 5.0�m membrane. Measurement with100 mm lens.
in droplets in the lower end of the distribution, sug-gesting that more droplets were broken up at thishigh pressure. However, what is not evident is that atthe higher pressure there was a significant bimodaleffect on a volume-basis (Fig. 18) not detected on
34 A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39
Fig. 18. Volume frequency droplet size distribution of bottom fraction permeate after treatment with 1.2�m membrane. Measurement with100 mm lens.
a number-based (Fig. 19) presumably because therewere only a few large drops.
A different scenario occurred when 0.45�m mem-branes were used.Fig. 20 shows the volume-baseddroplet size distribution when pressure was variedbetween 150 and 500 kPa. A bimodal distribution was
Fig. 19. Number frequency droplet size distribution of bottom fraction permeate after treatment with 1.2�m membrane. Measurement with100 mm lens.
obtained at the lowest pressure of 150 kPa, indicatingan intermediate step in membrane coalescence whereboth droplet break-up and coalescence is taking place.At higher pressure drops, the bimodal distribution isnot as pronounced and there is a significant shift indroplet size towards the higher end of the distribution,
A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39 35
Fig. 20. Volume frequency droplet size distribution of bottom fraction permeate after treatment with 0.45�m membrane. Measurementwith 100 mm lens.
suggesting that coalescence was favoured as pres-sure drop was increased with the 0.45�m membrane.When the number-based distribution is plotted asnumber frequency inFig. 21, the larger oil dropletsencountered in the volume-based distribution are not
Fig. 21. Number frequency droplet size distribution of bottom fraction permeate after treatment with 0.45�m membrane. Measurementwith 100 mm lens.
identified, presumably because the number of largecoalesced droplets is not significant. However, froman application point of view, the large coalesceddroplets represents the bulk of the system. There isstill a significant amount of small oil droplets present
36 A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39
in the system, and the number was highest whenthe operating pressure was only 150 kPa. When themembrane was operated at 400 or 500 kPa, similardistributions were obtained but they were both higherthan distributions obtained when the membrane wasoperated at 150 kPa. This indicates that when the0.45�m membrane was used at high shear rates, therate of droplet coalescence was dominant over therate of droplet breaking up. At lower shear rates, par-tial droplet breakage still occurs (although lessened),however, droplet coalescence is also reduced whencompared to operation at higher in-pore shear rates,since the in-pore shear rates (at lower pressure) arenot sufficiently high to overcome the repulsive forcesbetween the oil droplets.
Fig. 22shows feed emulsion droplet size and mem-brane pore size distributions. Due to the distributionof droplet diameters in the feed emulsion and the dis-tribution of pore sizes present in the membranes, aswe force oil droplets to flow through smaller pores,there will be inevitably some droplet breakage withinall the membranes used.Eq. (7) implies that for a
Fig. 22. Membrane pore size distributions for new PTFE 0.45, 1.20 and 5.00�m at reversed orientation. Feed 1.0 wt.% emulsion dropletsize distributions (number and volume based).
given membrane, where a distribution of pores ex-ists, the larger pores will have higher shear rates fora given pressure drop. However, for different mem-branes, and a given flux, the large pore membranewill have lower in-pore shear rates (assuming similarporosity) as described inEq. (8).
It is clear fromFig. 15 (separation efficiency ver-sus shear rate at different pore size) and the dropsize data inFigs. 16–21that the 0.45�m membranewas a more effective coalescer at a given shear rate.As presented inFig. 22, the 5.0�m membrane haspores substantially larger than the majority of thedroplets which could permit passage without signif-icant droplet–membrane interaction. However, the1.20 and 0.45�m membranes have pore sizes whichoverlap with the drop size distribution, which meansthat droplet–membrane interactions would be in-evitable. This suggests that coalescence could be as-sisted by droplet deposition and surface flow over theinternal surface of the membrane. Thus the mecha-nisms for coalescence involve drop–drop interactions(in the fluid stream) due to shear and surface–drop
A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39 37
Fig. 23. Effect of emulsion concentration on membrane coalescence.
interactions due to pore size effects. The role of thesemechanisms will ultimately depend on the droplet tomembrane pore size ratio.
3.4. Effect of feed emulsion concentration
Oil emulsions with oil concentrations of 0.5, 1.0and 5.0 wt.% were made to investigate the effect offeed concentration on membrane coalescence.Fig. 23shows the separation efficiencies obtained with a5.0�m membrane operated over a range of trans-membrane pressures.
Fig. 23shows that doubling the concentration from0.5 to 1.0 wt.% was not sufficient to provide a no-ticeable improvement in membrane coalescence andseparations were within experimental data deviation.However, when emulsions of 5.0 wt.% were treated,a clear improvement of about 10% in separation ef-ficiency was obtained compared to the more diluteemulsions over the range of operating pressures.
This is a result anticipated byEq. (3)for drop–dropcoalescence in the fluid. At higher feed concentrationthere is an increased number of oil droplets presentin the system and consequently, there is an increasednumber of oil droplets present inside the membranepores. As explained inSections 3.1 and 3.3, a higherconcentration of oil droplets inside the membranepores should lead to better membrane coalescence.Oil droplet coalescence improved with increasingin-pore shear rates. Such improvement in coalescence
will tend to occur in the larger pores of a given mem-brane, and consequently, the amount of oil in thesmaller pores present in the distribution is diminishedsuch that the droplet population and concentrationincreases in the larger pores. At higher concentra-tions, the probability of oil droplets colliding witheach other or colliding with the pore surface in-creases, improving the chances for coalescence to takeplace.
4. Conclusions
Stable oil-in-water emulsions can be effectivelydestabilised by passage through hydrophobic stretchedPTFE membranes. However, membrane and operatingparameters play important roles in performance.
Membrane orientation (i.e. ‘as received’ or re-versed) has a noticeable effect with higher fluxesand better coalescence performance for the reversedorientation. Electron microscopy and porosimetry re-vealed a higher porosity and pore size for the surfaceof reversed membranes.
Coalescence performance is strongly dependent onmembrane pore size (varied from 0.22 to 5.0�m)and imposed in-pore shear rates. Results indicate thatthere has to be an optimal in-pore shear rate such thatit will enhance droplet coalescence opposed to dropletbreak-up. The measured droplet size distributionscompared with the pore size distributions suggest that
38 A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39
when emulsions are forced through the membranes,three mechanisms could occur. The magnitude ofthe in-pore shear rates and the oil droplet to mem-brane pore size ratios will determine droplet breakup,droplet coalescence from drop–drop interactions dueto shear, and droplet coalescence from drop–surfaceinteractions due to pore size effects.
The 5.0�m membranes were the least efficient inpromoting droplet coalescence, and separation ef-ficiencies were only about 45% presumably fromdrop–drop coalescence due to shear. Separation ef-ficiencies of up to 66% were obtained with 1.2�mmembranes presumably from drop–drop coalescencedue to shear and partially from droplet coalescencedue to drop–surface interaction. 0.45�m membranesproduced better performance where the rate of dropletcoalescence was significantly higher than the rateof droplet break-up within the pore, separation effi-ciencies achieved were up to 81% presumably fromcoalescence due to both drop–surface and drop–dropinteractions. The best performance was achievedwith the 0.22�m membranes where separation effi-ciency of 86% was obtained at low operating pres-sures also presumably from coalescence due to bothdrop–surface and drop–drop interactions. However,at higher operating pressures with the 0.22�m mem-brane, droplet break-up becomes more pronouncedthan droplet coalescence and separation efficiencystarted to decrease.
The potential for using membranes to coalesce fineoil droplets present in stable oily emulsions is signif-icant. The need for expensive speciality chemicals iseliminated and separation efficiencies of up to 86%can be achieved at relatively high flux (330 l m−2 h−1)with 0.22�m membranes with only 1 h flotation time.
Acknowledgements
The authors would like to thank the CRC for WasteManagement and Pollution Control for providing ascholarship and project funding to A. Hong. SartoriusAG is acknowledged for providing the membranesused in this study. Worth Recycling Pty Ltd. is ac-knowledged for providing rolling oil used in thisstudy. Also Ashlee Goh is acknowledge for assist-ing with some of the experiments presented in thispaper.
Nomenclature
d oil droplet diameter (m)dp manufacture’s nominal membrane
pore size (m)dP/dx pressure drop across membrane
(kg (m s)−2)D particle diameter (m)g acceleration due to gravity (m s−2)J permeate flux (l m−2 h−1)N total number of particles�N/�t change in total number of particles
efficiency (%)µ water viscosity (kg (m s)−1)ν oil droplet rising velocity (m s−1)ξ rolling oil-in-water emulsion zeta
potential (mV)ρo density of oil (kg m−3)ρw density of water (kg m−3)ϕ particle collision efficiency (%)
References
[1] V.B. Menon, D.T. Wasan, Demulsification, in: P. Becher (Ed.),Encyclopedia of Emulsion Technology, Marcel Dekker, NewYork, 1985, p. 1–75.
[2] A.D. Nikolov, M. Randie, C.S. Shetty, D.T. Wasan, Chemicaldemulsification of oil-in-water emulsion using air flotation:the importance of film thickness stability, Chem. Eng.Commun. 152–153 (1996) 337–350.
[3] S. Dezhi, J.S. Jong, X. Duan, D. Zhou, Demulsification ofwater-in-oil emulsion by wetting coalescence materials instirred- and packed-columns, Colloids Surf. A: Physicochem.Eng. Asp. 150 (1–3) (1999) 69–75.
[4] T. Kataoka, T. Nishiki, Development of a continuous electriccoalescer of W/O emulsions in liquid surfactant membraneprocess, Sep. Sci. Technol. 25 (1–2) (1990) 171–185.
[5] H.M. Huotari, I.H. Huisman, G. Tragardh, Electricallyenhanced crossflow membrane filtration of oily waste waterusing the membrane as a cathode, J. Membr. Sci. 156 (1)(1999) 49–60.
A. Hong et al. / Journal of Membrane Science 222 (2003) 19–39 39
[6] N. Nabi, P. Aimar, M. Meireles, Ultrafiltration of an olive oilemulsion stabilized by an anionic surfactant, J. Membr. Sci.166 (2) (2000) 177–188.
[7] P. Srijaroonrat, E. Julien, Y. Aurelle, Unstable secondaryoil/water emulsion treatment using ultrafiltration: foulingcontrol by backflushing, J. Membr. Sci. 159 (1) (1999) 11–20.
[8] P. Wang, N. Xu, J. Shi, A pilot study of the treatment of wasterolling emulsion using zirconia microfiltration membranes, J.Membr. Sci. 173 (2000) 159–166.
[9] T.C. Arnot, R.W. Field, A.B. Koltuniewicz, Cross-flow anddead-end microfiltration of oily-water emulsions. Part II.Mechanism and modelling of flux decline, J. Membr. Sci.169 (2000) 1–15.
[10] I.W. Cumming, R.G. Holdich, I.D. Smith, The rejection ofoil by microfiltration of a stabilised kerosene/water emulsion,J. Membr. Sci. 169 (1) (2000) 147–155.
[11] B.A. Farnand, H. Sawatzky, M.A. Poirier, An evaluation ofthe use of porous membranes for the dewatering of wellheadbitumen/water/mineral emulsions, Sep. Sci. Technol. 20 (2–3)(1985) 193–203.
[12] P. Lipp, C.H. Lee, A.G. Fane, C.J.D. Fell, A fundamentalstudy of the ultrafiltration of oil-water emulsions, J. Membr.Sci. 36 (1988) 161–177.
[13] F.F. Nazzal, M.R. Wiesner, Microfiltration of oil-in-wateremulsion, Water Environ. Res. 68 (7) (1996) 1187–1191.
[14] U. Daiminger, W. Nitsch, P. Plucinski, S. Hoffmann, Noveltechniques for oil/water separation, J. Membr. Sci. 99 (2)(1995) 197–203.
[15] M. Hlavacek, Break-up of oil-in-water emulsions induced bypermeation through a microfiltration membrane, J. Membr.Sci. 102 (1995) 1–7.
[16] T. Kawakatsu, R.M. Boom, H. Navetani, Y. Kikuchi,M. Nakajima, Emulsion breakdown: mechanisms anddevelopment of multilayer membrane, AIChE J. 45 (5) (1999)967–975.
[17] P. Becher, Emulsions: theory and practice, second ed.,Reinhold, New York, 1965.
[18] I.W. Cumming, R.G. Holdich, I.D. Smith, The rejection ofoil using an asymmetric metal microfilter to separate anoil in water dispersion, Water Res. 33 (17) (1999) 3587–3594.
[19] T. Leiknes, M.J. Semmens, Membrane filtration forpreferential removal of emulsified oil from water, in:Proceedings of the International Specialized Conferenceon Membrane Technology in Environmental Management,IAWQ, Tokyo, Japan, 1999.
[20] H. Ohya, J.J. Kim, A. Chinen, M. Aihara, S.I. Semenova,Y. Negishi, O. Mori, M. Yasuda, Effects of pore size onseparation mechanism of microfiltration of oily water, usingporous glass tubular membrane, J. Membr. Sci. 145 (1998)1–14.
[21] B. Tansel, J. Regula, R. Shalewitz, Treatment of fuel oil andcrude oil contaminated waters by ultrafiltration membranes,Desalination 102 (1995) 301–311.
[22] G. Rios, C. Pazos, J. Coca, Destabilization of cutting oilemulsions using inorganic salts as coagulants, Colloids Surf.A: Physicochem. Eng. Asp. 138 (2–3) (1998) 383–389.
