analysis of droplet size during crossflow membrane emulsification using stationary and vibrating...

Journal of Membrane Science 261 (2005) 136–144

Analysis of droplet size during crossflow membrane emulsification usingstationary and vibrating micromachined silicon nitride membranes

Jie Zhu, David Barrow∗

Laboratory for Applied Microsystems, School of Engineering, Cardiff University, 1-5 The Parade, Cardiff CF24 3AA, UK

Received 1 July 2004; received in revised form 1 December 2004; accepted 21 February 2005Available online 21 April 2005

Abstract

Crossflow membrane emulsification is a promising method to achieve very small and uniform emulsions. The droplet size produced iscontrolled mainly by the choice of membrane. Using microengineering technology it is currently possible to produce membranes with precisiondefined parameters (uniform pore size, shape and inter-pore distance). In the work presented here, individual pore behaviour was studied usingmicromachined membranes with wider inter-pore distances (100�m). It was found that the diameter of droplets increased during an initialperiod of operation. Also, interaction between droplets formed at adjacent pores was observed to enhance the reduction of mean droplet sizea ved. It wasd anf ontrol overd e producedb©

K

1

nastmmeoivnip

ze isak-

in therousylin-

eenution

h dif--

thed to

i-e, re-

0d

nd negatively correlated with inter-pore distance. A ‘push-to-detach’ mechanism was proposed to explain the behaviour obseremonstrated that a micromachined membrane with pore diameter of 2�m and inter-pore distance of 20�m produced smaller droplets th

or membranes with larger inter-pore distances. To facilitate the droplet detachment from the membrane and provide additional croplet detachment, the effects of membrane vibration were investigated. Preliminary results showed that smaller droplets could by introducing low frequency (0–100 Hz) membrane vibrations without increasing their size distribution.2005 Elsevier B.V. All rights reserved.

eywords:Emulsification; Piezoactuated membrane; Micromac hined membrane

. Introduction

Crossflow membrane emulsification (CME) was origi-ally developed in Japan in the late 1980s[1]. Two maindvantages of CME are identified[2]. First, the energy con-umption per unit of product made using CME is much lesshan that using conventional methods, i.e., high-pressure ho-ogeniser, the diaphragm homogeniser and the more recenticrofluidizer. This is significant not only in term of energyfficiency but also in improving the quality and functionalityf delicate ingredients since the high shear and accompany-

ng temperature rise due to the viscous dissipation in con-entional methods (99.8% energy converted to heat) haveegative effects on the delicate ingredients. Secondly, CME

s a much more promising method to achieve very small (es-ecially diameter less than 1�m) and uniform droplets than

∗ Corresponding author. Tel.: +44 29 2087 5921; fax: +44 29 2087 4716.E-mail address:[email protected] (D. Barrow).

the conventional methods since the resulting droplet sicontrolled not by the generation of turbulent droplet breup, but primarily by the choice of membrane.

The most frequently used membranes as describedliterature are microporous glass (MPG) and Shirasu poglass (SPG)[3]. These membranes are characterised by cdrical, interconnected micropores. The ceramic�-Al2O3 or�-Al2O3 coated with titanium oxide or zirconia had also bused. The pore size can be reduced and pore size distribmade narrower by coating. These membranes come witferent nominal pore size (0.05–14�m) and distribution (usually claimed to be±15%). It is generally accepted thatdroplet diameter produced by CME can be linearly relatethe pore size under given operating conditions:

dd = xdp

wherex can range typically from 2 to 10[3–5] under opti-mised operating conditions anddd,dp are the mean droplet dameter of emulsions and the pore diameter of membran

376-7388/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
oi:10.1016/j.memsci.2005.02.038

J. Zhu, D. Barrow / Journal of Membrane Science 261 (2005) 136–144 137

Fig. 1. Scanning electron micrograph of (a) a section of a micromachined silicon nitride membrane showing 14 pores and (b) detail of an individual pore.

spectively. Therefore, ‘mono-dispersed’ emulsions can onlybe produced if the membrane pore size distribution is suf-ficiently narrow. The presence of coarse pores can lead tobimodal distribution[6]. It is believed that the availability ofwell-defined membranes (uniform pore size, shape and inter-pore distance) could provide a step-change improvement inCME. Microengineering technology now enables the fabri-cation of such well-defined membranes in various formats. Inthis study, such micromachined membranes were used for in-vestigation. An example of such a membrane (pore diameter2.0�m and inter-pore distance 20�m) is shown inFig. 1.

For a given system (continuous phase, dispersed phase,emulsifier and membrane are fixed), the most significant driv-ing force for droplet detachment in CME is the shear force(or drag force) produced by continuous phase flow[4,7,12].However, this provides only limited control over the process.It has been shown that membrane vibration can enhance localshear stress over the membrane[8]. Therefore, in the workreported here, membrane vibration, through piezoactuation,was investigated to help the detachment of droplets and pro-vide an extra control over the emulsification process.

During this study, the individual pore behaviour in theemulsification process was visualised and the droplets gen-erated were quantitatively analysed whilst using a speciallydesigned membrane with relatively large inter-pore distance(the distance between adjacent pores). The observed pore be-h er-i here-a withs esti-g cantf edm

2

2

andw rea:4 yA ifi-c

Fig. 2. A membrane pore array (drawing is two rows of pores).

Tween 20 (polyoxyethylene sorbitan monolaurate, Aldri-ch) dissolved in demineralized water was used as the con-tinuous phase (CP). Hexadecane (BDH) coloured with sat-urated Sudan Red 7B (Aldrich) was used as the dispersedphase (DP). Prior to experiments, the DP was filtered usingPolydiscTM PTFE membrane (Whatman) that has an aver-age pore size of 0.45�m. The CP was directly used withoutfurther treatment.

2.2. Experimental setup

The experimental setup is shown inFig. 3. A flat mem-brane (Fig. 2) was housed in the module. The flow-rate of theCP was monitored by a water flow meter (Key Instruments;range 5–50 mL/min). The height of CP solution column waskept constant, which maintained a constant pressure over themembrane and a stable flow-rate in the channel of the mod-ule. A large container filled with aqueous solution was used toprovide the continuous phase, without interruption over longperiods (days). This design enabled life-time experiments onthe membranes to be undertaken. The fluid channel for the CPin the module was 7.8 mm wide and 0.5 mm high. The dis-persed phase was delivered by a KD Scientific (model 101)syringe pump, which had a minimum step of 1.0�L/min witha 10 mL syringe. The formation of emulsions was observeda iable

TM nn

C

MMM

aviour provided valuable information for optimising expmental conditions and selecting membrane design. Tfter, the performance of micromachined membranesmaller inter-pore distance and pore size was also invated. The investigation demonstrated that whilst signifi

urther optimisation is required, vibrating micromachinembranes offer significant potential for CME.

. Experimental

.1. Materials

Micromachined (using both reactive ion etchinget etching) silicon nitride membranes (surface amm× 4 mm; thickness: 1.0�m, Fig. 2) were provided bquamarijn Microfiltration B.V. (Netherlands). The specations of these membranes are shown inTable 1.

nd recorded by videomicroscopy using a custom var

able 1orphological specifications of the 1.0�m thick micromachined silicoitride membranes

ode Pore diameter(�m)

Inter-pore distance(�m)

Porosity (%)

2.5L 2.5 100 0.052.5 2.5 25 0.792.0 2.0 20 0.79

138 J. Zhu, D. Barrow / Journal of Membrane Science 261 (2005) 136–144

Fig. 3. Schematic apparatus of the piezoactuated crossflow membrane emul-sion system.

zoom lens (Microinstruments Ltd., UK), Watec 203B cam-era and a JVC video recorder.

2.3. Membrane assembly and piezoactuater system

The membrane was glued into the recess of the sam-ple holder using Araldite epoxy resin (Fig. 4a). In orderto generate membrane vibration, two piezo stacks (PST150/10× 10/20; Piezomechanik GmbH) were assembledtightly as shown inFig. 4b and excited with square wavepotentials of the same amplitude but out of phase by 180◦.This was achieved by passing a single source ac voltage froma HP 33120A function generator, via two separate, parallel741 operational amplifiers, one in inverting mode, througha two-channel Piezomechanik power amplifier (model SUR150-3). The frequency and amplitude of the excitation signalwas controlled by the function generator. A hollow PTFE base(not shown) was assembled to form a channel for deliveringthe disperse phase.

2.4. Membrane pre-treatment

In order to produce oil/water emulsions, it is well acceptedthat the hydrophilicity of the membrane is crucial[1]. Ourprevious experiments showed that, without pre-treatment, thesilicon nitride membrane were not sufficiently hydrophilicto readily allow droplet detachment. Also, droplets formedwere observed to be randomly dispersed. To improve the hy-drophilicity of the membrane, a pre-treatment method usingoxygen plasma was implemented. Hence, prior to emulsi-fication experiments, the glued membranes (together withits holder) were pre-treated in an oxygen plasma (PolaronPT7160) for 10 min. This treatment significantly changed(two-tailed t-test;p< 0.001;n= 10) the water contact angleon the membrane from 60.3± 3.3◦ (untreated membrane)to 42.4± 3.8◦ (treated membrane) rendering it more hy-drophilic.

2.5. Measurement of the droplet diameter

By observing a standard-length unit (e.g. 1/2 mm) underthe identical conditions using the video-microscopy capturesystem, the magnification could be calculated from Eq.(1):

N = Ls (1)

w ex-p int

yedo ordedv iam-e

d

w td

F viewed d from theb o stack tor.

ig. 4. (a) A membrane glued in position within a membrane holder (ottom of the module with membrane holder inserted). Opposed piez

Lr

hereN is the magnification of the video system at theerimental conditions,Ls the length of the unit measured

he screen andLr is the real length of the unit.In the CME experiments, the droplet diameter displa

n screen was manually measured by replaying the recideo under frame control. Therefore, the real droplet dter was calculated by Eq.(2):

d = Ld

N(2)

heredd was the real droplet diameter andLd was the dropleiameter measured on screen.

from top); (b) membrane holder actuated by two piezo stacks (viewes are excited with 180◦ out of phase amplified signals from a function genera


3. Results and discussion

3.1. Emulsification on stationary membrane with largeinter-pore distance

To facilitate the study of individual pore behaviour inthe emulsification process, a membrane with large inter-poredistance (100�m) was selected. The formation of emulsionwas observed and recorded by videomicroscopy. Unlike thetypical emulsification process[9,10], a very low CP veloc-ity (25 mm/s) was applied to facilitate the measurement ofdroplet diameter, although this resulted in the production oflarge droplets. It was observed from video footage (an ex-ample single-frame shown inFig. 5a) that all pores underthe observed field of view were active, but that the frequencyof droplet generation from individual pores was different.For example, inFig. 5a, pore 4 had a droplet generationfrequency of 0.4 Hz whilst pore 6 was 0.1 Hz. It was alsofound that some pores produced much larger droplets thanothers. The average droplet diameter produced from pore 3was only 79± 1.2�m whilst those produced from pore 6was 111± 2.1�m. Accounting for all the observed pores,the droplet size distribution shown inFig. 5b indicated that atleast two types of pore behaviour existed. One type of poresgenerated droplets centred at 85�m, but another type gen-erated droplets centred at 105�m. This strongly suggested

that some of pores behaved substantially different from theothers under the experimental conditions despite the fact thatpores were well defined using microengineering techniques.Therefore, further experimentation is required to establishthe relationship between droplet size distribution and poregeometry, and understand the bimodal distribution observed.

To investigate the performance of the membrane underprolonged running time, the droplets generated from the sixthpore were monitored intermittently. FromFig. 6, it can beshown that the pore tended to produce larger droplets asa function of prolonged running time. In particular, duringthe first few hours, the trend was substantial. For example,the average droplet diameter increased from 111 to 118�min the first 80 min. This can be understood by the fact thatthe chemically active species, which were generated on themembrane surface during pre-treatment[11], tend to adsorbor react with other functional groups (i.e., impurity, surfac-tant) that existed in the emulsification system. Therefore,the membrane surface is likely to become less hydrophilicand resulted in producing larger droplets. This speculationwas supported by the results of contact angle measurements.These showed that the water contact angle on the membranechanged from an initial 42.4± 3.8◦ (after O2 plasma treat-ment, but prior to emulsification experiment) to 65.9± 4.2◦at the end of an emulsification experiment of 3 h (the re-covered membrane was exhaustively washed using acetone,

F ing tim erimc flow v

ig. 5. (a) Visualization of droplet formation after an initial 3-h runnonditions: M2.5L membrane; 3.0 wt% Tween 20 aqueous solution; CP

Fig. 6. The relationship of droplet size generated and duration of

e and (b) the droplet diameter distribution measured under the expentalelocity, 25 mm/s; DP flow-rate, 1�L/min, pore numbers marked.

continuous operation. Error bars = 1σ; n= 20–30 for each data point.


Fig. 7. The effect of CP velocity on the diameter of droplets produced.The experimental conditions were: M2.5L membrane; 3.0 wt% Tween 20aqueous solution; DP flow-rate, 1�L/min. For each data point,n= 5–10.

methane and HPLC water, respectively). A two-tailedt-testfound the data to be significantly different (p< 0.001;n= 10).However, the membranes continued to work well after 48 hwithout further significant increments in the droplet size. Itwas also observed that some of the originally active poresbecame inactive with increasing running time. This deactiva-tion of pores may be attributed to the pore blockage and/ormembrane fouling[8], which is well documented for micro-filtration. This reproducible dependence of droplet size oninitial running time confirmed that factors, which influencedroplet size are dynamically variable on a new membraneand in this case stabilised after 3 h.

An investigation of the effects of CP velocity and DP flow-rate was undertaken using the same membrane type but in sep-arate experiments. These data were collected from a singlepore (refer toFig. 9) after an initial 3 h operational period. Itcan be shown fromFig. 7that the CP velocity had a dramaticeffect on reducing the droplet size. Much smaller emulsionscan be obtained by increasing the CP velocity[3,7]. The ve-locity can be increased by decreasing the cross-sectional areaof the CP channel whilst leaving the CP flow-rate constant.CFD simulation can also provide design rules for maximisingthe CP velocity[7] through the module.

Although the crossflow membrane emulsification has beenidentified as a promising method to produce narrowly dis-

Fig. 8. The effect of DP flow-rate on the diameter of droplets produced.The data were recorded after 3-h running time with the conditions: M2.5Lmembrane; 3.0 wt% Tween 20 aqueous solution; CP velocity, 21 mm/s. Foreach data pointn= 50–60; error bar = 1σ.

persed emulsions with shear sensitive components using rel-atively low energy input, its practical industrial applicationsare mainly limited by the low disperse phase flux[3]. A num-ber of parameters (e.g. porosity, pore size, membrane thick-ness) can affect the disperse phase flux. For a given mem-brane, the disperse phase flux can be enhanced by elevatingthe transmembrane pressure. In this study, the disperse phaseflux was increased by increasing the DP flow-rate (no trans-membrane pressure was measured in this study). It is shownfrom Fig. 8 that the diameter of droplets produced becamelarger as the flow-rate increased in the range of 0–10�L/min.One explanation for this could be the fact that, under higherDP flux, the time for forming a new interface between aque-ous phase and oil phase is substantially shorter. This in turnresults in higher dynamic interfacial tension of Tween 20(whilst it takes 2–10 s for Tween 20 to reach a stable lowvalue of dynamic interfacial tension, it takes only 0.1–1 s togenerate a new droplet) and thus the retaining force[12]. Be-cause of this, the increased retaining force could contributeto the generation of larger droplets. Surprisingly, the dropletdiameter dramatically decreased with further increments inCP flow-rate from 10 to 20�L/min. By careful study of thevideomicroscopy files, it was found that the two adjacentpores around the studied pore became more active (produc-

F file and mec.
ig. 9. (a) An example image taken from a recorded videomicroscopy (b) a schematic illustrating the possible ‘push-to-detach’ detachmenthanism


Fig. 10. (a) Visualization of droplet formation and (b) the droplet diameter distribution under the experimental conditions: M2.5 membrane; 3.0 wt%Tween20 aqueous solution; CP velocity, 25 mm/s; DP flow-rate, 1�L/min.

ing droplets more rapidly than the studied pore). As dropletsgrew in size, the two droplets produced from adjacent poresappeared to ‘push’ the droplet produced from the studiedpore and contributed to its detachment (Fig. 9a). This obser-vation is consistent with that documented by Abrahamse etal. [13]. The detachment mechanism is schematically illus-trated inFig. 9b. However, coalescence, which was usuallyobserved in crossflow emulsification using traditional mem-branes (i.e., MPG and SPG), was not observed in this case.This is due to the fact that the inter-pore distance (100�m)of the used membrane was far larger than that of conven-tional membrane. When the dispersed phase emerges from apore, it is a small droplet and thus the interface increases veryrapidly due to its high surface to volume ratio (assuming theDP flux is fixed). Therefore, the dynamic interfacial tensionis rather high. If this small droplet contacts with another adja-cent droplet (also with high dynamic interfacial tension), thechance of coalescence is high. However, if the droplet growslarger before contacting adjacent droplets, its surface area in-creases slowly due to the relatively low surface to volumeratio. Thus, the dynamic interfacial tension at the interface ofthe droplet is small. Hence, the chance of coalescence is lit-tle. It was considered that under the experimental conditionsemployed here, the interfacial tension of the droplet was lowenough to avoid considerable coalescence before it contactedadjacent droplets. Therefore, adjacent droplets were forcedt t thatc voidc

3.2. Emulsification on a stationary M2.5 membrane

In order to test the ‘push-to-detach’ hypothesis, a mem-brane with smaller inter-pore distance (25�m) but identicalpore size and membrane thickness was selected. The experi-mental conditions were also kept identical to those presentedin Fig. 5. It was shown from the visualization inFig. 10a thatmost of the pores were active and no obvious droplet coales-cence was observed. The quantitative analysis of the dropletdiameters showed that the mean droplet diameter (d= 48�m;σ = 7.4�m) was dramatically and significantly (two-tailedt-test,p< 0.001;n= 50) smaller than that produced from mem-brane M2.5L (94�m; σ = 13.2�m). It was also shown fromFig. 10b that droplet size distribution was also substantiallynarrower (two-tailedF-test,p< 0.001;n= 50). This signif-icant observation suggests that it is possible to reduce thesize of emulsion droplets by reducing the inter-pore distance.Moreover, a higher DP flux can be expected when using sucha membrane (higher porosity).

One possible explanation for this observation is that thehigh surfactant concentration (3.0 wt%) used in the exper-iments may strongly prevent coalescence, thus causing the‘push-to-detach’ phenomena observed. A surfactant concen-tration of 1–2% was widely used for emulsification in the lit-erature[13,14]. Accordingly, the following experiments werecarried out using 1.0 wt% Tween 20 (keeping all other ex-p 3%T dt ed

F tion usi 0 aqueouss

o detach. Thus, an optimum inter-pore distance can exisan facilitate droplet detachment (small droplets) but aoalescence (narrow size distribution).

ig. 11. Image of droplet formation and the droplet diameter distribuolution; CP velocity, 25 mm/s; DP flow-rate, 1�L/min.

erimental conditions constant). A comparison of 1 andween solutions found no significant difference (two-tailet-est,p= 0.87;n= 50) in the size of emulsion droplets form

ng a M2.5 membrane under the following conditions: 1 wt% Tween 2


Fig. 12. Image of droplet formation and the droplet diameter distribution using a M2.0 membrane under the following conditions: 1 wt% Tween 20 aqueoussolution; CP velocity, 25 mm/s; DP flow-rate, 1�L/min.

(Fig. 11). This suggested that, above a certain concentrationof emulsifier, the dynamic interfacial tension were lowereddown enough during the droplet growth to avoid coalescence.In particular, for Tween 20, the dynamic interfacial tensiondoes not change substantially with variation of concentration[12].

Despite the similarity on mean size,F-test gives a verysmall value (probability = 1.9e−3). This suggested that thesize distribution was substantially different under the twoconditions. The difference was attributed to the fact that theadsorption of emulsifier molecules by the active species inthe pores could change the size and opening geometry ofpores, thus resulting in the wider size distribution. This couldhappen more possibly at high emulsifier concentration (3%).Indeed, this was consistent with our recent observation (notdescribed here) that pore fouling may be contributed to byadded emulsifier.

3.3. Emulsification using a stationary M2.0 membrane

To further reduce droplet size, a membrane with smallerpore size (2.0�m) and inter-pore distance (20�m) than previ-ous membranes were used for emulsification under otherwiseidentical experimental conditions. It was shown (Fig. 12) thatsmaller droplets (mean diameter = 38�m, σ = 8.6�m) werep rved.T tach’h e sizea to theu theya

3m

cu-m h lo-c f highc ayf ts ine twop ations( fre-

quency range 0–1 MHz with amplitudes of up to 50�m). Inthis study, a square wave actuation signal was used. Actualmembrane displacement could not be accurately measuredover the full frequency range. Therefore, droplet diametersare expressed as a function of potential voltage. The experi-ments were carried out using a membrane of M2.5L. It wasa surprise to observe (Fig. 13) that the droplet size increasedwith increasing applied voltage in the range of 0–75 V whilstthe vibration frequency was fixed at 10 Hz. It should be notedthat the vibration experiments (starting fromt= 0 h) were un-dertaken during the time period 0–3 h in the same experimentas that presented inFig. 6. Therefore, the droplet size increasecould be attributed to duration of operation. By analysis ofthe recorded videomicroscopy data files it took 80 min to readthe maximum droplet diameter of 121�m (Fig. 13). Undersimilar experimental conditions but with no membrane vi-bration (Fig. 6), a mean droplet diameter of∼118�m shouldbe obtained after 80 min. Considering the experimental error(σ > 4.8�m in previous discussions), these results matchedwell. When the excitation signal amplitude was increasedfrom 75 to 125 V, the droplet diameter was substantially re-duced (Fig. 13). This suggested that the reduction of dropletsize by membrane vibration could only be caused only abovea certain threshold value. At this applied potential (125 V),the vibration displacement amplitude (peak to peak) can beestimated to lie in the range of 10–12�m.

F zoac-t ment ast rdedf

roduced without significant coalescence being obsehese results further support the proposed ‘push-to-deypothesis. Further experiments using even smaller pornd inter-pore distances could not be undertaken duenavailability of membranes with such geometries sincere currently challenging to fabricate.

.4. Membrane emulsification using a piezoactuatedembrane

In crossflow microfiltration processes, it was well doented that vibrating membrane modules can induce hig

al shear stress near the membrane surface without use orossflow velocities[8,15,16]. The enhanced shear stress macilitate the detachment of droplets, which in turn resulmulsions with smaller droplet size. To test this notion,iezo stacks were assembled to cause membrane vibrFig. 4). The piezo stacks could generate vibrations in the

ig. 13. Emulsion droplet size as a function of voltage applied to pieuated membranes. Experiments were undertaken in the same experihat presented inFig. 6using a freshly pre-treated membrane and recorom t= 0 h. For data pointsn= 50–60; error bar = 1σ.


Fig. 14. Emulsion droplet size as a function of piezoactuated membrane vibration frequency for a given voltage of 125 V. For data pointsn= 50–60; errorbar = 1σ.

The effect of vibration frequency was also investigatedusing another freshly pre-treated membrane (M2.5L) for agiven applied potential of 125 V (Fig. 14). It was found thatthe minimum droplet diameter (d= 108�m) was obtained atapproximately 5 Hz. Further increments in the vibration fre-quency resulted in the droplet diameter increasing sharplyup to∼100 Hz, after which the effect reached a plateau. Themost likely explanation is that membrane displacement at fre-quencies beyond 5 Hz became progressively curtailed withincreasing frequency due to mechanical damping effects ofthe membrane holder and sliding mechanism employed. Re-cent data supporting this is under analysis and will be reportedelsewhere. Such a displacement–frequency relationship mayexplain the result shown inFig. 14.

4. Conclusions

In this work, CME was undertaken using well-definedmicromachined silicon nitride membranes of short path-length (1.0�m thickness). After being pre-treated by oxy-gen plasma, the membranes performed well for emulsifica-tion albeit with relatively large droplet sizes. It was foundthat the average diameter of droplets from a wide inter-poredistance membrane increased with duration of usage. Thiso pre-t em-b diame allerd ur-i ible‘ estedt de-t nismw em-b pores ranes ct them

To facilitate the droplet detachment from the membraneand provide additional control over the detachment, the ef-fects of membrane vibration were investigated. The prelimi-nary results showed that a vibrating membrane had a signifi-cant effect in reducing the average size of emulsion dropletsbut only at very low frequencies. Further work is underway to(i) characterise the frequency–displacement relationships ob-served here, (ii) employ a module with mechanical displace-ment amplification and (iii) use higher flow-rates with high-speed imaging, to quantitatively assess the demonstrated po-tential for piezoactuated membranes in CME.

Acknowledgements

This work was supported by the project ‘Towards HighlyAdvanced Membrane Emulsification System (THAMES,QLK1-CT-2001-01228)’ supported by the Commission ofthe European Communities. We kindly thank Mr. TyroneJones for his contribution in the construction of the mem-brane module, Aquamarijn Microfiltration B.V. for supply-ing the micromachined membranes and Dr. Jo Janssen andother partners for helpful discussions. The authors also thankthe anonymous referees for their helpful comments in revis-ing the manuscript and the Rutherford Appleton LaboratoryCentral Microstructure Facility for access to equipment foru

R

ion

tatustrial

ure

singore,

bservation may be attributed to the instability of thereated membrane. After the first 3 h of operation, the mranes appeared to become stabilised and the averageter of droplet produced became relatively constant. Smroplets could be produced using higher CP velocity. D

ng investigations on the effect of DP flow-rate a posspush-to-detach’ mechanism was observed. This sugghat relative short inter-pore distances could facilitate theachment of droplets from the membrane. This mechaas confirmed by the emulsification results using a mrane of smaller inter-pore distances but of the sameize. Amongst the membranes tested the M2.0 membhowed the best performance characteristics with respeean minimal droplet diameter.

-

ndertaking contact angle measurements.

eferences

[1] T. Nakashima, M. Shimizu, M. Kukizaki, Membrane emulsificatby microporous glass, Key Eng. Mater. 61–62 (1991) 513.

[2] A.J. Gijsbertsen-Abrahamse, A. van der Padt, R.M. Boom, Sof cross-flow membrane emulsification and outlook for indusapplication, J. Membr. Sci. 230 (2004) 149.

[3] S.M. Joscelyne, G. Tragardh, Membrane emulsification—a literatreview, J. Membr. Sci. 169 (2000) 107.

[4] S.J. Peng, R.A. Williams, Controlled production of emulsions ua crossflow membrane, Part I. Droplet formation from a single pChem. Eng. Res. Des. 76 (1998) 894.


[5] Y. Mine, M. Shimizu, T. Nakashima, Preparation and stabilization ofsimple and multiple emulsions using microporous glass membrane,Colloids Surf. B: Biointerfaces 6 (1996) 261.

[6] R.A. Williams, S.J. Peng, D.A. Wheeler, N.C. Morley, D. Taylor,M. Whalley, D.W. Houldsworth, Controlled production of emulsionsusing a crossflow membrane, Part II. Industrial scale manufacture,Chem. Eng. Res. Des. 76 (1998) 902.

[7] M. Rayner, G. Tragardh, Membrane emulsification modelling: howcan we get from characterisation to design? Desalination 145 (2002)165.

[8] J. Postlethwaite, S.R. Lamping, G.C. Leach, M.F. Hurwitz, G.J. Lye,Flux and transmission characteristics of a vibrating microfiltrationsystem operated at high biomass loading, J. Membr. Sci. 228 (2004)89.

[9] G.T. Vladisavljevic, H. Schubert, Preparation and analysis of oil-in-water emulsions with a narrow droplet size distribution usingShirasu-porous-glass (SPG) membrane, Desalination 144 (2002) 167.

[10] V. Schroder, H. Schubert, Production of emulsions using microp-orous, ceramic membranes, Colloid Surf. A 152 (1999) 103.

[11] L.T. Yin, J.C. Chou, W.Y. Chung, T.P. Sun, S.K. Hsiung, IEEE Trans.Biomed. Eng. 48 (2001) 340.

[12] V. Schroder, O. Behrend, H. Schubert, Effect of dynamic interfacialtension on the emulsification process using microporous ceramicmembranes, J. Colloid Interface Sci. 202 (1998) 334.

[13] A.J. Abrahamse, R. van Lierop, R.G.M. van der Sman, A. van derPadt, R.M. Boom, Analysis of droplet formation and interactions dur-ing cross-flow membrane emulsification, J. Membr. Sci. 204 (2002)125.

[14] N.C. Christov, D.N. Ganchev, N.D. Vassileva, N.D. Denkov, K.D.Danov, P.A. Kralchevsky, Capillary mechanisms in membrane emul-sification: oil-in-water emulsions stabilized by Tween 20 and milkproteins, Colloids Surf. B 209 (2002) 83.

[15] O. Al Akoum, M.Y. Jaffrin, L. Ding, P. Paullier, C. Vanhoutte, Anhydrodynamic investigation of microfiltration and ultrafiltration in avibrating membrane module, J. Membr. Sci. 197 (2002) 37.

[16] O. Al-Akoum, L.H. Ding, M.Y. Jaffrin, Microfiltration and ultrafil-tration of UHT skim milk with a vibrating membrane module, Sep.Purif. Technol. 28 (2002) 219.

analysis of droplet size during crossflow membrane emulsification using stationary and vibrating...

Documents