transport of bubbles in square...

PHYSICS OF FLUIDS VOLUME 16, NUMBER 12 DECEMBER 2004

Transport of bubbles in square microchannelsThomas Cubaud and Chih-Ming HoMechanical and Aerospace Engineering Department, University of California at Los Angeles,420 Westwood Plaza, Los Angeles, California 90095

(Received 24 February 2004; accepted 16 September 2004; published online 5 November 2004)

Liquid/gas flows are experimentally investigated in 200 and 525mm square microchannels made ofglass and silicon. Liquid and gas are mixed in a cross-shaped section in a way to produce steady andhomogeneous flows of monodisperse bubbles. Two-phase flow map and transition lines betweenflow regimes are examined. Bubble velocity and slip ratio between liquid and gas are measured.Flow patterns and their characteristics are discussed. Local and global dry out of the channel wallsby moving bubbles in square capillaries are investigated as a function of the flow characteristics forpartially wetting channels. Two-phase flow pressure drop is measured and compared to single liquidflow pressure drop. Taking into account the homogeneous liquid fraction along the channel, anexpression for the two-phase hydraulic resistance is experimentally developed over the range ofliquid and gas flow rates investigated. ©2004 American Institute of Physics.[DOI: 10.1063/1.1813871]

mi-

blemck-gasn-

icroof dble

uch

hectsthe

lyposchaex-ting

wittemtacttwo

on-tingrleon

aticted.

s

n

ve

drop

intouidvoid

licxture.ndi-chan-in ad ho-aseare

themaplocityeraged toeen

allyheirse oftantd asesenelfromwtionaulic

I. INTRODUCTION

Multiphase flow occurs in many operations in the checal, petroleum, and power generation industries(such asnuclear power plants and micro-fuel cells). Unlike large-scale systems, gas bubbles can present significant proin microfluidic systems by disturbing and eventually bloing the flow. Interactions on the boundaries betweenliquid, and solid introduce nonlinearity and instabilities. Uderstanding how bubbles affect the flow resistance in mchannels is, besides its fundamental aspect, a concerntermining the pumping or energy requirement for portamicrofluidic devices where two-phase flow is involved sas in a microdirect methanol fuel cell.1

When the sizeh of a channel is sufficiently small so tBond number Bo is smaller than unity, gravitational effeare likely to be small. In a square microchannel, whenbubble sized is larger thanh, gas and liquid are neutralbuoyant and the cross section of the liquid-air interfacesesses symmetries with respect to the center axis of thenel. While most of the two-phase flow literature reportsperiments conducted with a complete wetting fluid, wetoften takes place either on heterogeneous surfaces orcontaminant preventing a complete wetting of the sysand leading to the formation of a contact line. The conangle has been shown to have a significant effect onphase flow in circular capillary.2

The conditions for bubble displacement through cstricted polygonal capillaries were studied for a wetfluid.3–5 The shape of a meniscus inside anN-sided regulapolygonal capillary depends onN and on the contact angu.6,7 Whenu,p /N, the gas does not fill the entire polygcross section. In the case of a square section, foru,p /4,liquid fills the wedges and gas fills the center. In a stcondition and for a partially wetting liquidsu.0d, a contacline between liquid, gas, and solid is energetically privileg

When liquid and gas are flowing, the situation becomes

1070-6631/2004/16(12)/4575/11/$22.00 4575

s

,

-e-

-n-

h

-

rather complex because of dynamic contact angle(u,Ca1/3 where the capillary number Ca=hV/g with h theviscosity,V the velocity, andg the liquid/gas surface tensio)and hysteresis.

Liquid/gas and liquid/liquid flows in microchannels hareceived an increasing experimental interest.8–10 Differentmodels have been proposed to predict the pressureacross a bubble train in a square capillary5,11–13but, so far, nosystematic experimental studies were performed takingaccount the flow patterns. Interaction of gas and liqphases in small channels often results in pressure andfraction fluctuations.14–16 Fluctuations affect the hydrauresistance of the channel and average density of the mi

The objective of this work is to understand the cotions necessary to purge gas bubbles from square micronels. We first discuss how liquid and gas are mixedcross-shaped section in a way to produce a steady anmogeneous flow in a microfluidic device. The two-phflow map and the transition lines between flow regimesdrawn for 200 and 525mm square microchannels. Overrange of liquid and gas flow rates investigated; the flowremains the same regardless of channel size. Bubble vewas measured and was found to be equal to the avmixture velocity. The apparent liquid fraction is comparethe homogeneous liquid fraction and the slip ratio betwthe liquid and the gas velocity is shown to monotonicincrease with the mixture velocity. Flow patterns and tcharacteristics are then individually discussed. Becauthe small channel size, wettability effects play an imporrole in the system. Wedging and dry flows are describespecific to noncircular partially wetting capillaries. Thflows correspond to local or global dry out of the chanwalls by bubbles. Finally, the pressure drop is measureda single liquid flow to a single gas flow for different florates. Taking into account the homogeneous liquid fracalong the channel, expressions for the two-phase hydr

resistance are experimentally developed when the liquid is

© 2004 American Institute of Physics

http://dx.doi.org/10.1063/1.1813871

icro-si-forbled toex-perp r

exflowand

are

ainwelsea

sur-surerted

n islasse-

the

s isa

inia-e

nk,tureowr-

Theeren-end-ssured airsys-

tate isle dis-ragedwasure

ofo-

nted intflow

theg the

teadto

ventera

edflow

p ind ann aiquid/gasas

utlet.

theas

ric

4576 Phys. Fluids, Vol. 16, No. 12, December 2004 T. Cubaud and C.-M. Ho

mostly pushing the bubbles(high liquid fraction) and whenliquid is mostly flowing in the corners(low liquid fraction).

II. EXPERIMENTAL SETUP

A. Microchannels fabrication

Channels are made with glass and silicon using mfabrication techniques.17 Channel masks are printed in potive with a high-resolution printer on transparent paperlithography. Photoresist is spin coated on a 10 cm dousided polished silicon wafer. The wafer is then exposeUV light through the mask. The parts of the resin layerposed to UV light are removed by immersion in a develobath. Channels are etched at different depths using deeactive ion etching(DRIE). The sealing is made with Pyrglass using anodic bonding, providing optical access foranalysis. Flexible tubes are glued with epoxy on gasliquid inlets and on the pressure sensor inlet.

Figure 1 shows a typical channel module. Channelscomposed of two parts: a mixing section(50±5 mm squarechannel) and a test section(200±10 or 525±10mm squarechannel). A 50 mm square channel is connected to the mchannel to allow pressure measurements. Experimentsconducted using de-ionized(DI) water and air. As channewere partially wetting, contact angle hysteresis was msured using DI water droplets on progressively tiltedfaces. Advancing and receding contact angles were meaon glass and on DRIE silicon just before the droplet stamoving. The advancing contact angle on glass isua,glass=25±1° and the advancing contact angle on DRIE silicoua,silicon=9±1°. The receding contact angle for both gur,glass and ur,silicon were slightly above 0°. Dewetting phnomena were observed in the experiments confirmingnonstrictly zero receding contact angles.

B. Apparatus

A schematic diagram of the experimental apparatushown in Fig. 2. Liquid is injected into the channel fromreservoir, the pressure of which is adjustable with a mture regulator. The liquid flow rateQL is measured at thchannel inlet with a liquid volumetric flow meter(from0 to 1 ml/min). Gas is coming from a compressed air tathe pressure of which is also adjustable with a miniaregulator. Air flow rateQG is measured with gas mass flmeters (from 0 to 100 ml/min). Mass flow meters dete

FIG. 1. Two-phase flow microchannel module: mixing section(50 mmsquare channel), test section(200 or 525mm square channel).

mine the volumetric flow rate at the channel outlet based on

-

e-

re

-

d

the viscosity of the air at the measured temperature.pressure inside the test channel is measured with a difftial pressure sensor ranging from 0 to 1, 5, or 15 psi deping on the flow rate. Before each measurement, the presensor channel is flushed with water to remove all trappepockets. After each change of the gas inlet pressure, thetem was allowed to reach a steady state. The steady sdefined when pressure sensors, flow meters, and bubbtribution are stationary. The measured pressure is aveover 1000 points during 20 s. The liquid inlet pressurelimited to 50 psis

l inmm

nel.ifor

blesingis

mixhas

owfluidob

bbly

efinirge

tflow

nes

n

staninlet

ofFlowex

andi-ho

risnsi-some. Form are.sur-han-ase isnelsof

e the

han-tedns.

citytaken. For

--

Phys. Fluids, Vol. 16, No. 12, December 2004 Transport of bubbles in square microchannels 4577

liquid inlet, the liquid flow rate and pressure are identicaeach of the channels. Because of the square channel sytries, bubbles flow “naturally” in the center of the chanThe cross shaped was chosen to produce a steady unflow. This production technique of monodisperse bubpresents some similarities with a capillary flow focustechnique.18 In our geometry, the minimal size of bubblesabout the size of the channel. In these experiments, thising system is used as an effective way to produce two-pflows in microdevices.

IV. TWO-PHASE FLOW MAP

Two-phase flows are distributed into several distinct flpatterns depending on the liquid and gas flow rates andand channel properties. Five main flow regimes wereserved in the partially wetting square microchannels: buwedging, slug, annular, and dry flows(see Fig. 4). Flow pat-terns are described in Sec. V. Besides differences in dtion of flow patterns, some typical regimes observed in lachannels such as stratified, churn, or mist flows19–21were noobserved in our channels within the range of laminarrates investigated.

Figure 5 shows the flow map and the transition libetween flow regimes for 200 and 525mm channels. Liquidsuperficial velocity(JL =QL /A, whereA is the cross-sectioarea) is plotted against gas superficial velocitysJG=QG/Ad.Each set of experiments is conducted maintaining a conliquid inlet pressure while gradually increasing the gaspressure. The liquid flow rateQL then decreases becausethe increase of the hydraulic resistance of the channel.regime transitions are determined by visualization andamination of pressure drop data(see Sec. VI). Transitionsbetween regimes are predictable as a function of liquidgas flow rates. For both 200 and 525mm channels, transtions between each regime occur for fixed values of the

FIG. 3. Example of cross-shaped mixing section in 100mm square channels: (a) gas inlet,(b) and (c) liquid inlets, and(d) two-phase flowing towards test section.(1) and (2) wedging flow,(3) annular flow.

mogeneous liquid fractionaL defined as

e-

m

-e

-,

-r

t

-

-

aL =QL

QL + QG. s1d

The bubbly/wedging transition is foraL

ho-a-ityficia

to

at

rageation.

etero-uid

voidatedForated

th

spect

ee

lls is

ss-e


nor can the velocity be optically measured. For differentmogeneous liquid fractionsaL, Fig. 6 compares the mesured bubble velocityVB to the average superficial velocJ, defined as the sum of the liquid and the gas supervelocities:

VB < JL + JG = J. s2d

Over the range of measured velocities(from0.01 to 1 m/s), the bubble velocityVB can be assimilated

FIG. 5. Flow pattern map for horizontal fl

FIG. 6. Comparison of the bubble velocityVB (measured by image proceing) with the average superficial velocityJ (measured with liquid and thgas flow meters) as a function of the homogeneous liquid fractionaL for

different initial conditionssh=525mmd.

l

the average superficial velocityJ. This result suggests ththere is no corner flow. Contrary to circular channels,24 nocharacteristic drift velocity between bubbles and the avespeed of the fluid was measured in the present investigIn studies using circular channels, as the channel diamshrinks, drift velocity is reduced21 and removed under micrgravity conditions due to lack of buoyancy between liqand gas.25

B. Void fraction measurement

From image processing, the apparent time-averagedfraction «G is calculated. Time-space diagrams are crefrom movies with a line plotted normal to the channel.each liquid and gas flow rate, a composite image is crewhere thex coordinate corresponds to the channel widhand they coordinate corresponds to the timet [Fig. 7 (left)].As the liquid/air interface possesses symmetries with reto the center axis of the channel[Fig. 7 (right)], «G is calcu-lated as follows:

«G =

p

4 o dt2 + o s2l th − l t2d

Th2, s3d

wheredt=2r is the diameter of the curved interface,h is theheight(and width) of the channel,l t is the width of the corgas in “contact” with the walls, andT=St corresponds to thtime of acquisition. Except for the bubbly flow(wherel t=0),the liquid film between the gas and the center of the wa

n 200 and 525mm square channels, air-water.
ow i
neglected in the calculation of the total liquid volume be-

thes

quid

tica

etri-the

n anfacto-

oidde-

-ry

utedin-

eels.e

sizeiquid

as. The

rgeitiest of asion

ticalto

.se todif-urfaceoundeads

ned

en-avitywith

ti-oss


cause it is negligible compared to the liquid volume incorners. The apparent liquid fraction«L is simply defined a

«L = 1 −«G. s4d

Figure 8 shows the relation between the apparent lifraction «L and the homogeneous liquid fractionaL. Transi-tions between flow regimes are shown by straight verlines. Deviation from a linear relationship between«L andaLoccurs from the slug flow foraL ,0.02. Kolb and Cerro

26

showed the transition from an axisymmetrical to a symmcal semi-infinite bubble profile in square channels whencapillary number Ca increases. This behavior results iincrease of the liquid fraction with the gas flow rate. Thethat the apparent void fractionaL does not increase montonically with Ca is specific to polygonal channels.

From the calculation of the apparent liquid and vfractions, the slip ratioS between liquid and gas can befined as

S=VGVL

=JGJL

«L

«G. s5d

FIG. 7. Image processing to calculate the apparent void fraction«G. (a) Partof a raw time-space diagram(wedging flow), (b) threshold and area idenfied time-space diagram,(c) characteristic symmetries of a channel crsection.

FIG. 8. Apparent liquid fraction«L vs homogeneous liquid fractionaL for

different initial conditionssh=525mmd.

l

Figure 9 shows the slip ratioS as a function of the homogeneous liquid fractionaL. For the annular and the dflow, experimental data suggest a relationship as

S= 0.1aL−1. s6d

The large slip between the two phases can be attribto wall shear for the liquid, while the gas can flow unhdered in the channel core.S is meaningless foraL =0. In a100 mm circular channel, Kawaharaet al.16 reported a largslip ratio S compared to macrochannels and minichannFor example, they foundS

n reorneeds thbeddwa

in-tionsig-dissibluid.

dth

dginow,

n

r ofs in

he

ed

ebye is

d

heg on.ction

engle

ntact

n-drycon-

le toe

id

thetinghave

a


occurs in sharp corner bends[Fig. 10(a)]. As the liquid flowin the vicinity of the corner is very small, a small bubble(1)can easily be trapped; the effective channel size is theduced and a bubble too large to pass through the cbecomes trapped as well(2). Other bubbles become trappas a chain reaction until they eventually merge and pasbend. At higher liquid flow rates, bubbles are not perturby the sharp bend[Fig. 10(b)]. This effect can be removeusing a curved bend. The bend effect on two-phase flowinvestigated in larger channels28 and has been shown tocrease coalescence of small bubbles in some flow condiIn our experiments, bubble size and velocity were notnificantly perturbed by this geometry except for the casecussed above. We note that the bubbly flow is not accesto the system when a mechanical pump supplies the liq

B. Wedging flow

The wedging flow [Fig. 4(b)] consists of elongatebubbles, the size of whichd is larger than the channel widh sd.hd. When the flow is steadily developedsaL ,0.70d,bubbles are equally spaced and monodisperse. The weflow exhibits some differences from the Taylor bubble flusually reported in the literature.15 For a partially wettingsystem, as a function of the bubble velocityVB, bubbles cadry out the center of the channel creating triple lines(liquid/gas/solid). The liquid film between the gas and the centethe channel can be considered as static, while liquid flowthe corner. The film thicknessd was shown to depend on tbubble lengthd.29 A liquid film of thicknessd is known to bemetastable ifd,dC, dC being the critical thickness definby

dC = 2lC sinSur2 D , s7dwhere lC is the capillary lengthflC=sg /gsrL −rGddg1/2,whereg is the gravity andrL and rG are, respectively, thliquid and gas densities). In that case, the film can dewetnucleation and growth of dry patches when their siz

30

FIG. 10. Bubbly flow: effect of a sharp return;(a) low liquid velocity, (b)high liquid velocity.

greater than a critical valuerC defined as

-r

e

s

s.

-e

g

rC =d

l r sinur, s8d

where l r is a logarithmic function ofrC. For the capillaryregimeslC, rCd, the dewetting velocityUdew is constant anindependent of the film thickness,31

Udew=U*u3

6LL, s9d

where U* is the characteristic velocity of the liquidsU*=g /hd and LL is a prefactor depending on the liquid. Tdewetting velocity was measured from image processinglass in the channel and was found to beUdew

lugneldobe-tioness

wercificd b

hapled

etric

ithrenin-

t-xi-hin, Caange

icalln i

cir-ent

heannel

ndede

bendingceanceow

ated.gas

opa-f theasn goy cor-ein-

rede theThegasr,a.

ored onsta-

film

the


C. Slug flow

For aL ,0.20, bubbles distribute themselves in s[Fig. 4(c)]. The bubble size is far larger than the chanheight sd@hd. Bubbles are surrounded by liquid so theynot touch the channel walls. For long bubbles the filmtween gas and solid may dry out. In a previous investigaWonget al.29 demonstrated the decrease of the film thicknd along the bubble. In some case, growing dry patcheslocally observed at the bottom of the gas slug on spechannel locations, suggesting the nucleation was inducelocal surface defects.

The nose of the bubbles has a characteristic bullet sWhen a semi-infinite bubble is injected into a liquid filsquare capillary, Kolb and Cerro34 studied the transition froma nonaxisymmetric bubble nose profile to an axisymmbubble nose profile for a capillary number of Ca

e

asendi-s a., ex

and

wssluginlere

aturandmin

crossr the

ther atheumelsan

e drychesptoti-

the

l

e-

DI

-re-

es,

ed

t ofd tocap-

ow

the

alls

l-e a


stay on the solid wall(see Fig. 14). The radial size of thgas/solid contactl increases with the gas flow rateQG, asaLdecreases. Single gas flow occurs whenl =h.

VI. PRESSURE DROP

Pressure drop caused by frictional force in two-phflow is a parameter of prime interest to determine the cotions for the flow. A typical plot of the pressure drop afunction of the capillary number Ca=hLJ/g is shown in Fig15. Each flow regime has a different dependence on Cacept for the bubbly and the wedging flow. On the other hwhen DP is plotted versus the gas superficial velocity[Fig.15 (right)], the transition between bubbly and wedging floappears clearly. The transition from the wedging to theflow causes a large drop in pressure. At the same gasand liquid inlet pressure, the system switches from onegime to another. The slug flow presents an interesting feof decay of pressure with gas flow rate. RatulowskiChang11 proposed an expression for the pressure, assu

FIG. 14. Dry flow: the central gas core dries out the center of channel w

FIG. 15. Typical pressure drop measurementDP as a function of the capilary number Cash=200mmd. Transitions between flow regimes caus

change on the pressure drop dependence on Ca andJG.

-,

t-e

g

that the pressure drop is the sum of the pressure abubbles and liquid slugs. They found an expression fopressure that is proportional to the number of bubbles inchannelnb. As nb is decreasing with the gas flow rate fogiven channel of lengthL, the pressure drop decreases forslug flow. For the annular flow, the presence of a maximin the pressure gradient was observed in larger chann36

and it was attributed to a transition from a mist flow toannular flow. In our case, as the liquid volume fractionaLdecreases, the liquid friction decreases as well. For thflow, the effective channel size for the gas flow approathat of single gas flow and the pressure converges asymcally to the single gas one.

A. Single-phase flow

For a laminar steady state flow of fluidi in a squarechannelsRe,3000d, the pressure gradient is related toflow rate as38

Qi =h4

Ahi¹ P, s10d

where A

bbleowaleugh

entode

e-e

ere

penone

the, theix-s a

f theidpat-la-w is

lls,s-n oions-s-

atatwo

re onthe

ingleiq-ia-ses

l isever-n be

.a

ncelabletion.d

-g

asquid

ssurece


al.13 took into account the capillary geometry and the bulength for the pressure-velocity calculation for bubble flin polygonal capillaries. They found that pressure drop sclinearly with Ca when the liquid bypasses the bubble throcorner channels(corner flow) and that it scales with Ca2/3

when the liquid is essentially pushing the bubble(plug flow).Stanleyet al.12 analyzed their pressure drop data in differsquare microchannels using the homogeneous flow mwith the Fanning equation leading tof =C/Ren, where f isthe friction coefficient,C andn are numerical constants dpending on the flow(laminar or turbulent), and Re is thReynolds number.

Applied to our experimental data these correlations winaccurate in predicting the measured pressure dropsDP.This can be attributed to several factors. First, the dedence of the slip ratioSbetween liquid and gas velocitiesthe homogeneous liquid fractionaL [Eq. (6)] hinders the usof an apparent viscosity of the mixture used to calculatecapillary number Ca and the Reynolds number Re. Alsochange in pressureDP that is dependent on the average mture velocityJ between the annular and dry flows disablemodel that takes into account a monotonical change oslip ratio S with J. The Lockhart and Martinelli model dnot provide good prediction nor transition between flowterns. This is not surprising as Lockhart-Martinelli corretions are known to provide poor agreement when the flolaminar.22

Since liquid is mostly flowing along the channel wathe contribution of the liquid viscosity to the frictional presure drop is predominant. Figure 16 shows the evolutiothe pressure drop for 200mm square channels as a functof the homogeneous liquid fractionaL. The two-phase presure dropDP2-phasewas scaled by the single liquid flow presure dropDPL associated to the liquid flow rateQL in thechannelsDPL =RLQLd. As can be seen in the figure, dcollapse more or less on a single master curve with

FIG. 16. Two-phase flow pressure drop scaled by single liquid flow predrop as a function of homogeneous liquid volume fractionaL sh=200mmd.

distinct regimes. The bubbly and the wedging flows are de

s

l

-

f

picted on one side and the slug, annular, and dry flows athe other side. For the bubbly and the wedging flows,pressure drop associated can be written as

DP2-phase= DPLaL−1. s12d

In that case the flow can simply be considered as a sliquid flow with a velocityV corresponding to the mean luid velocityVL =JL /aL. As can be seen in Fig. 16, the devtion to Eq. (12) increases as the liquid velocity decrea(circles and squares) because of wetting phenomena(seewedging flow).

For smaller liquid fraction, the homogeneous modenot realistic because of the slip between the phases. Ntheless, a simple empirical correlation for the pressure cawritten as

DP2-phase= DPLaL−1/2. s13d

This equation better fits the data for high liquid flow rateAs aL changes from 1 to 0, the system goes from

single liquid flow to a single gas flow. The fluidic resistaR=DP/Q is an interesting parameter because it is calcufrom direct measurements without assuming any correlaThe two-phase flow fluidic resistanceR2-phasecan be defineas

R2-phase=DP2-phaseQL + QG

. s14d

Figure 17 shows the variation ofR2-phasewith the homogeneous liquid fractionaL. For high liquid fractions, usinEq. (12), R2-phasecan be written as

R2-phase= RL . s15d

The deviation for slow liquid velocity appears clearlycontact lines increase the resistance to flow. For low lifractions, using Eq.(13), R2-phasecan be written as

R2-phase= RLa1/2. s16d

FIG. 17. Two-phase fluidic resistanceR2-phasecompared to liquid resistanRL as a function of homogeneous liquid fractionaL sh=200mmd.

- L

ws,e osis-

bu-flowe caresiscir-

sedin ary imapn fo

ofno

nelandThe

s liqgasoidnded.im-et-

ndco

ntacthe

yriestheurentliquhicher inc-nel

s exflow

o

cell

hase

ith

ttinghem.

longolloid

vity,”

po-

micRev.

pat-

“An

ries,”

icro-

s inluid

ow,”

es in

aset. J.

mi-

as-

cro-

ase

flowMass

ns in

rav--

are

cro-

two,

s in

oles

ting,”


At the transition between the annular and the dry flothe flow can be seen as a gas flow in a circular tubdiameterh and a liquid flow in the corners. The gas retance in a circular tube of diameterh is RG,circular=128hGL /ph

4. In this situation, assuming that the contrition to the pressure drop is essentially due to the gasbecause of the large slip ratio, the two-phase resistancbe writtenR2-phase

Pro

ns in

of

uid

ular

erseys.


faces,” Fourth European Coating Symposium, Advances in Coatingcesses, edited by J.-M. Buchlin and J. Anthoine(Brussels, 2001).

33L. Schwartz, R. V. Roy, R. R. Eley, and S. Petrash, “Dewetting pattera drying liquid film,” J. Colloid Interface Sci.234, 363 (2001).

34W. B. Kolb and R. L. Cerro, “The motion of long bubbles in tubessquare cross section,” Phys. Fluids A5, 1549(1993).

35T. C. Thulasidas, M. Abraham, and R. L. Cerro, “Flow pattern in liq

slugs during bubble-train flow inside capillaries,” Chem. Eng. Sci.52,

- 2947 (1997).36G. F. Hewitt and A. H. Govan, “Phenomena and prediction in ann

two-phase flow,”Advances in Gas-Liquid Flows, 1990, ASME-FED, Vol.99 (ASME, New York, 1990).

37J. M. Gordillo, A. M. Gañán-Calvo, and M. Perez-Saborid, “Monodispmicrobubbling: Absolute instabilities in coflowing gas-liquid jets,” PhFluids 13, 3839(2001).

38
F. M. White,Viscous Fluid Flow(McGraw-Hill, New York, 1991).

transport of bubbles in square...

Documents