Physical aspects of liquid-impelled

loop reactors

Ontvanyen
0
9SEP 1992
UB-CARDe-

: LANDBOUWCATALOGUS

0000 0490 9749

4C*Y 5 t
Promotor: dr.ir.J.Tramper
hoogleraar indebioprocestechnologie
 S
mjQc*^ >>~^

HenkvanSonsbeek

Physical aspects of liquid-impelled

loop reactors

Proefschrift
terverkrijging vandegraadvan
doctor inde landbouw- en milieuwetenschappen,
opgezagvande rector magnificus,
dr. H.C.vander Plas,
in hetopenbaar te verdedigen
opwoensdag 2September 1992
des namiddagstevier uur inde aula
van de Landbouwuniversiteit te Wageningen
 uwuomtm
LANDBOUWUNIVERSOia
SVAGENlNGE&l
 Imagination createsscienceandart,
justskiptheemptyphrases;
it'sallcommonsenseandheart,
hiddenbehindfaces.

(vrij naarWilliamWordsworth's "Thetablesturned").

 STELLINGEN

1. Hetopsommenvan voor- en nadelenvan systemen en/of werkwijzen kan een

vertekend beeld geven omdat de nadelen de, meestal eerder genoemde,
voordelenkunnenopheffen.

Ditproefechrift.

M.H. Cho andS.S.Wang (1988). Enhancement of oxygentransfer in hybridomacellculture by usinga

perfluorocarbonasanoxygencarrier.Biotechnol.Lett., 10,855-860.

2. Het zou beter geweest zijn als Garcia Calvo de frictiecoefficient in het
gepresenteerde hydrodynamicamodelmetbehulpvandatzelfde model bepaald
hadinplaatsvangebruiktemakenvaneenandermodel.

E.GarciaCalvo.Afluiddynamicmodelforairliftloopreactors(1989).Chem.Eng.Sci.,44,321-323.

3. De mechanistische basis voor de empirische machtsrelatie tussen holdup en

disperse fase debiet in pneumatische reactoren, gegeven door Chisti and
Moo-Young,iszelfgebaseerdopeenempirischerelatie.

M.Y.ChistiandM.Moo-Young.Gasholdupinpneumaticreactors.(1988).Chem.Eng.J.,38,149-152.

4. Hetdunnerwordenvandeozonlaagzalertoeleidendatweindetoekomstmeer
zullengenietenvanbewolktweer,danvanstralendhelderedagen.

5. HetgebruikvanEngelsewoordenindeNederlandsetaalheeftalzulkeextreme
vormen aangenomen dat woorden zoals deeltijdhoogleraar het gewichtigst
klinkenindezuiverNederlandsevorm.

6. HetwervingsbeleidvandeLandbouwuniversiteitWageningenkomteropneerdat
discriminatiemag,mitsdaareengoederedeneneenmeerderheidvoorzijn.

7. Eenexpertsysteem iseencomputerprogrammawaarop de maker of bedenker

ergtrotsis.
 fijiJoVzo*) )SZH

8. DeeenheidvandeEuropesegemeenschapzalpasdanbereiktzijn,wanneerde
eerstelidstaatzichonafhankelijkverklaart.

9. Geziendekwaliteitvandeprogramma'sopdeNederlandsetelevisiekanverdere
uitbreidingvandereclamezendtijdalleenmaartoegejuichtworden.

10. Dewaarde van "Who's afraidof Red,Yellow and Blue III" kanalleen nog inere
hersteldwordendoordeontdekkingvanoudeverfrollertjes vanNewman.

Stellingen behorend bij net proefschrift: 'Physical aspects of liquid-impelled loop

reactors'.

HenkvanSonsbeek,2September 1992
 CONTENTS

Shortintroduction 1

Chapter 1: Two-liquid-phase bioreactors 3

Chapter 2: Hydrodynamic modelforliquid-impelledloopreactors 21

Chapter3: Liquid-circulation modelsfortwo-phaseloopreactors 39

Chapter 4: Mixinginliquid-impelledloopreactors 67

Chapter 5: Steady-state methodfor kadetermination inmodelsystems 91

Chapter6: Oxygentransfer inliquid-impelledloopreactors using 101

perfluorocarbon liquids

Generaldiscussion 119

Summary 125

Samenvatting 129

Nawoord 133

Curriculumvitae 135
 SHORT INTRODUCTION

Theresearchthatformsthebasisforthisthesisoriginatesfromacombinationof
advancesinbiotechnology.Theattractivecharacteristicsofanair-liftloopreactor(ALR)
andthepromisingfieldofbiocatalysisinorganicmediaarecombinedinthedevelopment
ofthe liquid-impelled loop reactor (LLR).Otherthanthework described inthisthesis,
thereismoreresearchgoingon,inwhichtheLLRplaysanimportantrole.Theconversion
ofbenzenetocis-benzeneglycolbyamutantofPseudomonassp.isanexampleforthe
suitability of an LLRfor biological processes.3 By using an LLRwith hexadecane as
rising dispersed phase, lethal effects of benzene on the biocatalyst are prevented
becausemostofthebenzeneispresentinthehexadecanephase.Researchhasbeen
carriedoutforthecultivationofplantcellsinanLLR.1Theaimistoworkunderverylow
shear stress conditions and to extract the secondary metabolites. Furthermore, the
conversion of tetralin to tetralol byAcinetobactor 75 is investigated, with the aim of
controlled supply of the highly toxic substrate.4 Ribeiro et al.2 are investigating the
production of L-tryptophan from L-serine and indole, which is poorly water soluble.
Escherichiacoli is used asthe biocatalyst and higher productivities arefound when
multi-phasesystemsareapplied.CarryingoutthisprocessinanLLRiscurrentlybeing
workedon.
The purpose of this thesis is to describe the studies carried out to gain more
fundamentalknowledge aboutthecharacteristics ofLLRs.Therefore itisorganizedin
thefollowingway. Inthefirstchapter aspectsofbioreactorswithtwo liquidphasesare
described and examples found in a literature study are discussed. This chapter is
intendedtogiveagoodimpressionofthepresentstateoftheartandtogetacquainted
with the relevant matter. In the second chapter, the liquid-impelled loop reactor is
introducedandahydrodynamic modelisdescribedthatcanbeusedfor predictionof
liquidvelocity anddispersed-phase holdup asafunction ofthedispersed-phase flow
rate.ExperimentalresultsobtainedwithanLLRwitharisingdispersedphaseareused
forverification ofthis model.Theexperiencewiththis hydrodynamic modeland astill
growing,alreadyhugeamountofsometimesconfusingliteratureaboutthistopic,mainly
with respect to air-lift loop reactors, has lead to chapter three. In this chapter
 SHORT INTRODUCTION

hydrodynamicmodelsareclassifiedaccordingtotheirphysicalprincipleonwhichthey
are based. It gives information that shows more clearly the limitation or weakness of
each type of model. The purpose is to make the essence of these models easier
identifiableandbetterjudgement possible.Althoughmerelyliteratureaboutair-lift loop
reactors isused,theapplicability isgeneralandthereforeitispartofthisthesis.
Agood-controlledflowisusuallyclaimedasoneoftheadvantagesofloopreactors.
Besidesthevelocityofthecontinuousphase,describedinthepreviouschapters,mixing
duringthisflowisacharacteristicofmajorimportancefordesignandcontrol.Therefore
chapter four contains the description of overall mixing and mixing per section inthe
continuousphaseofanLLR.Aconfigurationwithdescendingdispersedphaseisused
forexperimentaldeterminationofmixingparameters.Theresultsarecomparedtomixing
inALRs.
PerfluorocarbonliquidsareusedinLLRstoinvestigatemass-transfer.Thedynamic
method,that is used most oftento measure volumetric transfer coefficients, was not
easyapplicablefortheLLRduetopracticallimitations.Anewlydeveloped,steady-state
techniquetodeterminethevolumetricmass-transfer coefficientisdescribedinchapter
five. Inchapter sixtheactual masstransfer inan LLRisdescribed.The purpose isto
establishtheoxygentransferrateandsupplycapacityofLLRs.Attheendofthisthesis
thewholeresearch isevaluatedandconclusions andfutureperspectives aregiven.

[1] Buitelaar R.M.,Susaeta,I.,TramperJ. (1990).Application oftheliquid-impelled loop reactorfor

theproduction ofanthraquinones by plant cellcultures.694-699. in:H.J.J. Nijkamp,L.H.V.Van
DerPlas,andJ.VanAartrijk(eds.),Progressinplantcellularandmolecularbiology,KluwerAcad.
[2] Ribeiro M.H.L, Nunes I.M.L, Cabral J.M.S., Fonseca M.M.R. (1991).A multiphase system for
the production of L-tryptophan from L-serine and indole; studies on product separation and
recovery. NatoAdvanced Study Institute, 16-21September, Turkey.
[3] Van den Tweel W.J.J., Marsman E.H., Vorage M.J.A.W., Tramper J., De Bont J. (1987). The
application of organic solvents for the bioconversion of benzene to cis-benzeneglycol.
International conference on bioreactors and biotransformations, Gleneagles,Scotland,UK.
[4] Vermue M.H., Tramper J. (1990). Extractive biocatalysis in a liquid-impelled loop reactor. 5th
Europeancongress on biotechnology, Proceedingsvolume 1,Copenhagen, Denmark.
 CHAPTER1

TWO-LIQUID-PHASE BIOREACTORS

INTRODUCTION

Thischapter isdealingwithbioreactorsinwhichtwoliquidphasesareapplied.In
1982 Lilly42 already published an article entitled "Two-liquid-phase Biocatalytic
Reactions" and predicted new commercial applications for multi-liquid biocatalytic
reactors.In1987Lillyetal. 44published"Two-liquid-phase biocatalytic reactors",which
mainlygivesashortgeneraldescriptionofaspectsthatcanplayarolewhentwoliquid
phases are applied in bioreactors.The present chapter is about the sametopic and
givesanoverviewofwhathasbeenachievedwithtwo-liquid-phase bioreactors sofar.
Before discussing these bioreactors, a short consideration is given about the
introductionoftwoliquidphasesinbiotechnological production processes.
Newdevelopmentsinbiotechnologyarefocussedonnewproductsorintroduction
ofnewprocessesasasubstitutetotheconventionalway,orperformanceimprovements
ofexistingbiotechnologicalapplications.Inrecentyearsmuchattention isgiventothe
use of two-liquid-phase systems. Especially the application of organic solvents has
fascinated many researchers and isstillconsideredto beapromisingtopic, because
manybiocatalytic reactionsinvolveorganiccomponentsthatarepoorlywatersoluble.

Thischapter has beensubmittedfor publication bytheauthors :H.M.VanSonsbeek, H.H.

Beeftink andJ.Tramper.
 CHAPTER1

Forthisreasonadditionofwater-solubleorganicsolventsinsmallproportionsisalready
wellknownforalongtimeasatechniquetoimproveavailabilityofpoorlywater-soluble
reactantstocellsandenzymes.
More recent research concerns the use of two immiscible liquid phases in a
bioreactor. In most cases the aqueous phase contains the biocatalyst, possibly
immobilized.Thesecondliquidphasecanbeanorganicsolventoranotherwaterphase.
Incaseofthelatter,thesystem iscalledanaqueoustwo-phasesystem,thatconsists
oftwopolymer solutionsorapolymerandasaltsolution.
Reversedmicellesareacompletelydistinctcategoryoftwo-liquid-phase systems
andcanbeconsideredaspoolsofwaterinorganicsolventswithasurfactantinbetween.
Thiscategorywillnotbediscussedhere.
Inthis chapter research developments concerning two-liquid-phase bioreactors
willbediscussedwiththeaimtogiveaclearcomprehensionofthemotivesoftheiruse
andtogiveanoverviewofthestateoftheart.First,somegeneralaspectsofconventional
bioreactors will be highlighted, such as aeration, shear stress, continuous or batch
processing,andimmobilization.Theseaspectswillreturninthesubsequentpartwhere
bioreactors arediscussed inwhichanorganic solvent andacatalytic aqueous phase
arepresent.Theorganic solventcanbethecontinuous orthedispersedphaseandit
canbethereactantorasolventinwhichpartofthereactantand/orproductisdissolved.
Inaddition,the suitability of aqueoustwo-phase systems asabiological mediumand
some research examples are given. Finally, some conclusions are drawn and future
perspectivesaregiven.

CONVENTIONAL BIOREACTORS

Biotechnological possibilities are often shown on a laboratory scale inflasks or

tubes.Furtherdevelopmentofsuchprocessesmeansthatthesameorbetterispursued
onalargerscale,whichimpliesthat bioreactortechnologybecomesimportant.Agreat
varietyofbioreactortypesexists,ofwhichthestirred-tankreactor(STR)andthebubble
column (BC)arethemostsimpleandmostwidely usedones. Many biotechnological
processesrequireoxygenthatistransferredfromairtotheculturebroth.Instirred-tank
reactors air bubbles brought into the reactor are dispersed by the stirrer to obtain a
large exchange areafor mass transfer. Therefore high oxygen-transfer rates can be
 CHAPTER 1

achieved, which is important because this is often the limiting factor in aerobic
bioconversions. BCs are attractive, even at small scales, because of their simple
construction. At large scales this advantage becomes even more important because
aerationathighenergy inputratesismuchsimplerthanstirring.Besidesthesetwo basic
types, other bioreactors are developed such as bioscrubbers and membrane,
packed-bed,trickle-bed,orfluidized-bedbioreactors.Thesetypeswillnotbe discussed
indetail here.
Because the oxygen-transfer rate is often the rate-limiting step, many
biotechnological studies are focussed on this topic. The limited transfer rate is the
consequence of the contradiction between the need for oxygen of many biocatalysts
and the low solubility of oxygen in aqueous environments, that are used most often.
Closely relatedtothisaspect isthe issueofthe cultivationof shear-sensitive cells, such
as plant, animal and insect cells. Stirring enhances oxygen-transfer rates to a great
extent, but is only sparingly possible in these cultures, because of damaging effects.
Thereforedifferent kindsof reactors,such asair-liftloop reactors (ALRs),are proposed
to be suitable forthese purposes. 17 But even aeration by sparging air isfound to have
adetrimentaleffectoncells16,30,31 andvariouskindsofaerationmethodsaredeveloped,
e.g. bubble-free aeration using silicontubing. 4 7
Thedescribedcaseofobtainingappropriate oxygensupply showsthe importance
ofthe choice of the reactor type and adjustments of process conditions to the specific
needs ofthe application.The introduction of non-conventional media can be regarded
assuch an adjustment that improves processes. Examples of non-conventional media
are organic solvents, supercritical solvents or aqueous two-phase systems. The
bioreactors inwhich these media are applied can be the standard-type bioreactors or
standard types that are adjusted for use with non-conventional media. Examples can
befound inthe next sections.

ORGANIC SOLVENTS IN BIOREACTORS

The topic of organic solvents as biocatalytic reaction media has been reviewed
many times 1,10,14,22,28,38,40,41,42,43,46 and needs no further explanation here than a
shortsummary oftheargumentsthatareusedtoindicatetheusefulnessofthismethod.
The argument that is relevant most often, is a better solubility of substrates and/or
 CHAPTER1

products;anexampleisthesteroidtransformation.15Anotherargumentisthatreverse
hydrolyticreactions,thatarenotfeasibleinpureaqueoussystems,canbecomefeasible;
anexampleistheesterificationofafattyacidandanalcohol.66Insomecasesproduct
inhibitioncanbereducedbecauseoftheremovaloftheproductintotheorganicphase.
Transfer of products to the organic phase can shift the apparent reaction equilibria
calculatedonanoverallvolumebasis.23Ifthe productconcentration isincreaseddue
toabettersolubility intheorganic phase,product recovery becomeseasier.
Therearealsoafewaspectsoforganicsolventsthatgiverisetonewproblemsor
lessattractive phenomenasuch as inhibition or biocatalyst inactivation.Furthermore,
inextractivefermentationsunintentionalformationofemulsions,andremovalofprotein
fromthefermentationbrothcanoccur (proteinstripping).
Agreat number ofsolventtypescan beapplied inbioreactors,suchasalkanes,
alkenes, esters, alcohols, ethers, perfluorocarbons, etc. Only the group of the
perfluorocarbons will shortly be highlighted, because they are relatively unknown.
MattiassonandAdlercreutz50andKingetal. 39describetheuseofperfluorocarbonsin
biotechnology. Perfluorocarbons are ring or straight-chain hydrocarbons in which
hydrogenatomshavebeenreplacedbyfluorine.Perfluorocarbonsarebothstableand
virtuallychemically inertduetothepresenceofthestrongcarbon/fluorine bonds(ca.
487kJ/mol).Inaddition,thefluorineatomsofferstericprotectiontothecarbongroups.
Because ofthe low interactions between perfluorocarbon molecules themselves, gas
molecules caneasily occupy intra-molecular spaces and largevolumes ofgases can
thus dissolve. There is no chemical attraction of oxygen or carbon dioxide to
perfluorocarbons, which implies that Henry's law for dissolved gas is applicable.
Perfluorochemicalsshouldnotbeconfusedwitharelatedandcurrently,fromthepoint
of view of environmental care, somewhat controversial group of compounds, the
chlorofluorocarbons.
The choice of the organic solvent depends on the purpose of use. In case of
extractive biocatalysis, for instance, the product should dissolve preferently in the
solvent. Besidesproductsolubility,toxicityofasolventforthe biocatalyst isofgeneral
importance and much research is carried out on this topic. In general a positive
correlationisfoundbetweenhydrophobicityofsolventsandnon-toxicityforbiocatalysts.
A measure for hydrophobicity, that was found very suitable for characterization of
solvents,isthelogPvalue,whichisthelogarithmofthepartitioncoefficientofasolvent
over awater octanoltwo-phase system.Solventswithalog Pvalueabove4arevery
 CHAPTER1

hydrophobicandgenerallyshownotoxiceffectsonbiocatalysts.BruceandDaugulis11
recentlypresentedsolventselectionstrategiesforextractivebiocatalysis,andconcluded
thatitispossibletocomposemixturesofsolventsthatshowgoodextractionproperties
aswellasgoodbiocompatibility,evenatrelativelyhighconcentrations oftoxicsolvent.
For the application of organic solvents it is obvious that existing standard-type
bioreactorscanbeused,suchasstirred-tank, packed-bed,fluidized-bed,andcolumn
reactors.Smalladjustmentsintheexperimentalset-upcanoftenbesufficienttooperate
atwo-liquid-phase biocatalyticalprocess.
Inmany investigations oftwo-liquid-phase biocatalysis, stirred-tank reactorsare
used.6'69Infigure1 and2thetwobasicexperimentalset-upsforextractivefermentation
withstirred-tankreactorsaregiven.Thefirstset-upisthemostsimpleone,butcanlead
to arather stable emulsion that requires along settlingtime. Inthat case anexternal
extractioncolumnismorefavorable because emulsionformationislessvigorous.An
externalextractioncolumnisalsomoreeffective becauseitisamulti-stage contactor,
whereasfordirect solventadditiononlyoneequilibriumstagecanbeachieved.

regenerated
solvent settler
solvent to
•Q:::.o:
, product
'•'• ' Q \ \ 7 . recovery
substrate O

—>•
broth
o-.-o-
recycle

stirred
=€>
pump discharge
tank
reactor

Figure 1: Extractionwithdirect solvent addition (adaptedfrom Roffler etal.55).

 CHAPTER1

solvent to
product
recovery

substrate

regenerated
stirred
pump solvent
tank
reactor
discharge

Figure2: Extractionwith contacting inanexternalvessel (adaptedfrom Roffler etal.55).

Packed-bedreactorscanalsobemadesuitableforapplicationoftwoliquidphases.
Belletal.6comparedastirredtankreactorandapacked-bedreactorforthehydrolysis
oftriglycerideandconcludedthatinthiscasethepacked-bedreactorwassuperior.
Withinthegroupoftheauthors BrinkandTramper8.9investigatedtheapplication
oforganicsolventsinanimmobilized-cellfluidized-bedreactorand usedthe epoxidation
ofpropene byimmobilizedMycobacterium cellsasamodelsystem.
Concerningcolumn-typebioreactorsafewproposalscanbefoundinthe literature
fortheset-upwithanorganic solvent asoneofthetwo liquidphases.Chibataetal. 18
describe in a patent a bioreactor system with external aeration of hydrocarbons or
perfluorocarbons. Pumping these air-saturated solvents via a liquid sparger in the
bioreactor insteadofairsparging,enhances oxygen-transfer rates.The backgrounds
and literature concerning this aspect are described in the paragraph "aeration of
two-liquid-phasesystems".Tramperetal. 64 introducedtheliquid-impelledloopreactor
(LLR),whichisschematically drawninfigure3,andVandenTweeletal. 65 showedan
experimentinwhichthisreactorwasusedforthe conversionofbenzenetocis-benzene
glycol. Furthermore, Buitelaaretal. 13 havebeenworkingontheuseofthe LLRforthe
 CHAPTER1

cultivation of plant cells and at present the conversion of tetralin to tetralol by

Acinetobacterisinvestigated.Gianettoetal. 27introducedacontinuousextractionloop
reactor,whichwillbediscussedintheparagraph"batchorcontinuous".

organic
solvent o^)f>gSS<o~

aqueous
phase

circulation
main tube
tube

liquid buffer
sparger
t=€> vessel

pump

Figure3: Liquid-impelled loop reactor (LLR) (Tramperetal.64).

Finally,themembranereactorisabioreactorinwhichtwoimmiscibleliquidphases
canbeappliedandmustthereforebementionedhere.Anessentialdifferencewiththe
othertypes isthatthetwo liquid phases are not indirect contact, but separated bya
membrane.The membrane also servesasimmobilization supportforthe biocatalyst,
10 CHAPTER1

usually an enzyme. Van der Padt et al. 66 demonstrate the suitability of a membrane
bioreactor for the enzymatic esterification ofglycerol andfatty acidswiththe enzyme
lipaseimmobilizedonahollow-fibre membrane (figure4).

glycerol/water A

membrane reactor

adsorption
column

glycerids/fatty acid

Figure4: Schematic processdesignfor esterificationwithamembranesystem(Vander Padtetal.66).

AQUEOUSTWO-PHASE SYSTEMS
Besidestheuseofanorganicsolventincombinationwithanaqueousphase,the
aqueoustwo-phase system isanother possibility of usingtwo liquidphases.This isa
systemthat consistsoftwo immiscible aqueous phases,whichare polymer solutions
orapolymer andasaltsolution.Thephenomenon haslongbeen knownandagood
overview of the principle is given by Albertsson.3 Inthe last decade the interest for
applicationsofaqueoustwo-phasesystemsinbiotechnologyhasincreased,duetothe
combination of two attractive properties. Like in other two-liquid-phase systems,
partitioningofcomponents andbiocatalystsbetweenthetwo phasesoccurs,whichis
importantfordownstream processingandbiocatalyst retention.Thesecondattractive
aspect in aqueous two-phase systems isthe very low interfacial tension (0.0001-0.1
 CHAPTER1 11

mN.nr 1 ) compared to that of organic solvent/water systems (0.5-50 mN.m- 1 ). This

keeps proteins in solution and unfolding at the interface is prevented. Due to this
property, research of biotechnological applications of aqueous two-phase systems is
often focussed on isolation of proteins, cells or organelles from the broth.^. 6 1 The
partition coefficient for these components in aqueous two-phase systems is normally
in the order of magnitude of 0.1-10, but in some cases can be improved by affinity
interactions. However, this application asisolation method isnot considered here as a
replacement of existing techniques, but as complementary to more-selective isolation
methods. An attractivefeature ofthis system as compared to other isolation methods,
isthatthescaleupseems easy.Another way inwhichaqueoustwo-phase systems can
be applied for biotechnological purposes is to use them as medium for extractive
biocatalysis.Usuallythebioconversion preferentlytakes placeinoneofthetwo phases,
and the product can be extracted in the other phase. 4 . 48 . 49 . Andersson and
Hahn-Hagerdal 4 gave an extensive review of accomplishments and prospects for the
application of aqueous two-phase systems, which are at this moment still in a rather
experimental stage, lacking fundamental knowledge of partitioning.

BATCH OR CONTINUOUS

Inspiteoftheconsistent claimsoftheadvantages ofcontinuous processing, such

asahigher productivity andamore uniform product, most biocatalytical processes are
still carried out as batch processes, especially in the food industry. The enormous
amount of work that is done through the years to achieve continuous processing for
several applications, did not yet prove the suitability that is often claimed. An example
ofacontinuous process thatfailedinpractice isfound inthe beer-brewery history. The
complexity of acontinuous brewing process requiredthe close attention of skilled staff
and labour costs turned out to betoo high. 63 Onthe other hand developments are still
goingon,which isillustrated bythe recent presentation of acontinuous processfor the
production of soy sauce. 29
Improving the feasibility of continuous processing is often used as an argument
for the use of two-liquid-phase systems. Although it seems not very easy to set up a
continuous process with two-liquid phase systems, several articles are published on
extractive biocatalysis. 2,5,11,18,21,24,25,26,27,52,53,54,56,62This is a type of continuous
12 CHAPTER1

biologicalproductionprocessinwhichtheprocessisnotdisturbedorsometimeseven
improvedbyinsiturecoveryofthe product. Especiallyethanolproductionisapopular
topic inthis field.5'25'27*53'54. A recent example isthe work of Gianetto et al.27, who
developed andusedacontinuous extraction loopreactor (CELR)whichcanbeseen
infigure5.Thisreactorisanexternal-loopreactor,inwhichliquidcirculationisinduced
byagasflow inoneoftheverticaltubes andethanolisextractedbya countercurrent
extraction in the other tube. The experiments concern glucose fermentation by
Saccharomyces cerevisiae entrapped in calcium-alginate spheres. The authors
observedproblemswiththecalcium-alginatematrix,suchaslowmechanicalresistance,
aleakageofviablecells,asignificantinternaldiffusionlimitationandadecreaseinviability
oftheentrappedmicroorganisms.Theyemphasize,however,thatthereactor isstillin
apreliminarystage.

to
solvent product
out recovery
grid
^

extraction reaction
tube tube

liquid liquid
circulation circulation

solvent
in substrate
in
A--
g r i d \ ~/cglass
sparger
g.ast
in l

Figure5: Continuous extractionloopreactor (CELR) (adaptedfromGianettoetal.27).

 CHAPTER 1 13

Mattiasson 49 described a classical example of a product-inhibited biocatalytic

process, i.e. the production of acetone and butanol from glucose usingClostridium
acetobutylicum. From batch experiments in aqueous two-phase systems it was
concluded that butanol production should preferably be performed in a continuous
processinwhichtheproduct isremovedandfreshsubstrateadded.
Ingeneral,mosttwo-liquid-phaseresearchiscarriedoutinbatchexperimentsand
not much attention isgiventothe aimoftheway ofapplication. Claimed advantages
areoftennotprovenandprocessesstillneedtobedesigned.Althoughapplicationhas
notyetbeenverysuccessful,thereisstillfaiththattheintroductionoftwo-liquid-phase
systemsissuitablefortheoldidealofcontinuousprocessing.

IMMOBILIZATION

An extensive overview of immobilization of biocatalysts is given by Rosevear.60

Black7andVenkatasubramanian etal. 68givegooddescriptions oftheapplications of
immobilized biocatalysts andthe reactorsthat can beinvolved.There areafewvalid
motivesforimmobilization ofbiocatalysts.Themostimportant onefor growingcellsis
the increase of productivity because it is possible to work with dilution rates above
washout conditions.7'68 Furthermore, facilitated recovery of the biocatalyst and
protectionofthebiocatalystagainstadamagingturbulentenvironment canplayarole
aswell.Alsobecauseofdiffusionlimitationitispossiblethatanimmobilisedbiocatalyst
willnotbeexposedtoaninhibitingsubstrateconcentration.Intwo-liquid-phasesystems
gelentrapment of biocatalysts canprotectthemfrom denaturation atthe liquid/liquid
interface.Fortheapplication intwo-liquid-phase systems itisofimportanceto realize
thatimmobilizationusedasatechniquetopreventfreebiocatalystsinthesolution,can
onlybesuccessfulifitconcerns non-growing biocatalystsorahighdilutionrate.

AERATIONOFTWO-LIQUID-PHASE SYSTEMS

Inmanyaerobicbioprocessestheoxygen-supplyratetotheaqueousphaseisthe
limiting factor. The use of two liquid phases can contribute in two different ways to
14 CHAPTER1

improvethis.First,itispossibletoaerateoneliquidphaseandleaditthroughtheother
thatcontainsthebiocatalyst.Second,anemulsioncanbecreatedthatformsthemedium
andthat isaeratedbyspargingairoroxygen-enrichedair.
DamainoandWang20presentedamethodinwhichaperfluorocarbonwas used
asanoxygencarrier, continuously removedfromthe bioreactor after coalescence of
the droplets at the bottom. One circuit of perfluorocarbon was pumped back in the
bioreactor through a nozzle at the top. A second circuit of perfluorocarbon was
re-aeratedandthenpumpedbackinthebioreactor.Oxygenwastransferredbymeans
of liquid-liquid contacting. Escherichia coli was cultivated and it was shown that
operationofacolumnwithexternallyaeratedperfluorocarbonwascapableofsupplying
oxygenatsignificantlyhigherratesthanthebestrunofabubblecolumn.Thisisascribed
to amuchlarger surface areaavailable for masstransfer, which ispossibleforthree
reasons:1)atequalworkinputs,dropswillbesmallerthanbubblesduetolowersurface
tension;2)bubblesgenerally coalesce intolarger bubbles muchmorerapidlythando
drops;3) liquidholdups can befixedat levelsmuchhigherthantypical gas holdups.
Theprocess described by DamainoandWang20waslimited bythe coalescence rate
ofthesolventdropletsatthebottomofthereactor.
ChoandWang19alsousedasecondliquidphaseasanoxygencarrierinsteadof
thedirectspargingofagasintoahybridomacellculturetoincreasetheoxygen-transfer
rate.Overallvolumetricmass-transfercoefficientsweremeasureduptofivetimeshigher
thaninadirectlyaeratedsystem.Itisconcludedthattheinterfacialareaismadelarger
quiteeasily.Noharmfuleffectsweredetectedoftheperfluorocarbononthehybridoma
cells.Unfortunately nopositiveresultsontheapplicationcouldbereportedduetocell
aggregationattheinterfaceofthemediumandtheperfluorocarbon.
Inourlaboratoryadistinctionwasmadebetweenthemass-transfercoefficientand
theavailableexchangeareaofaliquid/liquid system. Mass-transfer coefficientswere
measuredthatarelowerthaningas/liquidsystems67(chapter6).However,becauseof
a higher specific exchange area inthe experimental set-up,the mass-transfer rateof
theliquid/liquidsystemisfavorable,whencomparedwithgas/liquidsystemsatequal
dispersed-phase flow rates. Because of slow coalescence the maximum
dispersed-phaseflowrateislimited,whichmeansasevererestrictionforthemaximum
oxygen-transfer rateinliquid/liquidsystems.
 CHAPTER1 15

Juet a l . 3 5 emphasize intheirwork that incaseof applicationfor low-shear, gently

mixedsystems,thegeneralefficiencyofexclusivelyexternallyaerated perfluorocarbons
as oxygen-transfer enhancers is unacceptably low.They also point out the problem of
protein stripping fromthe growth medium by the circulated perfluorocarbons.
Thesecondwaytoimprovetheoxygen-transfer ratebyapplyingtwoliquid phases
is as an emulsion. It has been reported many times that very small droplets of
hydrocarbonsorperfluorocarbonscanhaveanenhancingeffectonthe oxygen-transfer
rate. 12,32,33,34,35,36,37,45,57,58,59 Linek and Benes45 studied the mechanism of gas
absorption into organic solvent water emulsions. They distinguished organic solvents
onthe basis oftheir spreading coefficient:

S +
OV = *VA - (°OA °Ow) 0)

where0istheinterfacialtensionandthesubscriptsO,W, andAstandfororganicsolvent,
water and air, respectively. Organic solvents with a negative spreading coefficient
showed no influence on the mass-transfer coefficient, but had a negative influence on
thespecific exchange area.Organicsolventswithapositivespreading coefficient gave
positive effects on both the mass-transfer coefficient and the specific exchange area.
Hassanand Robinson 32foundresultsthatwereinagreementwiththedataof Linek and
Benes. 45 The increase of the specific exchange area in case of a positive spreading
coefficient is assumed to be caused by spreading of the organic solvent as athin film
onthe gas aqueous interface, loweringthe surface tension,andthereby increasing the
specific interfacial area of the gas dispersion. The decrease in case of a negative
spreading coefficient isassumedto becaused bystaticaccumulation of droplets atthe
interface,thus blocking partofthesurfaceareaavailableforoxygen masstransfer. Rols
etal. 5 7concludedthattheeffectoforganic-solventdropletsonthespecific gas-emulsion
exchange area is usually of minor importance comparedtothe enhancement effect on
the mass-transfer coefficient, and denoted the principle as oxygen vector. Several
mechanisms of the increased mass-transfer coefficients for organic solvents with a
positive spreading coefficient have been proposed 12 . 32 . 35 . 45 . 57 , but there is still much
uncertaintyleft.Itismostlikelythatinsomewaythereisaregularlyoccuring interchange
of droplets between the air aqueous interface and the bulk region. Some examples in
the literature are found that clearly demonstrate the positive effect of organic-solvent
droplets onvolumetric transfer coefficients. RolsandGoma5 9 found apositive effect of
16 CHAPTER1

soybeanoilonthevolumetricoxygen-transfercoefficientduringcultivationolAerobacter
aerogenes. Hoetal. 33foundanenhancedoxygensupplybyaddinghexadecanetothe
oxygen-limited penicillin fermentations withPenicilliumchrysogenumin shake flasks
andinastirredtank.Theydenotethetransfer-enhancement effectasahighereffective
oxygen solubility of the fermentation medium. Ju et al. 35 used Escherichia coli to
demonstratethe mass-transferenhancementeffect,wheninsteadofanaqueousphase
perfluorocarbonemulsionsareusedinastirred-tank reactor.
Uptoafive-foldincreaseofthe oxygen-transfercoefficientisreportedintheliterature58.
Therefore,thestrategyofaddingsmalldropletsofanorganicsolventtoafermentation
mediumseemstobeveryusefulforimprovingtheoxygen-transfer process.

CONCLUSIONSANDFUTURE PERSPECTIVES

Thenumber ofpublicationsaboutresearchwork onbiocatalysisinnon-aqueous

mediaisenormousandeventhereviewarticlesonthistopicarenumerous.Inspiteof
allthe advantages that the researchers come up with, serious applications are still
scarce.Itisgoodtorealizethatfromascientific pointofviewtremendous progressis
made,butthat the commercial success isstilllimited.Theoveroptimisticview,thatis
shownbymanyauthorswithrespecttoapplicationsoftwo-liquid-phasesystems,suits
therelativationonbiotechnology ingeneral,thatisgivenbyMosesandCape51 inthe
book entitled "Biotechnology, the science and the business". There is a world of
difference between laboratory achievements and development of processes for
commercialpurposes.
So far most laboratory biocatalytic reactions intwo-liquid-phase systems have
been restricted to batch processes. Interfacial effects make continuous processing
difficult.Continuous processing canonly beachieved ifcontinuous phase separation
isfeasible.Centrifuges or membranes can beagreat helpto achievethis. Ingeneral
thecomplexity of biocatalysis intwo-liquid-phase systems isasevere obstacle inthe
further development ofwhathasbeenachievedinthelaboratories.Afewconceptsof
two-liquid-phase bioreactors exist, butthese arestillinarather experimental stage :
thecontinuous extractionloopreactor (CELR)ofGianettoet al. 27 ,the liquid-impelled
loopreactor (LLR)ofTramperetal. 64 ,thestirred-tankreactor6.10,andthepacked-bed
reactor6.
 CHAPTER 1 17

REFERENCES

[I] Adlercreutz P., Mattiasson B. (1987). Aspects of biocatalyst stability in organic solvents.
Biocatalysis, 1,99-108.
[2] Aires Barros M.R. Cabral J.M.S., Novais J.M. (1987). Production of ethanol by Immobilized
Saccharomyces Bayanus in an extractive fermentation system. Biotechnol. Bioeng., 29,
1097-1104.
[3] Albertsson P.A. (1986).Partitioning ofcellparticlesand macromolecules.JohnWiley&Sons.
[4] AnderssonE.,Hahn-HagerdalB.(1990).Bioconversionsinaqueoustwo-phasesystems.Enzyme
Microb.Technol., 12, 242-254.
[5] BarR.,GainerJ.L.(1987).Acidfermentation inwater-organic solventtwo-liquid phasesystems.
Biotechnol.Prog.,3,109-114.
[6] BellG.,ToddJ.R., BlainJ.A., PattersonJ.D.E.ShawC.E.L (1981).Hydrolysis oftriglyceride by
solid phaselipolytic enzymes of Rhizopusarrhizus in continuous reactor systems. Biotechnol.
Bioeng,23,1703-1719.
[7] Black G.M. (1986).Characteristics and performance of immobilised cell reactors. In: C.Webb,
G.M. Black, B.Atkinson (eds.), Process Engineering aspects of immobilised cell systems,The
Institutionof Chemical Engineers,Warwickskire.
[8] Brink L.E.S., Tramper J. (1986). Modelling the effects of masstransfer on kinetics of propene
epoxidation of immobilized Mycobacterium cells: Pseudo-one-substrate conditions and
negligible product inhibition.EnzymeMicrob.Technol.,8,281-288.
[9] Brink L.E.S.,Tramper J. (1986). Modelling the effects of masstransfer on kinetics of propene
epoxidationof immobilized Mycobacterium cells:product inhibition.Enzyme Microb.Technol.,
8,334-340.
[10] BrinkL.E.S.,Tramper J., Luyben K.Ch.A.M.,Van't RietK. (1988).Biocatalysis inorganic media.
Enzyme Microb.Technol., 10,736-743.
[II] Bruce L.J., Daugulis A.J. (1991). Solvent selection strategies for extractive Biocatalysis.
Biotechnol.Prog.,7,116-124.
[12] Bruining W.J., Joosten G.E.H., Beenackers A.A.C.M., Hofman H. (1986). Enhancement of
gas-liquid masstransfer byadispersed second liquid phase.Chem.Eng.Sci.,41,1873-1877.
[13] Buitelaar R.M.,Susaeta I.,Tramper J. (1990).Application oftheliquid-impelled loop reactor for
theproduction ofanthraquinones by plant cellcultures.694-699. in:H.J.J. Nijkamp,L.H.V.Van
DerPlas,andJ.VanAartrijk(eds.),Progressinplantcellularandmolecularbiology,KluwerAcad.
[14] Carrera G. (1984). Biocatalysis in water-organic solvent two-phase systems. Trends in
Biotechnology, 2,102-106.
[15] Carrea G., Cremonesi P. (1987). Enzyme-catalyzed steroid transformations in water-organic
solventtwo-phasesystems.151-157In:K.Mosbach,Methodsinenzymologyvol136Immobilized
enzymesandcellspart C,Academic press.
[16] ChalmersJ.J.,BavarianF. (1991).Microscopicvisualizationof insectcell-bubble interactionsII:
The bubblefilmand bubble rupture. Biotechnol.Prog.,7,151-158.
18 CHAPTER1

[17] Cherry R.S., Papoutsakis E.T. (1986). Hydrodynamic effects on cells inagitated tissue culture
reactors.Bioprocess Engineering, 1, 29-41.
[18] ChibataI.,YamadaS.,Wada M.,IzuoN.,YamaguchiT. (1974).U.S. Patent 3,850,753
[19] ChoM.H.,WangS.S. (1988).Enhancementofoxygentransferinhybridomacellculturebyusing
a perfluorocarbon asanoxygencarrier. Biotechnol.Lett., 10, 855-860.
[20] DamianoD.,WangS.S.(1985).Noveluseofaperfluorocarbonforsupplyingoxygento aerobic
submerged cultures. Biotechnol.Lett.,7,81-86.
[21] DaugulisA.J. SwaineD.E. KollerupF. GroomC.A. (1987).Extractivefermentation-integrated
reactionand product recovery. Biotechnol. Lett.,9,425-430.
[22] DickinsonM.,FletcherP.D.I.(1989).Enzymesinorganicsolvents.EnzymeMicrob.Technol., 11,
55-56.
[23] Eggers, D.K., Blanch, H.W., Prausnitz, J.M. (1989). Extractive catalysis: solvent effects on
equilibriaofenzymatic reactionsintwo-phasesystems.EnzymeMicrob.Technol., 11,84-89.
[24] Evans P.J.,Wang H.Y. (1990). Effects of extractivefermentation on butyric acid production by
Clostridium acetobutylicum.Appl. Microbiol.Biotechnol.,32,393-397.
[25] Eiteman M.A., Gainer J.L. (1989). In situ extraction versus the use of an external column in
fermentation.Appl.Microbiol. Biotechnol.,30,614-618.
[26] FourierR.L.(1986).Mathematical modelofextractivefermentation:applicationtothe production
of ethanol.Biotechnol.Bioeng.,28,1206-1212.
[27] Gianetto A., Ruggeri B., Specchia V., Sassi G., Forna R. (1988). Continuous extraction loop
reactor (CELR): alcoholic fermentation by fluidized entrapped biomass. Chem. Eng. Sci.,43,
1891-1896.
[28] Hailing P.J. (1989). Organic liquids and biocatalysts: theory and practice. Trends in
biotechnology, 7,50-52.
[29] HamadaT., Sugishita M., FukushimaY., FukaseT., Motai H. (1991).Continuous production of
soysauce byabioreactor system.Process Biochemistry, 26,39-45.
[30] HandaA., Emery A.N., Spier R.E. (1987). Onthe evaluation of gas-liquid interfacial effects on
hybridomaviability in bubble column bioreactors. 7thGeneral Meeting of ESACT on advances
in animal cell technology: cell engineering, evaluation and exploitation Baden, Austria, 1985.
Develop,biol. Standard,66,241-253.
[31] Handa-Corrigan A.,EmeryA.N., Spier R.E. (1989).Effect of gas-liquid interfacesonthe growth
of suspended mammalian cells: mechanisms of cell damage by bubbles. Enzyme Microb.
Technol., 11,230-235.
[32] HassanI.T.M.,RobinsonC.W.(1977).Oxygentransferinmechanicallyagitatedaqueous systems
containingdispersed hydrocarbon.Biotechnol.Bioeng., 19, 661-682.
[33] Ho C.S., Ju L, Baddour R.F. (1990). Enhancing Penicillinfermentations by increased oxygen
solubilitythroughtheadditionof n-hexadecane. Biotechnol. Bioeng.,36,1110-1118.
[34] JuL-K. LeeJ.F.,ArmigerW.B.(1991).Effectoftheinterfacialsurfactantlayeronoxygentransfer
through the oil/water phase boundary in perfluorocarbon emulsions. Biotechnol. Bioeng.,37,
505-511.
 CHAPTER 1 19

[35] Ju L.-K., Lee J.F., Armiger W.B. (1991). Enhancing oxygen transfer in bioreactors by
perfluorocarbonemulsions. Biotechnol.Prog.,7,323-329
[36] Junker B.H., Hatton T.A., Wang D.I.C. (1990). Oxygen transfer enhancement in
aqueous/perfluorcarbon fermentation systems: I. Experimental Observations. Biotechnol.
Bioeng.,35,578-585.
[37] Junker B.H., Hatton T.A., Wang D.I.C. (1990). Oxygen transfer enhancement in
aqueous/perfluorcarbonfermentationsystems:II.Theoreticalanalysis.Biotechnol.Bioeng., 35,
578-585.
[38] KhmelnitskyY. L, LevashovA.V.,Klyachko N.L., MartinekK. (1988).EnzymeMicrob.Technol.,
10,710-724.
[39] KingAT.,MulliganB.J.,LoweK.G.(1989).Perfluorochemicalsandcellculture, Bio/Technology,
7,1037-1042.
[40] KlibanovA.M.(1989).Enzymaticcatalysisinanhydrousorganicsolvents.TrendsinBiochemistry,
14,141-144.
[41] LaaneC. (1987). Medium-engineeringfor bio-organic synthesis.Biocatalysis, 1,17-22.
[42] LillyM.D.(1982)Two-liquid-phasebiocatalyticreactions.J.Chem.Tech.Biotechnol.,32,162-169.
[43] LillyM.D.,BrazierA.J.,HocknullM.D.,WilliamsA.C,WoodleyJ.M.(1986).Biologicalconversions
involvingwater-insoluble organic compounds.3-17. In: C. Laane,J.Tramper, M.D. Lilly (eds.),
Biocatalysis inorganic media,Elsevier Science Publishers.
[44] Lilly M.D., Harbron S., Narendranathan T.J. (1987). Two-liquid phase biocatalytic reactors.
138-149. In: K. Mosbach,Methods inenzymologyvol 136Immobilized enzymes and cells part
C,Academic press.
[45] LinekV.,BenesP.(1976).Astudyofthemechanismofgasabsorption intooil-wateremulsions.
Chem. Eng.Sci.,31,1037-1046.
[46] LuisiP.,LaaneC.(1986).Solubilizationofenzymesinapolarsolventsviareversemicelles.Trends
inBiotechnology, 4,153-160.
[47] ManoT., KimuraT., lijima S.,Takahashi K., Takeuchi H., Kobayashi T. (1990). Comparison of
oxygen supply methods for cultures of shear-stress sensitive organisms including animal cell
culture.J.Chem.Tech.Biotechnol.,47,259-271.
[48] Mattiason B. (1983). Application of aqueous two-phase systems in biotechnology. Trends in
Biotechnology, 1,16-20.
[49] Mattiasson B. (1988). Bioconversions in aqueous two-phase systems: an alternative to
conventional immobilization. 657-667. In: K. Mosbach, Methods in enzymology vol 137
Immobilized enzymesandcellspart D,Academic press.
[50] Mattiasson B., Adlercreutz P. (1987). Perfluorochemicals in biotechnology. Trends in
Biotechnology,5,250-254.
[51] MosesV.,CapeR.E. (1991).Biotechnology,thescienceandthe business.Harwood Academic
Publishers.
[52] ParkY.S., Chang H.N., Kim B.H. (1988).Adaptation of Saccharomyces cerevisiaeto solvents
used inextractivefermentation. Biotechnol.Lett., 10, 261-266.
20 CHAPTER1

[53] Ramelmeier R.A., Blanch H.W. (1989). Mass transfer and cholesterol oxidase kinetics in a
liquid-liquid two-phasesystem.Biocatalysis,2,97-120.
[54] RofflerS. BlanchH.W. WilkeC.R.(1987).Extractivefermentationofacetoneandbutanol:process
designandeconomic evaluation.Biotechnol. Prog.,3,131-140.
[55] Roffler S.R., Randolph T.W., Miller D.A., Blanch H.W., Prausnitz J.M. (1992) Extractive
bioconversionswithnon-aqueoussolvents. 133-171In: B.Mattiason,0. Hoist (eds.), Extractive
bioconversions,Marcel Dekker.
[56] RofflerS.R. WilkeC.R. BlanchH.W. (1988).Designandmathematicaldescriptionofdifferential
contactors usedinextractivefermentations. Biotechnol.Bioeng.,32,192-204.
[57] RolsJ.L,Condoret J.S.,Fonade C, GomaG.(1990).Mechanism of enhanced oxygentransfer
infermentation usingemulsified oxygen-vectors. Biotechnol. Bioeng.,35,427-435.
[58] Rols J.L., Goma G. (1989). Enhancement of oxygen transfer rates in fermentation using
oxygen-vectors. Biotech.Adv.,7,1-14.
[59] Rols J.L, Goma G. (1991). Enhanced oxygen transfer rates in fermentation using soybean
oil-in-waterdispersions. Biotechnol.Lett., 13,7-12.
[60] RosevearA.(1984).ImmobilisedBiocatalysts-acriticalreview.J.Chem.Tech.Biotechnol.,34B,
127-150.
[61] SikdarS.K.,ColeK.D.,StewartR.M.,SzlagD.C.,ToddP.,CabezasH.(1991).Aqueoustwo-phase
extraction inbioseparations:anassessment.Biotechnology,9,253-257.
[62] Sola C. Casas C. Godia F. Poch M., Serra A. (1986). Continuous ethanol production by
immobilized yeast cells and ethanol recovery by liquid/ liquid extraction. Biotechnol. Bioeng.
Symp., 17,519-533.
[63] Springham D.G. (1991). The established Industries. 249-296. In: V. Moses, R. E. Cape.
Biotechnology,thescienceandthe business,HarvardAcademic Publishers.
[64] Tramper J., Wolters, I., Verlaan P. (1987). The liquid-impelled loop reactor: a new type of
density-difference-mixed bioreactor. 311-316. In: C. Laane, J. Tramper, M.D. Lilly (eds.),
Biocatalysis inorganic media,Elsevier Science Publishers.
[65] VandenTweelW.J.J.,MarsmanE.H.,VorageM.J.A.W.,Tramper J., DeBontJ.A.M. (1987).The
aplication of organic solvents for the bioconversion of benzene to cis-benzeneglycol.
International conference onbioreactors andbiotransformations, Gleneagles,Scotland,UK.
[66] Vander Padt A., Keurentjes J.T.F., SewaitJ.J.W, Van Dam E.M.,Van Dorp L.J., Van 't Riet K.
submittedfor publication
[67] VanSonsbeek H.M., De Blank H.,Tramper J. Oxygentransfer in liquid-impelled loop reactors
using perfluorocarbon liquids,submittedfor publication
[68] Venkatasubramanian K., Karkare SB., Vieth W.R. (1983). Chemical engineering analysis of
immobilized-cellsystems.Appl. BiochemBioeng.4,311-349.
[69] WoodleyJ.M., Cunnah P.J., Lilly M.D. (1991).Stirredtanktwo-liquid phasebiocatalytic reactor
studies: kinetics,evaluationand modelling ofsubstrate masstransfer. Biocatalysis,5,1-12.
 21

CHAPTER 2

HYDRODYNAMIC MODEL FOR LIQUID-IMPELLED

LOOP REACTORS

SUMMARY

Anewtypeofbioreactor, inwhichtwoliquidphasesareused,isintroducedand
amodeldescriptionofitshydrodynamic behavior isgiven.Themodelisbasedonthe
two-phase drift-flux model of Zuber and Findlay and on afriction coefficient derived
from one-phase flow theory. Drop sizes, necessary for model calculations, were
estimatedfromliteraturecorrelationsandphotographicallyverified.Themodelpredicts
the continuous-phase circulation velocity and dispersed-phase holdup within an
accuracyof5%forapilot-plantreactor,andwithin 10%foralab-scale reactor.Onlyat
verylowflowranges,especiallyonlabscale,isthisaccuracy notattained.

INTRODUCTION

Biotechnological developments concern manipulating biological materials to

produce high-value products. To achieve optimum performance of biotechnological
processes, the biocatalyst can be modified in addition to adjustments of the
environmental conditions. Traditionally, most of these processes are carried out in

Thischapter hasbeenpublished inBiotechnology &Bioengineering,vol.36,940-946(1990)

bytheauthors :H.M.VanSonsbeek, R.E.M.Verdurmen, P.VeriaanandJ.Tramper.
22 CHAPTER 2

aqueous solutions, though introduction ofnon-conventional media, such asorganic

andsupercriticalsolvents,cangiveconsiderableadvantages.2Thesearemerelybased
onthedistributionofsubstratesandproductsovertheaqueousphaseandthesolvent.
Theperformanceofbiologicalsystemscanalsobeaffectedtoalargeextentbythefluid
dynamicsasbiocatalystsareoftensensitivetohydrodynamicforcesandconcentration
gradientsofsubstratesandproducts.Theair-lift loopreactor (ALR)isanexampleofa
system inwhich improvement ofthe performance can beaccomplished. Adequate
mixingandmasstransfercharacteristicsarecombinedwithlowshearforcesinthistype
of reactor.11

OUTLET

ORGANIC SOLVENT

CONTINUOUS PHASE

00WNC0MER
D d = 0.04 m

INDUCTIVE FLOWMETER
STAINLESS STEEL

INLET

Figure 1 : Lab-scale liquid-impelled loop reactor (LLR).

 CHAPTER2 23

The same principle of the ALR is used in the liquid-impelled loop reactor (LLR).9
Insteadofagasphase,adispersedliquidphaseinducesthecirculationofthe continuous
liquidphase inthistypeofreactor. Severalconfigurationsofthese reactorsarepossible:
internalor external loop,upflowor downflowofthedispersed phase,or a combination
ofthese.Itisalso possibleto combine liquidinjectionwithgas injection.There areafew
constraints for the two liquid phases that can be used inthe LLR. The phases should
be immiscible and must have a different density. Furthermore, the phases should not
have a serious toxic effect on the biocatalyst used inthe reactor.
Inthe LLRthe advantages of organic media andoftheALR are combined.As the
biocatalyst can easily be retained inthe continuous phase andthus inthe reactor, the
dispersed phase can be used, in addition to supplier of substrate, for downstream
processing. For these reasons this type of reactor has been developed. To meet
requirements of a biotechnological process and to optimize the use of the specific
advantages of the LLR, a good description of the reactor and the operation
characteristics is essential. Hydrodynamic properties, such as circulation velocity,
holdup,anddrop sizeareof great influence on shearforces, mixing,and mass-transfer
performance.Therefore, inthis chapter ahydrodynamic model ispresented describing
the circulation velocity and the dispersed-phase holdup for different flow rates of the
dispersed phase for an up-fiow LLRwith external loop.

THEORY

The model
The hydrodynamic model presented inthis chapter is based on the assumption
that friction in the LLR can be characterized by a friction coefficient obtained from
one-phaseflowcalculations.Furthermore,themodelisbasedonthetwo-phase drift-flux
theory ofZuber and Findlay 13 givinganestimateforthevelocity ofthe dispersed-phase
drops.
Thepressure dropA Pofaone-phaseflowthrough astraight pipewith obstacles,
suchassudden changes incross-sectional area,fittings,andvalves, canbe described
by the following equation:
24 CHAPTER2

A P = l i / i P < , > f ( | -Lj\( + X

W l ^ p <, , .>
e
2^ (1)

wherethesubscript/denotesthepertinentpipesectionandj thepertinentobstacle ,f
isthe wall-friction coefficient for straight pipes,p isthe density of the fluid, <v> the
averagevelocity,LandR^ arethelengthandthehydraulic radiusofthepipeandevis
thefrictioncoefficientforobstacles.
Astheflowthrough each pipesectionandobstacle isthesame,allthe different
velocities used inequation (1)areproportionally interrelated.Therefore,thisequation
canberewrittentoaforminwhichonlyonearbitraryvelocitytermisused:

1 2 (2)
AP = -Kfp<v>2

whereK\istheoverallfriction coefficient ofthe systembasedonthe averagevelocity

<v> inoneofthepipesections.
The same equation can be used for an LLR,which can be considered as two
straightverticalpipesectionswiththebottomandtopconnectingobstacles (figure1).
Whenitisassumedthatthedispersed-phase holdupisconstantoverthelengthofthe
riserandthatthereisnodispersedphaseinthecirculationloop,thehydrostaticpressure
difference betweenthetwocolumnsis:

AP=aApgH (3)

wherea isthe holdup,Apisthe density difference of thetwo liquid phases,g isthe

gravitationalconstant,andHisthelengthoftheriser.Combinationofequations(2)and
(3)gives:

1 2 (4)
aApgH =-K fp <v>

Inthisequationtherelationshipbetweenthedispersed-phaseholdupandthecirculation
velocityofthecontinuousphaseisshown.Arelationshipwiththedispersed-phaseflow
rateasthe main controlling operation condition is desiredfor predicting thevalue of
 CHAPTER2 25

bothhydrodynamicpropertiesunderagivenset-up.Therefore,arelationbasedonthe
theory of Zuber and Findlay13, is introduced, giving an estimate for the velocity of
dispersed-phase drops (v^):

u d = C O s d +t;sc)+od.* (5)

where Vsdandvsc are the superficial velocities of the dispersed and the continuous
phaseintheriser,respectively;Cisthedistribution parameter,andVd<x> istheterminal
velocityofasingledropinaninfinitevolumeofthepertinentcontinuous phase.HereC
correctsfor nonuniformflow andconcentration profiles ofthe dispersed phase. Ifthe
concentrationofdispersedphaseisuniformlydistributedoverthecrosssectionofthe
tube,thevalueof Cis 1. Iftheconcentration ofthedispersed phase atthewallofthe
tubeiszeroandthevelocityprofileispronouncedparabolic,Chasavalueofabout1.5.
Withequation (5)therelationbetweentheholdupandthedispersed-phaseflowrateis
foundastheholdupcanbedescribedbythefollowingequation:

(6)
a= —

Estimatesforthehydrodynamic properties atvariable dispersed-phaseflow ratesare

obtained by combining equations (4),(5), and (6). From the resulting equation the
dispersed-phase holdup andthecirculationvelocity caneasilybederived.

Dropsize
Thesizeofthedispersed-phasedropsplaysanessentialroleinthehydrodynamic
model,asitisoneofthemainpropertiesthatdeterminesv^ooinequation (5).Forthis
purposefigure2isderivedfromagraphical correlationgivenby Huand Kintner4 and
givestherisevelocityofasingledropasafunctionofthedropsizeforahexaneinwater
system.Thedropsizeisalsoofmajorimportanceforgoodperformanceofthereactor.
Toosmalldropletsmayenterthecirculationloop,resultinginadecreaseofthedriving
forceforliquidcirculation.Increasingdropsizewilldiminishthemass-transfercapacity.
26 CHAPTER2

Thesizeofdropsformedatanozzledependsontheliquidvelocity inthenozzle.
Clift et al.3 distinguish four flow regimes each being responsible for a characteristic
drop-formationmechanism.
Atverylowvelocitiesinthenozzlethedropsareformedatthenozzleonebyone,
being rather big and uniform. Scheele and Meister8 proposed a general theory to
calculatethedropsizeinthisregime,whichwillnotbediscussedfurtherinthischapter.

0.15
v
dm [ms"1]

0.05-

Figure2: Terminal risevelocityof hexanedrops inwaterasfunctionofdropdiameter; derivedfromHu

and Kintner.4

Inthesecond regimeajetisformed,whichbreaksupinsmallerdroplets.Thejet
breakupismainlycausedbyaxisymmetricamplificationofjetsurfacedisturbances.This
so-calledRayleighinstabilityleadstoauniformdropsize.ScheeleandMeisterSgivethe
following equation to estimate at which liquid velocity inthe nozzlejet formationjust
startsandregime 1 changesintoregime2:

(7)
V
jetting 1 .r O
PaDt d
 CHAPTER2 27

wheredisthedropdiameteraccordingtothetheoryforthefirstregime,0isthe interfacial
tension, pcisthe density of the continuous phase and D N is the nozzle diameter. The
size of drops inthe second regime can be estimated by thefollowing equation: 7

- - (8)
d = (1.5Xm)3JD3N

where d is the diameter of the drop and xm is the wavelength of the fastest growing
disturbance of the jet surface. Meister and Scheele5 give a graphical correlation to
determineAmfromthephysical properties ofthetwo liquidphasesandtheradiusof the
jet. The latter can be assumed to be V2DN.6 Inthis regime thejet length first reaches a
maximum and then decreases again at increasing velocity inthe nozzle.
Inthethirdregimeasymmetricdisturbancesofthejetsurfacebecome increasingly
significant and the jet length decreases further. The drop formation process becomes
rather erratic and the drop-size distribution widens. 6
Inthe fourth regime thejet has completely disappeared and many small droplets
of non-uniform size areformed atthe nozzle. 3
The formation of very small satellite droplets occurs in each regime and is
unpredictable because of the very low forces involved, compared to the forces that
causethedescribedjetbreakup. 7Generallyitcanbestatedthatthedropsizedecreases
andthe drop-size distribution widens for increasing liquid velocity inthe nozzle.

MATERIALS AND METHODS

The experiments have been carried out at 30 °C in a lab-scale and a pilot-plant

reactor with external loop, constructed of glass cylinders, with volumes of 0.004 and
0.165 m 3 , respectively (seefigure 1).The diameter of the riser and downcomer of the
pilot plant reactor are0.2 and0.1 m, respectively. In both cases external loop reactors
wereusedwithlengthsof0.55and3.14m.Theliquidspargersweredesignedto produce
uniformdrops.They consisted of stainless steelpipeswithan inner diameter of 1.4mm
and an outer diameter of 2 mm and were evenly distributed over the cross section of
the main reactor tube. A chamber at the inlet of the pipes gave a uniform pressure
distributionoverthepipes.Thespargersofthelab-scaleandthepilot-plantreactor were
28 CHAPTER2

equippedwith 12pipesof0.2 mand49pipes of0.3 mlength,respectively. Boththe

reactorswereusedwithwaterasthecontinuous phaseandhexaneastheascending
dispersedphase.Thiscombinationwaschosenasthetwophasesareimmiscibleand
have arather different density, 1000and660kg/m 3 , respectively. Thetoxic effect of
hexane was irrelevant in these experiments as no biocatalyst was used. For this
combinationofliquidphasesaAmvalueof6.7x10-3mwasderived.5
Tomeasuretheholdupofthedispersed phaseinthemaintubeofthereactor,a
reversedU-tubemanometerwasused.Thesensitivityofthemanometerwasexpanded
byusingethyl-acetateontopofthewaterinbothtubesofthemanometer. Ethyl-acetate
isnotmisciblewithwaterandhasadensity (990kg/m3) closetothedensityofwater.
Thecirculationvelocityofthecontinuousphasewasmeasuredbymeansofaninductive
flowmeter,locatedinthecirculationloop.
Drop-sizedistributionsweredeterminedbymeasuringdiametersofphotographed
drops.Thephotographsweretakenthroughtheflatsurfaceofasquareboxfilledwith
water, and put up around the cylindrical maintube ofthe reactor to eliminate errors
causedbyrefraction.

RESULTSAND DISCUSSION

Frictioncoefficients
Equation(4)wasusedtoderivefrictioncoefficientsfromtheexperimentaldataon
holdup andcontinuous-phase velocity. Theaverages ofthesecoefficients were used
forfurther modelcalculations.
Theflow regime inthe lab-scale reactor rangedfrom lowturbulent to turbulent.
TheReynoldsvaluesRevariedfrom1.3x103to104fortheriserandfrom2x103to15x103
for the downcomer. Inthe low turbulent regimes it can be difficult to predict friction
coefficients asthesemay notbeconstant. Infigure3indeedaslightinitialincreasein
frictioncoefficientwithReisshown.However,thefrictioncoefficientforthisreactorcan
beapproximatedbyaconstantvaluewhenthevaluesforlowRenumbers(<8x103)are
excluded.
 CHAPTER2 29

Forthepilot-plant reactor Rerangedfrom8x103to 24x103fortheriserandfrom

16x103to48x103forthe downcomer.Intheseturbulentflowregimesfrictioncoefficients
may beassumedtobe independent ofthevelocity.Theexperimental results confirm
thisassumption (figure4).
Inthis way thefriction coefficients for the lab-scale and pilot-plant reactor were
foundtohaveavalueof1.94and2.05,respectively.Thesevaluesagreeverywellwith
thepredictedvaluesof1.74and1.96,respectively (seeAppendix).

3.0
K f[-]
2.5

2.0 •
•nn nu °-tP^

1.5H
a •
1.0 D

0.5H

0.0
0 5000 10000 15000
Re[-]
Figure3: Friction coefficient asfunction ofRe inthe downcomer for the lab-scale LLR (experimental:
• ; averageforRe > 8x103: ).
30 CHAPTER2

3.0
K f [-]
2.5

2.0H " • , - . ^

1.5

1.0

0.5

0.0
0 25000 50000
Re [-]

Figure4: Frictioncoefficient asfunction ofReinthedowncomerforthe pilot-plant LLR (experimental:

• ; average: ).

Dropsize
TheSauter meandiameter (cfo) ofthedrops inthe pilot-plant reactor hasbeen
determinedatvariousflowratesandcomparedwiththepredictedvalues (figure5).As
expected,the drop sizedecreasesfor increasingflow rate.Atvery lowflow ratesthe
drops are formed separately, and using Scheele and Meister's theory7 for the first
drop-formationregime,thepredicteddropsizedeviatesonly5%fromtheexperimental
average.
Fromequation (7)itfollowsthatthesecondregimestartswhentheliquidvelocity
inthe nozzle exceeds 0.39 m.s-1,corresponding withthe pilot-plant reactor having a
superficialvelocityof9x10"4m.s-1.Atthisflowratejetformationatallnozzlesisnoticed.
However,jetformation isalready observed atsomenozzlesatasuperficialvelocityof
 CHAPTER2 31

5X10- 4 m.s-1,where at others drops are stillformed separately. Therefore the region
from this velocity up to the calculated jetting velocity can be considered as the
intermediateregion.

u.uuo

0.005-
D
d [m]
0.004 -
•
a
0.003- D a
D

0.002-

0.001-

0 i

C) 0.0025 0.0
1
V
sd [m.s- ]

Figure5: Dropdiameter asfunctionofthesuperficialvelocity ofthedispersed phaseintheriserofthe

pilot-plant reactor (experimental values for d 3 2 : o; predicted drop diameter using Scheele
and Meister'stheory 8 : — ; predicteddropdiameter using equation (8): ).

Increasingtheflowrateofthedispersedphaseinthesecondregime,theaverage
dropsizedecreasesandthedrop-sizedistributionwidensconsiderably. Especiallythe
numberofverysmalldroplets(<0.5mm)increasesdramaticallyforincreasingflowrate.
Obviously regimes 2 and 3 are hardly distinguishable and appear at the sametime.
However,theestimateddropsizefromequation(8)describestheSautermeandiameter
inthisflow regionwithin 10%accuracy.Therefore,thisequationcanbeveryusefulfor
furthermodelcalculations,astheterminalvelocityisgivenasafunctionofthedropsize
infigure2.
32 CHAPTER2

Themaximumflowrateinthenozzleswhichstillgivesanacceptable performance
oftheLLRwasvisuallydeterminedtobeabout1.7 m.s-1forboththelab-scaleandthe
pilot-plantreactor.Abovethisflowratetoo manysmalldropletswereformed,causing
separation problems.

Themodel
Theexperimentshaveallbeencarriedoutin alimitedflowrangeasthemaximum
liquidvelocity inthe nozzlewaskeptcloseto 1.7m.s-1for acceptable performanceof
thereactor. Fromvisualobservationnochangeofflowregimecouldbeobserved.The
shape ofthe drops was rather spherical and did not change muchat increasingflow
rate.Therefore,thedistribution parameter Cwas assumedto beconstant inthisflow
range and was determined from the experimental measurements of holdup and
circulationvelocity,whereforv^ aconstantvalueof0.12m.s-1wasusedderivedfrom
figure 2. For both lab-scale and pilot-plant reactor, a value of 1.6 was found for C,
suggestingavery pronounced parabolicvelocity andconcentration profileacrossthe
duct.Thiseffectcaneasily beseeninthe lowerpartofthepilot-plant reactor,wherea
veryunevenconcentrationdistributionandhighlocalvelocitydifferencesoccurreddue
to the unestablished flow after the bottom section. At greater height, the effect
diminished.
Thevalue1.6forCisveryhighcomparedtothevalueof1.07fortheALRusedby
Verlaanetal. 10 However,ReintheriseroftheALRvariedfrom25x103to105,wherefor
theLLRtheRerangewas8x103to24x103.Theconsiderablylowerflowrangeverylikely
influencesthevalueofC. Atverylowflowranges,localcirculationcellscanbeinduced,
which cannot be represented by the distribution parameter C. In that case the
hydrodynamic modelisnotvalid.13
Infigures6and7modelpredictionsandexperimentalresultsarepresented.The
modeldescribesthevelocityandholdupofthepilot-plantreactorwithinanaccuracyof
5%, except for a dispersed-phase superficial velocity below 5x10-4 m.s-1. Lab-scale
predictions are less accurate.Above adispersed-phase superficialvelocity of3x10"3
m.s-1thedeviationofthepredictedcirculationvelocity islessthan 10%.Fortheholdup
predictionaccuracyof10%isonlyattainedaboveadispersed-phasesuperficialvelocity
of9x10'3m.s-1.Thedisturbancesofflowprofilesinaverticaltubecausedbythebottom
section can be the reason for the deviation from the model, which is based on
fully-establishedflow.Onlabscaleagreaterpartofthereactorisinvolvedinsmoothing
 CHAPTER2 33

outthiseffect,whichcanthusexplaintheloweraccuracyonthisscale.Nevertheless,
the model can be usedfor suitable prediction onthis scaleaswell,especially inhigh
flowranges.

0.06
0.30- a [-]
v
csd [ m s " 1 ]

0.005 0.01 0.015

v
sd [ m s " 1 l

Figure6: Thesuperficialvelocity ofthecontinuous phaseinthedowncomerandthedispersed phase

holdupintheriser,bothforthelab-scaleLLR(experimentalcirculationvelocity:• ; experimental
holdup:O; model predictions: ).
34 CHAPTER2

0.02

v
a [-]
csd [m.s- 1 ]
-0.015

-0.01

0.005

v s d [m.s" 1 ]

Figure7: Thesuperficialvelocity ofthecontinuous phaseinthedowncomerandthedispersed phase

holdup in the riser, both for the pilot-plant LLR (experimental circulation velocity: n;
experimental holdup:O; modelpredictions: ).

CONCLUSIONS

Thehydrodynamicbehaviorofaliquid-impelledloopreactorcanbedescribedby
themodelbasedonone-phaseflowfrictionlossesandontwo-phasedrift-fluxflow.The
valueofthedistribution parameter Cunder ourworkingconditions is 1.6,indicatinga
pronounced parabolic velocity profile and volumetric concentration distribution. The
estimation of drop size is reliable and can be usedfor derivation of theterminal rise
velocity inan infinite medium.The holdup andvelocity for pilot-plant reactors can be
predicted within an accuracy of 5%for a dispersed-phase superficial velocity above
 CHAPTER 2 35

5x10-4m.s-1.Forlab-scalereactorsthemodelisalsoapplicable,butlessaccurate,due
toflow disturbance causedbythebottomsection.Accuracy of 10% isattainedabove
adispersed-phase superficialvelocityof9x10"3m.s-1.

ACKNOWLEDGEMENT

ThisworkwaspartlysupportedbyApplikon DependableInstruments B.V.,Schiedam,

TheNetherlands.TheauthorswishtothankH.H.Beeftinkforthevaluablediscussions.

NOMENCLATURE

Ad downcomer cross-section m2

Ar riser cross-section m2

A, top cross-section m2

C distributionparameter -
D tubediameter m
DN nozzlediameter m

d dropdiameter m
d-32 Sauter meandropdiameter m

eu frictioncoefficientfor obstacles -
f wall-frictioncoefficient -
g gravitational constant m.s
H distancefromnozzleopeningtoseparationlevel m

Kf overallfriction coefficient -
L tube length m
R, Hydraulic radius m
Re Reynoldsnumber -
36 CHAPTER2

vcsd superficialcontinuous-phasevelocity inthedowncomer m.s-'

vsd superficialdispersedphase-velocity intheriser m.s-1
vd velocitydispersed-phase dropsintheriser m.s-
vsc superficialvelocitycontinuous phaseintheriser m.s-
vda, terminalvelocityofasingledropinaninfinitemedium m.s-'
f jetting lower limitvelocity innozzleatwhichjettingoccurs m.s-'
< v> averagevelocity m.s-'
a holdup
AP pressuredifference Pa
Ap densitydifference dispersedandcontinuous phase kg.m-3
P density kg.m-3
pd densitydispersed phase kg.m-3
a interfacialtension N.nr1
Xm wavelengthoffastestgrowingdisturbanceofjetsurface m

REFERENCES

[1] BeekW.J., MuttzallK.M.K. (1975).Transport phenomena,JohnWiley&Sons.

[2] BrinkL.E.S.,TramperJ.,LuybenK.Ch.A.M.,Van't RietK(1988).Biocatalysisinorganicmedia.
Enzyme Microb.Technol., 10,736-743.
[3] CliftR., GraceJ.R.,Weber M.E. (1978).Bubbles,dropsand particles,Academic Press.
[4] HuS.,Kintner R.C.(1955).Thefallof singleliquiddropsthroughwater.AlChEJ., 1,42-48.
[5] Meister B.J., Scheele G.F. (1967). Generalized solution of theTomotika stability analysisfor a
cylindricaljet.AlChEJ., 13, 682-688.
[6] MeisterB.J.,ScheeleG.F.(1969).Predictionofjetlengthinimmiscibleliquidsystems.AlChEJ.,
15,689-699.
[7] MeisterB.J.,ScheeleG.F.(1969).Dropformationfomcylindricaljetsinimmiscibleliquidsystems.
AlChEJ., 15,700-706.
[8] ScheeleG.F.,MeisterB.J.(1968).Dropformationatlowvelocitiesinliquid-liquidsystems.AlChE
J., 14, 9-19.
 CHAPTER 2 37

[9] Tramper J., Wolters, I., Verlaan P. (1987). The liquid-impelled loop reactor: a new type of
density-difference-mixed bioreactor. 311-316. In: C. Laane, J. Tramper, M.D. Lilly (eds.),
Biocatalysis inorganic media, Elsevier Science Publishers.
[10] Verlaan P.,Tramper J.,Van 't Riet K., Luyben K. Ch.A. M. (1986).A hydrodynamic model for
airlift-loop bioreactorwithexternalloop.Chem. Eng.J.,33,43-53.
[11] Verlaan P. (1987). Modelling and characterization of an airlift-loop bioreactor. Ph.D. thesis,
Agricultural UniversityWageningen,TheNetherlands.
[12] WallisG.B. (1969).Onedimensional two phaseflow,McGraw-Hill.
[13] Zuber N., FindlayJ.A. (1965).Averagevolumetric concentration intwo-phase flow systems.J.
HeatTransfer, 87, 453-468.

APPENDIX

Calculationofthefrictioncoefficient

Thefrictioncoefficientforone-phaseflowcanbecalculatedtheoreticallyusingthe
expressionsfor straight pipes andobstacles.Thefriction coefficient for straight pipes
isgivenby:

(A.1)
K,,pipe =\ - 1D,* 4 * /

whereLandDarethelengthanddiameterofthepipe,respectively,andfisthefriction
factor for straight pipes. Wallis12 proposed anf-value of 0.005for turbulent flow and
smoothpipes.TheK\valuesforobstaclesaregivenby BeekandMuttzall.1
Thefrictioncoefficientsarebasedonthevelocityinthepertinentreactorsections.
In the different sections different velocities appear and therefore an overall friction
coefficient,asexplainedinthechapterTheory,iscalculated.Basedonthedowncomer
velocity,thefollowingformulaforK\canbeused:

A2d A\ (A-2)
IS — # IS + _ * IS _j_ IS
f A2 f •riser .2 f < t°P f' downcomer
Ar At

whereAisthecrosssectionandKfisthefrictioncoefficient.
38 CHAPTER2

Intable ItheKfvalues usedfor calculation ofthe overallfriction coefficients are

given.Theabove equation gives afriction coefficient of 1.74 for the lab-scale reactor
and1.96forthepilot-plant reactor.

TableI Dimensionlessfrictioncoefficientsforall reactor sections.

Reactor section Pipeor obstacle Lab-scaleLLR Pilot-plant LLR

Riser Cross-sectional change 0.32 0.59

Bend 0.40 0.40
Straight pipe 0.19 0.31

Top Cross-sectional change 0.23 0.09

Bend 0.40 0.40
Straight pipe 0.05 0.05

Downcomer Cross-sectional change 0.05 0.36

Bend (2x) 0.80 0.80
Straight pipe 0.44 0.70
 39

CHAPTER3

LIQUID-CIRCULATION MODELS FOR TWO-PHASE

LOOP REACTORS

SUMMARY

The conformity of the great number of hydrodynamic models that exist for
two-phase loop reactors is investigated. At least two independent equations are
necessary to calculate the two desired quantities: the holdup and liquid velocity as
functionofthedispersed-phaseflowrate.Theseequationscanbechosenfromempirical
correlations andfromtheconservation lawsofmass,momentumandenergy. Inmost
casestwooftheconservationlawsarecombinedwithempiricalelementsandusedas
thetwoequationsthatgivethehydrodynamicdescription.

INTRODUCTION

The increased interest for biotechnological processes has also induced the
development of newtypes of reactors. Besides the conventional stirred and aerated
vessel, the bubble column (BC) has gained great popularity due to its simple
construction. By means of mounting an extra tube in the BC, circulation flows are

Thischapter isto bepublished bytheauthors:H.M.VanSonsbeek,W.A. Beverlooand K.Van

't Riet.
40 CHAPTER3

enhanced andthe resulting device iscalled air-lift loop reactor (ALR).The extratube
maybeinthecentreoftheBC(internal),whereitusuallyfuntionsasariser. Itmayalso
beoutsidethe BCasaby-pass (external),whereitfunctionsasdowncomer.
Despite of the enormous amount of research done on ALRs, different and
ill-founded opinions exist about the advantages of ALRscompared to BCs or stirred
aeratedvessels.AmainadvantageofALRscomparedtostirredvesselsistheabsence
of mechanical moving parts, which is especially desirable on very large scales.
Furthermorefermentationheatremovalismoreconvenient. InanALRhighergasflow
ratesarepossiblethaninaBC,whichincreasesthecapacity ofthereactor. Although
mixingtimesarelonger,ALRshaveabetter-controlled mixingbehavior.
Analogousto BCs,liquid/liquid extraction columnscanalso beadjustedto loop
reactors.Thecirculationoftheliquidcontinuousphaseisinducedbyaliquiddispersed
phase,whichmaybeascendingsuchasairinanALR,ordescendingincasethedensity
ishigherthanthedensity oftheliquidcontinuous phase21.
Inrecent literature many articles can befound on modelling hydrodynamics of
ALRs.Hydrodynamicsareofgreatinfluenceonmassandheattransfer. Dropsizeand
holduptogetherdefinethetotalexchangeareaandthereforealsotheoverallvolumetric
masstransfercoefficient {ka).Theliquidvelocityisamixingcharacteristicandisalsoof
importanceforshearstresses,whenitconcernsshear-sensitive organisms2.
Thischapter ismainlyfocussedonmodelsto predict average liquidvelocity and
volumetric dispersed-phase concentration (holdup) in ALRs but the principles are
generally valid. Less attention is given to descriptions of mixing characteristics. The
subject is also restricted to Newtonian flow, because non-Newtonian flow is a vast
researchareabyitself,withmany unclarifiedaspectsandmanyempiricalcorrelations.
Theaimofthischapteristogaininsightinthedifferenttypesofmodelsandthegeneral
problems that make hydrodynamic description oftwo-phase flow inloop reactors as
difficult asitis.

FLUIDDYNAMICS

Thephysicalaspectsofanyfluidflowaregovernedbythreefundamentalprinciples:
theconservation lawsofmass,momentumandenergy. Formodelling offluidflowthe
appropriatefundamental physicalprinciples haveto bechoosen.Byapplyingthemto
 CHAPTER 3 41

asuitableflow model,mathematical equations canbededuced,which canbesimple

analyticalequations,integralequationsorpartialdifferentialequations.Thelatteristhe
mostgeneralform.
The integral and partial differential form of equations are mostly solved with
numericaltechniques. Forone-phaseflowthis hasbeensuccessful inmany fields of
application.Stillthistechnique isnot easily applicablefor many systems.Thebiggest
obstacle is the proper numerical implementation of the boundary conditions, and
sometimes ofthe initial conditions aswell. Furthermore turbulence is a phenomenon
thatispoorlyunderstoodandthereforedifficulttoincorporateinmodels. Anewfieldof
research has developed under the name computational fluid dynamics (CFD) that is
dealingwiththiskindofproblems.Two-phaseflowisevenmuchmorecomplicatedand
lessunderstoodthanone-phaseflow.Theinteractionbetweenthetwophasesisdifficult
todescribeindetailandthisisthereasonwhythemodelsproposedfortwo-phaseflow
intheliteratureareusually basedonequationsofthesimpleanalyticaltype.
A completely different type of equation is the empirical correlation, which is a
mathematicaldescriptionofmeasureddataandnotbasedonanyofthebasicphysical
principles. Equations of this type can either be used as independent models, or in
combinationwiththephysically-basedequations.
Inthischaptertwo-phasefluid-flowmodelsarediscussed.Itisshownhowphysical
principles and empirical correlations are used to obtain at least two independent
equationsthatgiveafulldescriptionoftheholdupandtheliquidvelocity asfunctionof
thegasinputrate.

TWO-PHASE FLOW

Behavior oftwo-phaseflow isstrongly dependent onfluxesof bothphases.The

flowpatternsthatcanbeobservedvisually,areagoodcriteriumtodistinguishdifferent
flowranges,whichhaveeachtheirowncharacteristichydrodynamic behaviour.These
flowrangesarecalledflowregimesandareoftenfurthersubdivided.
Wallis24distinguishesforgas-liquidsystemsbubble,slug,annularanddropflow.
For ALRs only the first flow regime is relevant. Bubble flow is characterized as a
suspensionofdiscrete bubbles inacontinuous liquid. Typicalfortheslug-flow regime
are the big cylindrical bubbles that almost fill the duct. The bubble-flow regime is
42 CHAPTER3

subdividedbyZuberandFindlay26inanidealbubbleflowregimeandinachurn-turbulent
flow regime, also called heterogeneous flow regime. Inthe first regime bubbles rise
uniformly and steadily. Inthe second regimethree-dimensional effects are important
andbubbleentrainmentineachother'swakesissignificant26.Thisregimeisatransition
fromidealbubbleflowtoslugflow.VanderLans22concludesthattheidealbubbleflow
infact impliesabsenceofallinfluencesthatmakeflowheterogeneous, suchasradial
variations,andthatitonlyexistsundercarefullycontrolledlaboratoryconditions.From
thesedescriptions itbecomesclearthatforALRstheheterogeneousflow regimeisof
maininterest.

MASSCONSERVATION :DRIFT-FLUX EQUATIONS

Whentheprincipleofmassconservationisappliedtotwo-phaseflow,anequation
isobtainedthatcanbeusedinhydrodynamicmodels.Forthesakeofsimplicityisobaric
conditions areassumedfortheequations givenhere,which meansthatthe densities
areconsideredtobeconstant. Modelsbasedontheseequationscaneasilyberefined
by non-isobaric modelling as shown byVerlaan et al. 23 (seeAppendix A).Whenitis
assumedthatalldispersed-phase bubbleshavethesamevelocityandthecontinuous
phasevelocity isuniform,continuity considerationsfortwo-phaseflow ingenerallead
to:

<*>d (1)

and:
<t>c (2)
,4(1 - a )

wherethesubscriptsdandcdenotedispersedandcontinuousphase,respectively.4>
isthevolumetricflowrate,visthesuperficialvelocitycorrectedforholdup,Aisthecross
sectionalareaofthetubeanda isthedispersed-phase holdup.Therelativevelocityor
slipvelocitybetweenbothphasescanbedefinedas:

vs = vd - vc (3)
 CHAPTER 3 43

Fromequations (1),(2)and(3)thefollowingequationcanbederived:

<*>« •$>* *c ,, (4)

+ + u (1 a)
U ' T 1 - -
Thisisthebasicformofthe continuityequationfortwo-phaseflowandequationsofthis
kindareseenmanytimesin the literature,sometimeswithminoradjustmentspresented
as empirical correlations (for instance Hills12).There is nodoubt about thevalidity of
this equation in case pressure differences can be neglected but there are two
complicatingphenomena.Whenequation(4)isapplied, the valueofthetermvs(1 - a )
isassumedtobe constantwithineachflowregimeand hastobe estimated.Furthermore
radialconcentrationand velocityprofilesofthe dispersedphasecan be present.Uniform
profiles are prerequisites for equations (1)-(4). Nicklinis is one of the investigators,
havingproposedmodelsbasedonequation (4)withestimatorsforthetermvs(1 - a )
and correction factors for concentration profiles and non-uniform flow. Zuber and
Findlay26generalizedthistheoryand defineddriftvelocitiesand volumetricfluxdensities
to make it valid for any two-phase flow system and for any flow regime. Wallis24
denominated models basedonthistheory asdrift-flux models.The drift-fluxtheory is
giveninAppendix B.Thefinalequationisessentiallythesameasequation (4):

vsd _.^ ^ ^, <audy> (5)

- C{<vsd+vsc>)
<a> "- <a>

wherevscjandv scarethesuperficialvelocitiesintheriserofdispersedandcontinuous
phase,respectively, <> meanstheaverageoverthecross-sectionalareaand Vdj isthe
drift velocity of the dispersed phase (seeAppendix B).The lastterm in equation (5),
< avdj > / < a >,is calledthe weighted mean drift velocity.Thisterm isassumed to
account for the velocity difference between thetwo phases dueto buoyancy, andis
assumedtobeconstantwithineachflowregime.
Thedistribution parameter Ctakes into accountthe effect ofthe combination of
radial concentration and velocity distributions of the dispersed phase. For fully
established,i.e.,constant,velocityand concentrationprofiles, the valueofC is assumed
tobeaconstant andwillonly changewithaflow regime26.UsuallyCishigherthan 1,
44 CHAPTER3

duetohighvelocitiesandconcentrations inthecentreofthetube.AlthoughZuberand
Findlay26 indicatewiththeoretical calculationsthat 1.5 isvery high, amaximumvalue
forCcannot begiven.
Theweighted meandriftvelocityisassumedtobeindependent of concentration
andtherefore equaltotheterminalvelocity ofthebubbleinaninfinite medium:vb„ , 26
Correlationsfortheweighted meandriftvelocity<avdj > / < a >aregivenbyZuber
andFindlay26. For churn-turbulentflow:

(6)
< avd, > IcrgAp
' - " = 1.53<
<a> "~ I pc2

whereaisthesurfacetension,gisthegravitationalconstant,Apisthedensitydifference
betweenbothphasesandpcisthedensityofthecontinuous phase.Foran air/water
systemthisresultsinavalueof0.25m.s-1.
Manyappliedmodelsfortwo-phaseflowarebasedondefinitionsofdriftvelocities
and flux densities. Infigure 1the typical drift-flux plot for literature datafrom several
sources is given. In such plots linear correlations between the datum points can be
distinguished,wheretheslopeofthelinerepresentsCandtheinterceptrepresentsthe
weighted-meandriftvelocity.Inmostcasesvaluesofabout1 and0.25,respectively,for
bothparameters cangiveacceptable results. Infigure 2holdup predictions aregiven
whentheseparametervaluesareusedandthemeasuredliquidvelocity isusedasan
inputparameter. Holdupdataagreewithin25%ofmodelprediction,yetinsomecases
oftheexaminedliteraturedatainfigure1 andfigure2significantdeviationscanbeseen.
This can be corrected by choosing other values for both model parameters and by
including possible dependence on flow rate and concentration. Correlations for the
modelparameterstoincorporatethisdependence,andotheradjustmentsofequation
(5) are proposed by some authors.5Theresulting equations become more empirical
andwillnotbediscussedanyfurtherhere.
 CHAPTER3 45

v
sd
[m.s- 1 ]

v
sd+ v
sc [m.s-1]
Figure 1: Typical driftflux plotforliteraturedata:GarciaCalvo9:O; GarciaCalvo9:• ; Verlaanet al. 23 :
A; Verlaanet al. 23 : *; MerchukandStein1®: • ; Jones13:• ; Hatch™: A; Frohlichet al. 7 :-.
Equation (5)with C=1 and <a VM>/«X> =0.25: .

0.15
a ["]
predicted

0.05

a [-]
0.05 0.1 measured
Figure2: Holduppredictionwithdrift-flux equationsforC=1and <a vyj>/<<*> =0.25 :GarciaCalvo9:
O; GarciaCalvo9:• ; Verlaanetal. 23 : A; Verlaanetal. 23 : *; Merchukand Stein18:• ; Jones13:
• ; Hatch1<>: A- Frohlichet al. 7 :-.
46 CHAPTER3

The continuity equations in terms of the drift-flux quantities as described by

equation(5)canserveasoneofthetwoindependentequationsthatarenecessaryfor
prediction of both holdup and liquid velocity. Either holdup or liquid velocity can be
calculated when the other value is known. However, for the second equation other
physical principles or empiricalcorrelations shouldbeused,whichshouldnot involve
the slip velocity vs. This is caused by the similarity of the assumption of constant
< audy > / < a >inequation(5)andtheassumptionofconstantvs(1 - a ) inequation
4.

MOMENTUM CONSERVATION

Momentum is mass multiplied by velocity or force multiplied by time. In steady

statethismeansthatthegoverningequationformomentumconservation isthesame
as for a force balance. The principle of momentum conservation is often used in
hydrodynamicmodelsforALRs,butlittleisknownaboutfrictionandpressureforcesin
thedeviceandbetweenthetwophases.Thisleadssometimestotheopinionthatfriction
coefficients areadjustable parameters.14

Gasphase
Asimpleexampleofhowamomentumbalancecancontributetoahydrodynamic
modelforALRsisgivenhere.Itconcernsaforcebalanceforasinglerisingbubble.In
steadystatethebuoyancyforceequalsthefrictionforce:

1 3 1 2 2 (7)
-nD3(pc-Pd)g = -CwnD2pcv2b.
6 Z

wherepdisthedensity ofthedispersed phaseandC w isthefriction coefficient. C wis

dependent on bubble risevelocityvba and bubble diameter D.Therefore correlations
forvbx havebeendeterminedexperimentally24. Ingeneralavalueofabout0.25m.s-1
isfound forvb„ in air/water systems (equilibrium bubble diameter 6 mm).Whenthis
valuefor the rise velocity of asingle bubble inan infinite medium is used for the slip
velocity as described inthe equations (1)-(3) this can be useful for a hydrodynamic
model. However,thiscannotbecombinedwiththecontinuity equation (4) becauseit
 CHAPTER3 47

isinfactanapplicationofequation (4)withanassumptionforvs. Infigure2, equation

(4)isused,wherevs(1 - a )isassumed0.25m.s-1 (derivedfromequation(6)).Infigure
3 equation (4) is used,wherevsisassumedto be0.25 m.s-1 (derivedfrom equation
(7)). Inbothcasesthemeasuredcontinuous-phasevelocityisusedasinputparameter
becauseonlyapartofacompletehydrodynamicmodelisexaminedhere.Itisobvious
that both equations are essentially very similar and therefore give nearly the same
results.

0.15
CX [-]
predicted

0.05

a [-]
measured
Figure3: Holduppredictionbasedonequation(7)forvs=v^=0.25: GarciaCalvo 9 :0; GarciaCalvo9:
• ; Verlaanetal. 23 : A; Verlaanetal. 23 : *; Merchukand Stein' 3 : • ; Jones 1 3 :*; Hatch1°:*•;
Frohlich etal. 7 : -.

Liquidphase
Another possibilityfor usingthemomentum conservation lawisaforce balance
overtheliquidphase.Thedrivingforce forliquidcirculation isassumedtobeequalto
thehydrostatic pressuredifference betweenriseranddowncomer:

1 (8)
(ar-ad)Cpc-pd)gH0 I/.oPc"! £ * X< '2>
K
/pyScd
48 CHAPTER3

wherea r isthe holdup inthe riser,a d isthe holdup inthedowncomer, /denotesthe

successivereactorsections,k denotesthesuccessiveflowrestrictions,K\istheoverall
friction coefficient for the reactor, based on the superficial liquid velocity in the
downcomervsccj,fand e varefrictionfactorsforliquidflow,Listhelengthand ftn isthe
hydraulicradius.Adistinctionmustbemadebetweenexternalandinternalloopreactors.
The downcomer holdup (aJ in external loop reactors is usually small, and in the
examined literature cases neglected. Insmall-scale internal-loop reactors<xd is often
foundtobeupto90%ofarand mustbetakenintoaccount,whichmakesthe calculated
friction coefficients much more sensitive to measurement errors. The investigated
literaturedataonexternalloopreactorsgivealwaysagoodlinearrelationship between
ar andv S cd 2 - lntableItheaverageoverallfrictioncoefficientsandthestandarddeviations
derivedfromthedataaregiven. Thesmallstandarddeviations,especiallyforexternal
loopreactors,indicateagoodsuitabilityforliquid-circulationmodels. Thisisconfirmed
by using the K\ values of table I and the measured continuous-phase velocities to
describetheholdupofexternalloopreactors.Incaseofinternalloopreactorsthe same
methodgivestheholdupdifferenceinriseranddowncomer. Figures4and5showthis
suitabilityofequation (8)todescribearor( ar - ad ).

Table I: Average overallfriction coefficients derivedfromliteraturedata.

derivedfrom equation derivedfrom derived from

(8) equations (12) equations(14)
and(13) and(16)

Reference average Std average std average std

Kf in% Kf in% Kf in%

Garcia Calvo9 33.3 5 67.1 9 25.8 18

Garcia Calvo9 19.3 5 35.4 13 16.3 12
Verlaanet al. 2 3 2.05 11 4.91 26 2.9 9
Verlaanet al. 2 3 4.32 9 11.4 40 6.1 41
Merchukand Stein18 11.2 3 12.6 5 9.3 4
Jones 13 2.84 11 41.1 10 22.7 7
Hatch10 16.8 19 342.5 40 206.7 26
Frohlich et al. 7 2.35 38 4.9 5 4.91 5
 CHAPTER3 49

0.2

0.15-
A •

a [-]
predicted
0.1 - V
A, *

0.05-

1 1 1 1 ' 1 a [-]
0 0.05 0.1 0.15 measured
Figure4: Predicted holdup compared to measured holdup for literature data. Prediction based on
equation (8): Garcia Calvo9: O; Garcia Calvo9: • ; Verlaan et al. 23 : A; Veriaan et al. 23 : *;
Merchuk andStein18: • .

/
//
/ //
0.06- /
_ /
a r - a d [-] //
predicted /
/// •
0.04- / A_
//

,'A
/
/
0.02- /
/A
/ A
//
•

f
0
L , . ,
I ' I • ar-ad H
0.02 0.04 0.06 measured
Figure5: Predictedholdupdifferencebetweenriseranddowncomercomparedtomeasureddifference.
Prediction basedon equation (8):Jones13:• ; Hatch1°: * ; Frohlichetal. 7 : -.
50 CHAPTER3

Anotemustbemadeaboutthemeasuredlinearitybetweenar andvScd2> because

theremaybeasnakeinthegrass.Itisnotsurprisingthatthislinearityisfoundbecause
themethodofmeasuringholdupsthatisused,automaticallyimpliesit.Theholdupsare
directly derivedfrom pressure-difference measurements betweentwo locations inthe
riser.Alinear relationship betweenthispressure dropinthewholesystem,exceptfor
theriserpartbetweenthetwolocations,andthesquareoftheliquidvelocityiswhatcan
be expected according to Bernoulli's law. This method is only valid for holdup
determinationwhenthepressuredropduetoflowintheriserisnegligible(seeAppendix
C).

ENERGYCONSERVATION :MECHANICAL-ENERGY BALANCE

ThemajorityofhydrodynamicmodelsforALRspresentedrecentlyintheliterature
usetheprinciple ofenergyconservation.Theequation,thatisusedmost often,isthe
mechanical-energybalanceforanisothermalsysteminsteadystate.Mechanical-energy
balancesdescribethevariouswaysinwhichdifferentforms ofmechanicalenergyare
interconverted or irreversibly converted to thermal energy. A simplified form of this
equationfor one-phase isothermalflowisgivenbyBirdet. al. 1 :

ri 1 , , + w E ^
-dP + g(/i„-/i() + -Oo-u?) — + — = 0
Jp 2 p <i> p<f>

whereWistherateatwhichthesystemperformsmechanicalworkonitssurroundings,
Eisthesumoftheconversionofenergytoheat,alsodenotedasthepowerdissipation
ofthe system and* isthe volumetric flow rate ofthefluidum.The subscripts /ando
denotetheentranceandexitof* toandfromthesystemconsidered.
Roelsand Heijnen20 havefound bythermodynamic reasoningthat entropy flow
fromaspargedcolumntypereactor,duetoentropytransfertothedispersedidealgas
phasefromthesystemASd, canbedescribedby:
 CHAPTER 3 51

RT. (PA (10)

TASd - P ^ d - l n ^ -

where7isthetemperature,misthemolarmassofthegasphase,Ristheuniversalgas
constantandP\andP 0aretheentranceandexitpressureofthegasphase,respectively.
Whenequation (9)isappliedonlyforthegasphaseandkineticandpotentialenergyof
the gas-phase are negligible,aswellasthe irreversible energy conversion inthegas
phase,anequationsimilarto(10) isobtained:

whereIV^istheworkofthegasphasedoneonthesurroundings,whichisinthiscase
the continuous liquidphaseonly. Whenthe liquidphaseof anALR isexamined, -W^
definedinequation (11),canbeusedforthetermIVinequation (9).Furthermore,the
termsinequation (9)concerningthechangeinpressure,potential,andkineticenergy
arezero,whenthereactorwallsareconsideredtobethesystemboundaryandthestart
andtheendoftheloopreactor isatthesameposition.Whenitisfurther assumedthat
thecontinuous phasecannotdoworkonthevesselwalls,equation (9)thenresultsin:

whereEcistheirreversibleconversionofmechanicaltothermalenergyinthecontinuous
liquidphase.Thisequationshowsthattheentropygainedbythegasphaseiscompletely
produced by irreversible conversionof mechanicalenergy inthe liquidphase. Forthe
descriptionofEc, thefollowingcorrelationisused, whichisbasedonone-phaseflow:

fv i 2 f n v i 2\ i (13)
52 CHAPTER3

Theenergyconversiondescribedinequation(13)isascribedtoenergyconversiondue
toflowin astraighttubeandenergyconversiondue toflowinobstaclessuchasbends,
valvesanddiameter changes.Thiscanalsobewrittenasanoverallfriction coefficient
Kf.The frictioncoefficientsfande vcan beestimated,butespeciallythe latterisoflimited
accuracy. It is difficult to use equation (12) and (13), becausefriction coefficients for
two-phase flow are different from that of one-phase flow. The presence of a second
phase causes different flow conditions, which lead to different values for the friction
coefficients.Precisedescriptionsofthe frictioncoefficientsand the influenceofasecond
liquidphasearestilllacking.
Studiesarefocussed onadjustingthefrictioncoefficientsfande v insuchaway
thattheydescribethetotalfrictionlossesintwo-phaseflow.Wallis2 4statesthatforlow
void fractions friction factors in one-phase flow are a reasonable approximation for
thoseintwo-phaseflow.Thereisalsothepossibilityofcalculatingatheoreticalviscosity
ofthe mixture.Thiscan be usedforthe calculation ofthe Reynolds number, whichis
correlated with friction coefficients. Another approach isto use correction factors to
correlate friction losses in one-phase flow to two-phase flow. This is proposed by
Martinelli and others16.17 and many studies follow this approach. Literature data are
analysedaccordingtothisapproach,andtheresultingfrictioncoefficients,derivedfrom
(12) and (13), are given in table I. The friction coefficients derived from the
mechanical-energy balance (column 3) are significantly higher than the friction
coefficientsderivedfromthemomentumbalance(column1).Thereasonforthisisthat
themomentumbalancefortwo-phaseflowisthesameasforone-phaseflow,butthe
energy balance isnot.Therefore another strategy isnecessary, which issplittingthe
irreversible conversion of mechanical energy intwo parts. In addition to the energy
conversionascalculatedforone-phaseflow,itisassumedthatthereisanextraamount
ofenergy/Afconvertedduetothepresenceofthedispersedphase:

RT (PA 1 2 (14)
p ln = +A
'*'-^ UJ P^z*'" '
TheresultsoftableImakeitquiteacceptablethatthesplittingofenergyconversionin
two parts makes sense.An example ofthis strategy isfound inthe work ofJones13,
whousesfortheadditionalenergyconversionthefollowingassumption:
 CHAPTER 3 53

RT, fPt\ 05)

Thisdivisionofenergyinputfortwopurposesseemsarbitraryanditisveryunlikelythat
itisuseful.Therefore,itisnosurprisethatthemodelbasedonthisequationdidnotgive
satisfyingpredictionaccuracy.
Leeetal. 15 ,andGarciaCalvo8>9alsoassignenergyforthetwodifferentpurposes
accordingtoequation (14)andthefollowingassumption:

Af = avspcgHAr (16)

GarciaCalvo8^9used afriction coefficient derivedfrom measurements, but insteadof

the equations (14) and (16), the momentum balance for the liquid phase was used
(equation(8)).Leeetal 15usedarbitrarychosenfrictioncoefficients,whichisnotuseful.
Toexaminethevalidityandusefulnessoftheassumptioninequation(16),friction
coefficientsKfarederivedfromliteraturedataofexternalloopreactorsandpredictions
ofholduparemadewiththeaveragefrictioncoefficients. IntableIcolumn5thefriction
coefficientsaregivenandinfigure6thepredictionofholdupwiththeaverageKfvalues
is shown. It is obvious that compared to the momentum balance, there is no
improvement.
Chistietal.3intendedtofollowthestrategyofsplittingtheirreversibleconversion
of mechanical energy in two parts, but derived equation (8) from equation (9) (see
AppendixC).
Fromthis section it can be concludedthat acorrelation betweenv^,a andvsc
basedonthemechanicalenergybalanceoftheliquidphaseexists,whichcanbeuseful
incirculationmodels.However,comparedtothemomentumbalanceoftheliquidphase
noimprovement isfound.
54 CHAPTER3

0.15
a [-]
predicted

0.05

measured

Figure6: Predictedholdupcomparedtomeasuredholdupforliteraturedata.Predictionwithequations
(14)and (16)andaverage Kfvaluesderivedfromeachdataset itself (table I).Garcia Calvo9:
O; GarciaCalvo9:• ; Verlaanetal.23: A; Verlaanetal.23:*; MerchukandStein13: • ; Jones13:
• ; Hatch1o: A- Frohlichetal. 7 : -.

EMPIRICAL EQUATIONS

Empiricalcorrelationsareeasytouseandcanbequitevaluable,butsometimes
only valid for exactly the same conditions as used to obtain the original data. No
information about basic physical phenomena is given and under different conditions
empiricalcorrelationscanbequitemisleading.Noindicationscanbederivedformodel
improvement.Becauseofthegeneralapplicability,themostgeneralformsofempirical
equationsaregiven.ThebestknownempiricalmodeltodescribeholdupinALRsorin
BCsis:

(17)
a = av.
 CHAPTER3 55

whereaisthedispersed-phaseholdup,vscjisthesuperficialdispersed-phase velocity
andaandb arecoefficients. HeijnenandVan 't Riet11 derivedfrom literature datafor
BCsthatholdupcanbedescribedingeneralwithequation(17)whena=0.6and b=0.7.
Vander Lans22claimsforALRsthatthevalueofthecoefficientb usually liesbetween
0.4and1.2.IntableIIanoverview isgivenofexperimentaldataforALRspublishedby
several authors. The values for the coefficients a and b areobtained by regression
analysis.Inallcasesagoodlinearrelationshipisfoundbetweenthelogarithmofaand
thelogarithm ofVsd> whichjustifiesthe useof equation (17). Infigure7theexamined
dataisgiveninaplotwithlogarithmicscales.Regressionanalysisofallthedatapoints
infigure 7learns that for a=0.8 andb=0.8 the best general description isobtained.
Thesevaluesareusedfor holduppredictionwiththemeasuredliquidvelocity asinput
parameter. Infigure8the results ofthese calculations aregiven.Only below 0.1,the
holdup predictionisacceptable.

Table II: Coefficients for equations (17)and (18)derivedfromliteraturedata.

a = c^L U*c = cvdsd

Reference a b c d
9
GarciaCalvo 0.86 0.78 0.97 0.37
GarciaCalvo9 0.60 0.72 1.35 0.40
Verlaanet al. 23 0.47 0.71 4.61 0.40
Verlaan et al. 23 1.08 0.96 6.17 0.44
MerchukandStein18 0.54 0.78 1.87 0.36
Jones13 0.83 0.82 1.25 0.49
Hatch1 ° 0.50 0.45 0.85 0.56
Weiland25 0.16 0.80 1.15 0.41
Frohlichetal. 7 0.24 0.37 3.24 0.30
56 CHAPTER3

u.o
^**
0.2-
o ( r [-]
0.1-
0.05 :

0.02- §7^
0.01- s ^ *

0.005- *

0.002- I ' I I I 111 - I

v 1
0.002 0.01 0.05 0.1 0.2 sd[m.s- ]
Figure7: Logarithmic plotofthemeasuredriserholdupversusthesuperficialdispersed-phasevelocity.
Garcia Calvo9: O; Garcia Calvo9:o; Verlaan et al. 23 : A; Verlaan et al. 23 : *; Merchukand
Stein18: • ; Jones13:• ; Hatch1°: * ; Frohlichet al. 7 :-.
Equation (17)fora=0.8and b=0.8: .

\J.H A«
//
//
• /
/
' •
0.15- //
A
• /
/
a [-] // •
/
predicted //
0.1 - •* //
//
A O

o
o
0.05- Sffo
^0O

•*r*
0 / I I ' I 1 '

0( [-]
0.05 0.1 0.15
measured
Figure8: Predicted holdup compared to measured holdup for literature data. Prediction based on
equation(17)anda=b=0.8.GarciaCalvo9:O; GarciaCalvo9:• ; Verlaanetal. 23 : A; Verlaan
etal.2 3 :*; Merchukand Stein13: • ; Jones13:• ; Hatch1o: A ; Frohlichet a l . 7 : - .
 CHAPTER3 57

ChistiandMoo-Young4useaforcebalanceandKolmogoroff'stheoryofisotropic
turbulencetoshowthatthistypeofequationisnotfullyempiricalbuthassomephysical
background. However, two other empirical parameters appear to be necessary to
describetherelationship betweenthedragcoefficientofabubbleandtheRe number.
Thereforethephysicalfoundationisstillquestionable.
Similartotheholdupinequation(17)theliquidvelocity inALRscanbedescribed:

= CV„ (18)

wherec anddarecoefficients.Thesameliteraturereferencesasforequation (17)are

usedtovalidateequation (18).Againinallcasesagoodlinear relationship isfoundin
the double logarithmic plot of vsc versusv^,which meansthat equation (18) isalso
suitable(figure9).Theregressionresultsindicatethatthevalueofdisalwaysbetween
0.30and0.56. Nogeneralvalueforc canbegiven (tableII).

v
sc[ms-1]

0.002 0.01 0.05 0.1 0.2 'sd [m.s-1]

Figure9: Logarithmic plotofthemeasured risersuperficialvelocityofthecontinuousphaseversusthe

superficialdispersed-phasevelocity. GarciaCalvo 9 :0; GarciaCalvo9:• ; Verlaanetal. 23 :A;
Verlaanetal. 23 : *; MerchukandStein1®: • ; Jones13:• ; Hatch1°: A ; Frohlichetal. 7 : -.
58 CHAPTER3

Becausethesimilarity ofcoefficientsbanddfor different experimentalset upsis

usuallybetterthanforthecoefficientsaandc,thelatterareassumedtobedependent
onreactor geometry.
A completely different category of empirical hydrodynamic equations is
distinguished byVander Lans22.Theequations haveincommonthattheyare based
onslipvelocityvs.Thegeneralformoftheequationinthiscategorycanbesplitupina
definitionforvs, which is given in equation (3),and anempirical equationfor the slip
velocity:

vs - vb„ + f(a,vsc,D) (19)

whereDisthebubblediameter.VanderLans22givesalonglistofempiricalcorrelations
forvs.Thecorrelationthatgivesthe bestdescription oftheinvestigated literaturedata
isremarkbly simple:

vs = vb„ + 0 . 0 6 5 (20)

Theseslip-velocity basedmodels givearelationship betweena,v sc andv^. Thiscan

beusedasoneofthetwoequationsthatarenecessaryfor hydrodynamic description
of ALR's, but can not be combined with other equations based on slip velocity. Itis
strikingthatthebestmodelbasedontheslipvelocityvsistheonethatisalreadyderived
frommomentumconservation principle (equation8).

CONCLUSION

Simple hydrodynamic modelling of two-phase loop reactors corresponds with

choosingandimplementingtwoindependent equationstosolvea andvscasfunction

Empirical equations describing an exponential correlation of holdup or liquid

velocitywiththedispersed-phaseflowrateshownoagreementwithequations based
onconservation lawsanddeviationsbetweenmeasuredholdupandpredictedholdup
sharply increase above a holdup of 0.1. The most accurate equation based on slip
 CHAPTER 3 59

velocity appearedtobethesameastheequationbasedonmomentumconservation.
Thepredictionofholdupwiththisequationstillgivesdeviationsuptoabout25%ofthe
measuredvalue, butisbetterthantheempiricalequation.
Equationsbasedoncontinuityconsiderationsgiveremarkablysimilarresultswith
equationsbasedonconstant slipvelocity.Closerexaminationoftherelevantequation
(20) learnsthat this isnot surprising. Besides astraightforward mass balance inthe
drift-fluxtheory itisalsoassumedthatvs(l - a) isconstant,whichisnearlythesame
asassumingthatvs isconstant.
Ingeneral hydrodynamic models with simple analytical equations arebased on
twoequations.Thefirstisanassumptionconcerningslipvelocity,writtenintheformof
acontinuity, momentum or empiricalequation.Thesecond equation isamomentum
balanceoramechanical-energy balance.Frictioncoefficientscanoftenonlyberoughly
estimated,whichlimitstheaccuracyofthefinalhydrodynamic model

NOMENCLATURE

A cross section m2
Af frictionduetothepresenceofthedispersedphase J.s-1
a coefficient inequation (17) (n.r.)
b coefficient inequation (17) (n.r.)
C distributioncoefficient indrift-fluxtheory
c coefficient inequation (18) (n.r.)
d coefficientinequation (18) (n.r.)
D bubblediameter m
E energyconversionduetofriction J.s-1
Ec energy conversionduetocontinuous-phaseflow J.s-1
Et energy input J.s-1

G,. frictioncoefficientfor obstacles

60 CHAPTER3

/ frictioncoefficientfortubes
g gravitationalconstant m.s-2
dispersionheight m
h height m
i index
J volumetricfluxdensity m.s-1
overallfriction coefficient
k index
I length m
m molarweightofthedispersedphase kg.mol-1
P pressure N.m-2
PH hydrostatic pressure N.m-2
R universalgasconstant J.moH.K-1
Rh hydraulic radius m
v velocity m.s-1
vsd superficialdispersed-phasevelocity m.s-

VSc superficialcontinuous-phasevelocity m.s-

V scd superficialcontinuous-phasevelocity indowncomer m.s-

Vc velocityofcontinuous phasecorrectedfor holdup m.s-

Vd velocity ofdispersedphasecorrectedfor holdup m.s-

'ci
driftvelocity ofcontinuous phase m.s-

V dj driftvelocity ofdispersedphase m.s-

slipvelocity m.s-

vb. terminalrisevelocity ofasinglebubbleinaninfinitemedium m.s-

relativevelocity m.s-
 CHAPTER3 61

w Energy exerted by a system J.s-1

a holdup inatube or inthe riser of anALR -

ar holdup riser -

« d holdup downcomer -

<t> flow rate m 3 .s- 1

ASd entropy transfer to the dispersed phase J.s-1.K-1

a surface tension N.m-1

P density kg.m- 3

Ap density difference of dispersed and continuous phase kg.m- 3

<> average (n.r.)

weighted average (n.r.)

subscri Pts:

c continuousphase
d dispersedphase
i entranceofthesystem
o exitofthesystem

REFERENCES

[1] Bird R.B., StewartWE, Lightfoot E.N. (1960).Transport Phenomena.JohnWiley&Sons.

[2] Cherry R.S., Papoutsakis E.T. (1986). Hydrodynamic effects oncells inagitated tissue culture
reactors. Bioprocess Engineering, 1, 29-41.
[3] Chisti M.Y. Halard B.,Moo-Young M. (1988). Liquid circulation inAirlift reactors. Chem. Eng.
Sci.,43,451-457.
[4] Chisti M.Y.,Moo-Young M. (1988). GasHoldup in Pneumatic reactors. Chem. Eng. J., 38,
149-152.
[5] ClarkN.N.,FlemmerR.L (1985).Predictingtheholdupintwo-phasebubbleupflowanddownflow
usingtheZuberand Findlaydrift-flux model.AlChEJ.,31, 500-503.
[6] ClarkN.N.,JonesA.G.(1987).Onthepredictionofliquidcirculationinadraft-tubebubblecolumn.
Chem. Eng.Sci.,42,378.
62 CHAPTER3

[7] FrohlichS., Lotz M., KorteT., Lubbert A.,Schugerl K., Seekamp M. (1991).Characterization of
apilotplantAirlifttowerloopbioreactor. I:Evaluationofthephasepropertieswithmodelmedia.
Biotechnol.Bioeng.,38,43-55.
[8] Garcia Calvo E. (1989). A fluid dynamic model for airlift loop reactors. Chem. Eng. Sci.,44,
321-323.
[9] GarciaCalvo E.,LetonP. (1991).Afluiddynamic modelfor bubble columnsandairlift reactors.
Chem.Eng.Sci.,46,2947-2951.
[10] HatchR.T.(1973).Experimentalandtheoreticalstudiesofoxygentransferinanair-liftfermenter.
Ph.D.thesis,M.I.T.,Cambridge.
[11] Heijnen J.J.,Van 't Riet K. (1980). Masstransfer, mixing and heat transfer phenomena in low
viscosity bubblecolumn reactors.Chem.Eng.J.,28,21-42.
[12] Hills J. (1976). The operation of a bubble column at high throughputs. I: Gas holdup
measurements. Chem.Eng.J.,12,89-99
[13] JonesA.G.(1985).Liquidcirculationinadraft-tubebubblecolumn.Chem.Eng. Sci.,40,449-462.
[14] Kawase Y., Moo-Young M. (1990). Liquid circulation in external-loop airlift bioreactors.
Biotechnol. Bioeng.,35,540-546.
[15] LeeC.H.,GlasgowL.A., EricksonL.E.,PatelS.A.(1987).Liquidvelocityinairliftfermentors.50-59.
in:C.S. Ho,J.Y. Oldshue, Biotechnology processes.
[16] Lockhart R.W., Martinelli R.C. (1949). Correlation of data for isothermal two-phase,
two-component flow inpipes.Chem.Eng.Progr.,45,39-48.
[17] Martinelli R.C, Nelson D.B. (1948).Trans.ASME,70,695.
[18] MerchukJ.C.,SteinY. (1981).Localhold-Upandliquidvelocity inair-lift reactors.AlChEJ., 27,
377-387.
[19] NicklinD.J. (1962).Two-phase bubbleflow. Chem.Eng.Sci.,17, 693-702.
[20] Roels J.A., Heijnen J.J. (1980). Power dissipation and heat production in bubble columns:
approach basedonnonequilibriumthermodynamics. Biotechnol.Bioeng.,22,2399-2404.
[21] Tramper J., Wolters, I., Verlaan P. (1987). The liquid-impelled loop reactor: a new type of
density-difference-mixed bioreactor. 311-316. in: C. Laane, J. Tramper, M.D. Lilly (eds.),
Biocatalysis inorganic media,Elsevier Science Publishers.
[22] Van der Lans R.G.J.M. (1985). Hydrodynamics of a bubble column loop reactor. Ph.D.thesis,
Technical University Delft.
[23] Verlaan P., Tramper J.,Van 't Riet K., Luyben K. Ch.A. M. (1986). A hydrodynamic model for
airlift-loop bioreactorwithexternalloop.Chem.Eng.J.,33,43-53.
[24] WallisG.B. (1969).One-dimensional two-phaseflow. McGraw-Hill.
[25] WeilandP.(1978).UntersuchungeinesAirliftreaktors mitausseremUmlaufimHinblickaufseine
Anwendungals Biorektor. DissertationChemietechnik Universitat Dortmund.
[26] Zuber N., Findlay J.A. (1965).Averagevolumetric concentration intwo-phase flow systems.J.
HeatTransfer, 87,453-468.
 CHAPTER 3 63

APPENDIXA: NON-ISOBARICMODELING

Thegasvelocity canbeexpressedasfollows:

vsa(0)P(0) (A.1)
vsd(z~) =
P(z)

wherevsd( 0 )andP( 0 ) arethesuperficialgasvelocityandthepressure,respectively,

atthebottomofthereactor.ThelocalpressureP ( z ) i n theALRisrepresentedby:

( *r \ (A-2)
F ( z ) = P(O) - Pcg z - J a(z)dz

To overcome that a ( z ) hasto be known to calculate P(z)and vice versa, as afirst

approximationa( z ) isapproximatedby:

a ( z ) = a(O)

wherePftisthe hydrostatic pressure, usedasanapproximation for the real pressure

anddefined as:

Ph(z) = F ( 0 ) - pcgz (A.4)

Substitutionofequation (A.4)and (A.3)intoequation (A.2)gives:

p
„<-^ (A.5)
P(z) = />„(*) - />(0)a(0) f _ L _ d ( P h ( z ) )

Integrationofequation (A.5)yields:
64 CHAPTER3

( 9cQz\ (A.6)
P{z) = P ( 0 ) - pcgz - P(0)a(0)ln^l-^-1

whichcanbeusedincombinationwithequation(A.1).
(Verlaanetal.23).

APPENDIX B: DRIFT-FLUX EQUATIONS

Volumetricfluxdensitiesaredefinedas:

je = ( l - a ) y c (B.1)
J, = <™d (B.2)

wherevisvelocityandcandd refertocontinuousanddispersedphase,respectively.
Driftvelocitiesaredefinedas:

veJ = vc - j (B.3)
"*, = v* - J (B-4)

relativevelocity:

vr = vd - Uc (B.5)

volumetricfluxdensityofmixture:

+
J = h U (B-6)

ZuberandFindlay26showwithcontinuityequationsthat/'isnotafunctionofplacebut
onlyafunctionoftimefor onedirectionalflow. Furthermore,iftherelativevelocityvris
zero,bothdriftvelocities,vCjandv^,arezeroandthetwophaseshavethesamevelocity,
whichisequaltothevolumetricfluxdensityofthemixture(/').Theproblemthatisalways
relevant for calculations, isthat it always concerns average values. To calculate the
continuous-phase velocitytherearetwopossibilities:
 CHAPTER 3 65

Ju (B.7)

<vda> <Jd> <aj> <OLvdi> (B.8)

vd = = = +
<a> <a> <a> <a>
Theaveragevelocity<vd >isnotthesameastheweightedmeanvelocityvd.Thelatter
ispreferredas< yd >isaninputparameterthatisreadilyavailableaccordingto:

Qd (B.9)
<;d> = <aud> = —

from8thefinaldrift-fluxequationfollows:

< aJ / > <aud/> <aud/> (B.10)

vd = —<j> + '— = C<j> + —
<a > < ] > <a> <a>

Ingeneral,insteadof<j d >,< jc >,and< j >thenotationsassuperficialvelocitiesare

used:Vsd, vsc> andVsd+vSC) respectively.

APPENDIXC: ELIMINATIONOFENERGY INPUTFROM

MECHANICAL-ENERGY BALANCE

Themechanical-energy balanceoveranALRcanbederivedfromequation (14):

Et - Ec+Af (C.1)

Amechanical-energy balanceovertheriser canbegiventhatdescribes energy input

£j,potentialenergygainofthecontinuousphase,pressure-energy lossofthemixture,
friction-energy lossdueto relative motionofthetwophasesAfandcontinuous phase
wallfrictionintheriser E cr j Ser :

Ei = pcgHdArvsc - (apd + (l-a)pc)gHdArvsc + A, + ECtHser (C.2)

66 CHAPTER3

After substitution of equation (C.2) inequation (C.1) this results in :

a(Pc-Pd)gHdArvsc = Ec - Ec,riser (C.3)

^criser is negligibly small compared to £ c .

 67

CHAPTER 4

MIXING IN LIQUID-IMPELLED LOOPREACTORS

SUMMARY

The liquid-impelled loop reactor isanewcolumn-type bioreactor. Thedesignof

this device is based on the principle of the air-lift loop reactor. In the external-loop
configuration used in this work, descending perfluorochemical drops bring about
circulation of the continuous aqueous phase. Mixing of this continuous phase is
characterized persectionofthe reactor.Axial-dispersion coefficients forthetubewith
two-phaseflow aredeterminedandcorrelatedwiththeenergydissipation inthetube.
Comparisonswithsimilarsystemssuchasbubblecolumnsandair-liftloopreactorsare
made.Overallmixingparametersarederivedandusedforcalculationofthenumberof
circulations neededto achieve acertain degree of mixing.The hydrodynamic model
frompreviouswork istestedforthereactorconfigurationsofthiswork. Itcanbeuseful
tocalculatecirculationtimes.

INTRODUCTION

Column-type reactorsareattractivefor biotechnological purposes becausethey

areof simple construction and usually easyto operate.To improve liquidcirculation,

Thischapter hasbeenpublished inBiotechnology &Bioengineering,vol.39,707-716(1992)

bytheauthors: H.M.VanSonsbeek, S.P.VanDerTuinandJ.Tramper.
68 CHAPTER4

recycleloopsareintroduced.Inthisway,theair-lift loopreactor (ALR) isderivedfrom

the bubble column (BC).Thesame principle isappliedfor extraction columns, which
recentlyhasresultedintheintroductionoftheliquid-impelledloopreactor (LLR).16The
LLRcombinesthe properties of loop reactorswiththe useof organic solvents,which
makesitextremelyattractiveformanybiotechnologicalprocesses.Applicationforplant
cell cultures is an example that has been tested.4^ Mixing under controlled-flow
conditions and low shear forces are combined with high solubilities of components
poorly soluble inwater, increased reaction rates,andfacilitated product recovery by
extraction. For biotechnological applications of the LLR, it is most likely that the
continuous phase isanaqueousphase,whichcontainsthe (immobilized) biocatalyst.
Therefore,themixinginthecontinuous phaseofanLLRisdiscussed.
Mixinginreactorsdeterminesperformanceandoperationrequirementsandthus,
isofmainimportancefor reactor engineers. Notwithstandingtheenormouseffortthat
istakeninthisfieldofresearch,mixingandturbulenceinfluidsarestillpoorlyunderstood
andoftendescribed byempiricalcorrelations.Thecomplexityofthismatterhasmade
turbulence to a popular topic in chaos theory. Kolmogoroff's theory of isotropic
turbulenceisthebest-knowntheorytodescribeturbulenceandfindsgeneralfavor.To
get abetter understanding of mixing andturbulence, computational fluid dynamics is
often used to solve the complete Navier-Stokes equations. However, accurate
descriptionoffluidflowisstillproblematic,becauseofthemanyunrevealedinfluences
andtheir significance. Besidestheory, experimentaltechniques are being developed
rapidly.Fluorometrytechniquesareimprovedbyusingfiberopticsensors. Measuring
theextintinctionofaheatpulsehasgainedgreatpopularity.Laserdoppleranemometry
isarelativelynewtechniquethatgivesnewimpulsestofluid-dynamics research,butis
stillnotaccurate enoughfor multi-phase systems. Inspite ofallthese developments,
the basic principles inthefieldofreactor engineering havenotyet beenrevolutionary
changed. Relatively simple approaches of mixing are still the most favorable ones,
becauseoftheir practicaluse.
Incaseofcolumn-typereactors,theaxial-dispersionmodelisthemostwidelyused
with mixing described as plug flow with diffusion-like disturbances. This mixing
mechanismisbasedontheassumptionthatfluidelementsbehavecompletelyrandom,
independant oftheir history. Inthiscasethe standard deviation ofthe residence time
distribution (RTD)function isproportionaltothe square root ofthevessel length. Itis
not correct to describe all pipe flow by diffusion laws. Levenspiel and Fitzgerald13
observedageneralmisuseofthedispersionmodel,whereaconvectivemodelshould
beused.Inthecaseofaconvective-mixingmechanism,fluidelementshavetheirown
 CHAPTER 4 69

characteristic behavior patternandbehave independently ofotherfluidelements.The

standarddeviationoftheRTDfunctionisproportionaltothevessellength.Anexample
isthe dispersed phase in a non-coalescing two-phase system.The different sizesof
dropletscausedifferentmeanvelocities.Accordingtothisdescriptiononlythedispersed
phaseinLLRsbehavesinaconvectiveway.
Lubbertand Larson14distinguish inadditiontothesetwo mixing mechanisms a
thirdmechanism,whichisbasedonKolmogoroff'stheory ofisotropicturbulence,with
a standard deviation ofthe RTDcurve proportional tothevessel lengthtothe power
1.5.Forone-phasepipeflowatRe= 60x103,LubbertandLarson14havefoundindeed
suchacorrelationandconcludedthatthemixingmechanism basedon Kolmogoroff's
theoryofisotropicturbulenceisvalid.Thismeansthattheaxial-dispersionmodelisnot
valid for one-phase flow under the conditions of Re = 60x103. This theory of athird
mixing mechanism, based on isotropicturbulence, seemsto be incontradiction with
the work of Baird and Rice,2who usethe same theory to describe a correlation for
axial-dispersioncoefficients intwo-phaseflows,whichleadstoastandarddeviationof
the RTD function proportional to the square of the vessel length. Obviously, the
interpretations of Kolmogoroff's theory of isotropicturbulence arenotyet inharmony
andmaystillleadtosomeconfusion.
Forthecontinuous-phasemixinginthemaintubeofanALR,LubbertandLarson14
havenotfound convincing evidenceforthedispersionortheconvection mechanism,
andhaveconcludedthatbothplayarole.Theconvectivemechanismpartisassumed
tobecausedbytheliquidentrainedinthebubblewakes.AufderheideandVogelpohl1
haveinvestigatedextractioncolumnsandhaveconcludedthatthedispersionmodelis
validforthecontinuous phasebutnotforthedispersed phase.Theconvective model
hastobeusedforthedispersedphase.However,theynotethatinmostcases,itdoes
not makemuchdifferencewhether ornotthedispersionmodelisused.
Itistheobjectiveofthischaptertocharacterizethemixingofthecontinuousphase
inLLRsandcompareittoothersystems,suchasBCsandALRs.Basedonthetheoretical
considerations earlier, the axial-dispersion model is chosen for characterization of
continuous-phase mixing in the LLR. It will give information that is valuable for
engineeringpurposes,andmuchdatatocompare isavailable.Forthispurpose, loop
reactors can be distinguished in sections with different mixing conditions, e.g., main
tube,circulationtube,topandbottomsection.Thereistwo-phaseflowinthemaintube
andone-phaseflowintherecirculationloop,thelatterhavingasmallerdiameter.Inthe
topandbottomsection,theflowdirectionisreversedbytwo90degreebendsconnected
byahorizontaltube.Furthermore,thesesectionscontainachangeindiameter,andin
70 CHAPTER 4

oneofthetwosectionsthereisalsophaseseparation.Becauseofthese differentmixing
conditions, characterization of each section separately can be more accurate andof
morepracticalusefordesignpurposesthandetermininganoverallmixing parameter.
However,tocalculatemixingtimes,itisnecessarytoderiveanoverallmixingparameter,
basedoncontributions ofeachsection.

THEORY

Axial-Dispersion Model
Themixingmodelthatwillbeusedistheaxial-dispersion model,whichdescribes
flowthroughtubesasplugflowwithdisturbancesfollowing Fick'sdiffusionlaws:

dc _ J_f^£ _ dc 0)
ae ~ BOTJZ* Jsc

where c , 9 and x are dimensionless quantities for concentration, time, and axial
coordinate,respectively,and Bo istheBodenstein number,definedas:

vL (2)
Bo =
Dax

wherev istheaveragevelocity inthetube, L isthelengthofthetubeand, Dax isthe

axial-dispersion coefficient.
Theexit-agedistributioninaloopreactor accordingtothedispersionmodelwith
open-vesselboundaryconditionscanbecalculated3:

C ( e) = exp : •(i + (3)

~*' (H) J0 ( x-Q)zBo'

40

where/isthecirculation number.
For high mixing performances, the diffusion-like mechanism becomes rather
dubious,andinthatcasethemore empiricaltanks inseries modelmay bepreferred,
becauseitdoesnotsuggestanyphysicalbackground.Bothmodelsdescribemixingin
 CHAPTER 4 71

the reactor or in the reactor section with one parameter. The exit-age distribution
functionsofbothmodelsareneverthesame,butforincreasingparametervalues,they
can become more alike.Correlations betweenthe two parameters are based onthe
exit-agedistributionfunctionsandareofthefollowingform:

Bo = 2n + p (4)

wheren isthe number oftanks inseries.Thevalue ofp depends onthe criteria and
boundary conditions for the axial-dispersion model. If one compares the surfaces
underneaththeexit-agedistributioncurvesofbothmodels,withopen-vesselboundary
conditionsfortheaxialdispersion model,avalueof 1 forp givesthe bestcorrelation.
Comparingthevariancesoftheexit-agedistributionfunctionsforopen-vesselboundary
conditions, minus 1isfoundfor p.12

OverallBovalue
ThederivationofoneBovalueforthewholereactorfromthevaluesoftheindividual
reactor sections is based on the additivity of the variances of atracer signal12.The
changeinvarianceforawholecirculationintheloopreactorcanthusbedescribedby:

Aa2UK = A < +A(^ +A(rf+AcTb2 (5)

wherem,c,f,andbrepresentmaintube,circulationtube,andtopandbottomsection
oftheLLR,respectively.Thegeneralformulaforthechangeinvarianceofasignalthat
passesthereactor orareactor sectionis12:

(6)
^1 = JL + JL
+
rt " Bo, Bo*

where/ denotesthe reactor orthereactor sectionandtisthecorresponding average

residencetime. Bycombiningequations (5)and(6)theoverallBovalueforthereactor
can bederivedfromthe contributions of each section.This overall mixing parameter
canbeusedforcomparisonwithotherreactortypesandforcalculationofmixingtimes.
72 CHAPTER4

Axial-dispersion coefficient
Several relationships for axial-dispersion coefficients in BCs are proposedin
previous research articles.2.6.11 Some are derived from experimental results,6 while
others are based on the circulation flows in bubble columns11 or on energy input.2
Becauseoftheconsiderablesuperficialcontinuous-phasevelocity,itisratherdubious
whether thecirculationflow patterns of BCsexist aswellinloop reactors.The mixing
mechanisminloopreactors maybedifferentfromthat inBCs,anditcanbewrongto
use just any existing correlation that is valid for BCs. Therefore, themost
general-applicableoptionischosen todescribemixinginthecontinuous phaseofthe
LLR. This is given by Baird and Rice,2 who derived by dimensional analysis from
isotropic-turbulencetheoryfortheaxial-dispersioncoefficientincolumncontactors:

i i (7)
Dnr = Kl3Pl

whereKisadimensionlesscoefficient, I thelengthscaleofthelargesteddiesandP m
theenergy-dissipationrateperunitmass.Byassumingtheenergyinputtobecompletely
dissipatedinliquidmotionand I takentobethecolumndiameterD,aKvalueofabout
0.35wasfoundforBCs(L/D>5) intheabsenceofslugging.FieldsandSlater7derived
acorrelationfortheenergydissipatedinliquidmotionforALRs,neglectingtheenergy
dissipated in wakes behind the bubbles, the wall friction, and the potential-energy
changeofthethedispersedphase.Similartothiscorrelation,amoregeneralapproach
is given below for two-phase flow in vertical tubes, taking into account the potential
energy ofthedispersed phase.Thehydrostatic pressuregradient insuchasystemis
givenby:

^PhydrostaUc = 9 H d{( 1 " <*) Pc + «P d} (8)

where AP isapressuregradient,pcandpdarethedensitiesofthecontinuous andthe

dispersed phase,gthegravitationalconstant,H^ istheaeratedliquidheight,andais
the volumetric dispersed-phase concentration in the tube. By flowing through this
pressuregradient, bothphasesgainorloosepressureenergy:

A£" pressure = AgHd{{\ - a ) p c + a p d } O s d +u s c ) (9)

 CHAPTER4 73

whereA isthe cross section ofthetube,andvscjandv s c arethe superficial velocities of

the dispersed and the continuous phase. Furthermore, both phases gain or loose
potential energy:

&E p^ntiai = AgHd(pdvsd + pcvsJ (10)

Assuming that the power input is totally dissipated in liquid motion of the continuous
phase, a general correlation is found from equations (7), (9) and (10), describing the
axial-dispersion coefficient intubes withtwo-phase flow:

Dax = KD3 \vsdg vscg (l-a)pr (11)

Mixing time
Besides the mixing parameter (Bo or A?), dimensionless mixing time (number of
circulations) or mixing time to achieve a certain degree of mixing can also be used to
describe mixing.A correlation between mixingtime, mixing parameter, and number of
circulations is of great practical use. To find this, a definition of well-mixedness is
necessary. Forthat, the inhomogeneity h is introduced:

h = \c - 1| (12)

Blenke 3 derived a correlation between Bo, the inhomogeneity, and the number of
circulations. Forthis purpose, upper-envelope curves are usedthat are determined by
the extremes of a response signal of a Dirac pulse. An example of of such an
upper-envelope curves is drawn in figure 1. Blenke's relationship is based on the
exponentialdecayoftheupperenvelopecurve.Murakamietal. 1 5 givethesameformula
in more general terms and derivefor the dimensionless mixing time:

Bo , (2\ (13)
74 CHAPTER4

TheupperenvelopecurvesareapplicableforallBovalues,butitshouldbekeptinmind
thatfor lowBovaluesthe mixing mechanism may not bedispersion orthe boundary
conditionofaDiracpulsemaynotbevalid. Infigure 1 itcanbeseenthatthisequation
describestheupperenvelopecurveforaBovaluesof30veryaccurately,butonlyafter
acertaindegreeofmixingisachieved.Ifwell-mixednessisdefinedbyaninhomogeneity
h lessthan0.05thenequation (13) givesthefollowing linear relationship between x m
andBo:

> 0.09Bo (14)

This gives a direct relationship between the overall mixing parameter and the
dimensionless mixingtime (number ofcirculations).

C [-]

e [-]
Figure 1: Upperenvelopecurvesdescribing inhomogeneityforareactorwithanoverallSo-valueof30
after a tracer Dirac pulse (calculated tracer response signal: ; upper envelope curve:
 CHAPTER 4 75

Circulationtime
If one uses the hydrodynamic model for loop reactors described in previous
work,16thecirculationtimecanbeestimated,andthemixingtimeintimeunitscanbe
derived.ThefrictioncoefficientK\ofthereactor,whichdefinesthereactorgeometry,is
derived from experimental measurements of the holdup and the circulation velocity
accordingto:

1 2 ( 15 )
ApagHd = -Kfpcvsc

whereApisthe density difference betweenthetwo liquid phases. For predictions, a

second equation is necessary to solve bothvariables, a andvsc. Based on drift-flux
theory,thefollowingequationisusedforthispurpose16:

vsd (16)
1.6(vsd + vsc) + vda

wherev , ^ istheterminalrisevelocityofadropinaninfinitemedium,andthesuperficial
dispersed-phasevelocity (vsd)isaninput parameter.

MATERIALSANDMETHODS

Inthiswork,a4.5-liter lab-scaleLLRanda35-liter pilot-plantscaleLLR,bothwith

external loopandofsimilar designareused.Infigure2aschematic drawing isgiven.
ThedimensionsofbothdevicesaregivenintableI.AKCIsolution,madefromtapwater,
which has a density of about 1000 kg.m-3, is used as continuous phase and the
perfluorochemicalFC40(3M),whichhasadensityof1870kg.m-3,isusedasdispersed
phase.Thedispersed phase is pumped inthe reactor viaaliquidsparger, locatedat
thetop ofthemaintube.Thissparger consists ofapressure chamber andpipesthat
areevenlydistributedoverthecrosssectionofthemaintube.
76 CHAPTER4

Table I: LLRdimensions.

lab scale pilot plant

3
maintube: volume (m ) 0.0012 0.0143
length (m) 0.44 1.82
diameter (m) 0.06 0.1
3
circulation loop: volume (m ) 0.0002 0.0023
length(m) 0.45 1.82
diameter (m) 0.025 0.04

bottom section: volume continuous phase(m3) 0.0022 0.0132

volumedispersed phase (m3) 0.0003 0.0030
diameter* (m) 0.06 0.1

top section: volume (m3) 0.00084 0.0052

diameter* (m) 0.025 0.04

distance maintubeandcirculation loop (m) 0.45 1.05

liquid sparger: number of pipes 12 24

innerdiameter of pipes(m) 0.0014 0.0020

* concerningthe major partofthis section

Mixinginthecontinuousphaseisinvestigatedwithtracerresponsemeasurements.
Threesectionsofthereactor aredistinguished:themaintube,thebottomsection,and
the circulation tube withtop section,and are investigated separately. The location of
two tracer detection probes mark begin and end of a section. Because of the low
residence time in the circulation tube and the top section, these two sections are
measuredasonesection.
Inthe lab-scale reactor a500mMKCIsolution isusedasthe continuous phase,and
about 1.5 mlof 2 MHCIis injected astracer. Forthe pilot-plant reactor,thesevalues
were50mM KCIsolutionand12mlof2MHCI.TwopHprobesareusedfordetection.
Adding KOHsolutionafterameasurement makes itpossibletorepeattheexperiment
several times without replacing the reactor contents. The tracer is injected at some
distancefromthefirstprobe.Thetracerresponsesatbothprobelocationsaremeasured
and stored as function of time. To calculate dimensionless concentrations from the
measured pHvalues,aconversion curve is used,which isdetermined by measuring
thepHchangesinasmallquantityoftheKCIsolutionduringadditionofsmallamounts
ofaHCIsolution.ThecalculationprocedureofVerlaanetal. 17 isused,whichisbased
 CHAPTER4 77

ontheaxial-dispersionmodel.Inthisproceduretheoutletsignalispredictedforagiven
(measured) inlet signal.The bestfittingBovaluefor thepertinent reactor section can
befoundinthisway.

r////// ////// S-7-7-T

VALVE

Figure2: Aschematic drawingoftheLLRwithdownflow ofthedispersed phase.

Dropsizesaredeterminedbytakingphotographsofthemaintubeandmeasuring
thesizeofallthedropsthat areinanarbitrary area,containingabout 100-200drops.
Circulation times are derived from tracer-response measurements and from
measurement by means of an inductive flowmeter in the circulation loop. Holdup is
78 CHAPTER 4

derived from pressure drop in the main tube, which is measured with a differential
pressuremeter.AgraphicalcorrelationgivenbyHuandKintner1°isusedtoderivethe
terminalvelocity of an FC40drop inwater asfunction ofthe drop diameter, which is
giveninfigure3.

0.15

'dco I m s 1

0.05

I I I I 1-

0.005 0.01
d [m]

Figure3: Terminal fallvelocity of FC40drops inwater asfunction of drop diameter, derived from Hu
andKintnerJO

RESULTSAND DISCUSSION

Dropsizesaremeasuredbymeansofphotographsandgiveninfigure4.Itshows
that the average drop size in the pilot-plant LLR slightly decreases with increasing
dispersed-phaseflow rate.Inpreviouswork16,thetheory ofchangein drop-formation
regime that causes this effect is explained. The decreasing drop size can limit the
maximum-possibledispersed-phaseflowrate,becauseoftheformationoftinydroplets
that show fluid-dynamic behavior similar to the continuous phase. Inthat case, the
danger ofagradualemulsificationispresent,whichunderminestheworking principle
of the LLR. However, in the experimental set-up described inthis chapter, no such
 CHAPTER4 79

limitationsarise.Thelimitationthatdoesdeterminethemaximumdispersed-phaseflow
rateisthecoalescencerateinthebottomsectionofthereactor.Thetimenecessaryfor
completesegregationofthetwo phasesnotonlyincreasesbecauseoftheincreasing
dispersed-phase flow rate,but alsobecause ofthe decreasing drop size. Forthe4.5
liter LLR a maximum dispersed-phase superficial velocity of 0.01 m/s is possible,
whereas0.015m/s isthemaximumforthe35-liter LLR.Visualobservationshowsthat
slugformation and coalescence inthe maintube can be excluded. Furthermore, no
significantdropentrainmentinthecirculationtubeisobserved.

0.005

0.004
d,
J 9 [m]
32

v sd [m.s-1]

Figure4: Dropdiameterasfunction ofthesuperficial velocity ofthedispersed phase inthemaintube

ofthe LLR (lab-scale LLR:O; pilot-plant LLR:p).

Axialdispersionandtanksinseries
Simple mass-balance equations at short time intervalsfor a number oftanks in
seriescanbeappliedaswellasthemethodtocalculateBovalues,developedbyVerlaan
etal. 17 Both calculation procedures giveequivalent results accordingto equation (4)
withavalueof minus2forp. Thisisfound bygenerating simulated responsesignals
basedontheaxial-dispersion model,andsubsequent examinationofthese simulated
signals with the calculation procedure according to the tanks in series model. The
80 CHAPTER4

disadvantageofthetanks inseriesmodelisthatforthecalculations,averyshorttime
interval is required and,thus, moretime to compute.Therefore,this method has not
beenusedfurther.

Mixingpersection
Anexampleofameasuredoutletsignalandthebest calculatedresponsesignal,
accordingtothe methodofVerlaan et al., 17 isgiven infigure 5. Bothsignals showa
goodresemblance,andinthiswaytheBovalue,whichgivesthebestmixingdescription,
isobtained.Toevaluateandcomparethemixingperformanceofthethreesectionsof
thereactor,theBovaluesdeterminedinthiswayareusedincombinationwithequation
(6).ThedimensionlesschangeofvarianceoftheRTDfunction,causedbyflowthrough
thepertinentreactorsection,iscalculatedfromtheseexperimentalresults.Thechange
ofvariance ismadedimensionless bydivisionbythesquareofthecirculationtime.

C [-]

e H
Figure5: Exampleofameasuredandfittedtracerresponsesignalinthepilot-plantLLRfor v^=0.0125
m.s-1 (measuredoutlet signal: ; calculated responsecurvefor Bo=19: ).
 CHAPTER4 81

ThemeasuredBovaluesforthelab-scaleLLRare 8,12,and8forthemaintube,
bottom section and circulation tube plus top section, respectively. Significant
dependenceonthedispersed-phaseflow rateisdifficultto discriminate. Deviationsof
3areobservedintheduplicateexperiments.Thecorrespondingdimensionlesschanges
ofvariance are :0.029, 0.051, and0.022.Thisindicates aninfluence onmixingofthe
sameorderofmagnitudeforallthreesectionsofthewholereactor.Thebottomsection
hasthegreatest impact,mainlybecauseofthehighestresidencetime.
Theresults ofthe pilot-plant LLRareplotted infigure6and 7.TheBovaluesof
bottom section and the circulation tube with top section are not influenced by the
dispersed-phaseflow rate,whereforthe maintube asignificant increase isobserved
for increasing dispersed-phase flow rate. All Bo values are in the same order of
magnitudefortheexperimentalconditions.

30-

Bo[-]

20 • o p-
,&'' o
• • .'~v o

10-
:'o ^..A t TS- A ----A--

0 0.005 0.01 0.015

vsd [m.s- 1 ]

Figure6: ExperimentallydeterminedBo valuesforthe35-literLLR(circulationloopandtopsection:

bottom section:A; maintube:O; overall: • ) .
82 CHAPTER 4

In a reproducibility experiment for the main tube of the pilot-plant LLR 10

measurements are carried out for a dispersed-phase flow rate of 7.5x10'3 m.s-1.An
averageSo-valueof 13.4isfoundwithastandarddeviationof 2.5.The lattergivesan
impression of the limited accuracy of the measured values. The actual weight of the
threesectionsofthepilot-plant LLRontheoverallmixingeffect isshowninfigure7by
thedimensionlesschangeofvariance.Thebottomsectionandthecirculationtubeplus
topsectionhaveapproximatelythesameweightonmixing.Theinfluenceofflowthrough
the maintube onmixing is highat lowflow rates, but decreasesfor higher flow rates
andbecomesevenlesserthanmixingintheothersections. Forfurtherscale-upofthe
LLR,themixingeffectinthemaintubewillbecomedominantoverthemixingintheother
sectionsduetoahigher residencetimeoftheliquidinthissection.

0.1- T
~>-T
change '*-».„

of
T
variance [-] o

0.05- ^^
^•- o o

A n A 0 0--.<>. a i A
•

5 o '--<?,•

0 • i • i • i
() 0.005 0.01 0.015
v
sd [m.s" 1 ]

Figure7: DimensionlesschangeofvarianceofRTDfunctionbyflowthroughthereactorsectionsofthe
35-liter LLR (circulation loopandtop section:• ; bottom section:A; maintube:O; overall:
 CHAPTER4 83

Axial-dispersion coefficients
Infigure8theaxial-dispersion coefficients ofthemaintubesofbothreactorsare
given. Equation (11) is applied for description ofthe measured values oftheaxial
dispersioncoefficient.Kvaluesof0.35and0.56givethebestresultsforthemaintubes
ofthe lab-scale LLRandthepilot plant, respectively.TheKvalueforthemaintubeof
the lab-scale LLR corresponds very wellwith thevalue that isusually foundfor BCs
(0.35).2.9Forthemaintubeofthepilot-plant LLRahighervalueisfound.Calculations
forliteraturedatalearnsthathighervaluesfortheriserofALRsarefoundaswell.From
dataofVerlaanetal., 18Weiland,19andGlennonetal.,8Kvaluesof0.49,0.51,and0.51
are derived, respectively. The scatter intheexperimental data isusually high, which
must be kept inmindwhen one isdrawing conclusions basedon these differences.
Fortheexperimentalresultsofthiswork,thelimitedsignificanceofthemeasuredvalues
isdemonstratedwiththe reproducibility experiment. FromthemeasuredBovaluesof
thereproducibilityexperiment,axial-dispersioncoefficientsarederivedwithanaverage
of 9.4X10- 3 m2.s-1andastandarddeviationof2.1x10"3m2.s-1.

0.02

Dax [ m - s " ' 3

0.01 •A A A
A--' A
A

o-Z-rrW
—i 1 1-1 r—

0 0.005 0.01 0.015

v
sd [m.s" 1 ]

Figure8: Dax valuesforthernaintubeofthe4.5-literLLR (experimental:O; equation(11)withK=0.35:

) andthe35-liter LLR (experimental: A; equation (11)with K=0.56: ).

84 CHAPTER 4

From the experimental results of LLRs and literature data on ALRs it seems
reasonableto useequation (11),whichisderivedfromthetheory of BairdandRice2,
for order of magnitude estimations of dispersion coefficients. Avalueof about 0.5 for
thedimensionlessconstantKisrecommendedforthemaintubeofloopreactors.This
isillustratedinfigure9.

0.05- ,'
o
0.04- /o °
v /
(measured)
0.03 - vo /

0.02- • ' '

o *'
0.01 -

0 1
' 1 • — 1 ' 1 ' 1 ' 1
0 0.01 0.02 0.03 0.04 0.05

Dax fms"2l
(predicted)

Figure9: ExperimentalD^valuescomparedtovaluescalculatedwithequation(11)forK=0.5(lab-scale
LLR:• ; pilot-plant LLR:O; ALRWeiland19: A; ALR Glennonetal. 8 : v; ALRVerlaanet al. 18 :

Mixingtime
FromtheBovaluesofthe individualsections,overallBovaluesarederivedwith
equation (6).The dimensionless mixing times are calculated from these values with
equation(14).
Inthis way an overall Bo value of 21 is derived for the lab-scale reactor. The
calculateddimensionless mixingtime is2.This isinagreementwiththe experimental
results,wheretracer detectionispossible uptoabout2.5circulations.Withcirculation
timesvaryingfrom20to50seconds,thismeansmixingtimesofabout 50secondsto
2minuts.
 CHAPTER4 85

OverallBo values for the pilot-plant reactor are given infigure 6 and circulation
timesbetween2and3arederivedfromthesenumbers.Experimentalcirculationtimes
arefoundaround3.5.Theactual-observedmixingtimeisvaryingfrom0.5to4minutes.
IngeneralthecalculatedoverallSovaluesseemtobealittleunderestimated,but
oneshouldkeepinmindthelimitedaccuracy intheselowBoranges.

Hydrodynamics
Averyaccuratelinearrelationshipbetweentheholdupandthesquareoftheliquid
velocityismeasured.Thisisingoodagreementwithequation (15),andoverallfriction
coefficients (basedonvscr)forthereactorsarederivedof2.0and4.2,forlab-scaleand
pilot-plant reactor respectively. In figures 10 and 11 the experimental results of
hydrodynamicmeasurementsandmodelcalculationsaregiven. Inmodelcalculations
a value of 0.17 is used for Vdoo- Infigure 3 this value is found by matching with the
measured drop size c/32 from figure 4. The model describes the hydrodynamics
reasonably well. For lower dispersed-phase flow rates both,the holdupandthe liquid
velocity,arealittleoverestimatedbythemodel.

O.b- -0.06
v
scr [m.s" ] 1 a [-]
-0.05
0.4-

* / A
-0.04
0.3- / A
/A
-0.03
/A /
0.2- / /o
* /o - -0.02
/A / /°
°
0.1 - /o -0.01
o
0
/° 0
C) 0.005 0.01

v sd [m.s-1]

Figure 10: Superficialcontinuous-phasevelocity(A) inthecirculationtube.anddispersed-phaseholdup

(O) inthe maintube ofthelab-scale LLRand model calculations ( ).
86 CHAPTER 4

1-0.06
V
scr [ m . S " 1 ] a [-]
0.05

0.04

0.03

— -0.02

0.01

0 0.005 0.01
v
sd [m-s-1]

Figure 11: Superficialcontinuous-phasevelocity(A)inthecirculation tube,anddispersed-phaseholdup

(O)inthemaintubeofthepilot-plant LLRandmodelcalculations ( ).

CONCLUSIONS

Ingeneral,calculationswiththetanksinseriesmodelgiveequivalent resultswith
themethodofVerlaanetal. 17 , inwhichBovaluesaredetermined.
OverallBovaluesofabout 20-30arefoundfor bothreactors, labscale and pilot
plant,andingeneral3circulationsareobservedbeforeaninhomogeneitylessthan5%
isreached,whichisingoodagreementwiththecorrelationbasedonenvelopecurves.
Theaxial-dispersion coefficient inthemaintubeoftheLLRwithtwo-liquid-phase
flow can be estimated with equation (11),which isbased onthetheory of Bairdand
Rice2. TheKvaluesthatgivethebestdescriptionfortheexperimentalresultsare0.35
and0.56forthelab-scale reactor andthepilot-plant reactor, respectively. Dataonthe
riserofALRsreflectasimilar behavior. Ingeneralequation (11)withaKvalueofabout
0.5 issuitablefororder ofmagnitude estimationsofaxial-dispersioncoefficients inthe
maintubeofloopreactors.
 CHAPTER 4 87

Thehydrodynamic modelbasedondrift-fluxtheory isacceptableto applytothe

LLRconfigurationsusedinthiswork.Atlowflowrangesmodelpredictionsoverestimate
holdupandcontinuousphasevelocity.Whenthedispersed-phaseflowrateisincreased
thismaybecomeanunderestimationduetoadecreaseofdropsize.

ACKNOWLEDGEMENT

Thisworkwas partly supported byApplikon Dependable Instruments BV.Schiedam,

The Netherlands.The authors wishtothank W.A. Beverloofor his comments onthe
manuscript.

NOMENCLATURE

A cross section rr.2

Bo Bodenstein number -
c dimensionless concentration -

Dax axial dispersion coefficient m.s-2

D diameter ofthe column m

d32 Sauter mean diameter of droplets m

i circulation number -

J index referingto a reactor section or to whole reactor -

g gravitational constant m.s-2

Hd dispersion height m

h inhomogeneity -
K coefficient -

Kf friction coefficient of the reactor -

L length ofthe column m

I length scale of the largest eddies m

n number of tanks in series .

88 CHAPTER4

Pm energy-dissipation rate J.s-1.kg-1

p coefficient
tj averageresidencetimeinreactor orreactorsectioni s
Re Reynolds number
v averagevelocity inatube m.s-1
vsd superficialvelocityofdispersedphaseinthemaintube m.s-1
vsc superficialvelocityofcontinuous phaseinthemaintube m.s-1
vscr superficialvelocityofcontinuous phaseinthecirculationloop m.s-1
vdm terminalvelocity ofadrop inaninfinitemedium m.s-1
x dimensionless axialcoordinate
a holdupdispersedphase
A£" change inpressureenergy J.s-1
AP pressuregradient N.rrr2
Ap absolutevalueofdensitydifference betweencontinuous and kg.m-3
dispersed phase
pc density continuous phase kg.m-3

pd density dispersedphase kg.m-3

a standarddeviation (ofaRTDcurve)
xm dimensionless mixingtime

9 dimensionlesstime

REFERENCES

[1] AufderheideE.,VogelpohlA.(1986).Aconvectivemodeltointerpretdispersed-phaseresidence
time measurements inpulsedliquid/liquidextractors. Chem. Eng.Sci.,14,1747-1757.
[2] Baird M.H.I., Rice R.G. (1975). Axial dispersion inlarge unbaffled columns. Chem. Eng.J.,9,
171-174.
[3] BlenkeH. (1979).Loop reactors.Adv.Biochem. Eng.,13,121-214.
[4] Buitelaar R.M., Langenhoff A.A.M., Heidstra R., Tramper J. (1991). Growth and thiophene
production by hairy roots cultures of Tagetes patula in various two-liquid-phase bioreactors.
Enzyme Microb.Technol., 13, 487-494.
 CHAPTER 4 89

[5] Buitelaar R.M.,Susaeta I.,Tramper J. (1990).Application oftheliquid-impelled loop reactor for

theproduction ofanthraquinones by plant cellcultures.694-699 In: H.J.J. Nijkamp, L.H.V.van
derPlasandJ.vanAartrijk(eds.),Progressinplantcellularandmolecularbiology,KluwerAcad.
[6] Deckwer W.-D., Burckhart R., Zoll G. (1974). Mixingand masstransfer intall bubble columns.
Chem. Eng.Sci.,29,2177-2188.
[7] Fields P.R., Slater N.K.H., (1983).Tracer dispersion ina laboratory air-lift reactor. Chem.Eng.
Sci.,38,647-653.
[8] Glennon B., Macloughlin P.F., Malone D.M. (1989). Mixing studies in an air-lift loop reactor.
Chimicaoggi, 1,25-28.
[9] Heijnen J.J.,Van't Riet K. (1984). Mass transfer, mixing and heat transfer phenomena in low
viscosity bubblecolumn reactors.Chem.Eng.J.,28, 21-42.
[10] HuS.,Kintner R.C.(1955).Thefallofsingleliquiddropsthroughwater.AlChEJ., 1,42-85.
[11] Joshi J.B., Sharma M.M. (1979).A circulation model for bubble columns. Trans. IChemE,57,
244-251.
[12] LevenspielO. (1972).Chemical reaction engineering.2ndedition.JohnWiley&Sons.
[13] Levenspiel O., Fitzgerald T.J.(1983).Awarning onthe misuseofthedispersion model.Chem.
Eng.Sci.,38,489-491
[14] Lubbert A., Larson B. (1990).Detailed investigations ofthe multiphaseflow inairlifttower loop
reactors.Chem.Eng.Sci.,45,3047-3053.
[15] MurakamiY.,HiroseT.,OnoS.,NishijimaT. (1982).Mixing propertiesinloopreactor. J.Chem.
Eng.Jpn, 15121-125.
[16] VanSonsbeekH.M.,VerdurmenR.E.M.,VerlaanP.,TramperJ. (1990).Hydrodynamic modelfor
liquid-impelled loopreactors.Biotechnol.Bioeng.,36,940-946.
[17] Verlaan P., Van EijsA.M.M.,Tramper J.,Van't Riet K., Luyben K.Ch.A.M. (1989). Estimation of
axialdispersion inindividual sectionsofanairlift-loop reactor. Chem.Eng.Sci.,44,1139-1146.
[18] Verlaan P.,Tramper J., Van 't Riet K., Luyben K.Ch.A.M. (1986).A hydrodynamic model for an
airlift-loop bioreactorwith external loop.Chem.Eng.J.,33,43-53.
[19] WeilandP.(1987).UntersuchungeinesAirliftreaktorsmitausseremUmlaufimHinblickaufseine
Anwendungals Bioreaktor. Ph.D.thesis, University Dortmund,Germany.
 91

CHAPTER5

STEADY-STATE METHOD FOR KA

MEASUREMENTS IN MODEL SYSTEMS

SUMMARY

Amethodtodetermineoverallvolumetricoxygen-transfercoefficientsisdiscussed.
Itisbasedonmeasuringsteady-statevalues,whichisconsideredtobemostaccurate.
Althoughthemethodisveryconvenient, itisseldommentionedaspossibletechnique.
Adefinedflowiswithdrawnfromthereactor,de-aeratedbystrippingwithnitrogen,and
recirculated.Fromtheflowandsteady-statevaluesoftheoxygentensioninthereactor
andtherecirculationflow,thekavaluecanbecalculated.Althoughgenerallyapplicable,
thismethodisespeciallysuitableformodelsystemswithoutrealoxygenconsumption.
Inthischapter itisillustratedforkadeterminationinaliquid-impelledloop reactor.

INTRODUCTION

Two types of methods can be distinguished for determination of the overall

volumetricoxygen-transfer coefficient (ka):thedynamicandthesteady-statemethod.

Thischapter hasbeenpublished inBiotechnologyTechniques,vol.5,157-162 (1991) bythe

authors :H.M.VanSonsbeek, S.J.GielenandJ.Tramper.
92 CHAPTER5

Inthedynamicmethodtherateofchangeofoxygenconcentrationismeasuredaftera
step-wise change in the inlet oxygen concentration. In the steady-state method,
concentration measurements areonlynecessaryatonemoment.
Thedynamicmethodismostpopularbecauseitiseasytocarryoutandaccurate
resultscanbeobtainedveryrapidly.However,thedynamicmethodisverycomplicated
insomecases.Whentheresponsetimeofanelectrodeisnotmuchsmallerthan 1/ka,
probedynamicshavetobetakenintoaccounttoavoidunderestimationof/fa.8.5Another
restrictionisdescribedbyHeijnenetal. 2 ,whoshowthatthedynamickadetermination
forgas/liquidsystemsmaybeproblematicif,besideslargebubbles,verytinybubbles
existinthesystem.Theequilibriumofoxygentensioninthesebubblesandthesolution
willbereachedsofastthatsuchaliquid/small-bubbledispersioncanbeconsideredas
a homogeneous liquid phase. As a result the kavalues determined by the dynamic
method will be underestimated. Both restrictions mentioned become of increasing
importanceforincreasingviscosityofthegas/liquidsystem.Inadditiontothis,insome
cases it can be rather difficult to accomplish a proper step change in the inlet
concentration.
Thelatter problem isespecially relevantforkadetermination ina liquid-impelled
loop reactor (LLR). An LLR is a newtype of bioreactor, that works according to the
principle of anair-lift loop reactor (ALR),with asecond liquid phase instead of agas
phaseandisschematically drawninfigure 1. 7 Besidespracticallimitations,analogous
tosmallbubbles,smalldropletsinthetwo-liquid-phasesystemofanLLRcanalsobring
about disturbance of the dynamic measurement results. Therefore, especially for
systems likethe LLR, itwould bevery convenient to have a better and more reliable
methodforkadetermination.
Manyinaccuraciesandproblemswiththedynamic methodcanbeovercomeby
steady-statemeasurementoftheka,whichisconsideredasthemostreliablemethod.3
Probeandmediumpropertieshave noinfluenceandverysmalldropletscannotdisturb
the measurement.8
Inthischapterasteady-statemethodispresentedthatisrarelyusedormentioned,
though it is a very suitable way to determine overall volumetric oxygen-transfer
coefficients. The method is based on de-oxygenation of reactor medium outside the
bioreactor. Moser6mentionedtheexistenceofthisde-oxygenationtechnique.Manoet
 CHAPTER5 93

al.4usedanexternalde-oxygenationvesselanddeterminedkvaluesfromsteady-state
measurements. Alvarez-Cuenca and Nerenberg1 determined oxygen concentration
profiles inabubblecolumnbymeansofanexternalvacuumde-aerator.
Inthe steady-state method described inthis chapter asmallflow of continuous
phase is pumpedfrom an LLRviaabuffer vessel back intothe reactor. Inthis buffer
vesselde-oxygenationbystrippingwithnitrogentakesplace.

de-oxygenation
vessel

pump

continuous phase
-dispersed phase

valve

Figure 1: The experimental set upfor dynamic and steady-state measurements inthe liquid-impelled
loop reactor.
Dynamic method :vessel 1 and2,and probeP3.
Steady-state method :vessel 1 and3,and probe P1, P2andP3.
94 CHAPTER5

THEORY

Oxygentransferinatwo-phasesystemusuallyconcernstransferfromadispersed
phasetoacontinuous phaseandisdescribedas:

dcc(t) fcd \ 0)
(1 + em) ' = ka\ — - c c (Q - OUR
at Vm J

whereeistheholdupoftinybubblesortinydropsofdispersedphase(d<<1mm),that
formahomogeneousdispersionwiththecontinuousphase,andthatdonotcontribute
totheoxygensupply.2Oxygenissuppliedbylargeinjectedbubblesordropswithoxygen
concentrationc^.

Dynamickadetermination
If the continuous phase is assumed to be ideally mixed and the oxygen
concentrationinthedispersedphaseisassumedtobeconstantandforOURtobezero
throughout theexperiment,kacanbederiveddirectlyfromthedynamicmeasurements
withthefollowingformula:

ka = (1 + em)ln —— fj 2 - t{)
\cdm - cc(t2)J

Ifmoreadvanced modelsfor mixing inthe continuous anddispersed phaseare

necessary,anumericalsolutionforka canbefound (curvefitting).

Steady-state ka determination
Ifsteady state isreached (dc/dt=0), kacanbederivedfrom equation (1) anda
mass balance to calculate the OUR. For the experimental set-up of figure 1the net
transport ofoxygeninthede-oxygenation circuit isequaltothetransfer inthereactor
andtothetransfer inthede-oxygenationvessel:

(1 + a m ) * ( c i „ - c ou( ) = l / r ( f c a ) J -~- - c c r j = I/„(/ca)„c c ,„

 CHAPTER5 95

This steady-state method for ka measurements is a good alternative forthe

dynamicmethod,incasetinybubblesortinydropsarepresentinthecontinuousphase.
Thetotal holdup a intheexternal de-oxygenation circuit plays arole inthismethod.
Unlikeeinequation (2),thispropertyiswelldefinedandeasytomeasure.
Theothervariablesthat needtobemeasuredtoderive(ka)rare*,cm,c out ,c c r ,
andCd,r-Incasethereactorisassumedtobeperfectlymixedc/n=cCil-,andifthevessel
isassumedtobeperfectlymixed,cout=ccy. Fromequation(3)requirementsforusable
operationofthesystemcanbederived.ForaccuratemeasurementcCil-andcCiVshould
deviateatleast20%.FurthermorecCJshoulddeviateatleast20%fromCdj/m toobtain
adrivingforceformasstransferthatishighenoughtoderivemass-transfercoefficients.
Coming uptobothrequirements leadstothefollowing constraint:

Vv(ka)vVr(ka)r (4)
< <i> < 5Vu(ka)v
4Vv(ka)u - Vr(ka)r

EXPERIMENTAL

Dynamicandsteady-statemeasurementswerecarriedoutforanLLR.Aschematic
overviewoftheexperimentalsetupisgiveninfigure 1.Thereactor hasaheightof0.7
mandavolumeof0.0045m3.Thespargerconsistsof12pipeswithaninnerdiameter
of0.0014m.Tapwater was usedascontinuous phase.Theperfluorochemical FC40
(3M)wasusedasdispersed phase.The dispersed-phaseflow ratewasvaried,which
changed theka,mainly dueto increased exchange area between thetwophases.
Oxygenconcentrationsweremeasuredwithpolarographic probes.
Forboth,dynamic andsteady-statemethod,thecontinuous phaseinthereactor
isassumedtobeperfectlymixed.Forthedynamicmethodequation(2)isusedtoderive
a value forkafrom response measurements. Forthesteady-state ka determination
equation (3)isusedandthede-oxygenationvesselisassumedtobeperfectlymixed.
No large drops were entrained inthe circulation loop andthus not inthe external
de-oxygenationcircuit.Neithersmalldropletsofdispersedphasecouldbedetectedin
theexternalde-oxygenationcircuit,whichmeansa = O.
96 CHAPTER5

RESULTSAND DISCUSSION

In figure 2 an example of a steady-state measurement is given. Oxygen

concentration is plotted as function of time. The measured steady-state value was
verifiedbyrepeatingtheexperimentwithadifferent initialoxygenconcentration. After
sometimethesameconcentrationasinthefirstexperimentwasreached,indicatinga
truesteady-state.

0.250

oxygen
concentration
[mol.rrf 3 ]

0.125

6000

Figure2: Exampleof asteady-state measurement (probe P3)fortwodifferent initialvalues of oxygen

concentration.
 CHAPTER5 97

Infigure3the results ofdynamic andsteady-state measurements aregiven.No

significant difference isfoundbetweentheresultsof both methods.Themeasuredka
ismuchtoolowforanysignificantinfluenceofprobelag.Inthiscasethemostimportant
advantage of the steady-state method was the possibility to measure at higher flow
rates,where the dynamic measurement could not be executed properly, caused by
restricted supply capacity of FC40 with a constant oxygen concentration. In the
steady-statemethodthesupplybufferofFC40needsnotnecessarilybesaturated,but
willinanycasereachasteady-statevaluethatcanbemeasuredaccurately (Cd).

0.006 - O

o
ka [s' 1 ] A

0.004 -

AO

0.002 - O
O

O
n
0.005 0.01

superficial dispersed-phasevelocity [m.s"1]

Figure3: kavaluesforthe liquid-impelled loop reactor asfunction of the superficial dispersed-phase

velocity inthe maintube (dynamic measurements: A; steady-state measurements:O).
98 CHAPTER5

Themethodforsteady-statemeasurementsusedinthisworkgivesaccurateand
adjustable oxygen uptake rates,whichcan becontrolled bytheflow rateofthe liquid
and of the nitrogen inthe external de-oxygenation circuit. The method can easily be
used for model systems, but for practical applications the biological oxygen
consumptionneedstobeknown.Forlargeapplicationsorhighkavalues,highdemands
areposedonthecapacityofthede-oxygenationvesselasisexpressedinequation(4).
Themixing inthe reactor andthe mixing interaction withthe by-pass needto bewell
described. However, in many cases assuming a perfectly mixed reactor and
de-oxygenationvesselwillbesufficientlyaccurate.

CONCLUSIONS

Thesteady-state methodofkadetermination isverysuitableandcanbeagood

alternativeforthedynamicmethodifprobelagdisturbsthemeasurement.Alsoincases
whereitisdifficult to generate aproper step inthe inlet concentration,suchasforan
LLR,thismethodcomesinuseful.Evenfortheproblemoftheexistenceoftinybubbles
ordropsinthesystem,thismethodcanbemoreaccuratethanthedynamic method.
Yet, inthis case the total holdup inthe external de-oxygenation circuit needs to be
determined.

NOMENCLATURE

a specificexchangearea m-1
c oxygenconcentration mol.m-3
k overalloxygen-transfer coefficient m.s-1
m distributioncoefficientofoxygenoverdispersedand -
continuous phase
OUR oxygenuptakerate mol.m-3.s-1
t time s
 CHAPTER 5 99

V volume m3
a holdupdispersed phaseintheexternal de-oxygenation
circuit
e holdupoftiny bubblesortinydropsofdispersed phase
4> flowratetode-oxygenationvessel m3.s-1

subscripts

c continuous phase
d dispersed phase
in from reactortode-oxygenationvessel
out fromde-oxygenationvesselto reactor
r reactor
v de-oxygenationvessel

ACKNOWLEDGEMENT

TheauthorswishtothankK.Van'tRietforhiscommentsonthemanuscriptandApplikon
Dependable Instrumentsfortheirfinancialsupport.

REFERENCES

[1] Alvarez-Cuenca M., Nerenberg M.A. (1981).Oxygen masstransfer inbubble columns working
at largegasandliquidflow rates.AlChEJ.,27,66-73.
[2] HeijnenJ.J.,Van'tRietK.,WolthuisA.J. (1980).Influenceofverysmallbubblesonthedynamic
Klameasurements inviscous gas-liquid systems. Biotechnol.Bioeng.,22,1945-1956.
[3] LinekV.,VacekV.,BenesP. (1987).Acritical reviewandexperimentalverificationofthe correct
useofthedynamic methodsforthedetermination ofoxygentransfer inaeratedagitatedvessels
towater, electrolyte solutionsandviscous liquids.Chem. Eng.J.,34,11-34.
100 CHAPTER5

[4] ManoT., KimuraT., lijima S.,Takahashi K., Takeuchi H., Kobayashi T. (1990). Comparison of
oxygen supply methods for cultures of shear-stress sensitive organisms including animal cell
culture.J.Chem.Tech.Biotechnol.,47,259-271.
[5] MerchukJ.C.,Yona S.,Siegel M.H., BenZviA. (1990).Onthefirst-order approximation tothe
response of dissolved oxygen electrodes for dynamic kla estimation. Biotechnol.Bioeng.,35,
1161-1163.
[6] MoserA.(1981). BioprocessTechnology. Springer-Verlag.
[7] VanSonsbeek H.M., Verdurmen R.E.M.,Verlaan P.,Tramper J. (1990). Hydrodynamic model
forliquid-impelled loop reactors. Biotechnol. Bioeng.,36,941-946.
[8] Van't Riet K. (1979). Review of measuring methods and results innonviscous gas-liquid mass
transfer instirredvessels. Ind. Eng.Chem.Process Des.Dev., 18, 357-364.
 101

CHAPTER 6

OXYGEN TRANSFER IN LIQUID-IMPELLED LOOP

REACTORS USING PERFLUOROCARBON LIQUIDS

SUMMARY

Oxygentransferintheliquid-impelledloopreactorisdescribedforaset-upinwhich
theperfluorochemicalFC40isaeratedexternally.Twosizesofreactorsareinvestigated.
Themass-transfercoefficientkappearstobelowerwithafactorofabout0.6compared
togas/liquid systems.Thespecific exchange areainthe present experimental set-up
isfound to be favorable compared to gas/liquid bioreactors at the same superficial
dispersed-phasevelocities.However,slowcoalescenceofthedispersed-phasedrops
inthephase-separation section limitsthedispersed-phaseflow rateseriously. Incase
this becomes crucialfromthe pointofviewof oxygensupply,specialmeasures need
to befoundoralternativessuchascombinedspargingofairandsolvent.

INTRODUCTION

Oxygen is in many cases of biotechnological processes an essential nutrient.

Becauseofitslowsolubilityinaqueousmedia,oxygenisusuallysuppliedcontinuously
byspargingairinthemedium.Therateofoxygentransferfromtheoxygen-richdispersed

Thischapter isacceptedfor publication inBiotechnologyandBioengineeringasanarticleby

theauthors:H.M.VanSonsbeek, H. DeBlankandJ.Tramper.
102 CHAPTER 6

phasetothemediumisoften ratelimitingforthewholeprocess.
Other restrictions inbiotechnological processescanbecausedbythesensitivity
of the biocatalyst for shear stresses or concentration gradients. In general, high
oxygen-transfer rates go together with high shear forces, which gives conflicting
requirementsforshear-sensitiveorganisms,suchasplantandanimalcells.Toprevent
celldamageinthesecases,mechanicalforcesinthebrothmustbekeptlow.Thisleads
to poor mixingand lowtransfer coefficients, which inturncan bedetrimentaltocells.
Obviously specialmeasuresarerequired.
The best performance with respect to aeration and mixing can be obtained in
air-sparged,stirredtank reactors athighimpeller speeds, butthis isaccompanied by
high shear forces. Bubble columns (BCs) have approximately the same volumetric
energy efficiency for mass transfer9 and can provide high mass-transfer rates with
considerable lower shear forces than stirred tanks.4 However, mixing in BCs is less
favorable compared to stirred tanks. An advantage for applications is the simple
construction,whichbecomesreallydiscriminatingatlargescales.Inair-liftloopreactors
(ALRs)themeangasresidencetimeislowerthaninBCsandthereforetheyshowlower
transferratesatthesamegasinputrates.Howeverthemaximumgasinputrateishigher
forALRs,whichmeansthatthemaximumcapacity ishigherthanforBCs.Mixingtimes
are longer inALRsthan in BCs but particles are kept more easily insuspension and
stagnant zones inthe reactor are prevented bythe circulation flow. Shear forces are
lowwhichmakesALRsveryattractivereactorsforgrowingshear-sensitiveorganisms.4
AlthoughtheALRseemstobethebestcompromiseoftheconflictingrequirements,
severalinvestigators report about acell-damaging effect of sparging ingeneral16and
ofbubblebreakupatthesurfaceinparticular8.12.Manyvariationsonbasicreactortypes
have beentestedand insomecasesair bubbles havebeencompletely bannedfrom
thesystem.Manoet al. 13 describefour oxygen-supply techniquesfor shear-sensitive
organisms:free-surfaceaeration,supplyviaporousteflontubing,externalaeration,and
theuseofperfluorocarbonasoxygencarrier.DamainoandWang6discussthepossible
advantages of the use of perfluorocarbons ina liquid/liquid spray column fermentor
system compared to air-sparged stirred-tank reactors and BCs.The oxygen-transfer
ratecanbehigherfortheliquid/liquidsystemduetothepossibilityofalargerexchange
areafor masstransfer.Thisexchange areacanbehigherthaninagas/liquid system
becauseofformationofdropsthataresmallerthanbubbles.Furthermore,liquidholdups
can be maintained at a higher levelthan gas holdups and drops do not coalesce as
 CHAPTER6 103

quicklyasgasbubbles.DamainoandWang6alsoclaimthatshearforcesinliquid/liquid
systems are extremely low. Perfluorocarbons are found to be non-toxic for
microorganisms2, which makes them useful as oxygen carrier for biotechnological
purposes. Cho and Wangs demonstrate an increase in volumetric mass-transfer
coefficientwhensprayingofairsaturatedperfluorocarbonisaddedtoabioreactorwhich
isdesignedforcultivationofhybridomacells.
The research described inthischapter concerns oxygentransfer inthe recently
introduced liquid-impelled loop reactor (LLR).This new type of bioreactor operates
accordingtothesameprincipleasanALR.Insteadofagasphase,awater-immiscible
liquidphaseisusedasdispersed phase.20Analogoustotheexperiments ofChoand
Wangs and Damaino and Wang6 the application of the LLR can be to pursue the
cultivationofshear-sensitivecells.OthermotivesfortheintroductionoftheLLRthatare
evenmoreimportant,arerelatedtotheadvantagesoforganicsolventsinbiocatalytical
media. Limitations caused by solubility or reaction equilibria can be eliminated and
substrateandproductinhibitioncanbereducedwhenthesecompoundsdissolvevery
well in organic solvents.11 Furthermore product recovery is facilitated.11 In previous
work hydrodynamics andmixingaredescribed22>23anditistheobjectiveofthiswork
to investigatethe general suitability of an LLRasabioreactor with respect to oxygen
supply.Forthisreason,thetransferofoxygenfromtheperfluorocarbon FC40towater
in an LLR has been characterized by measuring overall oxygen-transfer coefficients.
Theroleofaxialmixing,endeffects,anddrop-sizedistribution isdiscussed.

THEORY

The rate of mass transfer between two phases is usually defined by the
concentrationgradient,theexchange areaandtheoverallmass-transfer coefficientk.
Partialmass-transfer coefficients ofbothphasescontribute inthefollowingway:

(1)
I = J_ + 1
K m KH Kr
104 CHAPTER 6

wherem isthepartitioncoefficientofoxygenoverdispersedandcontinuous phase;k^

andkcarethe partialmass-transfercoefficientsinthe dispersedand continuousphases,
respectively. When gas isthedispersed phase inagas liquidsystem,k$isvery high
compared tokc. This is dueto amuch higher oxygen diffusion coefficient inthe gas
phase. Inadditionto this effect,thevalue ofmfor oxygen is32,which meansthatat
equilibriumthe gas phasecontainsmuchmoreoxygenthanthewaterphase.Inequation
(3) itcan beseenthat botheffects makethe overall mass-transfer coefficient k inthe
gas/liquidsystemsmainlydependentofthepartialmass-transfercoefficient intheliquid
phase.
Thevalueofmfor oxygen inFC40andwater is 12,which isnot highenoughto
excludeaneffect onthe overall mass-transfer coefficient k.Thisalsodepends onthe
valueofkd, whichismuchlowerthanforagasbecauseofalowerdiffusion coefficient
of oxygen in FC40 compared to air. Therefore it can be expected that the overall
oxygen-transfer coefficient for an FC40/water system is lower than for an air/water
system.

Table I: Correlationsfor partial mass-transfer coefficients inliquid/liquid systems.

Rednumber stateof drop Equationstoderivepartialmass-transfercoefficientfor reference eq.

dispersedandcontinuous phase no.

Newman15
Red<10 Stagnant
VB (7)
*•- -H$i*-"{-<" '*)) Ranz and (8)
Shc = 2 + 0.57#e°' 5 0 Sc°' 3 3 Marshall17

k
-HiiBUA-64KD^))
10<Red<200 Circulating Kronig and (9)
*- Brink"

Shc = -126 + \.8Re°csSc°cA2 Garner, (10)

Foord and
Tayeban1

Red>200 Oscillating . 5.5 Dd( eg V - Rose and (11)

n 2 d 0 ' 7 l 3 p d +2pj KintneMS

Shc = 50 +0.0085Re cSc°c-70 Garner and (12)

Tayeban7
 CHAPTER6 105

Intable Iafew correlations that can easily be applied are given for kaandkc.
Notwithstandingthesecorrelations,therearestillalotofaspectsthatmakeestimation
of mass-transfer coefficientsvery unreliable.Masstransfer during dropformationand
coalescencemayhavegreatinfluence,butisstillnotclearlyunderstood.Todetermine
theinterfacialarea,thedrop-sizedistributionandthedispersed-phase holdupneedto
be measured. Especially the former is always of limited accuracy.10.19 Another
complication is that the individual mass-transfer coefficients ka and kc can not be
measuredbutarerepresentedbyanoverallcoefficientk. Furthermore,theinfluenceof
surfacetensionandsurfaceactiveagents,thelattereveninverysmallconcentrations,
may affect mass transfer considerably, but agood description isstill lacking. Finally,
axial dispersion in column contactors has influence on the ka measurements, but is
usuallynotexactlyknown.

MATERIALSANDMETHODS

LLR
Tworeactorsofsimilardesignbutdifferentscaleareused.Thedimensionsofboth
reactorsaregivenintable II. Bothreactorsaremadeofglass.Aschematicdrawingof
thereactorcanbeseeninfigure 1,thatgivesthecompleteexperimentalset-upforthe
pilot-plantLLR.TapwaterisusedascontinuousphaseandFC40isusedasdispersed
phase.Forpracticalreasonsallmeasurements arecarriedoutat25°C.

FC40
FC40 is the trade name (3M) of a mixture of perfluorocarbons. The physical
properties are given intable III.This hydrophobic solvent mixture consists of ring or
straight-chainhydrocarbons inwhichhydrogenatomshavebeenreplacedbyfluorine.
Perfluorocarbons are quite stable to heat and chemical reagents, because of the
presence ofthevery strong carbon/fluorine bonds14.Thesolubility ofseveralgases,
forinstanceoxygenandcarbondioxide,isveryhighrelativetowater.Becausethereis
no chemical attraction of oxygen or carbon dioxideto perfluorocarbons, Henry's law
fordissolvedgasisapplicable.
106 CHAPTER6

Table II: LLRdimensions.

labscale pilot plant

maintube: volume*(m3) 0.0018 0.0181

length(m) 0.64 2.3
diameter (m) 0.06 0.1

circulation loop: volume (m3) 0.0004 0.0030

length(m) 0.74 2.39
diameter (m) 0.025 0.04

bottom section: volumecontinuous phase(m3) 0.0013 0.0082

volumedispersed phase (m3) 0.0003 0.0030
diameter** (m) 0.06 0.1

top section: volume (m3) 0.001 0.0057

diameter** (m) 0.025 0.04

distance maintubeandcirculation loop(m) 0.45 1.05

liquid sparger: number of pipes 12 24

innerdiameter ofpipes(m) 0.0014 0.0020

' definedwithrespecttothe mass-transferfromdropsto continuous phase.

'*concerningthemajor partofthissection.

Dynamickadetermination
The most widely used method for determination of the overall oxygen-transfer
coefficientkaisthedynamicmethod.Theresponseonastep-wisechangeintheoxygen
inletconcentration ismeasuredasafunctionoftimeandfromthisthekaisderived.21
To carry out this dynamic measuring methodtwo buffer vessels with FC40 are
used.One isaeratedandtheother isspargedwith nitrogen.Thestepchange ininlet
oxygenconcentrationisobtainedbyswitchingthefeedfromonevesseltotheother.A
difficulty in this system is the generation of a proper step change. Residence-time
distributionintubesandinthepressurechamberoftheliquidsparger maysmooththe
stepchange.Duetothe relativelyshorttimescaleofthiseffect,howevertheinfluence
isexpectedtobeofminorimportancefortheexperimentalresults.Asystemisdesigned
with two external polarographic dissolved-oxygen electrodes in flow cells, which
guarantees constant and sufficiently high flow rates at the electrode membrane at
 CHAPTER 6 107

different modesofoperation ofthe LLR.Oneelectrode isusedto monitorthe oxygen

concentration intheaeratedbuffervesselthatcontainsthe FC40;theother isusedto
record the change of oxygen concentration as function of time in the reactor. The
dynamicmethodisusedforthelab-scaleLLR,butnotforthepilot-plant LLR,because
it has been too difficult to generate a proper step-wise change in dispersed-phase
oxygenconcentrationthatlastlongenoughonthisscale.

Table III: FC40physical properties (3M).

density (25°C) 1.870x103kg.m-3

dynamicviscosity (25°C) 4.5x10-3Pa.s
interfacialtensionwithwater (20°C) 34.9x10-3N.m-1
surfacetension (25°) 16.0x10-3N.m-1
oxygensolubility (25°C) 15.1mol.rrr3
boiling point 155°C
average molecular weight 0.650kg.moH

Steady-state ka determination
The inaccuracies and problems associated with the dynamic method can be
overcomebysteady-statemeasurement ofka.Thegenerationofastepchange inthe
inletoxygenconcentration becomes redundant.21 Forsteady-state measurements a
constantuptakeofoxygeninthereactor needstotakeplace.Forthe LLRanartificial
oxygen-consuming systemisbuilt,whichisdescribed inmoredetailinpreviouswork
and is schematically drawn infigure 1. 21 A small flow of continuous phase from the
reactor iscirculatedviaavessel, inwhichde-oxygenation by sparging nitrogentakes
place.Thesystemgivesaccurateandadjustable oxygen-uptake rates.Theflowrates
of the liquid and the nitrogen through the de-oxygenation vessel are the control
parameters.Theoxygen-uptakeratecanbecalculatedfrommeasurementsofflowrate
and oxygen concentrations of the inlet and the outlet flow of the vessel, which are
measured with two polarographic dissolved-oxygen electrodes. The oxygen
concentration in the aerated FC40 buffer vessel is monitored by means of a
polarographic dissolved-oxygen electrode in an external circuit. The steady-state
methodisusedfor boththelab-scaleandthepilot-plantLLR.
108 CHAPTER 6

flowmeter

condenser
1. buffer vessel
2. oxygen probe
3. valve
4. pump
5. magnetic stirrer

Figure 1: Aschematic drawingofthe LLRwiththeartificialoxygen-consuming system.

 CHAPTER 6 109

RESULTSAND DISCUSSION

Infigure 2 an example is given of two steady-state measurements of ka inthe

pilot-plantLLR.Theoperatingconditionsforbothrunsarethesame.Theonlydifference
is the concentrations at the beginning of the experiment. As expected for a proper
steady-statemethod,thesamesteadystateforbothrunsisreachedafter awhile.

0.250

oxygen LLR
concentration
[mol.rrf 3 ]

0.125

de-oxygenationvessel

1
1

1 i i i , i
1
0 500 1000 1500
time [s]

Figure2: Anexample oftwo steady-state measurements ofka inthe pilot plant LLR atdifferent initial
concentrations (experiment 1: ; experiment 2: ).

Infigure3theaveragesofkavaluesthatarederivedfromexperimentalresultsare
given. Forthecalculation ofthesekavalues,thecontinuous phase isassumedto be
perfectlymixed.Duplicatemeasurementsarecarriedoutforthelab-scale LLR,andthe
resultingkavaluesshowamaximumdeviationof 15%oftheaveragevalue.
110 CHAPTER 6

0.006- a
D
ka [s 1 ] o A
oa o
0.004-
A

oa A
A
0.002- •
D
• O
D
0 1
1 1
l
0 0.005 0.01
v
sd [m.s 1 ]

Figure3: ka as function of the superficial dispersed-phase velocity in the main tube (lab-scale LLR,
dynamic method: O; lab-scale LLR, steady-state method:• ; pilot-plant LLR, steady-state
method: A).

Asalready concluded inprevious work21 it isseen infigure 3thatthe dynamic

andsteady-statemethodgiveequivalentresults.Thepilot-plantexperimentsarecarried
outuptofivetimes,andtheresultingkavaluesshowastandarddeviationof10%ofthe
averagevalue.Theassumptionofaperfectlymixedcontinuousphaseisvalidwhenthe
reciprocalvalueofthevolumetricoxygen-transfer coefficient issmallerthanthemixing
time.Thepresentexperimentsdidnotmeetthisrequirementandfrompreviousworkit
isknownthattheflow inthereactor hasastrongplug-flowcharacter.23Therefore,the
maximumpossible influenceoftheassumptionofperfectly mixedcontinuous phaseis
investigated.Numericalcalculationswiththeothermixingextreme,plugflow,arecarried
out.Infigure4theresultsforthepilot-plantreactoraregivenanditcanbeseenthatthe
difference betweenbothassumptions islessthan 13%ofthecalculatedvalues,which
meansthat for these calculations no serious errors are madewiththe assumptionof
perfectmixing.
 CHAPTER 6 111

0.006 -

ka [s~1] •

A
0.004 -
o
A
o
A

0.002 - 2

0 i 1
1
0 0.005 0.01

v s d [m.s-1]

Figure4: kavaluesforthepilot-plant LLRasfunctionofthesuperficialdispersed-phasevelocity inthe

maintube (perfectly mixed continuous phaseassumed:O; plugflowfor continuous phase
assumed.A).

When all the kavalues of figure 3 are correlated according to an exponential

equation/this givesthefollowing result:

ka = 0.32u (2)

Intheliteratureasimilar correlationisfoundfor BCs:9

ka 0.36u (3)

Equations (2)and (3)indicatethatthe LLRsinthis researchand BCsingeneral,give

aboutthesamekavaluesasfunctionofthesuperficialdispersed-phasevelocity.This
isaremarkablefactbecauseinloopreactorslowerkavaluesareexpectedthaninBCs.
112 CHAPTER 6

It probably resultsfrom a larger interfacial area at equal superficial dispersed-phase

velocities and confirms what Damaino and Wang suggested.6 One of the possible
reasonsforthelargerinterfacialareaisahigherholdup.MorecertainisthesmallerFC40
drop sizecomparedtotheairbubblesize.Thegeneralapplicability ofequation (2)is
limitedbecauseitisdependentonmanypropertiessuchasdensity,viscosity,interfacial
tensionof bothliquidphases,drop sizeandreactor geometry. Although equation(2)
isverypromisingformass-transferintheLLR,themaximumpossibledispersed-phase
flowratethatcouldbeappliedsofar,isoneorderofmagnitudelowerthanforgasliquid
systems,suchas BCsandALRs.Specialmeasurestoovercometheseproblems are
required.
Calculations of kvaluesfrom the measuredkavalues is as already stated very
dubious,becauseofinaccuratedeterminationoftheinterfacialareaandendeffects.It
isnotinthescopeofthisworktoinvestigatethisfurther.Keepinginmindtheseeffects,
/rvaluesarederivedfrommeasuredkavalues, dispersed-phaseholdups,andaverage
drop sizes.The holdupa, that isderivedfrom pressure difference measurements in
previous work23, is given infigure 5 as afunction of the superficial dispersed-phase
velocity inthe maintube (i/sd)- A good linear correlation with slope 3 s.m-1 isfound,
whichisremarkablysimilarforbothreactors.

0.05
A
a [-] 0.04-
A''
A
0.03-
O-A
0.02- ,a A
P*
0.01-
JO
,0
0
() 0.005 0.01

v
sd [m-S"1]
Figure5: Holdupasfunctionofthesuperficialdispersed-phasevelocityinthemaintube(lab-scaleLLR:
O; pilot-plant LLR: A).
 CHAPTER 6 113

Fromthesamework23theSautermeandiameter032thatisdeterminedbymeans
ofphotographsisfoundtobeabout2.5x10-3and3.0x10-3mforthelab-scaleLLRand
thepilot-plantLLR,respectively.Infigure6thecalculatedkvaluesaregivenandaverages
ofabout3x10-4and2x10~4m.s-1arefound,respectively.

0.0005 -
k [m.s" 1 ]
0.0004 - 0 °
00
0 ° 0 0
0.0003 -
0
O
0.0002 - A A
A A O

0.0001 -

0 ' 1 ' 1
0 0.005 0.01

v s d [m.s--11]I

Figure6: kvalues,derivedfromkameasurementsasfunctionofthesuperficialdispersed-phasevelocity
inthe maintube (lab-scale LLR:O; pilot-plant LLR: A).

This is a little lower, but still inthe same order of magnitude as mass-transfer
coefficients ingas liquidsystems9andalittle higherthanthevaluefound by Manoet
al. 13for asimilar system (0.9X10-4m.s-1).Calculationswiththeformulas intable Ifor
thepilot-plantLLRresultsinkvaluesofabout1x10"4m.s-1 forstagnantdropsand3x10-4
m.s-1 for circulatingandoscillatingdrops.Althoughfledinthe pilot plant LLRisabout
350, masstransfer seemsto behave more likestagnant drops.The calculations give
good order of magnitude estimations but a distinction between different state of the
dropscannot bederivedfromtheseexperiments.
114 CHAPTER 6

CONCLUSIONS

Volumetricoxygen-transferratesintheLLRaredeterminedandfoundcomparable
to oxygen-transfer rates inBCsatequaldispersed-phase flow rates. Oxygen-transfer
coefficientsarefoundtobealittlelowerthaningas/liquidsystemsbutstillinthesame
orderofmagnitude.Literaturecorrelationsformass-transfercoefficientsleadtoagood
orderofmagnitudeestimation.Theremarkablygoodoverallvolumetricoxygen-transfer
coefficient inthe LLRscomparedtogas liquiddevices is caused bythesmaller drop
size compared to the bubble size. However, the maximum possible superficial
dispersed-phase velocities accomplished in the present configurations, are
considerably lower thanwhat can be attained ingas/liquid systems. If mass-transfer
limitations in general threaten to play an unacceptable limiting role, the present
configurationsareunsuitable.Improvementsofthesystemtoincreasethetransferrate
mustthenbeconsidered.Formationofsmaller droplets byabetter controlled droplet
formationand/orincreaseofthemaximumdispersed-phaseflowbyanimprovedphase
separation must be strived for. If it does not concern shear-sensitive organisms, an
alternativecanbealoopreactorinwhichspargingwithairinonetubeiscombinedwith
aerationviaair-saturatedorganicsolventintheother.

ACKNOWLEDGEMENT

Thisworkwas partlysupported byApplikon Dependable Instruments B.V.Schiedam,

TheNetherlands.

NOMENCLATURE

a specificexchangearea nr 1
B coefficient
D diffusioncoefficient ofoxygen m2.s-1
d dropdiameter m
 CHAPTER 6 115

d32 Sautermeandiameter m

k partialoroverallmass-transfer coefficient m.s-1

m partitioncoefficient ofoxygenoverdispersedandcontinuous
phase
n index
Rec Reynoldsnumber continuous phase = pcvsdr\~'

Red Reynoldsnumber dispersedphase = pdvsdr\ ~d'

5c c Schmidt number continuous phase = Ti c (p c D c ) _ 1

Shc Sherwoodnumbercontinuous phase = kcd(Dcy1

t time s
1
vsd superficialdispersed-phasevelocity inmaintubeoftheLLR m.s-

vsc superficialcontinuous-phasevelocity inmaintubeoftheLLR m.s-1

vs slipvelocity =v^a'1 - vsc( 1 - a ) " 1 m.s-1

a holdup inthemaintubeoftheLLR
r| dynamicviscosity Pa.s
A.„ n theigenvalue

P density kg.m-3
a interfacialtension N.nr1

subscripts:
c continuous phase
d dispersedphase
116 CHAPTER 6

REFERENCES

[1] Al-AswadK.K., MumfordC.J.,JeffreysG.V. (1985).Theapplication ofdropsizedistributionand

discretedropmasstransfermodelstoassesstheperformanceofarotatingdisccontactor.AlChE
J., 31,1488-1497.
[2] Brink L.E.S., Tramper J. (1985). Optimazation of organic solvent in multiphase biocatalysis.
Biotechnol.Bioeng.,27,1258-1269.
[3] BrinkL.E.S.,TramperJ., Luyben,K.Ch.A.M.,Van'tRiet,K.(1988).Biocatalysisinorganicmedia.
EnzymeMicrob.Technol.,10,736-743.
[4] Cherry R.S., Papoutsakis E.T. (1986). Hydrodynamic effects on cells inagitated tissue culture
reactors. Bioprocess Engineering, 1, 29-41.
[5] ChoM.H.,WangS.S.(1988).Enhancementofoxygentransferinhybridomacellculturebyusing
aperfluorocarbon asanoxygencarrier. Biotechnol.Lett., 10,855-860.
[6] DamianoD.,WangS.S.(1985).Noveluseofaperfluorocarbonforsupplyingoxygentoaerobic
submerged cultures. Biotechnol.Lett.,7,81-86.
[7] GarnerF.H.,Tayeban,M.(1960).Continuousphasemasstransfercoefficientsinoscillatingdrops.
An. R.Soc.Esp.Fis.Quim. B56,479-490.
[8] Handa-CorriganA., EmeryA.N., Spier R.E. (1989).Effect of gasliquid interfaces onthe growth
of suspended mammalian cells: mechanisms of cell damage by bubbles. Enzyme Microb.
Technol., 11,230-235.
[9] Heijnen J.J., Van 't Riet K. (1984). Mass transfer, mixing and heat transfer phenomena in low
viscosity bubble column reactors.Chem.Eng.J.,28,21-42.
[10] HughmarkG.A.(1967).Liquidliquidspraycolumndropsize,holdupandcontinuousphasemass
transfer. Ind. &Eng.Chem.Fundamentals,6,408-413.
[11] KronigR.,BrinkJ.C.(1950).Onthetheoryofextractionfromfallingdroplets.Appl.Sci.Res.,A2,
142-154.
[12] KunasK.T.,PapoutsakisE.T.(1990).Damagemechanismsofsuspendedanimalcellsinagitated
bioreactorswithandwithout bubbleentrainment.Biotechnol. Bioeng.,36,476-483.
[13] ManoT., KimuraT., lijima S.,Takahashi K., Takeuchi H., Kobayashi T. (1990). Comparison of
oxygen supply methods for cultures of shear-stress sensitive organisms including animal cell
culture.J.Chem.Tech.Biotechnol.,47,259-271.
[14] Mattiasson B., Adlercreutz P. (1987). Perfluorochemicals in biotechnology. Trends in
Biotechnology, 5,250-254.
[15] Newman A.B. (1931). The drying of porous solids: diffusion and surface emission equations.
Trans.Am.Int.Chem.Engrs.,27,203-220.
[16] OhS.K.W.,Nienow A.W.,Al-RubeaiM., Emery M. (1989).Theeffects of agitation intensity with
andwithout continuous spargingonthegrowthandantibody production ofhybridoma cells.J.
Biotech., 12,45-60.
[17] Ranz W.E., Marshall W.R. (1952). Evaporation from drops l+ll. Chem. Eng. Progr., 48,
141-146,173-180.
 CHAPTER6 117

[18] RoseP.M.,KintnerR.C.(1966).Masstransferfromlargeoscillatingdrops.AlChEJ.,12,530-534.
[19] Slater M.J., Baird M.H.I., Liang T.-B. (1988). Drop mass transfer coefficients for liquid liquid
systemsandthe influenceof packings.Chem. Eng.Sci.,43,233-245.
[20] Tramper J., Wolters I., Verlaan P. (1987). The liquid-impelled loop reactor: a new type of
density-difference-mixed bioreactor. 311-316. In: C. Laane, J. Tramper, M.D. Lilly (eds.),
Biocatalysis inorganic media, Elsevier Science Publishers.
[21] VanSonsbeek H.M., GielenS.J.,Tramper J. (1991).Steady-state methodforka measurements
inmodel systems. Biotechnol.Techn.,5,157-162.
[22] VanSonsbeekH.M.,VerdurmenR.E.M.,VerlaanP.,TramperJ. (1990).Hydrodynamicmodelfor
liquid-impelled loopreactors.Biotechnol.Bioeng.,36,940-946.
[23] VanSonbeekH.M.,VanDerTuinS.P.,Tramper J. (1992).Mixinginliquid-impelledloopreactors
1992.Biotechnol.Bioeng.,39,inpress.
 119

GENERAL DISCUSSION

In 1987 the liquid-impelled loop reactor (LLR) was introduced.9 This thesis
representsthefirstcomprehensivestudyonthisdevice.Theresultscanbeusedforthe
descriptionofthephysicalbehaviorandforthedeterminationofdesignconstraintsand
suitabilityfor biotechnological purposes. Inthis lastpart ofthethesis remainingfacts,
potentialitiesandlimitationsarediscussed.

Two-phase bioreactor systems

The thesis starts with a review about two-liquid-phase bioreactors, which was
writtenattheendofthestudytogiveanimpressionofthestateoftheart.Someofthe
devices described in the literature resemble the set-up of the LLR. The continuous
extraction loop reactor of Gianetto et al.4 (chapter 1,figure 5) is very similar to the
external-loop LLR.Theextractivefermentationthatwastestedinthisdevice appeared
to befeasiblebutalsocomplications didarisewiththeimmobilization matrix.Juetal.5
mentionintheirworksomeexperimentswithadevicethatisalikeaninternal-loopLLR.
Difficulties with protein accumulation at the liquid/liquid interface were encountered
which ledto protein stripping.Therefore the investigators stopped their work on this
deviceandswitchedtheirattentiontotheuseofaeratedemulsionsofaperfluorocarbon
and an aqueous phase. Damiano and Wang3 used a liquid/liquid spray column
fermentorsystemandduetoalargerinterfacialareatheyachievedhigheroxygensupply
rates compared to a bubble column of similar design.The maximum capacity of the
device appeared to be limited by slow drop coalescence at high cell concentrations.
ChoandWang2addedanindirectaerationviaaperfluorocarbontoanexistingreactor
typeandmeasuredanincreasedoverallmass-transfer coefficientbutcouldnotrunthe
system because of cell aggregation at the interface of the medium and the
perfluorocarbon.TheseexamplesofdevelopmentssimilartotheLLRshowthegeneral
interest in the subject and the possible advantages. However, it shows also new
difficultiesthatcanbeintroduced.
120 GENERAL DISCUSSION

Choiceoforganicsolvent
The continuous phase in the LLR is supposed to contain the biocatalyst and
because no application of (immobilized) biocatalysts, which preferably stay in the
organic phase, seemed to be within reach, only configurations with an aqueous
continuous phase were investigated. The most important criteria for suitability of a
solventinanLLRaredensitydifferencewiththeaqueousphase,interfacialtensionand
thelog Pvalue. Extractionpropertiesare notdiscussedhere,butbecomereallyrelevant
if actual applications are developed. The first experiments were carried out with
hexane/water systems, because of the high density difference and the fact that the
relative easy design of an upflow configuration could be used. Also good
phase-separationbehaviorwas foundinpreliminaryexperiments.Experimentswiththis
configuration resulted inthe hydrodynamic model asdescribed inchapter 2.FC40,a
solventthatwas usedinthe plantcellproject1,isasolventthathas evenbetterproperties
tobeusedintheLLR.Especiallythelog Pvalueof 11.2comparedto3.5forhexaneis
aconsiderable improvementwithrespecttoretentionofbiocatalyst activity. FC40has
ahigherdensitythanwater,whichimpliedthatadownflowconfigurationhad tobeused.
Bydesigning, building andtesting new reactor set-ups,two well-to-handle downflow
configurationswerebuiltandfurtherexperimentswerecarriedout withthe modelsystem
of FC40andwater.
Somelab-scaleexperimentsweredonetotestothersolvents.Alltestedsolvents
aregivenintableI.Onlysiliconoil and parafinoilappearedto beunsuitableforapplication
intheLLRbecauseofgradualemulsification,evenatverylowflowrates.
Insome preliminary experiments the possibility of aqueous two-phase systems
weretested.Itisnotsurprisingthatpropertiessuchastheextremelowsurfacetension,
the highviscosity andthe low density difference, result inuphillwork. Drop formation
andcoalescencebecomecriticalfor properoperation.
 GENERAL DISCUSSION 121

Table I: Solvents used as dispersed phase In the liquid-Impelled loop reactor, where water is the
continuous phase.

solvent logP density viscosity interfacialtension

* kg m-3**(20°C) 10-3Pa.s**(20°C) N.m-1**(20°C)

FC40 11.2 1870 4.5 0.035***

Hexane 3.5 660 0.3 0.05**
Hexadecane 8.8 770 3.3 -
Petroleum-ether - 640 - -
Nonane 5.1 718 0.7 -
Dodecene 6.2 758 - -
oleic acid 7.9 890 38.8 -
siliconoil - 937 12 -
parafinoil - 840 - -
8
*Pisthe partitioncoefficient inn-octanol/water calculatedaccording Rekker.
" datafromsupplier or Handbook of chemistryand physics10
***determined by3M

Fluiddynamicsanddesign
Thecirculation modelasdescribedinchapter 2andthemixingmodelofchapter
4canbeusefulforestimationsoffluid-dynamicbehaviorofanLLR.Itisstrikingthatthe
results of such complex physical processes, which seem so chaotic by visual
observation,canbeestimatedwithsuchsimpleexpressions.Demandsontheaccuracy
ofthesemodelsshouldthereforenotbetoohigh.
Both topics of fluid dynamics have resulted in useful methods to estimate the
hydrodynamicpropertiesofanewdesignbeforeactuallybuildingit.Themostdelicate
question istheestimation oftheoverallfrictioncoefficient ofthereactor. Calculations,
basedonempiricaldataofallkindofobstaclesintubes, cangivegoodresultsaswas
foundinchapter2,butonemusttakeintoaccountalimitedaccuracy.Whentheoverall
frictioncoefficientcannotbecalculatedaccuratelyenoughbecauseofthepresenceof
unconventional geometries or obstacles, it can still be useful to determine it
experimentally andapplyitforfurther descriptionoffluiddynamics.
122 GENERAL DISCUSSION

An important design aspect isthe ratio of the main-tube cross section andthe
circulation-tubecrosssection.Inthisworkforthelab-scaleexperimentstworatioswere
used:2.25:1and5.77:1.Forthepilot-plantexperimentsalsotworatioswereused:4:1
and6.25:1.Theoverallfrictioncoefficient increaseswithincreasingratioandtheeffect
on the overall mass-transfer rate is advantageous. Some unpublished experimental
resultsshowedthatchoosingahighcrosssectionalratiocanhaveconsequenceswhen
workingwithathirdphasesuchasgelbeads.Evenwhenthesuperficialliquidvelocity
forfluidizationishighenough,thecross-sectionchangecancausedeadspaces,where
thefluidisationofthebeadsisnotsufficient.

Masstransfer
Biocatalytic processes usually have much lower reaction-rate constants than
chemicalprocesses.Thelimitationisoftencausedbymass-transfer rates.Thenumber
of investigations concerning the oxygen-transfer rate from a gaseous phase to an
aqueousphaseisstillendless.Organicsolventsaresometimesintroducedtoimprove
oxygen-transfer rates. Usinganair-saturateddispersedphaseinsteadofairtosupply
oxygentothe aqueous phasewillleadto anincreased barrierfor oxygento moveto
theaqueous phase.Thiscanbecompensatedandresultinanoverallimprovementif
theexchangeareaisincreased.BotheffectsareindeedfoundfortheLLRanddescribed
in chapter 6 and result in an overall improvement compared to gas/liquid systems.
Becauseoflimiteddispersed-phaseflowthemaximumkathatisattainedismorethan
oneorderofmagnitudelowerthanwhatispossibleforgas/liquid bioreactors.Forfast
growing cells such asyeasts, moulds and bacteria,the demand on oxygen-transfer
ratecan be high. Inthat case,withthe state ofthe art ofthis moment, othertypes of
bioreactorsthantheLLRaremoresuitable.Whendemandsonoxygen-transferrateare
notdominatingthewholebiocatalyticprocess,theLLRcanbesuitable.Anexampleis
the cultivation of shear-sensitive cells such as plant, animal and insect cells, where
besides oxygen supply also shear stress is playing an important role. Loop reactors
andindirectaerationareconsideredtobeadvantageousinthatcase.Tofindoutabout
thesuitabilityforaerationofshear-sensitive culturesaroughestimationcanbemade.
Anexample isthecultivation of insectcellswithatypical oxygen-consumption rateof
4.3x10"17mol.s-1.celH andtypicalmaximumcelldensitiesof5x106cells.mH.7With ka
valuesdeterminedinchapter 6themaximumpossiblecelldensity is29x106cells.mH,
whichindicatessufficientaerationcapacity.
 GENERALDISCUSSION 123

Coalescence
Becauseofapossiblesuitabilityforcultivationofshear-sensitivecellsexperiments
aimed at insect-cell cultivation were carried out. Coalescence was found to be
unacceptableslowbecauseofmediumcomponents.Itisassumedthatthemaincause
isprotein accumulation atthe liquid/liquid interface.Theexperiments showed strong
hysteresiseffectsandthemaximumapplicablesuperficialdispersed-phasevelocitywas
decreasedoneorderofmagnitude.
ItisbeyonddoubtthatcoalescenseisacriticalaspectfortheapplicationofLLRs.
The limited flow rate of the dispersed phase limits mass transfer and thus reactor
productivity. Becauseproteinscannotbeavoidedinfermentations itcanbeexpected
thatslowcoalescencewillalwaysplayadominatingrole.
Asimple improvement canbemade by enlargingthe phase separationsection,
inotherwordstoincreasethesettlingtime.Destabilizationofemulsionswithchemical
agentsisstillratherempiricalandunexplored.Inaninitialstudy Lennieetal.6showed
that small molecule surfactants, which ontheir ownwould stabilize anemulsion,can
destabilizeemulsions,probably bydisplacingthe protein atthe liquid/liquidinterface.
Controllingemulsionformationinthisway isstillratherempiricalandunreliable.Alsoa
grid or a packing material in the phase separation section of an LLR can improve
coalescence and mustbeseriously considered. Furthermore, amembranefor phase
separationseemsapossibility,whichisshownbysomeexperimentsbyM.M.R.Fonseca
(Portugal).

References
[1] Buitelaar R.M.,SusaetaI.,Tramper J. (1990).Application oftheliquid-impelled loop reactor for
theproduction ofanthraquinonesbyplantcellcultures.694-699.In:H.J.J. Nijkamp,L.H.V.Van
DerPlas,andJ.VanAartrijk(eds.),Progressinplantcellularandmolecularbiology,KluwerAcad.
[2] ChoM.H.,WangS.S.(1988).Enhancementofoxygentransferinhybridomacellculturebyusing
a perfluorocarbon asanoxygencarrier. Biotechnol. Lett., 10,855-860.
[3] DamianoD.,WangS.S.(1985).Noveluseofaperfluorocarbonforsupplyingoxygento aerobic
submerged cultures. Biotechnol. Lett.,7,81-86.
[4] Gianetto A., Ruggeri B., Specchia V., Sassi G., Forna R. (1988). Continuous extraction loop
reactor (CELR): alcoholic fermentation by fluidized entrapped biomass. Chem. Eng. Sci.,43,
1891-1896.
[5] Ju L-K., Lee J.F. Armiger W.B. (1991). Enhancing oxygen transfer in bioreactors by
perfluorocarbon emulsions.Biotechnol.Prog.,7,323-329
124 GENERALDISCUSSION

[6] LennieS.,HailingP.J.BellG.(1990).Causesofemulsionformationduringsolventextractionof
fermentation brothsanditsreduction bysurfactants. Biotechnol.and Bioeng.,35,948-950.
[7] Maiorella B., Inlow D., Shauger A., Harano D. (1988). Large scale insect cell-culture for
recombinant protein production. Bio/Technology,6,1406-1410
[8] RekkerR.F.,DeKortH.M. (1977).Thehydrophobic fragmentalconstant;anextensiontoa1000
datapoint set.Eur.J.Med.Chem.Chim.Ther., 14, 479-488.
[9] Tramper J., Wolters, I., Verlaan P. (1987). The liquid-impelled loop reactor: a new type of
density-drfference-mixed bioreactor. 311-316. In: C. Laane, J. Tramper, M.D. Lilly (eds.),
Biocatalysis inorganic media, Elsevier Science Publishers.
[10] WaestR.C.,AstleM.J.(eds.) (1978-1979)Handbookofchemistryandphysics,5 9 * edition,CRC
Press.
 125

SUMMARY

Theliquid-impelledloopreactor (LLR)isareactorthatconsistsoftwo parts : the

maintubeandthecirculationtube.Bothpartsareinopenconnectionatthebottomand
atthetop.Thereactor isfilledwithaliquidphase:thecontinuousphase.Anotherliquid
phaseisinjectedinthemaintubebymeansofpumping.Thisliquidphaseisimmiscible
withthecontinuousphaseanditsdensityissignificantlydifferent.Ifthedensityishigher
thanthedensityofthecontinuous phase,injectiontakesplacesatthetopofthemain
tube.Foralowerdensityinjectiontakesplaceatthebottom.Duetothedensitydifference
thedispersed-phasedropletsthatareformedwillfallorrise,respectively,andcoalesce
attheother endofthetube.Thecoalesced liquidisdischargedfromthe reactor. Due
tothepresenceofdispersedphaseinthemaintubeapressuredifferenceexistswhich
causescirculation ofthecontinuous phase inthe reactor.Thisresults ingoodmixing
withouttheuseofanimpeller. Forbiotechnological purposes itismost likelythatthe
continuous phase is an aqueous phase that contains the biocatalyst, possibly
immobilized.Theworkontheliquid-impelledloopreactororiginatesfromtwoprevious
research studies. First, the physical characterisation and modelling of air-lift loop
reactorsfor applicationincultivatingshear-sensitivebiocatalysts.Second,theresearch
focusedonapplicationoforganicsolventsinbiologicalprocesses,which isapromising
areaformanyyearsalready.
Inareviewthecurrentstateoftheartisgivenwithrespecttobiocatalysisinmedia
consisting of two liquid phases that are not miscible.This research area has shown
much progress in recent years, however, industrial applications seem still not very
numerous. To carry out two-liquid-phase experiments on a scale bigger than shake
flasks,inmostcasesexistingbioreactorsareused.Adjustmentsaremadetothereactor
and the processing to make them suitable for use with two liquid phases. The
liquid-impelledloopreactorcanbeseenasaspecialcase,whereanair-liftloopreactor
isadjustedfor usewithtwoliquidphasesinsteadofaliquidandagasphase.
Theresearchwasstartedwithcharacterisationofimportantphysicalaspectssuch
asdrop size,dispersed-phase concentration (holdup) andcontinuous-phase velocity
asfunctionofthedispersed-phaseflowrate.Descriptionofdropsizesthatareformed
126 SUMMARY

in the liquid-impelled loop reactor at the liquid sparger, show good agreement with
theoreticalpredictions.Thehydrodynamicmodelthatwasusedintheair-liftloopstudy
isapplied. Itshowsto beagood methodto describe holdup and liquidvelocity.The
bestresultswiththismodelareobtainedwhenitisassumedthatthecontinuousphase
flowsfastestinthecentreofthetubeandthatthedispersedphaseconcentrates inthe
centreofthetube.
Onthetopic of hydrodynamic modelsfor air-lift loop reactors many articles are
published.Becausethisstillcontinues,thisliteratureisanalysedandthebasicprinciples
ofthe several models aredescribed and compared. Itappears that in allmodelsthe
relative velocity between dispersed-phase bubbles and continuous phase plays an
essentialrole.Thisquantityisdifficulttodetermineandshowsawidespreadduetothe
distributioninbubblesize.Furthermore,velocity profilesorturbulencecanhavemuch
influence but are nottaken into account inthe described models. Comparison ofthe
modelsbymeansofusingliteraturedatadidnotyieldaclearpreferenceforoneofthe
modelsnorforaparticular basicprinciple.
Todescribemixinginthecontinuousphase,theone-dimensionalaxialdispersion
model is used, which is in general suitable for flow in tubular devices. The mixing
parameter isdeterminedper reactor section.Forthemaintubeacorrelation between
mixingparameter andenergy dissipation isgiven.Themixingparameter canbeused
to describe the flow of the continuous phase as a plug flow with axial disturbances.
Furthermore,dimensionlessmixingtimescanbeestimated.Thedimensionlessmixing
timeisthenumber of circulationsthat isnecessaryto achievecomplete mixingofthe
continuous phase,wherethecriteriummustbedefinedbytheuser.
Masstransfer is investigated inan FC40water system. For this purpose a new
methodisdeveloped basedonthe principle of asteady-state measurement, instead
ofthe mostwidely useddynamic measurement. Comparedto agas/liquid systemat
equaldispersed-phase flow rates,the mass-transfer rateinthe liquid/liquid systemis
favorable.Thisisduetothelargerexchangearea,becausethedropsaresmallerthan
the bubbles and the drop velocity relativeto the continuous phase is lower thanthe
relativevelocityofthebubbles.Themass-transfercoefficientfortheliquid/liquidsystem,
derivedfromexperimentalresults,however,islowerthanliteraturevaluesforgas/liquid
systems.Thisisprobably causedbythelowerdiffusion coefficient of oxygeninliquid
than of oxygen in gas.The transfer capacity can often bethe highest for gas/liquid
 SUMMARY 127

systems because the maximum dispersed-phase flow rate inliquid/liquid systems is

limitedwithrespecttodropformationandcoalescence.Furtherphysicalreseachmust
befocussedonthislimitation.
 129

SAMENVATTING

Deliquid-impelledloopreactor (LLR)iseenreactordiebestaatuittweedelen:een
hoofdbuiseneencirculatiebuis.Beidedelenstaanaandeboven-enonderzijdeinopen
verbindingmetelkaar.Dereactorisgevuldmeteenvloeistoffase:decontinuefase.Met
behulp van een pomp wordt in de hoofdbuis een vloeistoffase geTnjecteerd die niet
mengbaar is met de continue fase en waarvan de dichtheid significant anders is.
Wanneer diedichtheid hoger isdande dichtheidvandecontinue fase,wordt aande
bovenkant van de hoofdbuis geTnjecteerd. Bij een lagere dichtheid wordt aan de
onderkant geTnjecteerd. De gevormde disperse fase druppels dalen, respectievelijk
stijgen,alsgevolgvanhetdichtheidsverschil encoalescerenaanhetandereeindvan
dehoofdbuis,waarnaafvoervandezevloeistoffaseplaatsvindt.Doordeaanwezigheid
vandedispersefasedruppelsindehoofdbuisbestaatereendrukverschildatcirculatie
van de continue fase in de reactor teweeg brengt. Hierbij komt een gecontroleerde
mengingtotstand,zonderdatdaarvooreenroerdernodigis.Voor biotechnologische
toepassingen ligt eenwaterige continue fase het meestvoor de hand. Hierin bevindt
zichdande biokatalysator, alof nietgeTmmobiliseerd.Hetideevande liquid-impelled
loop reactor is voortgekomen uit twee voorgaande onderzoeksprojecten. Enerzijds
was dat het fysisch karakteriseren en modelleren van air-lift loop reactoren, dat
bedoeldwasvoortoepassing bijhet kwekenvanafschuifkrachtgevoelige biologische
katalysatoren, zoals plantecellen. Anderzijds was dat het onderzoek gericht op
toepassing vanorganische oplosmiddelen inbiologische processen, dataljaren een
veelbelovendonderzoeksgebiedis.
Allereerstwordtinditproefschriftdestandvanzakengeschetsmetbetrekkingtot
biokatalytische processen in media,bestaande uittwee vloeistoffasen die slecht met
elkaar mengbaar zijn. Het onderzoek op dit terrein heeft de laatste jaren veel
vooruitgang geboekt, al lijken daadwerkelijke toepassingen nog niet zo talrijk. Om
experimenten uittevoerenopeenschaalgroterdanschudflesjes wordt indemeeste
gevallengebruik gemaaktvan bestaande bioreactoren. Hierbijworden de reactor en
de procesvoering aangepast voor gebruik met twee vloeistoffasen. De LLR kan
130 SAMENVATTING

beschouwd worden als een bijzonder geval waarbij een air-lift loop reactor is
aangepast voor gebruik met twee vloeistoffasen in plaats van een vloeistof- en een
gasfase.
Ineersteinstantiezijnbelangrijkefysischeeigenschappenzoalsdruppelgrootte,
disperse fase concentratie (holdup) en continue fase vloeistofsnelheid gekarak-
teriseerd als functie van net disperse fase debiet. Voor de beschrijving van de
druppelgrootte,diemetdeontworpenvloeistofverdelers indeLLRontstaan,blijkenin
de literatuur beschreven theorieen goed bruikbaar. Het hydrodynamicamodel
afkomstig uit het air-lift loop reactor onderzoek is toegepast en blijkt een goede
methode om de holdup en de vloeistofsnelheid te beschrijven of te voorspellen. De
besteresultatenmetditmodelwordenverkregenwanneer aangenomenwordtdatde
vloeistof inhet midden hetsnelst stroomtendatdedispersefasezich bijvoorkeur in
hetmiddenvandebuisbevindt.
Voorair-lift loopreactorenzijnalveelartikelenverschenenoverhydrodynamica-
modellen. Omdat er nog steeds gepubliceerd wordt over dit onderwerp, is deze
literatuurgeanalyseerdenzijndebasisprincipesvandediversemodellenaangegeven.
Het blijkt dat in alle modellen de relatieve snelheid tussen de disperse fase en de
continuefaseeenessentielerolspeelt. Ditisjuisteenmoeilijktebepalengrootheiden
vertoont een grote spreiding door de heterogeniteit in beldiameter. Ook wordt in de
beschrevenmodellennietofnauwelijks rekeninggehouden metstromingsprofielenof
turbulenties die beide de hydrodynamica sterk kunnen bei'nvloeden. Onderlinge
vergelijking van de modellen met behulp van literatuurdata resulteert niet in een
duidelijkevoorkeurvoor eenbepaaldmodelofbasisprincipe.
Voordebeschrijvingvandemengingindecontinuefaseisgebruiktgemaaktvan
heteendimensionaleaxialedispersiemodel,datinhetalgemeenheelgeschiktisvoor
strominginbuisvormigeapparaten.Demengparameterisperreactorsegmentbepaald
en voor de mengparameter van de hoofdbuis, waarin zich twee vloeistoffasen
bevinden,iseencorrelatie opgesteld metdeenergiedissipatie indiebuis. Metbehulp
van de mengparameters kan de stroming van de continue fase in de reactor
beschrevenwordenalseenpropstroommetkleineaxialeverstoringen.Verderkunnen
dimensieloze mengtijden geschat worden. De dimensieloze mengtijd is het aantal
circulaties dat de continue fase nodig heeft om volledig op te mengen, waarbij het
criteriumvoorvolledigemengingzelfgedefinieerdmoetworden.
 SAMENVATTING 131

De stofoverdracht is onderzocht in een FC40/water systeem. Hiervoor is een

nieuwemethodeontwikkelddiegebaseerdisopeensteadystatemeting,inplaatsvan
de meestaltoegepaste dynamisch meetmethode. Bijvergelijkbare debieten disperse
fase wordt een goede stofoverdracht waargenomen ten opzichte van gas/lucht
systemen. Dit is te danken aan een groter uitwisselend oppervlak, dat wordt
veroorzaakt doordat de gevormde vloeistofdruppels kleiner zijn dan luchtbellen en
omdatdesnelheidvandruppelstenopzichtevandecontinuefaselagerisdandievan
luchtbellen.Destofoverdrachtscoefficient voorhetvloeistof/vloeistofsysteem,afgeleid
uitdeexperimentelewaarnemingen,isechter lager danvooreengas/water systeem.
DitiswaarschijnlijkhetgevolgvandeveellagerediffusiecoefficientvanzuurstofinFC40
vergeleken met die in lucht. Vaak kan de overdrachtscapaciteit van gas/vloeistof
systemenveelgroterzijndanvaneenvloeistof/vloeistof systeemomdat hetmaximale
toepasbare debiet metbetrekkingtot druppelvormingencoalescentie een beperking
vormtvoorvloeistof/vloeistof systemen.Verderfysischonderzoek dientgerichttezijn
opdezebeperking.
 133

NAWOORD

AllereerstwilikAnjabedankenomdatikdankzijhaar nuweetdataldiecliches,dieik
zovaaklees,helemaalgeenclicheshoeventezijn.HansTramperwilikbedankenvoor
deprettigevriendschappelijkewerksfeer,diehijhooginhetvaandelhoudt.Zijndirecte
begeleidingensnellereaktiesophetschrijfwerk compenserenhetsomslatenuitlopen
vanvergaderingen ruimschoots. Heel belangrijk voor mijzijn ook de discussies met
Klaasvan 't Riet geweest. Inveelopzichten is dat een complementaire begeleiding
geworden.Wanneerhetmogelijkzouzijnvantweepersonenalleeigenschappensamen
tevoegenomvervolgensdaaruittweenieuwepersonentecreeren,danzoudaarinhet
gevalvan Hans en Klaas beslist een hoogleraar uit voortkomen die in alle opzichten
ideaalis.Detweedehoogleraar diedannogoverblijft,zouiklieverniettegenkomen.
Ikwilalmijncollega'sbedankenvoordegezelligetijdenbeperkmetothetnoemenvan
een paar namen: Wim Beverloo, die erg veeltijd aan mijn sprookjes besteedt heeft;
Albertvander Padt, diegezorgd heeftvoor degezonde hap eneenforse uitbreiding
vanmijnmuziekkennis;MarcelZwietering,dieervoorgezorgdheeftdatikoptijdnaar
Loburgging; MoniekVerbeek, metwiehetgezelligwerkenwasinonzekantoortuin.
Natuurlijk ben ik ook veel dank verschuldigd aan de studenten die zich voor mijn
onderzoek ingezet hebben: Sylvia de Hoop, met haar streven naar perfectie; Ruud
Verdurmen,hetdropjesmonster;ErwinBonarius,diemuziekzietinproceskunde;Erna
Noordkamp, met een hoog cpu-timeverbruik; Stephen Gielen,met dynamischwerk
aandestatische methode; MartenJongsma,voorwietijdnooit eenprobleem is;Erik
van Dam,zijn boek zalik beslist gaan lezen; HeindeWilde,die emulsiesecht stabiel
weette krijgen;EricvanCan,voorwiegeenfrictiete hoog is;Henriette de Blank,die
zeindewerkplaats nogsteedsnietvergetenzijn;entenslotte Henk-Jan Meijer,dieze
oplab632nogsteedsnietvergetenzijn.
Complimentenhebikvoorhetvakwerkendeflexibeledienstverleningvandewerkplaats,
tekenkamer,fotolokatie,endeafdelingenelectronicaenautomatisering.MetnameJan
TheunissenenEvertJanssenhebbenvelemalengoedeideeenaangedragenvoorde
praktische uitvoering. Op het gebied van glaswerk heb ik vakkundige hulp van
134

Technoglasgehad.Tenslotte wilik hier nogvermeldendat dit project medemogelijk

gemaaktisdoorfinancieleondersteuningvan,ineersteinstantie,ContactFlowenlater
vanApplikonDependable Instruments.

HenkvanSonsbeek
 135

CURRICULUMVITAE

7-3-1960 1972 1978 1982 1983 1987 1988 1992

geboren V.W.O. Hogere Alpe start afgestudeerd promomotie
in Druten School d'Huez studie aan onderzoek
voor aan Landbouw- waarvan
Levensmidelen- Landbouw- universiteit dit
hogeschool proefschrift
technologie
het
studierichting:
resultaat
Levensmiddelen-
technologie
hoofdvakken:
proceskunde
informatica