^V\ (92© t -y&Z

J.A.M. Spaninks

Design procedures for solid-liquid extractors

and the effect of hydrodynamic instabilities
onextractor performance

Proefschrift
terverkrijging van degraadvan
doctorin de landbouwwetenschappen,
opgezagvan derector magnificus,
dr. H.C.van derPlas
hoogleraarin deorganische scheikunde,
inhet openbaar teverdedigen
opwoensdag6juni 1979desnamiddags tevieruur
in deaulavandeLandbouwhogeschool teWageningen

Centrefor AgriculturalPublishingandDocumentation

Wageningen — 1979

\0TJVA
VA
Abstract

Spaninks, J.A.M. (1979) Design procedures for solid-liquid extractors and the effect of
hydrodynamic instabilities on extractor performance. Agric. Res. Rep. (Versl. land-
bouwk. Onderz.) 885, ISBN 902200693 X, (xi + 100 p., 44 figs, 10 tables, 155 refs,
8app.,Eng.andDutchsummaries.
Also:Doctoralthesis,Wageningen.

A design method is proposed for countercurrent mass transfer cascades with cross-
flow stages, and unsteady operated fixed beds in a countercurrent arrangement. The
separation performance of these cascades is calculated from mathematical models and
compared with a purely countercurrent extractor. The models are based on simple
equations describingmasstransfer inasinglestagefixed bed.Theresultsarepresentedas
concisecorrelations between thenumber of trueand exterior apparent transfer units.
Asymptotic values of themasstransfer coefficient for transient diffusion processesin
cocurrent and countercurrent extractors are calculated for particles with simplegeome-
try. Comparing the extraction efficiency calculated from the exact solutions of the
diffusion equations and the values obtained from the asymptotic mass transfer coeffi-
cient shows that differences are small under normal operating conditions for counter-
current extraction. The use of the asymptotic values in design calculations is discussed.
Due to the commonly observed viscosity and density gradients in the solventphase,
non-uniform flow ofliquid through thelayer of solidssubjected to extraction can occur.
Then some fraction of the solids is not efficiently contacted with the solvent. Experi-
ments show that a concentration gradient rather thanaconcentrationjump in theliquid
flowing through aninert packingmaterialreducesthe effect of channelling.Theeffect of
mass transfer between the packing and the surrounding liquid on flow stability is stud-
ied. The results indicate that channelling can reduce the extraction efficiency in solid-
liquid extractors.

Freedescriptors:leaching,extraction efficiency.

BIBLIOTHEEK LH.
2 8 ME11979
0NTV. TfJDSCHR. ADM.

Thisthesiswillbepublished asAgricultural Research Reports885.

© Centrefor Agricultural PublishingandDocumentation,Wageningen, 1979.

No part of this book may be reproduced and/or published in any form, by print,photoprint, micro-
film orany other meanswithout written permission from thepublishers.
N/J fóo/ ^z
Stellingen

1. De gebruikelijke indelingvanvast-vloeistof-extractieapparatuur naar percolatie-enim-

mersietype isniet zinvol.

Ditproefschrift, hoofdstuk 2.

2. Ten onrechte stelt Genie dat de eerste-termbenadering van de oplossingvan dedif-

fusievergelijking niet toelaatbaar isvoor het ontwerp van vast-vloeistofextracteurs.

Genie,G.V., 1973.Int. SugarJ. 75:67-70.

Ditproefschrift, hoofdstuk 3.

3. Als gevolg van slecht gedefinieerde begincondities ondervond Rein moeilijkheden bij
de interpretatie van zijn experimenten gericht op verificatie van zijn model voor de ex-
tractie van suikerriet. Deze problemen konden voorkomen worden door de experimenten
uit te voeren analoog aan dein Appendix Evan dit proefschrift beschreven methode.
Rein,P.W.& E.T.Woodburn, 1974.Chem.Eng.J. 71:41-51.

4. Terecht stelt Mikhailov dat het oplossen van warmte- en stofdiffusievergelijkingen

vaak minder tijdrovend is dan het opzoeken van de oplossing in de bestaande literatuur.
Het valt echter te betwijfelen of de door deze auteur gepubliceerde 'algemene oplos-
singen' eenbruikbaar alternatief stellen.

Mikhailov,M.D., 1972.Int. J.Eng.Sei. 10:577-591.

Mikhailov,M.D., 1977.Int. J.HeatMassTransfer 20: 1409-1415.
/•

5. Het dispergeren van gassen in oplossingen heeft eennadelige invloedopdearomareten-

tie tijdens het vriesdrogenvan deze oplossingen: dit ishet gevolgvan demanierwaarop de
gasbellen in de ijs-vaste-stofmatrix worden ingebouwd.

Spaninks,J.A.M., 1974.Afstudeerverslag Afdeling ScheikundigeTechnologie,THEindhoven.

6. De door Hanna gesuggereerde analogie tussen dragreductie door het toevoegen van
polymeren ingepakte bedden en inpijpen geldt slechtsvoor volledig turbulente stroming.
Hanna,M.R.,W.Kozicki&C.Tiu, 1977.Chem.Eng.J. 13:93-99.

7. Bij het concentreren van eiwithoudende oplossingen zoals wei met behulp vanomge-
keerde osmose wordt de economie van het proces sterk beïnvloed door de kosten van
membraanreiniging. Dit kostenaspect wordt onvoldoende onderkend.
 8. De door Hayakawa gepubliceerde eigenwaarden voor molekulair stof- en warmte-
transport zijn praktischvanweinigwaarde.
Hayakawa, K., 1975.Lebensm.Wiss.u.Technol. 8:231-233.

9. Het nut van het gebruik van blokschema's ter verduidelijking van computerprogram-
mateksten inprogrammeertaal moet worden betwijfeld.

10. Behalve economische redenen zijn er uit het oogpunt van verkeersveiligheid vol-
doende argumenten om onder de huidige omstandigheden niet te streven naar de pro-
duktievanautomobielen met een gebruiksduurlanger dantienjaar.

Proefschrift vanJ.A.M. Spaninks

Design procedures for solid-liquid extractors and the effect of hydrodynamic instabilities
on extractor performance.
Wageningen, 6juni 1979.
Woordvooraf

Velenhebben bijgedragen aanhettotstand komen vanditproefschrift. Mijn promotor

prof. dr.ir. S.Bruin heeft door zijn voortdurende belangstelling ensuggestieseenbelang-
rijke invloed op de inhoud van dit werk gehad. Dank ookaanmijn copromotor prof.dr.
ir. H.A.C. Thijssen, diemetenthousiasme hetonderzoek heeft gevolgd.Metmijn collegae
van de Sectie Proceskunde heb ik menig interessante discussie over dit werk mogen
voeren. Metname wilik hier ir.K.J.Beukema noemen, dievaak op kritische wijze mijn
uiteenzettingen heeft willen volgen. Inhetbijzonder voorhunbijdragen aanhet opzetten
van de experimentele opstellingen ben ik dank verschuldigd aandeledenvandeCentrale
Dienst Biotechnion.Hetbelangrijkste deelvandeexperimentele resultaten dieinditwerk
worden gepresenteerd komt voort uitdedoctoraal onderzoeken vanJohan Bachman, Gijs
Dorrestein, Marco Meyer, Haitze Meurs, Frank Termote, Piet Grin en Gerard v.d.Hazel-
kamp. Voor de uiteindelijke vormgeving vanditproefschrift dankikdemedewerkersvan
Pudoc, alsmede deheren Rijpma en SchimmelvandeLandbouwhogeschool diemetzorg
de figuren hebben getekend. Tenslotte zou hetafronden vanditproefschrift veelmoeilij-
ker geweest zijn zonder de bereidwillige medewerking van mijn huidige werkgever,het
Koninklijke/Shell-Laboratorium Amsterdam.
Curriculum vitae

De auteur werd geboren op 31 oktober 1951 te Tilburg. Hij behaalde in 1969 het
diploma HBS-B aan het St. Odulphuslyceum aldaar. Inhetzelfde jaar begon hij zijn studie
aan de Technische Hogeschool Eindhoven. In juli 1972werd het kandidaatsexamen afge-
legd in de richting Scheikundige Technologie. In de laatste studiefase specialiseerde hij
zich in de fysische technologie. Na het behalen van het ingenieursdiploma in juni 1974
trad hij als wetenschappelijk medewerker in dienst bij de Sectie Proceskunde van de
Landbouwhogeschool te Wageningen. In deperiode 1974-1978 werd het promotie-onder-
zoek verricht. Sinds september 1978 ishij werkzaam bij de afdeling Equipment Engineer-
ingvanhet Koninklijke/Shell-Laboratorium in Amsterdam.
Contents

listofsymbolsmostfrequentlyused

1 Introduction 1
1.1 Solidliquidextractioninthefood industry 1
1.2 Organizationofthestudy 5
1.3 Somegeneralremarksonthemodelsystemstudied 10

2 Effectoflocalphasecontactonextractionefficiency 12
2.1 Classification ofsolid-liquidextractors 13
2.2 Diffusion battery 14
2.2.1 Mathematicalmodelofasingle-stagefixedbed 15
2.2.2 Mathematicalmodelofadiffusion battery 16
2.2.3 Efficiency ofadiffusion battery 17
2.3 Belttypeextractor 19
2.3.1 Mathematicalmodelofasingle-stagecross-flow device 19
2.3.2 Mathematicalmodelofabelttypeextractorwithoutliquidrecirculation 21
2.3.3 Efficiency ofbelttypeextractorswithoutliquidrecirculation 23
2.3.4 Efficiency ofbelttypeextractorswithliquidrecirculation 26

3 Estimationofmasstransfercoefficientsinsolid-liquidextractors 31
3.1 Introduction 31
3.2 Estimatingmasstransfer coefficients inthecontinuousphase 33
3.3 Estimatingmasstransfer coefficients inthedispersedphase 37
3.3.1 Surveyofliteraturedata 38
3.3.2 Sh d a valuesfrom solutionsundercoupledboundaryconditions / 39
3.3.3 Shd a valuesfrom solutionsundergeneralboundaryconditions 49
3.4 Asymptotic Shda numbersinengineeringcalculations 51

4 Liquidmaldistributioninsolid-liquidextractors: hydrodynamicinstabilities 59
4.1 Introduction 59
4.1.1 Stabilitycriteria 60
4.1.2 Flowbehaviourintheunstableregime 64
4.1.3 Experimentalverification fromliteraturedata 66
4.2 Experimental studyofmiscibleliquiddisplacement from packedbeds 68
4.2.1 Effect ofaconcentration gradientintheliquidphaseonflow stability 69
4.2.2 Effect ofmasstransfer fromthepackingtotheliquidonflowstability 73
S Conclusions 79

Summary 80

Samenvatting 81

References 82

Appendices 86

Appendix A. Numerical solution of the equations describingmasstransfer in 86

afixed bed
Appendix B. Criteria for thevalidity of the simplified model of abelt extrac- 88
tor 89
AppendixC. Approximate relations for estimating Shd a values 92
Appendix D. Masstransfer efficiency inthe penetration period
AppendixE. Apparatus and experimental procedures: measurement of break- 93
through curves
Appendix F. Determination of the moments of the measured breakthrough 96
curves
Appendix G. Locating the position of the displacement front in downward
miscibleliquid displacement from packed beds 97
Appendix H. Determination of the model parameters Dand (k 0 a) 100
List of symbolsmost frequently used

a' interfacial areaperunitofsolids (m2/m3 )

a specificinterfacial areaperunitofpackedvolume (m2 /m3)
A cross-sectionalareaof the extractorperpendiculartotheliquid flow (m2)
Ad particlesurface area (m2)
B dimensionpackedsectionoftheextractorperpendiculartothesolids
flowdirection (m)
Bi dimensionlessBiotnumber= m (Dp/2)kc/2Dd (1)
Bo dimensionlessBodensteinnumber= <v>Dp/E)E (1)
c' dimensional continuous phaseconcentration (kg/m3)
c dimensionlesscontinuousphaseconcentration,Eqn2.6 (1)
C dimensionlesscontinuousphaseconcentration,Eqn3.44 (1)
D distributionratio,Eqn2.9 (1)
E> diffusion coefficient (m 2 /s)
2D E effective dispersion coefficient (m 2 /s)
Dp equivalent particle diameter (m)
E frequency distribution of residence times (1)
ƒ separation factor = 77/A (1)
F cumulative distribution of residence times (1)
Fo dimensionless Fourier number =r IDJR2 (1)
g gravitational constant (m/s 2 )
h liquid holdup, interstitial void fraction (1)
hm height of the mixing zone (m)
H belt length (m)
/m Colburn j-factor for mass transfer (1)
k mass transfer coefficient / (m/s)
L extractor dimension in the direction of the liquid flow (m)
m slope of the equilibrium curve (1) .
M viscosity ratio (1)
M moments of the distribution functions
n number of mass transfer sections in acountercurrent cascade (1)
TV number of transfer units (1)
p pressure (Pa)
Pe dimensionless Péclet number = <v>L/E)E (1)
r distance coordinate (m)
R particle radius =D p / 2 (m)
Re dimensionless Reynolds number (1)
s Laplace transform variable
s0 initial saturation of interstitial pore liquid (1)
5 cross-sectional area of the packed section in the extractor
perpendicular to the solids flow (m2)
Sc dimensionless Schmidt number =ju/pE)c (1)
Sh dimensionless Sherwood number =k DJID (1)
r time (s)
r d dimensionlesstime= tIDJR.2 (1)
rf time required to fill the interstitial voids of the packing (s)
<v> superficial velocity (m/s)
vb belt velocity (m/s)
vc superficial critical displacement velocity,Eqn 4.3 (m/s)
vs superficial stable displacement velocity,Eqn4.6 (m/s)
V volume (m3)
W dimensionless dispersed phase concentration, Eqn 3.35 (1)
x,y,z distance coordinates (m)

? dimensionless coordinatex/L (1)

T? separation efficiency, Eqn 1.4 (1)
6 dimensionlesstime,Eqn2.11 (1)
0 dimensionlessvariable,Eqn 2.26 (1)
K permeability (m 2 )
Xj eigenvalue
X wavelength perturbation (m)
A extraction factor (1)
Aj Afor countercurrent extraction;— Afor cocurrent extraction (1)
p viscosity (kg/ms)
ß eigenvalue
v geometry factor; 1,2 and 3 for slabs,cylinders and spheresrespectively (1)
| dimensionless coordinate r/R (1)
If ratio ofmixervolume and free volume column section=Vm i/Vcolh (1)
p density (kg/m3)
T residence time (s)
0 flow rate (m 3 /s)
<j>" flux (m/s)
co' dimensional dispersed phase concentration (kg/m3)
co dimensionless dispersed phase concentration, Eqn 2.7 (1)

Subscript

a asymptoticvalue
b belt
c continuousphase;operationcycle
col column
d dispersedphase
ext exteriorapparent
/ interface;sequencenumber
in phaseinlet
L longitudinal, axial
m mixer; mass
o initial value; overall
oui phase outlet
ref reference value
t true;total flow system
T transversal, radial
v volumetric
A entrained by the dispersed phase
°° value at infinite time or position

Superscript

* equilibrium
— average
1 Introduction

1.1 Solid-liquidextractioninthefoodindustry

Solid-liquid extraction is a unit operation aiming at the selective removal of soluble

components from a solid matrix in a solvent phase. Many synonimus names for this
process are used in the literature, such asleaching,washingand lixiviation. Sometimes
these names are used in relation to the mechanism that is responsible for the solute
transfer. Thismechanism can be simple washingof adheringliquid from thesurface ofa
solidmatrix.Thisoccursfor exampleinthefirst stageoftheextraction ofoilfrom flaked
soybean, after the cellscontaining the oilhavebeenbrokeninthepretreatment stage. In
other mechanisms,solute diffuses through apermeablebarrier.Intheextraction ofsugar
from sugar-beet, the cell walls are deliberately kept intact to prevent the transfer of
colloidal components. Some examples of solid-liquid extraction processes relevant to
food industry are givenin Table 1.The examplesin thistableshowthatformostofthe
important applications the liquid extract isthe major productstreamfrom anextraction
plant. When theprocessisdirectedtowardstheremovalofundesirabletracecomponents,
the solid phase isthe important product. In some cases,boththesolidandliquidstream
are subjected to further processing. In the extraction of oil from soybean for example,
the upgrading of the extractedsolidsisofparamountimportancefor theeconomyofthe
process.

Table 1. Some examples of solid-liquid extraction processesinthe food industry.

Component to be extracted Solidcarrier References

Sugar Beets Silin, 1964

Cane BrünicheOlsen, 1969 ,
Fruits Wucherpfenning, 1976
Oilsand Fats Oilseeds(soybeans, cotton, Hutchins, 1976
peanuts,palmfruit, rapeseed), *
fish
Protein Vegetable seeds,greenleaves Bergv.d., 1972
Coffee solubles Roast andground coffee beans Sivetz, 1963
Teasolubles Tealeaves
Chichory Chichory root
Licorice Licoriceroot Molyneaux, 1975

Removalof trace components

Aflatoxin Cereals,seeds USPatent 4055674, 1977

Caffeine Greencoffee beans Bichsel, 1976
 In this work attention willbe focussed on solid-liquid extraction in the food industry,
though most of the results have general applicability. The main reason is the growing
interest in this unit operation. In the last decades, the extraction ofvegetable oil shifted
from mechanical pressing to solvent extraction. For oilseedswith ahigh oilcontent, the
press cake obtained after mechanical prepressing by expellers is subjected to solvent
extraction. Even direct solvent extraction has been proposed (Bernardini, 1976). Also
there is a growing interest in solid-liquid extraction or diffusion for the production of
sugar from cane (Rein, 1976),instead of the classical imbibition process. Sugar from beet
has been produced by diffusion for several decades. Further, protein and carbohydrate
extraction from vegetable food sources will become of increasingimportance in the near
future.
Compared with other unit operations such as liquid-liquid extraction, limited atten-
tion has been given in the literature to the design of solid-liquid extractors. Even though
the leaching of metal ores is one of the oldest applications of this unit operation, solid-
liquid extraction is of relatively minor importance to the chemical and petrochemical
industries compared with distillation and liquid-liquid extraction, for example. On the
other hand, in the food industry, where this processis of major importance, an overdesign
of the extraction equipment normally has little impact on the final costs of the product
as long asthe overdesign doesnot increase the variable costs.For production of sugar and
soybean oil, the capital cost involve only some percent of the total production costs.In
the near future, with rising energy prices and an increasing scale of operation, it can be
expected that amore accurate design of the extractors willbe necessary.
The specific features of the solid-liquid extraction process can be elucidated by an
inspection of the operation principle of the equipment used. Solid-liquid extractors are
designed to transport a solid phase with a specified velocity through the extraction
section. During the time that the solids reside in this part, an intensive contact between
the particles and the solvent is assured to promote mass transfer. Finally, the apparatus
must separate the phases efficiently before they are withdrawn from the extractor. In
order to make efficient use of the solvent, the overall phase contact generally is in
countercurrence. The mechanical problems associated with processing oflarge amounts of
raw material under the conditions just mentioned have been solved more or less satis-
factorily and haveresulted inmany widely differing designs.
In the literature, solid-liquid extractors are often classified asextractors of the percola-
tion type and immersion type, operating in batch or continuous flow (Rickles, 1965;
Milligan, 1976;Silin, 1964;Bernardini, 1976).Although thisclassification highlights only
one specific feature of the appartus considered, it willbe followed here for a description
of solid-liquid extraction systems. However, it must be kept in mind that this differen-
tiation does not provide any insight in the stage efficiencies that can be expected in a
certain extractor. This disadvantage is overcome in an alternative classification that is
proposed in Section 2.1.The extractortypes mentioned inthe following paragraphshave
been described in some detail by the authorsjust mentioned.
In immersion extractors the solids do not form a densely packed bed, but they are
more or less dispersed in a pool of solvent. In this kind of equipment very finely ground
solids can be processed, aswell asparticles that disintegrate upon extraction. Thisdisinte-
gration would certainly choke the bed of solid particles in a percolation type extractor.
Normally these extraction systems are designed to operate in continuous countercurrent
 flow. Examples are the Hildebrand extractor, that was in use for the extraction of oil
from soybean in the 1950s, the Bonotto extractor and the BMA diffusor. The latter is
used primarily for the extraction of sugar from beet. In some casesintermittent drainage
of the solids isachieved bylifting the solidsperiodically out of the solvent.This principle
is followed in the DDS and Niro diffusors, which are predominantly used for the extrac-
tion of sliced sugar-beet, and in the Kennedy extractor. The main advantages of the
immersion type extractors are their insensitivity to the pretreatment of the solids,and the
ease of operation as the extractor accomodates a wide range of solid-liquid flow ratios,
i.e. extraction factors. Disadvantages are alowsolidsholdup in the system and aconsider-
ableentrainment of fines by the extract flow.
In a percolation extractor the solids are transported on a perforated support as a
compact bed. The liquid flow is distributed over the bed, percolates downwards by the
action of gravity, and leaves the bed dripping through the perforations in the structure
supporting the bed of solids (e.g. a perforated belt). The relatively high solid phase
holdup makes this extractor attractive for large scale extraction of for example oil from
oil seeds and of sugar from cane.They arealsoinusefor extraction of components from
fishmeal, spices,and solubles from tealeaves.
The Bollman extractor is a well-known example of a percolation type extractor that
was used in Europe and the USfor the extraction of soybean oil.Most of them, however,
have been replaced by extractors with a moremodern design suchasDe Smet and Lurgi,
where the solids are transported laterally through the extractor on a perforated belt, or
similar types with perforated throughs connected to an endless chain. Also extractors
with rotating cells provided with a perforated bottom such asRotocell and Caroussel, or
with spargers rotating over stationary cells, such as the French stationary basket extrac-
tor, are in use on a large scale. The operation principle of these extraction systems is
elucidated in Fig. 1. Thesolvent isdistributed overasection ofthe extractor by spargers.
The liquid percolates through the bed of solidsby gravity and after dripping through the
perforated support, it is collected in reservoirs or pans. Thenet liquid flow from inlet to
outlet can be accomplished in severalways.The extract from the collection panin section
i can be pumped to the sparger in section /'+1, asindicated inthe figure. In other cases,
the extract from pan i is recirculated in section iwhile the net liquid flow isobtained by
overflow of the collection pans. A combination of recirculation and transport to the
sparger downstream is also possible (Milligan, 1976). To avoid entrainment of^heliquid
contained in the interstitial voids of the solids to the section upstream, the solids are
often left to drain before leavingasection. ,

fresh
solvent in solids
1 i Asparger
flakes
r^ «
flow
.collection .»< i >^N
\ p a n yS > i+1 s ^ ^
4•
spent
solids _ extract
"" out
pump
Fig. 1. Belt type percolation extractor.
 extract
out

column

solvent Fig. 2. Diffusion battery with 4 col-

inlet umns.

In a diffusion battery or Shanks extractor, the percolation velocity is controlled by

pumping the solvent through the layer of solids. The extractor consists of a series of
columns in which the solids are held stationary. The operation of the extractor isshown
in Fig. 2. The liquid flows allthe way down through the series of columns asindicated in
this figure. Fig.3 illustrates how a diffusion battery with four columns operates.After a
certain 'cycle time' tc the desired depletion of the solids in the first column is attained.
Theliquid inlet isthen switched to the second column. The first column isthen recharged
with fresh solids and is put in the seriesof columns downstream of the last one (position
IV in Fig. 3). After the extract has filled the interstitial pore volume of the freshly
charged column, extract is drawn from the extractor. This cyclic operation is repeated,
thus establishing an essentially countercurrent phase contacting system. This extraction
system was traditionally in use for the extraction of sugar from beet but has been
replaced almost entirely by continuous diffusors nowadays. The diffusion battery isstill
in use for the extraction of roasted and ground coffee, as the water for the extraction is

I I II n cycle-operation
1 I I i —».
1 2 3 4 t=0
1 I I i

1 i I
r
2 3 4 1 t =t c
* I I I
1 I I I
3 4 1 2 t=2 t c
I I i
I I I I
4 1 2 3 t=3tc
* I I I
Fig. 3. Diffusion battery with 4 columns; schematic indication of
operationcycleswithcycletimetc.

4
fed to the extractor at temperatures as high as 170°Cat apressure of 8x 105 Pa.Under
these conditions it is difficult to feed the solids continuously to the extractor. In most
cases between five and eight columns are used (Sivetz, 1963). This extractor is manu-
factured for small scale extraction of olive pomace (Bernardini, 1976) and consists of
four columns inseries.
In percolation extractors the solids are held stationary on arigid support. Thisdesign
implies that no mechanical forces act onthe particles,and consequently the formation of
fines is considerably reduced. Furthermore the extract is filtered while percolating
through the layer of solids,sothat the entrainment of fines by the liquid flow leaving the
extractor is relatively small. On the other hand, a carefully designed pretreatment is
required since both the particle shape and size affect the percolation velocity. As the
solids subjected to the extraction are not in general repacked non-uniform liquid flow
through the layer of solids may restrict the separation efficiency. In particular when
particle sizes are small with consequently low permeability of the packed bed, chan-
nellingmay occur.Thislatter phenomenon willbe discussed inmore detail in Section 1.2.
This description of extraction systems clearly shows that for most of the practical
solutions to the mechanical problems involved in handling large amounts of particulate
solids, the flow pattern of both phases through the extractor is quite complicated. Ob-
viously, this flow pattern must be taken into account in the mathematical models
underlying design procedures for solid-liquid extractors. The designmethods proposed in
this study incorporate these specific features of extraction into the available design proce-
dures for other unit operations, e.g. drying andliquid-liquid extraction.

1.2 Organizationof thestudy

The present study can be divided into two main parts: atheoretical part inwhich the
number of transfer unit (NTU) approach for the design of solid-liquid extractors was
analysed and an experimental part inwhichthe hydrodynamic stability of the liquid flow
in solid-liquid extractors wasmeasured and analysed.
Generally the design of mass transfer processes starts from specified or assumed con-
centrations in inlet and outlet streams,andthe flow rates of the phases.Aswillbe shown
in Chapter 3, the mass flux that occurs in solid-liquid extractors isoften proportional to
the overall driving force for mass transfer, i.e. the local concentration difference of the
solute in both phases. In such cases, the NTU concept, in which the mass transfer
resistance is lumped at the solid-liquid interface, is a convenient basisfor adesignproce-
dure. An outline of the NTUapproach in process designisgivenin Fig.4.The separation
to be attained can then be indicated by the number of exterior apparent or 'plug flow'
transfer units, defined by Miyauchi&Vermeulen (1963):

c
OUt J >
(11)
; W =/ -^rzë
c
in

where c'in and c'out are the inlet and outlet concentrations in the continuous phase,
respectively; c'* is the hypothetical local concentration which would be in equilibrium
with the dispersed phase if the phaseswere contacted in purely countercurrent plug flow.
 Design
specification

Colburn equation

N
ext,c (Semi)- Solution
empirical diffusion
mixing model correlations equation
local phase contact

N =k at k k
t , c o,c c c d

overall mass
transfer d d i t i v i t y rule -
°
coefficient k

T
c =mean residence
time in the
extractor

extractor capacity
!
V • * c Tc/h = extractor
volume

Fig. 4. Outline of the transfer unit approach to the design of solid

liquid extraction equipment.

Only for purely countercurrent contact can it be shown that the number of exterior
apparent transfer units isequal to the number of true transfer units(Colburn, 1939):

*oc«r
-^ext, plug flow ~-"t, c (1.2)

where koc is the mass transfer coefficient on overall liquid phase basis and a is the
effective interfacial area. Here combination of Eqn 1.1 with a mass balance yields for a
linear equilibrium relation the well-known Colburn equation which isshown in Fig.5:

l A
N. « , c -- _! _J A
_ 1[n( -f
\l-f
\
) (1.3)

The separation factor ƒ is related to the extraction efficiency 7? asƒ= 77/Aand 77 is de-
finedas:
 1 0 Fig- 5. Efficiency ofa purely countercurrent
A process. Parameter: extraction efficiency 17.

0c ( c ó u t - c 'in) _ c
out c
in
(1.4)
c m
<Pd ( " i n - i n / ) mO}[n-c[

When deviations from countercurrent plug flow occur, the number of true transfer units
Nt c exceedsNext c , the number of exterior apparent transfer units.In fact:

Table 2. Relation between the number of exterior apparent and true transfer units for some simple
mass transfer stages.

Cin-cout = 1 -exp{(A-l)JVext.c>
Phase contact Liquid Solid 1 -A
phase phase cin-majout

batch 1 -exp[-JVt>c(l+A)]
batch mixed irrelevant
1 +A

continuous flow operation 1 - e x p [-A^ tiC (1 + A)]

cocurrent unmixed unmixed
1 +A
AiVt.c
1
e_?
A KAT- f ĥ"
systematic mixed unmixed 0
1 - exp [- Nt> :A]
movement of the
solid phase l-(l-A)exp[-JVt;CA]

countercurrent unmixed unmixed \1 - exp [ - JV t c (1 - A)]}/(1 - A)

continuous flow mixed mixed Nicl(l+Nic)

1. Complete mixing of both phases at the stage inlet.

 ^ t , c=S(Next,c A> flow
situation) (1.5)

The flow situation of thephases inside the apparatus is determined on amacro-scaleby

the local mode ofphase contact,non-uniform flow ofsolvent through the solids etc.;and
on amicroscopic scale by dispersive mixing between thephases. Forsome simple phase
contacting systems the relation between Next andNt is given in Table 2.However,the
description of extractor types in Section 1.1 hasshownthat the complex flow situations
in practical extractors are not adequately covered by the simple examples. Deviations
from these simple cases are often described by dispersive mixing or backflow models.
Also these models are not likely to present a realistic picture of the flow situationin
diffusion batteries andbelt type extractors for example. In Chapter 2 of this worknew
ready-to-use correlations betweenNext andNt aredeveloped for diffusion batteriesand
countercurrently cascaded crossflow sections. These correlations are based on mathe-
matical models which account for thelocal mode ofphase contact within theapparatus.
In ordertocalculatetherequired residence time ofthe solidsintheextractor toattain
the specified separation, asophisticated guessofthe overallmasstransfer coefficient k0 is
inserted in the definition equation forNt. According totheLewistwofilm theory,kQis
dependent on both themass transfer coefficient in the continuous liquid phase, kc,and
the dispersed solid phase, kà. A constant averaged value of kc can often be assumed
throughout the extractor, asispointed outin Section 3.1.Abundant literature dataare
available for estimating kc values for widely differing process conditions and packing
geometries. Thedispersed phasemasstransfer coefficient kd istime dependent duetothe
transient nature of the diffusion process. In Section 3.1it isshown that the Sherwood
number kADJJDA assumes an asymptotic value for long contact times so that for a
constant diffusivity E)à in the solid phase kA has a constant value. These asymptotic
values are calculated in Chapter 3 from the analytical solution of thediffusion equation
with boundary conditions relevant topurely cocurrent andcountercurrent phase contact.
Using the additivity rule for mass transfer resistances, kQ can be determined from the
estimated value ofkc andthe asymptotic kd value. Although this rule is exact only for
a constant value offc<jandkc, the error introduced when it isapplied to transient mass
transfer is small (Beek &Muttzall, 1975). Thefinal step indicated in Fig.4 then isthe
calculation of the sizeofthe extractor from therequired residence time andthe specified
throughput.
In both Chapters 2 and3,concerned with theeffect oflocal phase contact onextrac-
tion efficiency and the calculation of mass transfer coefficients, plug flow of liquid
between thephases isassumed. Normally however, residence time distribution inoneor
both of the phases reduces the separation efficiency ofthe extractor. InChapter 2,some
models are discussed which canbe used to account for axial dispersion intheliquidand
solid phases. Another important effect that might causeresidence time distribution in the
liquid phase is flow maldistribution causedbyanon-uniform permeability distributionin
a packed bed(Stanek &Szelkeley, 1972). Awell-known example ofthisphenomenonis
the preferent flow alongthe wallofacolumn packed withparticulate solids.The porosity
gradient in apacked given by Pillai (1977) for example showsthat theporosity nearthe
wall exceeds theaverage bedporosity over the distance of about aparticle diameter,see
Fig. 6.Thefractional wallflow resultingfrom thislocalhigh permeability isillustratedin
Fig. 7. In general this effect can be neglected for single phase flow of a liquid with
(1-h)
(1-h)

Fig.6. Porosity gradients in packed beds,y isdistance from

y/R the wall.FromPillai,1977.

uniform physical properties when the ratio column to particle diameter is higher than
20-30. In solid-liquid extractors the physical properties of the liquid phase, i.e. density
and viscosity, often increase in the direction of the liquid flow. In this case, gravity
segregation can occur in upward liquid flow due to the unfavourable density gradient. In
both upward and downward flow, local permeability variations can trigger the formation
of channels or 'viscous fingers', through which the solvent will flow preferentially. The
mechanism responsible for this phenomenon is discussed at length in Chapter 4. The
existence of such stable and unstable flow regimes has been verified experimentally by
many authors. Viscous fingering plays an important role in secondary recovery of oil

Fig. 7. Effect of particle to column diameterra-

tio on wall flow in packed columns. From Rie-
D
p/Dcol tema, 1976.
from underground reservoirs, where a fluid is injected in the porous rock structure to
displace residual oil (Dumoré, 1964). Examples related to the food industry are the
separation of ice crystals from the mother liquor inwash columns(Vorstman &Thyssen,
1972), filter cake washing (Tondeur, 1970), sweetening on and sweetening off of char
columns for decoloration of sugar solutions (Hill, 1952), and ion exclusion operations
(Cooney, 1974).
Though many studies have been devoted to the onset and growth of flow instabilities
during displacement of liquid from inert packed beds,little attention has been paid to
systems where a concentration gradient instead of a discontinuity exists in the liquid
flowing through a porous medium, as is the case in for example solid-liquid extraction.
Also the effect of mass transfer between the porous packing and the liquid has not yet
been studied. Chapter 4 of this work is focussed on these phenomena which can be
important in solid-liquid extractors. Experiments were performed to check stability cri-
teria which should be met in order to prevent the onset of flow instabilities in these
systems.

1.3 Somegeneralremarksonthemodelsystemstudied

An effective study of the chemical engineering aspects of solid-liquid extraction is

only possible when the problem is reduced to its essential elements. For this reason, if
not stated otherwise, the following simplifications concerning the mass transfer process
were assumed to bevalid throughout this work.
— A linear equilibrium exists between the solid phase and the solvent.This assumption
implies that adsorption of soluble substances on the surface of the porous matrix accord-
ing to a non-linear equilibrium can be neglected. The validity of this approximation is
shown experimentally by Oplatka (1954), Yang&Brier(1958) andKrasuk et al.(1967).
— The diffusivity of the extracted species is constant. The mathematics involved in
solving the diffusion equation canbe significantly simplified when the masstransfer rates
can be described with a constant diffusion coefficient. In Table 3 some extraction pro-
cesses from the food industry aregiven,for which the extraction rate canbe described by
Ficks law with a constant effective diffusion coefficient. In some cases the diffusivity is
found to depend on concentration of the solute,particle size or time.Even in these cases
a proper analysis of the data might sometimes reveal that the extraction can still be
described by Ficks law with constant diffusion coefficient. For the extraction of flaked
soybean, Bernardini (1976) found that the extraction rate is essentially constant until
about 30%wt of the oil is transferred to the solvent. In a later stage of the process the
mass transfer rate isgoverned by molecular diffusion. Accordingto Bernardini, deviations
of the initial extraction rate amount to washing off the oil adhering to the surface of the
flakes. Brüniche Olsen (1962) observed that the diffusivity apparently varies with the
particle size for the extraction of sugar from sliced beet. Similar effects were observed for
the extraction of roast and ground coffee beans (Douwe Egberts Research, pers. com-
mun.). This effect has been explained by a change in particle shape and fraction of
disruptured cellsasaresult ofvariationsin particle size.
— Constant flow rates of both phases in the extractor. When the particles can be con-
sidered as an inert solid matrix with constant internal pore volume,this approximation is
valid exactly when the flow ratesare expressed on avolumetricbasis,e.g.m 3 /s.When the

10
Table 3. References that report mass transfer rates that canbedescribed with aconstant diffusion
coefficient.

Rawmaterial Extracted component Source

Tungseed Oil Krasuketal., 19751

Soybean Oil Beek&Muttzall, 1972
Peanuts OU Fan&Morris, 1948
Sugarbeet Sugar Brüniche Olsen, 1969
Pickles Salt Plugetal., 1967
Bomben et al., 1974
Pickles Sugar Eder, 1971
Greencoffee Caffeine Bichsel, 1976
Roast andground coffee beans Solubles DouweEgberts Research,pers.commun.
Solanumlaciniatum Solasodine Tettamantietal., 1975

1. At elevated temperatures only.

liquid densities are constant throughout the extractor it is alsovalid when the flow rates
and concentrations are expressed on weigth basis.The results arenot generally applicable
when the particles shrink or swell considerably during extraction.

11
2 Effect of local phase contact on extraction efficiency

Large scale solid-liquid extractors are commonly operated in countercurrence as it is

well understood that this design reduces solvent requirements. To attain pure counter-
currence between solid granules or powders and a liquid phase usually involves consider-
able mechanical problems. For this reason one often resorts to cascading mass transfer
sectionsin an 'overall' countercurrent fashion, while the localphase contact within amass
transfer section deviates from the overall counterflow. Aclassification of countercurrent
separation cascades relevant to solid liquid extraction is given inTable 4. In Section2.1.
this classification is discussed in some detail and available literature dealing with the
design of such mass transfer cascades is reviewed. In the subsequent sections, the effi-
ciency of countercurrently cascaded fixed beds and cross flow sections iscalculated from
mathematical models.The separation efficiency of these systemsisexpressed asarelation
between the number of exterior apparent transfer units, defined in Section 1.2, and the
number of true transfer units:

v
od O.V ]_
^ t . d / A= (2.1)
4>d A

The results are presented in concise correlations covering the range of normal operating
conditions.

Table4. Classification of countercurrently cascadedmasstransfer sections.

Localphase contact References covering theory

within amass transfer Examples on design ofmass transfer
stage cascades

distributed lumped
parameter parameter
model model

countercurrent U-typeandJ-type extractor seeTable 7 Colburn, 1939

cocurrent Multiplebatch extractors seeTable7
crosscurrent Rotary and belt extractors Thibodeaux, 1977
unsteady column Diffusion battery Svedberg, Chen, 1972
1976

12
2.1 Classification ofsolid-liquidextractors

The classification of extraction systems proposed in Table 4 does not necessarilly

coincide with those presented elsewhere,see Chapter 1.However,most of the percolation
extractors will be found under local cross current contact, and most of the immersion
extractors in the other groups. It should be kept in mind that, where asother classifica-
tions give more direct information on the raw materials that can be processed in the
apparatus, this classification isrelated to the stageefficiencies that canbe expected in the
extractor.
Countercurrently cascaded mass transfer sections with localphase contact in counter-
current obviously imply acountercurrent extractor. In practice thismodel can beused to
describe some commercial counterflow extractors approximately, such as the DDS dif-
fusor, the NIRO and Olier extractor. Moreover purely countercurrent contact limits the
separation efficiency of any extraction system.Masstransfer in countercurrent extractors
is undoubtedly analysed in most detail in the literature. When plugflow of both phases
occurs, the analytical solution of the distributed parameter model is given by Mikhailov
(1977). Tables of analytical solutions of the diffusion equations when mass transfer
rate is limited by internal diffusion have been given by Tettamanti (1975). Plachco
(1970) discussed the effect of an initial concentration gradient inside a particle on the
extraction efficiency for particles with slab geometry. The lumped parameter modelwas
first solved by Colburn (1939). This latter model has been extended to deviations from
pure plug flow in the phases.Exact solutions obtained by Miyauchi &Vermeulen (1963),
Sleicher (1959) and Hartland &Mecklenburg (1966) using the model of plug flow with
superimposed axial dispersion, and those derived by Sleicher (1960) and Mecklenburg&
Hartland (1968) using the model of backmixing between complete mixed stages are not
convenient for design calculations. Approximate solutions aregivenbyWatson&Cochran
(1971), Stemerding &Zuiderweg(1963),Pratt (1975),Tolic(1973) andby Pratt (1976).
Kerkhof &Thyssen (1974) used a dual-cascademodel,wherethe number ofideal mixers
inboth phasescan be different.
Countercurrent cascades of mass transfer sections with local phase contact in cocur-
rence aremathematically similarto multiple batch countercurrent extractors. Thisextrac-
tion system is mainly used for small scale extraction of for example pharmaceutical
products and spices. The operation of the extractor is described by Treyball (1968).
Analytical solutions of the diffusion equation with the proper initial and boundary condi-
tions have been numerically evaluated to obtain the efficiency of these cascades by
Plachco & Lago (1972). Similar resultswere obtained by Schwarzberg (1977) whouseda
different calculation procedure. Inboth cases,it wasassumed that geometry was spherical
and that internal diffusion limits mass transfer rate.Whenliquid clingsto the particlesso
that the phases are not completely separated when stage transfer occurs, the analytical
solution ofPlachco&Lago(1975) canbe applied.
Countercurrent cascades of mass transfer sections with local phase contact in cross
flow describes extraction systems that are in use for large scale extraction of oil seeds,
sugarcane etc. The operating principle of this extractor is discussed in Chapter 1. In the
literature little attention has been given to the mathematical modelling of these systems.
For pure cross flow cascades where no entrainment of one of the phases by the other
occurs, Thibodeaux (1977) recently proposed a design procedure. He assumed lumped

13
mass transfer resistance and complete mixing of the streams before they enter a new
stage. Pure cross flow is encountered in cooling orheating of granular products (McGaw,
1976), cross-flow cooling towers (Thibodeaux, 1969) and heat exchangers (Gardner &
Taborek, 1977), and is a good approximation for belt type solid-liquid extractors when
the value of the distribution ratio D, defined in Eqn 2.9 is low. In his simulation model
for the extraction of sugar from sugarcane,Rein(1976)accounted forentrainment ofsugar
juice by the moving solids. In his calculation he neglected concentration gradients over
the height of the bed, perpendicular to the flow direction of the solids. No systematic
analysis of the efficiency of cross-flow sections in countercurrent cascadesisavailable in
the current literature.
Countercurrent cascades of unsteady operated columns are often referred to as mul-
tiple column countercurrent extractors, diffusion batteries or Shanks extractors. Details
of the operation principle of the extractor are given in Chapter 1. Multiple column
countercurrent contactors are in use for ion exchange and adsorption processes.Most of
the literature dealing with the design of this column arrangement isrelated to adsorption
on active carbon from the liquid phase. The liquid residence timeinasingle periodically
operated column was compared with a countercurrent adsorber by Neretnieks (1975).
For alinear sorption isotherm,the residence timein the periodic column wasat least70%
longer for the same efficiency of carbon utilization. Fornwalt & Hutchins (1966) pro-
posed agraphicalprocedure to determine the numberofcolumnsinamulticolumnsystem.
They assumed that all the columnshave similarly shaped breakthrough curves,sothat the
apparatus is overdesigned. Chen (1972) studied adsorption frorh the liquid phase on
activated carbon in a multiple countercurrent contactor using the model of acascade of
perfectly mixed vessels to calculate the performance of the system.An empirical relation
for the equilibrium relation and the rate ofmasstransfer wasused, sothat it is difficult to
generalize his results. Svedberg (1976) thoroughly studied the efficiency of multiple
column adsorption from the liquid phase. Masstransfer rate wascalculated from amodel
that takes into account film diffusion andpore diffusion with aconstant diffusion coeffi-
cient. Plug flow of liquid through the column and a linear equilibrium relation were
assumed. He concluded that when four or more columns are used, carbon utlization is
wihin 10%of the carbon utilization in a purely countercurrent contactor, under normal
operating conditions. For two columns, this figure was about 15-30% lower than in a
truly countercurrent apparatus. Klaus et al. (1977) calculated carbon utilization for the
adsorption of binary mixtures on active carbon. They found that the transient phase
before the concentration profiles stabilize is much longer than in single component ad-
sorption,where aperiodic steady state isattained in one ortwocycles.

2.2 Diffusion battery

In the following section a mathematical model of a single-stage fixed bed is derived.

Subsequently it is shown how this basic model can be used as a building block in the
mathematical simulation of a diffusion battery. The effect of process variables on the
separation performance of the multicolumn countercurrent extractor is then calculated.

14
2.2.1 Mathematicalmodel of asingle-stagefixed bed

From Figure 8 it can be derived that mass transfer in afixed bed canbe described by
Eqns 2.2 and 2.3 when a linear driving force for mass transfer is used and plug flow of
liquid through the bed is assumed:

ALh^=kocaAL(c'*-c')-<t>c^I (2.2)

AL (1 -h)~- =- kocaAL (c'*- c') (2.3)

withinitial andboundary conditions

f=0 u' = <o'0(x) c' = c'0(x) (2.4)

x =0 c'=c[n(t) (2.5)

With a linear relationship between the equilibrium concentrations in both phases,c'* =

m CJ',the equations can bemade dimensionlessby introducing the following variables:

c -c0
liquid phase concentration (2.6)
muQ - c o
m<J-
solidsphase concentration (2.7)
— mu>.0 ~co

^oc aV
number of true transfer units on overallliquid phase basis (2.8)

Fig.8. Fixed bed extraction.

15
 D=j]—TT distribution ratio (2.9)

x
?—T dimensionless distance from liquid inlet (2.10)

t<t>c
dimensionless time (2.11)
Vh

Withthesevariables the differential equations with initial andboundary conditions canbe

written as:

9c be

d9+~d$=Nt'c("~c) (2 12)
-

30" =-NUcD(u-c) (2.13)

0=0 co=Wo(r) c=cS) (2.14)

f=0 c=cin(0) (2.15)
When at time 6 =0 the concentration in the solid and the liquid phase arein equilibrium
and the inlet concentration c i n is constant, the solution of the above equations is well
known (Anzelius, 1926):

s
c = e-eMf e- J0(2iVsd^)ds (2.16)
o
where

ÖM=^t,cfl(Ö-l) (2.17)

The above relation is tabulated by Furnas (1930).Approximate solutions aregivenin the

survey of Klinkenberg (1954) and in the paper of Roetzel &Nicole (1975).For arbitrary
initial concentration profiles in both phases andtime variable concentration at the liquid
inlet, Klinkenberg & Harmens (1960) gave an unavoidably complicated analytical solu-
tion. Asan alternative to this solution,the equations canbe solvednumerically.Detailsof
the computational scheme that wasused here aregiveninAppendix A.

2.2.2 Mathematicalmodel of adiffusion battery

In order to simulate a diffusion battery, a number of columns can be put in seriesas

shown in Fig. 9. Another approach that is somewhat easier toincorporate in acomputer
program, is to consider the diffusion battery as a single column in which the solids are
transported intermittently. Since we are interested in the concentration profiles in a
cyclic 'steady' state and the profiles at the beginning of a cycle are not known before-

16
 g=1/n

x=0
} ll-ôj
Fig.9. Diffusion battery withncolumns.

hand, theequations aresolved numerically. Starting from an arbitrary initial concentra-

tion profile, theunsteady state start-up ofthe system ofcolumnsissimulated. Atthe end
of every cycle the concentration profiles areshifted onecolumn andtheconcentrations
in both phases in thelast column, i.e.where thesolids arefedtotheextractor, arereset
according to the boundary conditions. This operation cycle isrepeated until theconcen-
tration profiles in the extractor andtheexit concentration ofthe columns averaged over
one cycle time no longer change significantly. In most calculations the concentration
profiles stabilized in n cycles. Toallow an accurate evaluation of7V*ext,the calculation
must be continued upto a dimensionless time 0 between 6 and 10.Thecomputationis
then halted andthemassbalance checked,which closed within 0.5%inmost cases.
In analogytothe extraction factor incontinuous countercurrent processes,we defined
an extraction factor or 'draw-off factor' for semi-continuousmultiple column extraction
as theratio of theliquid flow drawn from theextractor totheflow ofsolids throughthe
extractor, multiplied by the slope of the equilibrium curve. Theflow of solids is deter-
mined by thevolume ofsolidsinone column andthecycle timetc only.Bearinginmind
that the water filling the interstitial voids in the first column is discharged with the
exhausted solids,we can calculate thedraw-off factor, A from:

Vh 1

A=m ••D(n0c-l) (2.18)

V(l-h) 1

The above equation is used to determine the dimensionless cycle time 6Crequired to
attain achosen draw-off factorA.

2.2.3 Efficiency of adiffusion battery

The results of the calculations are shown inFig. 10.Theplots show an almost linear
relation between Nt d /A andNext. Increasing thenumber ofcolumns resultsinasmaller
difference between Nt d /A and Next, showing that the extractor approaches a truly

17
 Fig. 10. Number of exterior apparent
against true transfer unitsfor a diffusion
battery withn columns.Dis distribution
ratio;Aisextraction factor.

countercurrent system.Anincrease in A,the draw-off factor, showsthesametendency,

whichisdue to amore favourable ratio ofdraw-off andcycletimeathighervaluesofA.
Sincethe time required to fill theinterstitialvoidsofthefreshly chargedcolumnis equal
totf=(hV/n)/(j>c, onecanwriteforthisratiowithEqns2.9and2.18:

et 1/« D
(2.19)
0. (1+A/D)/n D+A

18
The effect ofD canbe made clearer by writing the expression for D as follows:

Vh
m
mh= _r% = _A^_.A (2 . 20)
\-h V(l-h) 4>c-<i>c,A

D/A can thus be interpreted as the ratio of the liquid flow entrained by the discharged
solids, <[>c A and the net liquid flow drawn from the extractor. Since the solubles in the
liquid in the first column will be lost with the spent solids at the end of the cycle one
expects a higher efficiency for alowervalue ofD, whichisconfirmed by the calculations.
The results of the calculations are given in aconcise form by Eqn 2.21,which covers the
whole range of input parameters used in the simulation runs.

^ — =1-0.59 n-1-11 A'0-3 N°ex\5A'n D0-33 (2.21)

•^t.d/A

2 < n <6 0.25 < A< 2 0.25 < D < 4

0.25<W e x t <10 T ? < 0 . 9 9

Within the indicated region this correlation gives an estimated value of(Nt d /A - Next)/
Next, within 5%of the numerically calculated value.

2.3 Belttype extractor

In this section a mathematical model of asingle stage cross-flow device is derived.The

analogy to an unsteady operated column is discussed. It is shown that the algorithm
proposed in Section 2.2.1 can be used with some modifications for the mathematical
simulation of a belt type extractor. For extractors without liquid recirculation, the effi-
ciency has been obtained from numerical calculations. The efficiency of extractors with
liquid recirculation iscalculated from analytical solutions of approximate models.

2.3.1 Mathematicalmodel of asingle-stagecross-flowdevice

«
From Fig. 11, the folic ing equations can be derived for cross-current extraction
under steady state conditions:
v Sh
<t>c be' b be' , ,
-HTx + — * = * o . « * ( ' " - O (2-22)

vhS(l-h) 3co'
1 fy =- Kc<M<P*' - c) (2.23)

withboundary conditions:

19
 *c c" in (y)

B
x=0 J&i.
-X
C
Ó(X) e
ü)'o(x) I*'*
y*dy
•4
Vb
x*L
y=H y«0
Fig. 11. Crosscurrent phase contact.

j =0 co'=coó(x) c'=c'0 (x) (2.24)

x =0 c'=4(y) (2.25)

The equations are transformed into a dimensionless form by introducing the following
variablein addition to Eqns 2.6-2.10:

(yftb)0c
0= dimensionless residence time of solidsin the extractor (2.26)
~Vh

The equations can thus be rearranged to:

aë + ä F = J v t , c ( « - ^ ) (2.27)

9co
30 =-NUcD(u-c) (2.28)

with boundary conditions:

0 =0 co=co0(?) c=c 0 (r) (2.29)

r=0 c=cin(@) (2.30)

The above equations are similar to those given in Section 2.2.1. For pure cross-current
contact in single stage devices,thisanalogyiswellunderstood (Nusselt, 1930).It involves
that the equations describing mass transfer inunsteady operated columns canbe reduced
to those describing pure cross-flow contactors, by transforming the Eulerian system to
Lagrangian coordinates. When liquid is displaced in the direction of the flow ossolids,as
in solid-liquid extraction, the analogyappliesin Eulerian coordinates.Thisanalogy willbe
extended here to multiple fixed bedsinseriesin order to simulate masstransfer efficiency
in abelt type extractor.

20
 sparger
•c.ci.

vbAh x-0
y,0 "
.vb
v„A<irKr
'"out y-tT / y-ri/!
'•H/n y»0 Win
tZ? 'L *-cout
collection pans belt
Fig.12. Beltextractorwithoutdrippingsections.

2.3.2 Mathematicalmode!of abelt type extractor without liquidrecirculation

A sketch of the extractor considered in this section is given in Figure 12. The mean
concentration of liquid leaving section /can be calculated by averagingthe concentration
of liquid dripping form the belt into the collection pan below, thus at x =L or % = 1,
fromy =0toy -H/n, orfrom 0 =0 to:

0c
©=0 = (2.31)
c
Svhhn
0 C , the dimensionless residence time of the solids in one section of the extractor, is
related to the draw-off factor A, defined analogously to the extraction factor in counter-
current extractors as:

<PC - vbSh
A =m (2.32)
vhS(l-h)

Combining Eqns2.31 and 2.32gives:

0 1+
c=( e)/" (2.33)

Fig. 13 gives the calculation analogue of the belt extractor from Fig. 12. The liquid
leaving section /'is collected in chamber Ai and mixed up. After one 'cycle timV 0 C the
liquid from A{ is transferred to B{. The liquid entering section (/+1) iswithdrawn from
the latter. The calculation proceeds as follows. Since the concentration profiles in steady
state are unknown apriori,Eqns 2.27-2.30 areintegrated startingfrom anarbitraryinitial
concentration profile in both phases. After one cycle time 0 C , the average concentration
leaving column i is determined, and the liquid concentration in Bi is set equal to this
value. Concentration profiles are shifted one column. The concentration in the last col-
umn downstream is reset according to the boundary conditions. After alarge number of
cycles has been worked through, the average liquid concentration leaving each section
reaches a constant value. The calculation is then halted and the massbalance is checked.
It should be realized that this calculation method does not give any information on the
unsteady operation of the extractor since we started from the steady equations 2.22 and
2.23.

21
 V}-^- V c
-*-cout
•c
Ci(6) Cj*1 (9) out (9)

*d Ci=/c|(9)d9
M
O

Fig. 13. Analogueof belt extractor of Fig. 12.

The mathematical simulation of a belt extractor without dripping sections has just
been discussed. However, in most industrial extractors dripping sections are present. In
the following the mathematical model willbe extended to take these into account.
A rigorous simulation of this case is rather complicated, because both drainage rate
(Beek & Muttzall, 1972) and static holdup (Dombrowski, 1954) are dependent on the
physical properties of the liquid and the bed geometry. Therefore the problem issimpli-
fied considerably here by assuming aconstant drainage rate,equal to the percolation rate.
It is further assumed that no liquid remains in the interstitial voids of the bed after
drainage. This simplified situation can be simulated by shifting the chambers collecting
the percolating liquid over adistanceyf asindicated in Fig. 14.This distance isrelated to
the time required to fill the interstitial voids in the fresh solids entering section n. In the
time required for the liquid front to reach the lower end of the packing, the belt has
moved over alengthy f

hL
y{: (2.34)
4>J{BH/n)

orin dimensionless form:

g
l c in Ai-2

/;
•
i ~?l
/
0
/
/ 1 / M / ' / n /
/ L / i
A 0f 0 - 0 y= H/n 4 y=0
.u-AM^Ü
M-1 Cout
Fig. 14. Belt extractor with drippingsections.

22
 9f=- (2.35)

Since noliquid isentrained bythe solids leaving theextractor, theextraction factorcan

now be calculated from:

A=m (236
^scr^y >
WithEqn2.9thedimensionless residence time ofthe solidsinone section is obtained:

&
o = Wn <237>

The calculation proceeds as discussed in the previous section. The average concentration
ofliquid collected inchamberA- iscalculated from:

<U,=\ ƒ Cout,,_! d0 + ƒ c o u t > , d 0 i / 0 C (2.38)

Note that anextra column 0isusedinthecalculation toaccount formasstransfer during

drainage. Since only a part of this column contributes to the extractor volume, the
number of true transfer units used in the simulation Nt s i m , is corrected: on inspecting
Fig. 14itisfound that:

«0 C + 0 f

2.3.3 Efficiency of belt type extractors without liquidrecirculation

The results of the simulation of a belt extractor without recirculation of liquidand

without dripping zones arepresented in Fig. 15.Itshowsthan anincreaseinthenumber
of sections results in a smaller difference between NexX andNt d /A, showing that the
extractor approaches a purely countercurrent contactor. Forahigher value ofD, more
liquid is entrained by the solids on the belt to the upstream sections. This backmixing
flow 0C A isrelated tothedistribution ratiointhefollowing way:

D=
n u? C = : T - ^ A (2.40)

Solubles contained in the liquid flow <j>c A attheliquid feed endofthe extractor willbe
lost with the spent solids. Compared with a diffusion battery discussed previously,the
belt extractor gives alower efficiency duetomixingofliquid inthecollection pans after
percolation through the bed.In Fig.13 this mixing isillustrated in the analogue where
the liquid leaving thecolumn i (i.e. section of the belt) ismixed inthevesselA{ (i.e. the
collection pan).Whentheextraction factor ishighorthenumber oftransfer unitsineach

23
 Fig. 15. Number of exterior apparent
against true transfer units for a belt ex-
tractor with n sections without dripping
zones. D is distribution ratio; A is ex-
traction factor.

section is low, the concentration of liquid entering a collection pan does not vary
significantly during a cycle. Then mixing hardly influences the extractor performance
which thus approaches the performance of adiffusion battery.
For higher values ofNt d /«A the adverse effect of mixing becomesmoreimportant, as
is clearly shown in Figs. 15 and 16. A correlation was fitted to the data so that the
results can be easily applied. Within the indicated region the estimated value of (Nxä -
ANext)jNt d iswithin 5%of the numerically calculated value.

24
 Fig. 16. Number of exterior apparent
against true transfer units for'a belt ex-
•n tractor with n sections with dripping
•"" zones.Disdistribution ratio;Aisextrac-,
^td/ tion factor.

1 + 1.035i t " 1 - « A " 0 - 7 1 D06S N°ex\4 (2.41)

ext

2 < n < 7 0.3<A<2 0.25<D<2

0.3<JV ext <10 i?<0.99

If dripping zones arepresent, no liquid isentrained to the upstream section.The effect of

the variable D/A is thus less pronounced than in the foregoing case.Thevalue of(Nt d -

25
ANext)/Nt d can be estimated with 5%accuracy from Fig. 16orEqn 2.42:

^ ^ = l + 1.00«- 1 - 4 4 A- 0 - 6 3 £>°- 4 S 7V° x 4 t 2 (2.42)

• /v ext

2 < « < 6 0.3<A<2 0.25<Z»<1.5

0.3<7V ext <10 TJ<0.99 A/£>>1

The above equation isvalid only forA/D> 1. Since from Eqns 2.35 and2.37:
0
c A/Dn A
(2.43)
0f Ijn D

the residence time in a section thus hasto belonger than the time required to fill the
interstitial voidsofthe packed sectioninone stage.

2.3.4 Efficiency of belt type extractors with liquidrecirculation

Although the belt extractor with recirculation can be simulated analogously tothe
procedures we followed for the extractor without recirculation, it often suffices to
apply ananalytical approach forthesimplified caseofacompletely mixed liquid phasein
each section. Under what conditions this approximation isvalidisdiscussed inAppendix
B. Wefirst note that the local mode of phase contact ineach stage isimmaterial because
the liquid phase is completely mixed. Theanalytical expressions derived in this section
can thus be used to estimate the efficiency ofthe other flow arrangements mentionedin
Table4aswell,aslong asauniform liquid phase composition ineach section prevails.
Hartland &Mecklenburg (1966) derived an analytical solution forwhen thesolidsare
completely separated from theinterstitial liquid before they leave theextractor. Inthe
present notation their relation reads:

hi-Bi h2 - B\
. (Bi-l)B"i (* 2 -1)2*3 (2.44)
"'"" AxO-Afi?) h2 (1-A53)
(fij-l)fi? (Ba-l)fiS'

whereB. isdefined inEqn 2.58 andh-s isgivenby:

^ =1 - ^ ( ^ - 1 ) (2.45)

In thepresent analysis the entrainment of liquid bythesolidsleavingthe extractor is set

equal to the interstage liquid entrainment. This assumption isbelieved tobe physically
more realistic andthe results of this model caneasily becompared with thecorrelations
derived earlier. Forthe derivation of the equations describing themodel we refer to Fig.
17.

26
•ri Wj U)'K1 _*5L- U)in
* ( )
<;A Ci <=i+1
. i
c
i-1 c
i
*
C
OUt

Fig. 17. Belt extractor with complete mixedliquid phaseineachstage.

In this simplified model of the belt extractor, plug flow of the solids and complete
mixing of the liquid within each stage is assumed. Then a mass balance on the solids in
slice dx resultsin:

dwj dojj
v 5 (1 h) (2.46)
» - ST ^ "d7 - *oc«S(*i*- cù
= =

which can be written in adimensionless form by introducing Eqns 2.6—2.10.

dcoj Ntt,d
(2.47)
IT ( « i - ct)

The number of true transfer units on solid phase basisisdefined as:

k
ozaV Ki^v
NtA
t , d= (2.48)
*d/« ^

Equation 2.47 can beintegrated togive:

— =exp(-JV t i d //i) (2.49)

The mass balance on stage /, assuming complete entrainment of the interstitial liquid by
the solids to the next section,gives:

0d(«/+1 - «/) + 0c,A<p'i+1 - c ' i ) + 0c(*/-1 - c ' i ) = ° (2.50)

where the backflow <j>c A is defined as:

0c,A = V b 5 A (2.51)

Equation 2.50 can be written in adimensionless form usingEqn 2.9 and:

27
 A=m ( 0 c - 0 C ; A ) / 0 d (2.52)

Finally the following equation results:

( u l + !- co,.)+D(ci+1- 2cf + c,_,) + A(c,._, - c,.)=0 (2.53)

For the last section the balance has to bemodified sincenoliquid iscarried by the solids
enteringthis stage,so:

("in - « « ) + £>(c„_!- c n ) + A(c„_j- c „ ) =0 (2.54)

Combining Eqns 2.53 and 2.54 results in the following boundary condition at the solids
inlet end of the extractor:

cn+ l=cn (2.55)

The solution of Eqns2.49, 2.53 and 2.55 with the additional boundary conditionsc j n = 0
and w i n = 1 yields:

= ( g | -1)/|*?(2? -1)1 - (B\ - \)l[Bn2(ß2 - 1)]

(2.56)
Ct
(ß? +1-gi)l\giB? ( 5 , -l)]-(B2+1-g2)l\g2Bn2(B2 - 1)]
where:

g. =l„a(B.-l)=l-(Bi-l)l[exV(Ntdln)-l] (2.57)

andB- are the roots of the equation:

1 + ( A + D)a+ Z)(l+«) ( l + a ) g ) + A)
i
' Da Da

The number of exterior apparent transfer units canbe calculated from thisusing Eqn 1.7.
The results of the simplified model for a belt extractor with liquid recirculation in
each section are shown in Fig. 18.The resultsare presented asgraphs ofNextA/n plotted
against N ( Jn, the number of true transfer units in each section on overall solid phase
basis in pure countercurrent flow and in the extractor under consideration, respectively.
For somelimitingcasesof themodel discussedhere,asimplified solution canbe given:
— No liquid entrainment to the upstream sections. Since aconstant liquid concentration
in each stage was assumed, the distribution ratio Dwill only affect liquid entrainment as
shown in Eqn 2.40. This case can thus be solved by taking D = 0 in the foregoing
equations. The resulting equation is:

This equation is shown in Fig. 18c. Equation 2.59 can also be derived from the solution

28
Next *ln
D=2
2 sections
10 sections

Fig. 18. Number of exterior apparent against

true transfer units for a belt extractor with
complete mixed liquid phase in each section,n
is the number of sections;D is distribution ra-
tio;Aisextraction factor.

29
of the cascade of mixers model proposed by Kerkhof & Thijssen (1974), by taking the
limit for n going to infinity in their equation (13a), since the solidsmoveinplug flow.
— For a large number oftrue transfer units each section canbeconsidered asanequilib-
rium stage.Then theresulting equation forNext is:

A^ext A /D+I\ ,„^N

=T-7 MT;—H (2-6°)
J
n 1-A \D +Aj
From the reasoning given above under case 1,itcanbeconcluded that theresultsofthis
model can be used to estimate the number of true transfer units when dripping sections
arepresent. Heretheparameter!) asdefined inEqn2.9hastobereplacedby
v
„ A bSe me
(2
0 =A -7 Ö-=T— -70)

where eisthe residual liquid hold-up after partial drainage.One should bearinmind that
the extractor volume calculated from Nt d when dripping sections are present does not
includethevolume ofthe dripping sectionsitself.

30
3 Estimation of mass transfer coefficients insolid-liquid extractors

3.1 Introduction

In this work the NTU approach has been adopted for the design of solid liquid
extractors. In the preceding chapter it has been shown how the number of true transfer
units required to meet the design specifications canbe determined. Thepresent chapter
deals with the estimation of masstransfer coefficients. Bycombining these results,the
required residence time of the solids in the extractor under givenprocessconditionscan
becalculated from:

Td= ni-A)/*d=*t,d/(*od«') (3-D

wherekod isthemasstransfer coefficient onoveralldispersedphasebasis.Obviously,the

use of masstransfer coefficients to describe transfer rates ismostconvenientwhenkais
independent of time, position, concentration and drivingforce for masstransfer. These
conditions areonlyfulfilled approximatelyinsolid-liquidextractors.Thetwofilmtheory
of Lewis states that k0 is dependent on the mass transfer coefficient in the continuous
phasekc, and in the dispersed phasekd. Thisrelation hasbeen derivedfrom theexpres-
sionsforthefluxthroughasolid-liquidinterface (seeFig.19):

bulk c oncentr tition

-solld
liquid bounckary
layer

—_^«£_ Ul'

L\
J C|
i

c
y ^ solid
liquid
interface

Fig. 19. Concentration profiles near a solidli-

quid interface.

31
 =* c (c;-c')=M«'-«î) (3.2)

Foralinearequilibriumrelationc'*=mco',weobtainthewell-knownequation:
1_ 1 _ 1 1
(3.3)
od "oc d
whichcanbewritteninadimensionlessformas

1 1 1 (3.4)
Sfcod Sh + 25/
where Sh=kDp/BD,k and 2D refer to the samephase;Bi=m(DpIZ)k0IIDd. Aswillbe
recalled in Section 3.2,Shc isdependent onthephysicalpropertiesandtheflowvelocity
of the liquid surrounding the particle. Due to the masstransfer processthesephysical
properties change with time, while in gravity percolation extractors the flow velocity
might change accordingly. Moreover, evenif the physicalproperties areuniform andthe
velocity is constant, kc varies over the surface of the particle (Gillespie et al., 1968;
Frössling, 1938).Fortunately, however,theBinumber which embodiesthe ratio of the
masstransfer resistances inside and outsidetheparticleisoften ratherhighinsolid-liquid
extraction processes so that variations of kc have a small effect on the overall mass
transfer coefficient. Even for moderateBinumberstheuseofavalueofkc averagedover
the particle surface and the time that the solids reside in the extractor willonlyslightly
affect the accuracy of the designcalculation.ThedispersedphaseSherwoodnumberShd
is time dependent due to the transient nature of the diffusion proces. The initial
decrease of Shis described by the penetration theory (Higbie, 1935).In alater stage of
the extraction, theShd against time curvelevels off and approaches afiniteasymptotic
valuewhenenteringthe so-called 'regular regime'(Kondratiev, 1964),Fig.20.Thevalue
of ShA in this regular regime depends on the boundary conditions or concentration
history at the surface of the particle, and on the geometry of the particle; but it is
independent of the initial concentration distribution inside the particle. It can be ex-
pected that for high extraction efficiencies a considerable part of the process can be
describedwiththisasymptoticmasstransfer coefficient.

Fig. 20. Transient Shd numbers for unsteady diffusional

mass transfer. (...) penetration period; ( ) regular
regime.

32
 In Sections 3.2 and 3.3,literature data on masstransfer coefficients in continuous and
dispersed phases are reviewed. This information is thereafter extended to provide acom-
plete picture of asymptotic ShA values in cocurrent and countercurrent extractors for
particles with simple geometry. In Section 3.3.4,the range of applicability of asymptotic
Sh numbers in mass transfer calculationsisdiscussed.Moreover it isshown in this section
how one could proceed in more complicated cases that are not covered by the previous
sections.

3.2 Estimatingmasstransfercoefficients inthecontinuousphase

Mass transfer coefficients in the continuous phase are commonly expressed by the
dimensionlessiS7z„cnumber, defined as:

SK = - ^ (3-5)

Dp denotes a characteristic dimension of a particle or a system that containsn particles,

normally the equivalent diameter \/ÄJrvn where Ad is the surface area of the particles.
The external Shc number is determined by the concentration distribution in the con-
tinuousphase.The equality(Fig. 19):

= ftc(cR-c) (3.6)
R.

yieldsthe following relation for Shc:

-3c/ar|
Shc=Dp - H (3.7)
cR-c

For a stagnant fluidum surrounding asinglesphericalparticle the asymptotic Shc value is

easily obtained from the solution of the diffusion equation:

with boundary conditions:

r=R c=cR (3.9)

r=R0 C=CRQ (3.10)

The steady state solution for transfer from a particle submerged in astagnant, infinitely
extending fluidum iswellknown;when bc/bt = 0andR0 -*• °°:

Shc=Dp *-=2 (3.11)

CR -COO

33
Table 5. Effect of distance between two sphericalparticleswith
diameter Dp on Shc in an infinite stagnant medium. Data from
Cornish 1965.

Distancebetween centers of spheres Shc

Dp 1.39
2D p 1.60
50Dp 1.98
2.00

When two particles are placed in an infinite stagnant medium,the external mass transfer
coefficient decreases with decreasing distance between the particles, see Table 5. For
multiparticle systems in astagnant liquid Miyauchi (1971) calculated the value of Shc
from the concentric sphere model. Inthis model the particle issupposed tobe sur-
rounded by aliquid shell; the dimensions of this shell are adjusted to match the void
fraction of the packing.Miyauchi assumed aconcentric shellrangingfrom r=R tor= R0
around aspherical particle. When the liquid concentration at the particle surface is set
equal tocR and the concentration atthe outerboundary of theliquid shellequalto cR
the result for <S7zc inthe steady stateis:
-dc/br
S
K=Dp- - ^ IR= (3.12)
C l
R -°R0 -(.K/K0)

Suzuki(1975) solved Eqn 3.8 with amore realisticboundary condition atr=R0 :

r=R0 (oc/ar)=0 (3.13)

thus avoiding the artificial sink term atr=R0. Thetransient solution isused to calculate
theShc value,defined in Eqn 3.7,where

c=3ƒ cr2dr/(Rl-R3) (3.14)

R
Thistime-dependentShc number approaches an asymptotic value for longcontact times r.
This asymptotic solution is shown in Fig. 21,where also somemeasuredvalues reported
inliterature are included.
For small .Re-numbers, correlations for the external Shc number for single particles
have been derived from Stokes' stream function for creeping flow around particles.Fora
thick concentration boundary layer or small Re Sc numbers, the resulting equations are
often presented assummation seriesof the form:

Shc =2 + faeSc) +$(ReSc)2 +... (3.15)

For thin boundary layers, the relation obtained after summingup the result of the above

34
Shc,a

50 : v

10
n \
5
^
Fig. 21. Asymptotic values of Shc for
spherical particles in a stagnant finite
fluidum as a function of the void fraction
1 _ _i i . i i i i i i i
h. Key: o Miyauchi et al., 1976b;AMiya-
0.2 0.4 0.6 0.8 1.0 uchi et al., 1975;v Miyauchietal., 1976a;
h. I Suzuki, 1975;IIMiyauchi,1971.

analysisandthelimitingvalue2foraquiescentfluidreads:

Shc =2+ b(ReScyi3 (3.16)

where b rangesfrom 0.5 to 1according to Sideman &Shabtai(1964).The fundamental

problems associated with the interpretation of measured masstransfer rates at verylow
Re numbers inmultiparticle systems are discussedbyGangwal(1977).Thisdataanalysis
iscomplicated by several mechanisms that mightoverridetheeffect ofmasstransfer rate
on themeasurements,e.g. dispersion effects (Wakao, 1976),channellingcausedbyvaria-
tion in the void fraction (Martin, 1978) and free convection (Mandelbaum &Böhm,
1973). Also the often assumed symmetry of the concentration or temperature profiles
inside the particles isnot valid in the lowÄe-numberregime,(Wakao,1976).Inorderto
provide a theoretical basis for mass transfer correlations in these systems, Pfeffer &
Happel (1964) solved the spherical shellmodel assumingcreepingflow around thepar-
ticle. Their results agree reasonably well with the measured data. Nelson &/Galloway
(1975) combined the solution obtained by Suzuki(1975) with Danckwerts' surface re-
newal theory. Thus, an analytical expression was obtained for Shc. After scalingthefr
solution with the Ranz Marshallcorrelationwhichappliesforsinglespheres,theyderived
a correlation that has been used to predict masstransfer coefficients inthelowRe-num-
ber regime. Measurements of mass transfer rates in this regime were re-evaluated by
Dwivedi&Upadhyay (1977), and byWakao&Funazkri (1978).FoiRe numbersbelow
10, the first authors proposed the following empirical correlation between the Colburn
/-factor andtheReynoldsnumberfor theliquidflow, Re = p<v>D /p(l —h):

hShc
him = 1.1068 Re-012 (Re>\0) (3.17)
1 3
Re (Sc) /

35
Forhighervalues ofRe anoverwhelming amount ofliterature data are available.Based on
Schlichtings boundary layer theory, correlations for singleparticles usually take the form:

Shc =2 + bRe1l2Sc1'3 (Re>l) (3.18)

From the survey of Sideman &Shabtai(1964)it isfound that thevalue ofb isusually in
the range 0.5 <b< 0.9. Correlations for external mass transfer coefficients in multi-
particle systems are reviewed (e.g. Barker (1975),Sideman (1966),Upadhyay &Tripathi
(1975), Dwivedi &Upadhyay (1977) andKumar et al.(1977)).Based on acritical exami-
nation of previousdata,Dwivediand Upadhyay proposed the following empirical correla-
tion for higherRe numbers:

hjm= 0.4548 Re-°A069 (Re>10) (3.19)

These authors also suggested a generalized equation that correlates data over a wide Re
number range:

0.765 , 0.365
h
U =—Re'—
0.82+
(10" 2 < J Re<15 10 3 ) (3.20)
Re 0 . 3 8 6
Thisrelation isshown in Fig.22together with someliterature data.
Even though no attempt has been made to be complete, the above survey clearly
shows that methods for estimating the external mass transfer coefficient for widely
varying conditions is extensively dealt with inliterature.Moreover,aswaspointed out in
Section 2.1, the Biot number Bi = m k^R/IDi for solid-liquid extraction is often very

Re=p<v>Dp/n
Fig. 22. Mass transfer factor in fixed beds. Solid line: Eqn 3.20; Goud: experi-
mental data.From Dwivedi&Upadhyay, 1977.

36
large. Thus arough guessofkwillprovetobesufficient inmost engineering calculations.
If kç variesover the length ofthe extractor duetochangesinviscosity orflow velocity,it
isrecommendable touseaconstant, averagedvalue ofkc inthedesign calculation.

3.3 Estimatingmasstransfer coefficients inthedispersedphase

Masstransfer inside rigid particlesisgenerally described byFickssecond law:

21)
9co
Mi^-Z) °-
where v,the geometrical shape factor, is 1,2 and 3 for aflat plate, cylinder and sphere,
respectively. In analogy totheShc number,themasstransfer coefficient inthedispersed,
solid phase isembodied in theShd number,whichcanbecalculatedfrom theconcentra-
tiondistributioninsidetheparticles.Anoverallmassbalanceyields:
_ 9co
=- V d § 0.22)
kdAd(u-u,) =-E)dAd —
R at
By substitution of rd = IDdt/R2, £=r/R, and v/2=Ad/Vd, the above relation canbe
rearrangedto:
-0«/a0| _ d ^ /d , d
Shd =_ 1-1 = (3.23)
1
CO —COI CO — c o l
U=i U=i
Since theconcentration profile inside theparticle, co (£, r d ), isdetermined bythe geom-
etry of the particle andtheboundary conditions imposed onEqn 3.21, we expect Shd to
be afunction ofthe following parameters:

Shd =f(V,r d , initial and boundary conditions) (3.24)

For long contact times between thephases,theShd number will approach anasymptotic
value Shd a , as pointed out in the introduction. This asymptotic value is essentially
/
independent ofthe initial concentration profile, thus:

Shd a =f(y,boundary conditions) (3.25)

Mathematically this stems from the fact that the solution of the diffusion equation,
which can be written as a summation of terms containing aweighting function exp(—
X n r d ) multiplied bythe related eigenfunction f(Xn), isgoverned by theterm containing
the smallest eigenvalue Xx. Thecontribution ofhigher terms isnegligible because ofthe
nature of the weighting function for higher values of r d . Many authors proposed this
'first-term approximation' for heat and mass transfer calculations (Pflug & Blaisdell,
1963; Brüniche Olsen, 1962; andothers). It canbe shown that the extraction rateinthis
regular regime canbedescribed withaconstant masstransfer coefficient, kd.

37
3.3.1 Survey of literaturedata

For aconstant surface concentration, Shdi3 valuesaregivenbyBeek&Muttzall(1975).

For a constant bulk liquid concentration withfiniteBi values,these data are reported by
Thijssen (1969). The solutions for constant surface concentration were extended to con-
centration-dependent diffusivities and shrinking systems by Schoeber (1976) (Fig. 23).
The results of the analysismentioned sofar are summarized in Table 6.For cocurrent and
countercurrent extraction of heating/cooling of spherical particles,severalauthors derived
the asymptotic transfer coefficients when he phases are contacted in pure plugflow.As-
suming ageometrical similar profile inthe regular regime:

CO
=f(0 (3.26)
CO

where f(£) is a function of % = r/R only, Gardner (1966) obtained analytical expressions
for Shd a by substituting the regular concentration profile Eqn 3.26 in the diffusion
equation 3.21and combining the resulting relation with an overallmassbalance:

(w0-<o)=-(cout-c)At (3.27)

where \ is + A and — A for countercurrent and cocurrent phase contact, respectively.

The same results were obtained byWartman & Mertes(1966) from the transient solution
of the diffusion equation with the appropriate boundary conditions.Vorstman &Thijssen
(1971) calculated asymptotic Shd a valuesfor countercurrent extraction of spherical par-
ticles by using a different mathematical approach. In solvingthe diffusion equation 3.21,
the following general boundary condition wasapplied atr= R:

9co
(3.28)
w =Ä { W-w*(fd)} 1=1

The bulk liquid concentration, m co*(rd), was then obtained from the solution of the

Sh,d.a

Fig.23. AsymptoticShda valuesforconcen-

tration dependent diffusivity IDd =cja and
constant surface concentration (Bi,A-* ~).
-10 2 4 Parameter isgeometricalshapefactor v. From
Schoeber,1976.

38
Table6. ReviewofShod valuesfoiconstantbulkconcentration oftheliquidphaseintheregularregime.

Diffusion Boundary Geometry Source

coefficient condition
Sphere Cylinder Slab

constant constant 5.96 7T2/2 e.g.Beek&

surface 3 Muttzall, 1975
cone, Bi=«o

finite Bi 2\J/3 2XÏ Thijssen, 1969

X,= x,J, (X,) = X: =Bi/tunX,
(1- Bi) tan Xt Bi J0 (X,)

concentration Bi = 00 Fig.23 Fig. 23 Fig.23 Schoeber, 1976

dependent,

2D = c j a , Bi=°° exp(8/3) 4e exp(2) Schoeber, 1976

a ->oo

lumped parameter model, i.e. the Colburn equation. Their results coincide with those
obtained bythe methods discussed earlier.
In the subsequent sectionsShd a values arecalculated for both cocurrent and counter-
current extraction of particles which can be approximated by slabs,cylinders or spheres.
Some general procedures for generating these asymptotic mass transfer coefficients are
discussed.

3.3.2 Shd a valuesfrom solutions undercoupled boundary conditions

During the extraction of rigid particles,the transient concentration distribution can be

obtained from the solution of the diffusion equation, which can be written in a dimen-
sionless form as:

du
^ =J—i— e"-» ^L\ (3.29)
9t, up-i\dt 31 /
«
In general,the following initial and boundary conditions apply:

td =0 w & 0 ) = f(Ö 0<É<1 (3.30)

? =0 oco/a?=0 td>0 (3.31)

%= 1 doildi = -Bi{u-c(tA)lm} rd>0 (3.32)

The concentration in the continuous phase, c(td), canbe calculated from amassbalance.
For cocurrent and countercurrent extraction thisresultsin the following relations:

39
 de 1 dû_
dtA Â dTT
- cocurrent or batch (3.33)
'd = 0 c = c
in

de _ 1 du
dT, " + Â dTT
countercurrent (3.34)
fd = 0 c=c
out

Wewill indicate this system of equations by the term 'coupled boundary conditions',i.e.
coupling of c(f d ) from the boundary condition Eqn 3.32 with a mass balance relation
3.33 or 3.34. Analytical solutions of the above equations are summarized inTable 7,with
special reference to cocurrent and countercurrent phasecontact. Only those solutions not
dealt with extensively in the handbooks (Luikov, 1968;Crank, 1956; Carslaw & Jaeger,
1959) are included in the table. From these solutions the Shd a values can be derived

Tabel 7. Analytical solutions of the diffusion equation withcoupled andgeneralboundary

conditions.

Geometry factor Bi f«) Remarks Source

cocuirent
1,2,3 oo
« 0 tables, graphs Tettamanti et al., 1975
1 oo
m Plachco &Krasuk, 1972
3 oo f«) Plachco & Lago, 1972
1,2,3 finite "o tables, graphs Mikhailov, 1966
2 finite "o Edeskuty & Amundson, 1952
3 finite "o Farritor &Tao, 1972
all finite f«) arbitrary geometry Mikhailov, 1977
countercurrent
1,2,3 oo
"o tables, graphs Tettamanti et al., 1975
1 oo f(«) graphs Plachco & Krasuk, 1970
1 finite "o Jeschar, 1966
3 finite "o mass generation Munro & Amundson, 1950
included
sorption included Kasten & Amundson, 1952
Barbouteau, 1956
Wartman & Mertes, 1966
axial dispersion Neretnieks, 1974
included
general
3 finite "o c = c (r) Vorstman &Thijssen, 1971
all finite f(£) c = c(t) Olçer, 1964
1,2,3 finite f«) c = c(t) Mikhailov, 1973

40
from Eqn3.23.
The solution of the transient diffusion equation canbe writtenin ageneralized form:

w'-w'(rd=0) , o» _ X 2*
W('d)= TT, 7, =A0+g(ß)eß '<•+ 2 f(X;)e / d (3.35)

where f(Xj),/=0, °° andg(/x)arefunctions that depend ontheboundary andinitial condi-

tions and Xj and p are the nonzero roots of the characteristic equation, if those roots
exist. W(td) is defined inTable 8for cocurrent and countercurrent flow. Thevalue of A0
is also given,which can beverified easily for cocurrent extraction, and for countercurrent
extraction with an extraction factor A > 1 by taking the limitforr d ->oo j n Eqn 3.35
since no root ß exists in these cases. CombiningEqn 3.35 with the massbalance on both
phases, the value of C(td) defined in Table 8isobtained. The definition of theSh number
givenin Eqn 3.23 can nowbe written as:

tfgM^'i-ï X* ffy)e~*>'d
Shod(td)= '— — — (3.36)
1 ( fd
( l - A - ) { ^ U ) e ^ d + £ f(X..)e~^ }

where \ =A or — A for countercurrent and cocurrent phase contact respectively. For

longer contact times between the phases, the solution willbe determined by the smallest

Table 8. Asymptotic 5Vz-numbers and some related properties of the differential equationsinvolved.
7=At/v; Jft istheBesselfunction ofthefirst kind andk-th order;1^isthe modified Besselfunction of
the first kind and k-th order;Xisthe first non-zeroroot of the related trancedental equation.

phase
w C A» ^od.a X or M
contact
ÜJ - u , c - w0 A 2^o 2 yBiX
cocurrent X
c0 - w » c0 - " o A+ l V Bi + yX2
counter-
current
w - w0 c - to 0 A 2 A 0 ^ J. w yBiX
A>1
c
out~ ow c
out-w<> A-l V Jo M Bi + yX*

A(Bi + 7 X2)
nX =
Bi + 7 (1 - Bi)X2

counter- y Bin
current tanhju
A<1 c
out~ ow c
out -tJ
o * Bi + yß2

i. GO yBi ß
Io (M) Bi + 7 ( i '

+ 1 nV. ,.
ß(Bi + yß2)
Bi+ y(l-Bi)ß2

41
eigenvalue X t , or by the eigenvalue p if the latter exists. Thus the calculation of the
asymptotic Sh numbers can be reduced to an analysis of these smallest eigenvalues.In the
subsequent treatment, we will use Laplace transform techniquesto determine the charac-
teristic equations that generate the eigenvalues.
The solution of Eqn 3.29 with initial and boundary conditions 3.30 and 3.31 in the
Laplace domain isgiven by:

W =/ e - s t d H / ( | , f d ) d f d
o
=A3 sinh (£V 7 )/Vs" when v = 3 (3.37)

=A2 I 0 (£Vs~) when v = 2 (3.38)

=Aysinh{%Vs ) when v = 1

The value of the constant Av can be obtained from boundary condition 3.32 and the
overall mass balance equation 3.33 or 3.34. Thus, the characteristic equation isobtained
by setting the denominator of the resultingexpression for Wequalto zero and solving for
s. These equations aregiven below:

f— (ßi ys~\•y j
tanh (Vs )= ^ /;• „.. when v= 3 (3.40)
v
' Bi-ys(\—Bi)
Ii(Vs") yBi^Is
when v=2 (3.41)
Io(v7)~ Bi-ys

yBiVs
tanh(v7) =-%r. when v~\ (3.42)
v
' Bi - y s '
where y = A^/v. It canbe shown that the above equationshave anon-zero real root ß only
in case A< 1in the countercurrent situation. This singleroot determines the asymptotic
solution for W(tA). In other cases,aninfinite number of rootsisobtained by substituting
X = iy/Z The smallest non-zero root determines the asymptotic solution inthis case.The
resulting characteristic equations are included in Table 8. The Shod a values obtained
from the equations in this table, together with the Shd a valuescalculated from Eqn 3.4
are shown in Fig. 24-29. Both for cocurrent and countercurrent extraction the solution
approaches the constant bulk concentration limit when the extraction factor A goes to
infinity. These solutions were already summarized in Table 5. The asymptotic mass
transfer coefficient increases from slab to cylindrical to spherical geometry of the par-
ticles subjected to extraction. In fact,-this is a consequence of the definition of the Sh
number given in Eqn 3.23. Due to the higher value of v = 2Ad/Vd, a certain flux -
IDd(bcj'lbr)R causes a more pronounced concentration reduction in terms of average
solute concentration for spherical particles. Decreasing the external mass transfer coeffi-
cient or Bi number shifts the Shd a value towards the value obtained for A = 1. For
countercurrent extraction, an increase in A causes a reduction in the asymptotic mass

42
18

16

14

12

10

8-
100
6
_i i i i i ' i ' _i l : I I—1—L
0 ^
0.1 0.2 0.4 0.2 0.4 10
A
Shod,a

12 \ parameter Bi

" 100
6

4
^^^^__2zzzii:
— • — _ _ _ _ _
• 10

n i i 1 1 1 1 . 1 1 i i I , . . J_I
0.1 0.2 0.4 10
A

Fig. 24. Asymptotic Sh& and ShQ&valuesfor countercurrent extrac-

tion ofsphericalparticles.Bi= mkcR/ID^ istheBiot number.

43
 10 100 parameter Bi

j i i i i
10
A

2\-

_i i ' ' i i i i _l I L_J—I—1_

"0.1 0.2 0.4 10
A

Fig. 25. Asymptotic Sh^ and Sh0^ valuesfor countetcurrent extrac-

tion of cylindrical particles.Bi=mkcR/E>^ isthe Biot number.

44
 parameter Bi
12
100 »

10 \

lOv
6

5.
4
3 ______^^

2
i

i i ' i J~ •••• - J— . i - i • i—i-

XX Q2 0.4 10
A

Fig.26. Asymptotic Sh^ and Sh0^ valuesfor countercurrent extrac-

tion of particles with slab geometry. Bi=mkcR/EI^ istheBiotnum-
ber.

45
Shda
10-1—
parameter Bi

2-

_J i i i_ -I I I I I I 1—L
0.1 0.2 0.4 10
A
Shod,a
parameter Bi 100
6-
40

ÏÖ
sT^^^-'
5
jZs^.
3
3- jé<\^ "~"

2-

i i i i i i i i i i i i 1 1 1 l 1
0.1 0.2 0.4 10
A

Fig. 27. Asymptotic Sh^ and Sh0^ valuesfor cocurrent extraction of

sphericalparticles.Bi=mkfJi/E)^ istheBiot number.

46
 -I 1 1 1—I I • • -I 1 1 I I L_l_l_

0.1 0.2 0.4 10

A

Fig. 28. AsymptoticSh^ andSh0(^ valuesfor cocurrent extraction of

cylindrical particles.Bi=mkcR/lD^ isthe Biot number.

47
Shd|<1
10|—
parameter Bi

_! I I I ' ' ' ' _1 I I I I U

0.1 0.2 0.4 10
A
Sh od,a

parameter Bi

100

10

s' ^^~ 5

s^s^Z^—-— " 3

1 1 1 ! 1 1 1 1 1 1 1 1 1 1 1 1 1

0.1 0.2 0.4 10

A

Fig.29. AsymptoticSh,}andSh0a valuesfor cocurrent extraction of

particleswith slabgeometry.Bi=mkcR/E>àisthe Biot number.

48
transfer coefficient. Due to the falling concentration in the liquid surrounding the par-
ticle, the concentration gradients at the solid-liquid interface (9co'/9r)R are steepened
with the result that was mentioned above. The reserve is true for cocurrent extraction
where the liquid concentration rises during the extraction process.

3.3.3 Shd a valuesfrom solutions undergeneralboundary conditions

The analytical solution of the diffusion equation 3.29 with boundary conditions 3.30
through 3.32 can be derived from the generalized solutions presented by e.g. Mikhailov
(1973). For the basicgeometries these solutions are summarized below:

c'(td = Q)/m-to'0

2vB] 2 [ f C0j)e-*-" ( ' d - , , ) dt J ]/[X2 +Bi(Bi +2-v)] (3.43)

/?—'1 O

where C(td), the bulk liquid concentration at the dimensionless timetd, isembodied by:

c'(td)/m - co'o = 1
C{ td) l c(f
(3.44)
- ~ c'(0)/m - co'o ~ à

and Xnare the positive,non-zero roots of the characteristic equations:

tan Xn=Bil\n when v= 1 (3.45)

3l(K)Uo(\) =BHK when v= 2 (3.46)

tan Xn=Xn(1- Bi) when v=3 (3.47)

For longer contact times, the regular regime solutions result in a mass flux that can be
described with a constant mass transfer coefficient k„A. For the calculation of ShnA „
OU / OQ,a
values, we are only interested in this limiting behaviour. Therefore, the time variant bulk
liquid concentration C(td) can be calculated from the solutions ofthe lumped parameter,
model. In Section 2.1,some solutions of this model for various flow systemswere given.
For cocurrent and countercurrent contact in plug flow , C(td) can be determined from
Table 2when the number of transfer unitsisreplaced by:
^t.d =koda' Td (x/L) =Shod td v\2 (3.48)

where x isthe coordinate inthe liquid flow direction and Shod =kod2R/E)d. Inthisway
the following relation is obtained:

49
 (v i - M
A, - e x p ^ - td Shod >
<**„)= ^ ^ — ' (3-49)

where \ equals A for countercurrent phase contact, and —Afor cocurrent extraction.
The extraction rate — dtö/dtd canbe calculated bysubstitution ofthe former equation
inEqn3.43,which yields:

dZÔ__2vBi2 g [ ( A t - D ^ + M l e - ^ ' d - X e ^ d
dfd Aj - 1«=i (X* +A) [X* +Bi(Bi + 2-v)]

where A =(i>/2)Shod (1 - At)l\. An expression for thedrivingforce formass transfer

isobtained from amassbalance3.27:

(w - w'o)/At + (c' out - c')/m =0 (3.51)

orindimensionless notation:

w - c =( A t - l ) c - A t (3.52)

The overall drivingforce cannowbederived from Eqns.3.49and3.52:

1
iv ~At \
co- c=- exp<- Sfcod rdV=- exp(-AtA) (3.53)

In fact, this expression isproportional to theextraction rate calculated from thelumped

parameter model. TheShod number cannowbeobtained easily from Eqns3.50and3.53
for acontact timetd =Fo:

+
^ ^TA^KF0-Al^eAFo
-2vBi2 2
-Sh „= "=' (X2n+A)[X2n+Bi(Bi+2-u)]
s\ — o'od
o
z
-exp(AFo)

(X2l +-^-A)e-^+^Fo-A 1
Aj-1 ' At-1
= 2i>522 2 — (3.54)
«=i (k2,+A)[\2„+Bi(Bi + 2-v)]
The asymptotic Shod number for long contact times is derived from this equation by
taking the limit for Fo -»oo. This raises the problem that no information is available
beforehand about the value of -(X2,+A). However, it isknown that theShod a valueis
the lower limiting value of the actual Shod value, and we can only accept finite mass

50
transfer rates. Thus —(\„+^4) must always belessthen zero.Thus the following relation
isderived for Fo-> oo;

At
2 {(\2n+A)[\2n+Bi(Bi +2-u)]}-1 (3.55)
2vBil n=l

where A is a function of the asymptotic Sh number. The iterative solution of this

equation yields Sh values for arbitrary values ofBi, \ and v.The solutions are identical
to those presented in the previous section.

3.4 Asymptotic Shd a numbersinengineeringcalculations

The practical use of the asymptotic Sh values obtained from the theory presented in
the previous sections is restricted to the extraction of uniform particles with simple
geometry, cocurrent or countercurrent phase contact inpure flug flow and relatively long
contact times between the phases. In this section attention ispaid to the applicability of
the asymptotic transfer coefficients in practical design calculations, where more compli-
cated cases are encountered. The error introduced when Shod a numbers are applied at
finite Fo values is discussed. The determination of these asymptotic transfer coefficients
for other boundary conditions will be outlined. Subsequently, the estimation of Shod a
for irregular particles and for non-uniform particle size distributions isdiscussed. Finally,
an approximate explicit relation is derived that predicts Shod a values in countercurrent
extractors when the extraction factor isclose to orlarger than unity.

Range of applicability. The practical significance of asymptotic 57iod a numbers can be

measured by the critical Fo value, above which the time-averaged Sh differs less than a
specified percentage from the asymptotic value.Onthisbasisthe range of applicability is
discussed in this paragraph. The approximate Fo time required to perform a certain
separation is calculated from the lumped parameter model using the asymptotic Shod a
value, see Eqn 3.48. Also, the exact Fo' value isdetermined from the analytical solution
of the diffusion equation. The relative difference between those two is then plotted
against Fo'. Due to the large number of parameters involved: Bi, A, v, mode of phase
contact and Fo', only some sample cases are covered by the figures. Fig. 30 shows the
results for cocurrent extraction of particles with slab geometry. It clearly shows the
expected increasing accuracy of the method with risingFo' value.ForhigherBi numbers,
the critical Fo' value increases since the overall mass transfer coefficient becomes more
and more determined by the time-dependent Shd ;the contribution of the constant valued
Shc is gradually reduced. The effect of the extraction factor is shown in Fig. 31. This
graph indicates that an increase in A reduces the critical Fo' value for cocurrent extrac-
tion. The reverse is true for countercurrent extraction. The latter conclusion is inagree-
ment with the results of Vorstman &Thijssen (1971) and of Brüniche Olsen(1962).The
curve for \ =—30 is essentially identical to the curve obtained for a constant bulk
concentration. The figure clearly shows thatthecriticalFo'value increasesfrom counter-
current to cocurrent phase contact. To determine whether the criticalFo' valuesare met
in a process, the extraction efficiency hasto be calculated from the analytical solution of
the diffusion equation as a function of the Fo time. The results are shown in Fig. 32,

51
Fo- Fo
100%>

Fig. 30. RelativeetioiintheFouriervaluecal-

culated from the Sh^a value for cocurrent ex-
0.1 Q2 0.4 traction. Slab geometry, v= 1;Extraction fac-
torA= 10;Parameter Bi.

Fo-Fo'
100%
Fo'
100

cocurrent

countcrcurrent

Fig. 31. Relativeerrorin the Fouriervaluecal-

culated from the Sh^ value for cocurrent and
0.1 Q2 0.4 3 countercurrent extraction. Spherical geometry,
Fo' v=3;Bi= 100;parameter At.

52
Fo- Fo'
100°/,

Fig. 32. Error in extraction calculations using asymptotic Shod values

for cocurrent and countercurrent extractors. Bi- °°;( )v=3;( )
v= 1;ParameterA^.

wheretherelativeerrorintheFovaluerequiredfortheseparation,ascalculatedfromthe*
asymptotic Sh number, isplotted against the fraction ofthe solublesremainingin the
solids, l-W/WFo _... =(w'ln - w' out )/(w', n - oj'out)Fo .„..For countercurrentextrac-
tion and plane sheet geometry, the relative error is less than 10%when theextracted
fraction is higher than 0.9,or 1-W/Wmax < 0.1.This value will often be attained in
solid-liquid extraction processes in the food industry. For either cocurrent or batch
extraction, Fig.32showsconsiderable discrepancy betweenactualandapproximatedFo
value;inparticular forsphericalparticlestheminimumextractedfraction willnotbemet
inpractice.
The asymptotic Shoi a value only determinestheextraction rate attained after long
contact times between the phases, Fig.33.Theextracted fraction / canbecalculated
more accurately when a fictious initial condition td =0,ƒ =f0 isincorporated inthe

53
In(1-f 0 )

0.215

Fig. 33. Graphical representation of calculation procedures for

heat and mass transfer problems. I first term approximation; II
penetration/regular regime;IIIexact solution diffusion equation;
IVasymptotic transfer coefficients.

calculation. In fact, thisprocedure forms the basisof the first term approximations (Pflug
& Blaisdell, 1963). It has the disadvantage that the solution of the diffusion equation
with proper initial and boundary conditions must be available. Here, a calculation pro-
cedure is adopted, which is similar to the one proposed for drying calculations by
Schoeber (1976), that can be used to increase the accuracy of thecalculationatsmallFo
values. The first part of the extraction processiscalculated from the penetration theory.
In a later stage the regular regime solution is applied. In Appendix D an analytical
solution for short contact times is derived, for a plane sheet subjected to cocurrent or
countercurrent extraction with finite mass transfer rate in the solvent phase. The time-
dependent Shod numbers obtained from this solution are shown in Fig. 34. From this
figure we learn, that the Shod value decreases rapidly from Shod =Bi at t^= 0to some
minimum value. In all cases, this minimum is located close to y/Fd =0.45 or r d = 0.215,
independent ofBi, Aand the mode of phase contact. Furthermore, thisminimum valueis
very close to the asymptotic Shod value.Inmost casesit isnot more than 3%higher.The
increase of the Shod number at longer times is caused by the fact that the boundary
condition (D.4) for a semi-infinite slab is applied to a finite particle with finite storage
capacity for the solute.Thisincreaseinthe masstransfer coefficient isphysically unrealis-
tic. The above results suggest the use of the solution of the diffusion equation in the
penetration period in the calculation up to td = 0.215. The subsequent part of the
extractor is designed with the lumped parameter model, using the asymptotic value of
Shod attained in the regular regime. The initial conditions of the solid and liquid phases
in the second stage of the process arethen obtained from the solution in the penetration
period. The extraction efficiency:

54
Fig. 34a. 5Ä 0( J valuesinpenetration
period for semi infinite slab geom-
etry.Bi=o»; Parameter 1/At.

Fig. 34b. 5A 0 j valuesin penetration

period for semi infinite slab geom-
etry.Bi= 10;Parameter 1/At.

55
 w= (3.56)
c'0/m-oj'0

at Fo = 0.215 is shown in Fig.35for both cocurrent and countercurrent extraction. The

overallresidence time required for the specified separation isthen obtained from:

W(R>=0.215)

Fig. 35a. Extracted fraction inpenetration period

0.1 0.2 0.4 at Fo=0.215 for semi infinite slab subjected to
countercurrent extraction. ParameterBi.

Fig. 35b. Extracted fraction inpenetration period

at Fo= 0.215for semi infinite slab subjected to
cocurrent extraction. ( ) extracted fraction at
0.1 0.2 0.4 equilibrium (Fo-><»), W =A/(l + A); Parameter
Bi.

56
 ( F O S O
F 0 =0.215 + i - i - i n ^ -215)-1 F O >0.215 (3.57)
P aod,a pW(FÓ)-\
where p = (1—1/At). The Fo value calculated by the above procedure agreesup to a few
percent with the exact solution for slabs. For Fo values smaller than 0.215 the Fovalue
hasto be determined from the equations given in Appendix D.

Irregular shapesand particle size distributions. From the graphs presented in this chap-
ter, the Shod a values for the extraction of particles with a simple geometry can be
determined. For irregularly shaped particles, the common procedure is to approximate
the particle shape by one of the basic geometries. Rutov (1958) suggested a method to
improve the estimated kd value for slightly irregular shapes. Forifr-numbers greater than
1,the diffusion coefficient hasto be replaced by an effective diffusivity JD'd:

E>'d=E>dAJAdtV (3.58)

where Ad is the surface area of the particle and Ad v isthe surface area equivalent regular
particle.
For non-uniformly sized particles, Gardner (1966) proposed a method to estimate an
effective overall transfer coefficient when the particle size distribution is known. This
procedure involves the determination of the particle size r% which follows the average
concentration of the particle cloud. When m is the cumulative particle size distribution
function, the following relation results:

d(m)
=1-At (3.59)
ƒ At kod/r0v
1+
1-At (kod/r0i>)*

wherev isthe particle velocity relative to the wall.r% canbe determined bytrialanderror.
WhenAt = 1,asomewhat simpler equation is obtained:

d(m) 1
ƒ V
r 'v -
Kdl o k r v
( odl o ~)*
(3.60)

Boundary conditions. When a specific flow pattern inthe extractor givesriseto aboun-
dary condition cj*(td) that cannot be adequately described by one of the cases treated in
this section, one of the methods discussed in Section 3.3.2 or Section 3.3.3 should be
used to calculate Shoda values. The coupled boundary condition solution method as
discussed in Section 3.3.2 produces relatively simple equations for.S7zod a . However, the
related characteristic equations may be rather complex, and the eigenvalues are normally
not found in literature. Fortunately only one eigenvalue is required for the calculations.
The general boundary condition method presented in Section 3.3.3 involves trial and
error solution of an equation containing an infinite summation series to determine
Shoda. The advantage of this method is that for numerous cases the lumped parameter
solutions are published in literature. Furthermore, the expression for.S7zod is applicable

57
to all kinds of boundary conditions C(td), without modification of the characteristic
equations. The first 7 eigenvalues are readily obtained from the common handbooks,
while additional roots aretabulated by Hayakawa (1975) for v=2 and v- 3 and by Bakal
et al. (1970) for v = 1. For these reasons the method discussed in Section 3.3.2 is only
recommended for a rather simple problem. When complicated boundary conditions are
involved, and consequently complex characteristic equations, the method outlined in
Section 3.3.3 ismore suitable.

Explicit relation for Shod a. A quick estimation of Shod a values without resorting to
the graphs presented in this chapter can be obtained from the approximate relation
explicit in Shod a . The equation is valid for countercurrent phase contact, when the
extraction factor A is close to or greater than 1. The derivation of the equation is
discussed in Appendix C.For plane sheet geometry the relation reads:

Bi {6 {Bi + 2.25) + (A - 1) {Bi + 3) TT2/2>

Shd a = i-^ ' —— (3.61)
Bi A(Bi +2.25) + (A - 1){3 (Bi +2.25) - (Bi + 3)n2l4}

The solution for other geometries canbegenerated by the conversion rule

S V a ^ ^ V a , - ! -4("-l) (3-62)

For extraction factor A= 1,these equations are exact.

58
4 Liquid maldistribution in solid-liquid extractors: hydrodynamic
instabilities

4.1 Introduction

It was mentioned in Chapter 1 that viscous fingering may occur in systems where
viscosity gradients exist in the liquid flowing through a densely packed bed. The mecha-
nism that is responsible for the formation of viscous fingers can be elucidated by means
of a simple model shown in Fig. 36. For the sake of simplicity, any interaction between
the liquid and the packing material is not considered. The figure shows the situation
where a liquid 2 is displaced downwards by a miscible liquid 1, which has the lower
viscosity. Since in general density increaseswithviscosity, the displaced liquid isalsosaid
to be the denser one. A protrusion of the displacing liquid is supposed to be formed at
the liquid-liquid interface due to, for example, a local permeability variation in the
packing material. The displacing liquid will now flow preferentially through this protru-
sion, since the mobility of the liquid in this region ishigher than that inthe surrounding
liquid. Hence, the disturbance tends to grow. On the other hand, especially at low
displacement velocities, the favourable density difference counteracts any distortion of
the displacement front. The flow canthus be stabilized by gravity segregation. The overall
effect of viscous and gravity forces depends on the displacement velocity, physical prop-
erties of the liquids and properties of the packing material. Table 9 gives a schematic
overview of the relative stability of several displacement conditions. In solid-liquid extrac-
tors viscosity differences may arise from concentration gradients in the solvent phase.
Since both the shape of the concentration gradient andmasstransfer between the solvent
and the packing material affect the flow stability,the hydrodynamics of the flow in solid
liquid extraction equipment present a much more complicated picture than the simple
casejust described.It isthe aim of thissection to provide information onwhether viscous
fingering affects extraction efficiency. In order to do so, the stabilizing effect of a
concentration gradient in the liquid flowing through an inert packed bed is first ex-
amined. Subsequently, mass transfer between the packing and the solvent is taken into

<v>A

Fig. 36. Definition sketch for thephenomenological description of

unstable displacement.

59
Table9. Effect of flow conditions on the stability of displacement from
packed beds.Index 1refers tothe upperliquid.

Flow Sign Sign Stable displacement

direction (P! - P 2 ) Oi, - M2) at flowrates

t + + none
+ - high
- + low
- - all
I + + high
+ - none
- + all
- - low

account.
Stability criteria for liquid-liquid miscible displacement from inert packed bedsare
discussed in this section. Some attention is given to flow phenomena in the unstable
regime, and some methods suggested inliterature to describe unstable displacementare
given. Stability critera were verified experimentally for a stepwise concentration change
of the solute whichincreasestheviscosity oftheliquidflowing throughadenselypacked
bed. Thisstudy wasthen extended to the caseofagraduallyfallingconcentrationinthe
displacingliquid in Section4.2.Thissituationisrelevanttosolid-liquidextraction,where
the concentration ofthesoluterisesinthedirectionoftheliquidflow.Finally,the effect
of masstransfer onthe flow stability wasstudied. Twodifferent caseswere considered.
Whenthe liquid in the interstitial voids of the packedbedisinitiallyinequilibriumwith
theliquidin theinternalporesofthepackingmaterial,masstransferisexpectedtostabi-
lizethe displacement.Dueto masstransfer in the protrusion, the viscosity ratio for the
liquids outside andinside the protrusionisdecreased.Thegrowthrateofthedisturbance
isthus reduced bymasstransfer.Whentheliquidintheinterstitialvoidsisinitiallynotin
equilibrium with the solids,masstransfer occursbothinsideandoutsidethedisturbance.
Thentheoveralleffect dependsondisplacementconditionsandthephysicalpropertiesof
the liquid. The extraction efficiency in apacked column hasbeen measuredoverawide
rangeofdisplacementvelocitiestoestablishthese effects.

4.1.1 Stabilitycriteria

For liquid-liquid displacement, both miscible and immiscible, stability criteria are
reported inliterature.Ingeneral,interactionbetweentheporousmediumandthe flowing
liquid, e.g. adsorption,isneglected.Whengravityeffects arenegligible,asinhorizontalor
matched density experiments, Muskat (1949) found that the flow is stable when the
mobility ratio of the displacing and the displaced liquid exceeds 1. The mobility is
defined as the ratio of permeability andviscosity.Hill (1952)included stabilization by

60
gravity during miscible displacement in the analysis. He assumed a sharp interface be-
tween the displaced and the displacing liquids, as shown in Fig. 37. At z the pressure in
both the displaced and the displacing phase, p2 and p2 respectively, isassumed equal to
p. The pressures in the phases at z + Az can be calculated from this value when the
pressure gradients in thephasesareknown.Thehypothetical disturbance indicated in the
figure will then grow when p\ <p'2 and will be suppressed when p\ > p'2 •According to
the expression for p\ and p\, the condition for marginal stability cannow be formulated
as:

dp dp
(4.1)
àz dz

The pressure gradients in both phases can be calculated from the Darcy Law.Whenvc is
the maximum stablevelocity we obtain:

"cM2
+ Pig =-• + P2g (4.2)
Kl

The displacement isthus stabilized when <v> <vc wherevc isgiven by:

g(ß2 - P i )
(4.3)

«•2 «1

It canbe verified easilythat thisresult isinagreement with Table 9.

During the flow through packed beds, any concentration jump between two miscible
liquids, one of which displaces the other, will besmoothed out by diffusion or dispersion
effects. Thus the stability criterion derived by Hill was reformulated by Dumoré (1964)
to account for a continuous concentration change in the liquids. Since for downward
displacement any disturbance in the equi-concentration planes will grow when the local
pressure gradient is lower than the pressure gradient in the liquid downstream,,the dis-
placement will be stable when the pressure gradient increases continuously in the flow
direction. Thecondition for marginalstability thusreads:

'1'

P,= I V P z

z*Az
dp >
'2=P-(£)>
•2' i Fig. 37. Definition sketch of a hypothetical distur-
P;-P-(-H£), A Z banceat thedisplacementfront.

61
Combining the Darcy Lawwith Eqn4.4 givesfor aconstant permeability K
1 d2p vs dp
—r~ TT = - — + T " (4-5)
dß dz2 Kg dß
g
Tz
where vs is the maximum stable displacement velocity in the situation under considera-
tion. When the viscosity in the liquid increases in the direction of the flow, dp/dz is
greater than zero and the stability condition that isobtained for this caseisthat <v><
vs, withvs given by:
vs =ng(dp/dp) (4.6)

The value of (dp/dp.) is only afunction of the concentration of the solute. Since Eqn 4.6
should be satisfied everywhere in the packed column to prevent the formation ofviscous
'fingers', the minimum value of (dp/dp) is inserted in Eqn 4.6. Thususually (dp/dp) has
to be determined for the concentration of the displaced liquid in the present case.
The stability criteria of Hill and Dumoré do not take into account the stabilizing
effect of axial and radial dispersionin the liquid phase.Smallamplitude distortions of the
displacement front can be smoothed out by axial dispersion. On the other hand, radial
dispersion tends to supress small wavelength perturbations irrespective of their ampli-
tude. Perkins(1965) stated that the smallest channel that wasnot smoothed out by radial
dispersionhasawavelength:

X min =2 V 3 . 9 5 / D E T r (4.7)

This result can be obtained assuminga Fourier number of 0.25 required for smoothing of
the radial concentration gradients in the protrusion. From this discussion it can be con-
cluded that the perturbations with a large amplitude and a long wavelength arethe least
affected by dispersion. The least stable distortion of the displacement front thus has a
wavelength that isbounded by the sizeof the apparatus under consideration.
Amarginalstabilityconditionthattakesdispersioneffects into account canbe obtained
from thesimultaneoussolutionoftheequation ofchangeforcomponentAassuming disper-
sion with adispersion coefficient E)E, the equation of motion with the Darcyconstitutive
equation for the velocity vector ]t and the continuity equation in a three-dimensional
velocity field. For abinary mixture of AandB:

- £ +V-pAv-V'lDEpV (p A /p)=0 (4.8)

v+ (K//zA)(Vp-p?+ pD*/Df) =0 (4.9)

^ + V p v =0 (4.10)
at
62
d/dt denotes the partial derivative with respect to t, D/Drthe Lagrangian or substantial
derivative. In these equations, both the liquid density p andviscosityju are dependent on
the concentration of component A. For the determination of marginal conditions for
stable displacement, only the stability of the displacement front when it is subjected to
small distortions has to be examined. This problem is usually solved by perturbation
analysis. For small differences between the pure component densities, Ap/p A , Perrine
(1961) derived an expression for the initial growth rate ya of aperturbation of the form:

p„ =p„ e ' ( a x + ^ + "' rz > e-™ (4.11)

where p n is the solution for stable flow; x, y and z are the spatial coordinates of a
rectangular coordinate system extending infinitely in the* a n d j directions while bound-
ed in the z direction between 0 andL. Thebulk liquid flow isin the positivex direction.
Thegrowth rate parameter yn characterizes the initial growth rate of aperturbationwitha
specific wavelength. For stability analysis in miscible displacements, only the realpart of
7 n has to be considered, because the flow instabilities observed inlaboratory experiments
are found to be of non-oscillatory nature (Schowalter, 1965). This result was derived
analytically for small density differences between the displaced and the displacing liquid
by Lapwood (1948). A condition for marginal stability can thus be obtained by setting
the value of the growth rate parameter yn for the least stable distortion equal to zero,
thus forcing all possible perturbations to be just suppressed. Only the solution for down-
ward, vertical displacement is given here. Bearing in mind that the least stable perturba-
tion has a maximum possible wavelength and amplitude, we obtain for the maximum
stabledisplacement velocity v':

Kg dp
ß de
r i (4.12)
ain/i * /£> n \ be
dc
BoT\L / \dx'
where BoT = V'SDP/E>E,T i s thetransverseBo number andx' =x/D ,x beingthe distance
coordinate in the direction of the liquid flow in a coordinate system moving with the
mean flow velocity. For negligible radial dispersion BoT-+<x>,thisequation reducesto the
stability condition that was later derived by Dumoré (Ï964). Schowalter (1965)analysed
the casewhere no restriction isput on the maximum pure component density differences.
He assumed that the effective dispersion coefficient E)E has the same value in both the.
transverse and longitudinal direction. Accordingto Perkins(1965)the following empirical
relation canbeused to estimate radialPevalues inporous media that canbe characterized
by atortuosity factor F and aninterstitialvoid fraction h:

BO 1
T = TT, — + 0.0157 a (4.13)
Fh <v>Dp
where E)c is the molecular diffusion coefficient of component Ainthe liquid phase and
the parameter ameasures the inhomogenity of the packing.Asimilar expression holds for
theBo number in the longitudinal direction:

63
 Bo
l =™ FT + 0 - 5 f f ( 4 - 14 )
Fh <v>Dp

From these equations it is observed that equal dispersion coefficients can only be as-
sumed in the low flow velocity range, where dispersion is controlled bymolecular diffu-
sion. From perturbation analysis Schowalter found that the condition for marginal sta-
bility was determined by four dimensionless groups, quantifying the relative importance
of the effects of gravity, viscosity, dispersion and permeability. For perturbations witha
wavelength smaller than some critical wavelength Xc,the displacement front is stabilized
by dispersion and gravity effects. In his paper the author presented graphs of Xcagainst
the parameters that characterize the displacement condition. Asimilar analysisisgivenby
Heller (1966) who did not assume the same value for the dispersion coefficients in both
the longitudinal and the transverse direction. For a ramp-shaped concentration distribu-
tion in the mixing zone at the displacement front, the value of the growth rate parameter
of the least stable perturbation isgiven for various displacement conditions.
In conclusion, simple stability conditions describing the stabilizing effect of gravity
segregation at the onset viscous fingering are directly obtained from the Darcy Law.A
more complete analysis that includes dispersion effects shows that the onset of viscous
fingeringdepends on the wavelength of the perturbations. The development of the least
stable distortion of the displacement front is retarded by transverse dispersion effects,
and also by longitudinal dispersion when small amplitude perturbations are considered.
The wavelength of the least stable distortion is determined by the size of the experi-
mental equipment, and consequently the onset of flow instabilities depends on the size of
the apparatus considered.

4.1.2 Flow behaviourin the unstableregime

In addition to the formulation of stability criteria,many attemptshave beenmade to

describe the flow phenomena in the unstable flow regime.Themathematical modelling of
viscousfingeringcommonly proceeds along one of the following two routes:
- Numerical solution of the equations of motion, dispersion and continuity, Eqns
4.8—4.10.Sincethe 'dispersion equation',i.e.the continuity equation for acomponent A,
is a mixed parabolic and hyperbolic partial differential equation, simple finite difference
methods lead to numerical smearing of the concentration profiles when the dispersion
term is small compared with the convective term (Chhatwal, 1973).Therefore, onehas to
resort to complex numerical schemes,seefor example Peachman &Ratchford (1962).
— Phenomenological description of the flow behaviour based on the Buckley-Leverett
model. This model was origionally derived for the prediction of breakthrough curves for
immiscible displacement of liquids with equal densities.Twobasic equations areinvolved
inthismodel. Firstly the frontal advance equation:

dx <v> bf
(4 15)
Yt " ^ " -
the result from a simple mass balance, that embodies the rate at which an equi-satura-

64
r tion plane with solvent saturation sproceedsthroughtheporousmediumintermsofthe
mean linear flow rate <v >/h andtherelation between thefraction ofsolvent thatis
actually flowing and the local saturation of solvent s. Secondly, the fractional flow
equationthat relatesthequantityƒtotheviscosity ratioofthe immiscibleliquids,M:

vxA ( 1 1 n-1
1
ƒ= HT 7 T = 1 + - 'TM'-\ (4-16)
<v> (At +A2) [ a M y) '
where a =Ai/A2, M=/X2/MIand T = K I / « 2 (Fig.36).Thisequationcanbederived from
the Darcy Lawassuming anequal pressuredropinboththepuresolvent channelandthe
displaced liquid regions. Koval (1963) followed this approach for simulating matched
density displacements. Inorder tofithisresultswithexperimentaldata,hereplacedMin
Eqn4.16bytheeffective viscosity ratioMe,whichisrelatedtotheactualviscosityratio
Mby:

Me =(0.78+ 0.22M 0 - 2 5 ) 4 (4.17)

Dougherty (1963) included bothgravityeffects andmixingintheanalysis.Thisresultsin

afractional flow equationofthe form:
( 1 1 1 ) - 1 / a ns(l-c)Ap/Ji \
f=\l +- • r - -}• <1- -\ (4.18)
(. a M y) \ a+1 <v> )
where alinear relation betweenpandcisassumed,whencisthelocalsolventconcentra-
tionintheliquidmixture;p0 isdefinedby:

Mo={cM s -°- 2 5 +(l-c)M 0 - 0 - 2 5 }- 4 (4.19)

ThemassbalanceonslicedzinFig.35nowreadsforthesolventchannels:

"ITf + Jt —B'Pl C 1 -')* 1 -')" 1 <4-20>

andforthesolventdissolvedinthedisplacedliquid:

<v> aO_^£ + a ( i _ £ ) i = 5 s P l ( 1
n oz at
Thedispersivemixingtermsarelumpedinthetermontherightintheequations.Claridge
(1972) solved these equations (4.18—4.21)numerically.Theparameters y,B,px andp2
were adjusted to obtain acceptable agreement between calculated andmeasured break-
through curves.Bwas found todependondisplacementvelocity,viscosityratio,concen-
trationofsolventintheoilphaseandthedegreeoffingering.
Both the methods result in fairly close agreement between calculated andobserved
breakthroughcurves,thelatterhoweveronlyafter scalingofseveralmodelparameters.
In addition totheabove studies ontheonsetandprogressionofviscousfingers,some

65
interesting features of fully developed fingers are reported inthe literature. Mathematical
relations were derived for the shape ofboth the front end (Saffman &Taylor, 1958)and
the rear end (Outmans, 1963) of the fingers. Apart from this front and rear end, the
channel occupies a constant fraction of the column cross-sectional area. This fraction
ranges from about 1for verylow flow ratesto an asymptotic value 0.5 at higher displace-
ment velocities (Saffman & Taylor, 1958). Numerical simulations and electric analogue
studies have shown that in this portion of the channel the flow isinthe direction of the
main liquid flow vector. Only at the front end and at the rear lateral flow ofliquidwas
found from the finger towardsthe surrounding liquid and the reverse.(Richardson, 1961;
Perkins, 1965).

4.1.3 Experimental verificationfrom literature data

Many experimental investigations on liquid displacement in packed beds have been

reported in theliterature.Most of these are concerned with the efficiency of the displace-
ment in the stable and in the unstable flow regime, when a liquid is displaced from an
inert porous medium (Hill, 1952; Brigham, 1961; Slobod & Howlett, 1964; King &
Denizman, 1961). The results of these measurements are in qualitative agreement with
Table 9. In most cases,the critical velocity vc where the flow instabilitiesjust developis
determined by plotting the length of the mixing zone against the displacement velocity
<v>. Amore or less sharp increase of thislength isobserved near <v>=vc. The actually
measuredvalue of thismixinglength isdependent onthe displacement conditions and the
experimental setup that was used. Slobod (loc. cit.) found that the mixing length corre-
lated well with the ratio of the actual displacement velocity and the critical displacement
velocity, as shown in Fig. 38. Thisratio <v>jvc canbeinterpreted asthe ratio ofviscous
and gravity force termsin the Darcy equation:

<v> <V>AJU/K
(4.22)
gkp

The effect of a graded mobility zone, i.e. a continuously falling viscosity of the injected
fluid, on the stability of the flow through aporous medium wasfirststudied by Slobod&
Lestz (1960).They reported macroscopic stability of the flow whenthe displacement was
carried out with a linear falling viscosity of the injected fluid. The displacement condi-

hm.fraction ofporevolume
1.0

0.6-

0.2-
V 7
<v>
Fig. 38. Transition zone length hm for displacement with
favourable density and unfavourable viscosity ratio. From
Vr Slobod, 1964.

66
tions were such that Eqn 4.12 predicted unstable displacement. Kyle & Perrine (1965)
applied a linear concentration gradient in the displacing liquid. They observed that any
tendency for viscous fingers to develop was completely suppressed at steeper concentra-
tion gradients than those predicted by theory, Eqn 4.12. They concluded that formu-
lating the stability criterion based on the least stable perturbation is quite conservative.
Increasing the steepness of the concentration gradient increased the tendency for viscous
fingers to develop. Mungan (1971), who used a Hele Shaw model in his experiments,
found qualitatively that an exponential falling concentration of the injected liquid in-
creased flow stability inhisimmiscible displacement experiments.
The effect of mass transfer from the packing to the liquid, as occurs in solid-liquid
extractors, on the stability of the flow has been given little attention.King&Denizman
(1972)mentioned the existence of unstable flow during solid-liquid adsorption chromato-
graphy. In their experimental study they measured the height of the mixing zone,hm, as
a function of the displacement velocity, for both porous and non-porous packing mate-
rial. hm was defined by the authors as the time interval between passage of the 10 and
90% concentration levels in the liquid effluent, multiplied by the mean displacement
velocity. In all cases they observed a decreased height of the mixing zonewhen aporous
packing was applied instead of a non-porous one. In their analysisthey did not discrimi-
nate between the effect of finite masstransfer ratesand hydrodynamic instabilities on the
value of hm. Nevertheless it can be concluded from their experiments that the mass
transfer process considerably stabilizes the flow. The conclusion agrees with the discus-
sion onthis subject in Section4.1.
The growth of viscous fingers once they are initiated at the liquid-liquid interface or
equiconcentration plane has been measured for both miscible and immiscible displace-
ment by many authors, e.g. Benham & Olsen (1963) and Gupta et al. (1974b). It is
generally concluded that the growth rate of the fingered region is virtually constant
during the major portion of the displacement. For short times after the onset of the
channels the growth rate may be apparently slower or faster, and at long times stabiliza-
tion by transversal dispersion may reduce finger growth rate, although theoretically the
mechanism seems rather ineffective in practical systems (Benham & Olsen, 1963;Wood-
ing, 1969). When gravity effects can be neglected in comparison with the viscous effects,
the growth rate of a channel becomes independent of the displacement velocity (Schei-
degger, 1960). For normal flow rates the fractional areathat isoccupied by the channels
in the mixing zone isapproximately 0.5;at verylow flow ratesthisvalue can be/consider-
ablyhigher (Perkins, 1965;Saffman &Taylor, 1958).
The observations just mentioned contribute to the basic understanding of the be,-
haviour of viscous fingers during liquid displacement in packed beds in the unstable
regime. On the other hand it should be realized that the onset and growth of thesefingers
is strongly dependent on the scale of the apparatus used in the experiments (Blackwell,
1959), local inhomogenities of the packing (Hawthorne, 1960), preferent flow along the
wall etc. Thus the quantitative result cannot easily be extrapolated to other displacement
conditions and packingmaterials.

67
4.2 Experimentalstudy ofmiscibleliquiddisplacement frompackedbeds

Experiments reported in the literature aremostly concerned with the determination of

critical displacement velocities and the growth of viscous fingers in the unstable flow
regime, when a liquid is displaced from an inert packed bed by a miscible or immiscible
solvent, see Section 4.1.However, little experimental evidence is available on the stabi-
lizing effect of concentration gradients between displacing and displaced liquid, especially
when those gradients arise from mass transfer between the solids that constitute the
porous medium and the interstitial liquid. It is the aim of this study to contribute to the
knowledge on flow phenomena under these conditions. To facilitate the interpretation of
the measurements, the effect of aconcentration gradient onboth the onset and growth of
viscous fingers is established in the first part of the study. The experimental setup that
was used is sketched in Fig. 39. Both the experimental procedure and the apparatus are
discussed in detail in Appendix E. The column packed with glass beads and the top
and bottom mixer were initially filled with glycerol solution. Water was injected in the
top mixer during the downward displacement experiments. The liquid concentration at
the bed inlet thus fell exponentially with time. In this way, an arbitrary exponential
concentration gradient can be applied by varying the size of the top mixed region. The
breakthrough curve was measured by taking samples at the column outlet and by con-
tinuously monitoring the sample density. In order to assure representative sampling, the

vtfelr
/
« 6
^^m mzmt.
5 5

,
2a 2b 10

4 4
3a *a 3b *

8 ff^
(a) (b)
Fig. 39. Schematic diagram of the experimental setup. Item
list:
1. storagevessel 7. samplepump
2. packed columnsection 8. densitymeter
3. mixedsampleregion 9. paper tapepunch
4. mesh wire 10. storagevesselviscoussolution
5. porousglassdisk 11. three-way cock
6. gradient mixer 12. drain

68
outlet stream was mixed in the lower mixed region so that any radial concentration
difference caused bychannelling was smoothed efficiently. Inthe second part ofthe
experimental study the more complicated case of displacement with simultaneous mass
transfer between thepacking and the interstitial liquid was investigated. The packing
consisted ofporous silica spheres, impregnated with sugar solutions prior to the experi-
ments. The apparatus used and the experimental procedure that was followed areanalo-
gousto the onejust described.

4.2.1 Effect of aconcentration gradientin the liquidphaseonflow stability

Experiments were designed to determine the effect of aconcentration gradient in the

displacing liquid on the stability of the displacement. The measured breakthrough curves
give information on the residence time distribution in the packed section of the column
which arises from dispersion effects and channelling. Oncesomemixingmodelis adopted
to describe the measured data, several methods are available to adjust the mixing para-
meters involved inthe model in order to obtain agreement with the experimental data.
These methods include time-domain, s-plane and frequency domain curve fitting tech-
niques (Johnston, 1971;Abbi & Gunn, 1976), as well as graphical procedures (Vergnes,
1975; Gupta, 1974a). Since inthe present experimental study the residence time distribu-
tion can originate from viscous fingering, wallflow, axial dispersion etc.,nomixing model
was assumed beforehand. Instead,the residence time distribution inthe packed section of
the apparatus is characterized by the relative variance of residence times in this column
section, (O/T)IQ1, where the variance and the mean residence time are given by respec-
tively:

a2 =ƒ 0 - T ) 2 E ( r ) d f (4.23)
o9
and: CO

T = ƒ tE(t)dt (4.24)
o

Later on, parameters involved inmixing models can be adjusted toproduce the same
value of T and (OIT)2CO1, when model fitting should be required. {olr)2col can easily be
derived from the relative variance of the frequency distribution of residence times E t (r)
of the total flow system, (CT/T)J . The latter can be calculated from the moments of the
distribution function, which are defined as: *
oo

Mn =f t"E(t)dt (4.25)
o

Combining Eqns(4.23- 4.25),we obtain the well-known relation:

M,

where M0 equals 1ifa normalized frequency distribution is used inthe calculation. The

69
procedure followed to calculate the moments from the measured residence time distribu-
tion function is discussed at length in Appendix F. The relative variance caused by the
packedsection(O/T)IO1 can be calculated from (.G\T)\ by making use of the additivity of
variance and mean residence times for cascaded equipment. Since for a complete mixed
region(ojif equals 1,the following relation holds:

(OIT)IO1 = (OIT)\ (1+1,+ hf • % - &

(4.27)

where %xis the ratio of the mean residence times in the bottom mixer and the packed
section, and £2 represents the samevalue for the upper mixed region.
As a base case the breakthrough curve for a steep concentration jump using an inert
packing material was measured. The results are shown in Fig. 40, where the value of
2/(O/T)IO1 is plotted against the dimensionless displacement velocity <v>/v s , represent-
ing the ratio of inert to viscous forces in the packed bed. For single phase flow through
packed beds, where the residence time distribution is determined by axial dispersion that
can be described with an eddy diffusion coefficient E>E L the following relation holds
(van der Laan, 1958):

1 + exp {-Peh)} (4.28)

ax. disp.
Pe2, h

Single phase flow PeL-measurements revealed that Peh = 0.25 (L/Dp). This value was
virtually independent of the /?e-number in the range where the experiments were per-
formed (Re =p<v>DJß < 10). This behaviour in the low -Re-number regime is in
agreement with literature data (Miller & King, 1966; Levenspiel, 1957). For the high

Pe,apparent

125-

l
i
Xo I
X|
75- *}JDo
t <x
x

x xx
X
V
25- XyÄ
x
x
Fig. 40. Efficiency of miscible liquid-liquid
displacement from packed beds: theeffect ofa
XX concentration gradientin the displacing liquid.
?! isratio of thevolumeof the gradientmixer
—I—
10 100 1000 and the pore volume of the packed sectionof
<v> the column; Column: length 0.7 m, diameter
v« 0.07m;Key: x - J, =0,o - £, =0.21.

70
PeL-numbers encountered in the experiments the above relation reduces to:

(4.29)
Pe,
ix.disp.

or PeL =2/(o/T)2 . For miscible displacement stabilized by gravity segregation somewhat

higherPeL numbers canbe measured resulting from the favourable densitydifference. The
Per number, characterizing the transversal or radial mixing in the column, can be esti-
mated from Eqn 4.13. For the high values of PeT calculated from this relation and
because of the steep concentration gradient that is applied at the liquid inlet end of the
bed, no stabilisation of the flow by transverse mixing isexpected ascan beverified from
Eqn 4.12. For displacement velocities below v s ,the maximum stable displacement veloc-
ity, we thus expect Peh values of about 125. In the unstable flow regime,where <v>/v s
> 1, viscous fingering is likely to occur, so that the variance of the residence time
distribution curve increases considerably. For very high values of <v>/v s , the effect of
gravity forces becomes negligible, and the growth growth rate of the disturbances is
essentially independent of the displacement velocity (Scheidegger, 1960). This can be
verified easily from a simple model shown in Fig. 41. Lumping all viscous fingers into a
single hypothetical parallel channel and assuming the same pressure drop over the parallel
channel and the 'main' flow, we can derive the following relation for the ratio of the
velocity inside and outside the channel from the Darcy Law:

Vi Mya + 1
(a+1) —a (4.30)
v2
My(a +1) y(M-l)
R
<V>aß
where a =Ai/A2, y = KX/K^, M = JU 2 /MI, and R = —, the ratio of viscous and
intertial forces. In essence, this equation is identical to the fractional flow equation
derived by Dougherty (1963), expressed in Eqn 4.18. The value of the relative variance
{aJT)2 of the residence time distribution function shown in Fig.41,resembling the pulse

|<V>
v,( LL-j^ E(t)
fco 1»

<J>v«<v>(A,*A2)

Fig. 41. Two-stream model for the description of liquid-

liquid displacement from inert packed beds in theunstable
flow regime;and thehypotheticalpulseresponseofthistwo-
streammodel.

71
response of such atwo parallel streammodel,isgiven by:

a 1 ,
1+a 1+ a
2 (4.31)
a
parr. 1
1+a 1 -fa

where we made use of the relation T 2 /TI = Vi/v2 = T. Thus, for high displacement
velocities <v>, the value of T is virtually independent of theR number. Eqn 4.31 shows
that then (a/r) 2 reaches an asymptotic value that is determined by the parameters y,M
and a. In Section 4.1.2 it wasmentioned that a reaches aconstant value at high displace-
ment velocities; Miscible displacement experiments in packed bedsthat were designed to
confirm this asymptotic value are reported in Appendix G. Thus for high R values the
relative variance of the distribution curve reaches a constant value. From the simplified
model wealsolearn that (O/T)IO1 isindependent of the column lengthwhen the residence
time distribution is controlled by channelling, while for flow of a liquid with uniform
physical properties through packed beds, the value of 2/(a/r) 2 o ) is known to increase
linearly with the column length in the high/teL-number range, as can be verified from
Eqn 4.29. Experimental results for two different column lengths are shown in Fig. 42.
The results arein qualitative agreement with the simple flow model.
From Eqn 4.12 it can be derived that the onset of channelling isnot retarded signifi-
cantly when a concentration gradient of the form c =c0 exp(—t/Tm) is applied at the
column inlet, and Tm is 21%of the mean residence time in the packed section of the
column. From the simple flow model we expect, however, that the growth rate of the
fingers will be attenuated due to a decrease of the effective viscosity ratio M. This

Ped.Om)
Pe(0.5m)

2.0

1.0-
ci o
o

10 100
<V_>
V«

Fig. 42. Effect of column length on the apparent Péclet

number in the unstable flow regime. The columns used
had diameter 0.054 m, and length 0.5 and 1.0 m,respec-
tively. ( . ) Dispersion is the controlling mecha-
nism; ( ) Channelling is the controlling mechanism.

72
parameter represents the ratio of the average viscosity in the channel and in the main
flow. Due to this decreased finger growth rate, the variance of the breakthrough curveis
reduced, as is shown in Fig.40.Applying an evenlesssteep concentration gradient in the
displacing liquid causes considerable problems in the data analysis. From Eqn 4.27, it is
observed that the contribution of the mixing zones inthe top andbottom section of the
apparatus to the total variance in residence time isquite significant. The small additional
variance caused by the packed section of the apparatus isthus more difficult to determine
accurately for higher r m values. This fundamental problem is discussed by Otto &Ge-
strich, (1974), who proposed the assemblage of a mixer-column-mixer system for resi-
dencetime distribution measurements in singlephase flow.
From the above experiments it isconcluded that inthe interpretation of the measure-
ments, the ratio of the actual and the maximum stable displacement velocity can beused
to discriminate between astable and unstable flow regime.Although asmooth concentra-
tion gradient with an exponential decaying solute concentration attenuates viscous finger
growth rate, channelling is not completely inhibited. The displacement efficiency in the
unstable flow regime canbe described qualitatively with avery simplemodelbased on the
Darcy Law.

4.2.2 Effect of masstransferfrom thepacking to the liquid onflow stability

The effect of mass transfer on flow stability was studied with the aid of a porous
packing. The experimental setup was quite similar to the one used with aninert packing
material, see Fig. 39 sub a. These porous particles wereimpregnated with a concentrated
sugar solution before the experiment. Dependent on the initial condition that was re-
quested the interstitial liquid was either washed out of the column with pure water, or
the liquid remained in the column until equilibrium between theinterparticle liquid and
the pore liquid was attained. Subsequently the breakthrough curveswere recorded when
pure water was injected in the apparatus. In allcases,the flow direction was downwards.
Properties of the packing material and details of the experimental procedure aregivenin
Appendix E. In order to detect flow instabilities in this system, the measured break-
through curves were compared with theoretical extraction curvesobtained from asimple
mathematical model. For plug flow of liquid through a fixed bed, such a model isgiven
by Eqns 2.12 to2.15, where aconstant mass transfer coefficient is assumed. These
equations were slightly modified to account for axial dispersion. Thefinalresult reads:

be 1 92c dc „ , .

ye=-ND(u-c) (4.33)

where Pe =<v>L/E)E L and the other dimensionlessvariables aredefined in Eqns 2.6 to

2.11. Theinitial and boundary conditions that apply in our caseare:

73
 0=0 c =s0 and o;=1 z>0 (4.34)
e>o c =0 z =0 (4.35)
9c/9z =0 z =1 (4.36)

In Eqn 4.34, s 0 embodies the initialsaturation of the interstitial liquid. Fors 0 = 1,initial
equilibrium existsbetween the pore liquid inside and outside the porousparticle.Whens0
= 0, the external pore liquid is free from the solute.To simplify the mathematics involved
insolvingthese equationsthe second boundary condition hasbeen replaced by:

d >0 c= s0 =finite z=°° (4.37)

Extrapolating from similar dispersion problems itisexpected that the effect of this
modification on the final expression c =c (0) issmall for the high Peclet numbers
encountered in the present study under stable flow conditions (Gerson & Nir, 1969).
Adjustment of the parameters involved inthe model to match the predicted breakthrough
curve with the measured data is achieved by moment analysis. The zeroth and the first
moments calculated from the analytical solution of Eqns 4.32 to 4.37 aregivenbelow:

M c 0 = ( l + s 0 D)/D (4.38)

1
M
c l -
c,l ,
D2 h+(-2+Je!(1+D){l +SoD) (4.39)

Theindex cimpliesthat themoments pertain to the breakthrough curve at the exit of the
packed section of the apparatus. ThemomentsMm {of the actually measured concentra-
tion-time curve in the mixed sample region are related to the moments ofthe break-
through curve at the exit of the packed section of the column Mc ;.Themassbalance on
the mixed region with volumeVm reads:

dc m
Vm - ^ - =* v ( c c - c m ) (4.40)

where the subscript m refers to the mixed region and cto the exit of the packed section
of the column. Introducing dimensionless variables 6 = t/(Vcolh/<j>y) and % = Vm/Vcolh
leads to the following equation:
dc
m
S^-=c c -c m (4.41)

After Laplace transformation and some rearrangement we obtain:

ce=cmi^s+l)-Uo (4-42)

From this equation the moments are readily obtained by the well-known method:

74
 . 3'"c(s)
M,. =lim (-1)' — (4.43)
s^o ds1
with the following result:

M
c,o = M m , o - ^ o *' = 0 (4.44)
M
c*=Mm,/-iÇMmi/_1 i>0 (4.45)

Combining Eqns 4.38 and 4.39 with Eqns 4.44 and 4.45 gives the normalizedfirstmo-
ment of the breakthrough curve:

M 1>m _ D> h+\2+Tel(1+D)(1+SoD)

V
ju 1 +s0D ' J+l (4.46)
M
o,m D° +6a0

Thus, the value of the Peclet number can be calculated oncethe other parametersD and
TVare given. Since D is only determined by the geometry of the packing, it remains
unchanged for all the runs. For stable flow, the Pe number can be estimated from
previous measurements, see Section 4.2,to be Pe = 100.For the samereason asdiscussed
in Section 4.2.1. this value is essentially independent of the Re number in the low Re
number regime where the measurements are carried out, as long asdispersion isthe only
mechanism that causes deviations from plug flow. Estimation of the other parametersD
and TV from stable flow displacements is discussed in Appendix H.The results show that
D=0.9, andTV=0.0125 T, when stable flow occurs.
Experiments.were performed with initial saturation s0 = 0 and s0 = 1, over a wide
range of displacement velocities and initial concentration levels. The measured break-
through curves were reduced to their zeroth and first moment. The Pe number was
determined from these assuming the theoretically predicted value for TV. Also, the ap-
parent number of transfer units TV is calculated from Eqn 4.46 assuming Pe- 100. The
results are summarized in Table 10. The Pe and TV values contain essentially the same
/
information; from Eqn 4.39 it canbe derived that:

=l+N(l+D)(l+SoD)(—^--^) . (4.47)

Therefore, only the Pe number will be considered in the further analysis. When chan-
nelling occurs, some fraction of the soluteisrecovered at alater stage of extraction. Thus
the first moment Mj m increases accordingly, resulting in a decrease of the apparent Pe
number, see Eqn 4.39.Thuswhen the Pe number falls below 100,or alternativelyTV/TVref
below 1,itcan be concluded that viscous fingering contributes to the residence time
distribution, together with axial dispersion and finite masstransfer rates.
The results of the s0 = 1 runs are shown in Fig. 43 and Table 10. The dimensiónless
flow velocity along thehorizontal axisiscalculated from Eqn 4.6.Theminimum value of

75
Pe apparent

100-

10

Fig. 43. Effect of hydrodynamic instabilities

in systems with simultaneous mass transfer on
10 100 1000 the apparent Péclet number: pore liquid ini-
<v> tially saturated (s0 =1).( ) Stabledisplace-
x g (dp/ d|i)min mentlimit.

(dp/dp) isinsertedinthisequation,whichisthevalueattained attheinitialconcentration

inboth the particles and the interstitial liquid.Thefigure clearlyshowsthatthePenum-
berisfar below 100.The existence offlowinstabilities isindicated by breakthrough of
displacingliquid before t/r =1 andbydeformation oftheregulars-shapeinthemoreun-
stable cases.Due to excessivetailingofthebreakthroughcurve,especiallyintheunstable
flow regime,thecurvesweresometimestruncatedbeforethesolutewascompletelyrecover-
ed.Thistailing can originate from limited masstransfer rates,but alsofrom inclusionof
displaced liquid when the leading edges of viscousfingerscoalesce (Perkins, 1965).In-
spection of Eqn 4.39 showsthat when the breakthroughcurveisprematurelytruncated,
thevalueofPeisoverestimated.ThedatashowninFig.43areconservativeinthesecases.
From the figure it canthusbeconcludedthatthestabilizingeffect ofmasstransferisnot
sufficient to prevent viscous finger growth in case s0 =1 under the prevailingexperi-
mentalconditions.
Fig.44 and Table 10showthe resultsof thes0 =0 runs.Intheinterpretationofthe
breakthroughcurvesanadditionalproblemarisesherecomparedwiththepreviouscase.In
the prewash step there issome diffusional lossof thesolutewhenthesucrosesolutionis
removed from theinterstitial voids of the packing.Since thislossisnot knownexactly,
the initial interparticle concentration hasto be determined from the zeroth moment of
the breakthroughcurve.Onlywhenthesoluteiscompletelyrecoveredfrom thebed,does
thevalue thus obtained equal the actualinitial concentration. Themaximumconcentra-
tion that occurs inside the packed column determines the value of the dimensionless
displacement velocity alongthehorizontal axis of Fig.44.Thismaximumconcentration
iscalculated from the solutionofthemasstransfermodelforapackedbed,Eqns2.12to
2.15. InthiscalculationthetheoreticalnumberoftransferunitsN= 0.0125r isassumed.
Thismethodwasbelievedtogiveamorereliablereference conditionforthe(dp/djit)-term
in the expression for the stablevelocity than the concentration at themaximum of the
actually measured breakthrough curve.Eventhough thismeasured maximum equalsthe
maximum concentration obtained in thepacked sectionofthecolumnunderstableflow
conditions,seeEqn 4.40,it may be considerablylowerwhenradialconcentration differ-
ences occur since these differences are efficiently smoothed in the sample region.The
results of the calculations show lessthen 20%difference betweenthemaximumconcen-

76
Table 10. Resultsof displacementrunswith simultaneousmass transfer.

». = 1

Run Initial Mean <v> N Pe

liquid residence "s tfref
cone. timer (1) (1) (1)
(kg/m 3 ) (s)

1 600 560 22 0.22 6.5

2 350 560 4 0.32 10.5
3 200 560 1.7 0.27 8.7
4 620 350 45 0.24 4.6
5 140 350 2 0.40 9.5
6 615 200 70 0.28 3.3
7 590 97 65 0.45 3.5
8 550 67 130 0.77 9.3
9 111 67 9 0.91 24.0
10 700 38 500 0.69 3.7
11 370 38 60 0.74 4.6
12 190 38 30 0.54 2.0

y„=o
Run Theore- Mean <v> N Pe Initial
tical residence ^ref particle
maximum timer (1) (1) (1) cone. ui0
liquid cone. (s) (kg/m 3 )

1 187 590 1.4 0.43 9.4 360

2 128 590 1.1 0.43 9.6 240
3 73 590 0.9 0.45 10.0 140
4 259 195 6.7 0.80 15.0 500
5 173 195 4.1 0.74 12.0 330
6 150 195 3.5 0.76 13.0 290
7 108 195 3.0 0.74 12.0 210
8 71 195 2.5 0.81 17.0 140
9 212 98 10.2 0.79 8.1 450
10 179 98 8.7 0.68 4.8 380
11 156 98 7.7 0.73 6.0 330
12 152 98 7.7 0.76 6.9 325
13 98 98 5.6 0.70 5.2 210
14 154 47 15.5 0.75 3.3 425
15 133 38 17.0 0.74 2.5 430
16 131 38 17.0 0.80 3.3 425

77
Fe,apparent

100

10-<bo
Xpo °

Fig. 44. Effect of hydrodynamic instabilities

in systems with simultaneous mass transfer on
10 100 1Ô00 the apparent Péclet number: pore liquid ini-
<v> tially free from solute (s 0 = 0). ( ) Stable
x g ( d p / d u ) ('est displacement limit.

tration obtained from the measurements and the calculated value.Thefinalresults of the
s0 = 0 runs, shown in Fig. 44, show that considerable effects of channelling are observed
in the s 0 = 0 case where miscible displacement isaccompagnied by transfer of aviscosity
increasingsolute.
In conclusion, without putting emphasis on the numerical value of the results, the
experiments clearly show that in systems where there is simultaneous miscible displace-
ment and mass transfer of a solute that increases the viscosity of the liquid, viscous
fingers may develop and reduce extraction efficiency.

78
5 Conclusions

From mathematical models the effect oflocalphase contact in countercurrent separa-

tion cascades is established for diffusion batteries and belt type solid-liquid extractors.
The resultsarepresented asconcise correlations between the number of exterior apparent
and true transfer units. These correlations show that the separation efficiency of a diffu-
sion battery with 4 or more columns in series is within 10%of a purely countercurrent
extractor under normal operating conditions. Similarly, a belt extractor asshown in Fig.
12 performs much like a countercurrent extractor when it contains 6or more extraction
stages. The beneficial effect of dripping sections is quantitatively discussed. When the
solvent is recirculated in each section, complete mixing of the solvent in the stages can
often be assumed. Liquid recirculation restricts the separation efficiency of an extractor,
especially when the.number of transfer units per section is high and mixing inthe liquid
phasebecomes controlling.
Asymptotic values of the mass transfer coefficient in the dispersed phase have been
calculated for purely cocurrent and countercurrent extraction. A simplified design pro-
cedure based on these values is proposed. Only a minor error is introduced when the
asymptotic values are used directly in the solutions of the lumped parameter model for
countercurrent extraction. The error in the estimated residence time required for aspeci-
fied separation in such an extractor decreaseswithincreasing dimensionless Fourier time,
decreasing ratio of mass transfer resistanceinsideand outsidetheparticle orBiot number
and decreasing extraction factor. For cocurrent extraction the use of asymptotic transfer-
coefficients in combination with the solutions of thelumped parameter models seemsless
promising; for the high Biot numbers that normally prevailin solid-liquid extractors, the
critical Fourier value above whichthe asymptotic solution appliesisonly inthe region of
practical interest for a high extraction factor and slab geometry. For those cases where
the approach just mentioned fails, an accurate short-cut calculation method is proposed
which combines the penetration and asymptotic solutions. ,
Hydrodynamic instabilities, i.e. viscous fingering, can occur in solid liquid extraction
systems, in spite of the stabilizing effect of viscosity gradients in the liquid phase. Such,
gradients, rather than viscosityjumps in the liquid flowing through adensely packed bed,
significantly reduce the channelling intensity in the unstable flow regime. However, a
significant effect on the point of onset of flow instabilities is not observed. For inert
packing materials, the onset can be predicted from simple stability criteria reported in
literature. More experimental results arerequired to warrant aconclusion about the onset
of channelling during displacement with simultaneous mass transfer. However,inthe un-
stable flowregimeviscousfingeringconsiderablyreducestheoverallmasstransfer efficiency
of the extractor. Finally it was concluded that the preferent flow along the wall of a
packed column can be very muchhigher inthe unstable flow regime than onewould esti-
mate from the usual correlations that apply for liquidswithuniform physical properties.

79
Summary

Large scale solid-liquid extractors often consist of countercurrently cascaded mass

transfer sections. The local mode of phase contact within each section usually differs
from the overall countercurrent. A design procedure for solid-liquid extractors, which
incorporates the mode of phase contact in the extraction stages, is presented for a
diffusion battery or Shanks extractor, and for a rotary orbelt type extractor. The effect
of dripping sections, which are often applied to prevent interstage liquid entrainment, is
included in the analysis. The mathematical models developed in this study are based on
the equations describing mass transfer in a single-stage fixed bed. The results are reduced
to ready-to-use correlations that can be applied over a wide range of process conditions.
The estimation of mass transfer coefficients is an essential part of the current design
procedures for solid-liquid extractors. Asymptotic values of the transfer coefficients are
calculated for cocurrent and countercurrent extraction of particles with simple geometry.
It is shown that these asymptotic values are a convenient tool in design calculations,
especially for countercurrent extraction. For countercurrent extraction, an explicit semi-
theoretical equation for the asymptotic mass transfer coefficient is derived which applies
for an extraction factor greater than 0.5.Very accurate short-cut calculation methods are
suggested based on the combination of mass transfer rates derived from the penetration
theory and the asymptotic solutions.
In percolation type extractors, which are often used for large scale extraction pro-
cesses, the solid particles form a densely packed bed. In this study the effect of non-uni-
form liquid flow through this layer of solids on extractor performance isstudied. Special
attention is given to channelling phenomena caused by differences in viscosity and den-
sity in the solvent. Measurements are performed to establish the effect of both the
concentration gradients in the liquid phase and the mass transfer between the packing
material and the solvent on the stability of the flow. The results qualitatively show that
channelling caused by viscosity gradients in the solvent flow direction can occur in many
solid-liquid extraction systems. The effect of this phenomenon on the extractor perfor-
mance wasstudied experimentally for asingle-stage fixed-bed extractor.

80
Samenvatting

Extractie van vaste materialen met behulp van een oplosmiddel iseen,met namevoor
de voedingsmiddelenindustrie, belangrijke unit operation. Dit onderzoek richt zich op
twee aspecten van het extractieproces. Eerst wordt ingegaan op het dimensioneren van
extractieapparaten, die gebruikt worden voor processen op grote schaal. In het tweede
deel wordt de stroming van het oplosmiddel door een laag vanvaste deeltjes bestudeerd.
Extractieprocessen op grote schaal worden vaak uitgevoerd in apparatuur die bestaat
uit in tegenstroom geschakelde secties, waarin de vaste en vloeistoffase met elkaar in
contact worden gebracht. Het fasecontact in elke sectie zalinhet algemeen afwijken van
zuiver tegenstroom. Twee voor de voedingsmiddelen industrie belangrijke voorbeelden
hiervan zijn de diffusiebatterij of Shanks-extractor, en de band-type-extracteur. Aan de
hand van mathematische modellen is het extractierendement van deze apparaten verge-
leken met het rendement van een pure tegenstroomextracteur. De bij deze berekeningen
gebruikte modellen zijn gebaseerd op de vergelijkingen die het instationair stoftransport
in een gepakte kolom beschrijven. Zowel voor de diffusie batterij als voor verschillende
uitvoeringsvormen van deband-type-extracteur zijn de resultatenverwerkt tot correlaties,
die over een groot gebiedvan extractiecondities bruikbaar zijn.
Voor het berekenen van extractieprocessen kan de lokale stoftransportsnelheid vaak
voldoende nauwkeurig worden beschreven met een constante stofoverdrachtscoëffïciënt.
Voor het bepalen van devoor een gespecificeerd extractierendement benodigde apparaat-
grootte iseennauwkeurige schattingvan dewaardevan deze overdrachtscoëfficiënt essen-
tieel. Voor zowel meestroom- als tegenstroomextractie van vlakke, cylindervormige of
bolvormige deeltjes is de asymptotische waarde berekend die de stofoverdrachtscoëffï-
ciënt na lange contacttijden aanneemt. Het gebruik van deze asymptotische waarde in
ontwerpberekeningen wordt besproken. Vooralvoor tegenstroomextractie isde afwijking
van de uit de exacte oplossingen berekende apparaatgrootte klein. Voor dit praktisch
interessante geval is een semi-theoretische vergelijking opgesteld, waarmee de asympto-
tische waardevan destofoverdrachtscoëfficiënt direct kan worden berekend.
In vast-vloeistof-extracteurs van het percolatietype vormt de vaste fase tijdens het,
transport door het apparaat een stationair gepakt bed. Het effect van niet-uniforme
stroming van het oplosmiddel door de laag vaste deeltjes op het extractierendement is
experimenteel bepaald in een gepakte kolom. Behalve door inhomogeniteit van de pak-
king kan kanaalvorming optreden ten gevolge van viscositeits- en dichtheidsverschillen in
de extractievloeistof. Met name aan dit laatste aspect is in dit onderzoek aandacht be-
steed. Zowel de invloed van een concentratiegradient in de vloeistoffase als van stof-
transport van de vaste deeltjes naar devloeistof op destroming door delaagvaste deeltjes
is bestudeerd. De resultaten laten zien dat kanaalvorming ten gevolge van concentratie-
verschillen in de vloeistoffase de scheidende werking van een extracteur nadelig kan
beïnvloeden.

81
References

Abbi,Y.P.& D.J. Gunn, 1976.Trans.Instn.Chem.Engrs.54: 225-231.

Acrivos,A., 1956.Ind. Engng.Chem.48:703.
Anzelius,A., 1926.Angew. Math.u.Mech.6:291.
Backstrom D.M., 1935.Int.Vet. Akad. Handlinger. 136:.
Bakal,A.,G.Timbers& K.Hayakawa, 1970.Can.Inst.FdTechnol. I.3:76.
Barbouteau, I., 1956.RevueInst. Fr. Petrol. 11: 358-388.
Barker, J.J., 1965.Ind. Engng.Chem.57:43-51.
Beckenbach, E.F., 1961.Modern Mathematics for the Engineer, McGrawHillN.Y.p.350.
Beek, W.J. &K.M.K. Muttzall, 1972. Paper presented at the Int. Symposium Heat and MassTransfer
Problemsin Food Eng.,Paper D5,Wageningen,the Netherlands.
Beek,W.J.& K.M.K. Muttzall, 1975.Transport Phenomena.Wiley&Sons,London.
Benham,A.L.& R.W.Olsen, 1963.Soc.Pet. Engrs.J.June: 138-144.
Berg,C.v.d., 1972.DechemaMonogr. 70:27.
Bernardini, E., 1973.Oiland Fat Technology.Technologie s.r.i.Rome. p.228.
Bernardini, E., 1976.J. Am.Oü Chem. Soc.53:275-278.
Bichsel,B.,S.Gâl&R. Signer, 1976.J. FdTechnol. 11: 637-646.
Blackwell,W.E.,J.R. Rayne&W.M. Terry, 1959.Trans.Am.Inst.Min.metall.Engrs.216: 1-8.
Bomben, J.L., E.L.Durkee,E.Lowe& G.E. Secor, 1974.J. Fd Sei.39:260-263.
Brigham,W.E.,P.W.Reed & J.N.Dew, 1961.Soc.Pet. Engrs.J. March: 1-7.
BrünicheOlsen, H., 1962. Solidliquid extraction. Nyt Nordish Verlag Copenhagen.
Brüniche Olsen, H., 1969. SugarTechnol. Rev. 1: 3-42.
Carslaw,H.S.& J.C.Jaeger, 1959.Conduction of Heat in Solids.Oxford UniversityPress.
Chen, J.W., F.L. Cunningham &J.A. Buege, 1972.Ind. Engng.Chem. ProcessDes.Dev. 11: 430-444.
Chhatwal, S.S., R.L.Cox,D.W.Green & B.Ghandi, 1973.Wat. Resour. Res.9: 1369-1377.
Chuoke, R.L.,P.v.Meurs& C.v.d. Poel, 1959.Trans.Am.Inst. Min.metall.Engrs. 216: 188-194.
Claridge,E.L., 1972. Soc.Petr. Engrs.J. April: 143-155.
Colburn, A.P., 1939.Trans. Am.Inst. Chem.Engrs.35:211.
Cooney,D.O., 1974.A.I.Ch.E.Jl. 20:1010-1011.
Cornish, A.R.H., 1965.Trans.Instn. Chem.Engrs. 43:T332-T333.
Crank,J., 1956.TheMathematics of Diffusion. Oxford University Press.
Dombrowski,H.S.& L.E.Brownell, 1954.Ind. Engng.Chem.46: 1207-1219.
Dougherty, EX., 1963.Soc.Pet.Engrs.J.June:155-163.
Dumoré,J.H., 1964. Soc.Pet.Engrs.J. Dec:357-362.
Dwivedi,P.N.& S.N. Upadhyay, 1977.Ind. Engng.Chem. ProcessDes.Dev. 16:157-165.
Eder, B.D., 1971.Ph.D.Thesis,St.Univ.East Lansing, Michigan,USA.
Edeskuty, F.J. & N.R.Amundson, 1952.Ind. Engng.Chem.44: 1698-1703.
Fan, H.P.& J.C. Morris, 1948.Ind. Engng.Chem.40: 195-199.
Farritor, F.J. &L.C.Tao, 1972.Int.J. Heat MassTransfer 15:1179-1183.
Fornwalt, H.J.&R.A. Hutchins, 1966.Chem. Engng.9: 155.
Frössling,N., 1938.Beitr. Geophys.52: 170-216.
Furnas,CG, 1930.Trans.Am.Inst.Chem.Engrs.24:142-193.
Gangwal, S.K., R.R. Hudgins & P.L. Silveston, 1977. Proceedings Europ. Congr. Particle Systems,
Neurenberg, March 1977,prl-pr29.
Gardner,G.C., 1966.Ind. Engng.Chem.ProcessDes.Dev.5: 275-280.
Gardner, K.& J. Taborek, 1977.A.I.Ch.E.Jl. 23:777-786.

82
Gerson,N.D.&A.Nir, 1969.Wat. Resour. Res.5: 830-838.
Gibüaro, L.G.& A.A.H.Drinkenburg, 1972.Chem. Engng. Sei.27: 445^47.
Gillespie,B.M.,E.D.Crandall& J.J. Carberry, 1968.A.I.Ch.EJl. 14:483.
Gunn,D.J., 1970,Chem. Engng. Sei. 25:53-66.
Gupta, S.P.&R.A.Greenkorn, 1974a.Wat.Resour.Res.10:839-846.
Gupta, S.P. & R.A.Greenkorn, 1974b.Wat.Resour. Res. 10:371-374.
Hartland, S.& J.C.Mecklenburg, 1966.Chem.Engng. Sei. 21: 1209-1221.
Hawthorne, R.G., 1960.Trans.Am. Inst. Min.metall.Engrs.219:81-87.
Hayakawa, K., 1975.Lebensm.Wiss.Technol. 8:231-233.
Heller,J.P., 1966.J. Appl.Phys.37: 1566-1579.
Higbie,R., 1935.Trans.Am.Inst. Chem.Engrs.31: 365.
Hill, S., 1952.Chem. Engng. Sei. 1:247-253.
Hutchins, R.P., 1976.J. Am.OilChem. Soc.53:279-282.
Jeschar, R., 1966.Arch.EisenhüttWes.37: 193-200.
Johnston, J.L., L.T. Fan&Y.S.Wu, 1971.Ind. Engng.Chem. ProcessDes.Dev. 10:425-431.
Kasten,P.R.&N.R. Amundson, 1952.Ind.Engng. Chem.44:1704-1711.
Kerkhof, P.J.A.M.& H.A.C.Thijssen, 1974.Chem. Engng. Sei.29: 1427-1434.
King,P.J. & E.Denizman, 1972.Chem.Engng.May: 177-181.
Klaus,R., R.C. Aiken& D.W.T.Rippin, 1977.A.I.Ch.EJl. 23:579-586.
Klinkenberg, A., 1954.Ind. Engng.Chem.46: 2285.
Klinkenberg,A.& A.Harmens, 1960.Chem.Engn. Sei.11:260-266.
Kondratiev, G.M., 1964.RegularThermal Regime,Gostekhizdat, Moscow.
Koval,E.J., 1963.Soc.Pet. Engrs.J.June: 145-154.
Krasuk, J.H., J.L. Lombardi & C.D. Ostrovsky, 1967. Ind. Engng. Chem. Process Des. Dev. 6:
187-195.
Kumar, S., S.N. Upadhyay & K.K. Mathur, 1977.Ind. Engng.Chem.ProcessDes.Dev. 16: 1-8.
Kyle,CR. &R.L. Perrine, 1965.Soc.Pet.Engrs.J. Sept: 189-195.
Laan, E.T.v.d., 1958.Chem. Engng. Sei. 7: 187-191.
Lapwood, E.R., 1948.Proc. Camb.phil. Soc.math. phys. Sei.44:508.
Levenspiel,O.&W.K. Smith, 1957.Chem.Engng.Sei.6: 227-233.
Luikov, A.V., 1968.Analytical HeatDiffusion Theory. AcademicPress,N.Y.
Mandelbaum, J.A. &U.Böhm, 1973.Chem. Engng. Sei. 28:569-576.
Martin,H., 1978.Chem. Engng. Sei.33:913-919.
McGaw,D.R., 1976.Powder Technol. 13: 231.
Mecklenburg,J.C.& S.Hartland, 1968.Chem.Engng. Sei. 23:1421-1430.
Mikhailov, M.D., 1966. Forsch. Ing.Wes.32: 101-110, 147-150.
Mikhailov,M.D., 1973.Int.J. Heat MassTransfer 16:2155-2164.
Mikhailov, M.D., 1977.Int.J. Heat MassTransfer 20: 1409-1415.
Miller,J.S.F.& C.J. King, 1966.A.I.Ch.EJl. 12:767-773.
Milligan,E.D., 1976.J. Am.OilChem. Soc.53:286-290.
/
Miyauchi,T., 1971.J. Chem. Engng.Jpn.4: 238-245.
Miyauchi,T., H.Kataoka& T.Kikuchi, 1976a.Chem.Engng. Sei. 31: 9-13.
Miyauchi,T.,T.Kikuchi& K. Hsu, 1976b. Chem. Engng.Sei. 31:493-498.
Miyauchi,T.,K.Matsumoto &T. Yoshida, 1975.J. Chem.Engng.Jpn. 8: 228-232.
Miyauchi,T.& T.Vermeulen, 1963.Ind. Engng.Chem.Fund. 2: 113-126.
Molyneaux,F., 1975.FdTechnol. Aust. 231-234.
Mungan, N., 1971.Can.J. Chem.Engng.49:32.
Munro,W.D.&N.R. Amundson, 1950.Ind. Engng.Chem.42: 1481-1488.
Muskat,M., 1949.PhysicalPrinciples of OilProduction. McGrawHill,N.Y.
Nelson, P.A.& T.R. Galloway, 1975.Chem.Engng.Sei.30: 1-6.
Neretnieks, I., 1974. SvenskPapptidn. 11: 407-411.
Neretnieks,I., 1975.Chem. Ing.Techn.47:773.
Nusselt,W., 1930.Tech. Math.Therm. 1:47.
Olçer,N.Y., 1964.Int. J. Heat MassTransfer. 7:307-314.
Oplatka, G., 1954. Z.Zuck Ind. 79: 471.

83
 Otto, E.& W.Gestrich, 1974.Chem.Engng. Sei.29: 1294-1296.
Outmans, H.D., 1963.J. Geophys. Res.68:5735-5737.
Paynter, H.M., 1957.Pioc. Symposium on Control, Heidelbergsept. 1956.p.243.
Peachman,D.W.& H.H.Ratchford, 1962.Soc.Pet.Engrs.J. Dec:338.
Perkins,K., O.C.Johnston &R.N.Hoffman, 1965.Soc.Pet.Engrs.J. Dec:301-317.
Perrine, R.L., 1961.Soc.Pet.Eng.J. March: 9-16.
Pfeffer, R.& J. Happel, 1964.A.I.Ch.E.Jl. 10:605-611.
Pflug, I.J.& J.L. Blaisdell, 1963.ASHRAEJ. 5:33-49.
Pflug, I.J., P.J. Fellers& D.Gurevitz, 1967.Fd Technol. 21: 90-94.
Pillai,K.K., 1977.Chem.Engng. Sei.32: 59-61.
Plachco, F.P.& M.E.Lago, 1972.Can.J.Chem.Engng.50:611-615.
Plachco,F.P.& M.E.Lago, 1975.Ind. Engng.Chem.Process Des.Dev. 15:361-365.
Plachco,F.P.& J.H. Krasuk, 1970.Ind. Engng.Chem.ProcessDes.Dev.9:419433.
Plachco,F.P.& J.H. Krasuk, 1972.Chem. Engng.Sei.27:221-226.
Pratt, H.R.C., 1975.Ind. Engng.Chem.ProcessDes.Dev. 14: 74-80.
Pratt, H.R.C., 1976.Ind. Engng.Chem.ProcessDes.Dev.15:544.
Rein, P.W., 1976.Sug.J. Dec: 15-22.
Richardson, R.G., 1961.Handbook of Fluid Dynamics.McGrawHillN.Y.,sect. 16.
Rickles, R.N., 1965.Chem.Engng. 15:157.
Rietema, K., 1976. Fysische Transport en Overdrachtsverschijnselen. Het Spectrum Utrecht, Prisma
Technica 60.p. 253.
RoetzelW. &F.J.L. Nicole, 1975.J. Heat.Transfer. Febr: 5.
Rosenbrock, H.H. &C. Storey, 1966. Computational Techniques for Chemical Engineers. Pergamon
Press,p.127.
Rutov, D.G., 1958.Annexe Bull.Int. Inst. Refr. (Moscow). 2:415-421.
Saffman, P.G.& G.Taylor, 1958.Proc. R. Soc.A245:312-329.
Scheidegger,A.E., 1960.Phys.Fluids 3:94-104.
Schoeber,W.J.A.H., 1976.Ph.D.Thesis,Eindhoven University of Technology.
Schowalter,W.R., 1965.A.I.Ch.E.Jl. 11:99-105.
Schwartzberg, H.A. &M. Desai, 1977. Paper no. 2054 Mass. Agr. Exp. Station, University of Massa-
chusetts.
Sideman, S., 1966.Adv.Chem.Engng.6: 207-286.
Sideman, S.& H. Shabtai, 1964.Can.J. Chem.Engng.June: 107-117.
Silin, P.M., 1964.Technology of Beet Sugar Refining. IsraelProgr.for Sei.Transi., Jerusalem.
Sivetz, M.S., 1963.Coffee ProcessingTechnology, AVIWestport Connecticut.
Sleicher, CA., 1959.A.I.Ch.E.Jl.5: 145-149.
Sleicher,CA., 1960.A.I.Ch.E.Jl. 6: 529-530.
Slobod, R.L.& W.A.Howlett, 1964. Soc.Pet.Engrs.J. March: 1-9.
Slobod, R.L.&S.J. Lestz, 1960.Producers Mon.Penn.Bil Prod. Ass.Aug: 12-19.
Stanek, V.& J. Szekeley, 1972.Can.J. Chem. Engng.50: 9-14.
Stemerding, S.& F.J. Zuiderweg, 1963.Trans.Instn.Chem.Engrs.,d e m . Eng.(London) CE156.
Suzuki,M.,1975.J. Chem. Engng.Jpn. 8: 163-165.
Svedberg, U.G., 1976.Chem.Engng.Sei.31: 345.
Tettamanti,K.,J. Manczinger,J. Hunek& R. Stomfai, 1975.ActaChim.Hung.85:27-47.
Thibodeaux, L.J., 1969.Chem.Engng. 2: 165.
Thibodeaux, L.J., D.R. Daner, A. Kimura, J.D.Millican& R.J. Parikh, 1977.Ind. Eng.Chem. Process
Des.Dev. 16: 325.
Thijssen, H.A.C., 1969. Lecture notes Mass Transfer Processes, Eindhoven University of Technology.
Tolic,A.,V.Rod&V.Jevtovic, 1973.Glasn.Hem.Dvust, Beogr.38:571-602.
Tondeur, D., 1970.Chim Ind. 103:2799-2808.
Treyball,R.E., 1968.MassTransfer Operations. 2nd. ed., McGrawHillN.Y.
Upadhyay, S.N.& G.Tripathi, 1975.J. Sei.Ind. Res.34:10.
Vergnes,F., 1976.Chem.Engng. Sei.31: 88-90.
Vorstman, M.A.G. & H.A.C Thijssen, 1971.Proc.Int. Solv.Extr.Conf. ISEC,April 1971.TheHague,
the Netherlands, Paper 166,p.1071.

84

t~ - -
Vorstman, M.A.G. &H.A.C. Thijssen, 1972. ProceedingsInt. Symposium Heat MassTransf. Problems
in Fd Eng.,Wageningen, the Netherlands. PaperD4.
Wakao,N., 1976.Chem.Engng. Sei.31:1115-1122.
Wakao,N.&T. Funazkri, 1978.Chem.Engng.Sei.33: 1375-1384.
Wartman,R.& H.Mertes, 1966.Arch.EisenhüttWes.37:201-207.
Watson,J.S.& H.D.Cochran, 1971.Ind. Engng.Chem.ProcessDes.Dev. 10:83.
Wooding,R.A., 1969.J. Fluid Mech.39:477495.
Wucherpfenning, K., P.Possman&T.Özmen, 1976.Flüss.Obst 43: 119-121.
Yang,H.H.& J.C.Brier, 1958.A.I.Ch.E.Jl.4:453-459.

85
Appendices

Appendix A. Numericalsolutionof theequationsdescribingmasstransferinafixedbed

InSection 2.2.1, the equations describingmasstransfer inafixed bed were derived:

9c 9c
+ = JVt>c(u-c) (2.12)
ae W
9co
= -NtteD(u-c) (2.13)
9~0

Solving these equations by finite difference techniques gives accurate results only when
the number of true transfer units,A^c is high and the step sizeusedin the integration is
very small. The reason for this unfortunate situation is that any discontinuity in the
concentration profiles in the liquid phase is suppressed by numerical dispersion (Rosen-
brock &Storey, 1966).Therefore, Acrivos(1956) suggested the method of characteristics
for solving the above equations. The principle of this method isthat first the position of
the solid and liquid elements are calculated as a function of time. The solution of the
equations 2.12 and 2.13 is then calculated along these curves. Since to = co(f,0) andc=
cQ>>0)°ne canwrite:

dto=(9to/9?) •df + (9co/90) • dö (Al)

dc = ( 9 c / 9 f ) - d f + (9c/90)-d0 (A.2)

The above equations can be considered as four simultaneous equations in the unknowns
9c/9f, 9c/90, 9to/9f and 9co/90. These equations can be summarized inmatrix notation
as [A] •b=c, or

1 1 0 0 3c/3f JV-tjC(co-c)
0 0 0 1 9c/90 -NUcD(co-c)
(A.3)
df d0 0 0 9w/9? dc
0 0 df d0 9co/90 dco

The characteristic directions can be obtained by setting the determinant of [A]equal to

zero (Beckenbach, 1961). Thus the two characteristic curves, indicated by I and II, are
obtained:

86
 ^ =1 (A.4)
de J

iL :0 (A.5)
de
The characteristic grid for the hyperbolic system (2.12) and (2.13) isshown in Fig. A.l.
The differential equations that hold along the characteristics can be obtained by
combining the origional PDE with the equations for the characteristic curves. For the
derivative ofcwith respect to time alongcharacteristic Ithe result is:

dc
=^ t , c ( « - 0 (A.6)

Similarly, the derivative of cowith respect to time along characteristic II is obtained by

combining Eqns2.13 and A.6:

dcol
Tf\
ao
=-NttCD(u-c) (A.7)
• ii

Eqn (A.6) is solved numerically along the characteristic I, by applying the following
finite difference approximation:

AT/de dc
0
/+l.i+i ">"•<" + 2 W + df
(A.8)
7+i,i'+i /,</
Substitution of (A.6) and rearrangement gives:

h' l | + w / t i +w / + l f , + 1
Ar^t.c
c
/ + i,J + i (A.9)
+i
Ar^ t;C

Similarly the resulting finite difference approximation of dco/dÔalongcharacteristic IIis

obtained:

y'
ftimd s

s
Ae y s

S
S
Fig.Al. Characteristic grid. (- — ) I-characteristic;( )
j (position) Il-characteristic.

87
 co,-,,,-/ -l)+c / + M +c/H

" ƒ + ! . / +! = 7 O' \ (A-10)

+ 1]
lAö7Vt>c£>

The above equations are solved simultaneously. Theinitial conditions (2.14) providethe
valuesofCj 0 andcoj 0 at thebeginningofacycle.Theliquid concentration c0 i isgivenby
Eqn 2.15.The concentration inthesolid phase at%=0,Wj0 , iscalculated from Eqn A.7:
dcj
o,f
= -Z)# t > c (co 0 > ,.-c 0 ) I .) (A.ll)
dô

During the time interval A0the concentration c 0 ; is assumed to remain constant. After
integration of the above equation the following boundary condition at the liquid inlet
sideofthe column results:

w
o , / =c o,/ + ( w o,i-i-*<>,/)• exp(-tf tfC Z>A0) (A.12)

The differential equations (2.12) and(2.13)withtheinitial andboundary conditionscan

thusbereplaced bythefinite difference equations (A.9)and(A.10).

Appendix B. Criteriaforthevalidityofthesimplified modelofabeltextractor

The concentration ineach sectionisnearly constant when theconcentration ofliquid,

sprayed on asection,c i n , andtheconcentration ofliquid leavingthesolidsbeds,c o u t ,do
not differ toomuch. From analytical solutions of Eqns 2.12and2.13,onecancalculate
the liquid concentration at the bedexit at themoment of breakthrough, c b , when the
solidsenteringthesection haveuniform initial concentration overthebedheight.

Cb-Cin =l-e~NU/n (B.l)

whereJV*t r/w, the number of true transfer units on overall liquid phase basis in onesec-
tion,isdefinedas:

AT k„aVI„ NKC !
— -^T'—R <*»
It will be clear that here cb —cin is the maximum concentration difference within a
section.
From Eqn B.l itcanbeseenthat forhigh recirculation ratesR =(j>c r /0 c andlow mass
transfer ratesA^tc , the concentration of liquid inthe bedisnearly constant. Ininterpret-
ing this result it should berealized that ingeneralR isnotanindependent variable,butis
normally adjusted to assure complete filling of the interstitial voids in the packing.A
maximumconcentration difference of,say,10%isattained whenNt Jn R = 0.1, ascan be
seen from Eqn B.l. Assuming no liquid entrainment, A^tc canbe calculated from Eqns

88
Table B.l. Recirculation ratio resulting
in 10% maximum concentration differ-
ence at 90% efficiency.

Number of sections n

A 5 8
1.5 9 5
2 5 3
4 2 1

1.3 and 2.59. For T? = 0.90, the resulting recirculation ratio required to meet the above
condition isshown in Table Bl. The situations areoften met in practice.
The simplified model will not be applicable in extraction of deep beds and when the
number of sectionsislow. For these situations,the equations describing masstransfer ina
fixed bed must be solved to account for concentration gradientsin theliquid phase,and
inthe solid phase perpendicular to the flow direction of the solids.

Appendix C. ApproximaterelationforestimatingShd a values

In Section 3.3.2, it is shown that theShod a value for countercurrent extraction with
anextraction factor A> 1isgiven by:

•S»od,. =— -X? (CD

A,Bi,v A - l V

where Xj is the first non-zero root of the related characteristic equation and the indices
A,Bi and v refer to arbitrary value of A, the Biot number Bi and the geometrical shape
factor v. For alinear equilibrium relation it wasfurther shown in Section 3.1 that:

ShîX =S»öd,. - ( 25/)- 1 (C2)

A,Bi,v A,Bi,v '

Combining the foregoing equations yieldsthe following relation for Sh d,a

^d,a =- l - r - :V--^] (C-3)

A,Bi,v

The aim of thisappendix isto provide asimple method for estimating the first eigenvalue
Xi, thus allowing astraightforward calculation ofShd a . For convenience this approach is
followed for slab geometry. The results for thiscase arethere after extended to the basic
geometries cylinder and sphere.
For infinite A, i.e. a constant bulk concentration, several authors presented simple

89
Table C.I. Approximaterelation for the
firstroot of tan \ , = Bi/\t

X
°°,Bi,l Source

2.5 Bi
Rutov, 1958
Bi + 2.4

2.67 Bi
Backstrom, 1935
Bi + 2.67
(ir2/4)Bi 1.02
Kondratiev, 1964
Bi 102 + 2.24

equations for estimating the first eigenvalue X^Bi x. For slab geometry, some equations
are summarized in Table C.l. In the present analysis the Kondratiev equation was
adopted since it produces the exact solution for infinite Bi numbers. The equation was
usedin asomewhat simplified form, without forcing the accuracy significantly:
_2 D.*
(C4)
K,Bi,^^ 5 / + 2.25

To extend this equation to finite valuesof the extraction factor, it waswritten as:
2 Oj
*A ni i = f(A,Bi) (C.5)
A,BI,I 4 Bi+ 2.2S

Since the Shd a value for a slab subjected to countercurrent extraction with extraction-
factor A = 1is6,irrespective of theBi number, acombination of Eqns C.3 and C.5 setsa
constraint to the function f(A,Bi):
A _ 1
*YA D ^ 1 Bi+ 2.25 A , (ncr.
i(A,Bi)= —— — ; A= 1 (C6)
A j. \_ Bi IT214
3 Bi
or
A „A N „ Bi +2.25 12 .„ „
Ä - T f ( A , 5 0 = P * =- ¥ T r - - ; A=I (c.7)
From the values of f(A,Bi) calculated from Eqn C5the Fig.CI isconstructed. It isnoted
that for any Athe following relation holdswith reasonable accuracy:

A
f v( A , Ä ) = l + 7 ( p * - l ) (C.8)
A-l ' ' A

Thefinal result isobtained from Eqns C3,C5and C8with v= 1

90
 A f(ABi)= U2 -A- Bi+2.25
A - 1 W' ' * A-1 (Ti2/4) Bi

1.20 ^° oo
^ x iao
- / ^ x '
1.10-- s^* ^ A ' ^-b
5

£_---~
1.0 1
V V V v-— V

0.9 '
0 0.25 0.5 0.75 1.0 1.25 Fig. CI. Graphical determination of f(A, Bi);
1/A ParameterBi.

Bi [6(Bi + 2.25) + (A - 1) — (Bi + 3)]

Sh d,a (C.9)
ABi(Bi + 2.25) + ( A - 1 ) [3( 5 / + 2.25) - (5z+ 3) TT2/4]
A,Bi",1

The deviation of Eqn C9 from the exact solution is shown in Fig. C2. The extension to
other geometries is based on approximate equations given by Schoeber (1976)whichhe
used for correlating Shd a values for concentration dependent diffusivities and constant
surface concentration:

Sh d,a 2.01 Sh d,a 4.45 (CIO)

A,Bi,2 A,Bi,\

Sh d,a 3.18 Sh d,a 2.30 (C.ll)

A,S i , 3 A,Bi, 1

ShdQ - S h d a I
100%
Sh"d,a
d

0 1 2 3 4 5 6

Fig.C2. Error bound of theapproximate EqnC.9.

91
These relations were reduced to a single equation, which increased the accuracy slightly
for the caseof variable extraction factor and constant diffusivity:

Sh d.a = " 5»d.. -(«'-D4 (C.12)

A,Bi,v A,Si,1

Ascanbeverified, the approximate equations presented here areexact for A= 1.Also for
extraction factors slightly below 1they canbe usedto estimateShd valueswith reason-
able accuracy.

AppendixD. Masstransferefficiency inthepenetrationperiod

In the following an analytical solution will be derived for the mass transfer problem
for a slab subjected to cocurrent or countercurrent extraction with finite mass transfer
resistancein the continuous phaseinthepenetration period.
Themasstransfer processcanbe describedwith Eqn3.29 with v= 1:

dW d2W
(D.l)
3f„ 3£2
Thedimensionless distancecoordinate isnow defined accordingto Fig.Dl. The boundary
andinitial conditions aregiven by:

fd=0 (D.2)

dW/d£= Bi{W-C(td)} (D.3)

3W/3£= 0 (D.4)

Again,the value of C(td) isobtained from amassbalance,with the following result:

dC dW
=0 (D.5)
ç=o
with the initial condition td = 0, C= 1.The solutionisderived using Laplace transforma-

c.(JÜ
liquid , solid
eft*) "N,

Fig. Dl. Concentration profile near the solid li-

i=o I quidinterface: semiinfinite solid.

92
-op-

tion. Thus, the expression for dW/dl-\0 and W are readily obtained, where Whas been
defined as

dt= àtA (D.6)

0
y ~J 9? £=o
Thefinalrelation for theShod number then reads:

rS*od =

(D.7)

Inthe above equation the following variables are used:

f(x) =ex erfc(x) (D.8)

(D.9)

a2
H--i^} (D.10)

For2?z'->oo, the relation can be reduced to:

vW7 ^
kSh ,=
2 J " ontd
(D.ll)
l-KAt-lMl-ZX-V^/At)}
The average concentration in the solid phase canbe determined from:

Bi
W= — <-l +ƒ(-«! V/ d )}- — .<_1 +/(-a aVTd)> (D.12)
a\ ~a2 ßl «2

Appendix E. Apparatus and experimental procedures: measurement of breakthrough

curves

Apparatus: The experimental setup used in this study is shown in Fig. 39. Two
columns which were kept at a constant temperature were used, one with a diameter of
0.07 m and a length of 0.7 m and asecond with diameter 0.054 m and length 0.5 or 1m.

93
TableEl. Properties of thepackingmaterialusedin theexperimental study.

Property of the packingmaterial Inert packing Masstransfer experiments

diameter (m) 1.4 x 10- 3 1.4 x 10- 3

permeability (m2) 0.49 x 10-' 0.28x 10-9
interstitialvoidfraction (1) 0.35 0.35
effective internal porosity
calculated from measured
value of the distributionratio (1) 0.6
material glassballotini poroussilica
geometry sphere sphere

The properties of the packing material used are summarized in Table E.l. Since the
concentration of the liquid flow leaving the packed section of the column may vary over
the cross section of the bed, a complete mixed sampling region was constructed at the
column outlet. The displacement velocity is kept constant by means of a positive dis-
placement plunger pump. The plunger was driven by an electrical motor with variable
speed, thus allowing easy control of flow velocity. A sample was withdrawn from the
sample region and pumped continuously through a PAAR digital density meter. The
density is measured over a time interval of about 4 seconds. With this time interval asa
minimum, the measured density isstored on paper tape.
For experiments directed towards the effect of a concentration gradient on the flow
stability the configuration shown in Fig. 39b was used. The mixer on top of the column
provides an exponentially decreasing concentration at the inlet of the packed section of
the column, when astep function in concentration isapplied on the inlet stream:

c -c„ t<t>y
exp (E.l)
Co

where c 0 and c„ are the initial and final concentration in the mixer respectively. Vm is
the volume of the mixed section. The resulting concentration gradient travelling down-
wards through the column is thus:

c -c„
•= exp (E.2)
Co

Vk = hAL is the free volume of the column andz' =z/L isthe distance coordinate in the
flow direction. The shape of the concentration profile thus obtained resembles the one
that is expected in countercurrent extractors with plug flow in both phases, when the
extraction factor is greater than 1, see Eqn 1.3. It should be noted that the profile is
independent of the displacement velocity whenplug flow of the liquid occurs.Both from
literature data (Miller & King, 1966; Levenspiel, 1972) and our ownexperiments,itwas
concluded that the axial Pe number is virtually independent of Re in the region of

94
interest. The former conclusion canthusbe extended to stable displacement conditions in
general. This allows us to evaluate easily the effect of displacement velocity on the
hydrodynamic stability of the flow for afixed concentration gradient.
The experiments designed to determine the effect of mass transfer on flow stability
were performed in a set up similar to the one shown in Fig. 39a. The properties of the
packingmaterial are summarized in Table E.1.
Experimental procedure: Experiments run with an inert packing material i.e. glass
beads proceeded in the following way. From a liquid reservoir a glycerol solution was
recirculated through the bed until a uniform concentration wasobserved throughout the
column. In order to assure complete deaeration of the packed section of the column, the
flow was directed upwards. After closing the circulation loop, water was injected witha
constant velocity in a downward direction, and the breakthrough curve was measured.
When a porous packing was applied, the experiments proceeded in three steps. First, a
sugar solution was recirculated through the bed until equilibrium between the liquid in
the poresoftheparticlesand the flowing liquidwasobtained. Thereafter the columnwas
left to drain for several minutes. In the second step, the liquid adhering to the particles
was washed off with water. To maximize the rinsing effect while reducing diffusional
losses of sugar from the interior of the particle,the washliquid wasfed to the column at
a relatively high speed. Measurements of the concentration of sugar in the wash water
indicated that the desired effect was obtained when 5 pore volumes of washwater were
collected. The flow direction wasthen reversed and thebreakthrough curvewasmeasured
at the selected flow velocity. This third step is analogous to the final step in the experi-
mentswhere aninert packingmaterialwasused.
Scouting experiments with inert packing material using dyed solutions indicated that
considerable wall flow occurred in the unstable flow regime: the shape of the break-
through curve showed a sudden decreasewhenbreakthrough at thewalloccurred. Thisis
caused by the higher permeability in this region, a fact that triggers the formation of
protuberances near the wall.Themagnitude of the wallflow wasdependent on the sizeof
the equipment and the displacement conditions. To avoid the above effect, glass beads
were mounted at the wall surface. It was ensured that a glue layer of Tensol cement
occupied the voids fraction up to about half a particle diameter from the wall. Reruns

Fig. EL Effect of preferent flow along the wall on the

variance ofthe residencetime distribution curve.

95
showed a considerable attenuation of the wall flow, see Fig. E.l. At the same time, the
scatter in experimental data was considerably reduced. Reinterpretation of the residence
time distribution curves that were affected bywall flow,which showed asignificant con-
centration jump when breakthrough in the wall region occured, produced results that
werevery closeto the measurements in the walltreated column.

Appendix F. Determination of themomentsof thebreakthroughcurves

Several methods have been suggested in literature to determine the moments of the
distribution function. Direct application of Eqn 4.23 has some significant drawbacks.
Firstly, since in the experiments a cumulative distribution function F(r) is measured,
differentiation of the experimental data is required to obtain the frequency distribution
E(r). Numerical differentiation is a cumbersome procedure that requires careful smooth-
ing of the measured data. Secondly,when the second moment Mj iscalculated,measure-
ments in the tail of the distribution curve may contribute significantly because of the
weighting function f2. Therefore experimental errorsin the trailing edgeof thiscurve are
largely amplified. Procedures that areproposed inliterature try to overcome the foremen-
tioned difficulties. The method adopted here was suggested by Gunn (1970) and by
Paynter (1957). Here the moments are obtained from the Laplace transform of the
frequency distribution of residence times E(t), that can bewritten as:
oo

£<E(f)} = E ( s ) = J s~stE(t)dt (F.l)

o
After seriesexpansion of the exponent andneglectinghigher order terms,whichiscorrect
for smallvalues ofs, we obtain
oo oo oo
2 2
E(s)= ƒ E(t)dt-s ƒ f E ( r ) d r + f s ft E(f)df (F.2)
0 0 0

The moments of the distribution function can be recognized on the right side of this
equation. After some rearrangement the following result isobtained for smallvaluesofs:

1 E(s)-M0 Mi M2
= 1-s v (F.3)
s Mo Mo 2M0 '

Mt and M2 canthus be easily determined by linear regression. Eventhough the range ofs
is restricted to small values, an optimum choice exists. For very low values of s the
truncation error of the computing device may set limits to calculation accuracy. More-
over, for higher s values the weighting function exp(-sr) slightly suppresses the effect of
errors in the tail of the breakthrough curve.Accordingto Gibilaro&Drinkenburg (1972)
the values of s should be selected in the range 0.02 < ST< 0.1,where Tis the mean
residence time inthe flow system. Sincetheweighting function exp(-sr)isalwaysclose to
or smaller than unity, the accuracy in the determination of the moments isconsiderably
increased compared to direct application of Eqn 4.23. In most of the experiments the
cumulative distribution of residence times F(r) is measured. By definition, the relation

96
between E(r)and F(f)is:

E(f) =dF(r)/df (F.4)

Since F(0)=0by definition, taking the Laplace transform of thisequation gives:

E(s)=s F ( s ) (F.5)

where F(s) is defined analogousto Eqn F.l. After substitution of EqnF.5 in Eqn F.3 the
final result readsafter slight rearrangement:

£ {M0 - F(f)}/M 0 =Mi - sM2/2 (F.6)

Note that Mt can be determined directly from this equation when s=0isinserted. The
resultingequation isthen identical to the oneproposed by Van der Laan (1958):

M.„ =ƒ t"E(f)df =- ƒ tnd(Mo - F ) = « ƒ (M0 -¥)tn~l dr (F.7)

For the calculation of M 2 , the weightingfunction t that occursinthe procedure proposed

byvan der Laanisreplaced by exp(-sr),with somegainin accuracy.

Appendix G. Locating the position of the displacement front in downward miscible

liquid-liquiddisplacementfrompackedbeds

In order to getinformation on the behaviour ofviscousfingersduringunstable flow in

liquid-liquid miscible displacement, an experimental setup was designed in which the
location of the displacement front could be monitored. As pointed out in Appendix E,
considerable wall flow may occur duringunstable displacement. For this reason weaimed
at a setup where measurements could be made in the interior of the column packing,
while wall flow was suppressed as much as possible. This technique is likely to provide
more reliable information onviscousfingergrowth thaninterpretation of frontal position
measurements near the wall,seee.g. Chuoke et al.,(1959).
The rectangular column used in these experiments is shown schematically iiiFig. Gl.
In order to suppress preferent flow along the wallasmuch aspossible,apretreatment of
the wall surface as discussed in Appendix Ewasfollowed. The column packingconsisted*
of glass ballotini with a diameter of 0.0014 m. Glycerol solutions are displaced down-
wards with pure water. The location of the displacement front is determined by mea-
suring the conductivity of the interstitial liquid between a common electrode and gold
plated electrodes with the same dimension as the packing material. 100 Of these elec-
trodes were mounted in a 10 x 10matrix in avertical planeinthe centre of the column.
To speed up the sample rate and to simplify recording of the data measured, only the
order of magnitude of the conductivity isdetermined. This signalisconverted to eithera
0 or 1logic level. Thus, at everymoment theliquid concentrations inside the column are
represented by a 10 x 10 matrix of 0 or 1coded elements. The electronics were cali-
brated to produce a 0 signalfor pure water and a 1signalfor the glycerol solution. Every

97
water °
supply - * - ( R -
vessel ^ ^?

front sideview
view

JC
-drain

glycerol
solution
storagevat TU
Fig.Gl. Schematicpicture of the experimental setup.
Dimensions (m) Itemlist
length 0.2 1. goldplated electrode
width 0.1 2. digitizer, 100channels
height 0.5 3. multiplexer, 10channels
a 0.01 4. common electrode
b 0.03 5. digitalprinter
Packing: glassbeads 6. positive displacement pump
diameter 0.0014 7. pulse damper
porosity 0.35 (-) 8. sinewavegenerator

row of 10electrodesthus constitutes abinary number. The decimal representation of this

binary number is stored for further analysis.Thisfigurecanlater beuniquely decoded to
the binary representation. From these data, the timewhen the displacement front passed
a specific electrode can be determined. Byinterpolation, the position of the front canbe
calculated at all times. Totransform these resultsinto smooth curvesfor further analysis,
splineinterpolation wasused. Typicalresultsare shown in Fig.G2for stable displacement
conditions, and in Fig. G3 and Fig. G4 for ambient and strongly unstable conditions,
respectively.
From the measured front positions, the growth rate of the viscous fingers and the
relative channel width were determined. At low displacement velocity, <v>/vc < 3, the
fingers once initiated did not grow in size during the displacement. For higher flow
velocities, the growth rate was essentially constant. The growth rate of the fingers is
further increased with flow rate,until an asymptotic valueisreached where the stabilizing
effect of gravity can be neglected. The results confirm data reported in literature (e.g.
Benham & Olsen, 1963;and others);Most of these data have been obtained from mea-
surements in Hele Shaw models, a hydrodynamic analogon of a packed bed. Channel

98
reduced vertical
position
1600
2200

2800

3400

4000
Fig. G2. Front positions under stable displacement condi-
5200 tions. M = 60; <v>/vc =0.6; <v>= 0.33 x 10"4 m/s;Para-
meter time.

Fig. G3. Front positions under slightly unstable displace-

ment conditions. M = 20;<v>/vc =5;<v>=6.2 x 10""m/s;
Parameter time.

Fig. G4. Front positionsunder unstable displacement condi-

reduced horizontal tions. M =60; <v>/vc = i0;<v> =51.0 x io-4 m/s; Para-
position metertime.

width was only well defined for high viscosity ratios and displacement velocities. The
fraction of the column cross section occupied by the channelsvaried from 0.4 to 0.7 with
increasing displacement velocity. This value compares reasonably with the asymptotic
value of 0.5 for high displacement velocities reported by Safmann &Taylor (1958). For
more ambient displacement conditions the area occupied by the fingers variesin the flow
direction.

99
AppendixH. Experimentaldetermination of modelparametersDand(k0a)

In principle the distribution ratioD canbe determined from properties of the packing
material, i.e. the internal and externalvoid fraction e{andA.Yet wepreferred to measure
the value of D in situ since the internal pore volume might not be occupied completely
with the presaturation liquid. Therefore experiments were designed with relatively low
initial sugar concentration and rather low displacement velocities. Experimental condi-
tions were chosen such that stable displacement could be expected. At the beginning of
the experiment equilibrium conditionsbetween thepackingandthe interstitialliquidwas
assured, s0 = 1. Under this condition the initial solute concentration in both phases is
known exactly and, moreover, mass transfer stabilizes the liquid flow. From the zeroth
moment of the breakthrough curve the value ofD iscalculated with the aid of Eqn 4.36.
The result, averaged over three runs,showed that the distribution ratio equals:

0 =0.90 ±0.05 (H.l)

From the first moment of the breakthrough curve the number of transfer units iscalcu-
lated with Eqn 4.37, assuming that Pe equals 100. From this value the apparent (fc0a)is
determined. The effective interfacial area a is dependent on the packing geometry only
and thus hasthe samevalue for allexperiments.The overallmasstransfer coefficient k0 is
dependent on kc and kd. The former coefficient isdetermined by thephysical properties
of the liquid and by the flow velocity. Fortunately it can be shown from Eqn 3.20 that
the Biot number Bi = mkcDp/E)d is higher than 50, even for the runs with high sugar
concentration and low flow velocity. Thisimpliesthat themasstransfer rate islimited by
interparticle diffusion. Although the value of k0 is thus in principle a function of time
and dependent on the concentration history in the liquid surrounding the particle, a
constant value of ko is assumed here. This figure serves as a reference value for other
experiments. Averaging overthree runswe obtained:

(M)ref =( ^ H e f =0-0 1 2 5 ±0-0015 (s" 1 ) (H.2)

100