Sedimentation of oblate ellipsoids [Elektronische Ressource] / vorgelegt von Frank Rodolfo Fonseca Fonseca
125 Pages
English
Downloading requires you to have access to the YouScribe library
Learn all about the services we offer

Sedimentation of oblate ellipsoids [Elektronische Ressource] / vorgelegt von Frank Rodolfo Fonseca Fonseca

Downloading requires you to have access to the YouScribe library
Learn all about the services we offer
125 Pages
English

Description

Sedimentation of Oblate Ellipsoids¨ ¨Von der Fakultat Mathematik und Physik der Universitat Stuttgartzur Erlangung der Wurde¨ eines Doktors derNaturwissenschaften (Dr. rer. nat.) genehmigte Abhandlungvorgelegt vonFrank Rodolfo Fonseca Fonsecaaus Bogota´ ColombiaHauptberichter: Prof. Dr. H. J. HerrmannMitberichter: Prof. Dr. Ing. Rainer HelmigTag der mundlichen¨ Prufung:¨ 14. Mai 2004Institut fur¨ Computeranwendungen 1 der Universitat¨ Stuttgart2004To Laura Sofia,a little part of heavenon earth ...Contents1 Deutsche Zusammenfassung 11.1 Simulationsmethode . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Die Phasen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3¨1.3 Der stationare¨ Zustand und periodische Phasen: Ahnlichkeitgesetz. . . . . 3¨1.4 Zustandsdiagramm und Ubergange.¨ . . . . . . . . . . . . . . . . . . . . 41.5 Sedimentationsgeschwindigkeit fur¨ oblate Ellipsoide . . . . . . . . . . . 51.6 Orientierungsverhalten . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.7 Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.8 Raumliche¨ Korrelationen . . . . . . . . . . . . . . . . . . . . . . . . . . 8¨1.9 Anderungen im Volumenanteil und Kollaps . . . . . . . . . . . . . . . . 9¨ ¨ ¨1.10 Anderung der Behaltergroße . . . . . . . . . . . . . . . . . . . . . . . . 102 Introduction 132.1 The falling objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.

Subjects

Informations

Published by
Published 01 January 2004
Reads 25
Language English
Document size 2 MB

Exrait

EllipsoidsOblateofSedimentation

VonderFakult¨atMathematikundPhysikderUniversit¨atStuttgart
zurErlangungderW¨urdeeinesDoktorsder
Naturwissenschaften(Dr.rer.nat.)genehmigteAbhandlung

vorgelegtvon

FrankRodolfoFonsecaFonseca

Hauptberichter:Mitberichter:

ausColombiaa´Bogot

Prof.Prof.DrDr.H..-Ing.J.RainerHerrmannHelmig

Tagderm¨undlichenPr¨ufung:14.Mai2004

Institutf¨urComputeranwendungen1derUniversit¨atStuttgart

2004

oT

a

aLaur

little

on

Sofia,

part

earth

...

of

heaven

Contents

1

2

3

4

ZusammenfassungDeutsche1.1Simulationsmethode............................
1.2DiePhasen.................................
¨1.3Derstation¨areZustandundperiodischePhasen:Ahnlichkeitgesetz.....
¨1.4ZustandsdiagrammundUberg¨ange.....................
1.5Sedimentationsgeschwindigkeitf¨uroblateEllipsoide...........
1.6Orientierungsverhalten...........................
1.7Diffusion..................................
1.8R¨aumlicheKorrelationen..........................
¨1.9AnderungenimVolumenanteilundKollaps................
¨1.10AnderungderBeh¨altergr¨oße........................

oductionIntr2.1Thefallingobjects.............................
2.2Manyparticlesedimentation........................
2.3Drivensuspensionandhydrodynamicdispersion.............
2.4Steadysedimentationandthefluidizedbedgeometry...........
2.5LowReynoldsnumberflow.........................
2.6Velocityfluctuationsinhard-spheresedimentation.............
2.6.1CaflischandLuke’swork......................
2.6.2Resumeofexperimentsandsimulations..............
2.6.3Sometheoreticalapproaches....................
2.7Non-Sphericalparticles...........................
2.8Overview..................................

Model3.1Navier-Stokesequations..........................
3.1.1Thegeneralequationforthedynamicsofthefluid........
3.1.2ThedimensionlessformoftheNavier-Stokesequation......
3.2Boundaryconditions............................
3.3Themodel..................................
3.4Contactfunction...............................

Phenomenology4.1Trajectoriesofafallingoblateellipsoid..................
4.2Steady-fallingoblateellipsoid.......................
4.3Oscillatoryoblateellipsoid.........................

i

113345678910

131315161719202021212223

27272829303033

35353739

ii

5

6

7

8

9

Contents

4.4Chaoticoblateellipsoid...........................
4.5ComparisonwithMahadevan´smodel...................
4.6Vortex....................................
4.7ConclusionsandOutlook..........................
Phases5.1Steady-FallingPhase.............................
5.1.1Changeintheinitialfallingheight..................
5.1.2Dependenceonthekinematicviscosity...............
5.1.3Changeintheellipsoidaspect-ratio.................
5.2PeriodicPhase................................
5.2.1Changeintheinitialorientation..................
5.2.2ComparisonwithBelmonte’sresults................
5.3Chaoticregime................................
5.3.1Sensitivitytothechangeintheinitialorientation.........
5.3.2Powerspectra,autocorrelation,Poincaresection..........
5.3.3Lyapunovnumber..........................
5.4ConclusionsandOutlook..........................
transitionsphaseandlawSimilarity6.1Steady-FallingOblateEllipsoid:Similaritylaw...............
6.2Periodicbehavior:Similaritylaw......................
6.3PhaseDiagram................................
6.4TransitionfromSteady-fallingtoOscillatoryphase............
6.5TransitionfromSteady-fallingtochaoticphase...............
6.6ConclusionsandOutlook..........................
articlesPMany7.1Results....................................
7.1.1Sedimentationvelocityforoblateellipsoids............
7.1.2Orientationalbehavior.......................
7.1.3Orientationalchanges........................
7.1.4ModerateReynoldsnumber....................
7.2OutlookandConclusions..........................
Diffusion8.1Introduction.................................
8.2Results....................................
8.2.1Changeindensity,viscosityandaspect-ratio...........
8.2.2Orientationaldiffusion.......................
8.2.3Non-diffusivedynamicalbehavior.................
8.2.4Similarity..............................
8.3OutlookandConclusions..........................
FluctuationselocityV9.1Spatialcorrelations.............................
9.1.1Changeinthevolumefraction...................

404041424343434447484850525253535457575959616263656565697273747777808184859192939394

Contents

9.1.2Collapsingofthespatialcorrelations.
9.2Changeofthecontainersize.........
9.3OutlookandConclusions...........

Conclusion1010.1OneOblateellipsoid.....
sedimentationellipsoidsyMan10.210.3Outlook............

yBibliograph

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

...

iii

9598101

103103104106

109

vi

Contents

1Kapitel

ZusammenfassungDeutsche

Newtonzeigte,dassK¨orpermiteinerkonstantenBeschleunigungaufdieErdefallen,aber
trotzderunleugbarenAnziehungderSchwerkraftbewegensichnichtallefallendenGe-
genst¨andeaufgeraderFlugbahnabw¨arts.DieBetrachtungderFl¨ussigkeitbeinhaltetsehr
schwierigeundnichtlineareInteraktionen.TrotzderbahnbrechendenBem¨uhungendurch
Maxwellallgemeine(1853),ProblemderalsohneerstesL¨osung.dieK¨orper-AndererseitsFl¨ussigkistdieeit-InteraktionSedimentbildungbetrachteteineshat,bliebSystemsdas
vonProblemPartikinelnderineinerHydrodynamikFl¨ussigkeitundinunterderdemStatistischenEinflussderPhysik.SchwerkraftDieseseinProblemsehrhatwichtigesviele
ten,derAnwendungenBiophinysik,dendergrundleKlimaforschunggendenWsowieissenschaftenaufdemwieinGebietdenderLuftftechnischenahrtwissenschaf-chemi-
schenReaktoren,z.B.derAusbreitungvonVerschmutzung,inTintenstrahldruckern,den
druckaufgeladenenWirbelschichtsystemen,etc..DiesesProblemweistschwierigeMul-
tik¨orper-InteraktionenwegenderweitreichendenHydrodynamikauf,dief¨urKugeln¨uber
weinenamyBereich(2001).vTonrozΨ/edieseabf¨allt,breitenwobeieAnwendungmderAbstand¨oglichkzwischeneitenfehltdenPderartikPartikelnist,elsediment-Ramas-
bleibtbildungeinungelweiterhin¨osteseineProblem.statistisch-mechanischeundhydrodynamischeBeschreibungund

Simulationsmethode1.1

DasModellwurdedurchH¨oflerandSchwarzer(2000)entwickelt,erweitertdurchKuu-
selaHerrmannet.al.((1)(2001)2004)undwirdangewinandt.einigenDieBeStudienwevgungonKvonuuselaFl¨et.ussigkal.eiten(2003)wirdundgelF¨ost,onsecaindemand
mandieinkompressiblenNavier-StokesGleichungenaufeinemdiskretenGitterl¨ost:

∂∂iv+Σv∙)v=−p+,E()v+Ω(1.1)
Φ=v∙wterneobeivKraftdiedarstellt,GeschwindigkdieineitderunseremFl¨ussigkProblemeitist,diesowieSchwerkraftpdenist.DruckDieundReΩeineynoldszahlex-

1

2

240(a1)180Position in Y12060

0

(b1)

(I)

(c1)

Simulationsmethode1.1.1

−258118178238298
Position in X

Abbildung1.1:TypischeFalltrajektorieninunserenSimulationen.Wirzeigendasun-
ver¨zillationanderlich-fmitν=allendeΦ.ΦΩ)Re,Δegime,=ΦΣ.NΨ)Ψ´´,.mitΣPΨ)Δ,eund=Φ.dieΩ),ν=chaotischeΦ.Φ´´;BeΣOweΨ)gungdiemitΔperiodischee=Φ.Os-Ω),
ν=Φ.Φ´´.

ES=iΩEMρM/Σν),wobeiidievertikaleoblateEllipsoidgeschwindigkeitist,ΩEMder
gr¨oßteDurchmesserdesoblatenEllipsoids,ρMdieDichteundν=µ/ρMdiekinemati-
istschedabeiViskdieosit¨at(µSchwerkraft.istdieDieScherviskosit¨Grenzbedingungenat).DieFroudezwischenZahlistder>Fle¨=ussigkΣi))eit/ΣUundΣΩEdenM)),ob-g
Fllaten¨ussigkeitaufEllipsoidpartikderPelnartiksindeloberflerf¨¨ullt,achewennabh¨angigmanvinonderBetrachtrutschfreienzieht,dassdieBeGrenzbedingungwegungist,der
vΣx)=vj+rΣx)CM×ω,wobeividietranslationaleGeschwindigkeitdesEllipsoidsist,
rWΣxinkCM)derVelgeschwindigkektorveitondesseinerEllipsoids.MittezumPunktxanderEllipsoidoberfl¨ache,undωdie

DieInteraktionzwischenderellipsoidalenOberfl¨acheundderFl¨ussigkeit,dieandasEl-
lipsoidangrenzt,wirderhalten,indemmaneinewiederherstellendeKrafteinf¨ugt,die
eineDiese“VVerteilungskrafterteilungskraft”inahmtdiedieVGegenwolumenkraftartdesderNaEllipsoidsvier-StokinesdemSinneGleichungnach,vdasserursacht.die
Fl¨ussigkeitinnerhalbdesEllipsoidssichwieeinsteiferK¨orperbewegt.Einenichtreiben-
deKraftwirdausge¨ubt,wenndiePartikelschabloneunddersteifeK¨orpernichtinder
gleichenPositionsind(H¨ofler(2000)).

DierepulsiveKraftzwischendenEllipsoidenwirdproportionalzuihrer¨Uberlappung
gekurzew¨ahlt.AbstW¨ennandedievermeidenoblatendieEllipsoidehnichtydrodynamischen¨uberlappendKr¨afte,sind,dieistdiedasVKraftnull,orhandenseinund¨uberder
Fldiese¨ussigkKrafteitwirdbeschreiben,eineKdenKontaktfunktionontaktgewzwischen¨ahlt,dendiePartikeingehendereln(KinuuselaKet.uuselaal.et.al.(2001)).(2001)F¨ur
undinPerramandRasmussen(1996)erkl¨artwird.DieGeometriedesoblatesEllipso-
idsminimalenwirddurchRadiusseinEbSeitenzumverh¨maximalenaltnisΔegekRadiusEMennzeichnet,,somitΔedefiniert=Ebals/EdasM.Verh¨altnisdes

ZusammenfassungDeutsche

System L"System L

66System LSystem L"System LSystem L5544Vertical velocity vyVertical velocity vy33221100101418225791113
Time t(ts)Time t(ts)

3

Abbildung1.2:AusgangsbedingungenimSystem:θd=Ω+.+Φ,Vd=ΩΩΦ,Δe=Φ.Ψ)und
kinematischeViskosit¨atν=Φ.Φ1´.InderAbbildungstellenwirdievertikaleGeschwin-
digkeitgegendieZeitinbeidenSystemen(durchgezogeneLinie),A,(c=Ω,gestrichelte
LiniemitQuadraten)dar.DieSuperpositionwirddurchAnwendungderUmkehrtranfor-
mationdurchgef¨uhrt,dieinTabelleΨ.Ψbeschriebenwird.
SystemA→SystemASystemA→SystemA
ASdg→c∗ASdgG∗→c∗G∗
EM→c∗EMiSdg→vchor
Eb→c∗EbivEgi→vvcert
U→1∗Rc
Tabelle1.1:TransformationsregelnderSedimentation.cisteinereelleZahl.

PhasenDie1.2

WirfandendreiunterschiedlicheArtenvonBewegung,genanntPhasen,inunserenSimu-
lationen:Derstation¨areZustand,dieseitlicheoderperiodischeBewegung(Belmonteet.
al.(1998))undeinechaotischeBewegung,dieinAbb.1.1gezeigtwerden.

1.3Derstation¨areZustandundperiodischePhasen:
¨eitgesetz.Ahnlichk

DieGrenzgeschwindigkeitiwirdbestimmtdurchdas√Gleichgewichtzwischenden
Tr¨agheits-undViskosit¨atskr¨aften>T=>C,daheristi∼UEb.Wirnehmenan,dassdie
charakteristischePeriodederBewegungvonderseitlichenAbmessungASdgdesBeh¨alters
abh¨angtundvondervertikalenAbmessungunabh¨angigist.Folglichnehmenwiran,dass
diePeriodesichwieG∗∼ASdg/i¨andernsollte.DieTransformationinTabelleΨ.Ψl¨aßtdie
FroudeundReynoldszahlkonstantunddieDynamikindenzweiSystemenist¨aquivalent,

4

1.1.4Zustandsdiagrammund¨Uberg¨ange.

110-1Ref. [14]Ref. [13]This workTumbling10
PSfragreplacementsefghPSfragChaoticreplacementsChaotic
10-2bcd∗I
I∗∗ISteadyPeriodicaIE∗I*100
FallingΨΨ2ΦΦ10-3ΨΨ2ΦΦSteady−Falling Periodic
Ψ3ΦΨ3Φ
ΨΨΨΦ102E103104ΨΦ101Re102103
Abbildung1.3:DielinkeAbbildungzeigtdasZustandsdiagrammderfallendenScheiben
wiesenrauminf¨urBelmontedasfet.allendeal.Ψ221oblateberichtet.EllipsoidIndardes,dasrechtenwirinAbbildungunserenstellenSimulationenwirdererhaltenPha-
haben.

wennsichdiedynamischenVariablenwieinTabelleΨ.Ψ(Abbildung1.2)¨andern.Auch
dieDynamikf¨urdenstation¨arenZustandunddieperiodischenPhasenwerdenaufdieser
Skalaunabh¨angigvoneinander.

1.4Zustandsdiagrammund¨Uberg¨ange.

ImPhasenraumdefinierenwireindimensionslosesTr¨agheitsmomentI=15ρρfellipluidΔe,das
dasVerh¨altnisdesTr¨agheitsmomentsdesoblatenEllipsoidszuderseineskugelf¨ormigen
¨AquivalentsbeidergleichenReynoldszahlES=>)νgM)ist.Esistwichtigzuerw¨ahnen,
dassdasExperimentf¨ureinefallendeScheibemitkleinenSeitenverh¨altnissendurch-
gef¨uhrtwurde,undwirerwarten,dassdieDynamikdesSystems¨ahnlichdereinesoblaten
wird.seinEllipsoidsWennwirunserDiagramminAbb.1.3rechtsmitdenexperimentellenResultatenlinksin
Abb.1.3(Fieldet.al.(1997))vergleichen,sehenwir,dassinbeidenAbbildungendieVer-
teilungderPhasen¨ahnlichist.DieUnterschiedebez¨uglichunseresDiagrammsmitdenen
vonFieldet.al.(1997)sind,dasssieScheibenbenutzenundnichtoblateEllipsoideund
dasTaumel-RegimenichtinunserenResultatenvorkommt.DieKoexistenzderdynami-
schenPhasenistvonderanf¨anglichenOrientierungdesoblatenEllipsoidsunabh¨angig.
LinksinAbbildung1.4zeigenwirdasVerhaltendercharakteristischenZeitG∗/G.Mit
Erh¨ohungdieReynoldszahlESgehtdiecharakteristischeZeitnachNullbeiESB≈´))
(Abb.1.3).NachdiesemPunktfindenwirdieperiodischePhase,diesichwiediePhase
desunver¨∗andlichenFallensmiteinerunendlichencharakteristischenZeitverh¨alt.Folglich
k¨onnenwirG/GalsOrdnungparameterbetrachten,undderKontrollparameterf¨urdie-
sen¨UbergangistdieReynoldszahl.Dieser¨UbergangistwieeinPhasen¨ubergangzweiter
Ordnung.DieinnereAbbildungstelltdasPotenzgesetzmiteinemkritischenExponenten

ZusammenfassungDeutsche

0.039

60

45

*/Ty=0.53*x−6.1*30T/TT0.0340.024/T)*ln(T150.009268.5818.58)ln(Re−Rec0.00911021031000.180.20.220.24
ReRecΔ r Δ rc

5

Abbildung1.4:DielinkeAbbildungstelltdenG∗/GgegendenOrdnungparameterES,
mitEExponentenSB≈´))nahedarbei.ΦDie.).DieinnererechteAbbildungAbbildungzeigtzeigtdasdenG/GPotenzgesetz-V∗gegendenerhaltenmitOrdnungspa-einem
rameterΔe,mitΔeB≈Φ.ΩΩ.
≈Φ.)dar.ImFalldesoberenTeilsdes¨Ubergangs,Abb.1.3,deutetdieVer¨anderungdes
oblatenEllipsoid-Seitenverh¨altnissesdie¨Anderungvoneinenstation¨arenZustandFallen
zueinemperiodischesRegimean.
Derrechts¨Uberdarganggestellt.vomWirstationv¨arenerwendenZustandG/G∗F,allend.h.diezumInversechaotischendeszuvReorgimevwirderwendeteninAbb.Ord-1.4
nungsparameters,umden¨UbergangzubeschreibenundalsKontrollparameterdasSei-
tet.tenvDieerh¨altnisΔecharakteristische.BeiΔeBZeitwirdGein/G∗vendlichererschwindetSprungwegendesdernicht-reOrdnungsparametersgelm¨assigenbebeobach-we-
sind.gungen,Dieserdie¨geUbergengangkleinescheintSchwdaherankungeneinderPhasenanf¨¨uberganglichenangersterOrientierungOrdnungzusehrsein.empfindlich

1.5Sedimentationsgeschwindigkeitf¨uroblateEllipsoide

InderlinkenAbbildung1.5zeigenwirdiemittlerevertikaleSedimentationsgeschwindig-
keitiΣg)alsFunktiondesVolumenanteilsΦVf¨urEllipsoideundKugelnundvergleichen
dannmitdemph¨anomenologischenRichardson-ZakiGesetzvΦ)=ΣΨ−ΦV)c(Richard-
sonandZaki(1954))mitc=).).DerGrenzwertΦV→Φv0entsprichtdemeinzelnen
fallendenEllipsoid,daswirinFonsecaandHerrmann((1)2004)studierten.Esistinter-
essantzuunterstreichen,dassdieSedimentationsgeschwindigkeitdesEllipsoids,diedem
ph¨anomenologischenRichardson-ZakiGesetzfolgt,verglichenmitderder¨aquivalenten
Kugel,kleinist.F¨urEllipsoidegehtdievertikaleMittelgeschwindigkeitdurcheinlokales
MaximumbeiΦV≈Φ.Φ).DiesesMaximumistziemlichinteressant,daesnichtf¨urKu-
gelnbeobachtetwird.¨Ahnlichenicht-monotoneSedimentation¨ubernicht-kugelf¨ormige
K¨orper(z.B.Fasern)istexperimentelldurchHerzhaftandGuazzelli(1999)undf¨urpro-

6

0.9 (vs)||0.6<v>0.3

Richardson−ZakiSim. Oblates3Sim. Spheres

2.4)s(v||<v>

1.8

erhaltenOrientierungsv1.1.6

SphEllip

00.050.1ΦV0.150.200.040.08ΦV0.120.16
dasAbbildungoblate1.5:EllipsoidLinkerPlot(gestricheltezeigtdieLiniemitmittlereQuadraten)unddieSedimentationsgeschwindigkKugel(gestrichelteeitiΣg)Linief¨ur
mitPunkten),alsFunktiondesVolumenanteilΦVbeiES=(∗ΨΦ−).DierechteAbbil-
dungzeigtdiemittlereSedimentationsgeschwindigkeitbeiES≈0.DasoblateEllipsoid-
Seitenverh¨altnisistΔe=Φ.Ω),die¨aquivalentKugelhatEEfjT=Φ.20.

lateEllipsoideinSimulationenvonKuuselaet.al.(2001)berichtetworden.
die¨Abbildungaquivalenten1.5Krechtsugelnstellt(◦)diealsmittlereFunktionvdesertikaleVolumenanteilGeschwindigkbeieitf¨urniedrigerEllipsoideRe(ynoldszahl)und
dar(ES≈7).IneinervorhergehendenArbeitistdieseSimulationsmethodemitErfolg
biszuES≈ΨΦ(H¨oflerandSchwarzer(2000)undKuuselaet.al.(2003))verwendet
worden.DasZwischenmaximumf¨urdieEllipsoidewirdnichtinAbb.1.5rechtsbeob-
achtet,wieinAbb.1.5,linksbeiderniedrigenReynoldszahlgezeigt.EinVergleichmit
demph¨anomenologischenRichardson-ZakiGesetz(durchgezogeneLinieimAbb.1.5
rechts)zeigteinenExponentenvonc<eS=´.Ωf¨urKugelnundcEaaTe=(.Φf¨urEllip-
soide.InbeidenF¨allenfolgendieDatendemRichardson-ZakiGesetzrechtgut.Diese
Exponenten(c=´.Ωundc=(.Φ)liegenzwischendemGrenzwertniedriger
Partikel-T<eSeilchen-Zahlen(c≈(E.)a)aTeundeinemturbulentenPartikelsystem(c≈>Ω.)),
(1954).ZakiandRichardson

erhaltenOrientierungsv1.6

DiemittlerevertikaleOrientierung(MVO)θwirdalsFunktiondesVolumenanteilsin
Abb.1.6linksgezeigt.F¨urkleinereVolumenanteilezeigtdasMVOeinest¨arkereAus-
mitrichtungdermitSchwerkraftderSchwerkraft,beobachtet,undimwelchesGrenzfdemallΦV→OrientierungsvΦwirdeineerhaltenfgenauere¨ureinoblatesAusrichtungEl-
lipsoidentspricht,dasinFonsecaandHerrmann((1)2004)beobachtetwurde.Wirsehen
auchf¨urdasMVOeinZwischenmaximumbeiΦV≈Φ.Φ),welchesdaslokaleMaxi-
mumdervertikalenGeschwindigkeitbeigleichemVolumenanteilerkl¨arenk¨onnte,dasin

ZusammenfassungDeutsche

7

900.50.280)−1>−0.1θ70Θ=<2cos(−0.460Ψ−0.7500.4/2.4−10.4/1.60.4/0.800.05ΦV0.10.1500.05Φ0.1V0.150.2
Abbildung1.6:DieVerteilungsfunktionPΣPdfΣθ))f¨urdiemittlerevertikaleOrientierung
θdief¨urReynoldszahlunterschiedlicheES=V(∗ΨΦ−olumenanteile.).DasOrdnungparameterEllipsoid-SeitenΨalsverhFunktion¨altnisdesistVΔe=Φolumenanteils,.Ω)und
ΦVf¨ur,dreiunterschiedlicheSeitenverh¨altnisseΔe=Φ.(/Ω.(;Φ.(/Ψ.+;Φ.(/Φ.1.

Abb.1.5dargestelltist.DiesesZwischenmaximumexistiertnichtf¨urKugeln.F¨urgr¨oßere
WertedesVolumenanteils(ΦV>Φ.Φ1)zeigtdieAbbildungeinemonotoneAbnahme.
GrUm¨oßedieΨ=σOrientierungΩPdfΣθ)der−Ψ>oblatenein,dieEllipsoideauchinKquantitatiuuselavet.zual.bestimmen(2003)alsf¨uhrenOrdnungspa-wirdie
rameterderOrientierungmitdenWerten−Ψ,Φoder+Ψverwendetwurde,wennalle
oblatenEllipsoidezurSchwerkraftsenkrecht,zuf¨alligorientiertoderentlangderSchwer-
kraftausgerichtetsind.Abbildung1.6rechtszeigtdasVerhaltenvonΨgegenΦV,f¨ur
kleinereVolumenanteilenimmtderOrdnungsparameterΦV≈Φ.ΦΦΨ¨−Φ.Φ1negative
Wertean,wasdieAusrichtungentlangderSchwerkraftbeweistundinUbereinstimmung
mitdemGrenzwert,ΦV→Φ(einoblatesEllipsoid)FonsecaandHerrmann((1)2004),
dieist.Ungeforientierung¨ahrbeiΦdeutetV≈beiΦ.Φ1positiistvderemOrdnungsparameterOrdnungparameterNull.senkrechtF¨urgr¨zuroßeresΦSchwerkraft.V≥Φ.Φ1Im,
woBereichwirveinonΦlokalesV≈Φ.MaximumΦΦΨ−Φin.Φ1,Abbhat.1.6ΨeinrechtslokalesundinAbbMinimum.1.5nahelinksfbeiΦanden.V≈DieΦ.Φ)Si-,
Δg=mulationenΦ.(/Φ.1;Φwurden.(/Ω.(mit,undzweiwiranderenbeobachteten¨unterschiedlichenahnlichesVL¨angenerhalten.verh¨altnissenwiederholt,

Diffusion1.7

Abbildung1.7(a)zeigtdieAbleitungdermittlerenquadratischenTeilchenverschiebung
RE/Rgf¨ursedimentierendeEllipsoide,wobeiEΣg)=σ[kΣg)−ΣkΣΦ)+σi>g)4)>
ist.DieKlammernbedeuten,dasseinMittelwert¨uberdasEnsemblejenerPartikel,die
sichimunterenBereichdesBeh¨altersbefinden,gebildetwird.σi>istdiemittelere
GeschwindigkeitallerPartikelmiti=Φ.DieGraphikenstellendiegr¨oßenAnisotropie
denparallelen()undsenkrechten(⊥)Anteiledar.ImAllgemeinenstellenbeideAnteile

8

replacementsPSfrag

210⊥ll110log(dR/dt)010

110))log(t(ts

1.1.8R¨aumlicheKorrelationen

210

Abbildung1.7:Abbildungzeigtdieparallelen()undsenkrechten(⊥)AnteilederM.S.D.
f¨ursedimentierendeEllipsoideineinerlog-logSkala.DiedickengepunktetenLinienda-
zwischenstellendasWachstumindenballistischenundnicht-diffusivenRegimendar.
DerKugelradiusistEEfjT=Ψ.ΦΨ.DieReynoldszahlistΩ×ΨΦ−)und0dasSeitenverh¨altnis
Δe=Φ.(/Ψ.+.DieZahlderEllipsoideistvonderGr¨oßenordnungΨΦ.
ZeitAdimensionalZeit→igi→sσk)>→1θk,22eqρ
viv→sTabelle1.2:Transformationsregelnf¨urdieSedimentation

einanf¨anglichessogenanntesballistischesRegimedar,wieinAbbildung1.7(a),durch
eineparalleledickegepunkteteLiniezwischendenKurvenbildlichdargestelltist.Dieses
ballistischeRegimeistzug)proportional.Wirfindeneinnicht-diffusivesVerhaltendefi-
αniertEllipsoiddichte,durchEΣg)der≡Vgiskundosit¨atundExponentendemSeitenzwischenverh¨Ψ.altnis(≤abhα¨≤angen.Ω.),derenWertevonder
InTabelle1.2stellenwirdieTranformationsregelndar,dieverwendetwerden,umdie
TGraphikranformationsreeningelnAbbildungunver1.8¨andert¨ubereinandergelassen.zuDieseslegen.ErDiegebnisRegiltf¨urynoldszahlkleinewirdESdurchΨΦ−).die

1.8R¨aumlicheKorrelationen

WirbeginnenunsereAnalyse,indemwirdier¨aumlichenKorrelationenindenGeschwin-
digkonsfunktioneitsfluktuationendesparallelenstudieren(||)(imAnteilsFderolgendenSCVF).GeschwindigkDienormalisierteeitsfluktuationenAutokwerdenorrelati-wie
folgtdefiniert(Segr´eet.al.(1997)):

ZusammenfassungDeutsche

450400400350350300(e)300(f)2502>*2002>*250
<x<x2001501501001.00.4/1.6100500.8/1.61.2/1.6501.41.2
1.61.6/1.60050010001500200000500100015002000
**time ttime t

9

Abbildung1.8:AlleKurvenliegenrelativgut¨ubereinander,wasdieTransformationsre-
gelninTabelleΩrechtfertigt.

,Σr)≡δiΣΦ)δiΣr)(1.2)
δiΣΦ))
wobeidieKlammern...einenEnsembledurchschnitt¨ubermehrereindividellunter-
schiedlicheKonfigurationenimRaumundindenOrientierungen(Ellipsoide)darstellen.
DabeistelltδvT=vT−vhECdieFluktuationeninderGeschwindigkeitundvhEC=iT
dieMittlereGeschwindigkeitderKonfigurationdar.FallsderAbstandrinderRichtung
genommenwird,diezurSchwerkraftparallelist,dannnennenwirdieparalleleKompo-
nente,Σk)=,oderfallssenkrechtdazu,dannnennenwirdiesenkrechteKomponente
,Σz)=,⊥.

1.9¨AnderungenimVolumenanteilundKollaps

Abbildung1.9zeigt,dassalleKurvendesSCVFf¨urKugelnundEllipsoideim(||)Anteil
zurSchwerkraft¨ubereinstimmen.HierbeiwirdderAbstandEEfΦV−(/0benutzt,wievon
¨esSegralse´et.bemerkal.(1997)enswert,vordassgeschlagen.sief¨ur¨BeiderAnderungendesUbereinstimmungVolumenanteilsderKurvbisenzumbetrachten60-fachenwir
g¨ultigist.DievonunsgefundenenKorrelationsl¨angenk¨onnenwiefolgtangegebenwer-
den:ξ⊥,heS=Ω2eEfφ−(/0;ξ,heS=Ψ´eEfφ−(/0,wassichnichterheblichvondenResul-
tatenfindetvonmanSegrauche´et.eineal.sehr(1997)gute¨unterscheidet.UbereinstimmungImderEllipsoidfKurvallen.(sieheDieWerteAbbildungf¨urdie1.9,Korre-(b)),
lationsl¨angesindξ⊥,Eaa=Ω)EEfΦ−(/0;ξ,Eaa=ΨΦEEfΦ−(/0.DieKorrelationsl¨angef¨ur
EllipsoideistinbeidenAnteilenkleineralsf¨urKugeln.
DietuationenAmplitudeistinderAbbildungparallelen1.10()darundgestellt.senkrechtenDie(⊥Graphik)enAnteilewerdenderineinerGeschwindigklog-logeitsfluk-Skala
dargestellt.F¨urΦ.ΦΦ)≤ΦV≤Φ.Φ0sindFluktuationengefundenworden,diesowohlf¨ur

10

1.20.8Sph,||0.4C

0

N()

1.20.0020.0040.0200.1200.8Ell,||0.4C

0

1.1.10¨AnderungderBeh¨altergr¨oße

0.0020.0040.0200.120)O(

replacementsPSfragPSfragreplacements−0.402040(N)−0.402040
−1/3−1/3r/(ReqΦV)r/(ReqΦV)
Abbildung1.9:Kollapsderr¨aumlichenKorrelationsfunktionf¨urden(||)Anteilf¨urKugeln
(b).Ellipsoideund(a)

. δδ||⊥ Ellip Ellip

010. δδ⊥|| Sph Sph . δδ||⊥ Ellip Ellip
−110−1102>Vδ⊥2Vδ>⊥
2>,<Vδ<||2>,<Vδ<||10−2
−210−310−31010−2ΦV10−110−2ΦV10−1
Abbildung1.10:DieoberenAbbildungenstellendier¨aumlichenKorrelationsfunktionen
f¨urKugelnundEllipsoideindenparallelen()undsenkrechten(⊥)AnteilenalsFunktion
)−EdesEfjTV=Ψ.ΦΨolumenanteilsderKinugelneinerunddaslog-logSkalaEllipsoid-Seitendar.DievReerh¨altnisynoldszahlistΔeist=Ω×Φ.(/ΨΦΨ.+,.derRadius

Kugeln,alsauchf¨urEllipsoidewie≈ΦV(/0(diegeradeLinie¨uberdenDaten)wachsen,
(Segre´et.al.(1997)).F¨urgr¨oßereVolumenanteileΦV>Φ.Φ0)werdendieFluktuationen
inbeidenAnteilen,f¨urKugelnundf¨urEllipsoide,verringert.

1.10¨AnderungderBeh¨altergr¨oße
WirstudierendieGeschwindigkeitsfluktuationenwieindervorhergehendenArbeitvon
Segre´et.al.(1997)undver¨anderndieKorrelationsl¨ange,umGr¨oßeneffektezuunter-
suchen.DieResultatesindinderAbbildung1.11f¨urKugeln(gestrichelt-punktierte

ZusammenfassungDeutsche

1.2

0.8>s/v||Vδ<0.4

0.4

>s/v⊥Vδ<0.2

11

0SphEllip0SphEllip
2060L/(ReqΦ100−1/31401804080L/(ReqΦ−1/3)120160
Abbildung1.11:DielinkeAbbildungzeigtdenparallelenAnteilderGeschwindigkeitsf-
luktuationenbeiver¨anderlicherBeh¨altergr¨oße.DierechteAbbildungzeigtdenvertikalen
Anteil.DieReynoldszahlistΩ×ΨΦ−),derRadiusderKugelnistEEfjT=Ψ.ΦΨ,unddas
Ellipsoid-Seitenverh¨altnisistΔe=Φ.(/Ψ.+.

Linie)undEllipsoidegezeigt(durchgezogeneLinie).DieBeh¨altergr¨oßenwerdenmit
A/ΣEEfjΦ−(/0)normalisiert.WieimFallvonSegre´et.al.(1997)undvonH¨ofler(2000)
zeigendieGeschwindigkeitsfluktuationeneineanf¨angliche¨Ubergangsregion,dieeine
starkeAbh¨angigkeitvonderBeh¨altergr¨oße,zwischenΩΦ≤A/ΣEEfΦ−(/0)≤ΨΦΦhaben.
DanachsinddieSimulationsdatenunabh¨angigvonderBeh¨altergr¨oße.ImAllgemeinen
zeigenEllipsoideundKugelndasgleicheVerhalten,abermiteinemkleinerenGesamt-
wertf¨urEllipsoide.WieerwartetwirdistdasVerh¨altnisdesParallelanteilsderGe-
schwindigkeitsfluktuationzumVertikalanteil⊥≈Ω.)f¨urKugelnundEllipsoide,Segre´
et.al.(1997).DergleicheWertf¨urbeidePartikelnformenbeweist,dassdasanisotro-
peVerhaltenderGeschwindigkeitsfluktuationvonderPartikelformunabh¨angigist.Die
durchdieSchwerkraftverursachteSymmetriebrechungwirktgleichermaßenaufKugeln
Ellipsoide.undCaflishandLuke(1985)fanden,dassdieGeschwindigkeitsfluktuationenbeiZunahme
derBeh¨altergr¨oßedivergierenk¨onnen.AndererseitswurdewederinExperimenten,Si-
mulationen,nochinderTheorieeinBeweisf¨ursolcheDivergenzengefunden.InH¨ofler
(2000)wirdargumentiert,dassSysteme,diedurchW¨andebegrenztsind,nichteinever-
gleichbareSkalierungderGeschwindigkeitsfluktuationenaufweisen.Stattdessenzeigen
sieeineS¨attigung,sobalddiekleinsteAusdehnungdesBeh¨alterseinekritischeGr¨oße
¨ubersteigt.EsbestehtdieSchwierigkeit,eineinzigesSkalierungesetzzufinden.Inunse-
renSimulationenwurdedieBeh¨altergr¨oßevariiert,indemmandiegesamtequadratische
Unterseite¨anderte.DieResultateinAbb.9.7(a)und(b)zeigenkeineDivergenzderGe-
schwindigkeitsfluktuationenf¨urKugelnoderEllipsoideundbewegensichsehrnahean
denResultaten,diedurchSegre´et.al.(1997)gegebenwerden.

12

1.1.10

¨Anderung

der

Beh¨altergr¨oße

2Chapter

oductionIntr

objectsfallingThe2.1

toThefollotreewleathevesshortestfluttertopath.theThegroundwayininwhichautumn,eobjectsxhibitingfallatothecomplexgroundmotionhasandbeenrefusingstudied
sinceGreeks.antiquityDuring.ObjectsRenaissance,wereGalileothoughttoGalileireturntodropped“theirtwonatural”metalballsplacesfrombythetheleaningancient
TNeowerwtonofshoPisawed,andthatshothewedthatbodiesthefyallfallonatearththedrisamevenratebyadespiteconstanthavingdifaccelerationferentandmasses.he
alsoobservedthecomplexmotionofobjectsfallinginbothairandwater(VietsandLee
(1971)).

Butdespitegravity’sundeniableattraction,notallfallingobjectstraveldownwardsin
straighttrajectories.Theconsiderationoffluidsurroundingtheobjects(figure2.1),in-
troducesaverycomplicatedandnonlinearinteractionbetweentheobjectandthefluid.
ThefirstpioneeringeffortwasmadebyMaxwell(1853),whowasthefirsttoconsiderthe
fluid-objectinteractionandproposedamodelforafallingpaperstrip.

Inthebeginning,theoreticiansmadefewassumptions
a)constrainedmotionin2-dwastakenintoaccount
c)b)vorticesconsideredinafluidfluidwerewithzeroignoredviscosity
Basedontheseassumptions,GustavKirchhofshowedthattheproblemreducestoasim-
plifiedsetofequationsthatcanbesolvedforsimpleparticlegeometries.Thismethod
alsoappearsintheHoracelamb’sclassictreatiseonhydrodynamics,(Lamb(1932)).

Adeeperunderstandingofthemotionoffallingobjectsinafluidisofgreattechnical
importance,andhasbeeninvestigatedinavarietyofcontexts,includingmeteorology
(Kajikawa(1982)),aircraftstability(Mises(1945)),powergeneration(Lugt(1983)),
chemicalengineering(Marchildonet.al.(1964)),andalsointhestudyofstabilityof
submarinesandthecentrifugationofcellsbybiologicaltechniques.

Inhoffthe’sninetiesequationsArefthatandtheJonestrajectory(1993),ofanfoundobjectbymomeansvingofthroughnumericalansolutionsincompressible,ofinKirch-vis-

13

14

objectsfallingThe2.2.1

Figure2.1:Imagesofthevorticessheddingby(a)arisingairbubble(KellyandWu1997)
theandfluid(b)amotionmetalduestriptoasaitfallsdispersionthroughofwsmallater(S.aluminumFieldetalparticles1998).intheItwisater.possibleThevtoorteseex
shedsineachcasearesynchronizedwiththezigzagmotionoftheobject.

cidandirrotationalfluidischaotic.TanabeandKaneko(1994,95)andMahadevanet.al.
(1995)usedaphenomenologicalmodelforthefallingofapieceofpaperin1-d.Theyin-
cludedtheliftandkinematicviscosity,butneglectedtheinertiaofthefluid,anddescribed
fivefallingregimesofwhichtwowerechaotic.Further,Mahadevan(1996)implemented
anolderworkpresentedinHoraceLamb’sclassictreatiseonhydrodynamicstotheprob-
lemoffallingcardsthattakesintoaccounttheeffectofthefluidasashape-dependent
renormalizationofthemassandthemomentofinertiatensor.Tumblinganddriftmotion
wereobservedwhichareaconsequenceoftheanisotropyinmassandviscosityinthe
model.Furthermore,Mahadevanet.al.(1999)alsomadeanexperimentbydroppingho-
rizontalcardsofthicknessRandwidthjshowingthatthetumblingfrequency,Ωscales
asΩ≈R(/)j−(,consistentwiththedimensionalargumentthatbalancesthedragagainst
.vitygra

inafluidAdditionallyand,Fieldidentifiedet.al.different(1997)invdynamicalestigatedreegimesasxperimentallyafunctiontheofbehatheviorofmomentfallingofinertiadisks
andtheReynoldsnumber.Theyobtainedexperimentalevidenceforchaoticintermittency
f(Bauerallinginet.aal.fluid,(1992)).observedtwBelmonteoet.motions:al.side(1998),tosideinaneoscillationxperiment(flutter)withandthinflatend-ovstripser-
endinertialrotationdragand(tumbling)liftwhichfigureΨ.reproducesΩ.TheythisproposedmotionaandyieldsphenomenologicaltheFroudemodelsimilarityincludingwhich
describesthetransitionfromflutteringtotumblingregime.

Inspiteofthislargeeffort,thegeneralproblemofthefallingofabodyininteractionwith
thesurroundingfluid(figure2.1),remainswithoutsolution.

oductionIntr

15

Figure2.2:Theimagesshowacollageofconsecutivevideofields(Belmonteet.al.1998)
oftwofundamentalmotions:side-to-sideoscillation(flutter)andend-over-endrotation
(tumble).

sedimentationparticleMany2.2

Thesedimentationofasystemofparticlesinafluidundertheactionofgravityisavery
importantprobleminfluiddynamicsandstatisticalphysics.Thisproblemhasmanyap-
plicationsinbasicsciencessuchasaerospacesciences,biophysics,environment,etc,and
inthefieldofengineeringe.g.chemicalreactors,contaminationspreading,ink-jetprint-
ing,fluidizedbeds,etc.Manynaturalprocessesinvolveparticlesimmersedinfluidsfor
example,bloodflux,particlesintheatmosphere,diffusionstudiesoforganellestransport
inlivingcells,papermaking,sedimentationinriversandlakes.Tounderstandsuchsys-
tems,abetterknowledgeinsedimentationandabilitytosimulatethesedimentbehavior
isverymuchrequired.Withthiswiderangeofapplications,theparticlesedimentation
stilllacksastatisticalmechanicalandfluiddynamicaldescriptionandremainsanopen
problem.

Sedimentationandstatisticalphysicshavealonganddistinguishedcommonhistory,be-
ginningwiththeclassicaltheoretical(Einstein(1906))andexperimental(Perrin(1916))
studiesofBrownianmotion.Theearlyworkdealtprimarilywithsuspensionsatornear
thermalequilibrium.Thismeantthatthesourceoffluctuationsinthesystemisathermal
bathcharacterizedbythetemperature.Thecorrelationandresponsefunctionsofphysical
observablesaretightlylinkedbyfluctuationdissipationrelations.Theinterestofphys-
icistsinthestatisticalmechanicsandhydrodynamicsofsuspensions(Pusey(1997))has
continuedtothepresentdaywiththefocusshiftingprogressivelytotheproblemsofsys-
temsfarfromequilibrium.Theeffectofshearflowonthestructureandcrystallizationof
suspensionshasreceivedagreatdealofattention(AckersonandClark(1984));thecon-

162.2.3Drivensuspensionandhydrodynamicdispersion
whereceptuallythereissimpleronavstateerageofnorelatisedimentationvemotion(Russeloftheet.al.particles.(1989),BlancandGuyon(1991))
2.3Drivensuspensionandhydrodynamicdispersion
Systemspresentingsteadilysedimentingsuspensionsareinanon-equilibriumstateand
thereforehavepropertiesqualitativelydifferentfromthesysteminthermalequilibriumin
twoimportantways.Firstly,thesuspendedparticlesaredenserthanthefluidandalsoin
ordertobalancebetweengravityandviscousdissipationonaveragethereisadownward
relatidisplayvespeed.randomparticleSecondly,motionirrespectievenveofwhenthetheparticlethermalbeingBrowniansedimentedmotionorisnesheared,gligible.they
Eachparticleinfluencestheotherinsuchawaythatthedynamicsishighlysensitiveto
initialconditions.TheresultingchaosasobservedinBradyandBossis(1985)andJ´anosi
et.al.(1997),impliesthatthetime-evolutionofthecoarse-grainedquantitiesmustbe
describeddynamicsinusingthedifabsencefusionofcoefthermalficientsBroandwniannoisemotionsources,iseentirelyventhoughdeterministic.theThismicroscopicphe-
nomenonofdiffusivebehaviorinducedbytheflowduetotheobjectsdriventhroughthe
fluid,intheabsenceofathermalnoiseiscalledhydrodynamicdiffusionorhydrodynamic
dispersion.Thisplacesmanyquestionsaboutthelarge-scalestructureandlong-timedy-
namicsofsedimentingorshearedsuspensionsmainlyinthedomainofnon-equilibrium
statisticalphysicsratherthantraditionalfluiddynamics.
Ingeneral,thesuspendedparticlesareacteduponbyBrownianandotherforces.If<iS
isthethermalBrowniandiffusivityofasoluteparticlewithradiusNinaflowwithtypical
velocitygradientγ˙aroundtheparticle,thenthePS´cletnumberPS,
PS=γ˙N)(2.1)
<iSisdifafusionindimensionlessthesuspension.measureofFortheaparticleimportancemoofvingparticle(foreflowxample,comparedsedimenting)tothewiththermala
speedithroughanunboundedfluid,
i(2.2)γ˙≈NFoforsolvaBroentwniandisplaced)sphereLof=bradius,UNsettlingandbuothroughyantaweightviscous(i.e.fluidweightatminustemperaturetheY.Gweightin
energyunits,thesettlingspeed,ihis:
ih=b,U(2.3)
ΓHere,UisaccelerationduetogravityandΓisthecoefficientofviscousdragofthefluid
(Γ=+πNνforasphere).TheEinsteinrelationtellsusthatthediffusivity,<iS=Y.G/Γ.
Thus,PS=b,UN/Y.Gisindependentofthekineticcoefficientsofthesystem.A

oductionIntr

replacementsPSfragFhpSecNcgNcg
>eSS−fSggaWcU
FSRWbScg
U

Batch2.3:Figuresedimentation.

17

suspensioninwhichPSisextremelylarge,(ΨΦ)ormore)isanon-Browniansuspension.
Insuchsuspensionsthephysicsisdominatedbythedrivingforce(gravityinthecaseof
sedimentation)andhydrodynamicswhereasthermalfluctuationsplayanegligiblerole.

InelectricalcontrastorDethermalGrootandcurrents,Mazurfluctuations(1984),andhnon-equilibriumydrodynamicsteadydiffusistatesvitiesinsuchdriasvenflownon-of
Browniansuspensionsaredeterminedbythedrivingforce.Theyhavenothingtodowith
thethermodynamictemperatureofthesystemandarehencenotconstrainedbyfluctu-
ationsdissipationrelations.EveninsuspensionswithPS=Ψ,therewillbesubstantial
non-equilibriumcontributionstodiffusion,fluctuationsandlinearresponseofthehydro-
interactions.dynamic

2.4Steadysedimentationandthefluidizedbedgeometry

Oneofthemostfamiliaraspectsofsedimentationistheseparation(figure2.3),ofasus-
pensionintosedimentandsupernatantwithafree-settlinglayerinbetween(Russelet.al.
(1989))understeadysedimentation.Thisstatecaninprincipleberealizedbystudying

18

2.2.4Steadysedimentationandthefluidizedbedgeometry

bed.Fluidized2.4:Figure

thefree-settlingregionalone,whilefeedinginparticlesfromthetoptocompensatethose
eouswhichformperpetuallythesediment.settlingstateAniselegtoantmowveaytheofachiereferencevingthisframeideawithinathespatiallysettlinghomogen-particles
inthe’fluidizedgeometry’(referenceXueet.al.(1995))asfollows.Subjectthesus-
pensiontoanupwardflowofspeediΦfrombelow(figure2.4).Thenumberdensity,cΦ
isspatiallyuniformandiscompatiblewiththeflowrate.Forsampleswhoselineardi-
mensionsarelargeinalldirectionsthebehavior,apartfromachangeofreferenceframe,
shouldsettlingbewithidenticalspeediinΦinthethebulktolaboratorythatofaframeincollectionanofunboundedparticlesfluid.withAlthoughnumbernotdensityallctheΦ
experimentsthathavebeenperformedinthefluidized-bedgeometryprobethestatistical
propertiesofsteady-statesedimentation,itisanidealsettingforsuchstudies.Thema-
jorityoftheexperiments,eventhoseperformedinconventionalbatchsedimentationare
arycarried.outThroughoutunderthethemanimplicity-ellipsoidassumptionpartofthatthisthethesisunderlyingwearestateconcernedisstatisticallywiththestation-nature
ofsuchasteadilysedimentingstateandfluctuationsaboutthisstate.

oductionIntr

19

Figure2.5:(a)Apairofparticlessettlingsidebysidesettlesfasterthananisolated
particle;(b)apairwithobliqueseparationvectorsettleswithasmallhorizontalcompon-
enttoitsvelocity.

2.5LowReynoldsnumberflow

TheReynoldsnumberES=IΣΩ∗EM)/νisameasureoftheratioofinertialtoviscous
forcesintheflowofafluidwithν=µ/ρthekinematicviscosityandµtheshearviscosity;
ρthemassdensity;withIandAbeingtheverticalvelocityandthelargestradiusofthe
ellipsoid,respectively.Forthesuspensions,ESrangesfromΨΦ−0upto(dependingupon
thevaluesofρandν.

IfweworkintheStokesianlimit,ES=Φ,severalimportantfeaturesofStokesianflowin
thepresenceofparticles(Russelet.al.(1989))canbesummarized.

(i)Theequationsofthefluidflowinthislimitarelinear.

(ii)AnisolatedsingleparticleofbuoyantweightWsettlingundergravityinanunbounded
containergivesrisetoavelocityfielddecayingasL/ewithdistance,efromtheparticle
toanypointinthefluid.

(iii)tlingAparticleslocalizedlikedensitywiseproducesfluctuationaΨ/eaboutvaelocitybackgroundfluctuation.ofuniformconcentrationofset-

(izero:v)Thetheyrelatineithervevelocityapproach,ofanrecedeisolatedfrompairnorofrotatesettlingaboutparticleseachinotheran.Iftheunboundedystartatfluidtheis
sameheight,theyfalltogether(figure2.5(a))ataspeedgreaterthanthatofanisolated
singleparticlesubjectedtothesameforce.Iftheyareinitiallyseparatedbothvertically
andhorizontally,theircenterofmassfallsnotverticallybutobliquelywiththevelocity
pointinginadirectionbetweentheverticalandthevectorjoiningtheupperparticleto
thatatthelowerposition(figure2.5(b)).

(v)Thedynamicsofthreeormoreparticlesiscomplexandchaotic(J´anosiet.al.(1997)).

20

2.2.6Velocityfluctuationsinhard-spheresedimentation

2.6Velocityfluctuationsinhard-spheresedimentation

AnapproximatetheoryinsedimentationbeginswiththepioneeringworkbyBatchelor
(Batchelor(1970)).Hecomputedtheparticlesedimentationvelocitybyassumingaho-
mogeneousspatialdistributioninthedilutelimittogetthemeanparticlesedimentation
velocity.Thetheorypredictsthatatsmallvolumefraction,φthesedimentationvelocity
be:will

(2.4)

IΣφ)/IΦ=Ψ−+.)φ+DΣφ))(2.4)
withIΦbeingtheequilibriumvelocityofasphere.
Ingeneral,thisproblemrevealsaverycomplicatedmulti-bodyinteractionduetothe
long-rangehydrodynamicinteractionthatdecaysforasphereasΨ/e(Batchelor(1970))
wherenumbere.isThetheparticledistancevfromelocitiesthecanparticlefluctuateforaaroundspatiallytheinfinitemean,systembothatalongsmalltheRegraynoldsvity
(theaboutvtheertical)natureandofthethesevperpendicularelocityfluctuationsdirectionRamasw(horizontalamyplane).(2001).Verylittleisunderstood
Apartfromrecentadvancesinthefieldthisoldsubjectofsuspensiondynamicsposesex-
citingproblemsinthefrontiersofnon-equilibriumstatisticalphysics.Muchprogresshas
eralbeenpuzzlesmadeonremainthisasproblemdivergingbotheviewsdoxperimentallyontheandbasicissues.theoreticallyWe.Atdescribethesameeachtime,problemsev-
firstandoutlinetheprogressmade.

2.6.1CaflischandLuke’swork

AsevereprobleminthestatisticalphysicsofsedimentationwaspointedoutbyCaflisch
andLuke(hereafterCLCaflishandLuke(1985)).Averybriefresumeisgivenhere.
Considerasteadilysedimentingfluid-likesuspension(seesection1.3)ofhardspheres.
Aconcentrationfluctuationneartheorigininthissuspensionisapoint-likeforcedensity
andinthreedimension,givesrisetoavelocityfluctuationdecayingasΨ/ewithebeing
ultingfrommanyspatiallydistributedconcentrationfluctuationsissimplythesum,iT
distancefromtheorigin.ThelinearityofStokesflowimpliesthatthevelocityfieldres-
oftheindividualcontributions.Ifthesefluctuationstakeplaceinarandomandspatially)
uncorrelatedmannerthroughoutthesuspension,theresultingvarianceinvelocity,σvat
anypointinthesuspensionwouldbeclearlythesumofthesquaresoftheindividualcon-
tributions.ThisiT)hasC≈A0termsifthereareCsoluteparticlesinacontainerof
lineardimension,Ainalldirections.TheΨ/eformmentionedaboveforthevelocityfluc-
tuationproducedbyalocalizedconcentrationfluctuationmeansthatiT)≈A−)sothat
σv)≈A.SuchadivergingvarianceintheinfiniteAlimit,posesseriousproblemsinthe
calculations(Batchelor(1970)),ofthemeansettlingspeedinanunboundedsuspension.
MostexperimentsfindsizeindependenceunlikethepredictionbyCaflischandLukebut
thereareseriousquestions(Brenner(1999)),thatcanbeaskedabouttheinterpretationof
themeasurements.Itisfairtosaythatexperimentshaveneitherconfirmedthe,Apredic-
tionsnordefinitelyruledthemout.Itshouldofcoursebenotedthat,Af’predictions’

oductionIntr

21

saythatiftheconcentrationfluctuationsarestatisticallyindependentfromonepointto
anotherinspacethenthevelocityvariancediverges.Sufficientlystronganticorrelations
intheparticleconcentrationfieldatlargelengthscaleswillsuppressthe,Adivergence.
Clearlywhatisnecessaryisatheorythattacklesconcentrationandvelocityfluctuations
onthesamefootinginsteadofpostulatingtheoneandinferringtheother.

2.6.2Resumeofexperimentsandsimulations

Experimentsonthevelocity-fluctuationsproblemuseawiderangeoftechniquesinclud-
ofingthetrackingparticlestheinvanelocityfieldilluminatedbyreparticlegion(Toryimaginget.(Seal.gre´et.(1992)al.and(1997)),Leiet.directal.counting(2001)),
’diftagged’fusing-waparticlesveinaspectroscopsuspensiony(Xueofet.otherwiseal.(1995)),indexmatchedtrackingthespheresmotion(Toryofet.al.individual(1992,
Nicolai(2000)).et.Ithasal.been(1995))claimedandbysingleSegrase´wellet.al.as(1997)multipleandNicolaisound-scatteringand(CoGuazelliwanet.(1995),al.
awithonesize-independentsignificantvealuebxceptionutthisofToryinterpretationet.al.has(1992),beenthatthecriticizedbyfluctuationsBrennersaturate(1999).to
NumericalsimulationsbyLadd(1996)showclearevidenceofsize-dependenceoverthe
theserangeofstudiesAewerexplored,probingalthoughscalesitisarsmallerguedbythanSeagrlare´geet.al.screening(1997)thatlength.thisH¨woflerasbecause(2000)
citypointedoutfluctuations.thattheIfthesmallestsmallestdimensiondimensionoftheiscontainerincreased,thecontrolvtheelocitymagnitudefluctuationsofthevincreaseelo-
uptoalimitandareindependentofthecontainersize.
ThereisalsoaclassofexperimentsbyRouyeret.al.(1999),whichseparatestheprob-
lemofhydrodynamicdiffusionandnon-equilibriumstatisticalbehaviorinfluidizedbeds
fromthequestionofwhetherthevelocityvariancediverges.Thisisaccomplishedby
workingwithasuspensioninaneffectivelytwo-dimensionalgeometry,i.e.withlength,
Aandwidth,Lmuchlargerthanthethickness,δ,andδslightlylargerthantheparticle
size.Thisyieldsasystemwhoselocalhydrodynamicsisthreedimensional,sothathy-
drodynamicdispersiontakeplacewithlong-rangeeffectsincludinganypossibleCaflisch
ityLukedistribdiverutiongenceofthescreenedvelocityoutonfluctuationslateralscalesofh>>yperdifδ.fusiTheveparticlemeasurementsintheseofethexperimentsprobabil-
thestilllackaparticle-imagingtheory.ThevworkelocimetryofXueofSeet.gral.´eet.(1995)al.(1997)usingdifalsofusing-wbroadlyavfeallinspectroscopthiscateygoryand.
Butwiththethesequestioninterestingoftheanddiverimportantgentveelocityxperimentsvariance.doThenotconfinedspecificallyexperimentalconcernthemselvgeometryes
hoandwethever,isnotionof(Serelevgre´anceetal.since(2001))thedeofvanelopmenteffectivofeourtemperaturesimulationsisarecentralmadetothebasedonstochasticthat
P</approachofLevineet.al.(1998).

2.6.3Sometheoreticalapproaches

al.Apart(2000)fromseemthetoideashavinevruledolvingoutparticleprettyorconclusifluidvelyinertia,therewhichhaveebeenxperimentspreciselybyCofourwanthe-et.

22

particlesNon-Spherical2.2.7

(∂Foretical)(KochattemptsandtogoShaqfehbeyond(1991))whatwereCaflischthefirstandtoLukare,gue,Athathavaedone.mechanismKochandanalogousShaqfehto
screeningoftheCoulombinteractioninelectrolytesmightworkinsedimentingsuspen-
sions.dynamicallyTheystartedinteractingfromaparticlesmicroscopicandshowedstatisticalthatdescriptionthree-particleoftheencounterssedimentingcouldhleadydro-to
aentscreeningmeasurementsofthe,Acoulddiverpredictgence.whetherTheyahagivveennot,hosuspensionwever,isscreenedmentionedorthatnot.independ-Brenner
ar(1999)guingthatassumedthethe,AinterpretationmechanismoftheebutxperimentsquestionedbySethegree´et.videnceal.in(1997)favorareofgreatlyscreening,com-
plicatedbytheproximatewallsofthecontainer.Thecoarse-grainedapproachofLevine
et.locityal.fields(1998)ofaconsistssedimentingofthestochasticsuspension.equationsThisofretainsmotiononlyforthosethetermsconcentrationwhichanddominateve-
atlargelengthscalesassumingnorelationsamongstthephenomenologicalparameters
otherunderlyingthanthosetheoriesforcedofbydynamicalthesymmetriescriticalofthephenomenaproblem.(HalperinTheandspiritisHohenberidenticalgto(1977)),those
thehydrodynamicsoforderedphases(Martinet.al.(1972))orindeedthefluctuating
Navier-Stokesequations(LandauandLifshitz(1969)).
Thestationaryimportantdifconfigurationferenceisthatprobabilitiesaresedimentationnotgiisvenabyanon-equilibriumBoltzmann-Gibbssteadystate,distribwhereution
withrespecttoanenergyfunctionbutcanbeobtainedbysolvingtheequationsofmotion.
Thisapproachyieldsaphasediagramforsteadysedimentationcontainingan’un-
screened’phaseinwhichthevelocityvarianceΩidivergesasaasobservedin,A,and
a’nally,Tscreened’ongandphaseAckinersonwhich(1998)Ωimadesaturatesanfor’intriguingA’greaterobservthanationthethatthescreeningmodellengthξequations.Fi-
forthermalconsedimentationvectionatatlarlargegePSPrandtlandsmallandReRayleighynoldsnumbers.numberESTheareidenticalsedimentationtothoseproblemfor
dealswiththeconcentrationfieldwhereastheconvectionproblemdealswiththetem-
peratureRayleigh-Bernardfield.TheyturbthenulencetotranscribedargueforresultsscreeningfromandKraichnan’hence,sforamixing-lengthfinitevelocitytheoryvari-for
anceinsteadylowReynolds-numbersedimentation.Animportantdifferencebetween
TanonguniformandAckersonconcentration,(1998)whileandthewhereaseconxperimentsvectionisisthatdrithevenlatterbyanwereimposedundertakingtemperaturewith
gradient.

2.7particlesNon-Spherical

Asmentionedabove,manyoftheinvestigationshavebeenmadeonspheresandina
reducedmanneronslenderbodies,(e.g.fibers)byRamaswamy(2001)andBatch-
elor(1970).Ellipsoidalparticles(figureΨ.+),havefoundanapplicationinthemodeling
ofthebloodflowOlla(1999).Fibers,ontheotherhand,havenumeroustechnological
applications,thepapermanufacturing,transportandrefiningpetroleum,pharmaceutical
smallerprocessingvolumeandenfractionsvironmentalwhichwisastecharacterizedtreatment..byThetheyshowsedimentationorientationalvelocitytransitionthathasfora

oductionIntr

23

Figure2.6:Snapshotoftheoblateellipsoidsfallinginafluid.Thepictureshowsthe
distance“cluster”isgivformationeninunitsalongofthetheflaralling.gerTheradius,Eellipsoidandtheaspect-ratioReynoldsisΔe=numberΦ.(/isΨ.E),Sthe=
M(∗ΨΦ−).

globalmaximumaroundthetransitioninstationarystate,Kuuselaet.al.(2003).

Withthisbackground,theinvestigationofthesedimentationofnon-sphericalparticles
isfundamental.Anunderstandingofthesettlementmovement,orientationandspatial
distributionofparticlesindrivenflowsiscrucialforthistask.

viewerOv2.8

InthisthesisIpresentanumericalstudyofthedynamicsofoneandmanyfallingob-
lateellipsoidsparticleinaviscousfluid,inthreedimensions,usingaconstrained-force

24

viewerOv2.2.8

Figure2.7:Asectionofatridimensionalfallingoblateellipsoid.Thesystemsizeis
Re)Φ×ynoldsΨ1Φ×)ΦnumbertheisoblateΩ∗ΨΦ−),aspect-ratioandtheisΦratio.Ω/ofΨ.Φtheinfluidunitsofdensitythelarovgerertheellipsoidellipsoidradius.densityThe
.).´is

technique(H¨oflerandSchwarzer(2000);H¨ofler(2000);Kuuselaet.al.(2001)),under
gravitytablishing.Wethestudytypestheoffallingdynamicalmotionsbehaviorandforfindingaatypicaldosimilaritywnwlaardwformotion,regularfig.Ψ.0motions.,es-
Wemotion.proposeInamanmechanismy-particlesfor(ellipsoids)understandthesedimentationtransitions(figurebetweenΨ.+),Ithestudydiftheferentssettlingtypesvofe-
locityandtheaverageorientationoftheellipsoidsasafunctionofthevolumefraction.
Wealsoinvestigatethediffusivebehaviorofasedimentingellipsoidatlowandmoderate
Reellipsoidsynoldsandnumberspheres.Finallyonthe,wesizeshowcontainerthe.dependenceofthesedimentationdynamicsof

oductionIntr

25

Inchapter2webrieflypresentthefluiddynamicalbasisofthemodelandweexplain
thebasicingredientsofthemodel.Inchapter3weshowthebasicphenomenologyof
thefallingofoneoblateellipsoidineachregime(steady-falling,periodic-oscillationand
chaoticmotion)ofthesystemandwecompareourresultswithpreviousworks.Inchapter
4weinvestigatethedynamicsbehindeachregime.Fortheperiodicandsteady-falling
regimewefindasimilaritylawderivedfromtheinvarianceoftheReynoldsandFroude
number.Inthechaoticregimethetrajectoryoftheoblateellipsoidischaracterizedbya
highsensitivitytotinyvariationsintheinitialorientation.
Inchapter5aphasediagramispresentedandcomparedtotheresultsofFieldet.al.
(1997).Thetransitionfromoscillatorytosteady-fallingoccursatESB=´)),where
thetransienttimeofoscillationinthesteady-fallingregimetendstoinfinity,beyondthis
valuethesystemisoscillatory.ThetransienttimehasapowerlawdivergenceatESBwith
anexponentof0.5.Thetransitionfromsteady-fallingtochaoticregimebecomesabrupt,
foracertainvalueoftheellipsoidradii.
Inchapter6westudythesettlingvelocityandtheaverageorientationoftheellipsoidsas
afunctionofvolumefraction.Weseethatthesettlingvelocityshowsalocalmaximum
attheintermediatedensitiesunlikethespheres.Theaverageorientationoftheellipsoids
alsoshowsasimilarlocalmaximumandweobservethatthislocalmaximumdisappears
astheReynoldsnumberisincreased.Also,atsmallvolumefractions,weobservethat
theoblateellipsoidsexhibitanorientationalclusteringeffectinalignmentwithgravity
accompaniedbystrongdensityfluctuations.Theverticalandhorizontalfluctuationsof
theoblateellipsoidsaresmallcomparedtothatofthespheres.
Inchapter7weinvestigatethediffusivebehaviorofsedimentingellipsoidsatlowand
moderateReynoldsnumber.Webeginintroducingthecommontheoreticaltoolsused
tostudythediffusivebehaviorofsedimentingparticles.Wealsodiscusstheresultsfor
ellipsoidsincomparisontotheequivalentspheres.Moreover,westudythebehaviorofthe
sedimentingspheresandellipsoidsundervariationsofthekinematicviscosity,ellipsoid
densityandaspect-ratio.Furthermore,wepresenttheorientationaldiffusionbehavior.
Additionally,weshowtheanomalousdiffusionforellipsoids.
Finallyinchapter8wepresentthedynamicalbehaviorofsedimentingellipsoidsand
spheresundervariationsofthecontainersize.Westudytheinfluenceonthespatialcor-
relationsastheparticulatevolumefractionischanged,comparingtheresultsforellipsoids
andspheres.Wealsopresentthestudyofthevelocityfluctuationsasafunctionofthe
volumefraction.Afterthat,weinvestigatethedivergenceofthevelocityfluctuationsas
changed.issizecontainerthe

26

2.2.8

viewerOv

3Chapter

Model

Thischapterlaysoutthegeneraltechniqueusedformodelingthephysicsofoneand
manysedimentingellipsoids.Weintroducetheimportantquantitiesandterminologies
thatareemployedintherestofthethesis.Inthefirstsectionwepresentbasicresultsof
fluiddynamics.Next,adescriptionoffluidandparticlecouplingispresented.Finally,
theellipsoidcontactmethodfortheellipsoid-ellipsoidinteractionissketched.

3.1Navier-Stokesequations

Fluidisdefinedasthestateofmatterthatcannotsustainanyshearstress.Foranelastic
solid,thestrainincreaseswithtimeandattainsasteadyvaluewhereas,forafluiditin-
creasesindefinitelywithtime.Inthemathematicaldescriptionofafluidflow,theflow
quantitiessuchasvelocity,pressureareassumedtovarycontinuously.Inviewofthe
particlenatureofmatter,thevalidityofthisassumptioncanbequestioned.Ifthemean
freepathofthemoleculesiscomparablewiththecharacteristicdimensionofthemac-
roscopicflow,thecontinuousapproximationbreaksdown.Standingbythiscontinuous
hypothesis,onecanderivetheconservationofmassinthefluidflowbywritingtheglobal
equationforthemassofthefluidinsideafixedvolume.ConsideranarbitraryvolumeK,
fixedinareferenceframeusedfordescribingtheflowoffluidandboundedbyaclosed
surfaceF.Ateveryinstantoftime,fluidentersandexitsfromthisvolume.Therate
ofchangeofthemass,bcontainedwithinthevolumeisequalandoppositetotheflux
leavingtheboundarysurface.Thenwehave
RbR
Rg=RgVρMRK=−<ρMv∙nRF(3.1)
wherenistheoutwardunitvectornormaltothesurface,FoftheboundaryandKisthe
volume.SinceKisfixedwecaninterchangetheorderofintegrationanddifferentiation
withrespecttotime.Further,byapplyingtheGauss’sdivergencetheorem,weobtain:
∂∂ρgM+∙ΣρMv)=Φ(3.2)
27

28

3.3.1Navier-Stokesequations

Makinganexplicitdifferentiationoftheterm∙ΣρMv)andgroupingweget,
ρ∂Σ∂gM+v∙ΣρM))+ρM∙Σv)=Φ(3.3)
Thefirsttermoftheaboveequationisknownastheconvectivederivativecorresponding
totheLagrangiandescriptionandthereforewecanrewrite3.3as
RρM+ρM∙Σv)=Φ(3.4)
Rgwhichshowsconservationofmass.

3.1.1Thegeneralequationforthedynamicsofthefluid
WesolveNewton’sequationofmotion

RgRρMvRτ=ρMΩRτ+[σ4∙nRΣ(3.5)
<VVwhereRτisthedifferentialvolumeoffluid,RΣisthesurfaceelementoftheclosedsur-
face,FthatboundsK,and[σ4isthetensorofallthe(pressureandviscosity)forcesacting
onRΣ.ThevolumeforceΩperunitmassoffluidcouldbethegravitationalforce.
Thewiththederivatifluid,vethenR/RgtheisaproductLagrangianρRτisderivaticonstant,ve,easvitaluatedmovesinthealongthereferencelocalvframeelocitymofieldving
Moftheflow.ApplyingGaussdivergencetheoremtothesecondtermontherighthandside
3.5wecanwriteequation3.5as
Mρ
VRgvRτ=VρMΩRτ+V∙[σ4Rτ(3.6)
Takingthelimitasvolumetendstozeroanddividingbythevalueofthevolumeelement,
weobtainthelocalequationofmotionforaparticleoffluid:
ρMv=ρMΩ+∙[σ4(3.7)
Rg[σ4canbewrittenas[σ4=[σ4−pδTU.
evhaweThen

pδ∂TU∙[σ4)T=∙[σ4T−∂kU

(3.8)

Model

becomes3.7equationThen

29

ρMRv=ρMΩ−p+∙[σ4(3.9)
RgThisequationisapplicabletoanyfluid.IfwestudythemotionofaNewtonianfluidthen
wecanexpressσas

)∙[σ4=µ∂iT+Σξ+µ)∂∂ia(3.10)
∂kT∂kU´∂kT∂ka
Onsubstitutingineq.3.6,weobtaintheequationofmotionforacompressibleorincom-
fluid:wtonianNepressible

ρM∂v+ρMΣv∙)v=ρMΩ−p+µ)v+Σξ+µ)Σ∙v)(3.11)
´g∂Ifthecompressibilityeffectsarenegligibleinthefluidflowthen∙v=Φ.Theresulting
equationistheNavier-Stokesequations:

ρM∂∂gv+ρMΣv∙)v=ρMΩ−p+µ)v(3.12)

3.1.2ThedimensionlessformoftheNavier-Stokesequation
WecanalsowritetheNavier-Stokesequation,eq.3.12,intermsofdimensionlesspara-
meters(thatarelabeledwith’primes’).LetAandIbetherespectivescalingfactors
forthespatialcharacteristiclength(particlediameter)andvelocityofthesedimenting
particle.Thenwehave:

and

r=r,v=v
IA

(3.13)

Fp−p
g=A/I,p=Ψ/ΩρMI)(3.14)
Indefiningp,thevalueofpΦhasbeensubtracted,whichisthehydrostaticpressure.After
dividingeachsidebyρMI)/A,theNavier-Stokesequationbecomes:

∂∂gv+Σv∙)v=−p+EΨS)v+>e)Σ−g)(3.15)

30

conditionsBoundary3.3.2

Intheaboveequation,theinverseoftheRe)ynoldsnumberES=AIρM/µassociatedwith
theflowappearsasafactorofthetermv.Itisfoundthatthisnumberrepresentsthe
ratioofthenon-linearconvectivetermΣv∙)vtotheviscoustermµ)v.Wedefine
>e)=UA/I)astheFroudenumber.Fromtheaboveequation,thevelocityandpressure
fields(vandp)thatsatisfytheappropiateboundaryconditionsforagivenproblem,are
form:theof

v=FΣk,l,z,g,ES,>e)
p=GΣk,l,z,g,ES,>e)

(3.16)(3.17)

where>andGarefunctionsthataredependentontheflow.Wemakeuseofthisin
.)Chapter

Boundary3.2conditions

teThegrationcompleteofthesolutionequationforofthemotionmotionofoftheafluidfluidvparticleselocityandfield,thevΣr,g)specificationincludesofboth,boundarythein-
conditions;i.e.,thevalueofthevariablesatalltheboundariesofthefluid.
Boundaryconditionsvarydependingonwhethertheboundaryissolidorafluid.Inthe
casesolidofarequiressolidthatwallthebeingcomponentaboundaryofthe,vthefelocityactthatnormalthetofluidthecannotboundarypenetratesurfaceintoshouldthe
beequalforthefluidandthesolid:

modelThe3.3

viodad∙n=vflkad∙n

(3.18)

Thegeneralideaofourapproach,proposedbyFogelsonandPeskin(Fogelson(1988)),
istoworkwithasimplegridbyresolvingthefluidmotionatalltimesandrepresentthe
particlesnotasboundaryconditionstothefluid,butbyavolumeforcetermorLagrange
multipliersintheNavier-Stokesequations.Thistechniquewasdevelopedintheworkby
Schwarzeret.al.(H¨oflerandSchwarzer(2000)),H¨ofler(2000),Kuuselaet.al.(Kuusela
et.al.(2001)),Wachmann,etal.(WachmannandSchwarzer(1998)).Thisemploys
anumericalsolverforthedynamicalsimulationofthree-dimensionalrigidparticlesina
Newtonianfluid,boundedbyarectangularcontainer.Theequation3.12isdiscretizedona
regular,marker-and-cellmeshtosecondorderprecisioninspace.Forthetimeevolution,
weemployanoperator-splitting-techniquewhichisexplicitandaccuratetofirstorder.
Thesuspendingfluidissubjectedtono-slipboundaryconditionsatthesurfaceofthe
particles.MoredetailsarepresentedinKuuselaet.al.(2001),H¨oflerandSchwarzer
(2000).ofler¨H(2000),

Model

Figure3.1:Thepictureshowsthedistributionforcebetweenparticleandthefluid

31

Anoblateellipsoidisrepresentedbyarigidtemplateconnectedtothefluidtracer
straintparticles.ontheThisfluidisindoneorderbytothebodydescribeforcetheoblateterm,intheellipsoid.NavierThe-StokforceesdensityequationΩB,isasachosencon-
tobespringlike.WedefinethisforcedensityΩBas:

ΩB=TBΣxab+Σxab))=−YΣxab)(3.19)
wherexabisthedisplacementfieldoftheseparationbetweenthemarkersWandtheir
correspondingreferencepointX.ThestiffnessconstantY,mustbechosenlargeenough
sothat|Σxab)|V,Vbeingthesizeofthegrid,holdsforalliterations.
Ingeneral,thedisplacementfieldΣxab)isdefinedas:

aΣxabb)=xabb−xabg(3.20)
Thevector,xabbisthepositionofafluidtracerwhosemotionisdeterminedbythelocal
i.e.,,elocityvfluid

x˙abb=IΣxbTU)(3.21)
Thexabgarethereferencepointsassociatedwiththetemplatehavingtheshapeofthe
particle:ysicalph

xabg=xT+DTΣg).rTU

(3.22)

HerexTisthecenterofmassofthetemplate,DTΣg)istherotationmatrixthatdescribes
theinstantaneousorientationoftheoblateellipsoidandrabdenotetheinitialpositions

32

The3.3.3model

ofthereferencepointswithrespecttothecenterofmass.Rotationisdescribedbythe
quaternionformulationasinAllenandTilsdesley(1987).Theequationsofmotionofthe
are:templateparticle

and

F˙=UB

˙I=Ω

(3.23)

(3.24)

whereBisthemassofthetemplateparticle;UandΩarethelinearandangularvelocities
ofthetemplateparticle,respectively;Iisthemomentofinertiaoftheoblateellipsoid
withonlythreenon-zeroelements,I((,I)),I00whereI((>I))=I00(Goldsteinet.al.
(2002))andisthetorque,(Kuuselaet.al.(2001),H¨oflerandSchwarzer(2000)).The
boundaryconditionsnearthecontainerwallare:thenormalvelocitycomponentofthe
fluidiszero,thewallsareassumedtobeimpenetrable,becauseofano-slipcondition
forthetangentialcomponent(H¨oflerandSchwarzer(2000),WachmannandSchwarzer
(1998)).Theinteractionbetweentheoblateellipsoidandthewallsisdefinedthrougha
contactforce(PerramandRasmussen(1996)andKuuselaet.al.(2001)),wherethewalls
aretreatedasaparticlewithinfinitemassandinfiniteradius.Avelocity-Verletintegrator
(Presset.al.(1992))servestointegratetheequationsofmotionforthetranslationanda
Gear-predictorintegrator(AllenandTilsdesley(1987))fortherotationofthetemplate:

F=−BUˆX+ρMKUˆX+ΩTB+ΩTe(3.25)
TwhereˆXistheunitvectoralongtheverticalw.r.tthetemplate’scenterofmassrBb.

(3.26)

=Σxa−xTm)×ΩTB(3.26)
Twithrespecttothetemplate’scenterofmassrBb.
Thedefinedasgeometrytheofratiotheoftheoblatesmallestellipsoidradius,isEbtocharacterizedthelargestbyΔe,radius,itsEM:aspect-ratiowhichis
Δe=EEb(3.27)
MWedefineanequivalentspherefortheoblateellipsoid,asthespherethathasthesame
volume,withanequivalentradius:
EEfjT=3EbE)M(3.28)

Model

TherespectiveStokesvelocity:

33

ΩUE)EfjTΣρρfellip−Ψ)
ih=2ν(3.29)
withρEaa,Tetheellipsoiddensity,ρMbeingthefluiddensity,νthefluidkinematicviscosity
andUthegravity.
WedefinetheunitStokestimeghasthetimeneededforanisolatedellipsoidtogoesover
adistanceofone-ellipsoidlargerradius,EMwithavelocityih,as:
EMgh=ih(3.30)

functionContact3.4

Thereisaconsiderableamountofliteratureinwhichmodelsforsoftpotentialsbetween
nonsphericalparticles(e.g.,oblateellipsoids)areemployedinordertosimulatethebeha-
itviorisofnecessarysimpletomoleculesdetermineandtheliquidforce,>ecrystals.duetoThetheinterestinellipsoid-ellipsoidthepotentialcontacts.arisesbecause
TThelubricationcontacttheoryatlowReynoldsnumberestablishesthatthepresenceof
fluidsmoothlyavoidsthetouchingofellipsoid-ellipsoidsurfaces.Inthealgorithmthese
forcesarecapturedcorrectlyonlyonscaleslargerthanthegridresolution.Sincewe
arewellipsoidalorkingsurfwithacesdilutearerare.systemsWithatvtheeryaimlowofvolumerestrictingfraction,significantcloseparticleencountersoverlappingbetween
arepulsiveforceisintroducedbetweentheellipsoidsandischosentobeproportional
totheirintermediateoverlap.distancesIfthethehoblateydrodynamicellipsoidsareforcesnon-odescribingverlapping,theethexistenceforceofisthezerofluidaandvoidat
thecontactbetweentheparticles(Kuuselaet.al.2001).
WeconsidertwoellipsoidslabeledAandBwithsemi-axesN,O,PandN,O,P,respect-
ively.Therotationalstateoftheellipsoidisexpressedbythe()sets,0h(,h),(h0)and0i(,i),i0
ofareatrorthonormalandrunitrespectivvectorsely,thealongtheintercenterprincipalvectoraxisisofdefinedthetwas:oellipsoids.Ifthecenters
fl

andthematricesψand.are:

R=r−rfl

ψ=NV−)IcIcT
V.=OV−)vcvcT
V

(3.31)

(3.32)(3.33)

34

3.3.4functionContact

Inthecasewherenoneofthesemi-axesvanishes,thesematriceshavetheinverses

ψ−(=NV)IcIcT
V

(3.34)

.−(=OV)vcvcT(3.35)
VPerramandWertheimhavederivedtheobjectfunction,FΣλ)(PerramandRasmussen
1996)and(Kuuselaet.al.2001):

FΣλ)=λΣΨ−λ)E/[ΣΨ−λ)ψ−(+Σλ).−(4−(E(3.36)

FΣλ)=λΣΨ−λ)E/[ΣΨ−λ)ψ−(+Σλ).−(4−(E(3.37)
whereλisaparameter.FΣλ)isnon-negativeforλ∈[Φ,Ψ4.Thenthecontactfunction
>Σψ,.)forthetwoellipsoidsis

>Σψ,.)=[bNkFΣλ)|λ[ΦΨ44(3.38)
If>Σψ,.)σΨ,thetwoellipsoidsoverlap,if>Σψ,.)>Ψ,theydonotandif>Σψ,.)=
Ψ,thetwoellipsoidsareexternallytangent.

4Chapter

Phenomenology

Inthisthechapterprecedingwewillchapterapplywethishavemodelandpresentedseethethemodelcommonofphtheysicalsedimentingphenomenaellipsoids.associatedIn
tothefallingoblateellipsoids.Wewillpresentthethreephasesthatwefound,andthe
terminologiesbidimensionalthatonewillbeelaboratedusedbyintheMahadenextvan.chapters.ThevWeelocityalsofieldcompareforthethisfluidmodelaroundwiththethe
fallingellipsoidisshownandcomparedwiththeworkbyBelmonteet.al.(1998).We
presentinaqualitativewaythepresenceofvorticesinthefluid.

4.1Trajectoriesofafallingoblateellipsoid

Wefoundthreedifferentkindsofmotioninoursimulations:steady-falling,side-to-side
orwhichareshoperiodic-oscillation,wninfig.kno4.1.wnTheseasflutterkindsof(Belmontemotionareet.al.called(1998))patterns,andreagimes,chaoticormotion,phases
intheliteratureandweshallusethetermphaseinthisthesis.

eIngeneral,xperimentalthesetup.numberButofitisphasespossibledependstoonclassifythethespecificphasesmodelintoandtwothelargeconditionssetsofnamelythe,
a)regularandsmoothoscillationsandb)irregularandchaoticoscillations.Forexample,
Tsionalanabeandmodel:Kanekthreeoregular(1994,95),andtwoidentifiedchaotic.fivInefaneallingxperimentalphaseswusingorkawithsimplifieddroppingbidimen-disks,
asFieldshoet.wnal.infig.(1997)4.2.reportedAndlastlyfour,phasesBelmonteofet.whichal.three(1998),werereobservgularedaandreonegularwasoscillatorychaotic
androtationalphaseintheirexperimentswiththinstrips.

Inthemajorityofcases,thetrajectoriesinoursimulationsdependstronglyontheinitial
cosityconditionsνandandthetheoblatepropertiesaspect-ratiooftheΔesystem).In(oblate’orderstoinitialreducetheorientationparameterΘd,spacekinematicinvis-our
system,wefixρMajTC=ΨBbR3andρd∂aiE=´.)BbR3inoursimulations.

35

36

240(a1)180Position in Y12060

0

4.4.1Trajectoriesofafallingoblateellipsoid

(b1)

(I)

(c1)

−258118178238298
Position in X

Figure4.1:Typicalfallingtrajectoriesobtainedinoursimulations.FigureIshowsthe
componentsYΣiSegWPNa)−MΣVdeWzdcgNa)inthesteady-fallingregime,ΣNΨ),withinitial
conditionsθd=Ω+.+Φ,Δe=Φ.Ω),ν=Φ.Φ´´andΣN)Vd=Ω(Φ;ΣOΨ)periodic-oscillation
withinitialconditions:ν=Φ.ΦΩ),Vd=Ω(Φ,Δe=Φ.Ψ´´andθΦ=+´.(d.ΣPΨ)chaotic
motion,withinitialconditionsVd=Ω(Φ,Δe=Φ.Ω),ν=Φ.Φ´´andΣN)θΦ=Ω+.+d.

TheFigureimages4.2:Twererajectoriesobtainedoffallingfromthedisks,sidebypresentedusinginathewvideoorkbycamara.BelmonteTheetal,trajectory(1998).in
(a)correspondstothesteady-fallingregime,(b),periodic-oscillatingmotion,(c)chaotic
motionand(d)thetumblingmotion.

Phenomenology

Comparison vertical velocity and trajectoryChange of Vertical Velocity4.8TrajectoryVyFirst peak−to−peak amplitudeL14210Fourth AmplitudeL23.2Terminal VelocityL3140Vertical Velocity VyPosition in Y2.4L4L51.670L60.8L7Decreasing peak−to−peak amplitude amplitude0048Position in X & Vertical velocity12160510Time t(t15)202530
s1

0.60.2Horizontal Velocity Vx−0.2−0.6

−107142128
)Time t(ts

37

theFiguresame4.3:height.(Left)(Right)ComparisonDecreasingbetweenamplitudethevoferticalthevvelocityerticalvandelocitythe.spatialInitialtrajectoryconditionsat
ofthesystemareθd=Ω+.+Φ,Δe=Φ.Ω),ν=Φ.Φ´´.FallinginitialheightVd=ΩΩ1in
gime.reallingsteady-fthe

oblateSteady-falling4.2ellipsoid

Whentheellipsoidbeginstofall,theverticalvelocityischaracterizedbyadampedwaver-
ing(transientoscillation)intimeresemblingthebehaviorofadampedoscillator(Gold-
steinet.al.(2002)).Ataverylongtime,theverticalcomponentofvelocitybecomes
ellipsoidconstant.Theapproachesamplitudetheofbottomthevoferticalthecontainercomponentfig.ofv4.3.elocityThevdecreaseserticalastrajectorytheoblatefig.
4.3(right),alsoshowsthedampedwavering.Thiswaveringinthetrajectorycurveis
composedofsuccessiveturningpoints.Eachoneofthem,inturn,correspondstopoints

38

ellipsoidoblateSteady-falling4.4.2

Figure4.4:Euleranglesφ,θandψusedforthedescriptionoftheoblate’sorientational
.viorbeha

wheretherateofchangeintheverticalvelocityiszero.Thisvanishingvalueintherate
ofhorizontalchangeoflinesthevL2,...erticalL5,vwithelocityviserticalevidentvelocityin.figureThere4.3isan(left),byinterestingtherelationintersectionofbetweenthe
Ftheorevnumbererytwofoturningsuccessivepointsturninginthepointstrajectoryintheandvtheerticalverticaltrajectoryvelocitythere(seearefig.four4.3,turningleft).
pointsintheverticalvelocity.Ifacharacteristiclengthisassociatedwiththevertical
trajectory,thenthecharacteristiclengthcorrespondingtotheverticalvelocityisreduced.
Theandinhorizontalgeneraltheyhacomponentsvereofgularthevoscillationselocityfollo(seewfig.a4.3,similarbottom).behaviorastheverticalones

Theoblate’sorientationisdescribedthroughthethreerotationaldegreesoffreedom,
calledtheEulerangles,fig.4.4.Wepresentthetimeevolutionoftheanglebetween
theoblate’snormalandtheverticaldirection.Thiswecalltheverticalorientation,fig.
4.5(top).Θ=Φimpliesthattheoblate’sprincipalaxiswillbeparalleltothecontainer’s
bottomfig.4.5(bottom).Atthebeginningofthemovement,thereisalargeangular
changeoftheoblate’snormalwiththevertical,ΔΘfig.4.5(bottom),whichischarac-
terizedbyalargepeak-to-peakamplitudeΘe−e.infig.4.5weillustratethedefinitionof
thepeak-to-peakamplitudeasthedistancebetweensuccessiveturningpoints,whichde-
creasesastheoblatesinks.Inthesteady-fallingregime,boththepeak-to-peakamplitude,
Θe−eandthechangeΔΘ=ΘM−ΘT,decaysastheoblateapproachesthebottomofthe
container.Theoblatetendstoalignitsmajoraxisalongthevertical(Huanget.al.(1998))
havingaverylowresistancetoitsdescentinthefluidacquiringtheterminalvelocityfig.
(top).4.3

Phenomenology

Angular VariationChange of Vertical Orientation8590θiChange vertical orientation Θ(t)
80Terminal Orientationθ75Θ756070Δθ=θf−θiθft
Vertical Orientation 30Position in Y60Θ=0
45Θ65First peak−to−peak amplitude p−p

60Decreasing peak−to−peak amplitude15550051015Time t(t20)2530354050−5051015202530
sPosition in X

39

Figure4.5:(left)Decreasingpeak-to-peakamplitudeintheangularoscillation.(right)
AngularchangeΔΘalongasegmentoftheverticaltrajectory.Initialconditionsofthe
systemθd=Ω+.+Φ,Δe=Φ.Ω),ν=Φ.Φ´´andVd=ΩΩ1Pb.
Vertical Orientation & time100150II)27090110240807021070Vertical velocity & time306018026101418
4.1150503.3(I)2.512040III)1.790304812162024
24681012141618261014
Figure4.6:InitialconditionsVd=Ω´Φ,Ak×Az=Ω)×Ω),Δe=Φ.Ψ´´,ES=(´).
InitialorientationΘΦ=+´d.I)VerticaltrajectoryII)Theverticalvelocity.III)The
orientation.erticalvellipsoidoblateOscillatory4.3

Inthedynamicalvoscillatoryariables.phase,Wetheobserveellipsoidsuchisabehacharacterizedviorinourbyaresimulationsgularatoscillationsmallinkinematicallthe

40

°Θ=00540Θ0=(10−2)°
360Position in Y

180

15

12
Position in Z

Position in Z9

ellipsoidoblateChaotic4.4.4

°=0Θ0Θ0=(10−2)°

01.56.511.516.561113151719
Position in XPosition in X

Figure4.7:Theleftpicturesho−)wsthechangeintheverticaltrajectorywhentheinitial
orientationchangesΔΘΦ=ΨΦ.Therightpictureshowsthehorizontalcomponents.
TheinitialconditionsVd=+(Φ,Ak×Az=Ω)×Ω),Δe=Φ.(´,ES=ΩΩ).
viscosityandsmallaspect-ratios.Infig.4.6weshowtheellipsoid’sverticaltrajectory,
velocityandorientation.ThisoscillatoryphasehasbeenstudiedexperimentallybyBel-
monteet.al.(1998).Thisphasehasadistinctivecharacterthattheperiodofoscillation
oftheverticalvelocityishalftheperiodofoscillationoftheverticalorientation.Oursim-
ulationagreesverywellwiththeexperimentwhichwillbediscussedinthefourthsection
4.chapterof

Chaotic4.4ellipsoidoblate

inWefig.observ4.7.eThechaoticfigshobehawsviortheforvprominentariationschangeintheinthevellipsoid’erticalsandinitialhorizontalorientationastrajectoriesshown
astheinitialorientationchangesbytheorderofΨΦ−).Inoursimulationsthechaotic
isphaseimportanthasbeentopointfoundoutforthatlargethesensitiaspect-ratiosvitytoeq.small3.28vandariationsforinalltheReinitialynoldsconditionsnumbers.ofIt
fallingbodies(inourcasetheellipsoidalorientationΘd)hasnotbeenreportedneitherby
simulations,theorynorbyexperimentsuntilnow.

4.5ComparisonwithMahadevan´smodel

Ashadevanmentionedet.al.in(1995)chapterandΨ,thereMahadearevanseveral(1996))modelsthatha(TveanabegivenandgoodKanekinsightotothe(1994,95),Ma-prob-

Phenomenology

−5

−15Vertical position−25

Vertical trajectory

−5

−20Vertical position−35−50

−0.80.1Horizontal position11.9−1.2

Vertical trajectory

2.41.20Horizontal position

41

Figure4.8:Steady-fallinginitialconditionsΔed=β=Φ.Ω),α=Φ.Ω2,ν(=Φ.Ψ,ν)=
ΦΦ..+Ω),ν,0α==ΦΦ..0,Ω2ν,TνT==Φ.ΦΦ.,λΦ´´,=νTΦ.=ΦΦ+Φ.,Φ,Θλd==+∗Ω+.+ΨΦ.−7F,orΘdthe=2Φd.oscillatoryThesolidregime:andΔdottede=βlines=
showMahadevan’sandourresults,respectively.

lem.WeuseasimplifiedmodelformulatedbyMahadevan(1996)thatoffersagood
absencequalitativofequantitaticomparisonvetoouragreementsimulationshastobeintheunderstoodoscillatoryintermsandofsteady-fthedifallingferencephase.betweenThe
itethetwsizedocmodels.ylinder,whereMahedevtheancylinderdescribesaxistheismotionassumedofatofbeallingstripperpendicularofpapertotheinvanertical.infin-
Thefield.systemTheisviscosityassumedistaktoenbeintobidimensionalaccountinwithtermsanoftheincompressiblegeneralizedfluidRayleighinagradissipationvitational
appliedfunction.alsoThisbywLamborkisa(1932),continuationwhosetheofthecentralmethodapproachusedbyKconsistedelvininandignoringKirchhofthefvandor-
ticesandassumingthatthefluidhaszeroviscosity.Thenewequationsbecomecomplex
sincethereareΩΦnewly’addedmass’terms(Lamb(1932),Mahadevan(1996)).Inour
case,wehaveathreedimensionalmodelwithboundariesasexplainedinchapterΩ.
teThegratorequationsandweofusemotiontheinitialforthisconditionsmodelareforsolvwhichedbysteady-fusingallingRunge-Kanduttaoscillatoryfourthreordergimesin-
areobserved,fig.4.1.Theresultsareshowninfig.4.8.

ortexV4.6

Thevortexproductionisanimportantpartofthefluiddynamicsthatmustbetakeninto
accountintheformulationofatheoryforfallingbodies.Willmarthet.al.(1964)made
apioneeringworkwheredye-laddendisksweredroppedintowaterandvorticeswere

42

OutlookandConclusions4.4.7

Figure4.9:Therightpictureshowsthevortexstructureoftheverticalandhorizontal
componentsofthefluidvelocityfieldΣh,i)generatedbythefallingoblatewithadiameter
3.2inacontainerofΩΦΦ×´Φ×´ΦandReynoldsnumber,ES=ΨΩ1,aspect-ratio,
Δe=Φ.).TheleftpictureshowssheddingvorticesreportedbyBelmonteet.al.,ref.[12],
strip.allingfafor

clearlyobserved.Inoursimulationsthefallingellipsoidgeneratessheddingvorticesin
thefluidalongitstrajectoryasshowninfigure4.9(right).Itshowsthefluidvelocityfield
aroundtheoblateandthevortexislocalizedjustintheregionabovetheoblateandwhere
itexperiencesalargeangularchange,ΔΘasshowninfig.4.5(top).TheReynolds
numbercalculatedfromtheoblate’sdiameterandterminalvelocity,ES=ΨΩ1.Wepoint
outthatthevortexisobtainedalsointheworkbyBelmonteetal(Belmonteet.al.(1998))
whereasheddingvortexiscreatedbythezigzagmotionofthefallingstripasshownin
4.9.figureoflefttheAsystematicstudyoftheboundarylayeroftheellipsoidisnotpresentedinthethesisand
itwillremainasafuturework.

Conclusions4.7Outlookand

Thetainerhasmotionbeenofastudied.singleWoblateefoundellipsoidthreebasicsettlingreingimesafluidforintheathreedynamicsofdimensionalthesystemcon-
(steady-falling,oscillatory,andchaotic).Wefoundthatourresultsareingoodqualitative
agreementwiththesimplifiedmodelproposedby(Mahadevan(1996)).Thesameinitial
conditionshavebeenusedinourworkinthestudyofoscillatoryandstead-fallingre-
gimes.Moreworkhastoberealizedinfutureinordertounderstandtheroleofthefluid
pressureandvelocityfieldsbetter.

5Chapter

Phases

Inthischapterweinvestigatethedynamicsofthephasesintroducedintheprevious
chapter.Weexploreeachphasebychangingdifferentparametersinthesystem.We
introducesomeimportantconceptsanddefinitionswhichwewilluseinthenextchapter.
Inthefirstsection,wediscussthesteadyfallingphasewhentheinitialheightofthefall,
aspectratiooftheellipsoidandthekinematicviscosityarechanged.Next,wedepictthe
oscillatoryphase.Furthermore,westudytheeffectonthedynamicsduetochangesinthe
initialorientationoftheellipsoid.AcomparisonwiththeexperimentsbyBelmonteet.
al.(1998)hasbeenmadeandoursimulationsagreequitewellwiththeexperiments.Fi-
nally,wepresentthechaoticregime,whichishighlysensitivetothechangeintheinitial
orientationoftheellipsoid,asshownintheLyapunovexponent.

Steady-F5.1Phase.alling

5.1.1Changeintheinitialfallingheight.

Figure5.1(I),showsthetrajectoriesfordifferentinitialfallingheightsbyfixingallother
parameters.Itisevidentfromthefigurethatthereisnochangeinthepeak-to-peakamp-
litude.Foralltheheights,thetrajectoriesbehavelikethatofadampedoscillator.

Infig.5.1(II),weseethattheellipsoidapproachesthesameterminalvelocityindependent
ofzerotheasftheallingoblateheight,touchesatthethesamebottomtime.oftheAlso,thecontainerverticalwhichviselocityindependentsharplyofconvtheerfgesallingto
height.

inFiguretheinitial5.1(III)fshoallingwstheheight.vFerticaloralltheorientationaltrajectoriesbehaviourshownofinthefig.ellipsoid5.1(I),withwetheseeinchangefig.
5.1which(III)isthatseenthebytheorientationsemi-majoroftheaxisellipsoidgettingoffersalignedwithminimumgravity.resistanceagainstthefluid,

43

)hgΣgreplacementsPSfrag04allingSteady-F5.5.1Phase.

1

0.75(c)0.5Vertical velocity0.25(II)(c)(b)(a)

(I)(b)

(a)

75Vertical position5025

100

44

Inthissection,westudythedynamicsofthesteady-fallingellipsoidasthekinematicvis-
θΦcosity=isΩ+.+d.changed.ItglidesThedoellipsoidwnwardsstartsanditstotheswingingsidemotionacquiringwithasomegivenamplitude,initialwhileorientation,the
kinematicviscosityνacts,thereby,reducingthesubsequentamplitudesofoscillationas
5.2(I).fig.inwnshoFig.orientation5.2(I,ofII,theIII)shoellipsoidwthefordiftrajectoryferent,verticalkinematicandviscosities,horizontalvrespectielocitiesvely.andCurvtheevNerticalwith

5.1.2Dependenceonthekinematicviscosity.

ΦTFigurerajectory5.1:hasInitialdifferentconditionsinitialofheighttheΣN)system.Vd=θd2+,=ΣO)Ω+V.d+,=Δ0+e,=ΣP)ΦV.dΩ)=,ν)+.=I)Φ.TheΦ´´.spatialEach
trajectoryintheverticalplane.II)Verticalvelocityvs.time.III)Verticalorientationvs.
time.

PSfragreplacements005101520253035
gΣgh)Time t(ts)

0replacementsPSfrag68Horizontal position1012gΣgh)051015Time t(t20)253035
s(a)100θ(c)(b)75Vertical Orientation 50(III)25

(.Φ−(.Φ0.Φ5.Φ7.Φ9.Φ)hgΣgreplacementsPSfrag (a)←Vertical Orientation Vertical Orientation vs. Time120110100908070605040302035302520151050(.Φ−(.Φ0.Φ5.Φ7.Φ9.Φ)hgΣgreplacementsPSfragVelocities Vx,Vy(a)(b)Velocity & Time(II)(c)Vy2520151050(.Φ−(.Φ0.Φ5.Φ7.Φ9.Φ)hgΣgreplacementsPSfrag(I)(c)(b)(a)Vertical PositionHorizontal PositionOblate Vertical Trajectory90807060504030201001614121086445

ν=Φ.ΦΩ),curveOwithν=Φ.Φ´´andcurvePwithν=Φ.Ψ.Weobservefromallthe
threeplotsthatasviscosityincreases,theattenuationbecomesstrongertherebybehaving
likeadampedharmonicoscillator.
Theattenuationintheamplitudeoftheverticalvelocitywithtimeandthetimeperiod
betweentwoconsecutiveturningpointsarenotverydifferentfromeachotherasthe
viscositychanges.Also,weseethatboththehorizontalandverticalvelocitiesconverge
tothesamevalue(eg.ik≈Φandil≈Φ.+)forallthevaluesofviscosity.
Thisstrongattenuationforlargeviscosityishighinthehorizontalcomponentofthe

ΦFiguretrajectory5.2:hasaInitialdifferentconditionskinematicintheviscositysystem.ΣθN)dν==ΦΩ+..+ΦΩ),,VΣdO)ν==1ΦΦ,.ΔΦ´´e,=ΣP)νΦ.Ω)=.Φ.EachΨΦΦ.
(I)(bottom)Projectionvelocitiesofthevs.time.trajectory(III)inVtheerticalverticalorientationplane.vs.(II)timeV.ertical(top)andhorizontal

Phases

← (b) (c)←(III)

Vx

46

2

1.8)oln(A/A1.6

Characteristic time & viscosity

1.4sν/ν0.71.40120145170195
T/ts1494)Time(ts

allingSteady-F5.5.1Phase.

(1)(2)(3)(4)(5)

2419

Figure5.3:Initialconditionsofthesystem.θd=Ω+.+Φ,Vd=ΩΩΦ,Δe=Φ.Ψ).Inthe
figureeachlineshowsthedecreasinglogarithmicamplitudeofoscillationagainsttime
fordifferentvaluesofkinematicviscosityν(=Φ.Φ(,ν)=Φ.Φ0,ν0=Φ.Φ1,ν1=Φ.Ω
andν5=Φ.).Intheinsetweplotthecharacteristictimeτofeachlinevsthekinematic
.viscosity

velocityasseeninthefigure.Thiscanbeexplainedbythefactthattheinteractionbetween
thewallsandtheoblateislesswhenthefluidislessviscousBrenner(1961).Thisisalso
trueforlargeangularvariationsΔΘinfig.5.2(II)cforsmallvaluesofviscosity.

Infig.5.2(III)weseetheorientationalbehaviourastheviscositychanges.Thefirstpeak-
to-peakamplitudesarenearlythesameforallviscosities.Forν=Φ.ΦΩ)thesubsequent
peak-to-peakamplitudesapproachaconstantvalueofΘe−e∼Ψ)dwhileforhighervis-
cosities,theoscillationsgetdampedandtheellipsoidreachesaconstantorientation.We
alsoobservethatthefinalorientationoftheellipsoidgetsalignedwithgravity.

Figure5.3explainsthattheamplitudeofoscillationdecaysexponentiallywithtimewhich
isrevealedbytheloglinearplot.Thedimensionlesscharacteristictime,τdecayslinearly
withthedimensionlesskinematicviscosity,νasshownintheinsetofisfigure5.3.The
solidlineintheinsetshowsalinearfitwithνsequation,iτs=−Φ.Φ´νν0+).).Thefigure
clearlyshowsthatviscosityplaystheroleasadampingfactorthatdeterminesthedecay
rateoftheverticalvelocityandposition.

Phases

100

8060Vertical Position402003

Increment in the Aspect−Ratio1

Vertical Velocity vs. Time

(a)0.75(b)(c)Vertical Velocity Vy0.5(c)II)I)(b)0.25(a)ΔΔΔ r=0.18 r=0.22 r=0.28
579005101520
)Time t(tHorizontal PositionsVertical Orientation vs. Time140 (c)←

100θAngle 60

(a) ↑ (b)↑

III)2008Time t(t16)24
s

47

Figure5.4:Initialconditionsofthesystem.Θd=Ω+.+Φ,Vd=2+,Ak×Al×Az=
ΣOΨΦ)Δ×e=ΨΦΦΦ×,ΩΩ,ΨΦ,ΣνP)Δ=eΦ=.Φ´´Φ,.Ψ1.EachI)Trajectorytrajectoryforhasthedifvferenterticalplane.aspect-ratioII)VΣN)Δerticalev=Φ,elocityΩ2,
vs.time.III)Verticalorientationvs.time.

5.1.3Changeintheellipsoidaspect-ratio.

Whentheellipsoidbeginstofall,itgainsalargeamplitudeofoscillation.Thetrajectory
ΣP),plottedinthefigure5.4I,isforanellipsoidwithaspectratio,Δe=0.18.Weseean
oscillatorytrajectorycharacterizedbyaconstantpeak-to-peakamplitudeof3.Asthe
behaaspectvior.ratioIfistheaspectincreasedratiotois0.22,furtheri.e.,increased,trajectorytheΣO)intrajectorythefigure,showsshoirrewsgularasteadyoscillations.falling
Atthebeginningofthefall,theellipsoidwithlargeaspectratio(trajectoryN),showsa
rapiddampingamplitudeinthefirsthalfofthetrajectorybutinthesecondhalf,doesn’t

48

P5.5.2Phase.eriodic

showanysteadyfallingbehavior,insteadchangesabruptlyintoanon-stationarystate.
Inaspectfigureratio,5.4(I),Δeforallincreasestrajectories,from0.18thetofirst0.29.amplitude,ψΦdecreasesfrom3.0to1.8asthe
As(minortheaxisofaspect-ratio,theΔoblateeisisfixedincreasedinourfrom0.18simulations)to0.29,therebytheprooblate’svidingarealessbecomesresistancesmallto
thefluid.Therefore,thefinalverticalvelocitydecreasesastheaspect-ratioisincreased,
fig.5.4II.Alsoastheaspect-ratioisincreasedthepeak-to-peakamplitudeofthevertical
decreases.elocityvTheratiodecreasespeak-to-peakortheamplitudeoblate’sforareathevincreases.erticalFororientationΔe=ΘΦ.e−Ω2etheincreasespeak-to-peakwhentheamplitudeaspect-
ddisforΘeΔ−ee==ΦΨ).Ψ1,.fig.Inall5.4cases(IIIa),theandoblatemuchatthesmallerendcomparedorientsvtoΘerticallye−e=fig.0Φ,5.4fig.(III).5.4As(IIIc),the
whichellipsoidwillbeaspectratio,discussedΔeindetailincreases,inthechaptersystem).transitsfromregulartoirregulardynamics

Phase.eriodicP5.2

Wesmallerfindperiodicaspect-ratiobeha(Δevior=forΦ.Ψ´´smaller).Thekinematicdynamicsofviscositythef(ν)alling=Φ.ΦΩ)ellipsoid→EisSgo=v(1Φerned)andby
inertialeffects.Infigure5.5(I),weshowthetransitionfromaquasi-periodic,oralong
steady-fallingtrajectory(ν(=Φ.ΨΦΦ),toaperiodicbehaviorfig.5.5(I,ν)=Φ.ΦΩ)),when
thekinematicviscosityisvariedfromν(=Φ.Ψtoν)=Φ.ΦΩ).Thetrajectorypresented
infig.5.5I,withkinematicviscosityν)=Φ.ΦΩ)hasawavelengthof20.
Theverticalvelocityshowninfig.5.5(II),hasthesametransitionfromalongsteady-
fhasallingareoscillationgimewithperiodafinalofa´v.´.eragevelocityof´.Φtotheperiodicregimewherethevelocity
Theverticalorientationshowninfig.5.5(III)hasalsothesametransitionfromalong
oscillatesteady-fallingaroundreΘgime=to2Φdperiodicwithangularbehaviorwithpeak-to-peakaperiodofamplitude6.6s,Θand=the+Φd.angularvalues
e−eΦtheThesystemdecreasingfrominthesteady-fkinematicallingtoviscosityoscillatory,foraphase.muchWesmallerwillinvestigaspect-ratioateΔthiseσΦ.transitionΨ,takines
.)chapter

5.2.1Changeintheinitialorientation

Inthissection,weinvestigatetheperiodicphasewithdifferentinitialorientations.The
correspondingtrajectoriesareshowninfig.5.6(I)forΘΦ=Ω+dandforΘΦ=2Φd
andthepeak-to-peakamplitudesare2.3and0.4,respectivelyandshowanoscillatory

=0.100ν=0.025νVertical OrientationVertical Orientation & Time1301109070)hgΣgreplacementsPSfrag−132241680)hgΣgreplacementsPSfragHorizontal Position25141185249

behavior.Thepeak-to-peakamplitudeoftheoscillationinthetrajectoryverticalvelocity
andorientationfig.5.6(I,II,III)decreasesastheinitialangleoforientationincreases.
Thefinalverticalvelocityandorientationforallthethreeinitialorientationsisonan
averageΦ.+and1)d,respectively.Thismeansthattheaveragelimitingvaluesofthe
verticalvelocityandorientationarenotaffectedbythevariationoftheinitialorientation
.ΘdWeobservealargepeak-to-peakamplitudeofoscillationintheverticalvelocityandver-
ticalorientation,equaltoΨ.Φand0Φd,respectively,withaninitialorientation,ΘΦ=Ω+d
anditreducesastheoblate’sinitialorientation,ΘΦtendsto2Φd.

Figure5.5:Trajectoriesgeneratedforν(=Φ.ΨΦΦ,ν)=Φ.ΦΩ).(I)Trajectoryinvertical
andhorizontalposition.(II)VerticalorientationΘvstime.(III)Verticalvelocityilvs
time.TheinitialconditionsareVd=2+,Δe=Φ.Ψ´´,θΦ=+´.(d.

40

Phases

III)PSfragreplacements500816243240
Σg)gh

Vertical Velocity & Time4=0.025ν3=0.100ν=0.025ν2Vertical Velocity10II)

Oblate Trajectory10085=0.100ν70Vertical Position5540I)

50

o26o63o90

Inthissection,wecompareourresultswithBelmonte’sresults.Figure5.7(a)(simulation)
showsthetimedependenceoftheverticalorientationwithν)=Φ.ΦΩ)andthevalueof
thepeak-to-peakamplitudeisΘd=+Φd.
Theverticalvelocityasshowninfig.5.7(b)(simulation)reachesitsmaximumvalue´.+
asΘapproachesΘbkthereby,showingaminimumdrag.Theminimumverticalvelocity
il=Ω.)isachievedatΘbTc∼0Φd.
Thebutterflyshapedcurve(fig.5.7(b))wasalsomeasuredintheexperimentalworkby
Belmonteet.al.(1998)inwhichtheverticalorientationΘoscillateswithtwicetheperiod

5.2.2ComparisonwithBelmonte’sresults.

15

Figure5.6:I)Trajectoriesforthreeinitialorientations.(a)ΘΦ=Ω+d(b)ΘΦ=+´d(c)
ΘΦ=2Φd.II)Thecorrespondingverticalvelocities.III)Theverticalorientations.Initial
conditionsVd=ΨΦΦ,Ak×Az=ΨΦ×ΨΦ,Δe=Φ.Ψ´´,ES=(´).

c)

Phase.eriodicP5.5.2

Vertical Velocity & Time

0.890c)b)a)0.660Vertical VelocityVertical Position0.4II)300.2I)00replacementsPSfrag2Horizontal Position11gΣgh)05Time t(t10)
s Vertical Orientation & Time140a)120100c)Vertical Orientation8060b)40III)replacementsPSfrag20051015gΣ20gh)25303540

Oblate Trajectory1

)hgΣgreplacementsPSfrag

51Phases80130120601104010020900(Degrees)Θ80 [degrees]Θ−20
70)θ−40a)60−60a)50024681012141618−8000.20.40.60.8
T(s)t[sec]7043.8603.6503.43.240Vy[cm/s]Vy(cm/s)3302.82.6202.4b)b)102.2250607080Θ90(Degrees)1001101201300−80−60−40−20Θ0 [degrees]20406080
601.550140300.520100Vx[cm/s]Vx(cm/s)0−0.5−10−20−1−30−40−1.5c)c)−50−25060708090100110120130−60−80−60−40−20020406080
[degrees]Θ(Degrees)ΘFigure5.7:ComparisonwiththeresultsofBelmonteetal.ref[12]forafallingstripin
theperiodicregime,forthe(a)VerticalorientationΘvstime.(b)Verticalvelocityilvs
Θ.(c)HorizontalvelocityikvsΘ.TheinitialconditionsareVd=2+,Δe=Φ.Ψ´´,
ν(=Φ.ΦΩ),θΦ=+´.(d.

52

Oblate Trajectory

egime.rChaotic5.5.3

105907560Position in Y→45c) →30a) b)←15044.555.566.577.5
Position in XFigure5.8:InitialconditionsVd=2+,Δe=Φ.Ω),ν=Φ.Φ´´andtinyvariationsofthe
initialorientationΣN)θΦ=Ω+.+d,ΣO)θd=Ω+.+ΦΦΨd,ΣP)θd=Ω+.+ΦΦΦΦΦΨd.

.ioflThehorizontalvelocityoscillatesaroundzerowithconstantperiodofoscillationandits
maximumvalueik,bk=Ψ.)andtheminimumik,bTc=−Ψ.)atΘ∼2Φdasseenin
fig.5.7(c).Whenthehorizontalvelocityiszerotheoblatetakesitsmaximum(ΨΩΦd)and
minimum(+Φd)valuesinΘwithanon-zeroverticalvelocityil=´.Ω.

egime.rChaotic5.3

5.3.1Sensitivitytothechangeintheinitialorientation

Wenowdiscusssthesensitivityofthedynamicstotinychangesintheinitialorientation.
Wehavesimulatedthreetrajectoriesshowninfig.(.1,whichhaveslightlydifferent
initialorientation.Atinyvariationintherelativeorientation(Δθd=ΨΦ−0)producesa
significantvariationintheshapeofallthecurves.Thesechangesareappreciatedinthe
lowerpartofthetrajectories.
Inordertogetabetterpictureofsensitivity,wehaveincrementedthefallheightVdto166.
Theresultingfourtrajectoriesforfourslightlydifferentinitialorientationsinthevertical
planeareplottedinfig.(.2andweobservehighsensitivityto−0theinitialorientation.For
thefourtrajectoriestherelativeangularvariationisΔθd=ΨΦ.

Phases

150

Trajectory Sensitivity to the Initial Orientation

125100b)a)Position in Y7550

25

d)c)

53

02468101214161820
Position in XFigure5.9:InitialconditionsVd=Ψ++,Δe=Φ.Ω),ν=Φ.Φ´´andtinyvariationsofthe
initialorientationΣN)θΦ=().´1(d,ΣO)θd=().Φ´´d,ΣP)θd=((.21Ψd,ΣR)θd=((.20+d.

5.3.2Powerspectra,autocorrelation,Poincaresection.

Duetothesensitivitytosmallchangesintheinitialorientation,weproceedtoanalysethe
systembytheFourierpowerspectrumtimeseriesofthehorizontalcoordinatekΣg),kΣg+
δg),kΣg+Ω∗δg),...andinourcaseδg=Φ.Φ)´)++.Abroadspectrumoffrecuencies
appearsasshowninfig.5.10IIthereby,indicatingachaoticmotion.
Theautocorrelationfunction,forthesametimeseries(fig.5.10I),doesnotfallquicklyto
zeroratherdecreaseslinearlywithtimeandthepointsarenotindependentofeachother.
Inthefigure5.10III,wepresentslicesorPoincare´sectionsΣpk,k),correspondingtothe
trajectoriesinfig.(.2awhicharequiteirregular.Theorbitsarequasi-periodicinthe
notsenseclosedthattheandyarepassnotrepeatedlyassociatedandwithirreagularlyparticularthroughtimetheperiod.wholeThedomain.sensitiThevitytoorbitsinitialare
conditionsisclearinthesefourfigures.Asmallchangeintheinitialorientationresultsin
largechangesinpositionandvelocity.

5.3.3Lyapunovnumber.

Wenowinvestigatequantitatively,thissensitivitybystudyingtheincrementintheEuc-
(.2lidean(a)anddistance,(c).Re1Figuree2=(.ΨΨ,Σk(sho−wsk)))that+theΣl(−ldistance)))betweenbetweenthenearbycurvespointspresentedhasaninovfig.er-
allexponentialtimedependenceRΣg)∼expΣλg)andthefitgivesanestimateforthe

54

OutlookandConclusions5.5.4

Power Spectra Z10^2

10^1P−S

10^0

II)

Power Spectra ZAutocorrelation Function10^210.90.80.7II)10^10.6P−SAutocorr0.50.4I)0.310^00.20.10238028803380i38804380488010^110^2i10^310^4
Phase Space10.8III) θ0=45,033o
0.60.40.2Velocity component Vz0−0.2−0.4−0.61.522.533.544.555.56
Position in Z

Figure5.10:Detectionofchaos.I)AutocorrelationfunctionforthetimeseriesofkΣg)
forthetrajectoryoffig.14a.II)PowerSpectraoffig.14aIII)PoincaresectionΣpk,k)for
14a.fig.oftrajectorythe

Lyapunovexponentλ=Φ.Φ)Ω±Φ.ΦΦ).ThepositiveLyapunovexponentgivesaclear
Chaos.ofindication

Conclusions5.4Outlookand

Wehaveoscillatory(periodic),observedthreeandphaseschaoticofrethegimes.Thedynamicssteady-foftheallingsystemandthenamely,periodicsteady-fregimealling,ex-
hibitBelmonteasimilaret.al.ph(1998).ysicalWbehaehaviorveasobservcharacterizededfortheflatteneddynamicsbodiesofthebyFieldsteady-fet.allingal.re(1997),gime
whenthekinematicviscosity,droppingheight,andoblate’saspect-ratioarechanged.
Someconclusionscanbedrawnfromthispartofthework.
(a)ThespatialtrajectoriesΣk,l)arecomposedofoscillationsthatcorrespondtoadamped

Phases

1

0.5

0ln(dist)−0.5−1

−1.5

Increment of the Distance in Time

55

PSfragreplacements−201020304050607080
)gΣghFigure5.11:Increasinglogarithmicbehaviorfortheseparationdistancebetweenthe
trajectoriesΣN)andΣP)infig.14,thatslightlydifferintheinitialorientationangleby
ΔΘ=Φ.(Φ´d.

harmonicmotion.ThisregimeispresentforsmallvaluesofI≈Φ.)−Ψ,ES≈ΨΦΦ
andisshowninfig.5.1-5.4Thereisnovariationinthetrajectorieswhenweincreasethe
initialheight.Theviscosity,foraconstantsmallaspect-ratio,determinesthedecayofthe
positionandthevelocityoftheellipsoidasshowninfig.5.2.Whentheaspect-ratiois
changed,thetrajectoriesvarysignificantly(fig.5.4).
(b)Thefinalverticalvelocity,ildoesnotdependontheinitialfallingheightandthe
.viscositykinematic(c)Theverticalorientation,Θoftheellipsoid,undergoesarotationalmotionuntillits
majoraxisgetsalignedwithgravity.Thistendencyofexhibitingminimumresistance
againstthefluidexistsforallReynoldsnumbersintherangeES≈Ψ−)ΦΦ.
TheperiodicbehaviorinoursimulationsisfoundforES∼(ΦΦandsmallaspectratios
ΣΔe≤Φ.Ψ).Theverticalorientation,Θoscillateswithtwicetheperiodofoscillation
oftheverticalvelocityilandatthesameperiodofthehorizontalvelocityik.This
periodicmotionhasalsobeenobservedby(Belmonteet.al.1998)andthisshowsthat
oursimulationsareessentiallycorrect.Inthisregimetheinitialorientationdetermines
thevalueoftheamplitudeofoscillationinthespatialtrajectoryΣk,l),velocityiland
orientationΘ,andplaysthesameroleasthephaseangleintheoscillatorymotion.For
Θd=2Φd,theamplitudesoftheabovequantitiesapproachasmallvalue.
Thechaoticbehaviorispresentforlargeraspect-ratios(Δe≥Φ.´)andintheentirerange
ofReynoldsnumbersusedinthesimulation.Theseparationbetweenthespatialtraject-
oriesofthefallingoblateellipsoiddivergesforsmallvariationsintheinitialorientation

56

Θ

λ

d

,=

andwsgroΦ)Ω.Φ±

xponentiallye.ΦΦ).Φ

intime.Thealuev5.5.4

foundforConclusions

theyapunoLand

veOutlook

xponentis

6Chapter

lawSimilaritytransitionsphaseand

Inphasethetwandothevprecedingariationinchapterseachwephasehavewithseenrespectthetothecharacteristicphysicalparameters.phenomenologyInoftheeachfirst
sectionofthischapter,wediscussthesimilaritylawthatdescribesintrinsicallythedy-
namicsofthesteady-fallingandoscillatoryregime.Aphasediagramhasbeencompared
totheresultsofFieldet.al.(1997).Furthermore,wepresentanovelmechanismthatex-
plainsthetransitionfromtheoscillatorytothesteady-fallingphaseandthentothechaotic
phase.

6.1Steady-FallingOblateEllipsoid:Similaritylaw.

Itiswellknownforlargevelocitiesthattheinertialdragisgivenby,

>C=,ρMK)F

(6.1)

where,istheformfactoroftheinertialdrag,Fisthecross-sectionalareaoftheoblate
ellipsoid,ρMthefluiddensityandKtheellipsoidvelocity.Theweightoftheoblate
toproportionalisellipsoid

>T≈ρEaEa)MEbU(6.2)
withρEaabeingtheellipsoidaldensity,EMandEbbeingtheminorandmajorradius,
respectively.Theterminaldownwardvelocityisdeterminedbytheequilibriumbetween
thesetwoforcesandsincebothdensitiesarefixed,theterminaldownwardvelocityis
,byenvgi

>T=>C=⇒K∼UEb
57

(6.3)

58

654Vertical velocity vy32

1

6.6.1Steady-FallingOblateEllipsoid:Similaritylaw.

System L"System L

018141022)Time t(ts

654Vertical velocity vy32

1

System L"System L

05791113
)Time t(ts

ΦFigurekinematic6.1:viscosityInitialν=conditionsΦ.Ψ.Ininthethefiguresystem.(left)θdwe=plotΩ+.the+,vVdertical=v1+,elocityΔe=agΦ.ainstΨ)timeand
inapplyingboththesystemsinvAerse(solidline)transformationandAof(dottedtableΨ.line),andthesuperposition(right)performed

Wesupposethatthecharacteristicperiodofoscillationdependsonthelateraldimension
whichinoursimulationsisASdgΣ=Ω))andindependentoftheverticaldimension(falling
height).Therefore,weassumethattheperiodofoscillationshouldchangeas,

G∗∼ASdg(6.4)
KThedynamicsofthesystem,ingeneral,dependsontheReynoldsandtheFroudenumbers
(Sec.Ω.Ψ.Ω,Belmonteet.al.(1998)).WecanrescaletheparametersofasystemofsizeA
toasystemofsizeAthrough

SystemA→SystemA
ASdg→c∗ASdg
EM→c∗EM
Eb→c∗Eb
U→1∗Rc

Table1.Transformationsrules.

transitionsphaseandlawSimilarity

59

ThetransformationspresentedintableΨkeeptheFroudeandReynoldsnumberscon-
stantandthedynamicsinthetwosystemsbecomeequivalentifthevelocitycomponents
changeasintableΩ.Asaconsequenceofthistransformation,thedynamicsinsystemA
undergoesachangeintheellipsoidperiod,theverticalandhorizontalcomponentsofthe
velocity,asshownintableΩ.

ASystemASystem→G∗→c∗G∗
VhorKSdgc→KvEgi→Vvert
c

Table2.PeriodandvelocityapplyingthetransformationrulesinTable1.

Infig.6.1wepresentthesuperpositionoftheverticalvelocitybyapplyinganinverse
transforminAwithn=2.Thethreecurvescoincidequitewellinagreementwiththe
scaling.

6.2Periodicbehavior:Similaritylaw.

Figure6.2showstheverticalvelocityagainsttimeforthethreesystemsA(solidline)
(cand=AΩ),(n=2)andasandinAthe(n=4),caseoftherespectivsteadyely.fWallingeapplyregime,thethetransformationdynamicsforrulestheofsystemstableAΨ,
andA,intheoscillatoryregime,arerelatedbyasimilaritylaw.

Diagram.Phase6.3

∗Inthephasemomentspace,ofweinertiadefineoftheaoblatedimensionlessellipsoidtomomentthatofofitsinertiasphericalIwhichequivisalenttheatratiosameof
Reynoldsnumber.WedoasimilaranalysisforourresultsasdonebyFieldet.al.(1997).
Itisimportanttoremarkthattheexperimentwasforafallingdisk,withsmallaspect-ratio
andweexpectthedynamicsofthesystemtobeclosetothatofanoblateellipsoid.
Thedefinitionsofthedimensionlessvariablesforoursystemare:
I=Id∂aiE=)ebρEaaTe=)ρEaaTeΔe(6.5)
IheSEgE(eMρMajTC(ρMajTC
ES=IΣΩeM)(6.6)
ν∗ReFigureynolds6.3leftnumbers,sho(highwsthekinematicresultsintheviscosity),log-logthescale.motionAtisolovwervalues-dampedofandItheandoblatesmall

60

543Vertical velocity vy21

002.51.50.5Horizontal velocity vy−0.5−1.50

, r g, rMm g/8, 2*rm, 2*rM 5
4Vertical velocity vy321

Diagram.Phase6.6.3

System L|System L||System L

7.51522.5300012345678
Time t(ts)Time t (ts)
System L g, rm, rM System L|
g/8, 2*rm, 2*rM System L||
1.8Horizontal velocity vy−0.2

(d)−2.27.51522.5300246810
Time t(ts)Time t(ts)

ΦFigurekinematic6.2:viscosityInitialν=conditionsΦ.ΦΩ2.inInthefiguresystem.(left)θwed=plotΩ+.the+,vVdertical=v1+,elocityΔeag=Φ.ainstΨ)timeand
forapplyingsystemstheAinv(soliderseline)andtransformationAof(dottedtableline).Ψ.Therightfigureshowsthesuperposition

ellipsoiddropstothebottomofthecontainerwithoutanyoscillationandthisisthebe-
ginningofthesteadyfallingregime.Thisisseeninthelefthandbottomcornerofthe
diagram.WhentheReynoldsnumberisincreasedΣES≥ΨΦ))andI∗isfixed,thetraject-
oryiscomposedofsuccessiveoscillationsthatdecreaseinamplitudeandeventuallythe
oblatereachesthebottomofthecontainer.Thisisthesteadyfallingregime.
ForsmallvaluesofI∼ΔeσΨ,wehaveaflatellipsoid.ForReynoldsnumber,
ES≥(ΦΦ,thetrajectory,velocityandtheorientationarecharacterizedbyoscillations
thatareperiodicintimeandspace.Thisphaseiscalledastheoscillatoryregime.As
weincreaseI,theobjectbecomesaspherethatisslightlyflattenedatthepolesandits
dynamicsbecomessensitivetosmallvariationsintheinitialorientationtherebyexhibiting
.trajectorychaotica

transitionsphaseandlawSimilarity

This work10-1Ref. [14]Ref. [13]Tumbling

110

Chaotic

Steady−Falling

Periodic

61

efghChaoticChaotic
PSfragreplacements10-2bPSfragcdreplacements
∗Ia∗II*100
SteadyPeriodicEFallingI∗10-3I∗Steady−Falling Periodic
Ψ2ΦΨ2Φ
ΨΨΦ3ΦΨΨ3ΦΦ
Ψ102103104Ψ101102103
ΨΦEΨΦRe
et.Figureal.6.3:(1997).TheInlefttherightpictureplotshowewsthepresentphasetherediagramgimesofofftheallingphasedisksspacereportedfortheinfallingField
oblatedimensionlessellipsoidmomentobtainedofininertiaourI∗,andsimulations.theInhorizontalbothaxispictures,ESisthethevReerticalynoldsaxisnumberisthe.

IfwecompareourdiagramwiththeexperimentalresultsobtainedbyFieldet.al.(1997),
fig.6.3leftweseethatinbothpictures,thedistributionofphasesisequivalent.The
differencesinourdiagramwiththatofField’sarethattheyusedisksandnotoblate
ellipsoidsandthetumblingregimeisnotpresentinourresults.Thecoexistenceofthe
dynamicalphasesasexplainedaboveisindependentoftheinitialorientationoftheoblate
ellipsoid.

6.4TransitionfromSteady-fallingtoOscillatoryphase

Infig.6.4weshowthebehaviorofthecharacteristictimeG∗/G(Sec.(.Ψ.Ω),adimen-
sionalizedusingeq.).(,asweincreasetheReynoldsnumberES,thecharacteristictime
goestozeroatESB≈´)),infig.6.3.Beyondthispointwefindtheoscillatoryregime,
thatbehaveslikeasteady-fallingregimewithaninfinitecharacteristictime.
Therefore,wecanconsiderG∗/Gastheorderparameterandthecontrolparameteristhe
Reynoldsnumberforthistransition.Thistransitionislikeasecondorderphasetransition.
Theinsetexhibitsthepower-lawbehaviorwithacriticalexponent≈Φ.).Inthecaseof
theupperpartofthetransition,infig.6.3,thevariationintheoblateellipsoidaspect-ratio
impliesthechangefromsteady-fallingtooscillatoryregime,whichisalsosupportedin
fig.5.4wherethetrajectoriesvaryfromsteady-fallingtooscillatoryastheaspect-ratiois
decreased.

62

0.039

6.6.5TransitionfromSteady-fallingtochaoticphase.

/Ty=0.53*x−6.1*T0.034/T)0.024*ln(T0.009268.5818.58)ln(Re−Rec0.009310210110ReRec

∗TheFigureinset6.4:shoThewstheorderpower-laparameterwbehaGvior/Gvs.withtheanecontrolxponentcloseparametertoΦE.)S,withESB≈´)).
60

45

*30T/T

15

00.180.2Δ r 0.22Δ r0.24
c

Figure6.5:TheorderparameterG/G∗vs.thecontrolparameterΔe,withΔeB≈Φ.ΩΩ.
6.5TransitionfromSteady-fallingtochaoticphase.
Thetransitionfromsteady-fallingtochaoticregimeispresentedinfig6.5.Weusethe
orderparameterG/G∗,i.e.,theinverseoftheoneusedbeforeinordertodescribethe
transition,andthecontrolparameteristheaspect-ratioΔe.AtΔeBafinitejumpinthe
orderparameterisobserved.ThecharacteristictimeG/G∗disappearsduetothenon-
regularoscillationsthatareverysensitivetosmallvariationsintheinitialorientation.
Thistransitionseemstobethereforeoffirstorder.

transitionsphaseandlawSimilarity

Conclusions6.6Outlookand

63

Forthesteady-fallingandoscillatoryregimeweobtainasimilaritylawexpressedintables
ΨandΩ,whichisadirectconsequenceoftheinvarianceoftheReynoldsandFroude
numbers.

Theconstructionofthephasediagramshowsthreewell-defineddynamicregionsas
shobehawnviorbyisFieldassociatedet.al.with(1997).theThetransitiondiftoferencechaosinthethroughaboveintermittencreferenceyiswhichthatisthenotchaoticseen
inoursimulations.Thephasediagramisindependentoftheinitialorientation.

Ouritionfromsimulationssteady-fshowallingthattothechaotictransitionregimefromcanbesteady-funderstoodallingtoasoscillatorysecondandandfirstthetrans-order
phasetransition,respectivelyandthecharacteristictransienttimebeingtheorderpara-
.meter

64

6.6.6

Conclusions

and

Outlook

7Chapter

articlesPMany

Intheprecedingchapterswepresentedthephysicsbehindthefallingofoneoblateel-
lipsoid.Now,inthischapterwearegoingtoexposethedynamicsofmanysedimenting
ellipsoidsinafluidandundertheactionofgravityusingthemodelgiveninchapterΩ,at
smallandmoderateReynoldsnumbers.Westudythesettlingorsedimentingvelocityas
functionofthevolumefraction.Also,weexaminetheverticalandhorizontalfluctuations
oftheellipsoidscomparedtospheres.Additionally,weconsiderthebehaviorofpaircor-
relationfunctionsasafunctionofthevolumefraction.Nextwepresentanorientational
studyusingtheEulerangles,andtheangulardistributionsandtheorientationalbehavior
withvolumefraction.Thesamestudyforthesedimentingvelocityandorientationare
madeastheReynoldsnumberincreases(ES≈0).Furthermore,wepresentthebehavior
oftheorientationalorderparameterwiththevolumefraction.Finally,aconclusionis
en.vgi

Results7.1

7.1.1Sedimentationvelocityforoblateellipsoids

vInolumefig.7.1fractionweshoΦVw,thewithinmeanavrangeerticalof0.01sedimentationto0.21,vforelocityoblateiΣg)asellipsoidsafunctionandofspheresthe
andthencompareittothephenomenologicalRichardson-ZakilawvΦ)=ΣΨ−Φ)c
RichardsonandZaki(1954)withc=).).ThelimitofΦV→Φcorrespondsv0totheVsingle
fallingellipsoidwhichwestudiedinthepreviouschapter.Itisinterestingtopointout
thatsphere,theeq.Ω.sedimentationΩ1,whichvfolloelocitywsofthetheellipsoid,phenomenologicalissmallcomparedRichardson-Zakitothatoflawthe.Thisequivisalentnot
thecaseforfibers(elongatedellipsoids),whereitisfoundthatthesedimentationvelocity
hasamaximumforsmallervolumefractionwhichcanexceedtheterminalvelocityofa
singlefiberKuuselaet.al.(2003).
Foroblateellipsoidsthemeanverticalsedimentationvelocitypassesthroughalocalmax-
imumatΦV≈Φ.Φ).Thismaximumisquiteinterestingsinceitisnotobservedfor
65

66

0.9

(vs)||0.6<v>

0.3

Richardson−ZakiSim. OblatesSim. Spheres

Results7.7.1

00.050.1Φ0.150.2
VandFiguresphere7.1:Mean(dash-dotline),sedimentationasvfunctionelocityofiΣgthe)vforolumetheoblatefractionΦellipsoid.Theoblate(dash-squaredellipsoidline)
Vaspect-ratioisΔe=Φ.Ω),the−)radiusoftheequivalentsphereEEfjT=Φ.20andthe
ReynoldsnumberES=(∗ΨΦ.

spheres.Similarnon-monotonicsedimentationofnon-sphericalbodies(e.g.fibers)has
beenreportedexperimentallybyHerzhaftandGuazzelli(1999)andforprolateellipsoids
insimulationsbyKuuselaet.al.(2001)duetoanorientationparalleltogravity.
Wechoosethedensityofthefluid,theStokesvelocityandthelargerradiusoftheellipsoid
equaltounityinoursystem.Inallcasesthecontainerhasaheight,A=1)andabase
ofΨΦ×ΨΦ,andthelatticeconstant,V=Φ.´.WechangeΦVaddingellipsoidsandthe
maximumnumberofellipsoidsinoursimulationsisoftheorderofseveralthousands.
Theratiobetweenthedensityoftheoblateellipsoidsandthefluidis(.
Infig.7.2wepresenttheparallel()andperpendicular(⊥)componentsofthevelocity
fluctuationswithrespecttogravityasafunctionofthevolumefractionΦVwhichare
as:definedδi)=σi)>−σi>)(7.1)
δi)⊥=σi)⊥>(7.2)
Theangularbracketsindicateaveragingovertheellipsoidsthathavenotreachedthefinal
bottompositionatthecontainer.Theaveragesweremadeoveratleast50realizations
startingwithdifferentrandompositionsandorientations.Infig.7.2thevertical(parallel
togravity)fluctuationsforspheresandforellipsoidsaremuchlargerthantherespective
horizontalcomponents.Thefluctuationsforellipsoidsdecreasewiththevolumefraction.
Foranequivalentsystemofspheres,thefluctuationsshowamaximumatintermediate
volumefractions(ΦV≈Φ.Φ0)(Kalthoffet.al.(1996)andNicolaiet.al.(1995)).In
allcasesthefluctuationsforthespheresareconsiderablylargerthanthefluctuationsfor
ellipsoids.oblate

articlesPMany

0.080.06>2⊥Vδ>,<2||0.04Vδ<0.02

. δδ⊥ Ellip Ellip
|| Sphδ. ⊥ Sphδ ||

67

0.010.060.11ΦV0.160.21
theFigurevertical7.2:V(solidelocityline),andfluctuationshorizontalforellipsoids(dash-dotline)(circle-line)andcomponentsspherescorresponding(squared-line)toforfig.
Ψ.Theoblateellipsoidhasanaspect-ratioofΔe=Φ.Ω)−,)theequivalentradiusofasphere
isEEfjT=Φ.20,andtheReynoldsnumberES=(∗ΨΦ.

16

>122⊥Vδ>/<2||8Vδ<4

Ratio SphRatio Ellip

0.010.06Φ0.11V0.160.21
Figure7.3:Ratiooftheverticaltothehorizontalvelocityfluctuationsforspheresand
oblateellipsoidsasfunctionofthevolumefraction,ΦV.Theoblateellipsoidaspect-ratio
isΔe=Φ.Ω),andtheReynoldsnumberES=(∗ΨΦ−).

Infig.7.3wepresenttheratio,δi||)/δi)⊥forspheresandoblateellipsoids.Forspheres
theratioshowsamaximumaroundΦV≈Φ.Φ0(Kalthoffet.al.(1996)andNicolaiet.

68

Results7.7.1

18y=88x+0.7611>2⊥Vδ4>/<2||20Vδ<−32>V⊥15
y=33x+3δ10−102>/<Vδ||50
<0.010.060.11ΦV0.160.21
0.010.060.11Φ0.160.21
VFigure7.4:Ratiooftheverticalvelocityfluctuationsforspherestothatoftheoblate
ellipsoidsasfunctionofthevolumefraction,ΦV.Theinsetshowsthecorresponding
ratioforthehorizontalfluctuations.−)Theoblateellipsoidaspect-ratioisΔe=Φ.Ω),and
theReynoldsnumberES=(∗ΨΦ.

al.(1995)).Forellipsoidstheratioshowsaslightlylargervaluethanthatofthespheres
forverysmallvolumefractionsandhasanoverallmonotonicdecreasewiththevolume
fraction.Wedisplaytheratiooftheverticalvelocityfluctuationsforspherestothatoftheellipsoids
infig.7.4.Thequotientexhibitsalinearbehaviorwiththevolumefractionfollowing
approximatelytherelationΣδiv)Egi,heS/δiv)Egi,Eaa)Te=11∗ΦV+Φ.0+.Theinsetshowsthe
horizontalcase,alsoalinearbehaviorΣδi)Sdg,heS/δi)Sdg,Eaa)Te=´´∗ΦV+´isfound.
Infigure7.5wepresenttheparallel,τandperpendicular,τ⊥componentsoftheautocor-
relationtimeforbothellipsoidsandspheres.Weusethedefinitionofthecorrelationtime
as:∞ΨτB=,ΣΦ)Φ,Σg)Rg(7.3)
Where,Σg)istheparticlevelocityautocorrelationfunctionwhichisdefinedas,Σg)=σ
δiΣg)δiΣΦ)>.HereδiΣg)=iΣg)−σi>isthelocalvelocityfluctuation,whereσi>
wastakenasthemean(horizontalorvertical)velocity.
Theverticalcomponentshowsanoveralllargevaluefortheoblateellipsoidscompared
tothatofthespheres.Thehorizontalcomponentsoftheautocorrelationtimebetween
oblateellipsoidsandspheresareindistinguishable.
Infigure7.5thefourcurvesdecayaspowerlawsgivenbyτ≈Φ−α,Weseethatthe
valuesofαfortheparallelandperpendicularcomponentsforellipsoidsareα≈Φ.ΨΨ
andα⊥≈Φ.Ψ+respectivelyandthatforthespheresareα≈−Φ.ΨΦandα⊥≈Φ.ΩΩ

articlesPMany

))s(tτd d →→
log(102c c →→
→b →b →a →a

Ellip ||⊥Ellip Esf ||⊥Esf

69

10−2log(Φ)10−1
VΦVFigurefor7.5:oblateellipsoidsAutocorrelationandthetimesτequiv(inalentunitsofspheresgh)assplitfunctionintoofcomponentsthevolumeparallelfraction(||)
andequivalentperpendicularsphere(⊥system)tohasgraavity.radiusTheEEfjToblate=Φ.20ellipsoidandtheEaspect-ratioS=(∗isΨΦΔ−e).=TheΦ.Ω),resultsthe
areplottedinalog-logscale.

respectively.Theseexponentsarevalidinthedilutelimit.
WecalculatethepaircorrelationfunctionfordifferentvolumefractionsΦVandtheresults
areshowninfig.7.6.Thepaircorrelationfunction,forsmallervolumefractionsΦ=
Φ.ΦΦΨ,clearlyshowslargeinhomogeneitiesinthesensethatthereisa“packingformation”V
asseeninfig.Ψ.+,ofoblateellipsoids.Theseinhomogeneitiesdisappearforlargevolume
thefractionpairΦVcorrelation≥Φ,Ω.functionFurthermore,thattheinthefirstpeakintermediateisclosecasetotheforΦVorigin,=Φ.locatedΦ)weatcane=seeΦ.+in,
whichisalsopresentatΦV=Φ.Ωbutsmaller.
ThisadditionallargerpeakatΦV=Φ.Φ)couldberelatedtothelocalmaximuminthe
sedimentationvelocityfig.7.1.Bylookingatthesnapshotsastheoneshowninfig.Ψ.+
oneseesthatentirebundlesofalignedparticlesseemtodetachandmovedownfaster
beenwhichobservmightedwellinthebethesedimentationoriginofofthisfibers,peak.ThisHerzhaftkindandof“bGuazzelliundlebeha(1999),vior”wherehasthesealso
bundlessettlefasterthantheindividualfibers.

viorbehaOrientational7.1.2ForthemeasurementoftheorientationweusetheEuleranglesdescribedinfig.4.4.The
meanverticalorientation(MVO),θ,asafunctionofthevolumefraction,isshownin
fig.7.7.ForsmallervolumefractiontheMVOshowsmorealignmentwithgravityandin

70

g(r)

g(r)

15

10

5

0015

10

5

002015

10g(r)

5

Pair correlation function=0.2ΦV

12r(R)345
MPair correlation function=0.05ΦV

12r(R)345
MPair correlation functionΦ=0.001V

Results7.7.1

0012r(R)345
MFigure7.6:Pairdistributionfunctionsforoblateellipsoidsfordifferentvolumefractions,
ΦV.TheReynoldsnumberES=(∗ΨΦ−).

articlesPMany

Θ

90

80

70

60

50

71

00.05ΦV0.10.15
Figure7.7:Meanverticalorientationθforoblateellipsoidsasafunctionofthevolume
fractionΦV.−)Theoblateellipsoidaspect-ratioisΔe=Φ.Ω),andtheReynoldsnumber
ES=(∗ΨΦ.

theorientationallimitΦV→behaΦaviorcloserforonealignmentoblatewithellipsoidgravityobservisedinobservedsection(which.Ψ.Ω;(.Ψ.corresponds´.tothe

WtheealsolocalvseeerticalforthevMVelocityOanmaximumintermediateatthesamemaximum,volumeatΦVfraction≈Φ.Φ)sho,wnwhichinfig.could7.1.explainThis
ΦV>intermediateΦ.Φ1theplotmaximumshoiswsanotpresentmonotonicfordecrease.spheres.Forlargervaluesofthevolumefraction,

Figure7.8showstheorientationaldistributionfunctionPΣPdfΣθ))fortheverticalangle,
θ,fordifferentvolumefractions,ΦV.Forsmallervolumefractions,ΦV=Φ.ΦΦ1the
orientationaldistributionshowsamaximumaroundPdfΣθ)≈Φ.Ψinagreementwithfig.
7.7.Thelimitingcase(ΦV→Φ),i.e.,onesedimentingoblateellipsoid,studiedbyusin
section(.Ψ.Ω;(.Ψ.´,presentsaverticalalignmentwithgravity(θ≈2Φd),andinfig7.7we
canseeavalueofθ≈1)d.
Asthevolumefractionincreases,thedistributionsflatten,andatΦV=Φ.Φ)thedistri-
butionshowsamoderatemaximumaroundPdfΣθ)≈Φ.(),correspondingtothesimilar
intermediatemaximuminfigures7.7and7.1.Weconcludethattheverticalvelocityis
influencedsignificativelybytheorientationalbehavioralonggravity,asitiswellknown
forotherspheroidsystemsKuuselaet.al.(2003).

Figure7.9showstheorientationaldistributionfunctionPΣPdfΣφ))fortheangleφ,for
differentvolumefractions,ΦV.Theorientationaroundtheverticalslightlyincreasesfor
smallervolumefractions,anddecreaseswithlargervolumefractions.Similarbehavioris
alsofoundforthethirdEulerangleψ.WeconcludethattheEuleranglesφandψarenot
muchinfluencedbythevolumefraction.

72

0.0080.050.173

Results7.7.1

0.008210.050.1731815))θ12P(cos(96300.20.4cos(θ)0.60.81.0
Figure7.8:ThedistributionfunctionPΣPdfΣθ))forthemeanverticalorientationθfor
difnumberferentEvS=olume(∗ΨΦ−fractions.).Theellipsoidaspect-ratioisΔe=Φ.Ω),andtheReynolds

))φP(cos(

10

5

0.0080.0500.173

00.20.4cos(φ)0.60.81.0
Figure7.9:ThedistributionfunctionPΣPdfΣφ))forthemeanverticalorientationφfor
oblateellipsoids.Theaspect-ratioisΔe=Φ.Ω),andtheReynoldsnumberES=(∗ΨΦ−).

changesOrientational7.1.3ToquantifytheorientationoftheoblateellipsoidsweintroducethequantityΨ=σ
ΩPdfΣθ)−Ψ>thatwasalsousedinKuuselaet.al.(2003),HerzhaftandGuazzelli(1999)

articlesPMany

0.5

73

0.2)−1>−0.1θ=<2cos(−0.4Ψ−0.70.4/2.4−10.4/1.60.4/0.800.05Φ0.10.150.2
VFigure7.10:OrderparameterΨasafunctionofthevolumefraction,ΦVforthreediffer-
entaspect-ratiosΔe=Φ.(/Ω.(;Φ.(/Ψ.+;Φ.(/Φ.1.

aswereorientationalperpendicularordertogravityparameter,.randomlyItwouldorientedgive−orΨ,Φalignedor+Ψwithifallgrathevityoblaterespectivelyellipsoids.
Figure7.10showsthebehaviorofΨagainstΦV,forsmallervolumefractions,ΦV≈
Φ.ΦΦΨ−Φ.Φ1theorderparametertakesnegativevaluesevidencingthealignmentalong
gravityandinagreementwiththelimit,ΦV→Φ(oneoblateellipsoid,section
(Φ.Ψ.Ω;≥(.Φ.Ψ.´Φ1).apositiveApproximatelyorderatparameterΦV≈Φ.impliesΦ1thetheorderorientationparameterisisperpendicularzero.Fortolargragerv-
V.ityIntherangeofΦV≈Φ.ΦΦΨ−Φ.Φ1,ΨhasalocalminimumclosetoΦV≈Φ.Φ)wherewe
foundalocalmaximuminfig.7.7andfig.7.1.Thesimulationswererepeatedwithtwo
otherdifferentaspectsratiosΔg=Φ.(/Φ.1;Φ.(/Ω.(andweobservedsimilarbehavior.In
thecaseofoneoblateellipsoid(ΦV→Φ)theorderparameterΨhasavalueverycloseto
−Ψastheellipsoidaspect-ratioisincreased.

numberReynoldsModerate7.1.4Figure7.11presentsthemeanverticalsedimentationvelocityforoblateellipsoids(
squaredline)andtheequivalentspheres(◦circlelined)asafunctionofthevolumefrac-
tionatmoderateReynoldsnumber(ES≈7).Inourpreviousworkthissimulationmethod
hasbeenusedwithsuccessuptoES≈ΨΦH¨oflerandSchwarzer(2000)andKuuselaet.
al.(2003).Theintermediatemaximumfortheellipsoidsisnotobservedinfig.7.11as
seeninfig.7.1atlowReynoldsnumber.
AcomparisonwiththephenomenologicalRichardson-Zakilaw(continouslineinfig
7.11)showsanexponentaroundc<eS=´.ΩforspheresandcEaaTe=(.Φforellips-
oids.Inbothbothcases,thedatafollowtheRichardson-Zakilawratherclosely.These

74

3

2.4)s(v||<v>

1.8

Outlook7.7.2Conclusionsand

SphEllip

00.040.08Φ0.120.16
Vline)Figureand7.11:asphereMean(dash-circlesedimentationline),vaselocityfunctioniΣg)offorthethevolumeoblatefraction,ellipsoidΦV.The(dash-squaredoblate
ellipsoidaspect-ratioisΔe=Φ.(/Ψ.),theequivalentspherehasEEfjT=Φ.20andthe
ReynoldsnumberES≈0.

elimitxponents(c≈((c.)<)eSand=´a.ΩturbandulentcEaaTeparticle=(.Φ)systemare(cbetween≈Ω.)),thelowRichardsonparticleandReZakiynolds(1954).number
Figure7.12(top)presentstheverticaldistributionfunction,PΣc6sΣθ))atmoderateReyn-
oldsnumber.Forallvolumefractions,PΣc6sΣθ))presentsalargerdistributionaround
PdfΣθ)≈Φ(θ≈2Φd),whichtendstobemuchflatter(PdfΣθ)≥Φ.Ψ))thaninfig.7.8.For
thec6sΣψ)other≈Φ,angularandforvlarariables,gervφolumeandψ,fractions,thedistribtheyutionsfollowshoawaconstantpeakbehaaroundvior.c6sΣφ)≈Φ,
Thebottomoffig.7.12showsthebehavioroftheorientationalparameterΨatmoderate
Reynoldsnumber.Foroneoblateellipsoid(ΦV→Φ),thevalueofΨiscloserto−Ψ
(verticalalignment),asinthecaseoflowReynoldsnumberfig.7.9.Theintermediate
maximum,forΔe=Φ.(/Ψ.),isnotobservedinfigure7.10,andthepointatwhichthe
(bottom).orientationalThisshiftparameterinΨΨisvalsoanishes,seenisinshiftedthecaseslightlyoftofibers,thewhenright(theΦVRe≈Φ.ynoldsΨ)fig.number7.12
increases,byafactorof5,Kuuselaet.al.(2003).

ConclusionsandOutlook7.2

Wehavesimulatedthesedimentationofoblateellipsoidsatsmallvolumefraction(ΦV≤
Φ.Ω)andsmallReynoldsnumber(ES≈ΨΦ−)).Wehavefoundthatatintermediatevolume
fractionthesettlingvelocityexhibitsalocalmaximumwhichtoourknowledgehasnever
beenreportedintheliterature.Itwouldbedesirabletoexperimentallyverifythismax-
imum.

articlesPMany

0.60.4)θcos(

0.0080.1100.173

1.00.8

75

0.110180.1731512))θP(cos(96300.00.20.4cos(θ)0.60.81.0
0.30.1)−1>−0.1θ−0.3=<2cos(−0.5Ψ−0.7−0.900.05Φ0.1V0.150.2
Figure7.12:Thetoppictureshowsthedistributionfunction,PΣPdfΣθ))ofthemeanver-
ticalorderorientationparameteratbehadifvesferentwithvtheolumevolumefractionfractionΦV.ΦV.TheThebottomoblatepictureellipsoidshowsaspect-ratiohowtheis
Δe=Φ.(/Ψ.),theequivalentspherehasEEfjT=Φ.20andtheReynoldsnumberES≈0.

Thislocalmaximuminthevelocitycanberelatedtothenonmonotonicbehaviorofthe
verticalorientationoftheoblateellipsoidsalonggravity,whichisshowninfigures7.7,
and7.8,andcanbeexplainedbythe“cluster”formationshowninfig.2.7,whichisalso
foundinfiber-likesuspensionsHerzhaftandGuazzelli(1999).
AtlowReynoldsnumbertheorientationalorderparameterΨvanishesaroundΦV≈Φ.Φ1
fig.7.7.AsΦVdecreasestheorientationalalignmentwithgravityincreasesasshownin
fig.7.7and7.12(bottom),asforlowandmoderateReynoldsnumberandinthelimit
ΦV→Φasingleellipsoidalignswithgravity,whichisadistinctivefeatureofthesteady-
stateregimeforasingleoblateellipsoidasreportedinsections4.1.2,4.1.3andGaldiet.
(2001).al.

76

7.7.2ConclusionsandOutlook

WealsopresentdataatmoderateReynoldsnumber(ES≈0)forthesedimentationve-
locityofoblateellipsoidsasthevolumefractionΦVisincreased.Asinthecaseoflow
Reynoldsnumbertheellipsoidshaveasmallersedimentationvelocitythantheequivalent
spheres,fig.7.1and7.11.ThedataforellipsoidsandspheresfollowtheRichardson-
ZakilawRichardsonandZaki(1954)withexponents(cEaaTe≈´.Ω,ES=ΨΦ−))and
(c<eS≈(.Φ,ES=0)respectively.ThePΣc6sΣθ))distributionpresentsalargeralignment
ofellipsoidswithgravitycomparedtothosewithsmallReynoldsnumber.Thevanishing
oftheorderparameterisslightlyshifted(ΦV≈Φ.Ψ)totherightastheReynoldsnumber
increases(seefig7.12,bottom).Thealignmentwithgravityispresentforsmallandmod-
erateReynoldsnumberasΦV→Φ,asshowninfig.7.12(top)andfig.7.8,whichisin
agreementwiththeorientationalbehaviorofasingleellipsoid(section4.1.2;4.1.3).All
thesimulationsinthisworkarelocatedinthesteady-fallingregime,chapters´and(,for
ellipsoid.oblatesinglea

8Chapter

Diffusion

ThislipsoidschapteratlowisanddedicatedmoderatetotheReinvynoldsestigationnumbers.oftheWedifwillfusivbeeginbehawithvioranofintroductionsedimentingtheel-
commontheoreticaltoolsusedtostudythediffusivebehaviorofsedimentingparticles.
eWexplorediscussthethebehaviorresultsofforthesystemellipsoidsunderinvcomparisonariationsoftothetheequidynamicalvalentspheres.viscosity,Also,ellipsoidwe
densitytionally,andwepresentaspect-ratio.theanomalousFurthermore,difwefusiongivefortheoblateorientationalellipsoids.difLastlyfusion,webehavior.summarize.Addi-

oductionIntr8.1

AswasexplainedinchapterΨ,inasedimentingsuspensiontherearelargechangesin
thesiondominatesconcentration,overthatthecanthermaleitherBrobewniantemporaldiforfusion,spatial.intheThissystemhconsideredydrodynamic-likhere.edifThefu-
dimensionlessquantitythataccountsfortherelativeimportanceofthishydrodynamicdi-
fussionprocessoverthethermaldiffusionistheP´ecletnumberPS.WhenPSismuch
largerthan1,BrownianmotioncanbeneglectedRamaswamy(2001),KuuselaandAla-
Nissila(2001).ForsmallervaluesofPSσσΨ,inturn,thehydrodynamicdiffusionisnot
ant.vreleInaregimewhere(PS>>ΨandESσσΨ),sedimentingspheresundergolong-ranged
hydrodynamicfluctuations(seechapter1,sec.1.6).Theycauseinthelongtimelimit,
thatthefluctuatingparticlemotionbecomesdiffusive,Nicolaiet.al.(1995).There-
fore,wecandeterminethelongtimebehaviorofthisrandom-likeparticlemotions,by
ebyxaminingcomputinghowthetheparticleparticlevvelocityelocityfluctuationbecomesuncorrelated.autocorrelationThisefunction,xaminationΣg),(seeisKrealizeduusela
andAla-Nissila(2001)andchapter+,sec.+.Ψ).Thisquantitywasdefinedinsec.+.Ψ,as
ws:follo

,Σg)=σΔiΣg)ΔiΣΦ)>

77

(8.1)

78

oductionIntr8.8.1

wheretheaverageistakenoveralltheparticlesinmotion.HereδiΣg)=iΣg)−σi>
isthelocalvelocityfluctuation,whereas,Σg)isusedtodefinethehydrodynamicdiffu-
sioncoefficient<,inanalogywiththeBrowniantracerdiffusioncoefficientofBrownian
particlesimmersedinafluid(KuuselaandAla-Nissila(2001)).<isdefinedbytheGreen-
as:formulauboK

∞Ψ<=RΦ,Σg)Rg
whereRisthespatialdimension,andforthediscretecaseasfollows:

∞ΨΨ<=R[Ω,ΣΦ)+,Σc−τ)4
=(c

(8.2)

(8.3)

whereτisthetimestep.
Anothermethodtoinvestigatethelongtimebehaviorofasedimentingparticles,istocom-
putethemeansquareparticledisplacement(hereafterM.S.D.),Kalthoffet.al.(1996),
i.e.thesecondmomentoftheparticledisplacement,HerzhaftandGuazzelli(1999).Itis
ws:folloascalculated

σk)Σg)>3=σ[kΣg)−ΣkΣΦ)+σi>g)4)>(8.4)
herethebracketsindicateanaverageovertheensembleofthoseparticlesthatareabove
thefinalbottompositioninthecontainer,andσi>istheirmeanvelocitytakenofall
particleswithi=Φ.Asimplediffusivebehaviorischaracterizedbyalineargrowthwith
time.Theself-diffusivity<,(Nicolaiet.al.(1995)),canbedeterminedfromtheslopeof
line.thisAponentslargetograanosotropvity,yhasbetweenbeenthereportedM.S.D.byforothertheparallelauthors()(Nicolaiandet.perpendicularal.(1995);(⊥)Laddcom-
(1996);Kalthoffet.al.(1996)),whofound<>><⊥.However,thisanisotropybe-
comeslessimportantforlargeReynoldsnumbersandvolumefractions(Kalthoffet.al.
(1996),Nicolaiet.al.(1995),KuuselaandAla-Nissila(2001)).
Both,<and<⊥areobtainedbyexamininglongtimebehavioroftheparticledisplace-
mentbythefollowingrelations:

σkΣg)⊥>)≡Ω<⊥g;σkΣg)>)≡Ω<g.(8.5)
Despitetheclearevidencefornormaldiffusivebehaviorfoundinthecurrentresearch,
someexperiments,(Rouyeret.al.(1999))andsimulations(MiguelandPastor-Satorras
(2001)),hadshownthatthevelocitiesandtrajectoriesofnon-Brownianparticlesina
quasicomponentbidimensionalandasuperdiffluidizedfusivebed,behaexhibitvioraparalleldiffusitovegrabehavity.viorThisalonglatterthesuperdifperpendicularfusive

Diffusion

40

35

30

25

200

3

6

a)

9

7260

48vertical36

24

8

b)

201612horizontal

79

Figure8.1:(a)Abidimensionalvelocitymapshowingfasterandslowerellipsoids.(b)
TypicalBrownian-liketrajectoriesinellipsoidssedimentationinthelaboratoryreference
frame.TheReynoldsnumberisΩ×ΨΦ−)andtheellipsoidaspect-ratioisΔe=Φ.(/Ψ.+.
behaviorisalsoknownas,“anomalousdiffusion”,anddefinedasthegrowthofthesecond
by:momentumorder

σk)Σg)>≡gα

(8.6)

withα=Ψ.
Furthermore,wecouldextendequation8.5totheangularcase,anddefinetheM.S.D.as:

σθ)Σg)>≡Ω<θg(8.7)

Withregardtosedimentingparticleswithnon-sphericalshape,apart,fromafewexcep-
tions(e.g.Kuuselaet.al.(2001)andchapters(,)),verylittleisknownaboutthediffusive
behavior,beitfromsimulation,experimentortheory.Non-sphericalparticles,withro-
tationalsymmetryaroundanaxis(oblateorprolateellipsoids),givesrisetorotational
degreesoffreedom,andgenerallyplaysanimportantroleinthesedimentation,(Kuusela
et.al.(2003)andchapter+).

80

600⊥ll450(a)>*2300<x150

3060)time t(ts32⊥Sph⊥Ellip24(c)>*216<x

8

210⊥ll110log(dR/dt)(b)(b)010

19010))log(t(ts600llSphllEllip450(d)>*2300<x

150

Results8.8.2

210

204060306090120
time t(ts)time t(ts)

Figure8.2:Figure(a)showstheparallel()andperpendicular(⊥)componentsofM.S.D.
forsedimentingellipsoids.Picture(b)showstheslopebehaviorintime,ofthecurvesin
pictureballistic(a),andinanon-diflog-logfusivescale.regimes.ThethickThelodottedwerlinespictures,in(c)between,and(d)presentshowthethegrowthcomparisoninthe
betweenellipsoidsandsphereswithregardtotheperpendicular⊥(c)andparallel(d)
componentsoftheM.S.D..ThesphereradiusisEEfjT=Ψ.ΦΨ.TheReynoldsnumberis
Ω×ΨΦ−)andtheaspect-ratioΔe=Φ.(/Ψ.+.Thenumberofellipsoidsisoftheorderof
0.ΨΦResults8.2Infigure8.1(a),wepresentbidimensionalsnapshotofthevelocitymapforsedimenting
ellipsoids.Themapischaracterizedbycomplexandhighlyfluctuatingvelocityvectors

Diffusion

81

pointinginawiderangeofdirections,withverydifferentmagnitudes.Thesnapshot
reflectsthecomplexityoftheswirlsandchannelsthatareverysimilartothosereported
veolumexperimentallyfractioninofSethegre´etparticlesal.in(2001)theandsysteminisΦsimulations=Φ.Ω.byRouyeret.al.(1999).The
VInfigure8.1(b)wepresentatypicalellipsoidtrajectoriesinacontainerscaledinunits
ofthecharacteristicsmayorradiiobservEMed.byTheNicolaitrajectorieset.al.are(1995)ratherandPecomplicatedyssonandandeGuazzellixhibitman(1999),yofsuchthe
asthepresenceofloopsandastronganisotropybetweentheparallelandperpendicular
fluctuationsintheellipsoidtrajectory.
exhibitPicturesthe(a)larandge(b)inanisotropfigurey8.2betweenshowthetheparallelM.S.D.(for)andsedimentingperpendicularellipsoids.(⊥)Thecomponents.graphics
Infiguregeneral,8.2(b),bothinalog-logcomponentsscale,presentbyaanparallelinitialthicksocalleddotted-lineballisticinrebetweengime,asofthedepictedcurves.in
Thisballisticregimeisproportionalto≈g).Then,wefoundanon-diffusivebehavior
whichisquiteinterestingandwillbediscussedinsection)ofthischapter.
Anotheraspectofthelong-timebehavioristhat,asaconsequenceofthefinitesize
container,theellipsoidsreachasaturationregime.Thisischaracterizedbyfluctuations
aroundthefinalpositionofeachellipsoidinthecontainerbottom)(Kalthof)fet.)al.(1996)).
Thetimeingraphicsunitsofaregh,(seenormalizedeq.Ω.in´Φ).bothcomponentsaccordingtok∗≡k/EEfjandthe
ThecomparisonbetweentheM.S.D.forellipsoidsandthatforspheresisgiveninfigure
8.2,numberin(theES⊥≈(c)ΨΦand−))and(d)forflatcomponents.ellipsoids,Inourthatthesimulations,componentswefoundforatloellipsoidswerReareynoldsmuch
largerthanthoseforspheres.Thisphenomenoncanbeexplainedbythepresenceofa
muchlargernetdisplacementforellipsoidsthanspheres,ascanbeseeninfigure8.1.
Itisinterestingtopointout,thattheperpendicularcomponentoftheM.S.D.,fig.8.2
(c),foraveryflatellipsoid,experiencesapeakwhichthendecreasesabruptly,beforethe
saturationregimeisreached.Thisbehavioriscompletelynewandcanonlybeassociated
totheparticleshape,sincethespheresonlyechibitamonotonicincrement,(seefig.8.2
with(c)).aThisfixedlargepeakaspect-ratio,inthebutvellipsoidariableviscosityperpendicularorparticlecomponent,densityis..presentinallsimulations

8.2.1Changeindensity,viscosityandaspect-ratio

Theslopeforeachoneofthecurvesinfig.8.3(a)and(b)becomeslargeastheellipsoid
densityisincreased,keepingconstanttherestofparametersinthesystem,showingthe
ponentsincreaseofincreasestheinertiamuchinfasterellipsoidsthanforandspheres.spheres.ForFortheparallelellipsoids,thecomponentslopetoingrabothvity,com-the
denserellipsoidsdiffusemuchfaster,quicklyreachingthesaturationregime.Thiscan
be8.3(c)seenandby(d)wecomparingpresentthethecurvesM.S.D.forρ(beha=).viorΦandasρthe1=ΨΨ.kinematicalΦinfigureviscosity8.3(a).isInchanged.figure
Forallofthecases,theincreamentintheviscositydecreasestheslopeofeachonthe

Results8.8.282400600(a)(b)4002>*2>*200
<x<x200ρ1=5.0ρ1=5.0
ρ2=7.0ρ2=7.0
ρ3=9.0ρ3=9.0
ρ4=11.0ρ4=11.0
020406080100120020406080100120
time t(ts)time t(ts)
µ1,Ell=7.7µ1,Sph=7.7
µ2,Ell=6.7300µ2,Sph=6.7
400µ3,Ell=5.9(c)µ3,Sph=5.9
µ4,Ell=5.3µ4,Sph=5.3
2>*2>*200(d)
<x<x20010001530450306090120
time t(ts)time t(ts)
6006001.00.4/1.61.20.8/1.64801.6/1.61.2/1.64801.61.4
3603602>*(e)2>*(f)
<x<x2402401201200408012016020024028004080120160200240280
time t(ts)time t(ts)
Figure8.3:Figuresa)showtheparallelcomponentoftheM.S.D.forellipsoidsand
equivalentspheresb),astheparticledensitychanges.Figuresc)andd)presentthevari-
ationforthedynamicalviscosityandfigurese)andf)showthechangeintheoblate
aspect-ratioandtheequivalentradius,respectively.TheReynoldsnumberisES≈ΨΦ−),
thespheresradiusisEEfjT=Ψ.ΦΨandtheellipsoidaspect-ratioΔe=Φ.(/Ψ.+.

Diffusion

>*2<x

1512

9

6

3

0

0.4/1.60.8/1.61.2/1.61.6/1.6

83

0306090120150180210240
)time t(tsFigure8.4:BehavioroftheperpendicularcomponentofM.S.D.forellipsoidsasthe
aspect-ratiochanges(Δe=Φ.(/Ψ.+,Δe=Φ.1/Ψ.+Δe=Ψ.Ω/Ψ.+andΔe=Ψ.+/Ψ.+).
TheReynoldsnumberisES≈ΨΦ−).

curvesInfigure8.3(c)and(d).TheincreaseoftheviscosityreducestheM.S.D.inthe
suspension.Also,asforparticledensityvariations,thechangesintheviscositylead,for
lowerReynoldsnumbers,toadirectlyproportionalincrementintheM.S.D..
Thepeaksintheellipsoidperpendicularcomponent(seefig.8.4),ispresentinallofthe
simulationswherethekinematicalviscosityorparticledensityarechanged,withafixed
aspect-ratioandlowerReynoldsnumber(ES≈ΨΦ−)).
Next,wepresenttheM.S.D.fordifferentellipsoidaspect-ratiosΔeandequivalentsphere
radiiEEf.Thevariationintheaspect-ratiogoesfromaflatellipsoiduptoasphere,while
thevolumefractionΦ≈Φ.Ωiskeptconstant.Theresponseundervariationsoftheoblate
aspect-ratio,presentsalinearincreaseintheslopeofeachcurve,inbothellipsoidsand
spheres,asisshowninfigures0.´(e)and(f),respectively.
TheslopegrowingintheM.S.D.astheaspect-ratioincreases,becomingone,showsthat
thesphereshaveamuchlargerM.S.D.thantheellipsoids.Theellipsoids,ontheother
hand,havemoreresistancetosediment.Thischaracteristicispresentinbothcomponents
(seefigure8.3(e)and8.4).Asfortheparallelcomponentfig.8.3(e)and(f),spheresand
ellipsoidspresentthesamemonotonicbehaviorbeforetheyreachthesaturationregime.
Infigure8.4,wecanseehowtheperpendicularcomponentoftheM.S.D.approachesthe
saturationregime,astheaspect-ratiogrows.Thepeakisshiftedtotheleftandbecomes

84

120

902>Θ<603000

60

452>φ<30(a)ρρ1=7.0=5.015
2=9.0ρ3=11.0ρ430609012000
)t(ts60

Results8.8.2

(b)=5.0ρ1=7.0ρ2ρ=9.03=11.0ρ410t(ts)2030

452>ψ<(c)3015ρρ12=7.0=5.0
=9.0ρ3=11.0ρ400612t(ts)182430
Figure8.5:AngularM.S.D.fortheEuleranglesθ-fig.(a),φ-fig.(b)andψ-fig.(c)under
variationsofellipsoiddensity,ρ(=),0,2,ΨΨ.TheReynoldsnumberisΩ×ΨΦ−)andthe
ellipsoidaspect-ratioisΔe=Φ.(/Ψ.+.

muchlargerandsharperastheaspect-ratiobecomesΨ(thesphere).

diffusionOrientational8.2.2Figure8.5showstheangularM.S.D.,σΔΘ>)forthethreeEulerangles(seefigure4.4
inchapter´),undervariationsoftheellipsoiddensity,whichcorrespondtothetransla-
tionalbehaviorshowninfigure8.3(a).
Itisinterestingtonote,thattheangularM.S.D.followsabehaviorwhichissimilarto
thecorrespondingtranslationaldegreesoffreedom,inthesensethattheangularM.S.D.
exhibitsatthebeginningofthesedimentationafastergrowthandthenapproachesto

Diffusion

85

thecomparedsaturationtotheregime.translationalThisapproachcomponentsto(seesaturationfig.is8.3quite(a)),fandasteris(seepresentfig.for8.5),theasitthreeis
angles.EulerAstheoblatedensityisincreased,theangularM.S.D.foreachoneoftheEulerangles
shiftingslightlyintheincreases.finalvThealuereleofvtheantvsaturationariationinrethegime(seeangularfig.M.S.D.8.5).Wisecanpresentobservinetheupw(approx-ard
imately),thatequalvariationsinellipsoiddensitycauseequalupwardincrementsinthe
alue.vsaturationIttheissamealsokindinterestingofupwtoardobservshiftinge,thatasthedescribeddecreaseaboofve,thebutwithkinematicalasmallerviscosityshiftinggivesvalueriseofto
thefinalsaturation(seefig.8.6).
reducingIncreasingthethekinematicalimportanceofviscosityinertial,increasesforcesbytheusingsaturationahigherstateinmasstheofEulertheangles.ellipsoidor

8.2.3Non-diffusivedynamicalbehavior

EWΣeg)no=wσkstudy)Σg)the>,behaeq.vior(8.4).oftheThistimetypederiofvativeofcharacterizationtheM.S.D.,ofthedefineddiffusiasveREΣgbeha)/Rgviorwithin
sedimentationisalsousedintheworkofRouyeret.al.(1999)andMiguelandPastor-
whoseSatorrasconstant(2001).vInalueaiscaseequivwherealentREto/Rgone≡halfofPdcfgNcgthe,difwefusionhaveaconstant,simpledif(seefusieq.ve8.5).behaIfviorwe,
findnon-diffusivebehavior,thenRE/Rg=PgSandthetemporalbehavioroftheM.S.D.is
givenbyeq.(8.6).

densityparticletheinChange

Figure8.7(a)presentstheperpendicularcomponent(⊥),ifthespheredensityischanged.
Forsimpleρ(=dif)fusi.Φvetheandevolutionsaturation.oftheAncurveinterestingintime,aspectexhibitsheretheconsistsabovinerethefgimes,actthatballistic,when
themuchparticleshorterindensitytime.ρisBasicallyincreased,the(ρ=)increment.Φ→ofρ=theΨΨ.Φinertial)theforces,simpledifreducesfusivetheretimegimeofis
thesimplediffusiveregime.Thisgeneralbehaviorforspheresisinagreementwiththe
resultspresentedby(Rouyeret.al.1999)and(MiguelandPastor-Satorras2001).
ForAfterthethat,aparallelnon-diffusicomponentvebeha()viorfigure8.7characterized(b),allbythegα,curvwithesaneexhibitxponentaαballistic=Ψ.(recangime.be
observed(Rouyeret.al.(1999)).Liketheperpendicularcomponent,thelengthintime
oftheparallelcomponentisreducedforlargespheredensities,asshowninfigure8.7(b).
Infigure8.8(a)and(b)wecanobservethebehaviorofthetimederivativeoftheM.S.D.
aforballisticellipsoids.regimeForandthethenaperpendicularnon-difandfusivereparallelgime,whichcomponentsupto(seenowfigureisnot8.8)knowewnfirstinfindthe

86

120

90>2Θ<603001

75

50>2φ<25=5.0νEll,1=7.0νEll,2=9.0νEll,3=11.0νEll,4021110t(t)s60

Results8.8.2

=5.0νEll,1=7.0νEll,2=9.0νEll,3=11.0νEll,4302010)t(ts

45>2ψ<3015ννEll,2Ell,1=7.0=5.0
ννEll,3Ell,4=9.0=11.0
0010t(t)2030
sFigure8.6:AngularM.S.D.fortheEuleranglesundervariationsinthekinematicalvis-
cosity.TheReynoldsnumberisΩ×ΨΦ−)andtheellipsoidaspect-ratioisΔe=Φ.(/Ψ.+.

literature.TheexponentsinbothcomponentsareequalαegaaEa=Ω.),αeEge=Ω.)and
largerthantheslopeoftheballisticregime.
Incontrasttothespheres,thedensityvariationdoesnotmodifysignificativelythenon-
diffusiveregime’scharacteristictime.Thiskindofbehaviorforthediffusionofnonspher-
icalparticleswasneverbeforereporteduptonow,andtheexperimentalverificationwill
benecessary.Fromtheplot,wecanalsoextractthat<vEgi<Sdg,thereforethediffusive
anisotropic.highlyisgimereForbothellipsoidsandspheres,theM.S.D.inbothcomponentsaredisplacedupwardas
theparticledensitygrows.AsexpectedatlowReynoldsnumber,anincrementinthe
inertialforces,producesalargegrowthoftheM.S.D..

.elyvrespecti,viorbehaevfusinon-diftheandevfusidiftherepresent,linesdashedandsolidThe.+.Ψ/(.Φ=eΔisaspect-ratioellipsoidthe,)−ΨΦ×ΩisnumberynoldsReThechanged.isdensitytheasellipsoidsforM.S.D.theofevativderitimetheofcomponents(c)paralleland(a)⊥perpendicularthewshofiguresThe8.8:Figure)OΣ)NΣreplacementsPSfrag=11.04ρ=9.03ρ=7.02ρ=5.01ρ/dt)log(dR))slog(t(t110010−110−210−310210110010)OΣ)NΣreplacementsPSfrag=11.04ρ=9.03ρ=7.02ρ=5.01ρlog(dR/dt)))slog(t(t010−110−210−310210110010.elyvrespecti,viorbehaevfusinon-difande87

vDiffusion

fusi)NΣ

difOΣ)

therepresentlinesdashedandsolidThe.ΦΨ.Ψ=jTEfEradiusspheresthe,)−ΨΦ×ΩisnumberynoldsReThechanged.isdensitytheasspheresforM.S.D.theofevativderitimetheofcomponents(b)paralleland(a)⊥perpendicularthewshofiguresThe8.7:Figure)NΣreplacementsPSfrag=11.04ρ=9.03ρ=7.02ρ=5.01ρ/dt)log(dR))slog(t(t110010−110−210−310210110010replacementsPSfrag=11.04ρ=9.03ρ=7.02ρ=5.01ρ/dt)log(dR))slog(t(t010−110−210−310110010

88

010

110

Results8.8.2

ΣΣOO))

10−1ΣΣNN))100ΣΣOO))
log(dR/dt)log(dR/dt)−11010−2ν1,Sph=7.7ν1,Sph=7.7
=6.7ν=6.7νν2,Sph3,Sph=5.910−2ν3,Sph2,Sph=5.9
PSfragreplacements−3ν4,Sph=5.3PSfragreplacementsν4,Sph=5.3
10101102ΣN)101102
ΣO)log(t(ts))log(t(ts))
Figure8.9:Thefiguresshowtheparallel(b)andperpendicular⊥(d)componentsof
RetheynoldstimederivnumberativeisofΩ×theΨΦ−)M.S.D.,theforspheresspheresradiusastheEEfjT=kinematicalΨ.ΦΨ.viscosityischanged.The

viscositykinematicaltheinChange

Nextwestudythebehaviorofthesystemunderchangesinthekinematicalviscosity.Fig-
ure8.9(a)and(b)showstheM.S.D.forspheresintheperpendicular(⊥)andparallel(||)
components.Figure8.9(a)showswelldefinedballisticandsimplediffusiveregimes.The
plotsshowaslightincrementintimeforthesimplediffusiveregimeasthekinematical
viscositydecreases.Also,thereisaclearupwardshiftingasthekinematicalviscosityde-
creases.ThespheresincreasetheirM.S.D.asthekinematicalviscositybecomessmaller.
Inthecaseoftheparallel(||)component8.9(b),wefoundaballisticandnon-diffusive
regimes.Theexponentαforthenon-diffusivebehavioristhesameinallthecurvesand
α=Ψ.((Rouyeret.al.(1999)).Suchasfortheperpendicular(⊥)component,the
upwardshiftingisalsopresentfortheparallel()component.
Theellipsoidssystemisstudiedunderthesamekinematicalviscosityvariations.Figure
8.10showsthebehavioroftheperpendicular(⊥)(a)andparallel()(b)componentsof
M.S.D..theTheperpendicularcomponent(seefigure8.10(a))exhibitsballisticandnon-diffusive
behavior.Thecurves,asinthecaseofspheres,showanupwardshiftingasthekinemat-
icalviscositydecreases.Astheviscositydecreasestheexponentαforthenon-diffusive
regimebecomesgraduallymuchsmaller:ν=0.0→α=Ω.(toν=).Φ→α=Ω.Ψ.
Fortheparallelcomponent(),thedifferencebetweentheballisticandthenon-diffusive
behaviorisnotpronounced,andforthelowerkinematicalviscosityν=).Φ→α=Ω.Φ
thenon-diffusivebehaviorinthesystempracticallydisappears.

.)−ΨΦ×ΩisnumberynoldsReThechanges.radiusspheretheasspheresforM.S.D.theofevativderitimetheofcomponents(b)paralleland(a)⊥perpendicularthewshofiguresThe8.11:Figure)OΣ)OΣ)NΣreplacementsPSfrag))slog(t(t−110210110)OΣ)NΣ)NΣreplacementsPSfrag))slog(t(t−210210110.+.Ψ/(.Φ=eΔisaspect-ratioellipsoidthe,)−ΨΦ×ΩisnumberynoldsReThechanged.isviscositykinematicaltheasellipsoidsforM.S.D.theofe89

vΣΣOO))

atiChangeintheellipsoidaspect-ratioandsphereradius
Inthecaseofvariationsoftheequivalentradii,eq.(3.28),figure8.11showsthebehavior
oftheparallel()andperpendicular(⊥)components.Astheequivalentradiidecreases,

v210

deriDiffusion

time110

the=1.0Req,1=1.2Req,1=1.4100RReq,1=1.6
eq,1log(dR/dt)−110

of2=1.0R10eq,1=1.2Req,1=1.4Req,1=1.6Req,1110log(dR/dt)010

components(c)⊥perpendicularand(a)parallelthewshofiguresThe8.10:Figure)NΣreplacementsPSfrag=5.34,Ellν=5.93,Ellν=6.72,Ellν=7.71,Ellνlog(dR/dt)))slog(t(t110010−110−210210110)OΣ)NΣ)NΣreplacementsPSfrag=5.34,Ellν=5.93,Ellν=6.72,Ellν=7.71,Ellνlog(dR/dt)))slog(t(t010−110−210−310210110

Results8.8.2

Furthermore,intheparallelcomponent(seefigure8.12(b)),wecanseethechangeinthe
difslopefusivthatebehaviorcharacterizesforthespheresv(ariationΔe→fromΨ).Thenon-difefusixponentsvebehagoesviorfromfor(Δe=ellipsoidsΦ.Ω),toαthe=Ω.non-´)
oblateellipsoidsto(Δe=Ψ,α=Ψ.(,Rouyeret.al.(1999))spheres.

Infigure8.12(a)and(b)wepresenttheM.S.D.behaviorforellipsoidsastheaspect-
ratioischanged.Herewewanttodrawattentiontoaninterestingchangefromanon-
diffusiveregimetosimplediffusivebehaviorthatcanbeobservedintheperpendicular
(⊥)component(seefigure8.12(a))astheellipsoidaspect-ratiotendstoone(Δe→Ψ,
sphere).Theslopeofthecurveinthenon-diffusivepartchangesfrom(Δe=Φ.Ω),α=
Ω.Ω)to(Δe=Ψ,α≈Φ).

Fortheperpendicular(⊥)component,allthecurveslookquitesimilartoeachother,with
aregimeslightisdifmuchferencefasterbeingthanthatintheforcasespheresofEEfwith=EΨ.EfΦ.=TheΨ.e+thexponentαapproachforthistothenon-difsaturationfusive
regimeisα=Ψ.(.

wecanseeinfigure8.11(a)thesamekindofupwardshiftinginthecurves,asviscosity
ordensity,ischangedinthesystem.Asthesphereradiusisincreasedthepresenceofthe
simplediffusiveregimeisincreasedduringaverylongtime.

90

102Δ r1=0.4/1.6
=0.8/1.6 rΔ2=1.2/1.6 rΔ3=1.6/1.6 rΔ4110

log(dR/dt)010

=0.4/1.6 rΔ1=0.8/1.6 rΔ2100Δ r3=1.2/1.6
rΔ=1.6/1.64−110log(dR/dt)

.)−ΨΦ×ΩisnumberynoldsReThechanges.aspect-ratioellipsoidtheasellipsoidsforM.S.D.theofevativderitimetheofcomponentsparalleland(a)⊥perpendicularthewshofiguresThe8.12:Figure)OΣ)OΣ)NΣreplacementsPSfrag))slog(t(t−110210110)OΣ)NΣ)NΣreplacementsPSfrag))slog(t(t−210210110

Diffusion

600400(a)3004002>*2>*200
<x<x200ρρ1=7.0=5.0100
2=9.0ρ3=11.0ρ40501001502000
)time t(ts=7.7ν1,Ell=6.7ν2,Ell400νν3,Ell=5.3=5.9300
4,Ell2>*2>*200
<x<x200(c)100

01057035time t(t)s480

360(e)>*2240<x

0480

360>*2240<x

1201200.4/1.60.8/1.61.2/1.61.6/1.600900600300)time t(ts

(b)

=5.0ρ1=7.0ρ2=9.0ρ3=11.0ρ450100150time t(t)200250
s=7.7ν1,Sph=6.7ν2,Sph=5.9ν3,Sph=5.3ν4,Sph

(d)

50100time t(t150)200250
s

(f)

1.01.21.41.6900600300time t(t)s

91

Figure8.13:Weshowthecollapseofthecurvesfromfig0.´.Allthecurvescollapse
quitewelljustifyingthetransformationrule,table0.Ψ.
Similarity8.2.4

Inpresentedtable8.1,inwefigurepresent8.3.ThetheReynoldstransformationnumberrulesReisthatleftareinvusedarianttobythecollapsethetransformations.pictures,

92

ConclusionsandOutlook8.8.3

Time→AdimensionalTime
igi→sσk)>→1θk,22eqρ
viv→sTable8.1:Transformationrulesinsedimentation.

)−noTheseanalyticalresults,ehowexpressionver,toonlyholdaccountforthesmallterminalReynoldssettlingvnumberselocityE,SatΨΦmoderate,sinceRethereynoldsis
numbers.Therefore,itisimpossibletocomputetheStokestime,eq.(3.30)atmoderate
Re.

ConclusionsandOutlook8.3

Inthischapter,thediffusionofoblateellipsoidsinsedimentationwasstudied.Wehave
foundanon-diffusivebehaviorinbothcomponents(parallelandperpendicular),which
iscompletelynewandnotreportedintheliterature.Ourcomparisonwiththeequivalent
spheresystemtoellipsoidsshowsthatthereisasimplediffusiveprocessfortheperpen-
dicularandanon-diffusiveprocessfortheparallelcomponents.Thisresultregardingthe
diffusivebehaviorinspheresagreeswiththeworkof(Rouyeret.al.1999)and(Miguel
2001).-SatorrasastorPandInaddition,thebehaviorforellipsoidsandsphereswasinvestigatedastheparticledensity,
kinematicviscosityandellipsoidaspect-ratioorsphereradiiwerechanged.Itwasfound
thattheincrementoftheinertialforces,bymeansofthegrowthinparticledensityorthe
decreasinginthekinematicalviscosity,reducesthepresenceofthediffusivebehavior,
andthesystemgoesfasterfromtheballistictothesaturationregime.Whenthesphere
radiiarechanged,wecouldobserveasimilarbehavior:Asthesphereradiusisincreased
theinertiagrowsandthusalsotheparticleM.S.D..
ItisimportanttopointoutthebehaviourofthetimederivativeoftheM.S.D.inboth
componentsastheaspect-ratiochanges(seefigure8.12).Fortheverticalcomponent,
withthegrowthoftheaspect-ratiotheM.S.D.goesfromthenon-diffusivebehaviorto
thesimplediffusiveregime.Intheparallelcomponent,theexponentαfornon-diffusive
regimeschangesfor(α=Ω.Ω,Δe=Φ(/Ψ.+)flattenedellipsoidsto(α=Ψ.(,Δe=Ψ)
spheres.Last,wefoundasimilaritylawforthesedimentationprocesswhichisvalidatsmall
Reynoldsnumbers.Itispresentedintable8.1.

9Chapter

FluctuationselocityV

Inthislastchapterweexaminethedynamicalbehaviorofsedimentingellipsoidsand
spheresunderchangesofthecontainersize.Inthefirstsectionwestudytheinfluence
onthespatialcorrelationsastheparticulatevolumefractionischanged,comparingthe
resultsforellipsoidsandspheres.Wealsopresentthestudyofthevelocityfluctuationsas
afunctionofthevolumefraction.Afterthat,weinvestigatethedivergenceofthevelocity
fluctuationsasthecontainersizeischanged.Finally,wesummarize.

Wechoosethedensityofthefluid,theStokesvelocityandthelargerradiusoftheellipsoid
equalsquaretobaseunityofinsideourA=system.ΩΩeInxtendedallofupthetoAcases,=theΨ0+,andcontaineralatticeheightisconstantA=ofVΨ)Φ=andΦ.0a.
Theratiobetweenthedensityoftheoblateellipsoidsandthefluidis(.

elationscorrSpatial9.1

Westartouranalysisbystudyingthespatialcorrelationsinthevelocityfluctuations(here-
afterSCVF).Thenormalizedautocorrelationfunctionoftheparallel(||)componentofthe
velocityfluctuationsaredefinedas,(Segre´et.al.(1997)):

,Σr)≡δiΣΦ)δiΣr)(9.1)
δiΣΦ))
1cmwherethebrackets...representanensembleaverageoverseveralindividualdif-
ferentconfigurationsinspaceandorientations(ellipsoids).WhereδvT=vT−vhEC,
representsthefluctuationsinthevelocityandvhEC=iTisthemeanvelocityoverthe
configuration.Ifthedistanceristakeninthedirectionparalleltogravity,k,thenwecall
theparallel,orperpendicular,l,,⊥component.
93

94

1

0.6||C

0.2

−0.201

0.6EllC0.2

||,Sph||,Ell1

(a)0.6⊥C

0.2

−0.26040200r,Ell⊥||,Ell1

(c)0.6SphC0.2

elationscorrSpatial9.9.1

,Sph⊥,Ell⊥

(b)

604020r⊥,Sph||,Sph

(d)

−0.20204060−0.20204060
rrFigure9.1:Thefiguresshowtheparallel(a)andperpendicular⊥(b)componentsof
theSCVFforellipsoidsandspheres.Figures(c)and(d)comparetheparallelandper-
pendicularcomponentsforellipsoidsandspheres,respectively.TheReynoldsnumberis
Ω×ΨΦ−),thespheresradiusEEfjT=Ψ.ΦΨandtheellipsoidaspect-ratioisΔe=Φ.(/Ψ.+.

9.1.1Changeinthevolumefraction
Infigure9.1,wepresentthespatialcorrelationsofvelocitiesforellipsoidsandspheres.
Figure9.1showsthecomparisonbetweentheparallel(a)andperpendicular(b)compon-
entsforspheresandellipsoids.Infigures(a)and(b)thecomponentsforspheresshowa
muchfasterrelaxationthanellipsoids.Figures(c)and(d)exhibittheanisotropy,charac-
teristictothesedimentationprocess,withaslowerdecayoftheparallelcomponentofthe
velocityautocorrelationfuntion.Thisanisotropyispresentinbothspheres,(Nicolaiand
Guazelli(1995),Segre´et.al.(1997),H¨ofler(2000))andellipsoids.

FluctuationselocityV

95

WepresenttheSCVFfortheparallel()andperpendicular(⊥)componentsfig.9.2for
ΦVspheres=Φ.(a,b)ΦΩ,Φ.andΦ(,Φ.Φ1,ellipsoidsΦ.Ψ+.(c,d)Theanisotropcorrespondingytobetweenfourtherepresentaticomponentsvesvisalsoolumepresentfractionsas
thevolumefractionΦVincreasesasisshowninfig.(a)and(b).

eThexponentialcomponentdecay,ofas,the||≈SVCFSkpΣfor−z/ξspheres,heS),(see(Nicolaifigureand9.2(a),Guazelliapproximately(1995),Segrfolloe´et.wsal.an
(1997),H¨ofler(2000)).Ontheotherhand,the⊥component(seefigure9.2(b))shows
avaluerapidofinitialthedecaycorrelationinarelength,gionξ⊥of,heSdifinferentthenegsystem.ativevSegralues.e´et.Thisal.(1997)minimummeasureddefinesthethe
0/(−the(dependence)oneξof,heSthe=ΨΨNcorrelationΦ−(/0,lengthwhereinNisthethe(⊥)spherecomponentradiusas(inξ⊥our,heS=notationΩ0NΦN=,EEfand)andfor
fraction.olumevtheΦ

InordertomakeoursimulationscomparablewiththeexperimentalresultsofSegre´et.al.
(1997)andalsowithprevioussimulations,H¨ofler(2000),weuseasystemsizeofsquared
basewithasideΨ((andheightΨ)ΦandavolumefractionofΦV=Φ.Φ´.Segre´et.
al.(1997)showthatthecorrelationlengthoftheperpendicularcomponent(⊥)becomes
ξ⊥,heS≈0´NforΦV=Φ.Φ).Inoursimulations,wefindacorrelationlengthforthe
equivalentspheresoftheorderofξ⊥,heS≈01N.

Inandthecaseperpendicularfor(d)ellipsoidsthecomponents,correlationweseealengthlarger(fig.value9.2),ofwiththeregardcorrelationtothelengthparallelinthe(c)
parallelcomponentξ⊥,heS≈10Nwhereastheperpendicularcomponentdoesnotshowan
ference.difappreciable

9.1.2Collapsingofthespatialcorrelations

Figure9.3(a)and(b)showthe−(/0collapsingoftheSCVFforspheres,inbothcomponents,
wscalingorksquitethewell,distanceandwithweEEfconsiderΦVitasnotewasworthproposedythatitbyisSevgralide´et.foral.changes(1997).ofThethevcollapseolume
fractionbyupto60times.Thecorrelationlengthsthatwefoundare:

ξ⊥,heS=Ω2EEfΦV−(/0;ξ,heS=Ψ´EEfΦV−(/0

whichdoesnotreallydifferfromtheresultsofSegre´et.al.(1997).

(9.2)

96

10.75

0.5Sph,||C0.25

0010.750.5Ell,||C0.25

0

40

(a)

0.0020.0040.80.0200.120

0.4⊥Sph,C0

80120160200−0.40
r)1.20.0020.0040.0200.1200.8(c)⊥Ell,C0.4

0

corrSpatial9.9.1elations

(b)

4016012080r(d)

0.0020.0040.0200.120

0.0020.0040.0200.120

0408012016020004080120160
rrFigure9.2:Spatialcorrelationsfunctionsintheparallel(a,c)andperpendicular(b,d)
componentsoftheSCVFforellipsoidsandspheres,withchangesofthevolumefraction.
TheReynoldsnumberisΩ×ΨΦ−),thespheresradiusEEfjT=Ψ.ΦΨandtheellipsoid
aspect-ratioΔe=Φ.(/Ψ.+.

Intheellipsoidcase(seefigure9.3),wepresenttheSCVFfortheparallel(c)andperpen-
dicular(d)component,respectively.Thecollapsealsoworkswellinthiscase.Thevalues
forthecorrelationlengtharethefollowing:

ξ⊥,Eaa=Ω)EEfΦV−(/0;ξ,Eaa=ΨΦEEfΦV−(/0(9.3)

Thecorrelationlengthforellipsoids,inbothcomponents,issmallerthanforspheres.

elocityVFluctuations

(a)

1.20.0020.0040.0200.1200.8⊥Sph,0.4C0

(b)

0.0020.0040.0200.120

1.20.0020.0020.0040.0040.0200.02010.1200.1200.8(b)(a)0.75⊥Sph,||Sph,0.40.5CC0.250002040−0.40204060
−1/3−1/3r/(ReqΦV)r/(ReqΦV)
1.20.0020.0020.0040.0040.0200.02010.1200.1200.8(d)(c)0.75⊥CEll,||0.5CEll,0.4
0.2500010203040−0.402040
r/(ReqΦV−1/3)r/(ReqΦV−1/3)
Figure9.3:Collapsingofthespatialcorrelationsfunctionfordatainfig.1.Ω.

−0.404020r/(ReqΦV−1/3)
1.20.0020.0040.0200.1200.8(c)⊥Ell,0.4C

0

80604020r/(ReqΦV−1/3)
0.0020.0040.0200.120(d)

97

Theamplitudeintheparallel()andperpendicular(⊥)componentsofthevelocityfluc-
tuations(seechapter+eqs.+.Ψand+.Ω),ispresentedinfigure9.4.Thefiguresarepresen-
ted(in/0alog-logscale.ForΦ.ΦΦ)≤ΦV≤Φ.Φ0,thevelocityfluctuationsgrowlinearlyas
≈ΦV(straightlinesuperimposedondata),asforspheresaswellasforellipsoids(Segre´
(1997)).al.et.Forlargervolumefractions(ΦV>Φ.Φ0),thefluctuationsarereducedinbothcomponents,
forspheresandforellipsoids.Thedecreaseinthevelocityfluctuationsasthevolume
fractioncontainerhasincreasesafixedvcouldolumebeeandxplainedthebyparticlethegroencounterswthofaretheaparticledissipativeencounters,processinsincenature.the
Thereforethevelocityfluctuationstendtoreduce(Kalthoffet.al.(1996)).
Itwasmentionedinchapter+above,thatthevelocityfluctuationscomponentsforspheres
aremuchlargerthanforellipsoids.Theappearanceofrotationsaroundtheellipsoid

98

010−110>2⊥Vδ>,<2||Vδ<−210−310

−1−21010ΦV010−110>2⊥Vδ2V>,<||10−2
δ<

−310

Sphδ. ⊥ Sphδ ||−110>2⊥Vδ>,<−2102||Vδ<−310

9.9.2Changeofthecontainersize
. δδ⊥ Ellip Ellip
||

−1−21010ΦV Sphδ. ⊥. δδ|| Ellip Sph
⊥ Ellipδ ||

10−2ΦV10−1
Figure9.4:Thetoppicturespresentthevelocityfluctuationsforspheresandellipsoidsin
theparallel()andperpendicular(⊥)componentsasafunctionofthevolumefraction,
inalog-logscale.Thebottompicturecomparestheresultsforbothkindsofparticlesand
components.TheReynoldsnumberisΩ×ΨΦ−),thespheresradiusEEfjT=Ψ.ΦΨandthe
ellipsoidaspect-ratioΔe=Φ.(/Ψ.+.

centerofmass,impliesaworkagainstthefluid,thenthedissipationofenergyispresent
nowfortranslationalandrotationaldegreesoffreedom.Therefore,thedecreaseofthe
velocityismoredramaticforellipsoidsthanspheres.

9.2Changeofthecontainersize
InfixedtheprecontainervioussidesectionofΨ0+weewherexaminedthethecharacteristicsystemasthecorrelationvolumelengthfractionwasfound,increaseseqs.with9.2a
9.3.andNowweturntotheanalysisoftheeffectsofvariationsinthecontainersizeonthevelocity

FluctuationselocityV

1(a)

0.6Ell,||C0.2

−0.220100r1(c)

0.6Sph,||C0.2

44881132176

0.6⊥Ell,C0.2

−0.24030044881132176

0.6⊥Sph,C0.2

(b)

10(d)

20r

4488132176

40304488132176

99

−0.20204060−0.2010203040
rrFigurecomponents9.5:ofSpatialthevelocitycorrelationfluctuationsfunctionsforfortheellipsoidsparallel(a,b)and(a,c)andspheres(c,d)perpendicularasa⊥function(b,d)
ofthecontainerside.TheReynoldsnumberisΩ×ΨΦ−),thespheresradiusEEfjT=Ψ.ΦΨ
andtheellipsoidaspect-ratioΔe=Φ.(/Ψ.+.

fluctuations,asexaminedintheworkby(Segre´et.al.(1997)andH¨ofler(2000)).Weuse
asmallcontainersize(closetothecharacteristiccorrelationlength),inordertoinvestigate
thesizeeffectsonthefluctuations.
Infigure9.5,weshowtheSCVFforspheres(a,b)andellipsoids(c,d)asthecontainer
sizeisincreased.Thecorrelationlengthξ⊥,heSforspheresandellipsoidsdecreasesin
bothcomponentsasthecontainerincreases.However,thedecayforspheresislargerthan
thedecayforellipsoidsintheparallelcomponent.Theperpendicularcomponent,onthe
otherhand,doesnotdisplayanydifference.
W9.6).ealsoFormeasurespherestheandpairellipsoidscorrelationwefindfunctionintheaspairthecontainercorrelationsideforisallchangedcontainer(seesizes,figurea
characteristiclengthoftheorder≈Ω,followedbyamonotonicdecaythatincreaseswith
thecontainerside(seefigure9.6(a)and(b)).Itisinterestingtopointoutthattheratioof

100

(a)

9.9.2Changeofthecontainersize

1761328844(b)

2017617613213268888164444(b)5(a)124EllSphg(r)g(r)38241002.557.51012.5002.557.51012.5
rr617613258844(c)4Sph3g(r)210012345
rFigure9.6:Paircorrelationfunctionforspheres(a),andellipsoids(b)andtheratioofthe
(c).ellipsoidstospheres

thecontainerpairsidecorrelationΨ0+.forΨ´Ω(seespheres9.6to(c)).thatofThisthepeakellipsoidsrevealsemuchxhibitslarhighergerdensitypeakforthefluctuationslarger
forspheresthanforellipsoids.Thebehaviorofthepaircorrelationfunctionsresemble
thepaircorrelationfunctionforliquids,wherethepositionsofneighboringmoleculesare
stronglycorrelated,leadingtoamodulationofthepaircorrelationfunctions(Barratand
Hansen(2003)),whichisverysimilartothatpresentedinfigure9.6(a)and(b).
Infromorderthetoinvcorrelationestigatelength,sizeefasinfectstheinwtheorkvofSeelocitygre´et.al.fluctuations,(1997).weTheundertakresultsearevariationspresent
infigure9.7forspheres(dashdottedline)andellipsoids(dashcontinuousline).The
containersidearenormalizedbyA/ΣEEfΦ−(/0).ThevelocityfluctuationsσδK⊥,/ih>
(Sepresentgre´et.anal.initial(1997)transitionandHre¨oflergion,(2000)),whichhavebetweenastrongΩΦ≤A/dependenceΣEEfΦ−(on/0)the≤ΨΦΦcontainer,aftersideit,
thespheressimulationshowadatasimilararebehaviorindependent,butofwiththeasmallercontainerovside.erallvInaluegeneral,fortheellipsoids.ellipsoidsTheratioand
oftheparallelvelocityfluctuationtotheperpendicular⊥component,forspheresand

elocityVFluctuations

1.2

0.8>s/v||Vδ<0.4

0.4

>s/v⊥Vδ<0.2

101

EllipEllipSphSph02060L/(RΦ100−1/314018004080L/(RΦ−1/3)120160
eqeqFigure9.7:Theleftgraphicshowtheparallelcomponentofvelocityfluctuationsasthe
numbercontainerisΩside×isΨΦ−),changed.theThespheresrightradiuspictureEEfshojTws=Ψthe.ΦΨvanderticalthecomponent.ellipsoidTheaspect-ratioReynoldsis
Δe=Φ.(/Ψ.+.

ellipsoidsis≈Ω.),inagreementwithSegre´et.al.(1997).
Thisequalvalueforbothkindofparticles,revealsthattheanisotropicbehavioronthe
velocityfluctuationcomponentsareindependentontheparticleshape.Thesymmetry
breakinginducedbygravityactsequallyonspheresandellipsoids.
Asmightisdiverdiscussedgewithinchapterincreasing1andbycontainerCaflishsize.andOnLuktheeother(1985),hand,theevelocityxperiments,fluctuationssimula-
Htions¨oflerand(2000),theoryitsec.isarΨ.+guedhavethatfoundsystems,noevidenceboundedforbysuchwdialls,verdogence.notshoInawapreviouscomparablework,
scalingofvelocityfluctuationsbutasaturation,ifthesmallestextensionofthecontainer
exceedsacriticalsize,andthenthedifficultytofindauniquescalinglaw.Inoursimula-
tions,thevariationsofthecontainersize,weremadebychangingtheentiresquaredbase.
Thefluctuationsresults,fig.neither9.7for(a)andspheres(b),nordon’tellipsoids,presenteandvidencebehaveforreallythedivcloseergencetotheoftheresultsvgielocityven
bySegre´et.al.(1997).

9.3ConclusionsandOutlook

WsmalleRestudiedynoldsthevnumberariationΩof×ΨΦthe−).Itcontainerwasfoundsizeforthatthespheresspatialandcorrelationsellipsoidsofthesedimentingverticalat
velocityforellipsoidsshowamuchslowerdecayastheperpendicularcomponent.The
equivalentspheresystemreproducesthesameanisotropy,thismatchingrathercloselyto
theresultsofSegre´et.al.(1997).
ThecollapsingemployedinthespatialcorrelationsforΦV≤Φ.Ψworksquitewellfor

102

ConclusionsandOutlook9.9.3

bothspheresandellipsoids,fig.9.3and9.1.Thecorrelationlengthforellipsoidshavea

smallervaluethanforspheres,eq.9.2and9.3.

Thevelocityfluctuations,forbothkindsofparticles,alsorevealtheanisotropypresent

inbyusedtransformationcollapseThecorrelations.spatialthe

inthespatialcorrelations.Thecollapsetransformationusedby
ΦV≤fluctuationsΦ.Ψ,agreesdecrease.quiteItiswellimportantwithourtodata,remark,fig.that9.4.thevAfterelocityΦV

are

much

gerlar

as

for

Itisimportant

ellipsoids.

to

remark,

that

the

velocity

Se>grΦ.e´Ψ,et.allal.thev(1997),elocityat
spheresforfluctuations

fluctuations

for

spheres

10Chapter

Conclusion

Theaimofthisthesisisthestudyofsedimentationofoblateellipsoidsusinganumerical
oblatesimulationellipsoidtechniqueatlowH¨andoflerhighandReSchwynoldsarzernumber(2000),andwhichmanyisoblateapplyedtoellipsoidstheatcaselowofandone
moderateReynoldsnumber,inthreedimensions.

OblateOne10.1ellipsoid

Themotionofapieceofpaperoraleaf,asitfallstotheground,isanoldandunsolved
probleminPhysics.Maxwell,HelmholtzandKelvinarejustsomewhohavestudiedthis
problem.Recentexperiments(Fieldet.al.(1997)andBelmonteet.al.(1998))and
simplifiedmodelsMahadevan(1996),confirmthatthemotionoffallingobjectsisstillfar
understood.beingfrom

Inourworkthefallingobjectsareconsideredaveryflatoblateellipsoid,suchasleavesor
asheetofpaper,settlinginafluid,inathreedimensionalcontainer.Wefoundthreebasic
regimesforthedynamicsofthesystem(steady-falling,oscillatory-periodic,andchaotic).

Thesteady-fallingexhibitsasimilarphysicalbehaviourasobservedexperimentallyby
Fieldet.al.(1997)andBelmonteet.al.(1998).Wehavecharacterizedthedynamics
ofthesteady-fallingregimewhenthekinematicviscosity,droppingheight,andoblate’s
aspect-ratioarechanged.Someconclusionscanbedrawnfromthispartofthework.This
regimeispresentforsmallvaluesofI≈Φ.)−Ψ,ES≈ΨΦΦandisshowninfig.5.1-5.4.
TheperiodicbehaviourinoursimulationsisfoundforES∼(ΦΦandsmallaspect-ratios
ΣΔe≤Φ.Ψ).Theverticalorientation,Θoscillateswithdoubletheperiodofoscillationof
theverticalvelocityilandatthesameperiodofthehorizontalvelocityik.Thisperiodic
motionhasalsobeenobservedexperimentallybyBelmonteet.al.(1998),showingthat
correct.essentiallyaresimulationsour

Wequalitatifindvethatouragreementresultswithinthethecaseofsimplifiedthesteady-fmodelallingproposedandbyoscillatory(Mahadevanphases,(1996)).areingood

103

104

sedimentationellipsoidsMany10.10.2

Thechaoticbehaviourispresentforlargeraspect-ratios(Δe≥Φ.´)andintheentirerange
ofReynoldsnumbersusedinthiswork.Theseparationbetweenthespatialtrajectories
ofthefallingoblateellipsoiddivergesforsmallvariationsintheinitialorientationΘd,
andgrowsexponentiallyintime.ThevaluefoundfortheLyapunovexponentisλ=
Φ.Φ)Ω±Φ.ΦΦ).Itisworthwhiletopointoutthatwegiveaquantitativemeasureofthe
sensitivitytosmallchangesintheinitialstateofthesystem.
Forthesteady-fallingandoscillatoryregimeweobtainasimilaritylaw,whichisadirect
consequenceoftheinvarianceoftheReynoldsandFroudenumbers.Also,thesimilarity
expressestheindependenceofthephysicalresultsofthegridsize,whichisagoodtest
forthedynamicsofthesetworegimes.
Weconstructaphasediagramthatshowsthreewell-defineddynamicregionsasisalso
shownbyFieldet.al.(1997).Thedifferencewiththeabovereferenceisthatthechaotic
behaviourisassociatedwiththetransitiontochaosthroughintermittencywhichisnot
seeninoursimulations.Thephasediagramisindependentoftheinitialorientation.
Thetransitionforsteady-fallingtooscillatoryandthetransitionfromsteady-fallingto
chaoticregimecanbeunderstoodassecondandfirstorderphase,respectivelyandthe
characteristictransienttimeanditsinversebeingtheorderparameter,respectively.

sedimentationellipsoidsMany10.2

Wehavesimulatedthesedimentationofoblateellipsoidsatsmallvolumefraction(ΦV≤
Φ.Ω)andsmallReynoldsnumber(ES≈ΨΦ−)).Wehavefoundthatatintermediatevolume
fractionthesettlingvelocityexhibitsalocalmaximumwhichtoourknowledgehasnever
literature.theinreportedbeenWealsopresentdataatmoderateReynoldsnumber(ES≈0)forthesedimentationve-
locityofoblateellipsoidswichfollowsamonotonicbehaviourasthevolumefraction
ΦVisincreased.AsinthecaseoflowReynoldsnumbertheellipsoidshaveasmaller
sedimentationvelocitythantheequivalentspheres.Thedataforellipsoidsandspheres
followtheRichardson-Zakilaw(1954)withexponents(cEaaTe≈´.Ω,ES=ΨΦ−))and
(c<eS≈(.Φ,ES=0)respectively.
Inaddition,thelocalmaximum,atlowReynoldsnumber,inthevelocitycanberelated
tothenon-monotonicbehaviouroftheverticalorientationoftheellipsoidsalonggravity,
andcanbeexplainedbythe“cluster”formation.Thisnon-monotonicbehaviourisalso
foundinfiber-likesuspensionsHerzhaftandGuazzelli(1999).AtmoderateESthereis
alsoalargeralignmentofellipsoidswithgravity,comparedtothosewithsmallReyn-
oldsnumber.ThealignmentwithgravityispresentalsoatmoderateReynoldsnumber,
whichisinagreementwiththeorientationalbehaviourofasingleellipsoidFonsecaand
2004).((1)HerrmannAtlowReynoldsnumbertheorientationalorderparameterΨvanishesaroundΦV≈Φ.Φ1
fig.7.7.AsΦVdecreasestheorientationalalignmentwithgravityincreases.Forlow
ReynoldsnumberinthelimitΦV→Φasingleellipsoidalignswithgravity.The

Conclusion

105

alignmentwithgravityisadistinctivefeatureofthesteady-stateregimeforasingleob-
lateellipsoidasreportedinreferencesFonsecaandHerrmann((1)2004)andGaldiet.
al.(2001).Thevanishingoftheorderparameter,atmoderateES,isslightlyshifted
(ΦV≈Φ.Ψ)totherightastheReynoldsnumberincreases(seefig7.12,bottom).All
thesimulationsinthisworkarelocatedinthesteady-fallingphaseforasingleoblate
ellipsoid.

difThefusidifvefusionbehaofviouroblateinbothellipsoidscomponentsin(parallelsedimentationandwasperpendicular),studied.Wehawhichveisfoundacompletelynon-
newandnotreportedintheliterature.Ourcomparisonwiththeequivalentspheresystem
tonon-difellipsoidsfusiveshoprocesswsthatfortheretheisaparallelsimpledifcomponents.fusiveThisprocessresultfortheregardingperpendicularthediffusiandvae
behaviourinspheresagreeswiththeworkof(Rouyeret.al.1999)and(Migueland
2001).-SatorrasastorP

Furthermore,thebehaviourforellipsoidsandsphereswasinvestigatedastheparticle
density,kinematicsviscosityandellipsoidaspect-ratioorsphereradiiwerechanged.It
wasfoundthattheincrementoftheinertialforces,bymeansofthegrowthinparticle
densityorthedecreasinginthekinematicsviscosity,reducesthepresenceofthediffusive
behaviour.Whenthesphereradiiarechanged,wecouldobserveasimilarbehaviour:as
thesphereradiusisincreasedtheinertiagrowsandthusalsotheparticleM.S.D..

TheverticalcomponentoftheM.S.D.passesfromthenon-diffusivebehaviourtothe
esimplexponentdifαfusiforverenon-difgimefusiasvetheregimesaspect-ratiochangesforincreases.(α=FΩ.orΩ)theflattenedparallelellipsoidscomponent,to(αthe=
spheres.)(.Ψ

Last,wefoundasimilaritylawthatcollapsequitewellthediffusionprocessanditisvalid
numbers.ynoldsResmallat

Westudiedthevariationofthecontainersizeforspheresandellipsoidssedimentingat
smallReynoldsnumberΩ×ΨΦ−).Itwasfoundthatthespatialcorrelationsoftheparallel
velocityforellipsoidsshowasmallerdecayastheperpendicularcomponent.Theequi-
valentspheressystemreproducesthesameanisotropy,thismatchingrathercloselytothe
resultsofSegre´et.al.(1997).

ThecollapsingemployedinthespatialcorrelationsforΦV≤Φ.Ψworksquitewellfor
spheresandellipsoids.Thecorrelationlengthforellipsoidshasasmallervaluethanfor
spheres.

Thevelocityfluctuationsforellipsoidsrevealtheanisotropypresentinthespatialcorrela-
welltions.withTheourcollapsedata.AfterΦtransformationV>Φ.Ψ,usedallthebyvSegrelocitye´et.al.fluctuations(1997),atΦVdecreases.≤Φ.ΨIt,isagreesimportantquite
toremark,thatthevelocityfluctuationsforspheresaremuchlargerasforellipsoids.

106

10.3Outlook

Outlook10.10.3

manDespiteyopenthefactsproblems.thatfallingUsingthisbodiesmodelhaveitbeenisnowstudiedpossiblefortosuchastudylongthetime,theresedimentationarestillof
differenttypesofparticlesandthestructureofthefluidsurroundingthem.Futurework
couldbeaddressedinseveraldirections:

Thexperimentseillmarth(Wet.al.(1964),Belmonteet.al.(1998)andellyKandWu(1997))haveconfirmed,thatthepresenceofvorticesisfundamental,asthe
objectsfallorriseinafluid.Vortexgenerationissuchanimportantpartoffluid
dynamicsthatacompletetheorymustbetakenintoaccountinordertounderstand
theroleofthefluidinthemotionoffallingobjects.Then,anaturalfollowingwork
istoundertakeasystematicresearchaboutthevelocityandpressurefieldsinthe
dynamicsoffallingobjects.Thisisataskthatouralgorithmisabletogive.

•Theautorotationortheangularmotionthattheflatobjectexecutesasitfall,is
animportantresearchinalargenumberoflaboratorystudies.Theeffortscome

animportantresearchinalargenumberoflaboratorystudies.Theeffortscome
mainlybypracticalconsiderationsinmeteorologyastheformationofhailstones;
thedynamicsofaircraftafteritstall,etc.Theautorotationhasbeenfoundinsome
experiments(Fieldet.al.(1997)orMahadevanet.al.(1999))forveryflatobjects.
Ofparticularnoteistheflatobjectcanoscillateseveraltimesasitfell,increasingits
amplitudeineachoscillationuntilitcompletelyturnedover.Thesimulationofthis
typeofmotionrequiresthatthethicknessobject(smallerellipsoiddiameter)hasto
be,atleastoftheorderofthegridsize.Therefore,thelargerobjectdiameter(small
aspect-ratio)andthecontainersizewillimplyahugenumberofgridpointsinthe
simulationfortheinvestigationofthismotion.Inordertocarryoutthesimulation
ofautorotation,theaboveconditionsmustdemandahighcomputationaleffort.
Thesimulationtechniqueusedinthisthesisinconnectionwiththeparallelized
algorithmversioncanovercomethesedifficulties.

•Inthedynamicsofonefallingflatobjectwefoundasimilaritylawforthesteady-
fallingandoscillatoryphases(Sections).Ψand).Ω).Theexperimentaldatasup-
portingtheserelationswouldbeveryimportant.

•Wefoundanon-monotonicbehaviourfortheoblateellipsoidsettlingvelocityas
thevolumefractionincreases.Alsotheincreasingellipsoidalignswithgravityas

thevolumefractionincreases.Alsotheincreasingellipsoidalignswithgravityas
thevolumefractiondecreases.Finallyinthedynamicsofmanyellipsoidssedi-
mentation,wefoundananomalousdiffusionbehaviourfortheparallelcomponent
togravity.Uptonow,wedon’tknowanexperimental,simulationortheoretical
resultrelatedwiththissettlinglaworanomalousdiffusion,andanexperimental
workratifyingthesebehavioursisnecessary.

•Suspensionsinnatureandindustrygenerallyinvolvethemixturesofparticlesof
differenttypes,shapesandsizes(e.g.,oblateellipsoids,prolateellipsoidsand

differenttypes,shapesandsizes(e.g.,oblateellipsoids,prolateellipsoidsand
spheres).Thebidispersesuspensionsarenormallymadeoftwodifferentsize
sphericalparticles,butnotalikeshaped.Inthestatisticalphysicsofbidisperse

Conclusion

107

hard-spheresmixturesisknown(AsakuraandOosawa(1954))the“depletioninter-

action”,whichitistheeffectofthelargeparticlestogetherincreasetheavailable

volume,thereforetheentropy,

thetodueinteraction”pletion

important

task

for

accomplish.

Then,particles.smallthefor

shape

and

not

only

to

olumev

studythe“de-thisof

ferencedif

is

another

108

10.10.3

Outlook

encesRefer

AckersonB.J.,andClarkN.A.(1984).Shear-inducedpartialtranslationalorderingof
acolloidalsolid.Phys.Rev.A.,30,906.
AdamsM.andFradenS.(1998).Phasebehaviorofmixturesofrods(tobaccomosaic
virus)andspheres(polyethyleneoxide,bovineserumalbumin).Biophys.J.,74,
669.AdrianR.J.(1991).Particle-imagingtechniquesforexperimentalfluidmechanics.
Annu.Rev.fluidMech.23,261.
AllenM.P.andTilsdesleyD.J.(1987).Computersimulationsofliquids.Clarendon
Press,Oxford,1987.
ArefH.andJonesS.W.(1993).Chaoticmotionofasolidthroughidealfluid.Phys.
3026.,5A.FluidsAsakuraS.andOosawaF.(1954).Oninteractionbetweentwobodiesimmersedina
solutionofmacromolecules.J.Chem.Phys.22,1255.
BallentsL.,MarchettiM.C.andRadzihovskyL.(1998).Nonequilibriumsteadystates
ofdrivenperiodicmedia.Phys.Rev.B.57,7705.
Barrat,J.L.andHansen,J.P.(2003).Basicconceptsforsimpleandcomplexfluids.
CambridgeUniversityPress,Cambridge,2003.
BasuA.,BhattacharjeeJ.K.andRamaswamy,S.(1999).Meanmagneticfieldand
noisecross-correlationinmagnetohydrodynamicturbulence:resultsfromaone-
dimensionalmodel.Eur.Phys.J.9,735.
BatchelorG.K(1970).Thestresssysteminasuspensionofforce-freeparticles.J.
FluidMech.41,545.
BatchelorG.K.(1970).Slender-bodytheoryforparticlesofarbitrarycross-sectionin
Stokesflow.J.FluidMech.44,419.
BatchelorG.K.(1972).SedimentationinaDiluteDispersionofSpheres.J.Fluid
Mech.52,part2,245.
BatchelorG.K.(1982).Sedimentationinadilutepolydispersesystemeofinteracting
spheres.Part1.Generaltheory.J.fluidMech.119,379.
BatchelorG.K.andJanseVanRensburgW.(1986).Structureformationinbidisperse
sedimentation.J.fluidMech.166,379.
BauerM.,HabipS.,HeD.R.andMartiensenW.(1992).Newtypeofintermittencyin
discontinousmaps.Phys.Rev.Lett..68,1625.

109

110

BeenakkerC.W.J.,VanSaarloosWandMazur,P.(1984).Many-SphereHydrodynamic
InteractionsIII.TheInfluenceofaPlaneWall.physicaA..127,451.
BelmonteA.,EisenbergH.andMosesE.(1998).FromFluttertoTumble:Inertial
DragandFroudeSimilarityinFallingPaper.Nature.81,345.
BerneB.andPecoraR.(1990).DynamicLightScatteringwhitapplicationstoChem-
istry,BiologyandPhysics.Malabar,Fl:Krieger.
BibenT.andHansenJ.p.(1991).Phaseseparationofasymmetricbinaryhard-sphere
fluids.Phys.Rev.Lett..66,2251.
BibetteJ.(1991).Depletioninteractionsandfractionatedcrystallizationforpolydis-
perseemulsionpurification.J.ColloidinterfaceSci..147,474.
BlakeJ.(1971).Anoteontheimagesystemforastokesletinano-slipboundary.Proc.
Camb.Phil.Soc..71,303.
BlancR.andGuyonE.(1991).Thephysicsofsedimentation.LaRecherche.22,866.
BradyJ.F.andBossisG.(1985).Therheologyofconcentratedsuspensionsofspheres
insimpleshearflowbynumericalsimulation.J.FluidMech.155,105.
BrennerH.(1961).Theslowmotionofaspherethroughaviscousfluidtowardsaplane
surface.Chem.Eng.Sci.16,242.
BrennerM.P.(1999).Screeningmechanismsinsedimentation.J.PhysFluids.11,4,
754.CowanM.L.,PageJ.H.andWeitzD.A(2000).Velocityfluctuationsinfluidizedsus-
pensionsprobedbyultrasoniccorrelationspectroscopy.Phys.Rev.Lett.85(2),453.
ChaikinP.M.andLubenskyT.C.(1998).PrinciplesofCondensedMatterPhysics.New
CambridgeuniversityPress.
ChandrasekharS.(1981).HydrodynamicandHydromagneticStability.NewYork:
.verDoChuX.L.,Nikolov.A.D.andWasanD.T(1996).Effectsofinterparticleinteractions
onstabilityaggregationandsedimentationincolloidalsuspensions.Chem.Eng.
123.,150.Comm.CaflishR.E.andLukeJ.H.C.(1985).Varianceinthesedimentationspeedofasus-
pension.Phys.Fluids.28(3),759.
DasD.andBarmaM.(2000).ParticlesSlidingonaFluctuatingSurface:PhaseSepar-
ationandPowerLaws.Phys.Rev.Lett.85,1602.
DavisR.H.andHassenM.A.(1988).SpreadingoftheInterfaceattheTopofaSlightly
PolydisperseSedimentingSuspension.J.FluidMech..196,107.
DeGrootR.H.andMazurP.(1984).Non-equilibriumthermodynamics.NewYork:
.verDoDebyeP.andH¨ukelE.(1923).Thetheoryofelectrolytes.I.Loweringoffreezingpoint
andrelatedphenomena.Z.Phys.24,185.
EinsteinA.(1906).InvestigatonsonthetheoryoftheBrownianmovement.Ann.
Phys..19,289.

yBibliograph

111

ErtasD.andKardarM.(1992).Dynamicrougheningofdirectedlines.Phys.Rev.
929.,69Lett.EvansM.R.,KafriY.,KoduvelyH.MandMukamelD.(1998).PhaseSeparationin
One-DimensionalDrivenDiffusiveSystems.Phys.Rev.Lett.80,425.
FessasY.P.andWeilandR.E...(1981).Convectivesolidssettlinginducedbyabuoyant
phaseAIChEJ..27,588.
FieldS.,KlausM.,MooreM.andNoriF.(1997).Chaoticdynamicsoffallingdisks.
Nature.388,252.
FogelsonA.L.andPeskinC.S.(1988).Afastnumericalmethodforsolvingthethree-
dimensionalStokesequationsinthepresenceofsuspendedparticles.Journalof
50.,79PhysicsComputationalFonsecaF.andHerrmannH.(1)(2004).SimulationoftheSedimentationofaFalling
OblateEllipsoid.acceptedforpublicationinPhys.A.
FonsecaF.andHerrmannH.(2)(2004).Sedimentationofoblateellipsoidsatlowand
moderateReynoldsnumber.acceptedforpublicationinPhys.A..
ForsterD.(1980).HydrodynamicFluctuations,BrokenSymmetryandCorrelation
Functions.Bejamin/Cummings,Reading,MA,1975.
ForsterD.,NelsonD.R.andStephenM.J.(1977).Large-distanceandlong-timeprop-
ertiesofarandomlystirredfluid.Phys.Rev.A..16,732.
GaldiG.P.,PokornyM.,VaidyaA.,JosephD.D.andFengJ.(2001).Orientationof
symmetricbodiesfallinginasecond-orderliquidatnonzeroReynoldsnumber.J.
Math.FluidMech..3,183.
GoldsteinH.,PooleC.andSafkoD.J.(2002).ClassicalMechanics.AddysonWesley,
SanFrancisco,2003.
GrinsteinG.,LeeD.H.andSachdevS.(1990).Conservationlaws,anisotropy,and
“self-organizedcriticality”innoisynonequilibriumsystems.Phys.Rev.Lett.,64,
1990.HalperinB.I.andHohenbergP.C.(1977).Theoryofdynamiccriticalphenomena.Rev.
435.,49PhysMod.HamJ.M.andHomsyG.M.(1988)..Int.J.multiphaseFlow.14,533.
HappelJ.andBrennerH.(1965).LowReynoldsNumberHydrodynamics.Englewood
Cliffs,NJ:prenticeHall,1965.
Herzhaft,B.andGuazzelli,E.(1999).J.FluidMech..384,133.
HoyosM.,BacriJ.P.,MartinJ.andSalin,D.(1994).Structure,density,andvelocity
fluctuationsinquasi-two-dimensionalnon-Browniansuspensionsofspheres.Phys.
3809.,6.Fluids.H¨oflerK.andSchwarzerS.(2000).Navier-Stokessimulationwithconstraintforces:
Finite-differencemethodforparticle-laddenflowsandcomplexgeometries.Phys.
Rev.E.43,761.
H¨oeflerK.(2000).SimulationandModelingofMono-andBidisperseSuspensions.
DoctoralThesis.StuttgartUniversity.

112

HuangP.Y.,HuH.H.andJosephD.D.(1998).Directsimulationofthesedimentation
ofellipticparticlesinOldroyd-Bfluids.J.FluidMech..362,297.
J´anosiI.M.,Tel.T.,WolfD.EandGallasJ.A.C.(1997).Chaoticparticledynamicsin
viscousflows:Thethree-particleStokesletproblem.Phys.Rev.E.56,2858.
JayaprakashC.,HayotF.andPanditR.(1993).Universalpropertiesofthetwo-
dimensionalKuramoto-Sivashinskyequation.Phys.Rev.Lett..71,12.
KajikawaM(1982).Observationofthefallingmotionofearlysnowflakes.J.Meteorol.
Soc.Jpn.(Suppl.)1,797.
Kalthoff,W.Schwarzer,S.Ristow,G.andHerrmann,H.(1996).Ontheapplicationof
anovelalgorithmtohydrodynamicdiffusionandvelocityfluctuationsinsediment-
ingsystems.Int.J.Mod.Phys.C7No.4,543.
KelleyErinandWuM.(1997).PathInstabilitiesofRisingAirBubblesinaHele-Shaw
Cell.Phys.Rev.Lett.79,1265.
KochD.L.andShaqfeh,E.S.G(1991).Screeninginsedimentingsuspensions.J.Fluid
275.,224h.MecKochD.L.(1994).Hydrodynamicdiffusioninasuspensionofsedimentingpoint
particleswithperiodicboundaryconditions.Phys.Fluids.6,2894.
KraichnanR.(1962).Eddyviscosityanddiffusivity:exactformulasandapproxima-
tions.Phys.Fluids.5,1374.
KuuselaE.,H¨oeflerK.andSchwarzerS.(2001).Computationofparticlesettlingspeed
andorientationdistributioninsuspensionsofprolatespheroids.J.ofEngineering
221.,41MathematicsKuuselaE.,Lahtinen,J.M.andAla-Nissila,T.(2003).Collectiveeffectsinsettlingof
spheroidsundersteady-statesedimentation.Phys.Rev.Letters.90,094502.
KuuselaE.andAla-NissilaT.(2001).Velocitycorrelationsanddiffusionduringsedi-
mentation.Phys.Rev.E.63,061505.
KynchG.J.(1952).Atheoryofsedimentation.faradaySoc.48,166.
LaddA.J.C.(1993).Dynamicalsimulationsofsedimentingspheres.Phys.Fluids
299.,5(2)A.LaddA.J.C.(1994).Numericalsimulationsofparticulatesuspensionsviaadiscretized
Boltzmanequation.Part1.Theoreticalfundation.J.FluidsMech.271,285.
LaddA.J.C.(1996).Hydrodynamicscreeninginsedimentingsuspensionsofnon-
Brownianspheres.Phys.Rev.Lett.76,1392.
LaddA.J.C.(2002).Effectsofcontainerwallsonthevelocityfluctuationsofsediment-
ingspheresPhys.Rev.Lett.88.048301-1.
LambH.(1932).Hydrodynamics.2nd.Cambridge:CambridgeUniv.Press.
LandauL.D.andLifshitzE.M.(1969).StatisticalPhysics.Reading,MA;Addison-
.yesleWLahiriR.andRamaswamyS.(1997).AreSteadilyMovingCrystalsUnstable?.Phys.
Rev.Lett.79,1150.

yBibliograph

113

LahiriR.,BarmaM.andRamaswamy,S(2000).Strongphaseseparationinamodelof
sedimentinglattices.Phys.Rev.E.61,1648.
LandauL.DandLifshitzI.M.(1959).FluidMechanics.oxford:Pergamond.
LandauL.DandLifshitzI.M.(1965).TheoryofElasticity.Oxford:Pergamond.
Langevin,P.(1908).SurlaTheorieduMouvementBrownienComptesrendus.146,
530.LeiX.,AckersonB.JandTongP.(2001).SettlingStatisticsofHardSphereParticles.
Phys.Rev.Lett.86,3300.
LeightonD.andAcrivosA.(1987).Theshear-inducedmigrationofparticlesincon-
centratedsuspensions.J.FluidMech.181,415.
LevineA.,RamaswamyS.,FreyE.andBruismaR.(1998).Screenedandunscreened
phasesinsedimentingsuspensions.Phys.Rev.Lett.81,5944.
LugtH.J.(1983).Autorotation.Annu.Rev.FluidMech.15,123.
MaS.K.(1976).ModernTheoryofCriticalPhenomena.ReadingMA:Benjamin.
MahadevanL.,ArefH.andJonesS.W.(1995).Commenton’BehaviorofaFalling
Paper’.Phys.Rev.Lett..75,1420.
Mahadevan,L.(1996).Tumblingofafallingcard.C.R.Acad.Sci.323,729.
MahadevanL.,RyuA.andSamuel,D.T.(1999).Tumblingcards.Phys.Fluids11,1.
MarchildonE.K.,ClamenandA.andGauwinW.H.(1964).Dragandoscillatorymo-
tionoffreelyfallingcylindricalparticles.Can.J.Chem.Eng.64,178.
MartinP.C.,ParodiOandPershanP.S.(1972).UnifiedHydrodynamicTheoryforCrys-
tals,LiquidCrystals,andNormalFluids.Phys.Rev.A.6,2401.
Maxwell,J.C.(1853).TheScientificPapersofJamesClerkMaxwell.Cambridgeand
DublinMath.J.9,115.
MisesVonR.(1945).TheoryofFlight.McGraw-Hill,NewYork,1945.
MiguelM.-C.andPastor-SatorrasR.(2001).Velocityfluctuationsandhydrodynamic
diffusioninsedimentation.Europhys.Lett.54(1),45.
NicolaiH.,HerzhaftB.,HinchE.J.,OgersL.andGuazzelliE.(1995).Particleve-
locityfluctuationsandhydrodynamicself-diffusionofsedimentingnon-Brownian
spheres.Phys.Fluids.7(1),12.
NicolaiH.andGuazzelliE.(1995).Effectofthevesselsizeonthehydrodynamic
diffusionofsedimentingspheres.Phys.Fluids.7(1),12.
OllaP.(1999).Asimplifiedmodelforredcelldynamicsinsmallbloodvessels.Phys.
Rev.Lett..82,453.
PaulinS.EandAckersonB.J.(1990).Observationofaphasetransitioninthesedi-
mentationvelocityofhardspheres.Phys.Rev.Lett.64,2663.
PanT.-W.,JosephD.D.,R.Bai,R.GlowinskiandV.Sarin(2002).Fluidizationof1024
spheres:simulationsandexperiments.J.FluidsMech.451,169.
PedleyT.J.andKessleyJ.O.(1992).Annu.Rev.fluidMech.24,313.

114

PerramJ.W.andRasmussenJ.(1996).Ellipsoidcontactpotential:Theoryandrelation
tooverlappotentials.Phys.Rev.E.54,6565.
PerrinJ.(1916).Atoms.OxbowPress,Amherst,MA,USA,1990.
PeyssonY.andGuazzelliE.(1999).Velocityfluctuationsinabidispersesedimenting
suspension.Phys.Fluids11,1953.
PressW.h.,FlanneryB.P.,TeukolskyS.A.andVetterlingW.T.(1992).Numerical
recipesinC.Cambridge:CambridgeUniversityPress.
PriestleyC.H.B.(1959).TurbulentTransferintheLowerAtmosphere.Chicago:Uni-
versityofChicagoPress.
PuseyP.andRamaswamyS.(1997).TheoreticalChallengesintheDynamicsofCom-
plexFluids,LecturesinaNATOAdvancedStudyInstituteSeries.1990.
QiDewei.,LuoLishi.,AravamuthanRajaandStriederWilliam(2002).LateralMigra-
tionandOrientationofEllipticalParticlesinPoiseuilleFlows.J.Statist.Phys.107,
101.RamakrishnanT.V.andYussouffM.(1979).First-principlesorder-parametertheoryof
freezing.Phys.Rev.B.19,2775.
RamaswamyS.(2001).Issuesinthestatisticalmechanicsofsteadysedimentation.
Adv.Phys.50,297.
RichardsonJ.F.andZakiW.N.(1954).SedimentationandFluidisation:PartI.Trans.
Instn.Chem.Engrs.32,35.
RouyerF.,MartinJ.andSalinD.(1999).Non-gaussiandynamicsinquasi-2dnoncol-
loidalsuspensions.Phys.Rev.lett.83,1999.
RusselW.B.,SavilleD.A.andSchowalterW.R(1989).ColloidalDispersions.Cam-
bridge:CambridgeUniversityPress
SaffmanP.G.(1975).Ontheformationofvortexrings.Stud.appl.Maths.52,115.
SchimittmannBandZiaR.K.P.(1995).PhaseTransitionsandcriticalPhenomena.
.vol.17Segre´P.E.,LiuF.,UmbanhowarP.andWeitzD.A.Aneffectivegravitationaltemper-
atureforsedimentation.Nature409,594.
Segre´P.N.,HerbolzheimerE.andChaikinP.M.(1997).Long-rangecorrelationsinsed-
imentation.Phys.Rev.Lett.79,2574.
SimhaR.A.andRamaswamyS.(1999).TravelingWavesinaDriftingFluxLattice.
Phys.Rev.Lett.83,3285.
TanabeY.andKanekoK.(1994).Behaviourofafallingpaper;reply.Phys.Rev.
1372.,7573Lett.TongP.andAckerson,B.J.(1998).Analogiesbetweencolloidalsedimentationandtur-
bulentconvectionathighPrandtlnumbers.Phys.Rev.E.58,6931.
ToryE.M.,KammelM.TandChanManFongC.F.(1992).PowderTechnol.73,219.
VietsH.andLeeD.A.(1971).MotionoffreelyfallingspheresatmoderateReynolds
numbers.A.I.A.A.Journal.232,2038.

yBibliograph

115

vonSmoluchowskIM.(1916).Threediscoursesondiffusion,Brownianmovements,
andcoagulationofcolloidparticles.PhysZeits.17,557.

WputingachmannofB.andSuspensions.SchwarzerInt.S.J.Mod.(1998).Phys.C9(5)Three-Dimensional,759.MassivelyParallelCom-

WachmannB.,SchwarzerS.andK.H¨oefler(1998).Localdraglawforsuspensions
fromparticle-scalesimulations.Int.J.Mod.Phys.C9(8),1361.

WillmarthW.,HawkN.andHarveyR.(1964).Steadyandunsteadymotionandwakes
offreelyfallingdisks.Phys.Fluids7,197.

WuX.-L.andLibchaberA.(2000).ParticleDiffusioninaQuasi-Two-Dimensional
BacterialBath.Phys.Rev.Lett.84,3017.

XueDifJ.-Z.,fusion,Herbolzheimeerdispersion,andE.,settlingRutgersofM.A.,hardspheres.RusselWPhys.B.Reandv.Lett.Chaikim69,P.M.1715.(1992).

XueJ.Z.,CrystallineHerbbolzheimerfluidizedbeds.E.,Phys.RutgersRev.M.E.A.,51,4674.RusselW.BandChaikinP.M.(1995).

116

wledgmentsAckno

II,wψouldΨandlikefortohisthankinvProf.aluableDr.supportHansJ.andHerrmannorientation,forthewithoutgreathim,opportunitythisworktowworkouldathavthee
impossible.beenDuringmyworkIhadalotofmoralsupportbymymother,>adeWcRN,shehelpedme
IalonghavethetoedistancexpressmyandingratitudeallthetoaspectsDianaofAlonso,lifethatmyaregreatmoreadvawentureayofpartnerPh,ysics.weAlso,were
alwaysside-by-sideinourlifeinGermany.

IamgratefultothemembersoftheICA1forthesupportandthecomradeship,specially
withReinmarM¨uckandStefanSchwarzer,whoguidedmeatthebeginningintheartof
simulations.IwouldliketothankT.Ihleforthephysicalandscientificcriticsconcerning
tomywork.MyspecialthankstoV.R.IyengarandF.Raischelforwrittingcorrections
tothiswork,andalsotoMarliesParsons,whosupportedmeattheICA1concerningall
administrativeaffairs.

Iofwasvfrienshiperyluckinymytolife,meetalotspeciallyofR.friendsCruz,thatA.madeMora,myA.stayinEhl,J.GermanGaribayya,greatL.Le´opez,xpierienceJ.D.
Mu˜noz,O.L.Castro,F.Alonso,J.Jovel,V.Schwaemmle,M.Hasse,L.Aguilera,O.
Pozos,M.SidriandingeneraltoallOLApeople.
Iamalsoverygratefultoallthelatinamericancommunity,hereinStuttgart,andin
specialtotheColombiancommunity,forthewonderfultimethatwehadtogetheralong
years.these

Morover,IwouldalsoliketothankR.Dielforherunderstanding,affection,generosity
andconstantsupportduringthelasttimeofmythesis.
Itionwould(DeutscheliketoFgivemygratitudeorschungsgemeinschaft)totheforprojecttheSBF404financialofthesupport.GermanResearchFounda-

Finally,IwanttoexpressmydeepestrespectandadmirationfortheGermanpeoplethat
makemepossiblelearnabitofthisgreatnation.

FrankStuttgartRodolfo14thFJuneonseca2004Fonseca